Advanced Structured Materials

Francesco Marmo
Salvatore Sessa
Emilio Barchiesi
Mario Spagnuolo Editors

Mathematical
Applications
in Continuum
and Structural
Mechanics
Advanced Structured Materials

Volume 127

Series Editors
Andreas Öchsner, Faculty of Mechanical Engineering, Esslingen University of
Applied Sciences, Esslingen, Germany
Lucas F. M. da Silva, Department of Mechanical Engineering, Faculty of
Engineering, University of Porto, Porto, Portugal
Holm Altenbach , Faculty of Mechanical Engineering, Otto von Guericke
University Magdeburg, Magdeburg, Sachsen-Anhalt, Germany
Common engineering materials reach in many applications their limits and new
developments are required to fulfil increasing demands on engineering materials.
The performance of materials can be increased by combining different materials
to achieve better properties than a single constituent or by shaping the material or
constituents in a specific structure. The interaction between material and structure
may arise on different length scales, such as micro-, meso- or macroscale, and offers
possible applications in quite diverse fields.
This book series addresses the fundamental relationship between materials and their
structure on the overall properties (e.g. mechanical, thermal, chemical or magnetic
etc.) and applications.
The topics of Advanced Structured Materials include but are not limited to
• classical fibre-reinforced composites (e.g. glass, carbon or Aramid reinforced
plastics)
• metal matrix composites (MMCs)
• micro porous composites
• micro channel materials
• multilayered materials
• cellular materials (e.g., metallic or polymer foams, sponges, hollow sphere
structures)
• porous materials
• truss structures
• nanocomposite materials
• biomaterials
• nanoporous metals
• concrete
• coated materials
• smart materials

Advanced Structured Materials is indexed in Google Scholar and Scopus.

More information about this series at https://link.springer.com/bookseries/8611

Francesco Marmo · Salvatore Sessa ·
Emilio Barchiesi · Mario Spagnuolo
Editors

Mathematical Applications
in Continuum and Structural
Mechanics
Editors
Francesco Marmo Salvatore Sessa
University of Naples Federico II University of Naples Federico II
Napoli, Italy Napoli, Italy

Emilio Barchiesi Mario Spagnuolo

Università degli Studi dell’Aquila Università degli Studi dell’Aquila
L’Aquila, Italy L’Aquila, Italy

ISSN 1869-8433 ISSN 1869-8441 (electronic)

Advanced Structured Materials
ISBN 978-3-030-42706-1 ISBN 978-3-030-42707-8 (eBook)
https://doi.org/10.1007/978-3-030-42707-8

© Springer Nature Switzerland AG 2021

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Contents

1 Usage of Guided Wave Resonance Phenomena for Defect

Detection in Laminate Elastic Structures . . . . . . . . . . . . . . . . . . . . . . . . 1
Artem Eremin, Evgeny Glushkov, Natalia Glushkova,
and Rolf Lammering
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Computational Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Experimental Evaluation of Resonance Frequencies . . . . . . . . . . . 6
1.4 Estimation of the Defect Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Modelling of Piezocomposites with Mechanical Interface
Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Andrey V. Nasedkin
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Effective Moduli Method for Homogenization
of Two-Phase Piezoelectric Nanocomposite . . . . . . . . . . . . . . . . . . 15
2.3 Dimensionless Homogenization Problem . . . . . . . . . . . . . . . . . . . . 19
2.4 Finite Element Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Modelling of Representative Volume Elements . . . . . . . . . . . . . . . 21
2.6 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3 A Mathematical Model for Bone Cell Population Dynamics
of Fracture Healing Considering the Effect of Energy
Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Mahziyar Darvishi, Hooman Dadras,
Mohammad Mahmoodi Gahrouei, Kiarash Tabesh, and Dmitry Timofeev
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.1 The Main Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.2 The Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 38
v
vi Contents

3.2.3 The Stimulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.2.4 The Function κ(ϕ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.5 The Mechanical Framework . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.6 Numerical Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.7 Healing of Bone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.8 Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.3 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4 Second Gradient Linear and Nonlinear Constitutive Models
of Architectured Materials: Static and Dynamic Behaviors . . . . . . . . 53
Yosra Rahali, Hilal Reda, Benoit Vieille, Hassan Lakiss,
and Jean-François Ganghoffer
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2 First- and Second-Order Effective Moduli of Periodic
Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.1 Analytical Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.2 Homogenized Viscoelastic Behavior . . . . . . . . . . . . . . . . . 62
4.2.3 Incremental Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.3 Wave Propagation Analysis Based on Nonlinear Models . . . . . . . 67
4.3.1 Strain Energy Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5 An Application of Coulomb-Friction Model to Predict
Internal Dissipation in Concrete . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Giuliano Aretusi and Alessandro Ciallella
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2 A Brief Synopsis of the Employed Model . . . . . . . . . . . . . . . . . . . . 76
5.2.1 3D Formulation of a Micromorphic
Concrete-Based Material . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.2.2 Simplified Formulation for the Case of a Pure
Compression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.3 Numerical Simulations and Discussions . . . . . . . . . . . . . . . . . . . . . 79
5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
6 From the Swarm Robotics to Material Deformations . . . . . . . . . . . . . 87
Paolo D’Avanzo, Alessio Ciro Rapisarda, and Salvatore Samuele Sirletti
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2 Other Models in Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2.1 Position-Based Dynamics (PBD) . . . . . . . . . . . . . . . . . . . . 90
6.2.2 Swarm Robotics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.3 The Model Here Proposed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3.1 A Recall About Graph Theory . . . . . . . . . . . . . . . . . . . . . . 93
Contents vii

6.3.2 Constructing the Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

6.3.3 Relationship with Other Models . . . . . . . . . . . . . . . . . . . . 96
6.3.4 Meaning of Neighbors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.4 Numerical Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.4.1 Standard Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.4.2 Second Neighborhoods and Exotic Simulations . . . . . . . 110
6.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7 A Review of the Class of Bouc-Wen Differential Models
for Simulating Mechanical Hysteresis Phenomena . . . . . . . . . . . . . . . . 127
Davide Pellecchia and Massimo Paradiso
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.2 Modeling of Symmetric Hysteresis Loops . . . . . . . . . . . . . . . . . . . 129
7.2.1 Bouc Model and Its Modified Versions . . . . . . . . . . . . . . . 130
7.2.2 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.3 Modeling of Asymmetric Hysteresis Loops . . . . . . . . . . . . . . . . . . 132
7.3.1 Asymmetric Bouc-Wen Models . . . . . . . . . . . . . . . . . . . . . 133
7.3.2 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.4 Modeling of Pinched Hysteresis Loops . . . . . . . . . . . . . . . . . . . . . . 136
7.4.1 Pinching Bouc-Wen Models . . . . . . . . . . . . . . . . . . . . . . . . 137
7.4.2 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.5 Modeling of Degrading Hysteresis Loops . . . . . . . . . . . . . . . . . . . . 140
7.5.1 Degrading Bouc-Wen Models . . . . . . . . . . . . . . . . . . . . . . 140
7.5.2 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8 A Generalized Formulation of Time Integration Methods
for Nonlinear Dynamic Analysis of Hysteretic Mechanical
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Davide Pellecchia and Pasquale Cesarano
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.2 Families of Time Integration Methods . . . . . . . . . . . . . . . . . . . . . . . 151
8.2.1 Nonlinear Equilibrium Equations . . . . . . . . . . . . . . . . . . . 151
8.2.2 Generalized Formulation of Time Integration
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.3 Conventional Time Integration Methods . . . . . . . . . . . . . . . . . . . . . 153
8.3.1 Newmark’s Family of Methods . . . . . . . . . . . . . . . . . . . . . 154
8.3.2 Some Instances of the NFMs . . . . . . . . . . . . . . . . . . . . . . . 157
8.3.3 Implementation Scheme of the NFMs . . . . . . . . . . . . . . . . 158
8.4 Structure-Dependent Time Integration Methods . . . . . . . . . . . . . . . 159
8.4.1 Chang’s Family of Explicit Methods . . . . . . . . . . . . . . . . . 160
8.4.2 Some Instances of the CFEMs . . . . . . . . . . . . . . . . . . . . . . 163
8.4.3 Implementation Scheme of the CFEMs . . . . . . . . . . . . . . 163
8.5 Numerical Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
viii Contents

8.5.1 Mechanical System Properties . . . . . . . . . . . . . . . . . . . . . . 165

8.5.2 Applied Generalized External Force . . . . . . . . . . . . . . . . . 166
8.5.3 Hysteretic Model Parameters . . . . . . . . . . . . . . . . . . . . . . . 166
8.5.4 Results of the Nonlinear Time History Analyses . . . . . . . 166
8.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
9 Quasi-Harmonic Solutions for Transversely Isotropic
Magneto-Electro-Thermo-Elasticity: A Symbolic
Mathematics Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Francesco Marmo and Massimo Paradiso
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
9.2 Field Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
9.3 A General Solution to the Field Equations in Terms
of Quasi-Harmonic Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
9.3.1 Inversion of the Differential Operator L . . . . . . . . . . . . . . 179
9.3.2 Factorization of the Differential Equation
|L| ϕ = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
9.4 Automatic Evaluation of |L| and L∗ and Relevant
Coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.4.1 Evaluation of |L| . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
9.4.2 Evaluation of L∗ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
9.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
10 Mathematical Tools for the Seismic Analysis of Reinforced
Concrete Structures: A Selected Review . . . . . . . . . . . . . . . . . . . . . . . . . 191
Cristoforo Demartino and Sicheng Zhou
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
10.2 Review of Strategies Accounting for Global Torsion
in Buildings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
10.2.1 Review of the Dynamic Equivalent Rotational
Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
10.3 Computation of Multicomponent Actions by Seismic
Envelopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
10.4 Capacity Checks of Reinforced Concrete Beams . . . . . . . . . . . . . . 201
10.4.1 A General Algorithm to Perform Capacity Checks
by the Supreme Envelope . . . . . . . . . . . . . . . . . . . . . . . . . . 204
10.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
11 Form Finding of Shell Structures by Using Membrane Theory . . . . 213
Francesco Marmo and Nicoló Vaiana
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
11.2 The Membrane Theory of Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
11.2.1 Global and Local Reference Frames . . . . . . . . . . . . . . . . . 216
Contents ix

11.2.2 Transformation Formulas for Lengths and Areas . . . . . . 217

11.2.3 Distributed Loads and Stress Components . . . . . . . . . . . . 218
11.2.4 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
11.3 Form-Finding Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
11.3.1 Discretization of the Equilibrium Equations
by the Finite Difference Method . . . . . . . . . . . . . . . . . . . . 222
11.3.2 Assigning the Distribution of Projected Membrane
Stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
11.3.3 Evaluation of the Shell Mid-Surface Height . . . . . . . . . . . 227
11.3.4 Iterative Procedure for Assigning Projected Loads . . . . . 229
11.4 Numerical Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
11.4.1 Shell with One Free Side . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
11.4.2 Shell Supported at Corners . . . . . . . . . . . . . . . . . . . . . . . . . 231
11.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
12 Influence of Non-structural Components on Equivalent
Linearization of Buildings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Salvatore Sessa and Luciano Rosati
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
12.2 Brief Review of Tail-Equivalent Linearization . . . . . . . . . . . . . . . . 241
12.3 Influence of Secondary Devices on TELS . . . . . . . . . . . . . . . . . . . . 242
12.3.1 Frequency Content Comparison . . . . . . . . . . . . . . . . . . . . . 243
12.3.2 First Excursion Probability Comparison . . . . . . . . . . . . . . 246
12.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
13 Do We Really Need Pantographic Structures? . . . . . . . . . . . . . . . . . . . . 253
Mario Spagnuolo and Emilio Barchiesi
13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
13.2 Metamaterials Are (Natural) Materials on Demand . . . . . . . . . . . . 254
13.3 Second Gradient Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
13.4 Microstructure in Continuum Mechanics . . . . . . . . . . . . . . . . . . . . . 257
13.4.1 The Synthesis Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
13.5 Why We Really Need Pantographic Structures . . . . . . . . . . . . . . . . 259
13.5.1 The Existence of Pantographic Metamaterial
Motivates the Need of Second Gradient Theories . . . . . . 259
13.5.2 A Mechanical Diode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
13.5.3 An Iterative Algorithm for Synthesising
Metamaterials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
13.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
Contributors

Giuliano Aretusi MeMoCS - University of L’Aquila, L’Aquila, Italy

Emilio Barchiesi International Research Center for the Mathematics and
Mechanics of Complex Systems, University of L’Aquila, L’Aquila, Italy;
École Nationale d’Ingénieurs de Brest, ENIB, UMR CNRS, IRDL, Brest, France
Pasquale Cesarano CAE Technologies s.r.l. – Simulia, Centro Direzionale isola,
Naples, Italy
Alessandro Ciallella DICEAA - University of L’Aquila, L’Aquila, Italy
Paolo D’Avanzo University of Naples Federico II, Naples, Italy
Hooman Dadras Department of Mechanical Engineering, Isfahan University of
Technology, Isfahan, Iran
Mahziyar Darvishi Department of Mechanical Engineering, Isfahan University of
Technology, Isfahan, Iran
Cristoforo Demartino Zhejiang University—University of Illinois at Urbana
Champaign Institute (ZJUI), Zhejiang, Haining, PR China
Artem Eremin Institute for Mathematics, Mechanics and Informatics, Kuban State
University, Krasnodar, Russia
Jean-François Ganghoffer Université de Lorraine, Metz Cedex, France
Evgeny Glushkov Institute for Mathematics, Mechanics and Informatics, Kuban
State University, Krasnodar, Russia
Natalia Glushkova Institute for Mathematics, Mechanics and Informatics, Kuban
State University, Krasnodar, Russia
Hassan Lakiss Faculty of Engineering, Section III, Lebanese University, Beirut,
Lebanon
Rolf Lammering Institute of Mechanics, Helmut-Schmidt-University/University
of the Federal Armed Forces, Hamburg, Germany

xi
xii Contributors

Mohammad Mahmoodi Gahrouei Department of Mechanical Engineering,

Isfahan University of Technology, Isfahan, Iran
Francesco Marmo Department of Structures for Engineering and Architecture,
University of Naples Federic II, Napoli, Italy
Andrey V. Nasedkin Institute of Mathematics, Mechanics and Computer Sciences,
Southern Federal University, Rostov-on-Don, Russia
Massimo Paradiso Department of Structures for Engineering and Architecture,
University of Naples Federico II, Naples, Italy
Davide Pellecchia Department of Structures for Engineering and Architecture,
University of Naples Federico II, Naples, Italy
Yosra Rahali GPM, INSA Rouen, Université de Rouen, Saint-Étienne-du-
Rouvray, France
Alessio Ciro Rapisarda University of Naples Federico II, Naples, Italy
Hilal Reda Faculty of Engineering, Section III, Lebanese University, Beirut,
Lebanon
Luciano Rosati University of Naples Federico II, Napoli, Italy
Salvatore Sessa University of Naples Federico II, Napoli, Italy
Salvatore Samuele Sirletti University of Naples Federico II, Naples, Italy
Mario Spagnuolo Dipartimento di Ingegneria Civile, Ambientale e Architettura
(DICAAR), Universitá degli Studi di Cagliari, Cagliari, Italy;
International Research Center for the Mathematics and Mechanics of Complex
Systems, University of L’Aquila, L’Aquila, Italy
Kiarash Tabesh Department of Mechanical Engineering, Isfahan University of
Technology, Isfahan, Iran
Dmitry Timofeev Dipartimento di Ingegneria e Scienze dell’Informazione e
Matematica, Universitá degli Studi dell’Aquila, L’Aquila, Italy
Nicoló Vaiana Department of Structures for Engineering and Architecture, Univer-
sity of Naples Federic II, Napoli, Italy
Benoit Vieille GPM, INSA Rouen, Université de Rouen, Saint-Étienne-du-
Rouvray, France
Sicheng Zhou Zhejiang University—University of Illinois at Urbana Champaign
Institute (ZJUI), Zhejiang, Haining, PR China
Chapter 1
Usage of Guided Wave Resonance
Phenomena for Defect Detection
in Laminate Elastic Structures

Artem Eremin, Evgeny Glushkov, Natalia Glushkova, and Rolf Lammering

Abstract Since the values of natural scattering resonance frequencies strongly

depend on the size and shape of the scatterer, it seems worthwhile to use this
information for the enhancement of damage characterization capabilities of SHM
systems. In this chapter, potential possibility of such approach is demonstrated and
discussed in the example of guided wave interaction with a flat bottom hole in a
metallic plate that simulates deep pitting corrosion. A good agreement of the theo-
retically predicted and experimentally obtained scattering resonance frequencies has
confirmed the capability of this approach, which is illustrated by examples of the
defect’s size reconstruction.

Keywords Resonance frequency · Scattering · Structural health monitoring ·

Corrosion · Guided wave

1.1 Introduction

Elastic guided waves (GW) have become a recognized structural health monitoring
(SHM) tool for elongated laminate structures manufactured from metals and polymer
composites (Lammering et al. 2018). The presence of scattered GWs in the sensor-
acquired signals is considered as an indication of the developing structure-specific

A. Eremin (B) · E. Glushkov · N. Glushkova

Institute for Mathematics, Mechanics and Informatics, Kuban State University, Stavropolskaya st.
149, 350040 Krasnodar, Russia
e-mail: [email protected]
E. Glushkov
e-mail: [email protected]
N. Glushkova
e-mail: [email protected]
R. Lammering
Institute of Mechanics, Helmut-Schmidt-University/University of the Federal Armed Forces,
Holstehofweg 85, 22043 Hamburg, Germany
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 1

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_1
2 A. Eremin et al.

macroscopic damage. However, the damage evaluation process is complicated by

multiple diffraction patterns of GW interaction with structural features. Therefore, the
knowledge of any damage-specific wave characteristics that could be extracted from
the measured signals would allow for the enrichment of quantification opportunities
of SHM systems.
As a non-destructive evaluation method, GW can be useful also for damage recog-
nition for the newest class of artificial composites with microstructure also called
metamaterials (Barchiesi et al 2018, 2020c; Del Vescovo and Giorgio et al. 2014) with
mechanical properties that cannot be found in nature. These (macroscopic) properties
are mainly determined by the micro- or nanostructure of the considered metamaterial
rather than by the chemical and physical properties of the materials constituting it
at the microscopic level. An example of mechanical metamaterials is a pantographic
structure (dell’Isola et al. 2017, 2016a, b, c, 2019a, b; Placidi et al. 2017, 2016,
Spagnuolo et al. 2020, 2021; Barchiesi et al. 2020a, b; Turco and Barchiesi 2019). In
order to account for multiscale mechanical interactions, which take place for meta-
materials, higher order gradient continuum theories can be a choice (Alibert et al.
2013; Auffray et al. 2013; dell’Isola et al. 2015, 2016a, b, c, 2012; Rahali et al.
2015; Sciarra et al. 2007). The methodologies involving GW may also be used in a
synergetic way with other experimental (Vaiana et al. 2017), numerical (Vaiana et al.
2019, 2021; Serpieri et al. 2018; Sessa et al. 2018; Greco et al. 2018; Marmo et al.
2018), and analytical (Marmo et al. 2020, 2016; Trotta et al. 2016) aspects.
A phenomenon of GW resonance interaction with localized obstacles is among
the potential candidates for the enrichment of quantification opportunities of SHM
systems (Glushkov et al. 2012, 2015a, b; Solodov 2017). It is characterized by the
capturing of incident wave energy in the defect’s vicinity and its prolonged local-
ization with a gradual re-emission at near-real diffraction resonance frequencies
(eigenfrequencies of the corresponding defect-structure boundary value problem).
Since these frequencies strongly depend on the defect’s type, shape, and location,
they may, if being detected, enhance damage characterization capabilities of SHM
systems.
In this chapter, a theoretical and experimental study of the GW resonance inter-
action with circular flat-bottom holes (FBH) simulating deep pitting corrosion in
metallic structures is presented. Employing semi-analytical and mesh-based numer-
ical tools, the existence of near-real resonance diffraction frequencies is predicted,
and their strong dependence on the damage severity is shown. Experimental investiga-
tions performed with surface-mounted piezoelectric wafer active sensors and contact-
less laser Doppler vibrometry (LDV) confirm the values of the computed eigen-
frequencies and the corresponding amplitude localization patterns (eigenforms).
Predicted prolonged GW re-emission is also clearly observed, providing the possi-
bility of resonance frequency estimation using the spectral and time–frequency anal-
ysis of signals acquired by remote sensors. Some test examples of damage geometry
reconstruction through the minimization of the discrepancy between the measured
data and numerical results are presented and discussed.
1 Usage of Guided Wave Resonance Phenomena for Defect Detection … 3

1.2 Computational Models

(a) Analytically based approach

To study complex resonance frequencies f n of the diffraction problem, two
approaches have been implemented. In the first case, approximate plate theories
are employed to simulate GW diffraction by a circular FBH of radius R and depth
d in an opened waveguide of thickness H (Fig. 1.1). The Mindlin approach is used
for antisymmetric wave motion (i.e., modes A0, A1, and SH1) and the Poisson
theory handles the lowest order compressional waves (fundamental symmetric S0
and shear horizontal SH0 modes) (Cegla et al. 2008; Grahn 2008; Vemula and Norris
1997). Inthe framework
 of these theories, the components of the displacement vector
u(x) = u x , u y , u z where x = {x, y, z} are approximated by linear functions of the
transverse coordinate z. In cylindrical coordinates {r, θ, z}, his approximation takes
the form

u r = u r,S − zr , u θ = u θ,S − zθ , u z = w (1.1)

where u r,S (r, θ ) and u θ,S (r, θ ) are radial and angular displacement components
governed by the equations of the Poisson theory while w(r, θ ), r (r, θ ) and θ (r, θ )
are vertical and rotation components obtained from the Mindlin theory. The time-
harmonic motion u·exp(−iωt) is further assumed; the vector u(x, ω) is the frequency
spectrum of a transient wave field, ω = 2π f angular frequency, and f is frequency.
The displacement functions can be expressed in terms of scalar potentials:

∂φ 1 ∂ψ 1 ∂φ ∂ψ
u r,S = + , u θ,S = − , (1.2)
∂r r ∂θ r ∂θ ∂r

w(r, θ ) = w1 (r, θ ) + w2 (r, θ ), (1.3)

Fig. 1.1 General view of the problem geometry: (a)—analytically based simulation, (b)—mesh-
based discretization; cross-sections by xOz plane in the case of the plate theory approximation
(c) and FEM-PML approach (d)
4 A. Eremin et al.

∂w1 ∂w2 1 ∂V 1 ∂w1 1 ∂w2 ∂V

r = A1 + A2 + , θ = A 1 + A2 + .
∂r ∂r r ∂θ r ∂θ r ∂θ ∂r
The potential functions φ, ψ, w1 , w2 , V obey the Helmholtz equations with the
wavenumbers k p , ks , k1 , k2 , k3 standing for S0, SH0, A0, A1, and SH1 modes, respec-
tively. These quantities as well as the constants A1 , A2 in Eq. (1.3) are expressed via
the elastic constants of the plate (Cegla et al. 2008; Grahn, 2008; Vemula and Norris
1997).
Considering the interaction of the incident disturbance u0 (x, ω) with the FBH, the
total wavefield could be split into two different parts—one is for the region outside
the defect (r > R) and the other is for the area r < R below the FBH:

u0 + u> sc , r > R, −H/2 < z < H/2,
u(x, ω) = (1.4)
u<sc , r < R, −H/2 < z < H/2 − d.

Here, u> <

sc (x, ω) is the scattered field in the outer region, usc (x, ω) is the motion
getting inside the FBH. The relation (1.4) produces analogous decompositions for
the potentials entering the total field u(x, ω). Recalling the problem geometry, it is
convenient to expand them in Bessel/Hankel series with unknown coefficients:
∞

    
φ>, ψ > = H {am , bm }Hm(1) k pH , ksH r eimθ ,
m=−∞
∞
(1.5)
     
φ<, ψ <
=H { f m , gm }Jm k pH −d , ksH −d r eimθ ,
m=−∞

∞

 > > >   
w1 , w2 , V = {cm , dm , em }Hm(1) k1H , k2H , k3H r eimθ ,
m=−∞
∞
(1.6)
 < < <    
w1 , w2 , V = {h m , i m , jm }Jm k1H −d , k2H −d , k3H −d r eimθ .
m=−∞

The superscripts H and H − d indicate whether the wavenumber is for the outer
plate or FBH; am , ..., jm are unknown coefficients.
If the incident field u0 is known (e.g., a plane A0 or S0 wave upcoming from
infinity), the expansion coefficients are obtained from the linear algebraic equations
resulting from the substitution of expansions (1.5), (1.6) into the boundary condi-
tions at the FBH border r = R. The latter include the continuity conditions for all
displacements, forces, and moments as well as the coupling conditions arising from
the fact that the neutral axis of the FBH is offset from the neutral axis of the plate
material surrounding it (Cegla et al. 2008; Glushkov et al. 2018). Such coupling
allows for considering mode conversion phenomenon (i.e., A0–S0 and vice-versa
conversion) conditioned by the non-symmetry of the obstacle.
1 Usage of Guided Wave Resonance Phenomena for Defect Detection … 5

Due to the orthogonality of circumferential orders in Eqs. (1.5), (1.6), the equa-
tions for estimating am , ..., jm can be set up for each integer circumferential order
m. Therefore, the problem is reduced to the infinite set of 10 × 10 linear alge-
braic equations with respect to the unknowns am , bm , cm , dm , em , f m , gm , h m , i m , jm .
Their solution quickly converges and only the lowest circumferential orders could
be preserved. In general, the truncation number in the corresponding series depends
on the FBH size and characteristic wavelengths of the incoming and scattered fields.
In the computations presented below, the convergence is achieved with 9 lowest
circumferential orders.
Scattering resonance frequencies f n could be formally evaluated as singular points
in the complex frequency plane of the corresponding set of homogeneous alge-
braic systems (i.e., searching for the zeros of their determinants). However, such an
approach turned up being numerically unstable. Alternatively, the incident field in
the form of plane A0 or S0 wave is considered, truncated set of algebraic systems
is solved, and f n are approximated by the frequencies, which provide unbounded
growth of displacements in the FBH region, employing expansions (1.5) and (1.6).

(b) Finite element analysis

A more sophisticated three-dimensional case is studied with the finite element method
(FEM). The FEM is applied to a bounded specimen of thickness H containing all the
defects, which is enclosed by perfectly matched layers (PML) to simulate an open
waveguide (Fig. 1.1c) (Hein et al. 2012) (COMSOL 5.3). Although the built-in solver
yields many spurious complex eigenfrequencies intrinsic to finite bodies, the required
scattering resonance poles, remaining in the open guide, could be straightforwardly
selected since their imaginary parts are sufficiently small (|Im f n | << Re f n ) and
the corresponding eigenforms are strongly localized in the FBH.
To compare the performance of the implemented approaches and investigate the
influence of the FBH geometric parameters on the diffraction eigenfrequencies, a
waveguide of thickness H = 2 mm manufactured from aluminium (Young’s modulus
E = 70 GPa, Poisson coefficient ν = 0.33, and density ρ = 2700 kg/m3 ) is considered.
The obtained results (trajectories of complex resonance frequencies) are summarized
in Fig. 1.2. The left subplot corresponds to the fixed obstacle depth d = 1.75 mm
and the varying radius R; in the right one, on the contrary, the radius R = 2.5 mm
is constant while the depth d varies. With the increase of damage severity (larger
values of either R or d parameters), the number of almost real resonance poles in
the frequency range considered increases while the magnitude of their imaginary
parts decreases. Moreover, monotonous dependence of resonance frequencies f n
from the obstacle geometry parameters is observed, which confirms their potential
for damage characterization. Regarding the model comparison, the results obtained
with both approaches are consistent with each other; hence, the simplified 2D plate-
based model is capable of the reliable evaluation of resonance frequencies for a
relatively broad range of the parameters R and d.
6 A. Eremin et al.

Fig. 1.2 Dependencies of complex resonance frequencies from FBH geometry parameters

1.3 Experimental Evaluation of Resonance Frequencies

Preliminary experimental investigations have been carried out with an aluminium

plate sample of dimensions 400 × 300 × 2 mm3 containing a circular FBH of radius
R = 2.45 mm and depth d = 1.68 mm. Guided waves are generated by prolonged
rectangular piezoelectric wafer active sensors (PWAS) of the dimensions 5 × 30 ×
0.25 mm3 . It is adhered to the specimen 70 mm away from the FBH in such a way that
the elongated PWAS side is parallel to the Oy axis of the introduced Cartesian coor-
dinate system (see Figs. 1.1 and 1.4). Transient out-of-plane displacement velocity
vz (x, t) is measured on the plate surface by a Polytec PSV-500 one-dimensional
scanning laser Doppler vibrometer.
Theoretical resonance frequencies are evaluated in advance, and the first four of
them (f n , n = 1, 2, 3, 4), obtained by both models, are summarized in the first two
rows of Table 1.1. Corresponding localization patterns (eigenforms) at the FBH in
the open waveguide are shown in Fig. 1.3. As expected, the location and number of
nodal points change with the increase of n.
Since a considerable increase of the oscillation amplitude is expected at the reso-
nance frequencies, a straightforward way to experimentally reveal their presence is
ex p
to measure transient oscillations inside the FBH and select such frequencies f n
at which strong local maxima of the signal spectra are observed (Glushkov et al.
2015a, b). Preliminarily, the analysis of the corresponding theoretical eigenforms
(Fig. 1.3a) suggests that a few measurement points located at the FBH bottom could
ex p
be enough for the estimation and selection of f n values. In accordance with the

Table 1.1 The first four predicted and measured resonance frequencies of the circular FBH
Employed approach f 1 , kHz f 2 , kHz f 3 , kHz f 4 , kHz
Plate theory 127.4–5.2i 255.1–13.5i 398.2–22.8i 455.9–13.8i
FEM + PML 112.4–3.7i 223.7–10.5i 351.5–17.8i 407.3–9.3i
Experiment 116.3 223.7 362.1 421.5
1 Usage of Guided Wave Resonance Phenomena for Defect Detection … 7

f1 f2 f3 f4

Fig. 1.3 Computed (upper row) and measured (lower row) resonance localization patterns
(amplitudes of out-of-plane velocities on the FBH surface) for the first four resonance frequencies

Fig. 1.4 Sketch of the experimental setup (a); points on the FBH surface for the frequency spectrum
evaluation (b); normalized spectra of the signals measured at these points (blue line—point C1, red
line—C2, green line—C3) (c)

antinode patterns, the local maxima of the frequency response at the central point
C1(0,0) (Fig. 1.4b) are expected at the first and fourth resonance frequencies, the
ex p
point C2(R/2,0) can provide the value of f 2 , while the third resonance frequency
ex p
f 3 should be dominant at the location C3(0,3R/4). To check this assumption, the
PWAS is excited with a broadband rectangular 1 µs pulse voltage, and the veloc-
ities Vz (x, t) are measured at the proposed points C1–C3. The evaluated spectra
|vz (x, f )| are summarized in Fig. 1.4c, and all the four expected local maxima are
clearly visible. Moreover, their location on the frequency axis is in good agree-
ment with the predicted values f n (the last row and the first two rows in Table 1.1,
correspondingly).
Within the SHM concept, the wave signals are measured at a limited set of
distributed active/passive sensors. Therefore, it is important to understand whether
ex p
the resonance frequencies f n can be detected outside the damaged region. Since
the resonance interaction of the incident GW with the FBH is characterized by a
prolonged localization of the wave motion at the obstacle with a gradual re-emission
into the structure, the presence of such narrow-band oscillations in the wave signals
measured at some remote points might serve as a reliable resonance indicator. To
illustrate such a possibility, transient wave signals have been measured at point
B(−35.5, −15.4) mm located aside from the straight path between the PWAS and
FBH (Fig. 1.4).
8 A. Eremin et al.

Figure 1.5 a gives an example of the shape of the velocity signal vz (x, t) obtained
at this point for the broadband 2 µs PWAS excitation. From this plot, it is difficult to
distinguish the waves scattered by the FBH from the initial PWAS generated signal
and spurious reflections from the specimen boundaries. Therefore, time–frequency
analysis based on the continuous Gabor wavelet transform has been applied (Kishi-
moto et al. 1995). The corresponding normalized scalogram truncated for better
contrast to the 0.5-level is provided in Fig. 1.5b. The incident waves are pronounced
here as a dark broad vertical segment appearing at 0.02–0.04 ms. Recalling disper-
sion properties of fundamental antisymmetric mode A0 in a 2 mm aluminium layer
and considering the location of point B, the dark area at 0.06 ms is identified as
the wavefield scattered by the defect. Multiple reflections from boundaries arrive at
point B from t = 0.115 ms (yellow area in the low-right corner of the plot). There-
fore, the wavefields re-emitted from the defect due to the trapped mode phenomena
might manifest themselves in the time interval between these two events, i.e., for
t ∈ (0.065, 0.113) ms (denoted by the vertical dashed-dotted lines in Fig. 1.5a,
b). Some elongated patterns within this segment are already visible in Fig. 1.5b.
The spectrum of this part of the measured signal is shown in Fig. 5c. It indicates
the presence of two strong local maxima at the frequencies f˜1 = 117 KHz and
f˜1 = 224 kHz, which coincide with the first two f n determined above. The latter
ex p

is believed to serve as an illustration of the capability for the damage scattering

resonance frequency estimation with sensing points located away from the damaged
area.

Fig. 1.5 Remote estimation of resonance frequencies. (a) Example of velocity signal (broadband
2 µs PWAS excitation); (b) corresponding normalized scalogram; the spectrum of the measured
part for t between 0.065 and 0.113 ms is plotted in (c)
1 Usage of Guided Wave Resonance Phenomena for Defect Detection … 9

Table 1.2 Measured and reconstructed radius and depth of a circular FBH for two experimental
samples
Waveguide Measured radius Measured depth Identified radius, Identified depth,
thickness H, mm R, mm d, mm mm mm
2 2.46 1.68 2.72 1.64
3 2.76 2.6 2.9 2.66

1.4 Estimation of the Defect Size

In this section, we discuss a possible application of the diffraction resonance

phenomenon for damage sizing. The geometry of a circular FBH is parametrized
by the radius R and depth d. The task is to estimate these two parameters from the
values of several first resonance frequencies. A natural way for solving this problem
ex p
is to construct a quadratic discrepancy functional between N experimental f n and
computed f n eigenfrequencies


N
Err (R, d) = ( f n − f nex p )2 (1.6)
n=1

and to minimize it with respect to the R and d variables, e.g., by employing the
microgenetic algorithm (Krishnakumar 1989).
The results of damage geometry reconstruction for the initial experimental sample
and the thicker second one (H = 3 mm) are given in Table 1.2. Here, the first
three experimental resonance frequencies have served as an input, and the numerical
evaluation of theoretical f n has been performed within the analytically based model.
Disregard its relative simplicity, the estimated values of R and d are in good agreement
with the results measured by the digital micrometer.

1.5 Conclusion

Resonance phenomena accompanying the interaction of elastic GWs with local thick-
ness changes in the form of circular FBH are investigated. The parametric analysis
performed within the approximate plate theory and FEM-PML computational model
has shown the presence of almost real natural scattering resonance frequencies and
revealed the general influence of FBH geometry parameters on their distribution in
the complex frequency plane. Experimental LDV measurements have confirmed a
strong and prolonged wave motion localized at the obstacle at the predicted reso-
nance frequencies. The possibility of resonance frequency estimation using wave
signals acquired at off-damage locations as well as the reconstruction of damage
10 A. Eremin et al.

geometry relying on the experimental eigenfrequencies is illustrated by numerical

and experimental examples for some specific samples.

Acknowledgements This work is supported by the Russian Science Foundation (Project No. 17-
11-01191).

References

Alibert JJ, Seppecher P, dell’Isola F (2003) Truss modular beams with deformation energy de-
pending on higher displacement gradients. Math Mech Solids 8(1)
Auffray N, dell’Isola F, Eremeyev V, Madeo A, Rosi G (2013) Analytical continuum mechanics à
la Hamilton-Piola: least action principle for second gradient continua and capillary fluids. Math
Mech Solids
Barchiesi E, Spagnuolo M, Placidi L (2018) Mechanical metamaterials: a state of the art. Math
Mech Solids
Barchiesi E, dell’Isola F, Hild F, Seppecher P (2020b) Two-dimensional continua capable of
large elastic extension in two independent directions: asymptotic homogenization, numerical
simulations and experimental evidence. Mech Res Commun 103
Barchiesi E, Eugster SR, dell’isola F, Hild F (2020c) Large in-plane elastic deformations of bi-
pantographic fabrics: asymptotic homogenization and experimental validation. Math Mech Solids
25(3):739–767
Barchiesi E, Yang H, Tran CA, Placidi L, Müller WH (2020a) Computation of brittle fracture
propagation in strain gradient materials by the FEniCS library. Math Mech Solids
Cegla FB, Rohde A, Veidt M (2008) Analytical prediction and experimental measurement for mode
conversion and scattering of plate waves at non-symmetric circular blind holes in isotropic plates.
Wave Motion 45:162–177
Del Vescovo D, Giorgio I (2014) Dynamic problems for metamaterials: review of existing models
and ideas for further research. Int J Eng Sci 80:153–172
dell’Isola F, Seppecher P, Madeo (2012) A How contact interactions may depend on the shape of
Cauchy cuts in Nth gradient continua: approach “à la D’Alembert”. Z Für Angew Math Und Phys
63(6)
dell’Isola F, Andreaus U, Placidi L (2015) At the origins and in the vanguard of peridynamics, non-
local and higher-gradient continuum mechanics: An underestimated and still topical contribution
of Gabrio Piola. Math Mech Solids 20(8)
dell’Isola F, Della Corte A, Giorgio I (2016a) Higher-gradient continua: the legacy of Piola, Mindlin,
Sedov and Toupin and some future research perspectives. Math Mech Solids
dell’Isola F, Della Corte A, Greco L, Luongo A (2016b) Plane bias extension test for a continuum
with two inextensible families of fibers: a variational treatment with Lagrange multipliers and a
perturbation solution. Int J Solids Struct
dell’Isola F, Giorgio I, Pawlikowski M, Rizzi N (2016) Large deformations of planar extensible
beams and pantographic lattices: heuristic homogenization, experimental and numerical examples
of equilibrium. Proc R Soc A 472(2185)
dell’Isola F, Cuomo M, Greco L, Della Corte A (2017) Bias extension test for pantographic sheets:
numerical simulations based on second gradient shear energies. J Eng Math
dell’Isola F, Seppecher P, Alibert JJ, Lekszycki T, Grygoruk R, Pawlikowski M, Steigmann D,
Giorgio I, Andreaus U, Turco E, Gołaszewski M, Rizzi N, Boutin C, Eremeyev VA, Misra A,
Placidi L, Barchiesi E, Greco L, Cuomo M, Cazzani A, Della Corte A, Battista A, Scerrato D,
Eremeeva IZ, Rahali Y, Ganghoffer JF, Müller W, Ganzosch G, Spagnuolo M, Phaff A, Barcz K,
1 Usage of Guided Wave Resonance Phenomena for Defect Detection … 11

Hoschke K, Neggers J, Hild F (2019) Pantographic metamaterials: an example of mathematically

driven design and of its technological challenges. Continuum Mech Thermodyn 31(4):851–884
dell’Isola F, Seppecher P, Spagnuolo M, Barchiesi E, Hils F, Lekszycki T, Giorgio I, Placidi L,
Andreaus U, Cuomo M, Eugster SR, Pfaff A, Hoschke K, Langkemper R, Turco E, Sarikaya R,
Misra A, De Angelo M, D’Annibale F, Bouterf A, Pinelli X, Misra A, Desmorat B, Pawlikowski
M, Dupuy C, Scerrato D, Peyre P, Laudato M, Manzari L, Göransoon P, Hesch C, Hesch S,
Franciosi P, Dirrenberger J, Maurin F, Vangelatos Z, Grigoropoulos C, Melissinaki V, Farsari M,
Muller W, Abali BE, Liebold C, Ganzosch G, Harrison P, Drobnicki R, Igumnov L, Alzahrani F,
Hayat T (2019) Advances in pantographic structures: design, manufacturing, models, experiments
and image analyses. Continuum Mech Thermodyn 31(4):1231–1282
Glushkov E, Glushkova N, Golub MV, Moll J, Fritzen CP (2012) Wave energy trapping and
localization in a plate with a delamination. Smart Mater. Struct. 21
Glushkov E, Glushkova N, Eremin A, Lammering R (2015) Trapped mode effects in notched
plate-like structures. J Sound Vib 358:142–151
Glushkov EV, Glushkova NV, Eremin AA, Lammering R (2015b) Guided wave propagation and
diffraction in plates with obstacles: resonance transmission and trapping mode effects. Phys
Procedia 70:447–450
Glushkov E, Glushkova N, Eremin A, Lammering R (2018) Trapped modes and resonance wave
transmission in a plate with a system of notches. J Sound Vib 412:360–371
Grahn T (2008) Lamb wave scattering from a circular partly through-thickness hole in a plate. Wave
Motion 37:63–80
Greco F, Luciano R, Serino G, Vaiana N (2018) A mixed explicit-implicit time integration approach
for nonlinear analysis of base-isolated structures. Ann Solid Struct Mech 10(1–2):17–29
Hein S, Koch L, Nannen L (2012) Trapped modes and Fano resonances in two-dimensional
acoustical duct cavity systems. J Fluid Mech 692:257–287
Kishimoto K, Inoue H, Hamada M, Shibuya T (1995) Time frequency analysis of dispersive waves
by means of wavelet transform. J Appl Mech 62:841–846
Krishnakumar K (1989) Micro-genetic algorithms for stationary and non-stationary function opti-
mization. In: Proceedings SPIE 1196 intelligent control and adaptive systems. Rodriguez G,
Philadelphia, PA, United States
Lammering R, Gabbert U, Sinapius M, Schuster T, Wierach P (eds) (2018) Lamb-wave based
structural health monitoring in polymer composites. Springer International Publishing AG, Cham,
Switzerland
Marmo F, Sessa S, Rosati L (2016) Analytical solution of the Cerruti problem under linearly
distributed horizontal loads over polygonal domains. J Elast 124(1):27–56. https://doi.org/10.
1007/s10659-015-9560-3
Marmo F, Toraldo F, Rosati A, Rosati L (2018) Numerical solution of smooth and rough contact
problems. Meccanica 53:1415–1440. https://doi.org/10.1007/s11012-017-0766-2
Marmo F, Sessa S, Vaiana N, De Gregorio D, Rosati L (2020) Complete solutions of three-
dimensional problems in transversely isotropic media. Continuum Mech Thermodyn 32(3):775–
802. https://doi.org/10.1007/s00161-018-0733-8
Placidi L, Barchiesi E, Turco E, Rizzi NL (2016) A review on 2D models for the description of
pantographic fabrics. Z Für Angew Math Phys 67(5)
Placidi L, Andreaus U, Giorgio I (2017) Identification of two-dimensional pantographic structure
via a linear D4 orthotropic second gradient elastic model. J Eng Math
Rahali Y, Giorgio I, Ganghoffer JF, dell’Isola F (2015) Homogenization à la Piola produces second
gradient continuum models for linear pantographic lattices. Int J Eng Sci 97
Sciarra G, dell’Isola F, Coussy O (2007) Second gradient poromechanics. Int J Solids Struct 44(20)
Serpieri R, Sessa S, Rosati L (2018) A MITC-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220. https://doi.org/10.1016/j.compstruct.2018.02.031
12 A. Eremin et al.

Sessa S, Marmo F, Vaiana N, Rosati L (2018) A computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthquake Eng https://doi.
org/10.1080/13632469.2018.1551161
Solodov I (2017) Resonant ultrasonic activation of damage: application for diagnostic imaging. Res
Nondestruct Eval 28:28–44
Spagnuolo M, Franciosi P, dell’Isola F (2020) A Green operator-based elastic modeling for two-
phase pantographic-inspired bi-continuous materials. Int J Solids Struct 188:282–308
Spagnuolo M, Yildizdag ME, Andreaus U, Cazzani AM (2021) Are higher-gradient models also
capable of predicting mechanical behavior in the case of wide-knit pantographic structures? Math
Mech Solids 26(1):18–29
Trotta S, Marmo F, Rosati L (2016) Analytical expression of the Eshelby tensor for arbitrary
polygonal inclusions in two-dimensional elasticity. Compos B Eng 106:48–58. https://doi.org/
10.1016/j.compositesb.2016.09.010
Turco E, Barchiesi E (2019) Equilibrium paths of Hencky pantographic beams in a three-point
bending problem. Math Mech Complex Syst 7(4):287–310
Vaiana N, Spizzuoco M, Serino G (2017) Wire rope isolators for seismically base-isolated
lightweight structures: experimental characterization and mathematical modeling. Eng Struct
140:498–514. https://doi.org/10.1016/j.engstruct.2017.02.057
Vaiana N, Sessa S, Marmo F, Rosati L (2019) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901. https://doi.org/10.1007/s11071-019-05022-5
Vaiana N, Sessa S, Rosati L (2021) A generalized class of uniaxial rate-independent models for
simulating asymmetric mechanical hysteresis phenomena. Mech Syst Signal Process 146: 106984
https://doi.org/10.1016/j.ymssp.2020.106984
Vemula C, Norris AN (1997) Flexural wave propagation and scattering on thin plates using Mindlin
theory. Wave Motion 26:1–12
Chapter 2
Modelling of Piezocomposites
with Mechanical Interface Effects

Andrey V. Nasedkin

Abstract This investigation is devoted to the determination of the material prop-

erties of piezoelectric mixed composites with stochastically distributed inclusions
or pores and with mechanical imperfect interface, which simulates the nanoscale
effects. Finite element package ANSYS was used to simulate the representative
volume element and to calculate the effective material properties of piezocomposite
material. This approach is based on the theory of effective moduli, the representa-
tive volumes simulation and the finite element technologies. The contact boundaries
between the materials of different phases were covered by the surface elastic shell
elements with membrane stress option in order to take the interface effects into
account. As an example, the results of the effective moduli calculation for a porous
ferrohard piezoceramics were presented. The results have shown that the surface
stresses on the pore boundaries could have a significant effect on the values of the
effective stiffness moduli and little effect on the values of the effective piezomoduli
and dielectric permittivities.

Keywords Piezoelectricity · Porous piezoceramics · Effective moduli ·

Nanomechanics · Surface stress · Interphase boundary · Finite element method

2.1 Introduction

Piezoelectric composite materials are widely used for the manufacture of high-tech
devices for hydroacoustics, nondestructive testing, medical diagnostics and therapy,
level and flow measurement, consumer, automotive and aerospace industries. The
analysis of these materials shows that the material properties remain to be the essential
limiting factor in the development of the most effective piezoelectric transducers. A
lot of new nanostructured piezoelectric composite materials, which were developed in
the recent years, have a range of important advanced features, such as the possibilities

A. V. Nasedkin (B)
Institute of Mathematics, Mechanics and Computer Sciences, Southern Federal University,
Rostov-on-Don, Russia

© Springer Nature Switzerland AG 2021 13

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_2
14 A. V. Nasedkin

of controllable variation of the functional characteristics within a wide range and the
ultra-low mechanical quality factor.
Furthermore, it should be noted that the simulation of composite micro- and
nano-materials has specific features. It is known that some nanomaterials have
unconventional physical properties that significantly differ from the characteristics
of usual macrosized bodies. For example, the known experimental fact is the stiffness
increases with the reduction of the nanoobject sizes. One of the factors responsible
for this behaviour can be the surface effect. As research of the recent years shows,
for the bodies of submicro- and nano-sizes, the surface stresses play an important
role and influence the deformation of the bodies. In connection to this, an interesting
problem can arise from the extension of this approach to the nanostructured piezo-
electric composite materials (Dai et al. 2011; Fang et al. 2018; Huang and Yu 2006;
Pan et al. 2011; Park et al. 2011; Wang et al. 2014, 2016a, b; Zhao et al. 2016).
Therefore, here it is logical to consider not only the mechanical surface effects but
also the surface effects for electric fields (Eremeyev and Nasedkin 2017; Nasedkin
2017).
Theoretical investigations of piezoelectric and magnetoelectric nanosized mate-
rials with surface effects and imperfect interface models were also presented in (Dai
et al. 2011; Pan et al. 2011; Eremeyev and Nasedkin 2017; Gu et al. 2014a, b; Gu
and Qin 2014; Nasedkin et al. 2014), etc. Formulae for the effective shear modulus
of a fiber reinforced piezoelectric composite was obtained in (Xiao et al. 2011) using
a self-consistent method. Investigations of effective moduli for nanosized piezoelec-
tric composites were continued in (Wang et al. 2014; Nasedkin 2017; Chen 2008;
Gu et al. 2015; Malakooti and Sodano 2013), etc.
The study of piezoelectric composite materials may also be linked to the reasoning
and methodologies used for other complex materials. More specifically, it has been
shown in literature that variational approaches have many epistemological and heuris-
tical advantages (dell’Isola et al. 2020; dell’Isola and Placidi 2011). This kind of
approach has been successfully applied to model complex materials (Placidi et al.
2019; Pideri and Seppecher 1997; Barchiesi et al. 2018; Camar-Eddine and Seppecher
2001; Milton et al. 2009; Alibert et al. 2003; Turco et al. 2019a, 2020), especially for
the so-called metamaterials (dell’Isola et al. 2019a, b; Placidi et al. 2018a, b, 2020;
Turco et al. 2019b; Spagnuolo et al. 2020, 2021; Spagnuolo, 2020). Many application
fields may be found, such as structural mechanics (Vaiana et al. 2021, 2019; Sessa
et al. 2015, 2017, 2018a, b, 2019a, 2019b; Serpieri et al. 2018; Cricrì et al. 2015).
In the present work, the models of two-phase piezoelectric composite materials are
developed in the framework of classic continuum approaches of solid mechanics and
methods of composite mechanics. Similarly (Nasedkin 2019) these models are used to
construct more complicated models of the micro- and nano-sized piezoelectric bodies
that additionally take into account the surface or interphase mechanical boundary
conditions.
We use an integrated approach to the determination of the effective moduli of
nanoporous piezoelecrtric composites with stochastically distributed porosity. In
order to take into account nanoscale level at the borders between material and pores,
the Gurtin–Murdoch model of surface stresses are used. ANSYS finite element
2 Modelling of Piezocomposites with Mechanical Interface Effects 15

package was used to simulate the representative volume elements and to calcu-
late the effective moduli. This approach is based on the theory of effective moduli of
composite mechanics, simulation of the representative volumes and the finite element
method. Here, we adopt that the contact boundaries between material and pores are
covered by the surface membrane elements in order to take the surface stresses into
account.

2.2 Effective Moduli Method for Homogenization

of Two-Phase Piezoelectric Nanocomposite

Let  be a representative volume element of a two-phase composite piezoelectric

material with nanodimensional inclusions or pores;  = (1) ∪ (2) ; (1) is the
volume occupied by the primary material of the first phase (matrix); (2) is the set
of the volumes occupied by the material of the second phase (inclusions or pores);
 = ∂  is the external boundary of the volume   s is the set of frontier surfaces
of materials with different phases ( s = ∂(1) ∩∂(2) ); n is the external unit normal
vector to the boundary, outward with respect to (1) , i.e., to the region occupied by
the primary material; x = {x1 , x2 , x3 } is the vector of the Cartesian coordinates.
We suppose that the volumes (1) and (2) are filled with different piezoelectric
materials. Then in the framework of linear static theory of piezoelectricity we have
the following system of equations:

σi j, j = 0, Di,i = 0, σi j = ciEjkl εkl − eki j E k , Di = eikl εkl + εik

S
Ek (2.1)

 
εi j = u i, j + u j,i /2, E i = −ϕ,i (2.2)

where σi j are the components of the second rank stress tensor σ ; εi j are the compo-
nents of the second rank strain tensor ε; Di are the components of the electric flux
density vector D; E i are the components of the electric field vector E; u i are the
components of the displacement vector u; ϕ is the function of electric potential; ciEjkl
is the forth rank tensor of elastic stiffness moduli at constant electric field; eikl is
the third rank tensor of piezoelectric moduli; εik S
is the second rank tensor of dielec-
(m)
tric permittivity moduli at constant mechanical strain; ciEjkl = ciE(m) jkl , eikl = eikl ,
(m) (m) (m) (m) (m)
ciEjkl = ciE(m) S(m)
jkl , εik = εik , σi j = σi j , εi j = εi j , Di = Di , E i = E i , u i = u i ,
S

ϕ = ϕ (m) , x ∈ (m) , m = 1, 2.
The material
 moduli of the piezoelectric medium  have usual properties of
symmetry ciEjkl = c Ejikl = ckliE
j ; eikl = e ilk ; ε S
ik = ε S
ki and positive definiteness, i.e.,
∃W > 0, ∀εi j = ε ji , E i :
   
U εi j , E i = ciEjk εi j εkl + εklS E k El /2 ≥ W (εkl εkl + E k E k )/2 (2.3)
16 A. V. Nasedkin
 
where U εi j , E i is the density of the intrinsic energy in the volume  .
In vector–matrix notation in R 3 the formulas (2.1), (2.2) can be rewritten in the
form

L T (∇) · T = 0, ∇ · D = 0, T = c E · S − eT · E, D = e · S + ε S · E (2.4)

⎡ ⎤
∂1 0 0 0 ∂3 ∂2
S = L(∇) · u, E = −∇ϕ, L T (∇) = ⎣ 0 ∂2 0 ∂3 0 ∂1 ⎦, (2.5)
0 0 ∂ 3 ∂2 ∂1 0

where L(a) is a matrix operator for the vector a; ∇ = {∂1 , ∂2 , ∂3 }; T =

{σ11 , σ22 , σ33 , σ23 , σ13 , σ12 } is the six-dimensional array of the stress components;
S = {ε11 , ε22 , ε33 , 2ε23 , 2ε13 , 2ε12 } is the six-dimensional array of the strain compo-
nents; c E is the 6 × 6 matrix of elastic moduli at constant electric field; e is the 3 × 6
matrix of piezoelectric moduli; ε S is the 3×3 matrix of dielectric permittivity moduli
at constant strains; cαβ E
= ciEjkl , eiβ = eikl , α, β = 1, 2, ..., 6; i, j, k, l = 1, 2, 3
with the correspondence law α ⇔ (i j) ∼ ( ji), β ⇔ (kl) ∼ (lk), 1 ⇔ (11),
2 ⇔ (22), 3 ⇔ (33), 4 ⇔ (23) ∼ (32), 5 ⇔ (13) ∼ (31), 6 ⇔ (12) ∼ (21); (...)T
is the transpose operation; (...) · (...) is the scalar product operation between two
vectors or a matrix by vector multiplication. The matrices c E and ε S are symmetric
(cαβ
E
= cβαE
, εik
s
= εki s
) and positive definite by virtue of the properties (2.3), i.e.,
∃W > 0, ∀S, E:
   
U (S, E) = ST · cE · S + E T · εS · E /2 ≥ W ST · S + E T · E /2 (2.6)

In accordance with Gurtin–Murdoch model for surface stresses, we will assume

that on the nanosized interphase boundaries  s the following equations are satisfied:
 
L T (n) · [T ] = L T ∇ s · T s , n · [ D] = 0, x ∈  s , (2.7)

where [T ] = T (1) − T (2) is the stress jump, etc.; ∇ s is the surface gradient operator,
associated with usual spatial nabla–operator ∇ by the formula ∇ s = ∇ − n∂/∂r ; r is
the coordinate, measured by the normal to  s ; T s = σ11 s
, σ22
s
, σ33
s
, σ23
s
, σ13
s
, σ12
s
is
the array of surface stresses σi j . Note that surface stresses σi j have the dimensionality
s s

(N/m) different from the dimensionality of usual stresses σi j (N/m2 ).

We adopt that the surface stresses T s are related only to the surface strains Ss by
the surface Hooke’s law
 
T s = c Es · Ss , Ss = L ∇ s · us , us = A · u, A = I − nn* (2.8)

where c Es is the 6 × 6 matrix of surface elastic moduli with dimensionality N/m; I

is the 3 × 3 ideity matrix.
2 Modelling of Piezocomposites with Mechanical Interface Effects 17

The properties of the matrix of surface elastic moduli c Es are similar to the corre-
Es
sponding properties of the matrix c E , i.e., cαβ = cβα
Es
, and the transformed matrix c̃ Es
in local coordinate system with the tangent orts τ 1 , τ 2 and normal n, has the feature
s
of positive definiteness: ∃W > 0,∀ S̃ = S̃1s , S̃2s , 0, 0, 0, S̃6s ,
 s sT s T
2U S̃ = S̃ · c̃ Es · S̃ ≥ W S̃ · S̃ (2.9)

that follow from the condition

 of the positive definiteness of the surface mechanical
s s
energy density U S̃ relative to surface strain S̃ .
Setting the appropriate boundary conditions on the external boundary  = ∂  ,
we can find the solutions of the problems (2.4), (2.5), (2.7), (2.8) for heterogeneous
piezoelectric medium in the representative volume element  . Then the comparison
of the averaged stresses, strains, electric fluxes and electric fields with the analogous
values for homogeneous medium (i.e., for the comparison medium) will allow us to
determine the effective moduli for the composite material. We note that, because the
piezoelectric material is an anisotropic medium, in order to determine the full set of
the effective moduli it is necessary to solve several problems (2.4), (2.5), (2.7), (2.8)
for different types of boundary conditions.
Here the main tasks consist in the choice of the representative volume element with
the specific structure and in the choice of the boundary conditions for the heteroge-
neous medium and the comparison medium, as well as the technologies for solving the
boundary problems for heterogeneous piezoelectric media. According to the previ-
ously developed methods for the simulation of the ordinary piezoelectric composites
(Nasedkin and Shevtsova 2011, 2013), we consider analogous approaches for the
piezoelectric problems with surface stresses.
For homogeneous piezoelectric comparison medium we adopt that the same
Eqs. (2.4), (2.5) are satisfied with constant effective moduli c Eeff , eeff and ε Seff ,
which are to be determined. In order to calculate these moduli, let us assume that on
the external boundary  the following relations take place

u = L T (x) · S0 , ϕ = −x · E 0 , x ∈ , (2.10)

where S0 = {S01 , S02 , S03 , S04 , S05 , S06 }; S0β are some constant values that do not
depend on x; E 0 is some constant vector. Then, u = L T (x)·S0 , S = S0 , ϕ = −x· E 0 ,
E = E 0 , T = T 0 = c Eeff · S0 − eeffT · E 0 , D = D0 = eeff · S0 + ε Seff · E 0 will give
the solution for the problem (2.4), (2.5), (2.7), (2.8), (2.10) in the volume  for the
homogeneous comparison medium.
Let us solve now problem (2.4), (2.5), (2.7), (2.8), (2.10) for heterogeneous
medium and assume that for the composite medium and for the comparison medium
the averaged stresses and electric flux densities are equal, i.e., < T > =< T 0 > ,
< D > =< D0 > . Hereinafter the angle brackets < ... > denote the values,
averaged by the volume  , and the angle brackets < ... > denote the values,
18 A. V. Nasedkin

averaged by the volume  and by the interface surfaces  s .

 
1 1
< (•) > = ∫(•)d, < (•) > = ∫(•)d + ∫ (•) d
s
(2.11)
||  ||  s

Therefore, we obtain that the equations T 0 = c Eeff · S0 − eeffT · E 0 =< T >

and D0 = eeff · S0 + ε Seff · E 0 =< D > are satisfied for the effective moduli of
the composite, where S0 and E 0 are the given values from the boundary conditions
(2.10).
Hence, even in the assumption of the anisotropy of the general form for the
Eeff eff
comparison medium, all the stiffness moduli cαβ , the piezomoduli eiβ and the
dielectric permittivity moduli εi j can be computed, if we solve nine boundary
Seff

value problems (2.4), (2.5), (2.7), (2.8), (2.10), having set in (2.10) only one of the
components S0ς or E 0k (ς = 1, 2, ..., 6; k = 1, 2, 3) not equal to zero:
– problems I – VI (β = 1, 2, ..., 6)

S0ς = S0 δβς , E ok = 0 ⇒ cαβ

eff
= Tα  /S0 ; eiβ
eff
= Di  /S0 (2.12)

– problems VII – IX ( j = 1, 2, 3)

jα = − Tα  /E 0 ; κi j = Di  /E 0 ,
S0ς = 0, E 0k = E 0 δ jk ⇒ eeff eff
(2.13)

where δi j is the Kronecker symbol; α = 1, 2, ..., 6; i = 1, 2, 3.

In the case of porous piezoceramics of 6 mm class, in order to determine its ten
independent effective moduli (c11Eeff Eeff Eeff
,c13 , c33 , e31 , e33 , e15 , ε11
Eeff eff eff eff
, c44 S eff
, ε33
S eff
), it is
enough to solve only five static problems (2.4), (2.5), (2.7), (2.8), (2.10), having
set one of the component S0β , E 0l (β = 1, 2, ..., 6; l = 1, 2, 3) in the boundary
conditions (2.10) not equal to zero k = 1, 2, 3

I.S0β = S0 δ1β , E 0 = 0 ⇒ c1k

Eeff
= σkk  /S0 , e31
eff
= D3  /S0 , (2.14)

II.S0β = S0 δ3β , E 0 = 0 ⇒ ck3

Eeff
= σkk  /S0 , e33
eff
= D3  /S0 , (2.15)

III.S0β = S0 δ4β , E 0 = 0 ⇒ c44

Eeff
= σ23  /S0 , e15
eff
= D2  /S0 , (2.16)

IV.S0 = 0, E 0l = E 0 δ1l ⇒ e15

eff
= − σ13  /E 0 ; ε11
S eff
= D1  /E 0 , (2.17)

V.S0 = 0, E 0l = E 0 δ3l ⇒ e3k

eff
= − σkk  /E 0 ; ε33
S eff
= D3  /E 0 , (2.18)

Thus, as we can see, the boundary value problems (2.4), (2.5), (2.7), (2.8), (2.10)
with (2.12), (2.13) or with (2.14)–(2.18) differ from the usual problems of linear
2 Modelling of Piezocomposites with Mechanical Interface Effects 19

piezoelectricity by the presence of the interface boundary conditions (2.7), (2.8) and
the average quantities < • > , which are typical for the Gurtin–Murdoch model of
surface stresses for nanosized bodies (Chatzigeorgiou et al. 2015; Javili et al. 2013).
Note that complete interface conditions should take into account the coupled
surface electromechanical fields according to the relations that follow from
(Eremeyev and Nasedkin 2017; Nasedkin 2017)
 
L T (n) · [T ] = L T ∇ s · T s , n · [ D] = ∇ s · Ds , x ∈  s (2.19)

T s = c Es · Ss − esT · E s , Ds = es · Ss + ε Ss · E s , E s = −∇ s ϕ (2.20)

Here, however, we use the uncoupled interface conditions (2.7), (2.8), which are
obtained from (2.19), (2.20) with zero surface piezomoduli and dielectric constants:
es = 0, ε Ss = 0.

2.3 Dimensionless Homogenization Problem

s
Dimension analysis shows that the surface stiffness moduli cαβ differ in dimension-
ality from the corresponding values cαβ in the volume. Thus, in the SI system, the
moduli cαβ , as well as the stresses σi j , are measured in N/m2 , whereas the moduli
s
cαβ , as well as the surface stresses σisj , are measured in N/m. Experimental data also
indicate that the values of the corresponding surface and bulk quantities differ in such
a way that they are comparable cαβ ∼ cαβ s
/d with typical spatial sizes d of the order
−9
10 (m) and less. In our problem such a parameter is the specific size of inclusions
or pores, and so we expect that the surface or interface effects will have an influence
only if pores or inclusions are nanoscale.
Then, to ensure the accuracy of the finite element computations due to the small-
ness of the geometric sizes of inclusions, here it is convenient to make a transition
to the dimensionless coordinates and parameters. For example, we can choose the
following characteristic dimensional parameters: the minimum size a of the pores
or inclusions, the stiffness modulus c = c11 E
, and the dielectric permittivity constant
ε = ε11 . Then we can introduce the following notations:
S


x̂ = x/a, û = u/a, ϕ̂ = ϕ/(E d a), E d = c/ε, (2.21)

ĉ E = c E /c, ε̂ S = ε S /ε, ê = eE d /c, ĉs = cs /(ac), (2.22)

s
T̂ = T /c, Ê = E/E d , D̂ = DE d /c, T̂ = T s /(ac). (2.23)
20 A. V. Nasedkin

Thus, problem (2.4), (2.5), (2.7), (2.8), (2.10) can be solved in dimensionless form
for the variables (2.21)–(2.23) marked with the ‘hat’ symbol. Then after solving this
problem we can return back to dimensional quantities.

2.4 Finite Element Modelling

For solving piezoelectric problems (2.4), (2.5), (2.7), (2.8), (2.10) with (2.12), (2.13)
or with (2.14)–(2.18), we can pass to their weak statements and use classical finite
element techniques. Let h be the region filled by the finite element mesh, h ⊆ ,
( j)
h = (1) (2) ( j)
h ∪ h , h ≈  , h = ∪k  , where  is a separate piezoelectric
ek ek

finite element with the number k. Let also h be a finite element mesh from surface
s

elements conformable with the volume mesh h , hs = (1) (2)

h ∩ h , h ≈  ,
s s

h = ∪m s , s is a separate elastic surface finite element with number m and

s em em

the elements sem are the edges of the suitable volume elements ek located on the
interphase boundaries.
We will use the classic Lagrangian or serendipity volume finite elements with
nodal degrees of freedom of displacements and electric potentials. Note that due
to the structure of the surface mechanical fields, for the elements sem we can use
standard shell or plate elements with elastic membrane stresses options, i.e., only
with displacements nodal degrees of freedom. For these elements, we can take a
fictitious unit thickness so that the surface moduli in (2.8) can be determined from
the product of specially defined volume moduli and shell thickness.
For static piezoelectric problems, we can find the approximate solution
{uh ≈ u, ϕh ≈ ϕ} at the finite element mesh h in the form

uh (x) = N Tu (x) · U, ϕh (x) = N Tϕ (x) · , (2.24)

where N Tu is the matrix of the basic or form functions for displacements, N Tϕ is the
row vector of the basic functions for the electric potentials, U is the vector of nodal
displacements and Φ is the vector of nodal electric potentials.
According to usual approach, we approximate the continuum weak statement of
the piezoelectric problem in finite-dimensional spaces related to the system of basic
functions N Tu , N Tϕ . Substituting (2.24) and similar representations for the projecting
functions into the weak statement of the piezoelectric problem for h , we get the
following finite element system

K uu · U + K uϕ · Φ = F u , (2.25)

−K ∗uϕ · U + K ϕϕ · Φ = F ϕ , (2.26)
2 Modelling of Piezocomposites with Mechanical Interface Effects 21

a 
a 
a
Here, K uu = uu , K uϕ =
K ek uϕ , K ϕϕ =
K ek K ek
ϕϕ are the global matrices,
 a 
obtained from the corresponding element matrices by ensemble operation .
The element matrices are given by the formulas

uu = K uu + K uu
K ek ek ek
(2.27)

uu = ∫ B u · c · B u d, K uu = ∫ B su · c · B su d

K ek eT E e ek eT s e
(2.28)
ek sek

ϕϕ = ∫ B ϕ · ε · B ϕ d, K uϕ = ∫ B u · e · B ϕ d
K ek eT S e ek eT T e
(2.29)
ek ek
 s
B ue = L(∇) · N eT
u , B ϕ = ∇ N ϕ , B su = L ∇
e eT e
· A · N eT
u , (2.30)

where N eT eT
u , N ϕ are the matrix and the row vector of approximating basic functions,
respectively, defined at separate finite elements. The vectors F u , F ϕ in (2.25), (2.26)
are obtained from the corresponding boundary conditions (2.10) with (2.12), (2.13)
or with (2.14)–(2.18).
The matrices K uu and K ekuu , defined by the surface stresses, are similar to the
stiffness matrices for surface or interphase elastic membranes. Hence, for imple-
menting the finite element piezoelectric analysis for the bodies with mechanical
surface and interphase effects, it is necessary to have surface or interphase structural
membrane elements along with ordinary solid piezoelectric finite elements.

2.5 Modelling of Representative Volume Elements

The presented approaches were implemented in the finite element software ANSYS
for porous piezoceramic composites. The representative volume element  ˆ in
dimensionless coordinates was taken in a shape of cube divided in L̂ × L̂ × L̂
geometrically equal small cubes, which were 8-node hexahedral structured solid
finite elements SOLID5. As a result, in the grand cube  ˆ there were n 3 hexahedral
ˆ
finite elements of the size l = L̂/n, where n was an integer. Here, we assume the
following dimensional geometric values: L = a L̂, l = al, ˆ a = l, i.e., the parameter
a is the size of a separate finite element, and thus for a nanoporous structure the value
a is the nanosize of smallest pore.
In the obtained two-phase composite, the finite elements of the first phase had
material properties of the original piezoelectric material, and the pores had negligibly
small material moduli. Then the elements with the pore properties were selected by
the simple random method. It can be noted that such model is easy to build but it
does not support the connectivity of the elements of the first phase and does not
reflect the connectivity structure of the elements of the second phase (closed or open
pores). Other methods that support the connectivity of the skeleton consisting of the
22 A. V. Nasedkin

elements of the first phase or supporting the cluster properties for the elements of
the second phase are described in (Nasedkin and Shevtsova 2013; Kudimova et al.
2018a, b; Kurbatova et al. 2018; Nasedkin and Eremeyev 2014; Nasedkin 2015).
For automatic coating of the internal boundaries of the pores, the algorithm similar
to the one described in (Nasedkin et al. 2014; Nasedkin and Kornievsky 2017a, b) has
been used. At first the finite elements with the pore material properties were selected.
Then the external boundaries of the array of elements with the pore properties were
covered by target contact elements TARGE170 using the command TSHAP, QUAD.
Hence, the edges of all porous finite elements lying on the external surfaces of
the set of these elements got covered by four node contact elements (TARGE170
of the shape QUAD). Then the contact elements, located on the external border
of the representative volume, were deleted and the remaining finite elements were
substituted by the four node shell elements SHELL63 with the membrane capabilities.
As a result, all contact boundaries, where piezoelectric structural elements were
in contact with pores, had been coated by the membrane finite elements for the
simulation of the surface stresses effect.
We note that in order to define a membrane element in ANSYS, it is necessary to
m
specify the stiffness moduli cαβ and the thickness h m . As it was noted in (Nasedkin
and Kornievsky 2017a, b), the membrane element in ANSYS can be used as the
interface element, obtained when taking into account conditions (2.7), (2.8), when
s
cαβ = h m cαβ
m
. Therefore, in ANSYS computations, the important values are the
m
products of the stiffness moduli cαβ of the membrane element by its thickness h m ,
m m
and not the values of cαβ and h . separately. In contrast to (Nasedkin and Kornievsky
2017a, b), here, according to the previous section, the representative volume was
defined in dimensionless way where the edge of one finite element was equal to one.
Therefore, the dimensioning by space parameter was equal to the minimal size of
the element of the second phase a. As there is not enough data on the values of
the surface moduli for different materials and this data is often contradictory, then
in accordance with the investigations on other materials in the case of nanoporous
piezoceramics, it was assumed that the surface stiffness moduli were connected to
the volume moduli of piezoceramics by the formula cαβ s
= lc cαβ , where lc = 10−10
(m). Let us also assume that h = a, cαβ = k cαβ , where k s is the dimensionless
m m s

coefficient. Then cαβs

= h m cαβm
= k s acαβ = (k s a/lc )cαβ
s
, i.e., a = lc /k s and thus the
s
coefficient k is inversely proportional to the minimal size of the pores a. Further
s
in numerical experiments with constant surface moduli cαβ the coefficient k s and
the porosity p were varied. As it can be seen, the growth of k s corresponds to the
decrease of the minimal pore size a, where at k s > 1 the value a becomes smaller
than lc = 10−10 (m).
For the shell elements ANSYS SHELL181, it is necessary to ensure the anisotropy
type, which corresponds to the anisotropy of the volume finite element. As it is known,
the material properties of finite elements in ANSYS are set in the local element
coordinate systems O x ek y ek z ek , whereas for the membrane element with number
k the material properties in the plane O x ek y ek are important. For the cubic finite
element with the cube elements, the edges of which are parallel to the coordinate
axes of the main Cartesian coordinate system, the membrane elements can be located
2 Modelling of Piezocomposites with Mechanical Interface Effects 23

Fig. 2.1 Example of representative volume element: (a) all elements; (b) pores; (c) interphase
boundaries

in the planes parallel to the planes O x y, O x z and O yz. Thus, in the case of the
piezoceramic material of 6 mm-class, in order to provide a correspondence between
the volume anisotropy with the anisotropy of the membrane elements, it is necessary
to vary the stiffness moduli of the membrane elements located perpendicular to the
O x y, which can be implemented by the way of the permutation of the initial stiffness
coefficients by the rows and columns, corresponding to the axes z nd y.
One example of the representative volumes is the representative volume, which
consists of 20 × 20 × 20 elements (n = 20)or the porosity p = 5%, is shown in
Fig. 2.1, where on the left the whole volume is shown (Fig. 2.1a); the middle figure
illustrates the elements of the second phase (Fig. 2.1b); and on the right the interface
membrane elements are shown (Fig. 2.1c). At that the membrane elements located
perpendicular to the plane O x y are shown in crimson, and the elements located
parallel to the plane O x y are shown in blue.
At the next stage, problems (2.4), (2.5), (2.7), (2.8), (2.10) were solved in ANSYS
package. Then in the ANSYS postprocessor, the average stresses were calculated
according to (2.11) both over solid and surface finite elements. In the end, using
formulas (2.12)–(2.16) for the averaged stresses and electric fluxes, the effective
moduli of porous piezoceramic materials with surface stresses were obtained.
We emphasize that as the pores were chosen with the help of the random
number generator, every launch of the program will give different distribution of
these elements. Therefore, for different location of the pores, the total area of the
membranes  s will change and the total stiffness of the volume will also change.
However, for a not very large porosity p these effects will not have significant effect
on the values of the effective moduli for each new launch of the program.

2.6 Results and Discussion

As an example we consider nanoporous piezoceramic material (material of 6 mm

class). The skeleton material is the ferroelectric piezoceramics PZT-8 with the
24 A. V. Nasedkin

following bulk moduli: c11 E

= 14.68 · 1010 , c12
E
= 8.108 · 1010 , c13
E
= 8.105 · 1010 ,
c33 = 13.17 · 10 , c44 = 3.135 · 10 (N/m ), e31 = −3.875, e33 = 13.91,
E 10 E 10 2

e15 = 10.34 (c/m2 ), ε11 S

= 904ε0 , ε33S
= 561ε0 , ε0 = 8.85 · 10−12 (F/m). For
the pores, the material constants (marked by the superscript p) were chosen equal to
; eiα = χ eiα ; χ = 1 · 10−10 ; ε11 = ε33 = ε0 .
Ep p Sp Sp
the following values: cαβ = χ cαβ E

For the analysis of the influence of the porosity and surface effect on the effective
moduli, we have carried out the calculations of the effective moduli for a fixed
number of the elements n l = 20 along the axes of the representative volume, but
for various porosity p and various but not very large values of the multiplier k s .
The results of the calculations are presented in Fig. 2.2, 2.3, 2.4 and 2.5. Here and
after r (...) denotes the relative values of the effective properties, with respect to the
corresponding values of the moduli for zero porosity. For example, r (c11 ) = c11 eff
/c11 ,
eff
where c11 is the effective stiffness modulus for the nanoporous PZT ceramics, which
takes the surface stress into account, c11 = c11 E
is the value of the stiffness modulus
for an ordinary material of piezoceramics and so on. The curves 1 correspond to the

eff (a) and ceff (b) versus porosity

Fig. 2.2 Dependencies of the relative effective moduli c11 33

eff (a) and ceff (b) versus porosity

Fig. 2.3 Dependencies of the relative effective moduli c12 13
2 Modelling of Piezocomposites with Mechanical Interface Effects 25

eff (a) and area of interface A p (b) versus

Fig. 2.4 Dependencies of the relative effective modulus c44
porosity

eff versus porosity

Fig. 2.5 Dependencies of the relative effective piezomoduli eiα

case k s = 0, the curves 2 correspond to the case k s = 0.1, the curves 3 correspond
to the case k s = 0.5, and the curves 4 correspond to the case k s = 1.
As these figures illustrate, for relatively small values of the multiplier k s (curves
1 and 2) the surface stresses have relatively small influence on the stiffness moduli.
However, for any porosity the surface stresses increase the effective stiffness of
porous material. Moreover, here there is the possibility of the cases, when the
nanoporous material will have greater stiffness than the corresponding dense mate-
rial. Such situation can be observed for k s = 1 for the stiffness moduli c11 eff eff
, c33 , and
c44 with the porosity p ≤ 33% Meanwhile, the stiffness moduli c12 and c13
eff eff eff
do not
exceed the values c12 = c12 and c13 = c13 correspondingly, even when k = 1.
E E s

As it was noted in (Nasedkin and Kornievsky 2017a, b; Duan et al. 2006, 2008;
Eremeyev 2016; Eremeyev and Morozov 2010) and other works, the porosity and
the surface stresses have opposite effects on the effective stiffness. It is obvious that
the porosity growth leads to the decrease in the stiffness moduli, whereas the surface
26 A. V. Nasedkin

Fig. 2.6 Dependencies of the relative effective dielectric permittivities εiieff (a) and piezomodulus
eff (b) versus porosity
d33

stresses increase the stiffness. For the nanoporous materials the surface stresses
will grow considerably for large values of k s , which is equivalent to the very small
nanosized of pores. Moreover, for nanoporous materials initial porosity growth will
lead to the growth of the area of the boundaries of the pores with surface stresses.
For example, from the graph in Fig. 2.4b, it can be seen that for the considered cubic
structure of the representative volume the areas of the interface boundaries grow
fast at the beginning, and then their growth becomes smaller with the maximum
at p = 50%. The trends noted above can explain the phenomenon of the stiffness
moduli increase in the nanomaterials with not very large porosity and very small
nanosized of the pores [see (Nasedkin and Kornievsky 2017a, b; Duan et al. 2006,
2008; Eremeyev 2016; Eremeyev and Morozov 2010)].
Meanwhile, the uncoupled surface stresses have much less effect on the effective
eff eff eff
piezomoduli e31 , e33 , and e15 (Fig. 2.5) and almost do not influence on the dielectric
permittivities κ11 and κ33 (Fig. 2.6a).
eff eff

As it can be seen from Fig. 2.5, the influence of the surfaces stresses is significant
mostly for the relative values of the transverse piezomodulus r (e31 ), whereas the
curves for other piezomoduli and dielectric permittivities (Fig. 2.6a) for different
multipliers k s (0≤k s ≤1) almost coincide and look like one curve both in Fig. 2.5b
and Fig. 2.6a.
It is obvious that the effective piezomoduli and dielectric permittivities should
change more significantly, if instead of uncoupled mechanical surface stresses we
take into account fully or partially coupled electromechanical surface effects using
the models (Eremeyev and Nasedkin 2017; Nasedkin and Eremeyev 2014; Nasedkin
2017).
eff
It is interesting to note that the behaviour of the thickness piezomodulus d33 for
the porous piezoceramic with surface stresses (Fig. 2.6b, curves 2–4), considerably
differ from the behaviour of the same piezomodulus for ordinary porous piezoce-
ramic (Fig. 2.6b, curve 1). Whereas for ordinary porous piezoceramic, the thickness
eff
piezomodulus d33 almost does not depend on the porosity (Nasedkin and Shevtsova
2 Modelling of Piezocomposites with Mechanical Interface Effects 27

2011, 2013; Getman and Lopatin 1996; Hikita et al. 1983; Rybyanets 2011), the
fact of taking the surface stresses into account leads to the decrease of effective
eff
piezomodulus d33 , which starts to increase only for highly porous piezoceramics.
The described methodology could be also applied for mixed anisotropic nanos-
tructured composites with other connectivity types for different physico-mechanical
fields, such as poroelastic, thermoelastic, magnetoelectric or magnetoelectroelastic
and other types of nanocomposites. In the obtained two-phase composite, the finite
elements of the first phase had material properties of the original piezoelectric mate-
rial, and the pores had negligibly small material moduli. Then the elements with the
pore properties were selected by the simple random method. It can be noted that
such model is easy to build but it does not support the connectivity of the elements
of the first phase and does not reflect the connectivity structure of the elements of
the second phase (closed or open pores). Other methods that support the connec-
tivity of the skeleton consisting of the elements of the first phase or supporting the
cluster properties for the elements of the second phase are described in (Nasedkin
and Shevtsova 2013; Kudimova et al. 2018a, b; Kurbatova et al. 2018; Nasedkin
2015; Nasedkin et al. 2014).

2.7 Conclusion

In the paper, the problems of the effective properties determination for the two phase
piezoelectric composite with surface (interface) stresses on the interphase bound-
aries were considered. The applied interface conditions describe the size effects for
nanostructured composites. The presented approach can be used for porous elec-
troelastic materials, when the second phase is the set of pores, which are simulated
as piezoelectric materials with negligibly small stiffness moduli and piezomodulil.
The model for the calculation of the effective moduli of nanoporous ferroelectric
piezoceramics PZT-8 was developed. The homogenization problems were solved
numerically in finite element package ANSYS for the representative volume with
uniform meshing in hexahedral piezoelectric elements with random porosity. In the
constructed volume at the interface boundaries additionally, the elastic membrane
elements were generated, which simulated the conditions at the interface.
For highly porous piezoelectric materials, the presented results should be
corrected, due to the use of the model of the representative volume with stochastic
porosity, which does not support the connectivity of the skeleton elements and does
not reflect the connectivity structure for the second phase (closed or open porosity).
Further investigations can be connected with the analysis of the influence of the
representative volume connectivity structures on the effective moduli of nanoporous
piezoelectric materials. Such analysis can be carried out similarly to the one presented
in (Nasedkin 2019; Nasedkin and Shevtsova 2011, 2013; Chatzigeorgiou et al. 2015;
Kudimova et al. 2018a, b) for the composites without interface effects . Other actual
28 A. V. Nasedkin

problems are the problems of the development of surface electromechanical coupled

finite elements, which can be used to take into account the surface interface effects
in more details in the homogenization problems for nanostructured piezoelectric
composites.

Acknowledgements This research was supported by the Ministry of Science and Higher Education
of the Russian Federation, project No. 9.1001.2017/4.6.

References

Alibert JJ, Seppecher P, dell’Isola F (2003) Truss modular beams with deformation energy de-
pending on higher displacement gradients. Math Mech Solids 8(1)
Barchiesi E, dell’Isola F, Laudato M, Placidi L, Seppecher P (2018) A 1D continuum model for
beams with pantographic microstructure: asymptotic micro-macro identification and numerical
results. In: Advances in mechanics of microstructured media and structures. Springer, Cham, pp
43–74
Camar-Eddine M, Seppecher P (2001) Non-local interactions resulting from the homogenization
of a linear diffusive medium. Comptes Rendus L’académie Sci-Ser I-Math 332(5):485–490
Chatzigeorgiou G, Javili A, Steinmann P (2015) Multiscale modelling for composites with energetic
interfaces at the micro-or nanoscale. Math Mech Solids 20(9):1130–1145
Chen T (2008) Exact size-dependent connections between effective moduli of fibrous piezoe-lectric
nanocomposites with interface effects. Acta Mech 196:205–217
Cricrì G, Perrella M, Sessa S, Valoroso N (2015) A novel fixture for measuring mode III toughness
of bonded assemblies. Eng Fract Mech 138:1–18. https://doi.org/10.1016/j.engfracmech.2015.
03.019
Dai S, Gharbi M, Sharma P, Park HS (2011) Surface piezoelectricity: size effects in nanostructures
and the emergence of piezoelectricity in non-piezoelectric materials. J Appl Phys 110:104305–
104307
dell’Isola F, Seppecher P, Alibert JJ, Lekszycki T, Grygoruk R, Pawlikowski M, Steigmann D,
Giorgio I, Andreaus U, Turco E, Gołaszewski M, Rizzi N, Boutin C, Eremeyev VA, Misra A,
Placidi L, Barchiesi E, Greco L, Cuomo M, Cazzani A, Della Corte A, Battista A, Scerrato D,
Eremeeva IZ, Rahali Y, Ganghoffer JF, Müller W, Ganzosch G, Spagnuolo M, Phaff A, Barcz K,
Hoschke K, Neggers J, Hild F (2019a) Pantographic metamaterials: an example of mathematically
driven design and of its technological challenges. Continuum Mech Thermodyn 31(4):851–884
dell’Isola F, Seppecher P, Spagnuolo M, Barchiesi E, Hils F, Lekszycki T, Giorgio I, Placidi L,
Andreaus U, Cuomo M, Eugster SR, Pfaff A, Hoschke K, Langkemper R, Turco E, Sarikaya R,
Misra A, De Angelo M, D’Annibale F, Bouterf A, Pinelli X, Misra A, Desmorat B, Pawlikowski M,
Dupuy C, Scerrato D, Peyre P, Laudato M, Manzari L, Göransoon P, Hesch C, Hesch S, Franciosi
P, Dirrenberger J, Maurin F, Vangelatos Z, Grigoropoulos C, Melissinaki V, Farsari M, Muller W,
Abali BE, Liebold C, Ganzosch G, Harrison P, Drobnicki R, Igumnov L, Alzahrani F, Hayat T
(2019b) Advances in pantographic structures: design, manufacturing, models, experiments and
image analyses. Continuum Mech Thermodyn 31(4):1231–1282
dell’Isola F, Placidi L (2011) Variational principles are a powerful tool also for formulating field
theories. In: Variational models and methods in solid and fluid mechanics. Springer, Vienna, pp
1–15
dell’Isola F, Seppecher P, Placidi L, Barchiesi E, Misra A (2020) 8 least action and virtual work
principles for the formulation of generalized continuum models. Discret Contin Models Complex
Metamaterials 327
2 Modelling of Piezocomposites with Mechanical Interface Effects 29

Duan HL, Wang J, Karihaloo BL, Huang ZP (2006) Nanoporous materials can be made stiffer than
non-porous counterparts by surface modification. Acta Mater 54:2983–2990
Duan HL, Wang J, Karihaloo BL (2008) Theory of elasticity at the nanoscale. Adv Appl Mech
42:1–68
Eremeyev VA (2016) On effective properties of materials at the nano- and microscales con-sidering
surface effects. Acta Mech 227:29–42
Eremeyev V, Morozov N (2010) The effective stiffness of a nanoporous rod. Dokl Phys 55(6):279–
282
Eremeyev VA, Nasedkin AV (2017) Mathematical models and finite element approaches for nano-
sized piezoelectric bodies with uncoupled and coupled surface effects. In: Sum-batyan, M.A.
(ed.) Wave dynamics and composite mechanics for microstructured ma-terials and metamaterials.
series advanced structured materials, vol 59. Springer, Singapore, pp 1–18
Fang XQ, Zhu CS, Liu JX, Liu XL (2018) Surface energy effect on free vibration of nano-sized
piezoelectric double-shell structures. Phys B 529:41–56
Getman I, Lopatin S (1996) Theoretical and experimental investigation of the porous PZT ceramics.
Ferroelectrics 186:301–304
Gu ST, Qin L (2014) Variational principles and size-dependent bounds for piezoelectric inhomo-
geneous materials with piezoelectric coherent imperfect interfaces. Int J Eng Sci 78:89–102
Gu ST, Liu JT, He QC (2014a) Piezoelectric composites: Imperfect interface models, weak
formulations and benchmark problems. Comp Mater Sci 94:182–190
Gu ST, Liu JT, He QC (2014b) The strong and weak forms of a general imperfect inter-face model
for linear coupled multifield phenomena. Int J Eng Sci 85:31–46
Gu ST, He QC, Pensee V (2015) Homogenization of fibrous piezoelectric composites with general
imperfect interfaces under anti-plane mechanical and in-plane electrical loadings. Mech Mater
88:12–29
Hikita K, Yamada K, Nishioka M, Ono M (1983) Piezoelectric properties of the porous PZT and
the porous PZT composite with silicone rubber. Ferroelectrics 49:265–272
Huang GY, Yu SW (2006) Effect of surface piezoelectricity on the electromechanical be-haviour
of a piezoelectric ring. Phys Status Solidi B 243(4):R22–R24
Javili A, McBride A, Mergheima J, Steinmann P, Schmidt U (2013) Micro-to-macro transitions for
continua with surface structure at the microscale. Int J Solids Struct 50:2561–2572
Kudimova AB, Nadolin DK, Nasedkin AV, Oganesyan PA, Soloviev AN (2018) Finite element
homogenization models of bulk mixed piezocomposites with granular elastic inclusions in
ACELAN package. Mater Phys Mech 37(1):25–33
Kudimova AB, Nadolin DK, Nasedkin AV, Nasedkina AA, Oganesyan PA, Soloviev AN (2018)
Models of porous piezocomposites with 3–3 connectivity type in ACELAN finite element
package. Mater Phys Mech 37(1):16–24
Kurbatova NV, Nadolin DK, Nasedkin AV, Oganesyan PA, Soloviev AN (2018) Finite element
approach for composite magneto-piezoelectric materials modeling in ACELAN-COMPOS
package. In: Altenbach, H, Carrera, E, Kulikov, G. (eds.) Analy-sis and modelling of advanced
structures and smart systems. Series advanced struc-tured materials, vol 81. Springer, Singapore,
pp 69–88
Malakooti MH, Sodano HA (2013) Multi-inclusion modeling of multiphase piezoelectric compos-
ites. Compos Part B 47:181–189
Milton GW, Seppecher P, Bouchitté G (2009) Minimization variational principles for acoustics,
elastodynamics and electromagnetism in lossy inhomogeneous bodies at fixed frequency. Proc R
Soc A Math Phys Eng Sci 465(2102):367–396
Nasedkin AV (2015) Finite element design of piezoelectric and magnetoelectric composites with
use of symmetric quasidefinite matrices. In: Parinov IA, Chang S-H, Theerakulpisut S (eds)
Advanced materials–studies and applications. Nova Science Publ, NY, pp 109–124
Nasedkin AV (2017) Some homogenization models of nanosized piezoelectric composite materials
of types ceramics—pores and ceramics—ceramics with surface effects. In: Güemes A, Benjeddou
A, Rodellar J, Leng J (eds) VIII ECCOMAS thematic conf. on smart structures and materials,
30 A. V. Nasedkin

VI Int. conf. on smart materials and nano-technology in engineering—SMART 2017. CIMNE,

Barcelona, Spain, pp 1137–1147
Nasedkin AV (2019) Analysis of surface stress influence on the effective properties of na-noporous
piezocomposites. Probl Prochnosti I Plast-Probl Strength Plast 81(1):5–18
Nasedkin AV, Eremeyev VA (2014) Harmonic vibrations of nanosized piezoelectric bodies with
surface effects. ZAMM 94:878–892
Nasedkin AV, Kornievsky AS (2017) Finite element modeling and computer design of ani-sotropic
elastic porous composites with surface stresses. In: Sumbatyan MA (ed.) Wave dynamics
and composite mechanics for microstructured materials and metamaterials. Series advanced
structured materials, vol 59. Springer, Singapore, pp 107–122
Nasedkin AV, Kornievsky AS (2017) Finite element modeling of effective properties of elastic mate-
rials with random nanosized porosities. Vycisl. meh. splos. Sred–Computa-tional. Continuum
Mech 10(4):375–387
Nasedkin AV, Shevtsova MS (2011) Improved finite element approaches for modeling of porous
piezocomposite materials with different connectivity. In: Parinov IA (ed) Fer-roelectrics and
superconductors: properties and applications. Nova Science Publ, New York, pp 231–254
Nasedkin AV, Shevtsova MS (2013) Multiscale computer simulation of piezoelectric de-vices with
elements from porous piezoceramics. In: Parinov IA, Chang S-H (eds) Physics and mechanics of
new materials and their applications. Nova Sci-ence Publ, New York, pp 185–202
Nasedkin AV, Nasedkina AA, Remizov VV (2014) Finite element modeling of porous thermoe-
lastic composites with account for their microstructure. Vycisl. meh. splos. sred–Computational.
Continuum Mech 7(1):100–109
Pan XH, Yu SW, Feng XQ (2011) A continuum theory of surface piezoelectricity for nanodielectrics.
Sci China Phys Mech Astron 54:564–573
Park HS, Devel M, Wang Z (2011) A new multiscale formulation for the electromechanical behavior
of nanomaterials. Comput Methods Appl Mech Eng 200:2447–2457
Pideri C, Seppecher P (1997) A second gradient material resulting from the homogenization of an
heterogeneous linear elastic medium. Continuum Mech Thermodyn 9(5):241–257
Placidi L, Misra A, Barchiesi E (2018) Two-dimensional strain gradient damage modeling: a
variational approach. Z Angew Math Phys 69(3):1–19
Placidi L, Barchiesi E, Misra A (2018a) A strain gradient variational approach to damage: a
comparison with damage gradient models and numerical results. Math Mech Compl Syst
6(2):77–100
Placidi L, Rosi G, Barchiesi E (2019) Analytical solutions of 2-dimensional second gradient
linear elasticity for continua with cubic-d 4 microstructure. In: New achievements in continuum
mechanics and thermodynamics. Springer, Cham, pp 383–401
Placidi L, dell’Isola F, Barchiesi E (2020) Heuristic homogenization of Euler and pantographic
beams. In Mechanics of fibrous materials and applications. Springer, Cham, pp 123–155
Rybyanets AN (2011) Porous piezoceramics: theory, technology, and properties. IEEE Trans
Ultrason Ferroelectr Freq Control 58:1492–1507
Serpieri R, Sessa S, Rosati L (2018) A MITC-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220. https://doi.org/10.1016/j.compstruct.2018.02.031
Sessa S, Marmo F, Rosati L (2015) Effective use of seismic response envelopes for reinforced
concrete structures. Earthquake Eng Struct Dynam 44(14):2401–2423. https://doi.org/10.1002/
eqe.2587
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through–the–thickness jacketed shells
for the elasto-plastic analysis of confined composite structures: theory and numerical as-sessment.
Compos B Eng 113:225–242. https://doi.org/10.1016/j.compositesb.2017.01.011
Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018a) Evaluation of the capaci-ty
surfaces of reinforced concrete sections: eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512. https://doi.org/10.1007/s11012-017-0791-1
2 Modelling of Piezocomposites with Mechanical Interface Effects 31

Sessa S, Marmo F, Vaiana N, Rosati LA (2018b) Computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthquake Eng https://doi.
org/10.1080/13632469.2018.1551161
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos B Eng 163:413–423. https://
doi.org/10.1016/j.compositesb.2019.01.018
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force–biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54(9):1451–1469. https://
doi.org/10.1007/s11012-019-00979-4
Spagnuolo M (2020) Circuit analogies in the search for new metamaterials: phenomenology of a
mechanical diode. In: Nonlinear wave dynamics of materials and structures. Springer, Cham, pp
411–422
Spagnuolo M, Franciosi P, dell’Isola F (2020) A Green operator-based elastic modeling for two-
phase pantographic-inspired bi-continuous materials. Int J Solids Struct 188:282–308
Spagnuolo M, Yildizdag ME, Andreaus U, Cazzani AM (2021) Are higher-gradient models also
capable of predicting mechanical behavior in the case of wide-knit pantographic structures? Math
Mech Solids 26(1):18–29
Turco E, dell’Isola F, Misra A (2019a) A nonlinear Lagrangian particle model for grains assemblies
including grain relative rotations. Int J Numer Anal Meth Geomech 43(5):1051–1079
Turco E, Misra A, Sarikaya R, Lekszycki T (2019b) Quantitative analysis of deformation mecha-
nisms in pantographic substructures: experiments and modeling. Continuum Mech Thermodyn
31(1):209–223
Turco E, Barchiesi E, Giorgio I, dell’Isola, F (2020) A Lagrangian Hencky-type non-linear
model suitable for metamaterials design of shearable and extensible slender deformable bodies
alternative to Timoshenko theory. Int J Non-Linear Mech 123
Vaiana N, Sessa S, Marmo F, Rosati L (2019) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901. https://doi.org/10.1007/s11071-019-05022-5
Vaiana N, Capuano R, Sessa S, Marmo F, Rosati L (2021) Nonlinear dynamic analysis of seismi-cally
base-isolated structures by a novel OpenSees hysteretic material model. Appl Sci 11(3):900
Wang Z, Zhu J, Jin XY, Chen WQ, Zhang Ch (2014) Effective moduli of ellipsoidal particle
reinforced piezoelectric composites with imperfect interfaces. J Mech Phys Solids 65:138–156
Wang KF, Wang BL, Kitamura T (2016a) A review on the application of modified contin-uum
models in modeling and simulation of nanostructures. Acta Mech Sin 32:83–100
Wang W, Li P, Jin F, Wang J (2016b) Vibration analysis of piezoelectric ceramic circular nanoplates
considering surface and nonlocal effects. Compos Struct 140:758–775
Xiao JH, Xu YL, Zhang FC (2011) Size-dependent effective electroelastic moduli of pie-zoelectric
nanocomposites with interface effect. Acta Mech 222:59–67
Zhao D, Liu JL, Wang L (2016) Nonlinear free vibration of a cantilever nanobeam with surface
effects: semi-analytical solutions. Int J Mech Sci 113:184–195
Chapter 3
A Mathematical Model for Bone Cell
Population Dynamics of Fracture
Healing Considering the Effect of Energy
Dissipation

Mahziyar Darvishi, Hooman Dadras, Mohammad Mahmoodi Gahrouei,

Kiarash Tabesh, and Dmitry Timofeev

Abstract The importance of mechanical modeling has been increasing in recent

years for almost every area of biological sciences. The process of bone recovery
is one of the issues to be addressed within a mechanical framework. In this study,
a model for the bone healing process is proposed taking into account the bone cell
population as well as the effect of energy dissipation. Numerical simulations for bone
under a cyclic external loading are performed in order to show predicting capabilities
of the model.

Keywords Bone tissue · Bone healing · Cell population dynamics · Bone

remodeling · Bone fracture healing · Energy dissipation · Mechanical modeling

3.1 Introduction

Developing mathematical models for biological phenomena is nowadays one the

most exciting research areas for scientists working on the field of modern applied
mathematics. A model accurate enough can be a very powerful tool in understanding
and anticipating functions of a biological system. In this study, we will focus on the
bone capacity to repair itself when it is fractured, i.e. bone healing. Accurate descrip-
tion of this process is important for medical applications such as bone surgery or bone
substitutes like prostheses, since it helps to avoid patient dependency (Sheidaei et al.
2019).
There are many factors which have to be taken into account for a proper description
of the bone healing process. Such factors as, for instance, an interaction between
osteoblasts and osteoclasts, and signal transmission through osteocytes are widely

M. Darvishi · H. Dadras · M. Mahmoodi Gahrouei · K. Tabesh

Department of Mechanical Engineering, Isfahan University of Technology, Isfahan, Iran
D. Timofeev (B)
Dipartimento di Ingegneria e Scienze dell’Informazione e Matematica, Universitá degli Studi
dell’Aquila, L’Aquila, Italy
e-mail: [email protected]
© Springer Nature Switzerland AG 2021 33
F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_3
34 M. Darvishi et al.

addressed in the literature (Klein-Nulend et al. 2012; Baiotto and Zidi 2004). But
effects of osteophytes as well as stem cell responsibilities in the process have not
been properly discussed yet. The more accurate and comprehensive the model is,
the better one understands the healing process in bones. In this study, a model to
describe healing process is presented improving the work in Giorgio et al. (2017d)
by considering effects of energy dissipation as well as a width of a fracture gap.
Nowadays, bio-material implants in bone for big fractures are widely used in
orthopedic, jaw, skull surgery, and dental implantation (Giorgio et al. 2017c) to
replace a missing bone or to treat bone fractures. However, small cracks and gaps in
a bone tissue are naturally being fixed during the process of healing, since the new
bone cells would refill the gaps without any use of implanted materials and scaffolds.
Approximately 10–20% of bone mass is water. A significant proportion (30–40%)
of the dry weight is constituted by organic components of an extra-cellular matrix
and the remaining part (60–70%) consists of inorganic mineral salts such as micro-
crystalline hydroxyapatite (Della Corte et al. 2019). Each bone cell has a specific
effect on the healing process. Osteoblasts are derived from mesenchymal stem pro-
genitors that are located close to the bone surface. They are growing cells which
synthesize and secrete organic components of the matrix. Osteocytes act as a kind of
mechanical sensors controlling calcium and phosphate levels in the microenviron-
ment. They detect exerted mechanical forces and transform them into a biological
activity (mechano-transduction). Osteocytes are in control of synthesis/resorption
activities in bone tissues by sending a proper signal to adjacent cells. This is essential
in establishing a connection between mechanical stimuli and bone healing. Osteo-
clasts are giant, motile cells with multiple nuclei and they play a major role in matrix
resorption during bone growth and remodeling (Lekszycki and dell’Isola 2012). All
of these factors have to be taken into account in order to develop a proper formal
description of bone formation and resorption. The formulation of our model can be a
link between what happens in a bone at the cellular level, and the macro-mechanical
properties of a bone tissue (Della Corte et al. 2019).
The modeling problem for the mechanical behavior of living tissues has been
confronted by many researchers. In this context, we can refer to the pioneering work
by Wolff (1892). In recent years, the discovery of osteocytes led us to develop better
and more advanced bone-tissue models. Bone healing studies are divided into the
following two categories: the studies that consider the bone tissue at macro-scale
and micro-level, respectively. Models assuming the linear Cauchy isotropic material
were the simplest mechanical models at the macro-scale point of view which were
capable to explain dry bone tissue under a constant load in a short period of time
(Della Corte et al. 2019). A time-dependent theory for the remodeling in the case of
a bone subjected to the daily loading history was presented using this concept (Lek-
szycki and dell’Isola 2012). In 2011, a continuum poro-elastic mixture model was
presented including an interaction between bone tissues and biocompatible materi-
als (Andreaus et al. 2015). That model considered bone as a porous material and
suggested an explanation of a stimulus in its formulation (Lekszycki and dell’Isola
2012). Also, a second gradient, two-solid, continuum mixture with variable masses
was proposed to describe effects of microstructure on mechanically driven bone
3 A Mathematical Model for Bone Cell Population Dynamics … 35

remodeling mixed with bio-resorbable material and it is shown that the microstruc-
tural effects on the overall process of remodeling both natural bone and bio-material
may be non-negligible (Madeo et al. 2012). The bone tissue synthesis and resorp-
tion were explained in a mixture model of bone tissue and bio-resorbable material
including mass density. Later a new model was developed by Andreaus et al. (2015)
providing a range of bio-mechanical parameters for residual regions of bio-materials
which remain in remodeled bone. Another model investigated the mixture model of
Lekszycki and dell’Isola (2012) with a biological stimulus of nonlocal nature playing
a role in the feedback control process which governs the generation, transmission,
and the effect on great relevance as the problem of the model (Giorgio et al. 2019).
In Park and Lakes (1986), Yang and Lakes (1982), the micromorphic continua
was used for saturated bone. One of the important works which has been done at the
micro-scale, is adding a rotational degree of freedom (Park and Lakes 1986).
Computational costs for numerical simulations at micro-scale are immense, so if
the characterization of a representative elementary volume (REV) would be used in
homogenization procedures, the method would be handy. Moreover, there is no need
to propose any specific functions in this method to represent the amount of available
surface for bone cells (Della Corte et al. 2019). We also refer to Marmo et al. (2020),
Vaiana et al. (2019, 2021b), Serpieri et al. (2018), Sessa et al. (2018b), Greco et al.
(2018), and Marmo et al. (2018c) for examples of alternative ways to handle the
computational costs. Applications in civil engineering may also be found in Marmo
et al. (2018a), Paradiso et al. (2019), Marmo et al. (2019, 2018b), Perricone et al.
(2020), Vaiana et al. (2017), Cricrì et al. (2015), Sessa et al. (2015), Sessa et al.
(2019b, 2018a), Sessa et al. (2017), Sessa et al. (2019a), Vaiana et al. (2021a).
Bone is well known as one of the organs of our body which is responsible for
damping and energy absorption. The soft collagen gives bones their elasticity and
the ability to dissipate energy under mechanical deformation, and toughening mech-
anisms are active during crack propagation rather than crack initiation (Ritchie et al.
2009). In An et al. (2014), a bone-like material is studied which is structured by
arranging hard platelets embedded in a soft matrix. In order to consider crack growth
resistance, a model within the thermodynamic framework was developed taking into
account microscopic deformation mechanisms (elastic and plastic deformations as
well as damage). Experimental data presented in that work show a very good agree-
ment with the proposed model. Another description was introduced by Elbanna and
Carlson (2013) who developed a kinetic model with random properties assigned
to fibrils, and that model examined the effect of loading rate using classical Bell’s
theory (Bell 1978). The results showed that large amounts of energy are lost at the
molecular level. In addition, increasing the loading rate, which causes the system to
move away from equilibrium, leads to higher strength and failure resistance (Lieou
et al. 2013). As it is reported in Lanyon and Rubin (1984), Turner (1998), the energy
dissipation may have an important and significant influence on the remodeling pro-
cess. A poro-elastic analysis was performed on a simplified geometric model that was
simulated with ABAQUS by looking into a variation of pressure and fluid velocity
in pores with time and location. The evaluation of the dissipation energy stimulus
demonstrates that it increases with low frequency range and saturates at the higher
36 M. Darvishi et al.

frequencies (Kumar et al. 2011). In Giorgio et al. (2016), the phenomena of resorption
and growth of bone tissue are studied by means of a numerical method based on finite
elements. It considers stimulus as a linear combination of strain energy density and
viscous dissipation. This model also takes into account a dissipation associated with
the micromorphic variable controlling the porosity change. In order to have a better
description for mechanical behavior of bone tissue at different scales, a constitutive
model of bone bio-material is presented by Giorgio et al. (2017a). In this study, the
second gradient model used for accounting size effects in visco-elastic material and
the model is compared with a simpler one. The second gradient causes a delay for
evolution process.
Mechanical properties affect bone healing as well as biological aspects and that’s
the reason why mathematical models in order to have predicting capabilities should
consider both (George et al. 2018). Moreover, it is shown that mechanical contri-
bution is more important at the beginning of bone reconstruction when high strain
energy is developed with low bone density (George et al. 2019). There are also
models having a bio-resorbable material in the gap part, and it is shown that the
final distribution of Young’s modulus does not have significant dependence on dif-
fusion rate for small gaps (Lu and Lekszycki 2017). A continuum model is used to
reproduce the behavior of bone tissue regeneration and growth inside a defect as a
result of LIPUS treatment; both 2D and 3D cases are considered for bone growth
simulations (Scala et al. 2017). In Lu and Lekszycki (2016), one can find models
that introduce constitutive equation for Young’s modulus evolution accounting for
its dependence on nutrient concentration and a variable number of active cells. In
this study, stimulus and energy dissipation are considered as mechanical parameters,
while the three types of bone cells (osteoblast, osteocytes, and osteoclasts) with their
specific functions are biological factors of the model. It will be demonstrated that
the dissipation term has an important effect on stimulus and bone healing. Moreover,
gap size, porosity, and the local density of cells are the other crucial factors of bone
healing in this model. It is worth underlining here that for living tissue (or bone tissue
in our specific case), the effect of microstructure is non-negligible, and it can be cap-
tured by modeling living material as a Cosserat or second-gradient material. Thus,
bone tissue even in the absence of grafted materials can be regarded as metamaterial
(Del Vescovo and Giorgio 2014) and models appropriate for metamaterials can be
fruitfully exploited to describe biological phenomena characterized by the presence
of (lattice) microstructure (Giorgio et al. 2017d, b).
Most of the researchers working in the field of bone fractures and toughness
are acknowledged energy loss and have proven their results with experimental data.
Moreover in this study, the next step will be taken toward advancing the science
of mechanical behavior of bone by considering an energy loss while investigating
the healing of a crack. Therefore, in order to achieve a more accurate prediction
of bone fracture recovery, the energy loss has been introduced in the descriptive
differential equations for bone density changes and its effect has been shown in
the amount of stimulus. We assert that adding this gap size factor and considering
the energy dissipation term both in stimulus and the energy equations made our
model more accurate and appropriate to simulate the natural bone healing process.
3 A Mathematical Model for Bone Cell Population Dynamics … 37

For some historical information on development of higher-gradient continua, we

can refer to dell’Isola et al. (2016), dell’Isola et al. (2014). Patient factors like the
nutritional status of a patient (e.g. lack of calcium, phosphorus, and vitamins C and D),
abnormal growth on the bone called osteophyte (Bednarczyk and Lekszycki 2016),
and pharmacological factors are neglected in this model for the sake of simplicity
(Webb and Tricker 2000).
The paper is organized as follows. First, we express the geometry of our model
and loading conditions. Then, the ODEs system is proposed describing the time
evolution of cells and bone density during the process of healing. The stimulus
function is defined in the next part by considering the energy dissipation. After
that, the mechanical framework and dissipation terms are discussed. Finally, we
demonstrate the value of our work by comparing the results for two cases: in the
presence and absence of energy dissipation. Also, it has been shown how important
are the critical gap size and the dependence of healing to the fracture width. The
results are presented with dynamic loading conditions.

3.2 The Model

3.2.1 The Main Assumptions

In this study, a rectangular 2D specimen of a living bone tissue with initial porosity
ϕ = 0.5 is considered to undergo a cyclic compression loading. In Fig. 3.1, one can
see a schematic for the specimen and boundary conditions. The loading has the
following form:

f x (t) = F0 + F1 sin(wt). (3.1)

The main assumptions accepted by the authors are the following:

• ‘stimulus’ is defined as a scalar field providing a connection between the sensor
and actor cells. It is defined at every point of the bone and depends on the overall
mechanical load and density of sensor cells;
• an actual state of the system at a given time is described by the macro-mechanical
variables just like internal state;
• inertial effects have not been taken into account because of the large time scale.
• elastic properties of the bone tissue are explained by means of an isotropic Cauchy
continuum;
• the precursor cells are indefinitely accessible.
Models in Giorgio et al. (2017d), Della Corte et al. (2019) did not consider any
information about energy dissipation. Thus, they could not predict what is going to
happen if the gap between 2 parts of bone tissue changes. Therefore, the aim of the
paper is to take into account the following features (Rapisarda et al. 2019):
38 M. Darvishi et al.

Fig. 3.1 Initial configuration of bone tissue under a cyclic compressing load (Giorgio et al. 2017a).
Rollers make the bone able to have different widths under this cyclic load (due to the presence of
Poisson ratio)

• energy dissipation;
• a novel parameter to the equation of density evolution to affiliate gap size and bone
tissue healing (Giorgio et al. 2017d).

3.2.2 The Governing Equations

The proposed system of ODE describes the evolution of cells and bone density in
time:
∂ xk
= −βk X k + γbk xb κ(ϕ) (3.2)
∂t

∂ xb
= −βb X b − γbk xb κ(ϕ) + αb S + xk (3.3)
∂t

∂ xc
= −βc X c − γc xc κ(ϕ) + αc S − xk (3.4)
∂t

∂ρ
= (αxb − bxc )H (ϕ) × pw(γ ). (3.5)
∂t
By xk , xb , and xc , densities of osteocytes, osteoblast, and osteoclasts are denoted,
respectively. The terms −βk X k , −βb X b , and −βc X c indicate apoptosis of the cells
which is activated when cell densities reach threshold values denoted by x̃k , x̃b , x̃c :
3 A Mathematical Model for Bone Cell Population Dynamics … 39

xk , if xk > x̃k
Xk = (3.6)
0, if xk  x̃k

xb , if xb > x̃b
Xb = (3.7)
0, if xb  x̃b

xc , if xc > x̃c
Xc = (3.8)
0, if xc  x̃c

Each of the constant parameters βk , βb , and βc are responsible for a fraction of

cells that go into apoptosis. The terms +γbk xb κ(ϕ) and −γbk xb κ(ϕ) are related, and
they are responsible for the differentiation from osteoblasts to osteocytes. The γbk is
assumed to be constant instead. Also, αb and αc are considered as constant parameters
which indicate the fraction of osteoblasts and osteoclasts that are created due to the
stimulus. In Eq. (3.5), there are two quantities, a and b, which represent synthesis
and resorption rates for osteoblast and osteoclast, respectively.
By γ , we denote the evaluation of the overall density in each part of a bone where
a stimulus is defined. The output of a piecewise function pw(γ ) could be either zero
or one according to the value of γ .
The quantity ϕ represents porosity and the function H is utilized in order to
consider the influence of effective porosity on the biological activity of cells. If the
porosity is too big, then there is not enough material for actor cells to stay on. If
the porosity is too low, then there is a lack of free space in pores, to allow for cells’
mobility and deposit.
The function H (ϕ) is defined as follows:

H (ϕ) = H (1 − ρ) = 2(1 − ρ)ρ (3.9)

where ρ is the mass density of the bone.

Finally, S + denotes the positive part of stimulus while S − represents the negative
part.

3.2.3 The Stimulus

The stimulus combination of strain energy and power dissipation in the actual con-
figuration Ct is defined as
  
Ct U (x0 , y0 , t) + (Ds × β)ηxk (x0 , y0 , t) f (x, y, x0 , y0 ) d x0 dy0
S(x, y, t) =  − S0
Ct f (x, y, x 0 , y0 )d x 0 dy0
(3.10)
40 M. Darvishi et al.

||x−x0 ||2 +||y−y0 ||2

where f (x, y, x0 , y0 ) = e− 2D 2 , U = U (x, y, t) is the volume strain-energy
density, S0 is the reference value of stimulus, xk is the density of osteocytes, η is a
constant that represents the capability of osteocytes to amplify mechanical signals,
D measures the range of action of sensor cells, and Ds is the dissipation power.

3.2.4 The Function κ(ϕ)

The function κ(ϕ) deserves special attention. In general, one may suppose that the
function κ(ϕ) can change between Eqs. (3.2)–(3.5) (Rapisarda et al. 2019). But for
the sake of simplicity, we started by considering just one κ(ϕ) function for Eqs. (3.2)–
(3.5). For this function, we used a parabola (same as for the function H ), since if the
bone is too dense or too rare the actor cell activity stops.
The constitutive equation for porosity is (Rapisarda et al. 2019):
ρ
ϕ =1−θ , 0 < x  1. (3.11)
ρmax

For the sake of simplicity, bone density ρ is considered to be normalized with

respect to the maximal value of bone density, and θ is assumed to be 1. Finally, the
relation between density and porosity can be written as (Rapisarda et al. 2019):

ϕ = 1 − ρ. (3.12)

Unless otherwise indicated, κ(ϕ) has the following expression (Rapisarda et al.
2019):
κ(ϕ) = κ(1 − ρ) = 2(1 − ρ)ρ. (3.13)

3.2.5 The Mechanical Framework

It has been assumed that the deformation energy density can be expressed as a
function of the Green-Saint Venant strain tensor, the apparent volume Lagrangian
mass densities, and the position of the considered material particle. The deformation
gradient F, its determinant J , and the Green-Saint Venant strain tensor G are defined
as follows (Rapisarda et al. 2019):

1 T 
F = ∇χ , J = det F, G= F F−I , (3.14)
2

where χ : B → R3 is the placement function.

It is remarkable that the assumed deformation energy does not pertain explicitly to
time and that the mass of the bone tissue changes very slowly in time. As already said,
3 A Mathematical Model for Bone Cell Population Dynamics … 41

the time changing is assumed to be slow enough to avoid considering any inertial
effects, thus the behavior can be considered quasi-static. It is assumed that there
exists a function U , representing the strain energy having the following expression
(Rapisarda et al. 2019):

λ
U (G, ρ, x) = μtr (G 2 ) + (tr (G))2 . (3.15)
2
Here, μ and λ are the Lamé parameters. Since the material is not homogeneous, and
its density changes with time, μ and λ are supposed to depend on t and x. Thus
(Rapisarda et al. 2019),

μ = μ̂(ρ, x), λ = λ̂(ρ, x). (3.16)

For our aims, it will be advantageous to use Young’s modulus and Poisson’s ratio,
related to Lamé parameters by the common relations (Rapisarda et al. 2019):

Y Yν
μ= , λ= , (3.17)
2(1 + ν) (1 + ν)(1 − 2ν)

where Y is the Young modulus and ν is the Poisson ratio. The Young modulus for
the bone tissue is assumed (Rapisarda et al. 2019) as
 c
ρ
Y = Ym . (3.18)
ρmax

Here, ρmax is the maximum possible value for ρ and Ym is the maximum value for
the Young modulus, while c is a constant parameter. The mechanical equilibrium of
the sample is controlled by the balance equations (Rapisarda et al. 2019):
 
∂U
Div T = Div F · = −bext (3.19)
∂G

∂U
T [N ] = F · · N = f ext , (3.20)
∂G
where T is the first Piola stress tensor. Equation (3.19) gives the force −bext acting
on a REV of the sample considered, while Eq. (3.20) gives the force acting on the
surface f ext (with outward unit normal N ). In a nutshell, the proposed model consists
of the system of ODE (Eqs. (3.2)–(3.5)) and balance equations (Eqs. (3.19)–(3.20)),
reminding that the energy density U depends on ρ through Eq. (3.18), while the
stimulus S appearing in Eqs. (3.2)–(3.4) depends on U .
42 M. Darvishi et al.

Table 3.1 Numerical data used for simulations

a=1 b=1 S0 = 0.6 ρ0 = 0.5
xk0 = 1 xb0 = 0.5 xc0 = 0.1 η = 1000
βk = 20 βk = 20 βk = 10 αb = 0.03
γbk = 0.1 γc = 0.1 αc = 0.1 x̃k = 1.1
x̃c = 0.11 x̃b = 0.75 Ym = 1 ν = 0.15
D = 0.02 L=1 l = 0.2 c=2

3.2.6 Numerical Data

The complexity of the phenomena investigated needs a large number of constitu-

tive parameters to be defined in order to model both the mechanical and biological
behavior. Values of the parameters exploited in numerical simulations are presented
in Table 3.1 (Rapisarda et al. 2019).
By Ymb , the maximum possible value for the Young modulus is denoted, ν is the
Poisson ratio, and L and l are the normalized length and the height of the sample of
considered bone tissue. All the parameters are dimensionless.
Validation of these parameters is difficult in the absence of experimental data.
Therefore, the values given here are chosen in order to obtain a qualitative behavior
in the largest possible set of different cases.

3.2.7 Healing of Bone

In order to show that there is a gap in the bone, the following initial values are used:
⎧
⎪
⎨xi0 if 0  x  0.45L
X i = 10 xi0
if 0.45L  x  0.55L (3.21)
⎪
⎩
xi0 if 0.55  x  L
⎧
⎪
⎨ρi0 if 0  x  0.45L
ρi0
ρi = if 0.45L  x  0.55L (3.22)
⎪ 10
⎩
ρi0 if 0.55L  x  L

In this paper, we utilize the same equations for cell population as the previous
model did. But we want to use a specific parameter which leads us to a better and
more realistic result. This new quantity, called γ , enables our model to distinguish a
healthy region of the bone from the one which is occupied by a fracture. Figure 3.2
shows the idea behind this: contribution of γ helps us to increase sensitivity of the
model with respect to different locations of the bone tissue.
3 A Mathematical Model for Bone Cell Population Dynamics … 43

Fig. 3.2 Schematic of operative function γ containing a the healthy bone region, b the both regions
of healthy and fractured part, And c the gap size region. Each circle evaluates the density in each
point and determines if the bone density should increase

When stimulus is calculated in a healthy zone, osteocytes can receive more signals
from adjacent cells since there are a lot of them in an area surrounding the evaluation
point. But when overlap reduces and becomes less than a specific threshold, signals
become less and healing does not occur. This is a biological feature that pushes us
to improve and generalized the previous modeling.
We consider γ as an integral over a ball (or a circle in our 2D case) evaluating
mass of osteocytes within the ball in every evaluation point of stimulus:

 
γ = ρ × (x − x0 )2 + (y − y0 )2 d x0 dy0 . (3.23)

When the point considered has a location within the healthy zone ρ̇, the equation
should remain the same but if the overlap is less than the threshold (γ̄ ) then ρ̇ turns
to zero.
We denote the density of osteocytes within the ball as xk , and x and y are coordi-
nates of the point where the stimulus is assessed in. The integral measures mass of
osteocytes then a piecewise function compares the value estimated with a threshold
to see if this point is located in the healthy zone or not. If the point is within the
healthy zone, ρ̇ should be the same as the equation of previous models, otherwise it
should be zero. The step function is defined as follows:

0 if ρ < γ̄
f = (3.24)
1 if ρ > γ̄

and
γ̄ = ξ × C × A R , (3.25)

where C(kg/m) is the largest value of γ when f function is not used in Eq. (3.5), and
A R is the area of the circle having radius r . The quantity ξ is used to show how much
the gap should be less than the radius to have healing on boundaries. We consider
ξ as a constant value of 23 . Adding γ to the ρ̇ equation should change the previous
model for which if the radius is greater than the gap, healing takes place and when it
is less than the gap, healing does not take place. Also, this function should cause the
44 M. Darvishi et al.

bone to start healing from the boundaries, since bone density there is higher than in
the fractured zone.

3.2.8 Dissipation

The basic kinematic field is the displacement u(X, t)—designated by the components
u i where t is time and X stands for the selected space parametrization, i.e. three
material coordinates for Lagrangian formulation. Then, to include the notion of
deformation, the strain tensor E is employed, and its components are expressed as
(Giorgio et al. 2016)
1
E i j (X, t) = (u i, j + u j,i ). (3.26)
2
Since bone is a hierarchical composite characterized by multiple structures at
different length scales, dissipation can arise from a variety of mechanisms. In this
work, dissipative phenomena occurring in bones are assumed to belong only to bulk
viscosity (Giorgio et al. 2016).
Bulk viscosity: a damping increase with frequency in both saturated and dry bone
was observed; as a result, it is reasonable to attribute the damping not only to the
fluid flow but also to other sources. Thus, due to its composition, trabecular bone
presents dissipation related to phenomena similar to those which have been already
observed to occur in polymers, and to an interstitial fluid flow in bone canaliculi.
The viscous stress can be expressed as (Giorgio et al. 2017a)
1
Tiνj ( Ė) = 2μν ( Ė i j − Ė kk δi j ) + κ ν ( Ė kk δi j ), (3.27)
3
where μν is the shear viscosity and κ ν is bulk viscosity evaluated for trabecular bone.
Consequently, the Rayleigh dissipation function in the solid matrix is (Giorgio et al.
2017a)
2Ds = Tiνj Ė i j . (3.28)

3.2.8.1 Dimensionless Form

A dimensionless form of the considered problem is obtained in order to be utilized for

numerical simulations. It has been done by normalization of the variables involved
in the analysis by reference quantities. In this regard, dimensionless quantities are
denoted with a superimposed tilde. In particular, the stored energy density E is
normalized with respect to the maximum bone stiffness YbMax = 18 GPa and therefore
takes the following form (Giorgio et al. 2016):

ε̃ = ε/YbMax (3.29)
3 A Mathematical Model for Bone Cell Population Dynamics … 45

As a consequence of this normalization, the material parameters related to Young

modulus become (Giorgio et al. 2016)

Ỹ = (ς ∗ )βb (3.30)

being
Ỹ = Y/YbMax , λ̃ = λ/YbMax

K̃ f = K f /YbMax , K̃ 2 = K 2 /(YbMax L 20 ),

where L 0 is a characteristic length, which is assumed to be equal to the length of the

sample. Dimensionless time is introduced as follows:

t˜ = t/tr e f , (3.31)

where a reference time tr e f is defined as the time interval in which the physiological
processes, leading to fulfillment of the phenomenon bone synthesis, are accom-
plished. In this work, the reference time is assumed as tr e f = 6.048 × 105 (s) which
is the specific number of seconds constituting a week. Similarly, by defining the
dimensionless form of the dissipation energy D as follows (Giorgio et al. 2016):

D̃ = (tr e f /YbMax )D, (3.32)

one gets
K̃ D = K D L 20 /(tr e f /YbMax ), K̃ B = K B /(tr e f /YbMax ),

K̃ ς = K ς /(tr e f /YbMax ), μ̃ν = μν /(tr e f /YbMax ),

κ̃ ν = κ ν /(tr e f /YbMax ).

The external applied force τi is normalized following the same criterion with
respect to the stiffness YbMax . The dimensionless mass densities of bone and material
are normalized with respect to ρ̂ = 1800 kg/m3 (Giorgio et al. 2016):

ρ̃ = ρ/ρ̂ = ζb . (3.33)

The dimensionless stimulus can be defined as (Giorgio et al. 2016)

S̃ = S̃x − S̃0 , (3.34)

where S̃0 = S0 /YbMax and S̃x comes from the assumptions ˜s = s /YbMax , D̃s =
(tr e f /Y Max )Ds , and b̃ = β/tr e f .
46 M. Darvishi et al.

3.3 Results and Discussion

In this study, we have developed a mechanobiological model of bone remodeling

in order to achieve an accurate prediction for bone fracture recovery. Moreover, the
effect of gap size, representing a scale of fracture on the bone surface, and energy
dissipation terms have been studied in the healing process.
It can be noticed from Fig. 3.3 that the influence range of osteocytes r can have
a significant effect on an outcome of calculations and it is shown clinically that the
large gap size, showing complete fracture of bone, causes the unsuccessful formation
of callus and failure of fracture healing (Lu and Lekszycki 2015). If we use a small
radius, the healing does not occur (the small blue region shown in the last time step in
Fig. 3.3a represents a very low density), since the circle of influence is totally inside
the gap that has low density. Increasing the radius, we increase the bone density
within the influence region and when the density threshold is reached, the healing
must take place. Additionally, when there is no healing, the mass density in the gap
and Young’s modulus are constants and equal to their initial values. Therefore, the
cyclic loading will cause the bone to be compressed only in the gap part.
Experimental studies are challenging at the level of fracture healing by consid-
ering the performance factor of all cells (dell’Isola et al. 2016). Thus, in this study,
simulation of bone healing in different crack dimensions has been investigated taking
into account energy loss in order to obtain a better prognosis for the healing process.
The energy loss in an organ such as bone has been proven, and many researchers

Fig. 3.3 Mass density in the process of bone remodeling for different time steps with a r = 0.005,
b r = 0.03; r represents non-dimensional radius of the influence range of osteocytes. The range
having a small value (shown in a) causes to a long healing time frame (blue region at t=2 in a which
illustrates a very low density). On the other hand, an appropriate radius (shown in b) makes the
bone healing time frame closer to the experimental observation
3 A Mathematical Model for Bone Cell Population Dynamics … 47

have shown this issue experimentally (Bednarczyk and Lekszycki 2016; George
et al. 2018, 2019; An et al. 2014; Ridha and Thurner 2013; Hosseini et al. 2012).
The remarkable point of this research and remodeling is to show the effects of energy
dissipation using the numerical solution on the process of bone recovery during cyclic
loading as well as its effect on the mechanism of cells affecting bone density changes.
The effect of the dissipation term on the total energy is presented in Fig. 3.4. It is
shown that taking into account the dissipation term, we obtain significantly reduced
total energy of the system. Since the healing process is obviously related to the energy
of our system, the energy has a key role in our modeling and cannot be neglected.
Figure 3.4a shows a sudden increase in the energy at first with respect to what we see
in Fig. 3.4b. In short, the dissipation term changes our model in such a way that the
damping becomes more evident and does not let the energy of our system increase
dramatically.
In this section, the radius of the circle equals 0.005 what is less than the critical
radius. According to Fig. 3.5, the bone tissue density in the middle of the gap in the
presence of γ is always zero, because the density of osteocyte, which is evaluated
over the small circle, is less than the piecewise threshold function. As data suggested
in Fig. 3.5, the density of osteoblasts increases sharply in the first period of 0.2s fol-
lowing an abrupt decrease from 0.2s to 0.5s; then a slight decline was experienced
by osteoblasts until 2s. It was also observed in Maes et al. (2006) that during exper-
imental investigations of fracture healing of knocked out mice, the osteoblast differ-
entiation was reduced. The same phenomenon was observed also in Mayr-Wohlfart
et al (2001). In Fig. 3.5a and Fig. 3.5b, the bone density of osteoblasts is higher than
for other kinds of cells, but in the presence of γ osteoblasts are not activated owing to
loss of osteocytes in that region. In other words, there are no cells to produce incited
signals to activate bone tissue generation. However, there exist osteoblasts. So, the
healing does not take place. On the other hand, due to Fig. 3.5b, in the absence of γ ,
the density of osteocytes increases activating osteoblasts. This leads to the increasing
of bone tissue density which, in turn, means that healing always takes place without
considering the gap size.

3.4 Conclusion

As previously discussed, experimental studies are challenging at the level of fracture

healing by considering the performance factor of all cells. Thus, in this study, sim-
ulations of bone healing in different crack dimensions has been investigated taking
into account energy loss in order to obtain a better prognosis for the healing process.
There is proposed a model which describes the healing process by considering
dissipation terms in energy equation which is capable of detecting different gap size
values by a quantity called γ . This novel quantity γ in the equation for density
evaluation helps us to avoid the healing process to occur for large fracture sizes. It
means that the existing gap size of the fracture should be less than the critical value
in order to activate bone fracture healing. On the other hand, ignoring γ causes to
48 M. Darvishi et al.

Fig. 3.4 Energy distribution: a without dissipation and b with dissipation. From the plots, one can
see that dissipation terms make a significant change in total energy of the bone and make the energy
damping much better, so it cannot be neglected in simulations
3 A Mathematical Model for Bone Cell Population Dynamics … 49

Fig. 3.5 a Cell population at the center point in the presence of γ . b cell population at the center
in the absence of γ . In the presence of γ (shown in a), there are not any cells to produce signals to
activate osteoblast so the density remains constant, but in the absence of γ healing takes place even
from the middle of the fractured zone

have healing for every gap size which is unrealistic. So γ increases the accuracy of
the model for mimicking the natural healing in one’s body. The energy loss in an
organ such as bone has been proven, and many researchers have shown this issue
experimentally. The remarkable point of this research and remodeling is to show
the effects of energy dissipation using the numerical solution on the process of bone
recovery during cyclic loading as well as its effect on the mechanism of cells affecting
bone density changes.

Acknowledgements We thank Dr. Ivan Giorgio for his hints and advise in this research. We are
also grateful to Dr. Alessandro Della Corte for his comments on an earlier version of the manuscript.
50 M. Darvishi et al.

We would like to thank the “International Research Center for the Mathematics and Mechanics of
Complex Systems” (M&MoCS), Universit‘a dell’Aquila, Italy, for all the support, and also Prof.
Mahmoud Kadkhodaei, Mechanical Engineering Department of Isfahan University of Technology,
Iran.

References

An B, Zhao X, Arola D, Zhang D (2014) Fracture analysis for biological materials with an expanded
cohesive zone model. J Biomech 47(10):2244–2248
Andreaus U, Giorgio I, Madeo A (2015) Modeling of the interaction between bone tissue and
resorbable biomaterial as linear elastic materials with voids. Z Angew Math Phys 66(1):209–237
Baiotto S, Zidi M (2004) Theoretical and numerical study of a bone remodeling model: the effect
of osteocyte cells distribution. Biomech Model Mechanobiol 3(1):6–16
Bednarczyk E, Lekszycki T (2016) A novel mathematical model for growth of capillaries and
nutrient supply with application to prediction of osteophyte onset. Z Angew Math Phys 67(4):94
Bell GI (1978) Models for the specific adhesion of cells to cells. Science 200(4342):618–627
Cricrì G, Perrella M, Sessa S, Valoroso N (2015) A novel fixture for measuring mode iii toughness
of bonded assemblies. Eng Fract Mech 138:1–18
Del Vescovo D, Giorgio I (2014) Dynamic problems for metamaterials: Review of existing models
and ideas for further research. Int J Eng Sci 80. https://doi.org/10.1016/j.ijengsci.2014.02.022
Della Corte A, Giorgio I, Scerrato D (2019) A review of recent developments in mathematical
modeling of bone remodeling. Proc Inst Mech Eng, Part H: J Eng Med, 0954411919857599
dell’Isola F, Andreaus U, Placidi L (2014) At the origins and in the vanguard of peri-dynamics, non-
local and higher gradient continuum mechanics. an underestimated and still topical contribution
of gabrio piola. Math Mech Solids 20:887–928. https://doi.org/10.1177/1081286513509811
dell’Isola F, Corte A, Giorgio I (2016) Higher-gradient continua: The legacy of piola, mindlin,
sedov and toupin and some future research perspectives. Math Mech Solids 22. https://doi.org/
10.1177/1081286515616034
Elbanna AE, Carlson JM (2013) Dynamics of polymer molecules with sacrificial bond and hidden
length systems: towards a physically-based mesoscopic constitutive law. PloS one 8(4)
George D, Allena R, Rémond Y (2018) A multiphysics stimulus for continuum mechanics bone
remodeling. Math Mech Complex Syst 6(4):307–319
George D, Allena R, Rémond Y (2019) Integrating molecular and cellular kinetics into a coupled
continuum mechanobiological stimulus for bone reconstruction. Continuum Mech Thermodyn
31(3):725–740
Giorgio I, Andreaus U, Scerrato D, Dell’Isola F (2016) A visco-poroelastic model of functional
adaptation in bones reconstructed with bio-resorbable materials. Biomech Model Mechanobiol
15(5):1325–1343
Giorgio I, Andreaus U, Dell’Isola F, Lekszycki T (2017a) Viscous second gradient porous materials
for bones reconstructed with bio-resorbable grafts. Extreme Mech Lett 13:141–147
Giorgio I, Andreaus U, Lekszycki T, Corte AD (2017b) The influence of different geometries of
matrix/scaffold on the remodeling process of a bone and bioresorbable material mixture with
voids. Math Mech Solids 22(5):969–987
Giorgio I, Andreaus U, Scerrato D, Braidotti P (2017c) Modeling of a non-local stimulus for bone
remodeling process under cyclic load: application to a dental implant using a bioresorbable porous
material. Math Mech Solids 22(9):1790–1805
Giorgio I, Rizzi N, Turco E (2017d) Continuum modelling of pantographic sheets for out-of-plane
bifurcation and vibrational analysis. Proc R Soc A: Math, Phys Eng Sci 473(2207):20170,636
3 A Mathematical Model for Bone Cell Population Dynamics … 51

Giorgio I, Dell’Isola F, Andreaus U, Alzahrani F, Hayat T, Lekszycki T (2019) On mechanically

driven biological stimulus for bone remodeling as a diffusive phenomenon. Biomech Model
Mechanobiol, 1–25
Greco F, Luciano R, Serino G, Vaiana N (2018) A mixed explicit-implicit time integration approach
for nonlinear analysis of base-isolated structures. Ann Solid Struct Mech 10(1):17–29
Hosseini HS, Pahr DH, Zysset PK (2012) Modeling and experimental validation of trabecular bone
damage, softening and densification under large compressive strains. J Mech Behav Biomed
Mater 15:93–102
Klein-Nulend J, Bacabac R, Bakker A (2012) Mechanical loading and how it affects bone cells: the
role of the osteocyte cytoskeleton in maintaining our skeleton. Eur Cells Mater 24:278–291
Kumar C, Jasiuk I, Dantzig J (2011) Dissipation energy as a stimulus for cortical bone adaptation.
J Mech Mater Struct 6(1):303–319
Lanyon LE, Rubin C (1984) Static vs dynamic loads as an influence on bone remodelling. J Biomech
17(12):897–905
Lekszycki T, dell’Isola F (2012) A mixture model with evolving mass densities for describing syn-
thesis and resorption phenomena in bones reconstructed with bio-resorbable materials. ZAMM-J
Appl Math Mech/Zeitschrift für Angewandte Mathematik und Mechanik 92(6):426–444
Lieou CK, Elbanna AE, Carlson JM (2013) Sacrificial bonds and hidden length in biomaterials: a
kinetic constitutive description of strength and toughness in bone. Phys Rev E 88(1):012,703
Lu Y, Lekszycki T (2015) Modeling of an initial stage of bone fracture healing. Continuum Mech
Thermodyn 27(4–5):851–859
Lu Y, Lekszycki T (2016) A novel coupled system of non-local integro-differential equations mod-
elling young’s modulus evolution, nutrients’ supply and consumption during bone fracture heal-
ing. Z Angew Math Phys 67(5):111
Lu Y, Lekszycki T (2017) Modelling of bone fracture healing: influence of gap size and angiogenesis
into bioresorbable bone substitute. Math Mech Solids 22(10):1997–2010
Madeo A, George D, Lekszycki T, Nierenberger M, Rémond Y (2012) A second gradient contin-
uum model accounting for some effects of micro-structure on reconstructed bone remodelling.
Comptes Rendus Mécanique 340(8):575–589
Maes C, Coenegrachts L, Stockmans I, Daci E, Luttun A, Petryk A, Gopalakrishnan R, Moermans
K, Smets N, Verfaillie CM et al (2006) Placental growth factor mediates mesenchymal cell
development, cartilage turnover, and bone remodeling during fracture repair. J Clin Investig
116(5):1230–1242
Marmo F, Masi D, Rosati L (2018a) Thrust network analysis of masonry helical staircases. Int J
Architect Herit 12(5):828–848
Marmo F, Ruggieri N, Toraldo F, Rosati L (2018b) Historical study and static assessment of an
innovative vaulting technique of the 19th century. Int J Architect Herit
Marmo F, Toraldo F, Rosati A, Rosati L (2018c) Numerical solution of smooth and rough contact
problems. Meccanica 53(6):1415–1440
Marmo F, Demartino C, Candela G, Sulpizio C, Briseghella B, Spagnuolo R, Xiao Y, Vanzi I, Rosati
L (2019) On the form of the musmeci’s bridge over the basento river. Eng Struct 191:658–673
Marmo F, Sessa S, Vaiana N, De Gregorio D, Rosati L (2020) Complete solutions of three-
dimensional problems in transversely isotropic media. Continuum Mech Thermodyn 32(3):775–
802
Mayr-Wohlfart U, Kessler S, Puhl W, Günther K, Knöchel W (2001) Bmp-4 of xenopus Laevis
stimulates differentiation of human primary osteoblast-like cells. J Bone Joint Surgery British
83(1):144–147
Paradiso M, Marmo F, Rosati L (2019) Consistent derivation of a beam model from the saint
Venant’s solid model. Int J Solids Struct 159:90–110
Park H, Lakes R (1986) Cosserat micromechanics of human bone: strain redistribution by a hydration
sensitive constituent. J Biomech 19(5):385–397
52 M. Darvishi et al.

Perricone V, Grun T, Marmo F, Langella C, Carnevali MDC (2020) Constructional design of echi-
noid endoskeleton: main structural components and their potential for biomimetic applications.
Bioinspirat Biomimet
Rapisarda AC, Della Corte A, Drobnicki R, Di Cosmo F, Rosa L (2019) A model for bone mechanics
and remodeling including cell populations dynamics. Z Angew Math Phys 70(1):9
Ridha H, Thurner PJ (2013) Finite element prediction with experimental validation of damage
distribution in single trabeculae during three-point bending tests. J Mech Behav Biomed Mater
27:94–106
Ritchie RO, Buehler MJ, Hansma P (2009) Plasticity and toughness in bone. Phys Today
Scala I, Spingarn C, Rémond Y, Madeo A, George D (2017) Mechanically-driven bone remodeling
simulation: Application to lipus treated rat calvarial defects. Math Mech Solids 22(10):1976–1988
Serpieri R, Sessa S, Rosati L (2018) A mitc-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220
Sessa S, Marmo F, Rosati L (2015) Effective use of seismic response envelopes for reinforced
concrete structures. Earthquake Eng Struct Dyn 44(14):2401–2423
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through-the-thickness jacketed shells for
the elasto-plastic analysis of confined composite structures: Theory and numerical assessment.
Compos B Eng 113:225–242
Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018a) Evaluation of the capacity
surfaces of reinforced concrete sections: Eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512
Sessa S, Marmo F, Vaiana N, Rosati L (2018b) A computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthquake Eng, 1–34
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos B Eng 163:413–423
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force-biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54(9):1451–1469
Sheidaei A, Kazempour M, Hasanabadi A, Nosouhi F, Pithioux M, Baniassadi M, Rémond Y, George
D (2019) Influence of bone microstructure distribution on developed mechanical energy for bone
remodeling using a statistical reconstruction method. Math Mech Solids, 1081286519828418
Turner CH (1998) Three rules for bone adaptation to mechanical stimuli. Bone 23(5):399–407
Vaiana N, Spizzuoco M, Serino G (2017) Wire rope isolators for seismically base-isolated
lightweight structures: experimental characterization and mathematical modeling. Eng Struct
140:498–514
Vaiana N, Sessa S, Marmo F, Rosati L (2019) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901
Vaiana N, Capuano R, Sessa S, Marmo F, Rosati L (2021a) Nonlinear dynamic analysis of seismi-
cally base-isolated structures by a novel opensees hysteretic material model. Appl Sci 11(3):900
Vaiana N, Sessa S, Rosati L (2021b) A generalized class of uniaxial rate-independent models for sim-
ulating asymmetric mechanical hysteresis phenomena. Mech Syst Signal Process 146(106):984
Webb J, Tricker J (2000) A review of fracture healing. Curr Orthop 14(6):457–463
Wolff J (1892) Das Gesetz der Transformation der Knochen (The Law of Bone Remodeling, trans-
lated by P. Maquet and R. Furlong, 1986). Springer, Beglin
Yang J, Lakes RS (1982) Experimental study of micropolar and couple stress elasticity in compact
bone in bending. J Biomech 15(2):91–98
Chapter 4
Second Gradient Linear and Nonlinear
Constitutive Models of Architectured
Materials: Static and Dynamic Behaviors

Yosra Rahali, Hilal Reda, Benoit Vieille, Hassan Lakiss,

and Jean-François Ganghoffer

Abstract We provide in this chapter a synthetic overview of homogenization meth-

ods for the setting up of second gradient linear and nonlinear anisotropic continuum
media representative of periodic network materials made of beam-type structural
elements, considering successively static and dynamic aspects in the context and
linear and nonlinear theories. Generalized continuum models accounting for either
nodal rotations or strain gradient contributions emerge from the extended asymptotic
homogenization schemes taking into account either additional nodal degrees of free-
dom or long-range interactions between neighboring representative unit cells. The
strain energy density of the effective continuum is derived relying either on analytic
methods or based on dedicated asymptotic homogenization methods specific to dis-
crete network materials. Strain gradient theories lead to dispersive wave propagation
features that are overlooked by classical elasticity and that reflect measurements.
Based on these models, we study the dispersion of elastic waves propagating in peri-
odic beam networks, considering the pantographic structures, the re-entrant hexago-
nal lattice, the diamond chiral lattice, plain weave fabric, and the 3D hexagonal unit
cell representative of trabecular bone. In the framework of nonlinear strain gradient
theories, different types of waves propagate depending on the degree of nonlinear-
ity; a supersonic mode occurs for a weak nonlinearity, whereas wave propagation
changes from a supersonic to an evanescent subsonic mode for a higher degree of
nonlinearity.

Y. Rahali (B) · B. Vieille

GPM, INSA Rouen, Université de Rouen, Avenue de l’université, Saint-Étienne-du-Rouvray,
France
e-mail: [email protected]
B. Vieille
e-mail: [email protected]
H. Reda · H. Lakiss
Faculty of Engineering, Section III, Lebanese University, Campus Rafic Hariri, Beirut, Lebanon
e-mail: [email protected]
J.-F. Ganghoffer
LEM3. Université de Lorraine, CNRS. 7, rue Félix Savart., 57073 Metz Cedex, France
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 53

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_4
54 Y. Rahali et al.

Keywords Periodic networks · Homogenization · Effective continuum · Second

gradient models · Dispersive waves

4.1 Introduction

The homogenization of network materials with a discrete topology toward an effec-

tive Cauchy continuum has deserved lots of research activities in the last two decades,
following different types of methods (Hubert and Palencia 1992; Bornert et al. 2001;
Jarroudi and Brillard 2001; Bouchitté and Bellieud 2002; Sili 2005; Dos Reis and
Ganghoffer 2010, 2012). Homogenization faces limitations when the wavelength of
the loading or of the strain field becomes comparable to the typical microstructure
size. One shall especially underline that size effects cannot be adequately captured
by the standard first gradient-based elasticity theory (Askes and Aifantis 2011). The
main motivation for scientists in 1960–1970s to develop non-classical continuum
mechanics theories like generalized continua is to extend their validity range from
a mathematical viewpoint beyond the strict assumption of scale separation (Trinh
et al. 2012). As mentioned in Forest (2006), criteria of choice of the homogenization
method include the ability to account for impact of the morphology and distribution
of phases on the material response and to capture scale effects.
Microstructured models or higher order/grade continuum models recourse to addi-
tional kinematic variables or additional intrinsic parameters like internal length scales
to account for the microstructure kinematics at the macrolevel via suitably introduced
macro-fields (Auffray et al. 2015; dell’Isola et al. 2015b). Higher order continuum
theories introducing additional degrees of freedom find their origin in the seminal
contribution of the Cosserat brothers (1909), Toupin (1962) and Mindlin (1964; 1965)
and have been properly formulated in a generalized sense in Germain (1973) using
the virtual power method (Jouan et al. 2014).
Higher order micro-continuum theories have been developed to account for
microstructure effects by introducing additional degrees of freedom like the Cosserat
medium (Cosserat and Cosserat 1896) and the micromorphic medium (Eringen and
Suhubi 1964) or additional higher gradients like the second gradient continuum
(Toupin 1962; Mindlin and Eshel 1968) and associated material constants (Chen
et al. 2004; Edelen 1969; Eringen 2012, 1966; Hadjesfandiari and Dargush 2011;
Lam et al. 2003; Polyzos and Fotiadis 2012; Yang et al. 2002; Forest 1998, 2002;
Kouznetsova et al. 2002; Mindlin 1964, 1965). Strain localization zones observed
in experimental tests (Desrues and Viggiani 2004) cannot be modeled with classi-
cal continuum mechanics models. Several authors developed novel homogenization
procedures to account for the heterogeneous nature of the material at the microlevel
involving a second gradient macroscopic constitutive law (Pideri and Seppecher
1997; Camar-Eddine and Seppecher 2003). In both the linear and nonlinear regimes,
higher order homogenization schemes have been built for architectured materials in
the work of Trinh et al. (2012). The homogenization of nonlinear elastic material
in contact with a set of more rigid nonlinear elastic fibers disposed periodically has
4 Second Gradient Linear and Nonlinear Constitutive Models … 55

been treated in El Jarroudi (2013). In Jouan et al. (2014), the second gradient model
developed by Chambon and co-workers is a specific higher order medium (Chambon
et al. 1996, 1998, 2001), aiming to regularize ill-posed strain localization problems
in soils and for geomaterials.
In Rahali et al. (2015b, 2017), second gradient continuum models of beam net-
works have been developed, considering two alternative formulations originating in
the heuristic homogenization methodology developed by Piola (see Mindlin 1965;
dell’Isola et al. 2015b, a; Alibert et al. 2003; Turco et al. 2017): (i) an analytical
method based on an evaluation of the strain energy density at the microscopic level
and (ii) the extension of the asymptotic expansion method up to the second gradi-
ent of the mesoscopic kinematic variables. Both identification procedures lead to
the same second gradient linear continuum. Three discrete homogenization models
have been developed in Rahali et al. (2017): the so-called simplified Euler–Bernoulli
model considering the local displacement as the sole kinematic variable, the com-
plete Euler–Bernoulli model incorporating both displacement and a local rotation
as kinematic descriptors, and a third model accounting for the lattice curvature and
another microstructural level. The first model ignores nodal rotations and proves
more adequate for tension dominated lattices. The consideration of the nodal rota-
tions in the second model leads to a more accurate model, since it can be used for
both bending dominated lattices (requiring a local microrotation) and tension dom-
inated lattices. The network curvature is incorporated in a third step, considering a
general parameterization of the material points with curvilinear coordinates. This
more general model may consider both lattice curvature and microstructure, or only
the microstructure (without curvature).
The aforementioned homogenization methods and their associated numerical
schemes are versatile enough to handle any lattice, periodical, or not. They lead to the
full set of first and second gradient effective moduli accounting for the full anisotropy
of the effective continuum. Applications have been made to both two-dimensional
(2D) (like pantograph and hexagonal lattice) and three-dimensional (3D) network
materials, including 3D prototype geometries of trabecular bone in the field of biome-
chanics (Giorgio et al. 2017; Madeo et al. 2012). The linearized homogenized model
of pantographic lattices represents an exact second gradient medium (Rahali et al.
2015b), as per the definition given by Germain (1973). Numerical simulations illus-
trate some peculiarities of the obtained continuum models, for instance, localization
of energy due to the strain gradient contributions.
Network materials are most often much more flexible in bending than in traction,
thus they are prone to geometrical nonlinearities. The homogenization methods to
identify first and higher gradient moduli have therefore been extended to account for
large deformations due to geometrical nonlinearities (Reda et al. 2016b, a).
Nonlinear wave propagation in periodic nonlinear lattice materials has been stud-
ied based on the derived homogenized continuum in Reda et al. (2016b). The nonlin-
ear wave propagation analysis is done relying on the identified effective strain energy
density including nonlinear kinematic variables. The ensuing analysis show that sub-
sonic or supersonic modes may propagate within the effective continuum. Subsonic
56 Y. Rahali et al.

waves become evanescent beyond a certain wavenumber, while the supersonic waves
are characterized by frequency increasing with the wavenumber.
Periodic lattice materials exhibit a strong contrast of their microscopic properties
and they constitute a special class of metamaterials, as explained in Del Vescovo and
Giorgio (2014), Hans and Boutin (2008), Boutin and Soubestre (2011), dell’Isola
et al. (2012), Kamotski and Smyshlyaev (2019). They witness heterogeneous defor-
mations and are likely to develop high strain gradients due to the voided matrix
(Forest and Trinh 2011; Hütter 2017). Pantographic sheets are exact strain gradient
materials under conditions of inextensible fibers; since then, most of the energy goes
into the bending mode; this entails that periodic pantographs deserve second gra-
dient continuum models at the continuum level (after homogenization) due to the
dominant flexural response of their fibers. The resistance in torsion at the junction
between both fibers of the pantograph is an essential parameter governing the transi-
tion from a Cauchy-type effective continuum to a second gradient effective medium
description (Coutris et al. 2019). The extension of linear models of metamaterials
toward the large deformation regime preserves the higher gradient effects, see the
recent contributions on the topic (Rahali et al. 2015a; Rokoš et al. 2019). The possi-
ble forms of the strain energy density for generalized pantographic sheets with non
orthogonal or inequivalent fibers has been done recently in Rahali et al. (2015a).
The chapter consists of four main sections: Sect. 4.2 provides a summary of the
developed anisotropic linear continuum higher order models accounting for the rota-
tions at the lattices nodes. An analytical method based on the derivation of the
strain energy density at microlevel is first exposed, followed by an extension of the
asymptotic expansion method up to the second gradient. Based on these models, the
dispersion of elastic waves in periodic beam networks is next analyzed. The propa-
gation of nonlinear waves in homogenized periodic nonlinear beam networks based
on second gradient nonlinear constitutive models deserves Sect. 4.3. The chapter
concludes with a summary of the work in Sect. 4.4.

4.2 First- and Second-Order Effective Moduli of Periodic

Networks

Different methods have been recently developed in Rahali et al. (2015b, 2017),
Giorgio et al. (2018), Abdoul-Anziz and Seppecher (2018) for periodic 2D and 3D
network materials, in order to construct second gradient continua (accounting for
the lattices nodes rotations, surface effects and the network curvature). The imple-
mentation of the discrete homogenization method uses a dedicated code written in
symbolic language.
We summarize, in the following, the main steps for the determination of the first-
and second-order effective moduli of periodic lattices considering the simplified
Euler–Bernoulli scheme.
 4 Second Gradient Linear and Nonlinear Constitutive Models … 57

Homogenization steps

1. Forces applied to a beam b to the end node E.

E b S b  b   
N Eb = b
e . u E − uO (4.1)
L

12E b I b  b⊥   
TEb = e . u E − uO (4.2)
(L b )3
 
2E b I b    
M Eb e3 = 
 2 −3e . u E − u O
b⊥
e3 (4.3)
L b

The subscripts O and E refer to the origin and extremity nodes of the beam, N,
T, E b , lbε , S εb and I εb are, respectively, the normal and transverse forces, Young
modulus, the beam length, the beam section, and the quadratic moment, eb the unit
director along each beam, eb⊥ the normal unit vector, e3 is normal to the planar
lattice, uεO , and uεE the displacements of the two extremity nodes of the beam. As
to notations, vectors and second-order tensors are here and in the sequel denoted
using boldface symbols.
2. Asymptotic expansion of the variables in curvilinear coordinates

• The beam width and length t εb = εt b , lb = L b

• The relative nodal displacement reads
 
∂u0 (λε )
(uεE − uεO ) = ε u1E − u1O − L i δib
∂λi
 ε  (4.4)
∂u (λ )
E
L δ ∂ 2 u0 (λε )
2 2
+ε2 u2E − u2O − L i δib 1 − i ib
∂λi 2 ∂λi2

with λ the curvilinear coordinates, δi ∈ {−1, 0, 1}. The index i ∈ {−1, 0, 1}

indicating the considered axis e1 or e2 .
3. Insert Eq. (4.4) into Eqs. (4.1)–(4.3).
4. The equilibrium of forces and moments in virtual power form
  
T b V̇ + N b U̇ = 0 (4.5)
v i ∈Z2 b∈B R

 
  b b
b · wb + M b · wb + l
MO eb ∧ FEb b −l
· wC e b ∧ FO
b b
· wC =0
O E E 2 2
v i ∈Z2 b∈B R
(4.6)
 58 Y. Rahali et al.

B R refer to the set of beams within the reference unit cell, w the virtual rotation
velocity and F b = N b eb + T b eb⊥ the force exerted on the beam b.
5. The virtual power of internal forces over an elementary cell
    
Pe = TE V̇E − V̇O + N E U̇ E − U̇ O (4.7)
b

with V̇i and U̇i therein the two components of the virtual velocity field
6. Continualization process

lim P = lim ε2 Pe = Pe dλ
ε→0 ε→0 
c∈Z

1 ∂ V̇o (λε ) ∂ U̇o (λε )
= TE1 L i δib + N E1 L i δib
 g b ∂λi ∂λi

L 2 δ 2 ∂ 2 V̇o (λε ) ∂ V̇o (λε )
+ ε TE1 i ib + T 2
L i δib (4.8)
2 ∂λi2 E
∂λi
ε 
L δ ∂ U̇o (λ )
2 2 2
∂ U̇o (λε )
+ N E1 i ib + N 2
L
E i ibδ
2 ∂λi2 ∂λi
 ε ε 
2 L i δib ∂ V̇o (λ ) 2 L i δib ∂ U̇o (λ )
2 2 2 2 2 2
+ ε TE
2
+ NE dV
2 ∂λi2 2 ∂λi2
   
with N E = εN E1 + ε2 N E2 , TE = εTE1 + ε2 TE2 and g is the Jacobian of the trans-
formation (Cartesian to curvilinear). One has to rewrite previous expression in
Cartesian coordinates. In the following, we do not study the couplings terms,
restricting to centrosymmetric networks.
7. Equivalence with a second-order gradient continuum in order to express the stress
and hyperstress tensors

Pi = ((σ − S · ∇) · ∇) · Ḋ d V

     (4.9)
∂ Ḋ ∂ 2 Ḋ
= F ·
q
−H · pq
dV
 ∂xq ∂x p ∂xq

With Ḋ the virtual rate of deformation, σ the Cauchy stress and S the third-order
hyperstress tensor with index symmetry Si jk = Sik j .
8. Calculation of the stress and hyperstress tensors
 
σ = σiq ei ⊗ eq = F q ⊗ eq (4.10)
 
S = Skq p ek ⊗ eq ⊗ e p = H pq ⊗ eq ⊗ e p (4.11)
4 Second Gradient Linear and Nonlinear Constitutive Models … 59

Fig. 4.1 Representative unit cell of some studied structures: a plain weave fabric, b 3D hexagonal
lattice, c Diamond chiral lattice, d re-entrant lattice (θ < 0), e pantograph

2D and 3D examples have been treated in Rahali et al. (2015b, 2017); Reda et al.
(2016c), some of them shown in Fig. 4.1. The method allows to treat any periodic
structure.
In Rahali et al. (2015b), relying on the complete Bernoulli schema (the rotation is
limited to the zero order), it has been shown that the linearized homogenized model
for the pantographic lattice must necessarily be a second gradient continuum. Some
numerical simulations are represented in Fig. 4.2 to illustrate some peculiarities of
the obtained continuum models, especially the ability of strain gradient models to
capture localization of energy in narrow bands.
FE computations performed over an elementary unit cell assess the validity of
the computed homogenized moduli and a satisfactory agreement is obtained with
13% as a maximum percentage of differences for the second-order shear moduli.
The effective properties evaluated by the asymptotic expansion have been verified
from an analytical evaluation of the strain energy density for the lattice, using the
Euler–Bernoulli model.
60 Y. Rahali et al.

Fig. 4.2 a Deformed shape in extension: the scale of colors measures bending energy, b shear
deformation field showing regions of constant shear and transition zones (plotted surface energy
densities are in J.m−2 )

4.2.1 Analytical Method

An analytical method has been developed in Reda et al. (2016a, b, 2017, 2018),
based on an evaluation of the lattice strain energy density in both the linear and non-
linear regimes, to evaluate the effective mechanical properties of a periodic network
materials. This section is focused on the linear regime. One considers a unit cell
made of 2D beam elements. U=(u,v) is the displacement field and θ the rotation, see
in Fig. 4.3.
M, F, and T are the moment, the normal, and transverse forces, respectively,
applied to the extremity nodes. The strain energy WS is expressed as follows (Rahali
et al. 2015b; Reda et al. 2018):

Fig. 4.3 Kinematics of a beam element

4 Second Gradient Linear and Nonlinear Constitutive Models … 61

1 E A i+1 2 1 E I i+1 2
WS = u j+1 − u ij + θ j+1 − θij
2 L    2 L  
Extension Energy Flexion Energy
 2 (4.12)
1 12E I L i+1
+ v i+1
j+1 − v ij − θ j+1 + θij
2 L3 2
  
Shear Energy

In Eq. (4.12), A, E, I and L stand for the beam section, the Young modulus, the
quadratic moment, and the beam’s length, respectively.
After development (Taylor expansion of the kinematic variables) the beam’s strain
energy contributions are written as follow, s is the curvilinear coordinate:
  i 2
1 ∂u i L 2 ∂ 2 u 
Wextension = kl L +
2 ∂s  j 2 ∂s 2  j
⎛ ⎞
⎜  2     2 2 i   2 2 i 2 ⎟
1 ⎜ ⎜ ∂u i ∂u i L ∂ u  L ∂ u  ⎟
⎟
= kl ⎜ L  +2 L   +
2 ⎝ ∂s j ∂s j 2 ∂s j
2 2 
2 ∂s j ⎟ ⎠
        
Linear Energy Coupling Energy SG Energy

Wflexion = 0
  i    2
1 ∂v i L 2 ∂ 2 v  L ∂v i
Wshear = kc L + − 2 
2 ∂s  j 2 ∂s 2  j 2 ∂s j
   2
L 2 ∂ 2 v 
i
1
= kc
2 2 ∂s 2  j
  
SG Energy
(4.13)
Using expressions Eq. (4.13), we distinguish between energy contribution terms asso-
ciated with the first-order gradient denoted as W SL , coupling terms W SC , and second
gradient terms W S2 . The total strain energy takes the form:

W S = W SL + W S2 + W SC (4.14)

One restricts the analysis to periodic structures thus the coupling term vanishes.
One then obtains the Cauchy stress tensor σi j and the hyperstress tensor Si jk
∂ (W s L ) ∂ (W s2 )
σi j =  Si jk = (4.15)
∂ ∂u i ∂x j ∂ ∂ 2
u i j 2
∂xk

Both identification procedures (asymptotic or analytic method) lead to the construc-

tion of the same second gradient linear continuum. The second gradient models
62 Y. Rahali et al.

obtained by the discrete homogenization method, see Sect. 4.1, are the basis for the
analysis (Reda et al. 2016c).

4.2.2 Homogenized Viscoelastic Behavior

Viscoelastic 2D beam elements obeying a Kelvin–Voigt rheological behavior have

been considered in Reda et al. (2016c). The difference with the exposed model in
subsection 1, lies in the expressions of the forces and moments, summarized below:

E b S εb  b  ε  μe Sεb  b  ε 
N Eεb = e · u − u ε
+ εb e · u̇ E − u̇εO
L εb E O
L
12E b I εb    12μe I εb   
TEεb =  3 eb⊥ · uεE − uεO +  3 eb⊥ · u̇εE − u̇εO
L εb L εb
 
εb 2E b I εb   ε ε
 2μe I εb   ε ε

M E e3 =  2 −3e · u E − u O +  2 −3e · u̇ E − u̇ O
b⊥ b⊥
· e3
L εb L εb
.
(4.16)
uε and μe are the velocity vector and the extensional viscosity, respectively. After
straightforward developments, the constitutive law (centrosymmetric lattices)
[]receives the following form:
    .
{σ} = Ae {} + Av 
     
elastic par t viscous par t
   (4.17)
{S} = D {κ} + D v {κ̇}
e
     
elastic par t viscous par t

. .
with σi j , Si jk ,  pq , κ pqr ,  pq , κ pqr , respectively, the stress and hyperstress tensors,
and their conjugated kinematic quantities, namely, the first and second deformation
gradients and their time derivatives, the first and second deformation velocity gradi-
ents. Ae i j pq , D e i jkpqr , Av i j pq , D v i jkpqr are, respectively, the first and second gradient
elasticity and viscosity coefficients.
The main steps to compute the dynamical equilibrium and the characteristic equa-
tion are summarized below:

1. The motion equations in the x j directions

 
∂σi j ∂ 2 Si jk
− = æ∗ ü j , j = 1, 2 (4.18)
∂x j ∂x j ∂xk

ρ∗ is the effective density.

4 Second Gradient Linear and Nonlinear Constitutive Models … 63

2. Write the generalized displacement field, for harmonic planar waves r

U = U e(λ t − i k .r) , V = V e(λ t − i k .r ) (4.19)

U , V are the wave amplitude and λ the complex frequency function, respectively,
and k = (k1 , k2 ) the wave vector.
3. Substitute Eq. (4.19) into Eq. (4.18), delivering
! "
U
[ D(k1 , k2 , ˘)] =0 (4.20)
V

4. The characteristic equation is written as

λ4 + a λ 3 + b λ 2 + c λ + d = 0 (4.21)

5. The roots of Eq. (4.21) are

#
λs (k) = −ζs (k) · ωns (k) ± i · ωns (k) 1 − ζs 2 (4.22)

where s represents the branch type, ωns (k), ωds (k) the natural and damped fre-
quency and ζs the damping factor
# #
ωns (k) = real(λs )2 + imag(λs )2 ; ωds (k) = ωns (k) 1 − ζs 2 ;

real(λs )
ζs = −
ωns

Relying on these expressions, one can represent the dispersion curve, Fig. 4.4.
The results show shifts in the frequency band diagrams due to damping.
Figure 4.6 shows a considerable difference between second gradient media
description with and without the coupling energy term contributions. We analyze
in the next section the acoustic properties of periodic network material undergoing
configuration changes associated to geometrical nonlinearities.

4.2.3 Incremental Scheme

We summarize below the main steps of the discrete homogenization method leading
to the nonlinear response of the homogenized continuum (Elnady et al. 2016).
For each iteration k:
1. Computation of the effective mechanical properties in the linear regime.
2. Computation of the incremental Second Piola–Kirchhoff stress tensor
64 Y. Rahali et al.

Fig. 4.4 a Damping ratio and b Dispersion relation with a direction of propagation θ = π/6 for
re-entrant lattice, Fig. 4.5, μe is given in MPa.sec and ωd in rad/sec

Fig. 4.5 Representative unit

cell of the re-entrant lattice

(k)
ΔS(k)
n = K T, n : ΔE Gn
S

ΔE (k)
Gn is the strain tensor
3. Go to next step if convergence is reached
4. Update Cauchy stress
 (k) 
σ (k) −1
n+1 = J n F n · Sn + ΔSn
(k)
· F nT
 −1    (k)  
= J n F n · S(k) T −1
n · F n + J n F n · ΔSn · F nT (4.23)
     
σ (k)
n Δσ (k)
n
4 Second Gradient Linear and Nonlinear Constitutive Models … 65

Fig. 4.6 Phase velocity for the three modes of propagation for the pyramid shaped unit cell (Reda
et al. 2018). Blue curve: second gradient medium with coupling energy terms. Red curve: second
gradient medium without coupling energy. Green curve: Cauchy medium

5. Repeat steps 1–4 up to the maximum applied strain and curvature over the unit
cell.
The strain energy of the second gradient medium undergoing large deformations can
be expressed in the one-dimensional (1D) context as follows:
 2  3  4  2
u ,x u ,x u ,x u ,x x (4.24)
W = Au ,x + B +C +D +β
2 3 4 2
where β is the factor of the linear second gradient term. The Cauchy stress σ and the
hyperstress tensor S are defined by

∂W ∂W
σ= ∂u
, S= ∂ u (4.25)
∂ ∂x
2
∂ ∂x 2

The wave propagation equation along the x direction

 
∂σ ∂2 S
− 2 = æ∗ ü
∂x  
∂x  (4.26)
SG

Inserting the constitutive law obtained combining Eqs. (4.24) and (4.25) into the
previous Eq. (4.26) leads to
      2   
∂2u ∂2u ∂u ∂2u ∂u ∂4u ∂2u
B 2 + 2C 2 + 3D 2 − β 4 = ρ∗ (4.27)
∂x ∂x ∂x ∂x ∂x ∂x ∂t 2
66 Y. Rahali et al.

Fig. 4.7 Dispersion relation with different values of s with second gradient terms equation for the
a Hexagonal lattice, b Re-entrant lattice, and c Textile Plane weave lattice

A similar problem was investigated in detail in Reda et al. (2016a), when neglecting
the C coefficient in the nonlinear form of the strain energy Eq. (4.24).
Doing the change of variable z = x − (w/k)t, in which w is the frequency and k
the wavenumber, following a simple transformation, PDE Eq. (4.27) is transformed
into an ordinary differential equation (ODE in short) for the new non-dimensional
strain of the wave function; its solution is of the solitary wave type:
 
A A s2 k0
N (z) = − + sn 2 z, s (4.28)
2 2 1 − KE(s) 2
(s)

Second gradient terms in the nonlinear motion equation lead to two different modes:
an evanescent subsonic mode (high values of s) and a supersonic mode (low
values of s). The dispersion relations are pictured in Fig. 4.7 for different values
of the nonlinear parameter s. In the sequel, parameters v and v0 refer to the phase
velocities in the nonlinear and linear effective medium, respectively
The propagation of nonlinear waves in homogenized periodic nonlinear beam
networks based on second gradient nonlinear constitutive models will be analyzed
in the sequel. The essential difference with the previous approach lies in the fact that
the geometrical nonlinearity is incorporated into the beam model.
4 Second Gradient Linear and Nonlinear Constitutive Models … 67

4.3 Wave Propagation Analysis Based on Nonlinear Models

The study of nonlinear elastic waves restrict most of the time in the literature to
the Cauchy-type elasticity theory, involving only the first displacement gradient.
Models based on Cauchy-type theory do not give realistic predictions of dynamical
behaviors such as the dispersion relation, since the Cauchy effective medium lacks
internal length parameters.
The Cauchy medium is by essence non dispersive, and waves propagate indepen-
dently of the wavenumber (Reda et al. 2016b). Experiments show to the contrary that
most waves are dispersive so that each wavenumber travels with a different phase
velocity (Vladimir et al. 2003; Jakata and Every 2008). This explains the success of
gradient-enriched theories in capturing dynamic behaviors overlooked by classical
elasticity. In order to circumvent this drawback, different methods have been devel-
oped (Reda et al. 2016b, a, 2017) to link the dispersive aspects of wave propagation
to gradient elasticity theories in a nonlinear effective medium in both 1D, 2D, and
3D situations.

4.3.1 Strain Energy Density

An analytical method based on the strain energy has been developed in Reda et al.
(2016b) to predict the effective properties of the nonlinear effective medium in the
context of a second gradient theory. The developments exposed in subsection 2 for
the linear case remain valid for the nonlinear regime, only the expressions of the
forces and moments are changed to incorporate nonlinear terms:
  2  2    
ΔU· eb 1 ΔU · eb ΔU. eb⊥ ψc Δψc 2
F = EA + + + + E I z
L 2 L2 L2 2 2
    (4.29)
ΔU· eb ΔU · eb⊥
T = G A −ψc +1 + (4.30)
L L
   
Δψc ΔU· eb ΔU · eb⊥
M = E Iz 1+ + ψc (4.31)
L L L

In previous expressions, Iz is the beam quadratic moment, G A = 12E L2

Iz
the shear
rigidity, ψc the central node rotation of the beam and U the 2D displacement vec-
tor. The symbol Δ (.) denoted the variation of any quantity (.) between both beam
extremities. The expression of the internal deformation energy of a single beam
element is obtained relying on previous expressions as

W s = W s1 + W s 2 + W s 3 (4.32)
68 Y. Rahali et al.

Fig. 4.8 Dispersion relation

for different values of s for
the hexagonal re-entrant
lattice θ = −π6

where W s1 and W s2 are the first and second gradient terms of the strain energy
density for small strains, and W s3 is the nonlinear contribution of the strain energy
density.
The strain energy density leads to

∂ (W s1 + W s3 ) ∂ (W s2 )
σi j =  ; Si jk =  (4.33)
∂ ∂u i ∂x j ∂ ∂ u i j ∂xk 2
2

The equations of motion for a second gradient homogenized medium are then written.
The methodology for determining the acoustic characteristics is exposed into more
details in Reda et al. (2016b).
The obtained second gradient nonlinear continuum is characterized by an evanes-
cent subsonic mode and a supersonic mode, both represented in Fig. 4.8.
Parameter s describes the degree of nonlinearity it accounts for the shape, the
period, and the velocity of waves.

4.4 Conclusion

A synthetic overview of homogenization methods employed to compute the static

and dynamic responses of periodic network materials is exposed in this contribution.
These effective homogenized models can account for additional degrees of free-
dom (like the Cosserat medium) or higher strain gradients, thereby reflecting the
discrete network kinematics. Strain gradient theories lead to dispersive wave prop-
agation reflecting the experimental observations. Since these networks have a low
bending rigidity of the inherent beams, they are prone to geometrical nonlineari-
ties, the impact of which on wave propagation is assessed, leading to different wave
propagation regimes. The powerfulness and versatility of homogenization schemes
4 Second Gradient Linear and Nonlinear Constitutive Models … 69

of network materials shall be exploited in future contributions to exhibit topologies

giving rise to band gaps (Misra and Poorsolhjouy 2016; Nejadsadeghi et al. 2019) and
other interesting phenomena in both the static and dynamic range that metamaterials
exhibit.

References

Abdoul-Anziz H, Seppecher P (2018) Strain gradient and generalized continua obtained by homog-
enizing frame lattices. Math Mech Complex Syst 6(3):213–250
Alibert JJ, Seppecher P, dell’Isola F (2003) Truss modular beams with deformation energy depending
on higher displacement gradients. Math Mech Solids 8(1):51–73
Askes H, Aifantis EC (2011) Gradient elasticity in statics and dynamics: an overview of formula-
tions, length scale identification procedures, finite element implementations and new results. Int
J Solids Struct 48(13):1962–1990
Auffray N, dell’Isola F, Eremeyev VA, Madeo A, Rosi G (2015) Analytical continuum mechanics
à la hamilton-piola least action principle for second gradient continua and capillary fluids. Math
Mech Solids 20(4):375–417
Bornert M, Bretheau T, Gilormini P (2001) Homogénéisation en mécanique des matériaux, Tome
1: Matériaux aléatoires élastiques et milieux périodiques
Bouchitté G, Bellieud M (2002) Homogenization of a soft elastic material reinforced by fibers.
Asymptot Anal 32(2):153–183
Boutin C, Soubestre J (2011) Generalized inner bending continua for linear fiber reinforced mate-
rials. Int J Solids Struct 48(3–4):517–534
Camar-Eddine M, Seppecher P (2003) Determination of the closure of the set of elasticity func-
tionals. Arch Ration Mech Anal 170(3):211–245
Chambon R, Caillerie D, El Hassan N (1996) Etude de la localisation unidimensionnelle à l’aide
d’un modèle de second gradient. Comptes rendus de l’Académie des sciences Série II, Mécanique,
physique, chimie, astronomie 323(4):231–238
Chambon R, Caillerie D, El Hassan N (1998) One-dimensional localisation studied with a second
grade model. Eur J Mech-A/Solids 17(4):637–656
Chambon R, Caillerie D, Matsuchima T (2001) Plastic continuum with microstructure, local second
gradient theories for geomaterials: localization studies. Int J Solids Struct 38(46–47):8503–8527
Chen Y, Lee JD, Eskandarian A (2004) Atomistic viewpoint of the applicability of microcontinuum
theories. Int J Solids Struct 41(8):2085–2097
Cosserat E, Cosserat F (1896) Sur la théorie de l’élasticité. premier mémoire. In: Annales de la
Faculté des sciences de Toulouse: Mathématiques 10(3-4):I1–I116
Cosserat E, Cosserat F (1909) Théorie des corps déformables. A. Hermann et fils
Coutris N, Thompson LL, Kosaraju S (2019) Asymptotic homogenization models for pantographic
lattices with variable order rotational resistance at pivots. J Mech Phys Solids, 103718
Del Vescovo D, Giorgio I (2014) Dynamic problems for metamaterials: review of existing models
and ideas for further research. Int J Eng Sci 80:153–172
dell’Isola F, Seppecher P, Madeo A (2012) How contact interactions may depend on the shape
of Cauchy cuts in nth gradient continua: approach “à la d’alembert”. Z Angew Math Phys
63(6):1119–1141
dell’Isola F, Andreaus U, Placidi L (2015a) At the origins and in the vanguard of peridynamics, non-
local and higher-gradient continuum mechanics: An underestimated and still topical contribution
of gabrio piola. Math Mech Solids 20(8):887–928
dell’Isola F, Giorgio I, Andreaus U (2015b) Elastic pantographic 2d lattices: a numerical analysis
on the static response and wave propagation. Proc Est Acad Sci 64(3):219
70 Y. Rahali et al.

Desrues J, Viggiani G (2004) Strain localization in sand: an overview of the experimental results
obtained in grenoble using stereophotogrammetry. Int J Numer Anal Meth Geomech 28(4):279–
321
Dos Reis F, Ganghoffer J (2010) Discrete homogenization of architectured materials: Implemen-
tation of the method in a simulation tool for the systematic prediction of their effective elastic
properties. Tech Mech 30(1–3):85–109
Dos Reis F, Ganghoffer J (2012) Construction of micropolar continua from the asymptotic homog-
enization of beam lattices. Comput Struct 112:354–363
Edelen DG (1969) Protoelastic bodies with large deformation. Arch Ration Mech Anal 34(4):283–
300
El Jarroudi M (2013) Homogenization of a nonlinear elastic fibre-reinforced composite: a second
gradient nonlinear elastic material. J Math Anal Appl 403(2):487–505
Elnady K, Goda I, Ganghoffer JF (2016) Computation of the effective nonlinear mechanical response
of lattice materials considering geometrical nonlinearities. Comput Mech 58(6):957–979
Eringen AC (1966) A unified theory of thermomechanical materials. Int J Eng Sci 4(2):179–202
Eringen AC (2012) Microcontinuum field theories: I Foundations and solids. Springer Science &
Business Media, Berlin
Eringen AC, Suhubi E (1964) Nonlinear theory of simple micro-elastic solids–i. Int J Eng Sci
2(2):189–203
Forest S (1998) Mechanics of generalized continua: construction by homogenizaton. Le Journal de
Physique IV 8(PR4):Pr4–39
Forest S (2002) Homogenization methods and the mechanics of generalized continua-part 2. Theoret
Appl Mech 28(29):113–144
Forest S (2006) Milieux continus généralisés et matériaux hétérogènes. Presses des MINES
Forest S, Trinh DK (2011) Generalized continua and non-homogeneous boundary conditions in
homogenisation methods. ZAMM-J Appl Math Mech/Zeitschrift für Angewandte Mathematik
und Mechanik 91(2):90–109
Germain P (1973) The method of virtual power in continuum mechanics. part 2: Microstructure.
SIAM J Appl Math 25(3):556–575
Giorgio I, Andreaus U, dell’Isola F, Lekszycki T (2017) Viscous second gradient porous materials
for bones reconstructed with bio-resorbable grafts. Extreme Mech Lett 13:141–147
Giorgio I, Harrison P, dell’Isola F, Alsayednoor J, Turco E (2018) Wrinkling in engineering fabrics:
a comparison between two different comprehensive modelling approaches. Proc R Soc A: Math,
Phys Eng Sci 474(2216):20180,063
Hadjesfandiari AR, Dargush GF (2011) Couple stress theory for solids. Int J Solids Struct
48(18):2496–2510
Hans S, Boutin C (2008) Dynamics of discrete framed structures: a unified homogenized description.
J Mech Mater Struct 3(9):1709–1739
Hubert JS, Palencia ES (1992) Introduction aux méthodes asymptotiques et à l’homogénéisation:
application à la mécanique des milieux continus. Masson
Hütter G (2017) Homogenization of a Cauchy continuum towards a micromorphic continuum. J
Mech Phys Solids 99:394–408
Jakata K, Every A (2008) Determination of the dispersive elastic constants of the cubic crystals
GE, SI, GAAS, and INSB. Phys Rev B 77(17):174,301
Jarroudi ME, Brillard A (2001) Asymptotic behaviour of a cylindrical elastic structure periodically
reinforced along identical fibres. IMA J Appl Math 66(6):567–590
Jouan G, Kotronis P, Collin F (2014) Using a second gradient model to simulate the behaviour of
concrete structural elements. Finite Elem Anal Des 90:50–60
Kamotski IV, Smyshlyaev VP (2019) Bandgaps in two-dimensional high-contrast periodic elastic
beam lattice materials. J Mech Phys Solids 123:292–304
Kouznetsova V, Geers MG, Brekelmans WM (2002) Multi-scale constitutive modelling of hetero-
geneous materials with a gradient-enhanced computational homogenization scheme. Int J Numer
Meth Eng 54(8):1235–1260
4 Second Gradient Linear and Nonlinear Constitutive Models … 71

Lam DC, Yang F, Chong A, Wang J, Tong P (2003) Experiments and theory in strain gradient
elasticity. J Mech Phys Solids 51(8):1477–1508
Madeo A, George D, Lekszycki T, Nierenberger M, Rémond Y (2012) A second gradient contin-
uum model accounting for some effects of micro-structure on reconstructed bone remodelling.
Comptes Rendus Mécanique 340(8):575–589
Mindlin RD (1964) Micro-structure in linear elasticity. Arch Ration Mech Anal 16(1):51–78
Mindlin RD (1965) Second gradient of strain and surface-tension in linear elasticity. Int J Solids
Struct 1(4):417–438
Mindlin RD, Eshel N (1968) On first strain-gradient theories in linear elasticity. Int J Solids Struct
4(1):109–124
Misra A, Poorsolhjouy P (2016) Granular micromechanics based micromorphic model predicts
frequency band gaps. Continuum Mech Thermodyn 28(1–2):215–234
Nejadsadeghi N, Placidi L, Romeo M, Misra A (2019) Frequency band gaps in dielectric granular
metamaterials modulated by electric field. Mech Res Commun 95:96–103
Pideri C, Seppecher P (1997) A second gradient material resulting from the homogenization of an
heterogeneous linear elastic medium. Continuum Mech Thermodyn 9(5):241–257
Polyzos D, Fotiadis D (2012) Derivation of Mindlin’s first and second strain gradient elastic theory
via simple lattice and continuum models. Int J Solids Struct 49(3–4):470–480
Rahali Y, Ganghoffer JF, Chaouachi F, Zghal A et al (2015a) Strain gradient continuum models for
linear pantographic structures: a classification based on material symmetries. J Geometry Symm
Phys 40:35–52
Rahali Y, Giorgio I, Ganghoffer J, dell’Isola F (2015b) Homogenization à la piola produces second
gradient continuum models for linear pantographic lattices. Int J Eng Sci 97:148–172
Rahali Y, Dos Reis F, Ganghoffer JF (2017) Multiscale homogenization schemes for the construction
of second-order grade anisotropic continuum media of architectured materials. Int J Multiscale
Comput Eng 15(1)
Reda H, Rahali Y, Ganghoffer J, Lakiss H (2016a) Nonlinear dynamical analysis of 3d textiles
based on second order gradient homogenized media. Compos Struct 154:538–555
Reda H, Rahali Y, Ganghoffer J, Lakiss H (2016b) Wave propagation analysis in 2d nonlinear
hexagonal periodic networks based on second order gradient nonlinear constitutive models. Int J
Non-Linear Mech 87:85–96
Reda H, Rahali Y, Ganghoffer JF, Lakiss H (2016c) Analysis of dispersive waves in repetitive
lattices based on homogenized second-gradient continuum models. Compos Struct 152:712–728
Reda H, Rahali Y, Ganghoffer JF, Lakiss H (2017) Wave propagation analysis in 2d nonlinear
periodic structures prone to mechanical instabilities. In: Bridging the gap, from microstructure
investigations to multiscale modeling, pp 67–97
Reda H, Karathanasopoulos N, Rahali Y, Ganghoffer JF, Lakiss H (2018) Influence of first to
second gradient coupling energy terms on the wave propagation of three-dimensional non-
centrosymmetric architectured materials. Int J Eng Sci 128:151–164
Rokoš O, Ameen MM, Peerlings R, Geers M (2019) Micromorphic computational homogenization
for mechanical metamaterials with patterning fluctuation fields. J Mech Phys Solids 123:119–137
Sili A (2005) Homogenization of an elastic medium reinforced by anisotropic fibers. Asympt Anal
42(1, 2):133–171
Toupin RA (1962) Elastic materials with couple-stresses. Arch Ration Mech Anal 11(1):385–414
Trinh DK, Janicke R, Auffray N, Diebels S, Forest S (2012) Evaluation of generalized continuum
substitution models for heterogeneous materials. Int J Multiscale Comput Eng 10(6)
Turco E, Giorgio I, Misra A, dell’Isola F (2017) King post truss as a motif for internal structure of
(meta) material with controlled elastic properties. R Soc Open Sci 4(10):171,153
Vladimir E et al (2003) Wave processes in solids with microstructure, vol 8. World Scientific,
Singapore
Yang F, Chong A, Lam DCC, Tong P (2002) Couple stress based strain gradient theory for elasticity.
Int J Solids Struct 39(10):2731–2743
Chapter 5
An Application of Coulomb-Friction
Model to Predict Internal Dissipation
in Concrete

Giuliano Aretusi and Alessandro Ciallella

Abstract To explore the possibility of proper modeling, the dissipation in specimens

made of concrete, a reduced two-degrees-of-freedom model, obtained from a nonlin-
ear micromorphic model is employed in the case of cyclic mono-axial compressive
tests. This model assumes that the nonlinear nature of the mechanical response arises
from the microstructure and specifically from the micro-cracks regularly present in
the cement matrix. The dissipation, therefore, is introduced as Coulomb-type fric-
tion associated with the relative motion of the faces of the micro-cracks. Numerical
simulations showing that the proposed model is fitting to describe some of the known
responses of the concrete are performed.

Keywords Continua with microstructure · Frictional sliding · Concrete-based

material · Energy dissipation under cyclic load

5.1 Introduction

It is astonishing that one of the most used materials in civil engineering, namely the
concrete, is still not utterly understood. Depending on how the term “concrete” is
interpreted, one can say that human constructions are made of concrete for more than
thousands of years. A long way has been covered since the dawn of civil engineering,
and the concrete that at the beginning was crude cement made by crushing and
burning gypsum or limestone nowadays is a more advanced and complex composite
material. As a matter of fact, it is the internal microstructure of the concrete that
bestows to the material its peculiar behavior. The various components that are used

G. Aretusi (B)
MeMoCS - University of L’Aquila, 18 Via Giovanni Gronchi, 67100 L’Aquila, Italy
A. Ciallella
DICEAA - University of L’Aquila, 18 Via Giovanni Gronchi, 67100 L’Aquila, Italy
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 73

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_5
74 G. Aretusi and A. Ciallella

in its recipe, i.e., Portland cement, sand, gravel with different sizes and shapes,
produce a very heterogeneous material whose strength is particularly influenced by
the water/cement ratio and the possibility of using additives that alter the properties
of the final concrete chemically (Misra 1998) and/or mechanically (Scerrato et al.
2016). From a modeling viewpoint, many challenging aspects must be addressed to
properly describe the behavior of such material, foremost among them there are the
nonlinear behavior even in small deformation regime, the dissipation effects due to
the microstructure both for the elastic and plastic regime, the evolution of the cracks
inside it.
The microstructure of the concrete as the first level of conceptualization can be
thought of as a matrix made by the solid phase of the cement paste that collects
within it, distributed in a more or less uniform way, pieces of inert which comprise
a quite large variety of sizes (Contrafatto et al. 2017; Bilotta et al. 2021). Moreover,
the microstructure is still more complex because during the curing for the effect of
the quantity and location of the chemical species and thermal cycles of heating and
cooling (day and night), naturally inside the material originate a certain quantity
of micro-cracks. These micro-cracks, generally speaking, affect the macroscopic
behavior of the concrete. Still, unless the external loads, both mechanical and thermal,
are not sufficiently large, they do not evolve, reducing, in turn, the strength of the
material. To predict the behavior of the concrete, different models in the literature
are exploited. Apart from the standard Cauchy model that decisively does not fit
the purpose, other possibilities have been explored in the past in the framework
of generalized continua. For instance, if one thinks of the stiffer aggregates inside
the cement paste, one of the most straightforward ways to model the concrete is a
Cosserat-type model (see, e.g., Giorgio et al. 2020a; Turco et al. 2019; De Borst and
Sluys 1991; Eremeyev and Pietraszkiewicz 2016; Altenbach and Eremeyev 2014). On
the other hand, the strong heterogeneity can suggest that some long-range interactions
may play a decisive role (Pideri and Seppecher 1997; dell’Isola et al. 2015a, b; Javili
et al. 2019), thus one can use a second gradient continuum (dell’Isola et al. 2017;
Germain 2020; Epstein and Smelser 2020; Jouan et al. 2014; Matsushima et al. 2002)
or rather a higher gradient continuum (Reiher et al. 2017; Alibert et al. 2003; Abali
et al. 2017, 2015). Besides, there is room to treat the concrete as a porous material
for the space left by the air in the curing process and for the micro-cracks (see, e.g.,
Coussy and Monteiro 2008; dell’Isola et al. 2009; Madeo et al. 2013; Gagneux and
Millet 2016; Giorgio et al. 2020b).
On top of that the micro-cracks have a twofold effect on the macroscopic behavior
of the concrete: the interaction between the opposite faces of them in compression—
when they are in contact—and tension—when they are separated, namely open
cracks—produces a lack of symmetry in the mechanical response of bulk mate-
rial; in a cyclic mechanical compressive load these faces can slide each over the
other and produce friction responsible of an internal dissipation (Pensée et al. 2002;
Scerrato et al. 2014; Giorgio and Scerrato 2017). This last point is one of the most del-
icate because modeling the dissipation in mechanics of continuous solids is exceed-
ingly tough (Bersani and Caressa 2020; Altenbach and Eremeyev 2008; Vaiana et al.
5 An Application of Coulomb-Friction Model … 75

2019). Typically, a viscous action is used to deal with dissipation before any dam-
age or plastic deformation occurs. However, albeit this model fits the description
of a creep phenomenon in many cases, it is a rough linearization of the dissipative
actions around a specific functioning point characterized by a fixed frequency of
the external load. In a common transient excitation, many frequencies are involved,
so this approximation is rather poor. On the other hand, the mechanism behind the
dissipation seems to be related to a different cause. In the case of the concrete, it is
most likely originated from the friction that the opposite faces of each micro-crack
experience. Therefore, this kind of dissipation, being of Coulomb-type, is almost
frequency-independent differently from viscous actions. Alternatively, rheological
models are adopted to tackle the problem of dissipation in solid material (Banks et al.
2011; Lekszycki et al. 2017). This aspect sometimes bring to generalizations that
conceive fractional derivatives (Smit and De Vries 1970; Rossikhin and Shitikova
2010; Carcaterra et al. 2015). An approach completely different instead makes use
of a thermodynamical setup by introducing a diffusive internal variable that is meant
to describe the dissipation (see for more details Maugin 2006; Ziegler and Wehrli
1987).
Finally, it is worth mentioning, even though it is beyond the purview of this
paper, the efforts spent by many researchers to face the aspects of the problem
related to plasticity and damage evolution. Herein, we recall some of these conceptual
results, especially selecting from the literature those works based on a variational
formulation. We believe indeed that this approach is always to be preferred because of
its logical consistency and because it allows us to dispense with a needless increase of
hypotheses (see, e.g., Contrafatto and Cuomo 2006; Contrafatto et al. 2012; Cuomo
et al. 2014; Placidi et al. 2018a, b, 2019; Placidi and Barchiesi 2018; Alam and Loukili
2017; Timofeev et al. 2020). Recently, some studies on continuum mechanics were
conducted using material particles acting in a sense like a swarm of robots. Some
algorithms involving these ideas also deal with multi-crack damage showing an
impressive computational efficiency (see, e.g., Della Corte et al. 2017; dell’Erba
2019a, b, 2018; Wiech et al. 2018).
In this context, a small notice of the issues of reinforced concrete also deserves
to be made for the sake of completeness (Kezmane et al. 2017; dell’Isola et al. 2019;
Franciosi et al. 2019; Spagnuolo et al. 2020).
In this work, we focus our attention on the dissipative effects occurring in the
concrete in a small deformation regime to underline the Coulomb nature of this
particular phenomenon. Specifically, we investigate a dissipative generalized model
of the concrete proposed in Scerrato et al. (2014) through numerical simulations in
order to illustrate the rate-independent behavior of the friction.
76 G. Aretusi and A. Ciallella

5.2 A Brief Synopsis of the Employed Model

5.2.1 3D Formulation of a Micromorphic Concrete-Based

Material

In this section, we briefly recall the model used to describe the behavior of a concrete-
based material (see Scerrato et al. 2014). This particular model considers the concrete
as a generalized material, namely a continuum endowed with a microstructure. Two
kinematical descriptors are introduced, a bulk displacement u, which is a vector,
and a scalar quantity representing the effect of the micro-cracks at a macroscopic
level of observations, namely ϕ. The phenomenology that this last variable is aiming
to capture takes into account the contact interaction between the opposite faces
of the typical micro-cracks within the concrete that may slip each over the other,
originating a friction interaction responsible for a loss of energy. In other words, the
scalar quantity ϕ can be interpreted as the magnitude of the total amount of relative
displacement of the micro-cracks lips in a given representative elementary volume
(REV). Naturally, the scalar nature of the variable implies an isotropic hypothesis.
A more general model can be conceived if one wants to consider the orientations
of the micro-cracks, but for achieving this, a tensorial quantity should be employed
instead (Adelaide et al. 2010).
In this framework, the considered energy density per unit of volume Ψ is assumed
as follows:
1  1 1 1 
Ψ (∇u, ϕ) = λ (tr E)2 + 2μ E · E + k1 ϕ 2 + k2 ϕ 3 + k3 ϕ 4 − αϕ I2(d)
2 2 3 4
(5.1)
where it is possible to notice a standard dependence on the classical linearized strain
tensor E = Sym(∇u), i.e., the symmetric part of ∇u, the nonlinear dependence on ϕ
that can induce nonlinear behavior also for the displacement u, which is similar to a
Duffing oscillator, and finally the coupling terms between ϕ and the second invariant
of the deviatoric strain tensor dev E = E − 13 tr E 1, i.e.,

1
I2(d) = tr (dev E dev E) . (5.2)
2
Here, the rationale behind the coupling term is that an exchange of energy between
the motion described by the two kinematic descriptors occurs when a shear bulk
deformation induces a relative movement of the micro-cracks faces or vice versa.
Therefore, the material parameters involved are the Lamé ones for linearly elastic
isotropic materials, i.e., λ and μ, to which k1 , k2 , and k3 are added to consider
the tendency of the micro-cracks faces to prevent the slip each over the other, and
α is the coupling coefficient between the micro-macro scales of the deformation.
All these material parameters can be evaluated with specific techniques as done in
5 An Application of Coulomb-Friction Model … 77

Scerrato et al. (2016), Giorgio et al. (2018), De Angelo et al. (2019), Abali et al.
(2016), Misra and Poorsolhjouy (2015), Bolzon et al. (2002), Rosi et al. (2018).
To introduce the dissipation, a Rayleigh functional R depending on the time rate
of the micro-slip ϕ is considered. Specifically, we use the expression:
 
Log( Cosh( η ϕ̇ ) )
R = ζ trE . (5.3)
η

This particular functional is conceived to obtain a dissipative internal force which

resembles as close as possible a signum function and, therefore, a Coulomb-type
friction force. Indeed, the derivative of R with respect of the velocity ϕ̇, i.e., the
friction force is a hyperbolic tangent, η being the slope of this function near zero,
or a coefficient of smoothness. This means that the dissipation is almost entirely
viscous for low micro-velocities while increasing the micro-velocities, the Coulomb
law becomes more significant. Besides, ζ is a material coefficient that takes into
account the bulk stiffness of the material and the friction. Indeed, the bulk stiffness
multiplied by the trE is nothing but the normal stress that close the micro-cracks and
allows an internal dissipation by friction.
The inertial effects are introduced by postulating a kinetic energy density as fol-
lows:
1 1
K= ρ u̇2 + ρϕ ϕ̇ 2 (5.4)
2 2
where ρ is the apparent mass density at the macro-level and the ρϕ is an effective
mass density related to the microstructure of the micro-cracks. It is worth noting that
for reasons of coherency among units of measure, the parameter ρϕ is also a mass
density.
By applying the generalized principle of virtual work, the equation of motion of
the considered material can be written as
(Elast) (I ner ) (Diss) (E xt)
δW + δW + δW = δW (5.5)

where the elastic and inertial virtual work are

 
(Elast) (I ner )
δW = δΨ dV, δW = ρ ü δu + ρϕ ϕ̈ δϕ dV, (5.6)
V V

the virtual work done by the friction forces is obtained by

  
(Diss) ∂R
δW (E, ϕ, ϕ̇) = δϕ dV = ζ tr E tanh(η ϕ̇) δϕ dV, (5.7)
V ∂ ϕ̇ V

and, finally, the virtual work done by the external force is


(E xt)
δW = bext · δu dV, (5.8)
V
78 G. Aretusi and A. Ciallella

where the integrations are performed on the volume of the considered sample.

5.2.2 Simplified Formulation for the Case of a Pure

Compression

To study a representative case, namely a pure compressive test, the Eq. (5.5) can be
reduced to a set of two ODEs because we partially know the solution. Thus, we can
employ this information to simplify the starting general PDE. As a matter of fact, the
strain tensor and consequently the displacement vector in a pure compression can be
represented in the form
⎛ ⎞ ⎛ ⎞
−νε 0 0 −νεx1
E = ⎝ 0 −νε 0 ⎠ , u = ⎝ −νεx2 ⎠ , (5.9)
0 0 ε εx3

for a cylindrical specimen whose longitudinal axis coincides with x3 -axis of a refer-
ence coordinate frame. For the hypothesis of isotropy, the deformation is independent
from the space along all the sample and can be made explicit as ε = E 33 = du 3 /dx3 .
In this formulation, also the Poisson effect is considered as highlighted by the relative
coefficient ν. The same isotropy allow us to assume uniform with the space also the
micro-displacement field ϕ. By substituting the Eq. (5.9) in the principle of virtual
work (5.5), one obtains the set of two ODEs in the variables ε(t) and ϕ(t) that are
here only time dependent as follows:
⎧
⎨ M ε̈(t) + K ε(t) + α̃ ϕ(t) = F(t)
⎩
m ϕ ϕ̈(t) + k̃1 ϕ(t) + k̃2 ϕ 2 (t) + k̃3 ϕ 3 (t) + α̃ ε(t) − ζ̃ tanh [η ϕ̇(t)] ε(t) = 0
(5.10)
where all the material parameters can be computed by
  √ 
 
K = λ + 2 μ + 4 (λ + μ) ν 2 − 4 λ ν dV, α̃ = 3/3 (1 + ν) α dV,
V V

   
k̃1 = k1 dV, k̃2 = k2 dV, k̃3 = k3 dV, ζ̃ = ζ (1 − 2 ν) dV,
V V V V

 
   
M= ρ ν 2 x12 + x22 + x32 dV, m ϕ = ρϕ dV,
V V

in terms of the previously defined ones. Eventually, the external applied force has been
 b = [0, 0, −b(t)] . Therefore, the generalized
ext T
chosen as a compression, hence
load F(t) = k b(t) being k = V x3 dV .
5 An Application of Coulomb-Friction Model … 79

Table 5.1 Material coefficients used in numerical simulations

K (Nm) α̃ (N) k̃1 (N/m) k̃2 (N/m2 ) k̃3 (N/m3 ) η (s/m) ζ̃ (N)
5.81 × 1010 1.5 × 1011 9.65 × 1011 1.2 × 108 7.78 × 1018 2.0 × 104 7.29 × 109

Finally, we mention that a similar model reduction for the case of bending and
compression has been made in Scerrato et al. (2015), but the complexity added with
the bending deformation seems not to be very significative as the pure compression.

5.3 Numerical Simulations and Discussions

To illustrate the behavior of the investigated model, we solve the Eq. (5.10) deploying
a Matlab code which makes use of the routine ode23s that evaluates the unknown
variables at the next time step using a modified Rosenbrock formula of order 2,
particularly fit to solve stiff problems. The external load is a cyclic sinusoidal load at
a fixed frequency and with a mean value different from zero to assure that the micro-
cracks responsible for the internal dissipation are closed, namely, the opposite faces
are in contact (see Fig. 5.1). Naturally, this is a quite common example of a structural
element when an earthquake occurs. Therefore, this kind of problem is of the utmost
importance in understanding the behavior of the concrete in the presence of cyclic
external loads. As a note, we consider a small initial transient which lasts one period
of the sinus evaluated as a spline (3rd order polynomial) that starts with zero value
and velocity and ends with the same value and derivative of the biased sinus. In other
words, we have a G 1 1 connection between the two pieces of the loading curve. The
reason for this choice lies in the fact that we want to avoid a sudden increase in
force and therefore reducing possibly negative inertial effects. The mass parameters
are roughly estimated as M = 0.4616 kg m2 , m ϕ = 0.5057 kg. The other material
parameters used in the numerical simulations are recorded in Table 5.1.
Specifically, we test the specimen under compression with an external load as
plotted in Fig. 5.1, with different frequencies, namely 0.1, 0.5, 1, 2 Hz. Figure 5.2
shows clearly that all the frequencies but the one at 0.1 Hz are characterized by an
almost coincident cycle and, therefore, by the same dissipation. Indeed, the figure
exhibits the stress-strain curves for the loading and unloading of the sample, and the
area of the cycle represents the dissipated energy at the specific frequency. Thus,
as expected, the investigated model has a threshold behavior that, once exceeded,
results in a dissipation that becomes almost rate-independent as predicted by the
Coulomb-friction model. Before reaching the threshold, the primary behavior of the
system is viscous and is governed by the parameter η, which represents the slope of
the hyperbolic tangent near zero in Eqs. (5.10).

1 Geometric continuity G 1 stands for the continuity of two curves together with their tangents in
the contact point.
80 G. Aretusi and A. Ciallella

7
10
0

-0.5

-1

-1.5
F(t), (Nm)

-2

-2.5

-3

-3.5

-4

0 0.5 1 1.5 2 2.5 3 3.5 4

time, (s)

Fig. 5.1 External mechanical load applied to the sample, F(t)

40 40

35 35

30 30

25 25
, MPa

, MPa

20 20

15 15

10 10

5 5

0 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
, strain , strain
a) b)
40 40

35 35

30 30

25 25
, MPa
, MPa

20 20

15 15

10 10

5 5

0 0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
, strain , strain
c) d)

Fig. 5.2 Dissipative cycles for different frequencies: a 0.1 Hz; b 0.5 Hz; c 1 Hz; d 2 Hz
5 An Application of Coulomb-Friction Model … 81

-200

-400
, strain

-600

-800

-1000

-1200
0 0.5 1 1.5 2 2.5 3 3.5 4
time, (s)

Fig. 5.3 Uniform compression ε(t) for an applied cyclic load at a frequency 1 Hz

From a model viewpoint, this is very useful because by varying the parameter η,
it is possible to increase or decrease the range of frequencies where the Coulomb
behavior is prevalent. In Figs. 5.3 and 5.4, the kinematical descriptors are plotted in
the case 1 Hz. They both show a nonlinear path, even though this more evident in the
case of the micro-crack sliding variable ϕ.
Finally, Fig. 5.5 plot the dissipative internal generalized force ζ̃ tanh [η ϕ̇(t)] ε(t)
also in this case for a cyclic external load 1 Hz. In this plot, we remark that the
characteristic trend of the generalized force is not as one can erroneously imagine
almost a square wave ‘signal’ governed by the smooth version of the function sign
(i.e., hyperbolic tangent), but rather a sawtooth wave. The reason behind this behavior
is the presence of ε(t) that modulates the amplitude of the signal. As a matter of
fact, the friction force depends on the normal stress, in turns proportional to the
compression ε(t), that pushes in contact the opposite faces of the micro-cracks.
Therefore, when the compression is low also the friction force is small in magnitude.

5.4 Conclusion

In this paper, we examine the predictive capabilities of a model conceived to describe

the complex behavior of concrete-based materials as proposed in Scerrato et al.
(2014). The model is quite useful because, regardless of its simplicity is able to
take into account the nonlinear behavior of the material, even in the small regime of
deformation, and its internal dissipation. This last effect is modeled with a Coulomb-
82 G. Aretusi and A. Ciallella

10 -4
1.4

1.2

0.8
, (m)

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4
time, (s)

Fig. 5.4 Uniform micro-crack sliding variable ϕ for an applied cyclic load at a frequency 1 Hz

6
10
8

4
friction force, (N)

-2

-4

-6

-8
0 0.5 1 1.5 2 2.5 3 3.5 4
time, (s)

Fig. 5.5 Dissipative internal generalized force ζ̃ tanh [η ϕ̇(t)] ε(t) for an applied cyclic load at a
frequency 1 Hz
5 An Application of Coulomb-Friction Model … 83

type model of friction and therefore turns to be rate-independent at least when a

certain tunable threshold is exceeded. The presented study is complementary to other
works on concrete (Sessa et al. 2018a, b, 2019b, 2015), with possible applications to
civil engineering (Raeder et al. 2008; Sessa et al. 2019a, 2017; Li et al. 2018; Cricrì
et al. 2015).

References

Abali BE, Müller WH, Eremeyev VA (2015) Strain gradient elasticity with geometric nonlinearities
and its computational evaluation. Mech Adv Mater Modern Process 1(1):1–11
Abali BE, Wu CC, Müller WH (2016) An energy-based method to determine material constants in
nonlinear rheology with applications. Continuum Mech Thermodyn 28(5):1221–1246
Abali BE, Müller WH, dell’Isola F (2017) Theory and computation of higher gradient elasticity
theories based on action principles. Arch Appl Mech 87(9):1495–1510
Adelaide L, Richard B, Ragueneau F, Cremona C (2010) Thermodynamical admissibility of a set of
constitutive equations coupling elasticity, isotropic damage and internal sliding. Comptes rendus
mécanique 338(3):158–163
Alam SY, Loukili A (2017) Transition from energy dissipation to crack openings during continuum-
discontinuum fracture of concrete. Int J Fract 206(1):49–66
Alibert JJ, Seppecher P, Dell’Isola F (2003) Truss modular beams with deformation energy depend-
ing on higher displacement gradients. Math Mech Solids 8(1):51–73
Altenbach H, Eremeyev V (2014) Strain rate tensors and constitutive equations of inelastic microp-
olar materials. Int J Plast 63:3–17
Altenbach H, Eremeyev VA (2008) Analysis of the viscoelastic behavior of plates made of function-
ally graded materials. ZAMM-Zeitschrift für Angewandte Mathematik und Mechanik: Applied
Mathematics and Mechanics 88(5):332–341
Banks HT, Hu S, Kenz ZR (2011) A brief review of elasticity and viscoelasticity for solids. Adv
Appl Math Mech 3(1):1–51
Bersani AM, Caressa P (2020) Lagrangian descriptions of dissipative systems: a review. Math Mech
Solids. https://doi.org/10.1177/1081286520971834
Bilotta A, Causin A, Solci M, Turco E (2021) Representative volume elements for the analysis of
concrete like materials by computational homogenization. In: Mathematical modeling in cultural
heritage. Springer, Berlin, pp 13–35
Bolzon G, Fedele R, Maier G (2002) Parameter identification of a cohesive crack model by Kalman
filter. Comput Methods Appl Mech Eng 191(25–26):2847–2871
Carcaterra A, Roveri N, Pepe G (2015) Fractional dissipation generated by hidden wave-fields.
Math Mech Solids 20(10):1251–1262
Contrafatto L, Cuomo M (2006) A framework of elastic-plastic damaging model for concrete under
multiaxial stress states. Int J Plast 22(12):2272–2300
Contrafatto L, Cuomo M, Fazio F (2012) An enriched finite element for crack opening and rebar
slip in reinforced concrete members. Int J Fract 178(1–2):33–50
Contrafatto L, Cuomo M, Greco L (2017) Meso-scale simulation of concrete multiaxial behaviour.
Eur J Environ Civ Eng 21(7–8):896–911
Coussy O, Monteiro PJ (2008) Poroelastic model for concrete exposed to freezing temperatures.
Cem Concr Res 38(1):40–48
Cricrì G, Perrella M, Sessa S, Valoroso N (2015) A novel fixture for measuring mode iii toughness
of bonded assemblies. Eng Fract Mech 138:1–18
Cuomo M, Contrafatto L, Greco L (2014) A variational model based on isogeometric interpolation
for the analysis of cracked bodies. Int J Eng Sci 80:173–188
84 G. Aretusi and A. Ciallella

De Angelo M, Barchiesi E, Giorgio I, Abali BE (2019) Numerical identification of constitutive

parameters in reduced-order bi-dimensional models for pantographic structures: application to
out-of-plane buckling. Arch Appl Mech 89(7):1333–1358
De Borst R, Sluys L (1991) Localisation in a cosserat continuum under static and dynamic loading
conditions. Comput Methods Appl Mech Eng 90(1–3):805–827
Della Corte A, Battista A, dell’Isola F, Giorgio I (2017) Modeling deformable bodies using discrete
systems with centroid-based propagating interaction: fracture and crack evolution. In: Mathemat-
ical modelling in solid mechanics. Springer, Berlin, pp 59–88
dell’Erba R (2018) Position-based dynamic of a particle system: a configurable algorithm to describe
complex behaviour of continuum material starting from swarm robotics. Continuum Mech Ther-
modyn 30(5):1069–1090
dell’Erba R (2019a) On how swarm robotics can be used to describe particle system’s deformation.
Continuum Mech Thermodyn, 1–21
dell’Erba R (2019b) Swarm robotics and complex behaviour of continuum material. Continuum
Mech Thermodyn 31(4):989–1014
dell’Isola F, Madeo A, Seppecher P (2009) Boundary conditions at fluid-permeable interfaces in
porous media: a variational approach. Int J Solids Struct 46(17):3150–3164
dell’Isola F, Andreaus U, Placidi L (2015a) At the origins and in the vanguard of peridynamics, non-
local and higher-gradient continuum mechanics: an underestimated and still topical contribution
of Gabrio Piola. Math Mech Solids 20(8):887–928
dell’Isola F, Seppecher P, Della Corte A (2015b) The postulations á la D’Alembert and á la Cauchy
for higher gradient continuum theories are equivalent: a review of existing results. Proc R Soc A:
Math, Phys Eng Sci 471(2183):20150,415
dell’Isola F, Della Corte A, Giorgio I (2017) Higher-gradient continua: the legacy of Piola, Mindlin,
Sedov and Toupin and some future research perspectives. Math Mech Solids 22(4):852–872
dell’Isola F, Bragov AM, Igumnov LA, Abali BE, Lomunov AK, Lamzin DA, Konstantinov AY
(2019) Mechanical response change in fine grain concrete under high strain and stress rates. In:
Altenbach H, dell’Isola F, Eremeyev V, Öchsner A (eds) Abali B. New achievements in continuum
mechanics and thermodynamics. Springer, Berlin, pp 71–80
Epstein M, Smelser R (2020) An appreciation and discussion of Paul Germain’s “The method of
virtual power in the mechanics of continuous media, I: second-gradient theory”. Math Mech
Complex Syst 8(2):191–199
Eremeyev VA, Pietraszkiewicz W (2016) Material symmetry group and constitutive equations of
micropolar anisotropic elastic solids. Math Mech Solids 21(2):210–221
Franciosi P, Spagnuolo M, Salman OU (2019) Mean green operators of deformable fiber networks
embedded in a compliant matrix and property estimates. Continuum Mech Thermodyn 31(1):101–
132
Gagneux G, Millet O (2016) Modeling capillary hysteresis in unsatured porous media. Math Mech
Complex Syst 4(1):67–77
Germain P (2020) The method of virtual power in the mechanics of continuous media, I: second-
gradient theory. Math Mech Complex Syst 8(2):153–190
Giorgio I, Scerrato D (2017) Multi-scale concrete model with rate-dependent internal friction. Eur
J Environ Civ Eng 21(7–8):821–839
Giorgio I, Harrison P, dell’Isola F, Alsayednoor J, Turco E (2018) Wrinkling in engineering fabrics:
a comparison between two different comprehensive modelling approaches. Proc R Soc A: Math,
Phys Eng Sci 474(2216):20180,063
Giorgio I, De Angelo M, Turco E, Misra A (2020a) A Biot-Cosserat two-dimensional elastic non-
linear model for a micromorphic medium. Continuum Mech Thermodyn 32:1357–1369
Giorgio I, Spagnuolo M, Andreaus U, Scerrato D, Bersani AM (2020b) In-depth gaze at the astonish-
ing mechanical behavior of bone: a review for designing bio-inspired hierarchical metamaterials.
Math Mech Solids. https://doi.org/10.1177/1081286520978516
Javili A, Morasata R, Oterkus E, Oterkus S (2019) Peridynamics review. Math Mech Solids
24(11):3714–3739
5 An Application of Coulomb-Friction Model … 85

Jouan G, Kotronis P, Collin F (2014) Using a second gradient model to simulate the behaviour of
concrete structural elements. Finite Elem Anal Des 90:50–60
Kezmane A, Chiaia B, Kumpyak O, Maksimov V, Placidi L (2017) 3D modelling of reinforced
concrete slab with yielding supports subject to impact load. Eur J Environ Civ Eng 21(7–8):988–
1025
Lekszycki T, Bucci S, Del Vescovo D, Turco E, Rizzi NL (2017) A comparison between different
approaches for modelling media with viscoelastic properties via optimization analyses. ZAMM-
Zeitschrift für Angewandte Mathematik und Mechanik 97(5):515–531
Li Y, Chen Y, Li T, Cao S, Wang L (2018) Hoberman-sphere-inspired lattice metamaterials with
tunable negative thermal expansion. Compos Struct 189:586–597
Madeo A, dell’Isola F, Darve F (2013) A continuum model for deformable, second gradient porous
media partially saturated with compressible fluids. J Mech Phys Solids 61(11):2196–2211
Matsushima T, Chambon R, Caillerie D (2002) Large strain finite element analysis of a local second
gradient model: application to localization. Int J Numer Meth Eng 54(4):499–521
Maugin GA (2006) On canonical equations of continuum thermomechanics. Mech Res Commun
33(5):705–710
Misra A (1998) Stabilization characteristics of clays using class C fly ash. Transp Res Rec
1611(1):46–54
Misra A, Poorsolhjouy P (2015) Identification of higher-order elastic constants for grain assemblies
based upon granular micromechanics. Math Mech Complex Syst 3(3):285–308
Pensée V, Kondo D, Dormieux L (2002) Micromechanical analysis of anisotropic damage in brittle
materials. J Eng Mech 128(8):889–897
Pideri C, Seppecher P (1997) A second gradient material resulting from the homogenization of an
heterogeneous linear elastic medium. Continuum Mech Thermodyn 9(5):241–257
Placidi L, Barchiesi E (2018) Energy approach to brittle fracture in strain-gradient modelling. Proc
R Soc A: Math, Phys Eng Sci 474(2210):20170,878
Placidi L, Barchiesi E, Misra A (2018a) A strain gradient variational approach to damage: a com-
parison with damage gradient models and numerical results. Mathematics and Mechanics of
Complex Systems 6(2):77–100
Placidi L, Misra A, Barchiesi E (2018b) Two-dimensional strain gradient damage modeling: a
variational approach. Z Angew Math Phys 69(3):1–19
Placidi L, Misra A, Barchiesi E (2019) Simulation results for damage with evolving microstructure
and growing strain gradient moduli. Continuum Mech Thermodyn 31(4):1143–1163
Raeder J, Larson D, Li W, Kepko EL, Fuller-Rowell T (2008) Openggcm simulations for the themis
mission. Space Sci Rev 141(1):535–555
Reiher JC, Giorgio I, Bertram A (2017) Finite-element analysis of polyhedra under point and line
forces in second-strain gradient elasticity. J Eng Mech 143(2):04016,112
Rosi G, Placidi L, Auffray N (2018) On the validity range of strain-gradient elasticity: a mixed
static-dynamic identification procedure. Eur J Mech-A/Solids 69:179–191
Rossikhin YA, Shitikova MV (2010) Application of fractional calculus for dynamic problems of
solid mechanics: novel trends and recent results. Appl Mech Rev 63(1)
Scerrato D, Giorgio I, Madeo A, Limam A, Darve F (2014) A simple non-linear model for internal
friction in modified concrete. Int J Eng Sci 80:136–152
Scerrato D, Giorgio I, Della Corte A, Madeo A, Limam A (2015) A micro-structural model for
dissipation phenomena in the concrete. Int J Numer Anal Meth Geomech 39(18):2037–2052
Scerrato D, Giorgio I, Della Corte A, Madeo A, Dowling N, Darve F (2016) Towards the design of
an enriched concrete with enhanced dissipation performances. Cem Concr Res 84:48–61
Sessa S, Marmo F, Rosati L (2015) Effective use of seismic response envelopes for reinforced
concrete structures. Earthquake Eng Struct Dyn 44(14):2401–2423
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through-the-thickness jacketed shells for
the elasto-plastic analysis of confined composite structures: Theory and numerical assessment.
Compos B Eng 113:225–242
86 G. Aretusi and A. Ciallella

Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018a) Evaluation of the capacity
surfaces of reinforced concrete sections: Eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512
Sessa S, Marmo F, Vaiana N, Rosati L (2018b) A computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthquake Eng, 1–34
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos B Eng 163:413–423
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force-biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54(9):1451–1469
Smit W, De Vries H (1970) Rheological models containing fractional derivatives. Rheol Acta
9(4):525–534
Spagnuolo M, Franciosi P, dell’Isola F (2020) A green operator-based elastic modeling for two-
phase pantographic-inspired bi-continuous materials. Int J Solids Struct 188:282–308
Timofeev D, Barchiesi E, Misra A, Placidi L (2020) Hemivariational continuum approach for
granular solids with damage-induced anisotropy evolution. Math Mech Solids. https://doi.org/
10.1177/1081286520968149
Turco E, dell’Isola F, Misra A (2019) A nonlinear Lagrangian particle model for grains assemblies
including grain relative rotations. Int J Numer Anal Meth Geomech 43(5):1051–1079
Vaiana N, Sessa S, Marmo F, Rosati L (2019) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901
Wiech J, Eremeyev VA, Giorgio I (2018) Virtual spring damper method for nonholonomic robotic
swarm self-organization and leader following. Continuum Mech Thermodyn 30(5):1091–1102
Ziegler H, Wehrli C (1987) The derivation of constitutive relations from the free energy and the
dissipation function. Adv Appl Mech 25:183–238
Chapter 6
From the Swarm Robotics to Material
Deformations

Paolo D’Avanzo, Alessio Ciro Rapisarda, and Salvatore Samuele Sirletti

Abstract In this chapter a discrete 2D kinematic system model is presented. Its aim
is to reproduce the behavior of several 2D continuum systems. We started from studies
on swarm robotics; in these papers, simple interaction laws among the elements of the
swarm are used to manage its behavior. We have employed them to simulate materials
deformation. This model seems to be promising because it is able to qualitatively
reproduce standard deformations and a lot of exotic phenomena that other methods
in literature can’t reproduce. Furthermore, it has a low computational cost and it is
parallelizable, allowing us to take profit of CUDA® architecture. Some numerical
simulations are provided and discussed using two different kinds of lattices and
changing some model’s parameters.

Keywords Discrete mechanical system · Second gradient continua · Elastic law ·

Swarm robotics · Numerical simulations · Exotic phenomena

6.1 Introduction

Simulation of non-standard continua behavior is gaining interest nowadays. This

happens because standard deformation models, like Cauchy, are unable to describe
the behavior of modern composite Casale et al. (1998) and structured materials
Dell’Erba (2001). Therefore, there is the need to find new numerical approaches
allowing us to obtain these exotic behaviors. In the literature, it is possible to find
some models with this purpose, for example, the Position-Based Dynamics (PBD)
(Müller et al. (2005, 2007)) and the swarm robotic model (dell’Erba 2015; Dell’Erba
2020a, 2021a; Şahin 2005; Battista et al. 2018): they are briefly described in this
chapter; our own model is inspired by these models and can obtain the researched
results with lower computational cost. An important aspect that must be considered is
that our model is purely kinematic, i.e., there are not forces or masses; there are only
laws that rule the motion of points. The algorithm developed is parallelizable and

P. D’Avanzo (B) · A. C. Rapisarda · S. S. Sirletti

University of Naples Federico II, Naples, Italy

© Springer Nature Switzerland AG 2021 87

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_6
88 P. D’Avanzo et al.

runs with CUDA® programming language, so we can obtain numerical simulations in

few seconds, also if they are very complex and using a modest desktop PC. Another
important aspect of our model to be considered is the possibility to correlate it to the
theory of generalized continua.
We are aware of the fact that a lot of mathematical results are demanding for
proving; we believe that the presented computation scheme is somehow equivalent
to those used in continuum mechanics. This hypothesis requires an in-depth inves-
tigation in this direction. But it is our opinion that the model can be useful in a
large variety of mechanical application. In particular, we believe that the demanding
computational problems to be faced in:
1. second gradient continuum theory (Abdoul-Anziz et al. 2019; Barchiesi et al.
2019a; Scerrato et al. 2016; Sciarra et al. 2007; Turco et al. 2016a; Epstein and
Smelser 2020; Germain 2020; Yang and Misra 2012);
2. peridynamics (Dell’Isola et al. 2015a; Javili et al. 2019);
3. explicit time integration (Vaiana et al. 2019; Greco et al. 2018a);
4. plasticity (Sessa et al. 2017, 2018a; Serpieri et al. 2018);
5. concrete structures (Sessa et al. 2019b; Marmo et al. 2019);
6. masonry structures (Marmo et al. 2018a, b);
7. bone growth phenomena and biomimetics (George et al. 2018, 2020; Giorgio
et al. 2016, 2017a, 2019b, 2020b; Rapisarda et al. 2018, 2020; Perricone et al.
2020);
8. generalized shell theories (Altenbach et al. 2010; Altenbach and Eremeyev 2011,
2015; Giorgio 2020b; Steigmann 2018; Spagnuolo et al. 2020; Giorgio et al.
2021; Altenbach and Eremeyev 2008a, b; Andreaus et al. 2018; Bilotta et al.
2019; Desmorat et al. 2020; Steigmann and dell’Isola 2015; Sessa et al. 2017,
2019a; Valoroso et al. 2015) and composite laminates (Cazzani and Ruge 2016;
Cazzani et al. 2018b, a; Giorgio et al. 2021; Browning and Askes 2019; Altenbach
et al. 2015);
9. beams in large deformations (Baroudi et al. 2019; Giorgio and Del Vescovo 2018,
2019; Barchiesi et al. 2020b, a, 2021a; Spagnuolo and Andreaus 2019; Paradiso
et al. 2021, 2019);
10. seismic analysis (Vaiana et al. 2017, 2021; Sessa et al. 2018b, 2015);
11. micro-morphic (Misra and Poorsolhjouy 2016b, a; Misra et al. 2020; Giorgio
and Scerrato 2017; Giorgio et al. 2019a; Abali et al. 2012; Gagneux and Mil-
let 2016; Rickert et al. 2019; Franciosi et al. 2019; Scerrato et al. 2014) and
micro-polar continuum theories (Altenbach and Eremeyev 2015; Eremeyev and
Pietraszkiewicz 2016; Giorgio 2020a; Greco 2020; Eugster et al. 2014; Olive
2019; Turco et al. 2019; Turco 2018b);
maybe dramatically simplified if the algorithm presented here will prove to be an effi-
cient integration scheme for all mentioned continuum models (see, e.g., also Wiech
et al. 2018). It has to be remarked that generalized continua already demanded the
introduction of complex integration schemes, among which we believe that isoge-
ometric ones are of particular relevance (Cuomo et al. 2014; Greco et al. 2018b;
Niiranen et al. 2016; Cazzani et al. 2016a, b, c; Yildizdag et al. 2020b; Schulte et al.
2020; Tran and Niiranen 2020).
6 From the Swarm Robotics to Material Deformations 89

Therefore, we believe that the algorithm presented here not only can represent
a powerful predictive tool in metamaterial theory (Barchiesi et al. 2019c, 2021b;
Dell’Isola et al. 2015b, 2019; Del Vescovo and Giorgio 2014; Eremeyev and Turco
2020; Eugster et al. 2019; Scerrato and Giorgio 2019; Spagnuolo 2020; Vangelatos
et al. 2019; Yang et al. 2019; Yildizdag et al. 2019; Alibert et al. 2003; Seppecher
et al. 2011; Giorgio et al. 2017b, 2019c, 2020a; Cuomo et al. 2016; Yildizdag et al.
2020a; Spagnuolo and Scerrato 2020), and in general in the theory of deformable
bodies, but also can represent a heuristic tool in helping in search of micro-macro
identification results (De Angelo et al. 2019a; Rosi et al. 2018; Giorgio et al. 2018;
Andreaus et al. 2016; Barchiesi et al. 2019b; Boutin et al. 2017; De Angelo et al.
2020; Placidi et al. 2017; Abali et al. 2016; Nejadsadeghi et al. 2019; Yang et al.
2018; Bolzon et al. 2002).
It has to be remarked that the presented algorithm can be used for calculating
families of equilibrium shapes by using variable load parameters, in a very similar
way as done in continuum mechanics (Turco et al. 2020; Turco and Barchiesi 2019;
Turco 2018a, 2019; Boutin et al. 2017). When damage phenomena occur in standard
continuum mechanics, some Karush–Kuhn–Tucker (KKT) conditions (Placidi 2016;
Placidi et al. 2018a, b; Timofeev et al. 2020) need to be used in order to take into
account for plastic and damage phenomena or crack onset (Cuomo 2019; De Angelo
et al. 2019b; Bilotta et al. 2018). All quoted dissipation phenomena can be more
easily modeled in the present algorithm, employing the purely deformation-based
criterion already in Spagnuolo et al. (2017), Turco et al. (2016b). We believe that
by introducing a similar threshold in our algorithm, it will be possible to predict not
only crack onset and damage progression, but also plastic phenomena (Contrafatto
and Cuomo 2002; Del Piero 2018; Cuomo 2017; Vaiana et al. 2018).
The chapter is organized as follows: in the second section, we recall the principal
and most important other models used in the literature from which our model takes
origin; in the third section, we construct step by step our own model. The fourth
section is dedicated to show some numerical simulations obtained; we divide them
into two groups: the standard ones and the exotic ones. In the last section, conclusions
with advantages and disadvantages are discussed.

6.2 Other Models in Literature

In literature there are several other models and methods that can approximate the
evolution of material under deformations; many of them share the use of points linked
by an interaction or a law that manages the motion of points in space. We started
from the study of these to develop a slightly different-mixed approach, so here we
briefly recap some important methods that inspired us.
90 P. D’Avanzo et al.

6.2.1 Position-Based Dynamics (PBD)

The Position-Based Dynamics (PBD) method of Matthias Müller was described in his
article (Müller et al. 2005, 2007). In this method, the dynamic object is approximated
by a set of vertices and the internal forces by constraints (the external forces are used
in the definition of the vertices velocities in a classical way). The vertices have a
position xi , velocity vi , and a mass, whose reciprocal is wi , meanwhile there are
different constraint functions C j that are used to simulate various types of materials
and particular behaviors in PBD. The main task of the PBD method is to handle the
positions of the vertices, in fact, every time step, as shown in Table 6.1, the velocities
are evaluated and slightly dumped only to get the “virtual” new positions x̃. These are
checked and rearranged to satisfy the constraints (for example, the distance between
two vertices) typically using a Gauss-Seidel-type iteration. This kind of solving
method is fast and allows to obtain good-like simulations in real time (this is the
reason why it is used to reproduce the physics in some video games).
This method is able to simulate collisions among objects (generating specific
constraints), simulate friction (updating the velocities after their upgrade), avoid
autocom penetrations, and simulate the breaking.

6.2.2 Swarm Robotics

Unlike the PBD method, in which the external forces are used, the swarm robotics
use a kinematic approach where a forced motion to some elements is imposed. In this
method, the material is seen as an ensemble elements that compose a swarm geometric

Table 6.1 The algorithm operation scheme of the PBD method shown in Müller et al. (2007)
(1) loop
(2) for-all the vertices i do vi ← vi + Δtwi f ext (xi )
(3) DumpVelocities(v1 , . . . , v N )
(4) for-all the vertices i do x̃i ← xi + Δtvi
(5) for-all the vertices i do generate Collision Constraints(xi → x̃i )
(6) loop solver Iterations times
 
(7) project Constraints C1 ; . . . ; C M+Mcoll ; x̃1 ; . . . ; x̃ N
(8) end loop
(9) for-all the vertices i
(10) vi ← (x̃i − xi )/Δt
(11) xi ← x̃i
(12) end for
(13) velocityUpdate(v1 ; . . . ; v N )
(14) end loop
6 From the Swarm Robotics to Material Deformations 91

Fig. 6.1 An example of the three types of points for a square lattice, in red the leaders, in blue the
followers, and in orange the frame points. We underline the presence of two shells of frame points
that means the use of an interaction that includes first and second neighbors of every follower

configuration (dell’Erba 2015); the idea is that these simple-acting elements together
can describe more complex behaviors. This approach is defined in Şahin (2005),
Moriconi and Dell’Erba (2014), Dell’Erba (2018, 2019a, 2020b, 2021b, c) as
Swarm robotics is the study of how a large number of relatively simple physically embod-
ied agents can be designed such that a desired collective behavior emerges from the local
interactions among agents and between the agents and the environment.

The works, which were our start point in this research, Battista et al. (2016; 2018)
and Dell’Erba (2019b) are based on this approach: “Can we describe a material under
deformations through tiny robots that control every time the relative distances as in
Dell’Erba (2020a, 2021a)?”. So inspired by the natural swarms, in which there are
leader elements that control the motions of the followers, and so of the entire swarm,
in these models, there are three principal categories of point as shown in Fig. 6.1:
• Leaders that have a forced movement simulating external interaction of the swarm,
for example, like a tensile test machine that brings in opposite direction two sides
of a metal bar;
• Followers, that are the majority of the swarm, represent tiny pieces of the cho-
sen material and every follower has a movement induced by the positions of its
neighbors points of the swarm as a centroid rule;
• Frame points, points added with the only purpose to balance the lack of neighbors
and make sure that the interaction always takes place among the same number of
points for the followers closer the edges of the swarm. Every frame point has one
or more linked points (followers and/or leaders) and their movement is defined as
follow them.
These models divide the neighbors points of a considered point in shell categorized
by the distance1 from it and the type of lattice used. So it is possible to identify first
neighbors, second neighbors, and virtually so on. The use of more shells of neighbors

1The kind of method used to evaluate the distance influences the shape of the shells. Usually for
swarm based on centered square cells (like in Fig. 6.1) is proposed the Chebyshev distance.
92 P. D’Avanzo et al.

Table 6.2 The algorithm operation scheme of the model proposed in Dell’Erba (2019b) with the
use of the Centroid rule
(1) for-all “time” step i
(2) for-all Leader j
(3) do x(i) (i−1)
j = xj + v(i)
j (the Leaders are translated)
(4) for-all Follower k
(i)
(5) do xk = Centroid [neighbhor sk ] (the Followers are moved)
(6) for-all Frame point h
(i)
(7) do xh = Average_mov [linked_ pointsh ] (the Frame is moved)
(8) end for

allows to obtain more features of the deformations and it is linked with higher gradient
order continua as explained in Battista et al. (2016). Then the algorithm proceeds for
every iteration following the instructions shown in Table 6.2. As the algorithm steps
proceed, the points are moved every iteration: first the leaders, then one by one the
followers, and finally the frame as a whole.

6.3 The Model Here Proposed

The model proposed here starts from an approach very similar to that used in Swarm-
based models, in fact, in our model is maintained a division in the role of the points,
but it eventually becomes similar to a PBD model in some features. Before we have
developed the model here presented, we have studied a quite similar model with
spring interactions with dynamical evolution carried out with differential equations;
this approach is hardly efficient for its computational costs but it gives us some
important hints about how to develop the model here proposed.
First of all, we have to make an important specification: In our model, we have only
a set of variables, which are points of a Cartesian plane, which change coordinates
with the increase of a certain discrete parameter. Therefore, physical concepts such
as “motion”, “time” or “velocity” do not exist in our model. The update calculation
of the coordinates of our set of points is done through an algorithm (explained in
detail in the next section).
We will explore in further papers in quantitatively way the parallelism among
some parameters and characteristics of our model and the physics of deformations,
so here we will use some terms only through analogy, such as the discrete parameter
that orders the iteration of the algorithm will be called “pseudo-time”, just as a single
execution of the algorithm (one run of the iteration) will be called “single pseudo-
temporal step” or simply “p-time step”, the “updated-trajectory” of a point will
indicate the updating of the coordinates of the point (that we call “updated-position”
or simply “coordinates” of a point) with a single execution of the algorithm (therefore
6 From the Swarm Robotics to Material Deformations 93

in a p-time step) and the “updated-velocity” of a point is the difference between the
actual and the previous updated-position. But, we remark again all the variables and
parameters that we use are dimensionless and devoid of any real physical meaning.
Therefore, it must not be made the mistake of apply standard concepts of physics
to our model. The latter, in fact, is composed only of points and an algorithm that
redefines their coordinates at each iteration. We will analyze in detail in the next
papers the connection between our model and a continuous model, and therefore,
among our parameters and those of the physics of deformations.

6.3.1 A Recall About Graph Theory

Our model is based on some concepts of the graph theory (i.e., the lattice, the neigh-
borhood of a point, and the adjacency matrix). A graph is a mathematical structure
used to model pairwise relations between objects. To be more specific, let us indicate
a graph G = (P, L) as an ordered couple of a set P of “points”, a set L “links”
(a pair of points). We will consider only undirected graphs, meaning that there is
no direction associated with links. Considering n points Pi , with i ∈ {1, . . . , n}, we
define the adjacency matrix that carries all the information on the connected points,
as the matrix A of dimension n × n defined by:

Ai j = 1 if Vi is linked to P j , 0 if Vi is not linked to P j (6.1)

The one given above is the standard definition of adjacency matrix, but as we will
see, in order to fulfill some properties in our swarm, we need to modify it to define
the neighborhood of a point.
It is possible to define a path on the graph. It is a finite or infinite sequence of
links that joins the sequence of points; this will be useful to define the concept of
neighborhood order.

6.3.2 Constructing the Model

Let us consider a swarm S constituted by a finite number of elements, and let us

indicate by C0 the reference configuration of S. In C0 , the elements of S occupy
all the points of a generic lattice with step σ. We consider a set of p-time steps, i.e.,
a set of ordered discrete values for the pseudo-time variable t that we will denote
with T = {t0 , t1 , . . . , tm , . . . }. We indicate the elements of S (at a pseudo-time tm )
by means of an index i like Pi (tm ) = (ri1 , ri2 )(tm ) that are the coordinates of the i−th
point in C0 at the pseudo-time tm in a orthogonal reference system. For each element
i in C0 , we define a chosen number of sets of points representing the k−th neighbors
of i.
94 P. D’Avanzo et al.

Fig. 6.2 Here we highlight

two different couples of
neighbors (one in blue and
one in orange), in the set of
the first and second
neighbors (N2 ) of the
considered point, marked in
red, on a hexagonal lattice.
In particular, we have
indicated in blue a couple of
first neighbors and in orange
a couple of second neighbors
of the red point

The first neighbors of a point Pi are all the points of C0 adjacent to the point Pi
in the meaning of the graphs (so it is linked with the coordination number of the
lattice). The second neighbors of a point Pi are all the points of C0 whose minimum
path without repetitions is composed of only two links. Therefore, defining l(i, j) as
the minimum path without repetitions between points i and j, the k−th neighbors
of a point i is the set
Nk (i) = { j ∈ C0 : l(i, j) = k} (6.2)

that we call “shell” of order k.

We will also consider pairs of opposite neighbors compared to a considered point.
The pairs collection of opposite k-order neighbors of a point Pi is the set

Pk (i) = {(h, g) ∈ Nk (i) × Nk (i) : (Ph − Pi ) = −(Pg − Pi )}, (6.3)

so the j-th pair of the point Pi will be indicated with ρi j , see Fig. 6.2.
Note that Nk (i) is the i-th row of the adjacency matrix: this means that our
adjacency matrix A is not n × n but n × m where m is the number of the total
neighbors considered until the k-th order. Let us select a set of Leaders L, they are a
subset of C0 to which is imposed an updated-trajectory, i.e., a collection of positions
in our reference system for each p-time step. Instead, all the remaining elements of C0
form the set of the followers F, so C0 = L ∪ F; the updated-trajectory of each point
Pi ∈ F is defined by the coordinate of its neighbors until the k−th order considered.
For constructing the model, we are interested to have an elastic interaction between
the points in order to obtain a swarm that exhibits a behavior qualitatively similar to
an elastic continuum. For this reason, we have started from a system of equations that
rules the kinematic a point Pi that belongs to a system of points with elastic interaction
among them; then, for not burden the model with the researching of ODE’s solutions,
we simplify the classic equations in order to obtain an approximation that leads to a
faster algorithm for our model. Moreover, we have enriched the model by introducing
another interaction aimed to emphasize the response of the swarm to the shear and
6 From the Swarm Robotics to Material Deformations 95

flexural simulations. There follow the equations that define the interaction law for
our model:
ri (tm+1 ) = ri (tm ) + Inti (tm ) (6.4)
⎛⎡ ⎤ ⎡ ⎤⎞

k    
Inti (tm+1 ) = ⎝⎣ Ai j (tm ) ⎦ + ⎣ (Bil (tm ))⎦⎠ (6.5)
h=1 j∈Nh (i) l∈Ph (i)

αj   ri (tm ) − ri j (tm )
Ai j (tm ) = − Ri j (tm ) − Ri j (t0 ) (6.6)
Ni |ri (tm ) − ri j (tm )|

βl
Bil (tm ) = − (ri (tm ) − ρil (tm )) (6.7)
Ni

where all the quantities in the equations are in Table 6.3.

Let a point Pi and its j-th neighbor, the term Ai j is proportional to the difference
between the actual distance ri (tm ) − ri j (tm ) and the same distance in the refer-
ence configuration C0 , and directed as (ri (tm ) − ri j (tm )), so its contribute to the
total displacement simulates an elastic interaction: repulsive if (Ri j (tm ) − Ri0j < 0),
attractive if (Ri j (tm ) − Ri0j > 0). The subscript j in the term α j is the label of the
considered neighbor in the set Nh . This means that, in general, we are assuming that
a point Pi has a different coupling constant for the interaction. In this chapter, we
will use always the same α j for every point of the h-th shell. So we will use a simpler
notation: α1 for the interaction with the first neighbor (i.e., 1-th order shell), α2 for

Table 6.3 List of quantities in the equations of the model

ri (tm+1 ) Is the updated-position of the point Pi at p-time step tm+1
Inti (tm ) Is the “interaction” among the point Pi and all its neighbors in the h-th
order shell until the k-th order
Ni Is the number of all the considered neighbors of the point Pi
ri j (tm ) Is the updated-position of the j-th neighbor of the point Pi at the p-time
step tm
Ri j (tm ) Is the Euclidean distance between point Pi and its j-th neighbor at the
p-time step tm
Ri j (t0 ) Is the Euclidean distance between point Pi and its j-th neighbor in C0
ρil (tm ) Is the average point of the pair ρil considering the position of the points at
time tm
αj Is a coefficient that modulates the intensity of the “elastic” interaction
(Ai j ) and the convergence velocity. It depends by the j-th neighbor’s shell
order
βl Is a coefficient that modulates the intensity of the“alignment” interaction
(Bil ). It depends by the j-th neighbor’s shell order; the l is the label of the
couple ρil
σ Is the step of the lattice
96 P. D’Avanzo et al.

Fig. 6.3 Here is an example of how the interaction Bil works on a hexagonal lattice for the first
neighbors. With respect to the red point, the blue and green points are its first neighbors. In green, we
have highlighted one of the couples of the set P1 (i) (i.e., the set of the first neighbors couples of the
considered red point) and marked in orange the centroid between the two points of the considered
couple. The interaction Bil pushes the considered red point, for the green couple, to be in the orange
spot. So this interaction will be proportional to the distance between the current position of the
considered point (ri (tm ) in red) and the aimed position (ρil (tm ) in orange) and the constant β1 .
This interaction is repeated on all the couples in the set P1 (i)

the interaction with the second neighbor (i.e., 2-th order shell), and so on. The same
considerations can be done for βl , but because, in this chapter, we will use the same
value for every pair and only for the first neighbors, we will omit the subscript.
The term Ai j and Bil contains also the factor Ni , it dampens the interaction and
may be interpreted as if there was a viscous friction. This is because, it reduces the
interaction in a way proportional to the number of the neighbors points, therefore,
innermost points will see their interaction reduced compared to outermost points.
Furthermore, the factor Ni makes the summation on the neighbors an average, so
untying it from the number of neighbors. If the point Pi does not have neighbors
(Ni = 0), the term Inti is null because the interaction can’t be defined. The term Bil
pushes the point i to place itself into centroid of the couple of neighbors considered
ρil as shown in Fig. 6.3. We have introduced this kind of interaction to have a way
to modulate the flexion response of the swarm and help to avoid the overlap of the
points as better explained in Sect. 6.4.
The algorithm operation scheme is shown in Table 6.4.
As we can see, for each point i of the lattice, the interaction is evaluated considering
only the neighbors up to the k-th order, this allows us to reduce the computational
cost, so the complexity of our algorithm is linear.

6.3.3 Relationship with Other Models

It is important to note that in the model, we propose there are only two main roles for
the points, unlike the (usually) three used in the swarm models, for example (Battista
et al. 2016, 2018; Dell’Erba 2019b). The points are divided into two roles:
6 From the Swarm Robotics to Material Deformations 97

Table 6.4 The algorithm operation scheme of the model here proposed (v j is the imposed “velocity”
of the j-th leader)
(1) for-all p-time step i
(2) for-all Leader j
(3) do x(i) (i−1)
j = xj + v(i)
j (the Leaders are translated)
(4) for-all Follower k Interaction phase
 
(i) (i) (i−1)
(5) do Intk = Functionk neighbor sk (The interactions are evaluated)
(6) for-all Follower k Spread phase
(i) (i−1) (i)
(7) do xk = xk + Intk (the Followers are translated)
(8) end

• Leader: points that have an updated motion assigned a priori for every time step.
In a physical metaphor, therefore, they represent the interaction between the object
modeled by our set of points and the cause of the deformation.
• Followers: points that through an algorithmic interaction “follow” the step-
movement due to the updated motion of the leaders.
In this model, we do not need the frame points. This is because, unlike (Dell’Erba
2019a, 2020b, 2021b, c), for the interaction used, we do not need that all the shells
are complete (i.e., every shell has the maximum number of neighbors expected for
the order considered) to balance the interaction on the edges points of the swarm.
Without the frame points, that usually had a proper rules of motion and were
moved in a different time respect to the other swarm points, all the non-Leader
points are moved simultaneously and they are treated in the same way. This makes
the algorithm leaner, more homogeneous, and faster.
In our model, the upgrade of the points coordinates is performed simultaneously
(the update takes place based on the coordinates of the previous p-time step without
using the updating positions of the current p-time step) this allows us to unbind
the result of the simulations with the update order of the points. So we divided the
positions upgrade in two steps:
• Interaction: in which, based on updated-positions and updated-velocities at the
previous p-time step, we calculate the interaction between the points using the
algorithm shown below;
• Spread: in which the updated-position (and updated-velocity) of the points is
updated in proportion to the calculated interaction.
A similar approach is already used in many numerical models such as the PBD.
It is clear that the computational cost of the evolution of the system, in our case,
increases approximately linearly with the number of elements of the system (as the
number of interactions considered for every element is fixed and the computation is
based upon linear operations) while employing a FEM, and the cost is nonlinear in
the number of mesh elements.
98 P. D’Avanzo et al.

6.3.4 Meaning of Neighbors

As we can see in the literature, the energy deformation E of a generalized continuum

depends on the high order gradient of the displacement field χ

E = E(χ, ∇χ, ∇ 2 χ, . . . ), (6.8)

in particular, in the papers (Della Corte et al. 2015), we can see that there is a link
between the energy of first gradient continua and the swarm with first neighbors
interaction; however, considering a swarm with second neighbors interaction, the
energy deformation is linkable to second gradient continua energy.
As we will see in the next section, our model is able to give results for standard
deformation (i.e., deformation of Cauchy continua) that are qualitatively similar to
those obtained with FEM, and was also able to simulate typical phenomena of the
second gradient continua. The breaking phenomena are not described in this chapter
to have a complete description in a future paper, but it can simply be added by
introducing a breaking threshold to the distance between the two points.
In the following, we will suppose:
• Ai j evaluated on the first neighbors, is linked to first gradient continua;
• Ai j evaluated on the second neighbors, is linked to second gradient continua;
• Bil evaluated on the first neighbors, is linked to second gradient continua.
In future works, we will try to support such hypothesis.

6.4 Numerical Simulations

We have divided the set of simulations into two sub-sets, as well as this chapter’s
section: the first set contains the whole standard simulations where they are called
standard because they contain only the interaction that can be related to the first
gradient continua, for this kind of simulations, we have chosen the traction, com-
pression, shear, and flexion deformations. The second set contains the simulation
that we will call exotic ones, this contains the interactions that can be related to the
second gradient continua; in this case, we perform only traction and compression
simulations.
In Table 6.5, the dimensions of our swarm and the kind of simulation performed
is reported. In Fig. 6.4, it is possible to see the initial configuration for traction and
compression, and in Fig. 6.5, for flexion and shear.
Since it is observed that, within some thresholds of the parameters (see Sect. 4.2),
stopping the leaders at a chosen pseudo time step and letting the system continue
to evolve, this converges to what seems an “equilibrium” in which all the points are
practically stationary. So all simulations were performed with the following scheme:
6 From the Swarm Robotics to Material Deformations 99

Table 6.5 Table of swarms’ dimensions used in our simulations. N E and N EH are clearly the number
of elements of the swarm, respectively, without and with hole
Dimension Deformation Lattice NE N EH
0.7 × 1.4 Traction and Hexagonal 2891 2768
compression
0.7 × 1.4 Traction and Squared 2556 2452
compression
0.56 × 2.8 Flexion and shear Hexagonal 4637 4554
0.56 × 2.8 Flexion and shear Squared 4089 4020

Fig. 6.4 The initial configuration swarm used for traction and compression simulations. The leaders
are in red, the followers are in blue. (Left: hexagonal lattice. Right: squared lattice)

Fig. 6.5 The initial configuration swarm used for shear and flexion simulations. The leaders are in
red, the followers are in blue. (Left: hexagonal lattice. Right: squared lattice)
100 P. D’Avanzo et al.

• Active deformation phase, in which the leaders are moved up to a certain pseudo
time step, at the same time, the followers adapt to this shift.
• Settling phase, in which the leaders are kept still and the followers continue to
move until they reach an “equilibrium configuration”, i.e., when the difference
between the pseudo time step position and the next pseudo time step is less than
a chosen tolerance ε value.
Considering only first neighbors interaction, the convergence velocity depends
only on α1 coefficient defined in the previous formula, so the latter increases with
the growth of α1 ; on the other side, the value of α1 does not influence the equilibrium
configuration, but only the dynamics, i.e., the way the system reaches the equilibrium
configuration. This happens unless α1 does not reach a particular value after that
exotic phenomena begin and so, growing further α1 reaches a critical value such that
the interaction term Ai j → ∞ and the swarm “explodes”, i.e., the swarm elements
leave away to the infinity or move themselves in a chaotic way in a limited domain
of space (for more information about this, see the Sect. 6.4, Fig. 6.21). Considering
also the second neighbors interactions, the convergence velocity still depends on
α1 as in the latter case, but in this case, we need to distinguish the α1 related to
the first neighbors, and the α2 related to the second neighbors, moreover, here the
equilibrium configuration depends from both α1 and α2 values, despite the first case
where the final configuration does not changes for α in the normal range, while now
there is a different final configuration for every couple (α1 , α2 ), as it is better argued
in Sect. 6.4.2.
To show how the swarm deforms itself under the leader’s motion, in addition to
its points positions, we have added the elementary area units that will be colored by
their deformation state.
Here we will define the elementary area unit of both hexagonal and squared
lattice: let a point i of the hexagonal lattice and its first neighbors, the elementary
area is any triangle which vertices are the point i itself and two of its first neighbors
that are adjacent to each other; similarly for the squared lattice, let a point i of the
squared lattice and its first neighbors, the elementary area is any quadrangle which
vertices are the point i itself and three of its first neighbors that are adjacent to each
other (see Fig. 6.6).
This process is done on the whole elements of lattice, so it is possible to cover it
with not superimposed elementary areas.
In the simulations that we will show, the elementary area’s colors represent the
absolute value of the percentage variation of the area A, with respect to the initial
area A0
|A0 − A|
Elementar y Ar ea Color = ; (6.9)
A0

the choice to highlight the elementary areas colors’ spread, in analogy with the
continuum mechanics physics, is due to its strong relation with deformation energy
distribution.
6 From the Swarm Robotics to Material Deformations 101

Fig. 6.6 Example of how

elementary area units are
considered for an elementary
cell of hexagonal and
squared lattices

The introduction of the elementary area units has the aim to show where high
deformations region occur, where a break is more likely. We will introduce this
feature in the next work.
In Fig. 6.7, we have show an example of the two principal phases of the algorithm,
the initial configuration and the equilibrium configuration, followed by the colored
one. In the rest of the chapter, we will show only the swarms at this latter phase for
every simulation because we are interested only in the shape that the system of points
assumes at this stage.
As we will see that some simulations will have more pseudo time steps than others;
this choice was been done to avoid superpositions of swarm points that would cause
undesirable effects (i.e., overlaps and compenetration of the points). These problems
can be avoided by lowering the leaders’ updated velocity and, in the same time,
growing the number of time steps to obtain the same deformation; this allow the
followers to better fit to the changing shape of the system. In future, we will modify
the interaction law in order to prevent this undesirable phenomenon.
Because we are using a model based on swarm, which does not have a conscience
of its own configuration and each of its element does not know as to a priori get the
equilibrium configuration, problems can arise due to the overlapping of the points.
An example of these problems is shown for a simple one-dimensional lattice of
three points without the implementation of the term Bil in Fig. 6.8. Indeed, the algo-
rithm considers the three configurations as equivalent equilibrium configurations.
Instead with a two-dimensional lattice, the situation is more complicated and,
without considering the interaction Bil , we provide an example in Figs. 6.9 and 6.10,
in which, we show two compression tests with the same initial configurations and
all the same parameters except leaders’ velocities.
In Fig. 6.10 is shown the response of the swarm to a low leaders velocities and
there is not the overlapping problem, instead in Fig. 6.9 the higher leaders velocities
cause an overlapping problem that leads to a new final configuration where the swarm
interpenetrates itself.
So the interaction Bil can help to avoid such kinds of situations by giving impor-
tance also to the directions and alignments of the link among the points (and not only
102 P. D’Avanzo et al.

Fig. 6.7 Here we show in (a) and (b) a sample of the true aspect of our model output. In particular,
in (a) is shown the initial configuration of a little squared-lattice-swarm subject to a traction test
which final equilibrium configuration is shown in (b). But in the follow of the chapter we do not
provide this kind of pictures but we provide (c)-like picture. In (c) we shown the same equilibrium
configuration of (b) but displaying also the elementary area variations and the associate colorbar.
This latter kind of picture is possible to get more information about the swarm deformation

Fig. 6.8 Example of a

superposition problem (I)

on the distances between them). It is important to note that the points on the edges of
the swarm perceive fewer interactions (because their shells are not complete, all the
possible couple ρil are not available), so we are working also on other interactions
of this type to show in future researches, an example is provided in Fig. 6.11 for a
one-dimensional lattice.
6 From the Swarm Robotics to Material Deformations 103

Fig. 6.9 Example of a superposition problem (II). The leaders’ high velocities cause an overlapping
within the swarm (order of images: top-left, top-right, bottom-left, and bottom-right)

Fig. 6.10 Example of an avoided superposition problem (order of images: top-left, top-right,
bottom-left, and bottom-right)

Fig. 6.11 On the top, we show an initial configuration considering the interaction acting on the
point P in red. With respect to P, the blue point is its first neighbor (1N ) and the green one is its
second neighbor (2N ). So an alignment interaction can be introduced by pushing the point P to
keep itself in the aimed position (A P), marked in orange, that is on the prolongation of the segment
2N − 1N as shown in the bottom. The advantage of this kind of interaction can be its capability to
work also with the points on the edges of the swarm
104 P. D’Avanzo et al.

6.4.1 Standard Simulations

As explained, we have used two different systems for the numerical simulations
shown in Fig. 6.4, for traction and compression, and in Fig. 6.5, for shear and flexion,
both with leaders in red and followers in blue. All these standard simulations have
been performed with only the interaction limited to the first neighbors and without
considering the alignment term Bil ; in other words, in this subsection, our purpose
is to show only the deformations attributable to first gradient continua.
In this chapter, we focused on traction and compression tests which are the simplest
simulations we can achieve. But in order to show that the model is qualitatively able
to simulate also flexion and shear tests, we insert some simulations of this kind only
for demonstration purposes.

6.4.1.1 Standard Traction

In these simulations, we hold the leaders of the right side, while pulled those of the
left one; we settled this model to obtain an elongation of 50%. It is possible to see the
final configuration in Fig. 6.12.2 The same simulation with a swarm with a centering
hole follows. This kind of simulations have been performed using 30,000 pseudo-
time steps, divided into the first 10,000 for active deformation phase, the second
20,000s for the settling phase. From these simulations, we can get the following
observations:
• The model is able to reproduce the Poisson effect. The two lattices used seems to
show slightly different Poisson effect.
• It is important to note the different behaviors of the lattice around the hole, in partic-
ular, on its upper and lower sides. In the first impression, we retained possible that
these differences were caused by the discretization around the hole (e.g., observe
the different shape of the hole in Fig. 6.4); so we performed the same simulations
with a finer discretizations obtaining the two same different behavior around the
holes; see Fig. 6.14. Furthermore, the coloration pattern (i.e., lower elementary
areas variations) of the hexagonal lattice on the top and bottom of the hole seems
compatible with the analog behavior of a isotropic material, meanwhile the same
coloration pattern of the squared one seems compatible with the behavior of an
orthotropic material (i.e., the stretching). These behaviors seem compatible with
our initial conjecture on the kinds of materials that the two lattice should simulate.
However, it is premature to do such hypothesis, so we will better argue it in the
future paper. Far from the hole, the variations of elementary areas are different, as
it may be seen from the differences between the two-color distributions.
• For the traction tests without hole and samples used in Fig. 6.12, the two kind of
lattice have approximately a similar response, as it could be seen in the figures

2We want to underline that, in this way, we are out of the region of elasticity because it is not a
small deformation.
6 From the Swarm Robotics to Material Deformations 105

Fig. 6.12 Standard traction simulations, respectively, for the hexagonal (images at the top) and
squared (bottom) lattices with α1 = 1, with and without hole

comparing two-color distributions (the areas of greater elementary areas variations

are the same in both lattices). It is important to note that, in general, two different
lattices can exhibit different behavior. An example is given in Fig. 6.15, where the
hexagonal lattice shows a different response if the orientation of discretization is
rotated by 90°(an example of such rotation is provided in Fig. 6.16). We want to
underline that to get the simulation of Fig. 6.15, we used cells of finite dimensions
and that all αs in the cells are the same. We also show how the two lattices can
give different results considering finer discretization in Fig. 6.14.
In order to understand if some of the features obtained in the previous simula-
tion are only an effect due to the discretization, we performed another simulation
for traction test but by using samples with a finer discretizations. The two initial
configurations are shown in Fig. 6.13.
• To get the simulations of the standard traction with a finer discretizations, there
were needed 2 million time steps in order to avoid superpositions.
• In Fig. 6.14, a situation similar to that shown in Fig. 6.12 occurs but with a more
homogeneous colors distribution in squared lattice.
• As the behaviors of the lattices with the hole in Fig. 6.12, also in Fig. 6.14 the
two lattices show a similar patterns around the holes but more homogeneous; this
contradicts the previous hypothesis of a mere mesh effect: in fact in this case, as
shown in Fig. 6.13, the holes are more round.
As mentioned before, changing the parameter α1 does not alter the equilibrium
configuration, so the magnitude of Poisson effect does not directly depend on α1 .
In order to modulate the Poisson effect, we have to define different values of α1 for
every neighbor in all the shells of the swarm. In particular, by considering a square
lattice, we assign a value α1vh for neighbors positioned in vertical and horizontal
direction with respect to the central point of the shell, and assign a different value
α1d for neighbors positioned in diagonal direction with respect to the central point of
the shell. The ratio (α1d /α1vh ) is linked to the intensity of the Poisson effect. We can
also obtain qualitative similarity with auxetic materials setting the two parameters
to obtain a negative ratio. We can proceed as the same way for the hexagonal lattice,
106 P. D’Avanzo et al.

Fig. 6.13 Here we show new initial configurations used to obtain new simulations of traction and
compression tests with a finer discretization. We want to underline that with a greater number of
elements, the holes are more round

Fig. 6.14 Standard traction simulations with a finer discretizations, respectively, for the hexagonal
(top) and squared lattices (bottom) both with α = 0.5

but it is more complicated owing to the different cell geometry (Figs. 6.14, 6.15 and
6.16).

6.4.1.2 Standard Compression

In these simulations, we hold the leaders of the right side, while pushing the leaders
of the left one towards the right side; we set this model to obtain a compression of
about the 18%.3 We performed the same simulations with a swarm with a centering
hole. It is possible to see the simulations in Fig. 6.17. The simulations of these kind

3 Also in this simulations we do not have small deformations.

6 From the Swarm Robotics to Material Deformations 107

Fig. 6.15 On the left, the standard traction simulations, respectively, for the hexagonal and squared
lattices. On the right, we rotated the system keeping the same orientation of discretization of the
horizontal samples and all the other parameters of the model

Fig. 6.16 Here is shown what we mean with the rotation of 90°of the orientation of discretization.
As it is possible to note in the hexagonal cell such rotation causes an huge change in the structure
and in the directions of the interaction between the black and red points. Meanwhile in the square
cell, there is no difference. But it is possible to guess that for a rotation of 60°, the situation is
reversed

have been performed using 120000 pseudo-time steps, divided into the first 100000
for active deformation phase, the second 20000s for the settling phase. From these
simulations, we can get the following observations:
• also in this case there is the Poisson effect, more accentuated in the squared lattice;
• Unlike the traction simulations, here is possible to observe a different color dis-
tributions between the two kind of lattices; in particular, the hexagonal one has a
more homogeneous color distribution;
• in the simulations with the holes, slight convexity on the top and bottom sides of
the swarm appear in both lattices;
108 P. D’Avanzo et al.

Fig. 6.17 Standard compression tests without hole (left) and with it (right), for the hexagonal lattice
(up) and squared one (down). We set α1 = 1 for tests without hole and α1 = 0.2 for tests with hole

• as in the traction test with hole, also Fig. 6.17, in the regions around the holes, two
different behaviors can be seen: in the squared lattice the regions of maximum
elementary area variation are concentrated in two spots, one on the top and one
on the bottom of the hole (orthotropic like materials), meanwhile in hexagonal
lattice, the same areas are distributed in four spots, two on the top and two on
the bottom of the hole (isotropic like materials). Also, in simulations with a finer
discretization, shown in Fig. 6.18, these responses occur.

In order to understand if some of the features obtained in the previous simulation

are only an effect due to the discretization, we performed other simulations for
compression test but by using samples with a finer discretizations. The two initial
configurations are shown in Fig. 6.13.
• in Fig. 6.18 a situation similar to that shown in Fig. 6.17 occurs but with a more
homogeneous color distribution in squared lattice;
• as the behaviors of the lattices with the hole in Fig. 6.17, also in Fig. 6.18, the two
lattices show a similar pattern around the holes but more homogeneous: also in
this case, we can exclude the previous hypothesis of a mere mesh effect.

6.4.1.3 Standard Flexion and Shear

In flexion simulations, shown in Fig. 6.19, we clamped the leaders of the right side,
meanwhile the leaders of the left side have been rotated around the lower one on
6 From the Swarm Robotics to Material Deformations 109

Fig. 6.18 Standard compression simulations with a finer discretizations, respectively, for the hexag-
onal (top) and squared lattices (bottom) both with α = 0.05

Fig. 6.19 Standard flexion tests without hole (left) and with it (right), for the hexagonal lattice (up)
and squared lattice (down). We set α1 = 0.2 for all the tests showed

the side by an angle of about 25°. The simulations of this kind have been performed
using more pseudo-time steps (i.e., 2 million from which the firsts 6,00,000 for
active deformation phase) in order to avoid the overlap problem as mentioned before.
From the color distributions, it seems that the two lattices give qualitatively the
same response and shape. In shear simulations, shown in Fig. 6.20, we clamped the
leaders of the right side, meanwhile the leaders of the left side have shifted down by
a distance equal to the length of the short side. Also, these simulations have been
performed using more pseudo-time steps (i.e., 2, 5 million from which the firsts 1
million for active deformation phase) in order to avoid the overlap problem. Also, in
this case, from the color distributions, it seems that the two lattices give qualitatively
the same response and shape. Note that in both the cases, we performed a huge total
deformation.
110 P. D’Avanzo et al.

Fig. 6.20 Standard shear tests without hole (left) and with it (right), for the hexagonal lattice (up)
and squared lattice (down). We set α1 = 0.2 for all the tests showed

6.4.2 Second Neighborhoods and Exotic Simulations

The simulations shown in the last subsection are made considering only the first
neighborhood interaction and a fixed a priori α1 .
In these type of simulations, changing the value of α1 does not alter the final
“equilibrium configuration” of the swarm if its value is kept between some threshold
values, so it is very important to get in which range it can be chosen without compro-
mising the algorithm’s work. It must be investigated in a schematic way, but we do
not want to burden this chapter with this kind of complicate and delicate study, and
so we reserve it for future investigations. Therefore, here we will show only some
heuristic considerations about the working region of the model.
When the parameter α1 is chosen out of this range, what happens is that, the swarm
“explodes” (i.e., the swarm becomes a chaotic swarm), so the interaction among the
points becomes huge and consequently they scatter around, far from the leaders,
oscillating. Instead, when α1 is in the range but near to the edges, the swarm assumes
exotic behavior, showing patterns and new kinds of “equilibrium configurations”.
We observed that, for traction and compression test, this range seems to be influ-
enced by the leaders motion, the number of points in the swarm, shape, and total
deformation of the swarm. But we want to underline that also negative values, very
little in module, usually seem to be allowed. This can be important because, espe-
cially in combination with the other parameters α2 and β, it seems promising in
order to simulate auxetic material. When we add the interaction with the second
neighborhood, the situation becomes even more complicated because the parameters
seem to be strictly correlated. In fact, the presence of α2 , β or both change the range
of non-explosion of α1 and, furthermore, α2 and β have their own working regions;
This overlap of regions makes the swarm gain new exotic behaviors. Furthermore,
unlike the case of only first neighbors interaction, varying the combination of the
parameters in their working regions causes a change of the equilibrium configuration.
6 From the Swarm Robotics to Material Deformations 111

Fig. 6.21 An example of various ranges of behavior: a standard range with α1 = 1.4; b exotic
range with α1 = 1.7; c explosion with α1 = 2.2

Our interest in exotic behavior is due to their qualitative similarities with particular
kinds of phenomena such as the formation of density patterns like stripes or holes
(Fig. 6.21).
In the following simulations, we have implemented the second neighbors and the
alignment interaction.4 The action of these interactions enriches the phenomenology
of the model. The principal observations that we get from these kind of simulations
are:
• the swarm is more reactive; the deformation propagates quickly because it is carried
through the swarm at the speed of two shells at algorithm iteration, instead when
the interaction is limited at the first neighbors, in which the speed is one shell per
iteration;
• now the Poisson effect depends from the combination of all the parameters α1 , α2
and β.
As said above for the following numerical simulations, we have used the initial
configurations shown in Fig. 6.4 and performed only traction and compression tests
to show the qualitative response of the model without going into too much detail
in commenting about them. We want to underline that changing the kind of lattice
changes also the ranges of the parameters α1 , α2 , and β, so setting the same parame-
ters doesn’t lead to the same result on different lattices (this also happen in standard
simulations). Furthermore, we want to remind that in case of only the first neighbors
interaction, the equilibrium configuration doesn’t depend on the value of α1 as long
as it is maintained in the normal range; instead, this invariance is not maintained
with the second neighbors interaction, so any set of α2 and β value makes changes.
For these reasons, we have performed many simulations by varying the parameters

4 This because as explained in Sect. 6.3.4 we are interested to get behavior qualitatively related with
the second gradient continua, and so we use both α2 and β.
112 P. D’Avanzo et al.

α1 , α2 , and β; then among these, we have chosen and reported only those which
seems more significant for us, but there are still a lot of parameters combinations to
be explored.
We have divided the simulations in “normal”, “exotic”, and “pre-explosion”. In
the first, we show the reaction of the swarm when all the parameters are in the working
regions; in the second, we show some exotic behavior when some parameters are
near the upper edge of their working regions; in the third, we show extreme behavior
when the parameters are at the upper limit of the working region.
We are interested in exotics and pre-explosions behavior because they are quali-
tatively similar to particular phenomena (e.g., deformation bands).
For the following simulations, we have ever used only traction and compression
tests with the same initial configurations shown in Fig. 6.4.

• Let start talking about Fig. 6.22 where we have done a simulation with a traction
tests without hole. Here we have used a low α1 and an high α2 to underline the effect
of second neighbors. In both the lattices, it’s possible to see an hardly flattering
of upper and lower edges carrying on to the center comparing to standard traction
Fig. 6.12. Furthermore, we have observed the arise of a concentric framing effect:
in the figure, it can be noted on the upper and lower sides of the hexagonal lattice
(a), where there is a succession of rows with a low color value (dark red) and high
color value (light orange). While in the hexagonal lattice, this effect is clearer, in
the squared one it is fewer. This effect will be more evident in the compression
tests. Comparing the final configuration of the hexagonal lattice (a) in Fig. 6.22
with the one in Fig. 6.12, it is possible to note how the second neighbors interaction
homogenize the elementary areas color, making a softer shade. Furthermore, the
elementary areas variation on the left and right edges (in yellow in (a) in Fig. 6.22
and green in Fig. 6.12) are slightly the same instead in the central part (dark orange
in both the figures); in the second neighbors case, it seems to have an higher value
(i.e., the inner cells undergo to a greater deformation). By comparing the squared
lattice (b) in Fig. 6.22 with the one in Fig. 6.12, it seems that with the second
neighbors interactions, the Poisson effect becomes slightly more intense and the
elementary areas variation in the middle region of the swarm are lower.
• Even in Fig. 6.23, we performed a simulation with a traction tests without hole,
but this time with such values of α2 and β to cause the arise of exotic behaviors.
In (a) there is a slight exoticism, in the other this is clearer. We observed that by
adding the α2 and β, it is more likely to obtain non-local structures as bands and
deformation localization.
• In Fig. 6.24, we used the swarm with an hole for the traction test. We highlight
the presence of a depression on the upper and lower edges by carrying on to
the center and comparing the standard traction with an hole; Fig. 6.12. The color
distributions are different but actually their numerical values are quite similar. The
main difference between them is that in the hexagonal lattice, the color distribution
is more homogeneous. Even here, there is an arise of concentric framing effect
as in the previous traction test. Furthermore, it seems that this new interactions
regularize the deformations around the holes.
6 From the Swarm Robotics to Material Deformations 113

• Even in Fig. 6.25, we have used the swarm with an hole for the traction test, but
this time, with such values of α2 and β to cause the arise of exotic behaviors with
vertical stripes structure. The presence of the hole here, also, causes depressions
on the upper and lower edges, but now there is the appearance of color stripes in
different ways in the two different lattices, how can be seen comparing (a) and
(b) with (c) and (d). Even more, the deformations around the hole are greater than
the same in Fig. 6.24. It’s important to highlight that these exotic structures (i.e.,
deformation bands) emerge spontaneously by only setting the right parameters.
• In Fig. 6.26, we performed a simulation with a compression tests without hole.
There are no such differences found between the two lattices. Even here, the colors
are different but the numerical value is quite the same, so the main difference is
the arise of concentric framing effect, which is more evident in the squared one
(and it explains why it is called “concentric”) instead of the hexagonal one, where
it is low or totally null.
• in Fig. 6.27, we performed a compression test simulation in exotic range. The
swarms exhibit an horizontal layering effect phenomena with different schemes;
• Figure 6.28 shows a compression test simulation with the hole. On comparing the
color bars, it is found that the two lattices exhibit the quite same elementary areas
variation, but even, in this case, the hexagonal one is more homogeneous (e.g., the
low variation areas, in red, at the sides of the hole have a softer shade). The square
lattice exhibits a stronger concentric framing effect with respect to the hexagonal
one. As in Fig. 6.17, also with the second neighbors interactions, we can observe
the arise of two convexities, but in this case, it is more pronounced. Also in this
case, it seems that this new interactions regularize the deformations around the
holes.
• In Fig. 6.29, we used the swarm with the hole for the compression test, but in
this case with such values of α2 and β to cause the arise of exotic behaviors with
horizontal layering effect. We want to highlight that choose of the kind lattice can
lead to different exotic behaviors, in particular the (b) exhibits the formation of
hexagonal regions of high elementary areas variations near the hole, effect that
the squared one does not show. This effect is accentuated in the pre-explosion
situation shown in (f) of Fig. 6.30.
• As we written before now we provide in Fig. 6.30 a mix of pre-explosion simula-
tions without explanations. In these, it is possible to observe some kind of effects
that there were in the previous simulations at all. In particular, in the (a), (b), and (c)
there are quickly colors variations; (b), (c), and (f) are not symmetrical, the (f) one
exhibits the formation of hexagonal regions of high elementary areas variations,
like the (b) of Fig. 6.29, but extended all over the lattice.
114 P. D’Avanzo et al.

Fig. 6.22 Example of a simulation obtained adding the second neighbors interaction. a Hexagonal
lattice with α1 = 0.2, α2 = 2.2, and β = 0.04; b Squared lattice with α1 = 0.2, α2 = 1.2, and
β = 0.12

Fig. 6.23 Exotic traction tests. a Hexagonal lattice with α1 = 0.4, α2 = 2.2, and β = 0.04; b
Hexagonal lattice with α1 = 1, α2 = 0, and β = 0.1; c Squared lattice with α1 = 2.2, α2 = 0.4,
and β = 0.08; d Squared lattice with α1 = 2.0, α2 = 0.8, and β = 0.08

Fig. 6.24 Standard traction tests with hole. a Hexagonal lattice with α1 = 0.2, α2 = 2.0, and
β = 0; b Squared lattice with α1 = 2.0, α2 = 1.4, and β = 0.08

Fig. 6.25 Exotic simulation with holes. a Hexagonal lattice with α1 = 0.2, α2 = 2.0, and β = 0.08;
b Hexagonal lattice with α1 = 0.2, α2 = 2.6, and β = 0; c Squared lattice with α1 = 0.2, α2 = 2.4,
and β = 0; d Squared lattice with α1 = 2.2, α2 = 0.4, and β = 0.08
6 From the Swarm Robotics to Material Deformations 115

Fig. 6.26 Standard compression tests. a Hexagonal lattice with α1 = 0.6, α2 = 1.2, and β = 0.15;
b Squared lattice with α1 = 0.6, α2 = 2.4, and β = 0

Fig. 6.27 Exotic simulations. a Hexagonal lattice with α1 = 0.6, α2 = 3.0, and β = 0; b Hexago-
nal lattice with α1 = 1.2, α2 = 2.4, and β = 0.05; c Squared lattice with α1 = 3.6, α2 = 0.6, and
β = 0.05; d Squared lattice with α1 = 3.6, α2 = 1.8, and β = 0

6.5 Conclusion

Our model seems to be able to qualitatively simulate the behavior of an elastic

continuum, it provide critical pre-failure configurations and work in case of large
deformations. It is also possible to implement failure phenomena, new kinds of
interaction among the points, and make its computations faster. We can sum up the
advantage of our model in:
116 P. D’Avanzo et al.

Fig. 6.28 Standard compression tests with hole. a Hexagonal lattice with α1 = 2.4, α2 = 1.8, and
β = 0; b Squared lattice with α1 = 0.6, α2 = 1.8, and β = 0

Fig. 6.29 Exotic simulations with holes. a Hexagonal lattice with α1 = 1.2, α2 = 2.4, and β =
0.05; b Hexagonal lattice with α1 = 0.6, α2 = 2.4, and β = 0.05; c Squared lattice with α1 = 1.8,
α2 = 2.4, and β = 0; d Squared lattice with α1 = 3.6, α2 = 1.2, and β = 0.05

• it is pretty fast; defining one frame as a pseudo-time step, so the computation of

the new position of all the points of the swarm, on our modest computer desktop,
it can provide about 14000 frame per seconds with swarm within 10000 elements
(e.g., less than 3 s for the simulations of standard traction provided), instead the
frame rate decreases to 2000 for swarm within 40000 elements (e.g., about 16
minutes for each simulation shown in Fig. 6.14).
• the algorithm complexity is linear, indeed for every point (for a swarm that contains
N points), the interaction in evaluated only with the first (and second in case)
neighbors, who’s number is fixed once the lattice is chosen. So if the number of
6 From the Swarm Robotics to Material Deformations 117

Fig. 6.30 Pre-explosion simulations. a Traction of a hexagonal lattice with α1 = 1.3, α2 = 0, and
β = 0.06; b Traction of a squared lattice with α1 = 2.6, α2 = 0.6, and β = 0.08; c Traction of
a holed hexagonal lattice with α1 = 2.6, α2 = 0.4, and β = 0.12; d Traction of a holed squared
lattice with α1 = 2.4, α2 = 0.8, and β = 0.08; e Compression of a hexagonal lattice with α1 = 0.6,
α2 = 3.0, and β = 0.05; f Compression of a holed hexagonal lattice with α1 = 1.8, α2 = 3.0, and
β=0

the mean considered neighbors is Nk , for every pseudo-time step the algorithm
evaluate N × Nk interactions and so its complexity is linear O(N );
• Our model is highly suitable for different kinds of situations: in fact, changing
the model’s parameters, like its structure or α and β, it is possible to obtain high
variety of responses different among them.
• it can describe complex exotic structure (i.e., deformation bands) by only setting
the right parameters and without external modifications.
But our model, as every other model, has also disadvantages that can sum up in:
• now it can only supply qualitative results. There isn’t yet a connection between
the constitutive parameters of continua and the parameters of the model. This
connection will be investigated in future works;
• the model suffers the superpositions of points; in order to avoid this problem, some
precautions have to be taken (e.g., low leader velocities and low α value), which
slow down the total computation’s time. These precautions and slow down begin
118 P. D’Avanzo et al.

Fig. 6.31 In orange circle, the auto-interpenetration problem occurs, meanwhile in the green circle,
we can see the superposition problem

more and more relevant as the number of points increases; another example of
superposition in shown in Fig. 6.31;
• in this chapter, there is no attempt to avoid auto-interpenetration of the swarm’s
pieces far each other, e.g., if in a beam the two opposite edges are stressed in a
way that they occupy the same region of the space (see Fig. 6.31).

References

Abali BE, Völlmecke C, Woodward B, Kashtalyan M, Guz I, Müller WH (2012) Numerical mod-
eling of functionally graded materials using a variational formulation. Contin Mech Thermodyn
24(4–6):377–390
Abali BE, Wu CC, Müller WH (2016) An energy-based method to determine material constants in
nonlinear rheology with applications. Contin Mech Thermodyn 28(5):1221–1246
Abdoul-Anziz H, Seppecher P, Bellis C (2019) Homogenization of frame lattices leading to sec-
ond gradient models coupling classical strain and strain-gradient terms. Math Mech Solids
24(12):3976–3999
Alibert JJ, Seppecher P, dell’Isola F (2003) Truss modular beams with deformation energy depending
on higher displacement gradients. Math Mech Solids 8(1):51–73
6 From the Swarm Robotics to Material Deformations 119

Altenbach H, Eremeyev V (2015) On the constitutive equations of viscoelastic micropolar plates

and shells of differential type. Mathematics and Mechanics of Complex Systems 3(3):273–283
Altenbach H, Eremeyev VA (2008a) Analysis of the viscoelastic behavior of plates made of func-
tionally graded materials. ZAMM-Zeitschrift für Angewandte Mathematik und Mechanik 88(5),
332–341
Altenbach H, Eremeyev VA (2008b) Direct approach-based analysis of plates composed of func-
tionally graded materials. Archive of Applied Mechanics 78(10), 775–794
Altenbach H, Eremeyev VA (2011) On the shell theory on the nanoscale with surface stresses. Int
J Eng Sci 49(12):1294–1301
Altenbach H, Eremeyev VA, Naumenko K (2015) On the use of the first order shear deformation
plate theory for the analysis of three-layer plates with thin soft core layer. ZAMM-Zeitschrift für
Angewandte Mathematik und Mechanik 95(10):1004–1011
Altenbach J, Altenbach H, Eremeyev VA (2010) On generalized Cosserat-type theories of plates
and shells: a short review and bibliography. Arch Appl Mech 80(1):73–92
Andreaus U, dell’Isola F, Giorgio I, Placidi L, Lekszycki T, Rizzi NL (2016) Numerical simulations
of classical problems in two-dimensional (non) linear second gradient elasticity. Int J Eng Sci
108:34–50
Andreaus U, Spagnuolo M, Lekszycki T, Eugster SR (2018) A Ritz approach for the static analysis
of planar pantographic structures modeled with nonlinear Euler-Bernoulli beams. Contin Mech
Thermodyn 30(5):1103–1123
Barchiesi E, Eugster SR, Placidi L, dell’Isola F (2019a) Pantographic beam: A complete second
gradient 1D-continuum in plane. Z Angew Math Phys 70(5):135
Barchiesi E, Ganzosch G, Liebold C, Placidi L, Grygoruk R, Müller WH (2019b) Out-of-plane
buckling of pantographic fabrics in displacement-controlled shear tests: experimental results and
model validation. Contin Mech Thermodyn 31(1):33–45
Barchiesi E, Spagnuolo M, Placidi L (2019c) Mechanical metamaterials: a state of the art. Math
Mech Solids 24(1):212–234
Barchiesi E, dell’Isola F, Hild F, Seppecher P (2020a) Two-dimensional continua capable of large
elastic extension in two independent directions: asymptotic homogenization, numerical simula-
tions and experimental evidence. Mech Res Commun 103(103):466
Barchiesi E, Eugster SR, dell’Isola F, Hild F (2020b) Large in-plane elastic deformations of bi-
pantographic fabrics: asymptotic homogenization and experimental validation. Math Mech Solids
25(3):739–767
Barchiesi E, dell’Isola F, Bersani AM, Turco E (2021a) Equilibria determination of elastic articulated
duoskelion beams in 2D via a Riks-type algorithm. Int J Non-Linear Mech 128:1–24
Barchiesi E, dell’Isola F, Hild F (2021b) On the validation of homogenized modeling for bi-
pantographic metamaterials via digital image correlation. Int J Solids Struct 208:49–62
Baroudi D, Giorgio I, Battista A, Turco E, Igumnov LA (2019) Nonlinear dynamics of uni-
formly loaded elastica: experimental and numerical evidence of motion around curled sta-
ble equilibrium configurations. ZAMM-Zeitschrift für Angewandte Mathematik und Mechanik
99(7):e201800,121
Battista A, Rosa L, Dell’Erba R, Greco L (2016) Numerical investigation of a particle system
compared with first and second gradient continua: deformation and fracture phenomena. Math
Mech Solids 22. https://doi.org/10.1177/1081286516657889
Battista A, D’Avanzo P, Laudato M (2018) Discrete systems with geometry-driven evolution: appli-
cation to 1d elasticity and granular media. Mech Res Commun 92:107–110
Bilotta A, Morassi A, Turco E (2018) The use of quasi-isospectral operators for damage detection
in rods. Meccanica 53(1):319–345
Bilotta A, Morassi A, Rosset E, Turco E, Vessella S (2019) Numerical size estimates of inclusions
in Kirchhoff-Love elastic plates. Int J Solids Struct 168:58–72
Bolzon G, Fedele R, Maier G (2002) Parameter identification of a cohesive crack model by kalman
filter. Comput Methods Appl Mech Eng 191(25–26):2847–2871
120 P. D’Avanzo et al.

Boutin C, Giorgio dell’Isola I F, Placidi L (2017) Linear pantographic sheets: asymptotic micro-
macro models identification. Math Mech Complex Syst 5(2):127–162
Browning F, Askes H (2019) Analytical solutions for the natural frequencies of rectangular sym-
metric angle-ply laminated plates. Mathematics and Mechanics of Complex Systems 7(1):51–61
Casale L, Avella M, Dell’Erba R, Martuscelli E (1998) Broom fibres as reinforcements for thermo-
plastic matrices. Macromol Symposia 127(1):211–218
Cazzani A, Ruge P (2016) Stabilization by deflation for sparse dynamical systems without loss of
sparsity. Mechanical Systems and Signal Processing 70:664–681
Cazzani A, Malagù M, Turco E (2016a) Isogeometric analysis of plane-curved beams. Math Mech
Solids 21(5):562–577
Cazzani A, Malagù M, Turco E, Stochino F (2016b) Constitutive models for strongly curved beams
in the frame of isogeometric analysis. Math Mech Solids 21(2):182–209
Cazzani A, Stochino F, Turco E (2016c) An analytical assessment of finite element and isogeomet-
ric analyses of the whole spectrum of timoshenko beams. ZAMM-Zeitschrift für Angewandte
Mathematik und Mechanik 96(10):1220–1244
Cazzani A, Rizzi NL, Stochino F, Turco E (2018a) Modal analysis of laminates by a mixed assumed-
strain finite element model. Math Mech Solids 23(1):99–119
Cazzani A, Serra M, Stochino F, Turco E (2018b) A refined assumed strain finite element model for
statics and dynamics of laminated plates. Continuum Mechanics and Thermodynamics pp 1–28,
https://doi.org/10.1007/s00161-018-0707-x
Contrafatto L, Cuomo M (2002) A new thermodynamically consistent continuum model for hard-
ening plasticity coupled with damage. International Journal of Solids and Structures 39(25),
6241–6271
Cuomo M (2017) Forms of the dissipation function for a class of viscoplastic models. Mathematics
and Mechanics of Complex Systems 5(3):217–237
Cuomo M (2019) Continuum damage model for strain gradient materials with applications to 1D
examples. Continuum Mech Thermodyn 31(4):969–987
Cuomo M, Contrafatto L, Greco L (2014) A variational model based on isogeometric interpolation
for the analysis of cracked bodies. Int J Eng Sci 80:173–188
Cuomo M, dell’Isola F, Greco L (2016) Simplified analysis of a generalized bias test for fabrics
with two families of inextensible fibres. Z Angew Math Phys 67(3):61
De Angelo M, Barchiesi E, Giorgio I, Abali BE (2019a) Numerical identification of constitutive
parameters in reduced-order bi-dimensional models for pantographic structures: application to
out-of-plane buckling. Arch Appl Mech 89(7):1333–1358
De Angelo M, Spagnuolo M, D’Annibale F, Pfaff A, Hoschke K, Misra A, Dupuy C, Peyre P, Dirren-
berger J, Pawlikowski M (2019b) The macroscopic behavior of pantographic sheets depends
mainly on their microstructure: experimental evidence and qualitative analysis of damage in
metallic specimens. Contin Mech Thermodyn 31(4):1181–1203
De Angelo M, Placidi L, Nejadsadeghi N, Misra A (2020) Non-standard timoshenko beam model
for chiral metamaterial: Identification of stiffness parameters. Mech Res Commun 103(103):462
Del Piero G (2018) The variational structure of classical plasticity. Mathematics and Mechanics of
Complex Systems 6(3):137–180
Del Vescovo D, Giorgio I (2014) Dynamic problems for metamaterials: review of existing models
and ideas for further research. International Journal of Engineering Science 80:153–172
Della Corte A, Battista A, Dell’Isola F (2015) Referential description of the evolution of a 2D
swarm of robots interacting with the closer neighbors: perspectives of continuum modeling via
higher gradient continua. Int J Non-Linear Mech 12. https://doi.org/10.1016/j.ijnonlinmec.2015.
06.016, https://hal.archives-ouvertes.fr/hal-01236021
Dell’Erba R (2001) Rheo-mechanical and rheo-optical characterisation of ultra high molecu-
lar mass poly(methylmethacrylate) in solution. Polymer 42:2655–2663, DOI: 10.1016/S0032-
3861(00)00636-4
Dell’Erba R (2015) Determination of spatial configuration of an underwater swarm with minimum
data. Int J Adv Robot Syst 12(7):97
6 From the Swarm Robotics to Material Deformations 121

Dell’Erba R (2018) Position-based dynamic of a particle system: a configurable algorithm to

describe complex behaviour of continuum material starting from swarm robotics. Contin Mech
Thermodyn 30. https://doi.org/10.1007/s00161-018-0663-5
Dell’Erba R (2019a) On how swarm robotics can be used to describe particle system’s deformation.
Continuum Mech Thermodyn. https://doi.org/10.1007/s00161-019-00845-4
Dell’Erba R (2019b) Swarm robotics and complex behaviour of continuum material. Continuum
Mech Thermodyn 31:1–26. https://doi.org/10.1007/s00161-018-0675-1
Dell’Erba R (2020a) The distances measurement problem for an underwater robotic swarm: a semi-
experimental trial, using power leds, in unknown sea water conditions. Contin Mech Thermodyn
1–9. https://doi.org/10.1007/s00161-020-00923-y
Dell’Erba R (2020b) A tool to describe particle system evolution from swarm robotics behavior. In:
Developments and novel approaches in nonlinear solid body mechanics. Springer, pp 191–217
Dell’Erba R (2021a) Distance estimations in unknown sea underwater conditions by power LED
for robotics swarms. Contin Mech Thermodyn
Dell’Erba R (2021b) Flocking rules governing swarm robot as tool to describe continuum defor-
mation. In: Dynamics, strength of materials and durability in multiscale mechanics. Springer, pp
223–243
Dell’Erba R (2021c) A plausible description of continuum material behavior derived by swarm
robot flocking rules. In: Dynamics strength of materials and durability in multiscale mechanics.
Springer, pp 329–378
Dell’Isola F, Andreaus U, Placidi L (2015a) At the origins and in the vanguard of peridynamics, non-
local and higher-gradient continuum mechanics: an underestimated and still topical contribution
of Gabrio Piola. Math Mech Solids 20(8):887–928
Dell’Isola F, Lekszycki T, Pawlikowski M, Grygoruk R, Greco L (2015b) Designing a light fabric
metamaterial being highly macroscopically tough under directional extension: first experimental
evidence. Z Angew Math Phys 66(6):3473–3498
Dell’Isola F, Seppecher P, Alibert JJ et al (2019) Pantographic metamaterials: an example of mathe-
matically driven design and of its technological challenges. Contin Mech Thermodyn 31(4):851–
884
Desmorat B, Spagnuolo M, Turco E (2020) Stiffness optimization in nonlinear pantographic struc-
tures. Math Mech Solids 25(12):2252–2262
Epstein M, Smelser R (2020) An appreciation and discussion of paul germain’s “the method of
virtual power in the mechanics of continuous media, i: second-gradient theory.” Mathematics
and Mechanics of Complex Systems 8(2):191–199
Eremeyev VA, Pietraszkiewicz W (2016) Material symmetry group and constitutive equations of
micropolar anisotropic elastic solids. Mathematics and Mechanics of Solids 21(2), 210–221
Eremeyev VA, Turco E (2020) Enriched buckling for beam-lattice metamaterials. Mechanics
Research Communications 103:103,458
Eugster S, Hesch C, Betsch P, Glocker C (2014) Director-based beam finite elements relying on
the geometrically exact beam theory formulated in skew coordinates. Int J Numer Meth Eng
97(2):111–129
Eugster S, dell’Isola F, Steigmann D (2019) Continuum theory for mechanical metamaterials with
a cubic lattice substructure. Math Mech Complex Syst 7(1):75–98
Franciosi P, Spagnuolo M, Salman OU (2019) Mean Green operators of deformable fiber networks
embedded in a compliant matrix and property estimates. Contin Mech Thermodyn 31(1):101–132
Gagneux G, Millet O (2016) Modeling capillary hysteresis in unsatured porous media. Mathematics
and Mechanics of Complex Systems 4(1):67–77
George D, Allena R, Remond Y (2018) A multiphysics stimulus for continuum mechanics bone
remodeling. Math Mech Complex Syst 6(4):307–319
George D, Allena R, Bourzac C, Pallu S, Bensidhoum M, Portier H, Rémond Y (2020) A new
comprehensive approach for bone remodeling under medium and high mechanical load based on
cellular activity. Math Mech Complex Syst 8(4):287–306
122 P. D’Avanzo et al.

Germain P (2020) The method of virtual power in the mechanics of continuous media, i: Second-
gradient theory. Math Mech Complex Syst 8(2):153–190
Giorgio I (2020a) A discrete formulation of Kirchhoff rods in large-motion dynamics. Mathematics
and Mechanics of Solids 25(5), 1081–1100
Giorgio I (2020b) Lattice shells composed of two families of curved Kirchhoff rods: An archetypal
example, topology optimization of a cycloidal metamaterial. Contin Mech Thermodyn. https://
doi.org/10.1007/s00161-020-00955-4
Giorgio I, Del Vescovo D (2018) Non-linear lumped-parameter modeling of planar multi-link manip-
ulators with highly flexible arms. Robotics 7(4):60
Giorgio I, Del Vescovo D (2019) Energy-based trajectory tracking and vibration control for multilink
highly flexible manipulators. Mathematics and Mechanics of Complex Systems 7(2):159–174
Giorgio I, Scerrato D (2017) Multi-scale concrete model with rate-dependent internal friction.
European Journal of Environmental and Civil Engineering 21(7–8), 821–839
Giorgio I, Andreaus U, Scerrato D, dell’Isola F (2016) A visco-poroelastic model of functional
adaptation in bones reconstructed with bio-resorbable materials. Biomech Model Mechanobiol
15(5):1325–1343
Giorgio I, Andreaus U, dell’Isola F, Lekszycki T (2017a) Viscous second gradient porous materials
for bones reconstructed with bio-resorbable grafts. Extreme Mech Lett 13:141–147
Giorgio I, Rizzi NL, Turco E (2017b) Continuum modelling of pantographic sheets for out-of-plane
bifurcation and vibrational analysis. Proc R Soc A Math Phys Eng Sci 473(2207):20170,636
Giorgio I, Harrison P, dell’Isola F, Alsayednoor J, Turco E (2018) Wrinkling in engineering fabrics:
a comparison between two different comprehensive modelling approaches. Proc R Soc A Math
Phys Eng Sci 474(2216):20180,063
Giorgio I, De Angelo M, Turco E, Misra A (2019a) A Biot–Cosserat two-dimensional elastic
nonlinear model for a micromorphic medium. Contin Mech Thermodyn 1–13
Giorgio I, dell’Isola F, Andreaus U, Alzahrani F, Hayat T, Lekszycki T (2019b) On mechanically
driven biological stimulus for bone remodeling as a diffusive phenomenon. Biomech Model
Mechanobiol 18(6):1639–1663
Giorgio I, Rizzi NL, Andreaus U, Steigmann DJ (2019c) A two-dimensional continuum model of
pantographic sheets moving in a 3D space and accounting for the offset and relative rotations of
the fibers. Math Mech Complex Syst 7(4):311–325
Giorgio I, Ciallella A, Scerrato D (2020a) A study about the impact of the topological arrangement
of fibers on fiber-reinforced composites: some guidelines aiming at the development of new
ultra-stiff and ultra-soft metamaterials. Int J Solids Struct 203:73–83
Giorgio I, Spagnuolo M, Andreaus U, Scerrato D, Bersani AM (2020b) In-depth gaze at the astonish-
ing mechanical behavior of bone: A review for designing bio-inspired hierarchical metamaterials.
Math Mech Solids. https://doi.org/10.1177/1081286520978516
Giorgio I, Varano V, dell’Isola F, Rizzi NL (2021) Two layers pantographs: a 2D continuum model
accounting for the beams’ offset and relative rotations as averages in SO(3) Lie groups. Int J
Solids Struct 216:43–58
Greco F, Luciano R, Serino G, Vaiana N (2018a) A mixed explicit-implicit time integration approach
for nonlinear analysis of base-isolated structures. Ann Solid Struct Mech 10(1):17–29
Greco L (2020) An iso-parametric G 1 -conforming finite element for the nonlinear analysis of
Kirchhoff rod. Part I: the 2D case. Contin Mech Thermodyn 1–24
Greco L, Cuomo M, Contrafatto L (2018b) A reconstructed local B formulation for isogeometric
Kirchhoff-Love shells. Comput Methods Appl Mech Eng 332:462–487
Javili A, Morasata R, Oterkus E, Oterkus S (2019) Peridynamics review. Math Mech Solids
24(11):3714–3739
Marmo F, Masi D, Rosati L (2018a) Thrust network analysis of masonry helical staircases. Int J
Archit Herit 12(5):828–848
Marmo F, Ruggieri N, Toraldo F, Rosati L (2018b) Historical study and static assessment of an
innovative vaulting technique of the 19th century. Int J Archit Herit
6 From the Swarm Robotics to Material Deformations 123

Marmo F, Demartino C, Candela G, Sulpizio C, Briseghella B, Spagnuolo R, Xiao Y, Vanzi I, Rosati

L (2019) On the form of the musmeci’s bridge over the basento river. Eng Struct 191:658–673
Misra A, Poorsolhjouy P (2016a) Elastic behavior of 2D grain packing modeled as micromorphic
media based on granular micromechanics. J Eng Mech 143(1):C4016,005
Misra A, Poorsolhjouy P (2016b) Granular micromechanics based micromorphic model predicts
frequency band gaps. Continuum Mech Thermodyn 28(1–2):215–234
Misra A, Nejadsadeghi N, De Angelo M, Placidi L (2020) Chiral metamaterial predicted by granular
micromechanics: verified with 1d example synthesized using additive manufacturing. Contin
Mech Thermodyn 1–17
Müller M, Heidelberger B, Teschner M, Gross M (2005) Meshless deformations based on shape
matching. ACM Trans Graph 24:471–478. https://doi.org/10.1145/1186822.1073216
Müller M, Heidelberger B, Hennix M, Ratcliff J (2007) Position based dynamics. J Vis Commun
Image Represent 18:109–118. https://doi.org/10.1016/j.jvcir.2007.01.005
Moriconi C, Dell’Erba R (2014) Social dependability: a proposed evolution for future robotics.
https://doi.org/10.12910/EAI2014-63
Nejadsadeghi N, De Angelo M, Drobnicki R, Lekszycki T, dell’Isola F, Misra A (2019) Parametric
experimentation on pantographic unit cells reveals local extremum configuration. Exp Mech
59(6):927–939
Niiranen J, Khakalo S, Balobanov V, Niemi AH (2016) Variational formulation and isogeometric
analysis for fourth-order boundary value problems of gradient-elastic bar and plane strain/stress
problems. Comput Methods Appl Mech Eng 308:182–211
Olive M (2019) Effective computation of SO(3) and O(3) linear representation symmetry classes.
Mathematics and Mechanics of Complex Systems 7(3):203–237
Paradiso M, Marmo F, Rosati L (2019) Consistent derivation of a beam model from the saint venant’s
solid model. Int J Solids Struct 159:90–110
Paradiso M, Sessa S, Vaiana N, Marmo F, Rosati L (2021) Shear properties of isotropic and homo-
geneous beam-like solids having arbitrary cross sections. Int J Solids Struct
Perricone V, Grun T, Marmo F, Langella C, Carnevali MDC (2020) Constructional design of echi-
noid endoskeleton: main structural components and their potential for biomimetic applications.
Bioinspiration Biomim
Placidi L (2016) A variational approach for a nonlinear one-dimensional damage-elasto-plastic
second-gradient continuum model. Continuum Mech Thermodyn 28(1–2):119–137
Placidi L, Andreaus U, Giorgio I (2017) Identification of two-dimensional pantographic structure
via a linear d4 orthotropic second gradient elastic model. J Eng Math 103(1):1–21
Placidi L, Barchiesi E, Misra A (2018a) A strain gradient variational approach to damage: a compar-
ison with damage gradient models and numerical results. Math Mech Complex Syst 6(2):77–100
Placidi L, Misra A, Barchiesi E (2018b) Two-dimensional strain gradient damage modeling: a
variational approach. Z Angew Math Phys 69(3):56
Rapisarda A, Della Corte A, Drobnicki R, Cosmo F, Rosa L (2018) A model for bone mechanics
and remodeling including cell populations dynamics. Zeitschrift für angewandte Mathematik und
Physik 70. https://doi.org/10.1007/s00033-018-1055-1
Rapisarda A, Almasi M, Almasi N, Barchiesi E, Della Corte A, Scerrato D (2020) Bone mechanics
and cell populations: mathematical description and parametric study of the model, pp 107–126.
https://doi.org/10.1007/978-3-030-50464-9_7
Rickert W, Vilchevskaya E, Müller W (2019) A note on Couette flow of micropolar fluids according
to Eringen’s theory. Math Mech Complex Syst 7(1):25–50
Rosi G, Placidi L, Auffray N (2018) On the validity range of strain-gradient elasticity: a mixed
static-dynamic identification procedure. Eur J Mech A/Solids 69:179–191
Şahin E (2005) Swarm robotics: from sources of inspiration to domains of application, pp 10–20
Scerrato D, Giorgio I (2019) Equilibrium of two-dimensional cycloidal pantographic metamaterials
in three-dimensional deformations. Symmetry 11(12):1523
Scerrato D, Giorgio I, Madeo A, Limam A, Darve F (2014) A simple non-linear model for internal
friction in modified concrete. Int J Eng Sci 80:136–152
124 P. D’Avanzo et al.

Scerrato D, Zhurba Eremeeva IA, Lekszycki T, Rizzi NL (2016) On the effect of shear stiffness
on the plane deformation of linear second gradient pantographic sheets. ZAMM-Zeitschrift für
Angewandte Mathematik und Mechanik 96(11):1268–1279
Schulte J, Dittmann M, Eugster S, Hesch S, Reinicke T, Dell’Isola F, Hesch C (2020) Isogeometric
analysis of fiber reinforced composites using kirchhoff-love shell elements. Comput Methods
Appl Mech Eng 362(112):845
Sciarra G, dell’Isola F, Coussy O (2007) Second gradient poromechanics. Int J Solids Struct
44(20):6607–6629
Seppecher P, Alibert JJ, dell’Isola F (2011) Linear elastic trusses leading to continua with exotic
mechanical interactions. J Phys. Conference Series, IOP Publishing, vol 319, p 012018
Serpieri R, Sessa S, Rosati L (2018) A mitc-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220
Sessa S, Marmo F, Rosati L (2015) Effective use of seismic response envelopes for reinforced
concrete structures. Earthq Eng Struct Dyn 44(14):2401–2423
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through-the-thickness jacketed shells for
the elasto-plastic analysis of confined composite structures: Theory and numerical assessment.
Compos B Eng 113:225–242
Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018a) Evaluation of the capacity
surfaces of reinforced concrete sections: Eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512
Sessa S, Marmo F, Vaiana N, Rosati L (2018b) A computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthq Eng 1–34
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos B Eng 163:413–423
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force-biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54(9):1451–1469
Spagnuolo M (2020) Circuit analogies in the search for new metamaterials: Phenomenology of a
mechanical diode. In: Nonlinear wave dynamics of materials and structures. Springer, pp 411–422
Spagnuolo M, Andreaus U (2019) A targeted review on large deformations of planar elastic beams:
extensibility, distributed loads, buckling and post-buckling. Mathematics and Mechanics of Solids
24(1), 258–280
Spagnuolo M, Scerrato D (2020) The mechanical diode: on the tracks of James Maxwell employing
mechanical–electrical analogies in the design of metamaterials. In: Developments and novel
approaches in biomechanics and metamaterials. Springer, pp 459–469
Spagnuolo M, Barcz K, Pfaff A, dell’Isola F, Franciosi P (2017) Qualitative pivot damage analysis in
aluminum printed pantographic sheets: numerics and experiments. Mech Res Commun 83:47–52
Spagnuolo M, Yildizdag ME, Andreaus U, Cazzani AM (2020) Are higher-gradient models also
capable of predicting mechanical behavior in the case of wide-knit pantographic structures? Math
Mech Solids. https://doi.org/10.1177/1081286520937339
Steigmann DJ (2018) Equilibrium of elastic lattice shells. Journal of Engineering Mathematics
109(1), 47–61
Steigmann DJ, dell’Isola F (2015) Mechanical response of fabric sheets to three-dimensional bend-
ing, twisting, and stretching. Acta Mechanica Sinica 31(3), 373–382
Timofeev D, Barchiesi E, Misra A, Placidi L (2020) Hemivariational continuum approach for gran-
ular solids with damage-induced anisotropy evolution. Math Mech Solids 1081286520968149
Tran LV, Niiranen J (2020) A geometrically nonlinear Euler-Bernoulli beam model within strain
gradient elasticity with isogeometric analysis and lattice structure applications. Mathematics and
Mechanics of Complex Systems 8(4):345–371
Turco E (2018a) Discrete is it enough? The revival of Piola-Hencky keynotes to analyze three-
dimensional Elastica. Continuum Mech Thermodyn 30(5):1039–1057
Turco E (2018b) In-plane shear loading of granular membranes modeled as a Lagrangian assembly
of rotating elastic particles. Mechanics Research Communications 92:61–66
6 From the Swarm Robotics to Material Deformations 125

Turco E (2019) Numerically driven tuning of equilibrium paths for pantographic beams. Continuum
Mech Thermodyn 31(6):1941–1960
Turco E, Barchiesi E (2019) Equilibrium paths of Hencky pantographic beams in a three-point
bending problem. Mathematics and Mechanics of Complex Systems 7(4):287–310
Turco E, dell’Isola F, Cazzani A, Rizzi NL (2016a) Hencky-type discrete model for pantographic
structures: numerical comparison with second gradient continuum models. Z Angew Math Phys
67(4):85
Turco E, dell’Isola F, Rizzi NL, Grygoruk R, Müller WH, Liebold C (2016b) Fiber rupture in
sheared planar pantographic sheets: numerical and experimental evidence. Mech Res Commun
76:86–90
Turco E, dell’Isola F, Misra A (2019) A nonlinear Lagrangian particle model for grains assemblies
including grain relative rotations. Int J Numer Anal Meth Geomech 43(5):1051–1079
Turco E, Barchiesi E, Giorgio I, dell’Isola F (2020) A Lagrangian Hencky-type non-linear model
suitable for metamaterials design of shearable and extensible slender deformable bodies alterna-
tive to Timoshenko theory. Int J Non-Linear Mech 123(103):481
Vaiana N, Spizzuoco M, Serino G (2017) Wire rope isolators for seismically base-isolated
lightweight structures: experimental characterization and mathematical modeling. Eng Struct
140:498–514
Vaiana N, Sessa S, Marmo F, Rosati L (2018) A class of uniaxial phenomenological models for sim-
ulating hysteretic phenomena in rate-independent mechanical systems and materials. Nonlinear
Dyn 93(3):1647–1669
Vaiana N, Sessa S, Marmo F, Rosati L (2019) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901
Vaiana N, Capuano R, Sessa S, Marmo F, Rosati L (2021) Nonlinear dynamic analysis of seismically
base-isolated structures by a novel opensees hysteretic material model. Appl Sci 11(3):900
Valoroso N, Marmo F, Sessa S (2015) A novel shell element for nonlinear pushover analysis of
reinforced concrete shear walls. Bull Earthq Eng 13(8):2367–2388
Vangelatos Z, Gu GX, Grigoropoulos CP (2019) Architected metamaterials with tailored 3D buck-
ling mechanisms at the microscale. Extreme Mech Lett 33(100):580
Wiech J, Eremeyev VA, Giorgio I (2018) Virtual spring damper method for nonholonomic robotic
swarm self-organization and leader following. Contin Mech Thermodyn 30(5):1091–1102
Yang H, Ganzosch G, Giorgio I, Abali BE (2018) Material characterization and computations of a
polymeric metamaterial with a pantographic substructure. Z Angew Math Phys 69(4):1–16
Yang H, Abali BE, Timofeev D, Müller WH (2019) Determination of metamaterial parameters by
means of a homogenization approach based on asymptotic analysis. Contin Mech Thermodyn
1–20
Yang Y, Misra A (2012) Micromechanics based second gradient continuum theory for shear band
modeling in cohesive granular materials following damage elasticity. International Journal of
Solids and Structures 49(18), 2500–2514
Yildizdag ME, Tran CA, Barchiesi E, Spagnuolo M, dell’Isola F, Hild F (2019) A multi-disciplinary
approach for mechanical metamaterial synthesis: A hierarchical modular multiscale cellular struc-
ture paradigm. In: State of the art and future trends in material modeling. Springer, pp 485–505
Yildizdag ME, Barchiesi E, dell’Isola F (2020a) Three-point bending test of pantographic blocks:
numerical and experimental investigation. Math Mech Solids 25(10):1965–1978
Yildizdag ME, Demirtas M, Ergin A (2020b) Multipatch discontinuous galerkin isogeometric anal-
ysis of composite laminates. Contin Mech Thermodyn 32(3):607–620
Chapter 7
A Review of the Class of Bouc-Wen
Differential Models for Simulating
Mechanical Hysteresis Phenomena

Davide Pellecchia and Massimo Paradiso

Abstract One of the most popular hysteretic models used in many areas of engi-
neering, especially the civil one, is the Bouc-Wen model. Although this model is able
to simulate several types of hysteretic phenomena, it cannot describe some typical
phenomena, such as cyclic degradation of strength and stiffness, pinching effect, and
so on. For this reason, many researchers have proposed several variants of the original
Bouc-Wen model. We present a review of the Bouc-Wen model and its most signif-
icant enhanced versions, utilizing the same technical terminology for all models in
order to clarify and to shed some light on the number and physical significance of the
parameters that the models require as input. Sensitivity analyses are also illustrated
with respect to the input parameters.

Keywords Civil engineering · Hysteresis · Bouc-Wen model · Cyclic

degradation · Sensitivity analysis

7.1 Introduction

Hysteresis is a complex phenomenon that can be experienced in many fields of science

and technology; undoubtedly it represents the predominant typology of nonlinear
constitutive behavior. The importance of properly reproducing hysteretic responses
in engineering has been highlighted by several contributions available in the litera-
ture (Visintin 2013). The main research fields include civil applications (Bahn and
Hsu 1998; Lima et al. 2018; Zuccaro et al. 2017), magnetism (Bai et al. 2019), as
well as higher-gradient materials (Alibert et al. 2003; Pideri and Seppecher 1997;
Barchiesi et al. 2018), and mechanics of thermal and porous media (Altenbach et al.

D. Pellecchia (B) · M. Paradiso

Department of Structures for Engineering and Architecture, University
of Naples Federico II, via Claudio 21, 80125 Naples, Italy
e-mail: [email protected]
M. Paradiso
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 127

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_7
128 D. Pellecchia and M. Paradiso

2012; Eremeyev and Morozov 2010). Beyond basic applications such as the model-
ing of seismic devices (Kikuchi and Aiken 1997), dampers (Nuzzo et al. 2019), and
concrete (Sessa et al. 2018, 2019b), hysteresis plays a significant role also for the
analysis of more complex mechanical systems (Badoni and Makris 1996; Song et al.
2007; Greco and Cuomo 2013) including applications concerning framed (Marmo
et al. 2011; Marmo and Rosati 2012a, b, 2013) and shell structures (Caggegi et al.
2018; Serpieri et al. 2018; Ascione et al. 2017; Sessa et al. 2017, 2019a; Valoroso et al.
2014, 2015), structural identification (Noël and Kerschen 2017), and random vibra-
tions (Baber and Noori 1985; Sessa 2010; Wen 1976). More recent developments
concerns meta-materials (Turco et al. 2017, 2018; De Angelo et al. 2019; di Cosmo
et al. 2018; Andreaus et al. 2018) based on pantographic microstructures (Barchiesi
et al. 2020; dell’Isola et al. 2019a, 2019b; Nejadsadeghi et al. 2019) as well as the
modeling of damage (Contrafatto and Cuomo 2002; Contrafatto et al. 2012; Placidi
et al. 2018, 2019).
The output of hysteretic systems and materials typically depends on present and
past histories of the input variable and can exhibit different peculiar features: in
particular, when the first time derivative of the input variable does not influence the
output, this hysteresis phenomenon is denominated rate-independent.
The development of mathematical models able to describe such nonlinear phe-
nomena is very complicated. In particular, in the last few years, many researchers
have proposed different models whose common objective was not to explain the
physical origin of the hysteresis but to try to reproduce the overall experimental
behavior (Mayergoyz 2003). These models are called phenomenological models.
It is possible to classify the phenomenological models according to the nature of
the equation to solve for the evaluation of the output variable, namely the generalized
force or the generalized displacement. In particular one has:
• algebraic models, such as the ones developed by Ramberg and Osgood Ramberg
and Osgood (1943), Menegotto and Pinto Menegotto (1973), and Vaiana et al.
(2019a, b, c, 2020, 2021b, a, c);
• trascendental models, such as the ones introduced by Kikuchi and Aiken Kikuchi
and Aiken (1997) and Sessa et al. (2020); Vaiana et al. (2018);
• differential models, such as those formulated by Bouc (1967, 1971),
Özdemir Özdemir (1976), and Wen (1976, 1980);
• damage-based models, such as the one proposed by Fedele et al. (2012), Sessa
and Valoroso (2017), Valoroso and Fedele (2010), Valoroso et al. (2013).
Among existing models, the differential ones are currently the most used models
to reproduce the behavior of mechanical systems and materials. These models are
typically based on the Duhem hysteresis operator (Duhem 1897) whose formulation
is defined by a Cauchy problem of the form:

ż(t) = g1 (x, z) ẋ(t)+ + g2 (x, z) ẋ(t)− ,
(7.1)
z(0) = z 0 ,
7 A Review of the Class of Bouc-Wen Differential Models … 129

in which z(t) and x(t) denote the hysteretic functions and the generalized displace-
ment, respectively, and the superimposed dot denotes the derivative with respect
to time t; g1 and g2 are continuous functions whereas ẋ(t)+ = max(0, ẋ(t)) and
ẋ(t)− = min(0, ẋ(t)); finally z 0 represents the value of the function z(t) at the time
t = 0. All models based on Duhem’s class are characterized by a peculiar prop-
erty: the output value can be evaluated if the current values of the input and output
variables (x, z) as well as the sign of the first derivative with respect to time of the
input variable (sign(ẋ)) are known Dimian and Andrei (2014). Examples of differ-
ential models based on the Duhem hysteresis operator are the ones proposed by Jiles
and Atherton (1983, 1984), Hodgdon (1988), Bouc (1971), Wen (1976, 1980), and
Özdemir (1976).
The present work aims to illustrate the evolution of the Bouc-Wen model in the
area of mechanics. In particular, such an evolution is described with reference to
the modeling of symmetric and asymmetric hysteresis loops, hysteresis loops with
pinching, and hysteresis loops with strength and/or stiffness degradation typically
displayed by rate-independent mechanical systems and materials. The influence of
the input parameters on the dimension and/or shape of the hysteresis loops is shown
and discussed for each hysteretic model.
This paper is organized into four parts. In Sect. 7.2, we review some models able
to describe symmetric hysteresis behaviors, namely the Bouc model (Bouc 1967,
1971) and the Wen model (Wen 1976), the latter currently known as the Bouc-Wen
model. In Sect. 7.3, we review some modified Bouc-Wen models able to describe
asymmetric hysteresis behaviors. In Sect. 7.4, the modified versions of the Bouc-Wen
model, proposed by some researchers in order to account for the pinching effect, are
illustrated. Finally, in Sect. 7.5, some models able to simulate both the strength and
stiffness degradation are described.

7.2 Modeling of Symmetric Hysteresis Loops

From a mathematical point of view, symmetric hysteresis loops are characterized

by odd functions with respect to the origin of the reference frame, i.e. hysteretic
functions z fulfilling the condition

z(x) = −z(−x). (7.2)

Among several mechanical systems and materials that exhibit a nonlinear behavior
characterized by symmetric hysteresis loops, we recall, as an example, the typical
cross section of steel and concrete filled steel elements (Colombo and Negro 2005;
Shih and Sung 2005), metal devices, such as wire rope isolators deforming along
their transverse directions (Vaiana et al. 2017), and seismic protection devices, such
as isolators (Greco et al. 2018; Hadad et al. 2017; Losanno et al. 2019a, b, 2021;
Sierra et al. 2019; Vaiana et al. 2019d; Pellecchia et al. 2020) and dampers (Nuzzo
et al. 2018, 2019). In Fig. 7.1, some examples of symmetric hysteresis loops obtained
in experimental tests are shown.
130 D. Pellecchia and M. Paradiso

(a) (b)

Fig. 7.1 Some symmetric hysteresis loops obtained in experimental tests. a The hysteresis behavior
of a bearing—Hadad et al. (2017). b The hysteresis behavior of a rhombic steel plate Shih and Sung
(2005)

The restoring force of the above-described mechanical systems and materials is

typically computed as follows:

f (x) = f e (x) + f h (x), (7.3)

where f e (x) is the elastic component whereas f h (x) is the hysteretic one. In turn,
the restoring force can be described in the following way:

f (x) = α k x + (1 − α) k z(x), (7.4)

in which α is ratio between the post-yield and pre-yield stiffness whereas k is defined
as the stiffness at yield, i.e. the ratio between the yield force and the generalized yield
displacement.

7.2.1 Bouc Model and Its Modified Versions

The full class of Bouc models is described by the following general nonlinear first-
order ordinary differential equation:

ż = B ẋ, (7.5)

in which ż denotes the time derivative of the hysteretic variable, required to evalu-
ate the rate-independent hysteretic component f h (x) = (1 − α) k z(x), whereas ẋ is
the generalized velocity. The hysteretic function z basically depends on the system
behavior, material properties, and response amplitude.
The nonlinear function B has been assumed of different forms over the years Bouc
(1967, 1971):
7 A Review of the Class of Bouc-Wen Differential Models … 131

Bouc model 1 (1967): B = A − z β sign(ẋ), (7.6a)

 
Bouc model 2 (1967): B = A − |z| γ + β sign(ẋ z) , (7.6b)
 
Wen model (1976): B = A − |z n | γ + β sign(ẋ z) (7.6c)

where A, n, and γ are material parameters that tune the size and/or the shape of the
hysteresis loops.
Equation (7.6a), that defines the original Bouc model, is characterized by two
model parameters, namely A and β. Subsequently, Bouc has modified such an equa-
tion by adding a new parameter, that is γ , as shown in Eq. (7.6b). Finally, Wen has
extended the class of the Bouc differential models by adding the parameter n in
order to smooth the hysteretic curve predicted by the original Bouc model. Note that
equation f (x) = α k x + (1 − α) k z and Eq. (7.6c) define the so-called Bouc-Wen
model.
The smooth nature of the Bouc model modified by Wen makes it particularly con-
venient for addressing several engineering problems especially when several dynamic
analyses are required. This includes the case of flutter analysis (Carboni et al. 2018)
and Random Vibration analysis of structures (Broccardo et al. 2017; Fujimura and
Der Kiureghian 2007; Sessa 2010).
The Bouc-Wen model is capable of reproducing several behaviors depending on
the parameters A, β, γ , and n, whose influence on the hysteretic variable z is illustrated
in Sect. 7.2.2.

7.2.2 Sensitivity Analysis

A parameter sensitivity analysis was carried out to evaluate the effect of each param-
eter on the hysteretic variable z(x) obtained by adopting Eqs. (7.5) and (7.6c).
The relationship between the hysteretic variable z and the generalized displace-
ment x is shown in Fig. 7.2 for different combinations of the constitutive parameters.
All hysteretic loops are obtained by applying a generalized displacement described
by the following sine wave:
x(t) = 2 sin(t) (7.7)

and integrating differential Eq. (7.5) by MATLAB ® using the solver ode45.
The top left plot shows that the tangent stiffness at the origin of the hysteresis loop
increases when the parameter A is increased and its sign is the same as that of A; for
negative values of parameter A, the tangent stiffness at the origin becomes negative.
The top right plot shows that the hysteretic energy dissipation increases as β
increases. In particular, an elastic nonlinear constitutive law can be obtained by
setting β = 0.
The bottom left plot shows that the hysteresis loop is bounded between two parallel
straight lines and it rotates clockwise when the parameter γ is increased. On the other
132 D. Pellecchia and M. Paradiso

1.5 1.5

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
-2 -1 0 1 2 -2 -1 0 1 2

6 1.5

4 1

2
0.5
0
0
-2
-0.5
-4

-6 -1

-8 -1.5
-2 -1 0 1 2 -2 -1 0 1 2

Fig. 7.2 Sensitivity analysis of the Bouc-Wen model with respect to the material parameters

hand, by decreasing the value of γ , the hysteresis loop is bounded by two parallel
curves such that, for high values of the displacement x, the hysteresis loop exhibits
a work hardening behavior.
Finally, the bottom right plot shows that the hysteresis loop gets smoother with
decreasing n so that such a parameter can be related to the smoothness of the hysteresis
loop. In the limit case n → ∞, the constitutive law becomes elastic-perfectly plastic.

7.3 Modeling of Asymmetric Hysteresis Loops

There exist several mechanical systems and materials displaying a nonlinear response
characterized by asymmetric hysteresis loops; typical examples are some materials,
such as metals (Dobson et al. 1997), polymers (Hossain et al. 2012), and shape
memory alloys (Graesser and Cozzarelli 1991), as well as some devices, such as
7 A Review of the Class of Bouc-Wen Differential Models … 133

(a) (b)

Fig. 7.3 Some asymmetric hysteresis loops obtained in experimental tests. a The asymmetric
hysteresis behavior of a wire rope isolator along their axial direction—Demetriades et al. (1993). b
The asymmetric hysteresis behavior of a Nitinol—Dobson et al. (1997)

wire rope isolators deforming along their axial direction (Demetriades et al. 1993)
and dampers (Kwok et al. 2006, 2007). Asymmetric hysteresis loops obtained in
experimental tests and retrieved from the literature are shown in Fig. 7.3.

7.3.1 Asymmetric Bouc-Wen Models

The differential models described in Sect. 7.2 are not able to reproduce rate-
independent asymmetric hysteresis phenomena. Hence, to simulate the typical
asymmetric hysteresis loops, some researchers (Sireteanu et al. 2012; Song and
Der Kiureghian 2006; Wang and Wen 1998) have proposed the following general-
ized expression for the function:

B = A − |z n | Ψ, (7.8)

where Ψ assumes different forms according to the specific improved model that is
adopted. In particular, one has
134 D. Pellecchia and M. Paradiso

Wang and Wen (1998) : Ψ = γ + β sign(ẋ z) + φ sign(ẋ + ż)

(7.9a)
⎧
⎪
⎨Ψ = β1 sign(ẋ z) + β2 sign(x ẋ)
Song and Der Kiureghian (2006) : +β3 sign(x z) + β4 sign(ẋ) (7.9b)
⎪
⎩
+β5 sign(z) + β6 sign(x)
⎧
⎪
⎨Ψ = β1 sign(ẋ z) + β2 sign(x ẋ)
Sireteanu et al. (2012) : −β2 sign(x z) + β4 sign(ẋ) (7.9c)
⎪
⎩
+β5 sign(z)

where β1 , . . . , β6 and φ are material parameters.

Equation (7.9a), introduced by Wang and Wen, includes an additional parameter
φ that takes into account the asymmetric behavior; being independent of the sign of
the generalized displacement x, Eq. (7.9a) cannot describe the asymmetric hysteresis
due to cyclic phenomena since, during them, the sign of the generalized displacement
x changes.
For this reason, Song and Der Kiureghian have developed Eq. (7.9b); basically,
the function Ψ defined by this equation can assume different values in six phases
depending on the signs of x, ẋ, and z and on the values of six fixed parameters,
namely β1 , . . . , β6 . Consequently, this model has six degrees of freedom that affect
the complexity of the parameter identification.
Subsequently, Sireteanu et al. have modified the Song and Der Kiureghian model
by imposing the following conditions:

β3 = −β2 and β6 = 0, (7.10)

that represent a continuity condition of the hysteresis loop at the points of intersec-
tions with the axis of ordinates. Equation (7.9c) describes four different behaviors
depending on the signs of x, ẋ, and z and on the values of only four fixed parameters,
namely β1 , β2 , β4 , and β5 .
Figure 7.4 shows the four different curves of the Ψ function defined by Sireteanu
et al. The continuity condition (7.10) involves that the hysteretic loop is characterized
by the conditions:
Ψ3 = Ψ4 and Ψ6 = Ψ1 . (7.11)

Table 7.1 lists the sign combinations of x, ẋ, and z for the different curves showed
in Fig. 7.4.
7 A Review of the Class of Bouc-Wen Differential Models … 135

z
Ψ1
Ψ2
x
Ψ5
Ψ4

Fig. 7.4 Values of the Ψ function in the model by Sireteanu et al.

Table 7.1 Sign combinations of the Ψ function in the model by Sireteanu et al.
Phase x ẋ z ψ(x, ẋ, z)
1 [−] + + ψ1 = β1 + β4 + β5
2 + − + ψ2 = −β1 − 2 β2 − β4 + β5
3 [−] − − ψ4 = β1 − β4 − β5
4 − + − ψ5 = −β1 − 2 β2 + β4 − β5

7.3.2 Sensitivity Analysis

A parameter sensitivity analysis was carried out to evaluate the effect of each
parameter of the Ψ function on the hysteretic variable z(x) evaluated by adopting
Eqs. (7.5), (7.8), and (7.9c).
Figure 7.5 shows the relationship between the hysteretic variable z and the gener-
alized displacement x. All hysteretic loops have been obtained by applying a gener-
alized displacement described by the sine wave (7.7).
The top left plot shows the influence of the first value of the Ψ function, namely
Ψ1 , on the hysteretic loop. Such a value modifies the shape of the hysteretic loop
in the first and second quadrant when ẋ > 0 and z > 0: the hysteretic variable z is
prone to decrease with an increasing value of Ψ1 ; the hysteretic loop exhibits work
hardening(softening) when the Ψ1 value is negative (positive).
The influence of the Ψ2 value on the hysteretic loop is shown in the top right
plot. The value Ψ2 modifies the trend of the hysteretic loop in the second quadrant
when x > 0, ẋ < 0, and z > 0: the hysteretic variable z is prone to decrease with an
increasing value of Ψ2 ; the hysteretic loop is concave (convex) when the Ψ2 value is
negative (positive).
The Ψ4 value modifies the shape of the hysteretic loop in the third and fourth
quadrants when ẋ < 0 and z < 0; the influence of such a value is shown in the
bottom left plot: the hysteretic variable z is prone to increase with an increasing
value of Ψ4 ; the hysteretic loop shows work hardening (softening) when the Ψ4 value
is negative (positive).
Finally, the bottom right plot shows the influence of the Ψ5 value on the hysteretic
loop. This value modifies the shape of the hysteretic loop in the third quadrant
136 D. Pellecchia and M. Paradiso

4 2

1.5
3
1
2
0.5

1 0

-0.5
0
-1
-1
-1.5

-2 -2
-2 -1 0 1 2 -2 -1 0 1 2

2 2

1 1.5

1
0
0.5
-1
0
-2
-0.5
-3
-1
-4 -1.5

-5 -2
-2 -1 0 1 2 -2 -1 0 1 2

Fig. 7.5 Sensitivity analysis of the model by Sireteanu et al. with respect to the material parameters

when x < 0, ẋ > 0, and z < 0: the hysteretic function z is prone to increase with an
increasing value of Ψ5 ; namely, the hysteretic loop is concave (convex) when the Ψ5
value is positive (negative).

7.4 Modeling of Pinched Hysteresis Loops

The pinching effect is a physical phenomenon, observed in many experimental

results (Kreger and Abrams 1978), in which a very low incremental stiffness near the
origin is followed by a stiffening under grater generalized displacements. In partic-
ular, we can observe the pinching effect in reinforced concrete structures due to the
high shear loads, the slippage of longitudinal reinforcement (Banon and Veneziano
1982), the opening and closing of cracks in the compression zones (Park and Paulay
1975), Y-braced steel frames (Zamani et al. 2012), and confined walls of masonry
7 A Review of the Class of Bouc-Wen Differential Models … 137

(a) (b)

Fig. 7.6 Some hysteresis loops, having the pinching effect, obtained in experimental tests. a The
hysteresis behavior of a steel frame having single bays with symmetrical y-shaped concentric
bracings—Zamani et al. (2012). b The hysteresis behavior of a concrete column section—Park and
Paulay (1975)

structures (Ahmad et al. 2012). Some hysteresis loops with the pinching effect,
obtained in experimental tests, are shown in Fig. 7.6.

7.4.1 Pinching Bouc-Wen Models

To account for the above-described pinching effects, some researchers (Baber and
Noori 1985, 1986; Foliente 1995; Sivaselvan and Reinhorn 2000) have modified the
expression of the function employed by the symmetric differential models, described
in Sect. 7.2, to evaluate the hysteretic variable z. In particular, some modified expres-
sions of such a function are
⎧
⎪ Bh B p
⎪
⎪B =
⎪
⎪ Bh + B p
⎨ 
Baber and Noori (1985): Bh = A − |z n | γ + β sign(ẋ z)
⎪
⎪
⎪
⎪ 1 z2
⎪
⎩Bp = √ Δx exp −
2 π Zσ 2 Z σ2

(7.12a)
⎧  
⎪
⎨ B = A h − |z n | γ + β sign(ẋ z)
Baber and Noori (1986): z2 (7.12b)
⎪
⎩h = 1 − ζ1 exp − 2
2 ζ2
138 D. Pellecchia and M. Paradiso
⎧  
⎪
⎨ B = A h − |z n | γ + β sign(ẋ z)
Foliente (1995):
⎪ (z − z̄ sign(ẋ))2
⎩h = 1 − ζ1 exp −
ζ22
(7.12c)
⎧ Bh B p
⎪
⎪B = B + Bp
⎪
⎪
⎪
⎪
h
z n 
⎪
⎪ = − γ + β sign(ẋ z)
⎪
⎪
Bh A
⎨ zy
−1
Sivaselvan and Reinhorn (2000): 1 (z − z̄ sign(ẋ))2
Bp = √ Δx exp −
⎪
⎪ π 2 Z σ2
⎪
⎪
2 Z σ
⎪
⎪
+
Δx = Rs (xmax −
− xmax )
⎪
⎪
⎪
⎪ Z = σ zy
⎩ σ
z̄ = λ zy
(7.12d)

where Bh , B p , γ , β, etc. are material parameters.

Baber and Noori (1985) take into account the pinching effect in their model
through a pinching spring with stiffness B p in series with the hysteretic element
associated with z, see Eq. (7.12a). The parameter Δx represents the length at which
the variable z of the pinching spring tends to +∞(−∞), namely:

lim z(x) = +∞, lim z(x) = −∞, (7.13)

x→+Δx − x→−Δx +

and is associated with the energy dissipation ε. Z σ is related to the sharpness of

pinching; in particular, a higher Z σ implies a more uniform pinching effect.
In 1986, Baber and Noori proposed a different strategy to account for the pinching
effect (see Eq. (7.12b)); in particular, it amounts to multiply the A parameter by a
pinching function h that depends on the energy dissipation ε, the hysteretic function z,
and two parameters, namely ζ1 and ζ2 : the first one controls the severity of pinching,
whereas the second controls the spread of the pinching region.
Foliente followed the same strategy proposed by Baber and Noori (1986) but
proposed a pinching function h that depends on ε, z, ζ1 , and ζ2 , as well as the
pinching function h in Eq. (7.12b); in addition, the function h also depends on the
parameter z̄ that corresponds to a fraction of z at dz/dx = 0.
Finally, Sivaselvan and Reinhorn proposed a model similar to the one by Baber
and Noori (1985), the main difference lying in the fact that in the former model the
pinching parameter Δx depends on the maximum generalized displacement reached
on the positive and negative sides during the response, weighed through the parameter
of the model Rs , and no longer on the energy dissipation ε. Z σ and z̄ are two variables
that depend on the yield value of the hysteretic variable, i.e. z y through two parameters
of the model, namely σ and λ (see Eq. (7.12d)).
7 A Review of the Class of Bouc-Wen Differential Models … 139

0.6 0.6

0.4 0.4

0.2 0.2

0 0

-0.2 -0.2

-0.4 -0.4

-0.6 -0.6
-4 -2 0 2 4 -4 -2 0 2 4

0.6 2

1.5
0.4
1
0.2
0.5

0 0

-0.5
-0.2
-1
-0.4
-1.5

-0.6 -2
-4 -2 0 2 4 -4 -2 0 2 4

Fig. 7.7 Sensitivity analysis of Sivaselvan and Reinhorn model parameters

7.4.2 Sensitivity Analysis

A sensitivity analysis was carried out to evaluate the effect of each parameter on the
hysteretic variable z(t) evaluated by adopting Eqs. (7.5) and (7.12d). The effects of
the parameters Rs , σ , λ, and z y on the hysteretic function z are shown in Fig. 7.7.
All hysteretic loops have been obtained by applying a generalized displacement
described by the sine wave (7.7).
The top left plot shows the variation of the σ parameter: this parameter controls
the pinching region and increasing σ causes the pinching region to spread. The top
right plot shows the variation of the Rs parameter: the intensity of the pinching effect
is prone to decrease with a decreasing Rs ; specifically, when Rs approaches 0, the
pinching effect is null. The bottom left plot shows the variations of the λ parameter:
the hysteresis loop tends to become more asymmetric when λ increases. Finally, the
bottom right plot shows the variation of the z y parameter: the tangent stiffness at the
origin of the hysteresis loop increases with an increasing value of z y .
140 D. Pellecchia and M. Paradiso

(a) (b)

Fig. 7.8 Hysteresis loops with the pinching effect obtained from experimental tests. a The hysteresis
behaviour of a well confined reinforced concrete column—Colombo and Negro (2005). b The
hysteresis behavior of a unreinforced masonry panels—Liberatore et al. (2019)

7.5 Modeling of Degrading Hysteresis Loops

It is well known from the scientific literature that many mechanical systems can
exhibit strength and stiffness degradation. In general, we can see a hysteretic degrad-
ing behavior when the systems are subjected to cyclic loads like earthquakes, winds,
and so on. For instance, in concrete (Loh et al. 2011; Sengupta and Li 2013, 2014)
and masonry (Liberatore et al. 2019; Tomaževič and Lutman 1996) structures, it is
possible to observe a progressive loss of stiffness due to the opening and closing
of cracks when the applied loads change direction. Wooden structures also exhibit
hysteretic degrading behavior (Xu and Dolan 2009; Zhang et al. 2002). In Fig. 7.8,
hysteresis loops exhibiting strength and stiffness degradation are shown.

7.5.1 Degrading Bouc-Wen Models

Similar to the asymmetric and pinched hysteresis phenomena, several researchers

(Baber and Wen 1981; Baber and Noori 1985; Foliente 1995) have modified the
differential models described in Sect. 7.2 to allow for the simulation of the stiffness
and strength degradation effects. Specifically, they have modified the expression
employed for evaluating the hysteresis function, required to compute the model
output, as follows:
7 A Review of the Class of Bouc-Wen Differential Models … 141
⎧  
⎪
⎪ =
A
− ν |z n | γ + β sign(ẋ z)
⎪
⎪ B
⎪
⎨ η
Baber and Wen (1981): A = A[ε(t)] = A0 − δ A ε(t) (7.14a)
⎪
⎪
⎪
⎪η = η[ε(t)] = η0 + δη ε(t)
⎪
⎩ν = ν[ε(t)] = ν0 + δν ε(t)
⎧  
⎪
⎪
A
= − ν |z n | γ + β sign(ẋ z)
⎪
⎪ B
⎪
⎨ η
Baber and Noori (1985): A = A[ε(t)] = A0 − δ A ε(t) (7.14b)
⎪
⎪
⎪
⎪η = η[ε(t)] = 1 + δη ε(t)
⎪
⎩ν = ν[ε(t)] = 1 + δν ε(t)
⎧ A  
⎪
⎪ = − ν |z n | γ + β sign(ẋ z)
⎨B η
Foliente (1995) : η = η[ε(t)] = 1 + δη ε(t) (7.14c)
⎪
⎪
⎩
ν = ν[ε(t)] = 1 + δν ε(t)

where A, η, ν, etc. are material parameters.

All models consider strength, stiffness, or combined degradation, from the initial
time t = 0 to the present one, as a function of the dissipated energy associated with
the hysteretic displacement z; its expression is given by

ε(t) = z ẋ d t. (7.15)
0

In particular, a convenient measure of degradation is the cumulative hysteretic

dissipated energy ε(t) since degradation depends on the intensity and duration of the
phenomenon under investigation.
Baber and Wen defined two new parameters: η and ν controlling in turn the
stiffness and the strength degradation. Moreover, the amplitude of the hysteresis loop,
controlled by the A parameter, can change. In the Baber and Noori 1985 model, the
η0 and ν0 parameters, which represent the initial values of the degradation functions,
are set to 1. Finally, in the Foliente model, the degradation law related to the A
parameter is null. All models assume that both parameters η and ν depend linearly
on the hysteretic energy ε(t).
It is worth being emphasized that the energy dissipated by the material coincides
with ε(t) in Eq. (7.15) only for peculiar values of the parameters. In general, ε(t)
does not necessarily fulfill the thermodynamic compatibility (Drucker 1957). For
this reason, the use of degrading Bouc-Wen formulations should carefully account
for such an issue.
142 D. Pellecchia and M. Paradiso

1 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1
-2 -1 0 1 2 -2 -1 0 1 2

1 1

0.9
0.8
0.8

0.6 0.7

0.6
0.4 0.5

0.4
0.2
0.3

0 0.2
0 5 10 15 20 0 5 10 15 20

Fig. 7.9 Sensitivity analysis of Foliente model parameters

7.5.2 Sensitivity Analysis

Sensitivity analysis has been carried out on the Foliente model (Foliente 1995) and the
results are shown in Fig. 7.9. The top left and the top right plots show the hysteresis
loops that exhibit a strength and stiffness degradation, respectively. For both top
plots, there are two black colored hysteresis loops without any kind of degradation,
obtained by setting to zero the values of δη and δν . Conversely, when the latter
parameters are greater than zero, a degrading hysteretic behavior is obtained; in
particular, the rate of the strength and stiffness degradation gets stronger when the
δη and δν parameters, respectively, increase. All hysteretic loops have been obtained
by applying a generalized displacement described by the sine wave (7.7).
The two bottom plots show the relationship between the number of hysteresis
loops and the value of the hysteretic function z associated with the maximum positive
7 A Review of the Class of Bouc-Wen Differential Models … 143

displacement x. In these plots, the effect of δη and δν parameters on the hysteresis

behavior is more evident.

7.6 Conclusion

Over the years, the Bouc-Wen model has evolved to describe some physical phe-
nomena observed in experimental tests. In this paper, the evolution of the Bouc-Wen
model in the field of rate-independent mechanical hysteresis phenomena has been
presented. In particular, we have reviewed some models based on the Bouc-Wen
model able to take into account some mechanical phenomena such as asymmetric
hysteresis, the pinching effect, and the strength and stiffness degradation of hystere-
sis. In all sections, we have used the same technical terminology to help the reader to
get acquainted with the physical meaning of the proposed models, to compare each
model, and to shed light on their differences, such as the number of parameters that
the models require as input.
Finally, for all models, some sensitivity analyses have been carried out to evaluate
the effect on the model’s result when the input parameters are modified.

References

Ahmad N, Ali Q, Ashraf M, Khan AN, Alam B (2012) Performance assessment of low-rise confined
masonry structures for earthquake induced ground motions. Int J Civil Struct Eng 2(3):842–859
Alibert JJ, Seppecher P, dell’Isola F (2003) Truss modular beams with deformation energy depending
on higher displacement gradients. Math Mech Solids 8(1):51–73
Altenbach H, Bîrsan M, Eremeyev VA (2012) On a thermodynamic theory of rods with two tem-
perature fields. Acta Mech 223(8):1583–1596
Andreaus U, Spagnuolo M, Lekszycki T, Eugster SR (2018) A Ritz approach for the static analysis
of planar pantographic structures modeled with nonlinear Euler-Bernoulli beams. Contin Mech
Thermodyn 30(5):1103–1123
Ascione L, Berardi V, Feo L, Fraternali F, Tralli AM (2017) Probabilistic assessment of historical
masonry walls retrofitted with through-the-thickness confinement devices. Proc AIMETA
Baber TT, Noori MN (1985) Random vibration of degrading, pinching systems. J Eng Mech
111(8):1010–1026
Baber TT, Noori MN (1986) Modeling general hysteresis behavior and random vibration application.
J Vibr Acoust Stress Rel Des
Baber TT, Wen YK (1981) Random vibration of hysteretic, degrading systems. J Eng Mech Div
107(6):1069–1087
Badoni D, Makris N (1996) Nonlinear response of single piles under lateral inertial and seismic
loads. Soil Dyn Earthq Eng 15(1):29–43
Bahn BY, Hsu CTT (1998) Stress-strain behavior of concrete under cyclic loading. ACI Mater J
95:178–193
Bai XX, Cai FL, Chen P (2019) Resistor-capacitor (RC) operator-based hysteresis model for mag-
netorheological (MR) dampers. Mech Syst Signal Process 117:157–169
Banon H, Veneziano D (1982) Seismic safety of reinforced concrete members and structures. Earthq
Eng Struct Dyn 10(2):179–193
144 D. Pellecchia and M. Paradiso

Barchiesi E, dell’Isola F, Laudato M, Placidi L, Seppecher P (2018) A 1d continuum model for

beams with pantographic microstructure: asymptotic micro-macro identification and numerical
results. In: Advances in mechanics of microstructured media and structures. Springer, pp 43–74
Barchiesi E, Eugster SR, dell’Isola F, Hild F (2020) Large in-plane elastic deformations of bi-
pantographic fabrics: asymptotic homogenization and experimental validation. Math Mech Solids
25(3):739–767
Bouc R (1967) Forced vibrations of mechanical systems with hysteresis. In: Proceedings of the
fourth conference on nonlinear oscillations, Prague
Bouc R (1971) Modèle mathématique d’hystérésis. Acustica 21:16–25
Broccardo M, Alibrandi U, Wang Z, Garrè L (2017) The tail equivalent linearization method for
nonlinear stochastic processes, genesis and developments. In: Risk and reliability analysis: theory
and applications. Springer, pp 109–142
Caggegi C, Sciuto D, Cuomo M (2018) Experimental study on effective bond length of basalt textile
reinforced mortar strengthening system: contributions of digital image correlation. Measurement
129:119–127
Carboni B, Lacarbonara W, Brewick PT, Masri SF (2018) Dynamical response identification of a
class of nonlinear hysteretic systems. J Intell Mater Syst Struct 29(13):2795–2810
Colombo A, Negro P (2005) A damage index of generalised applicability. Eng Struct 27(8):1164–
1174
Contrafatto L, Cuomo M (2002) A new thermodynamically consistent continuum model for hard-
ening plasticity coupled with damage. Int J Solids Struct 39(25):6241–6271
Contrafatto L, Cuomo M, Fazio F (2012) An enriched finite element for crack opening and rebar
slip in reinforced concrete members. Int J Fract 178(1–2):33–50
di Cosmo F, Laudato M, Spagnuolo M (2018) Acoustic metamaterials based on local reso-
nances: homogenization, optimization and applications. In: Generalized models and non-classical
approaches in complex materials 1. Springer, pp 247–274
De Angelo M, Spagnuolo M, D’annibale F, Pfaff A, Hoschke K, Misra A, Dupuy C, Peyre P,
Dirrenberger J, Pawlikowski M (2019) The macroscopic behavior of pantographic sheets depends
mainly on their microstructure: experimental evidence and qualitative analysis of damage in
metallic specimens. Contin Mech Thermodyn 31(4):1181–1203
dell’Isola F, Seppecher P, Alibert JJ, Lekszycki T, Grygoruk R, Pawlikowski M, Steigmann D, Gior-
gio I, Andreaus U, Turco E et al (2019a) Pantographic metamaterials: an example of mathemati-
cally driven design and of its technological challenges. Contin Mech Thermodyn 31(4):851–884
dell’Isola F, Turco E, Misra A, Vangelatos Z, Grigoropoulos C, Melissinaki V, Farsari M (2019b)
Force-displacement relationship in micro-metric pantographs: experiments and numerical simu-
lations. Comptes Rendus Mécanique 347(5):397–405
Demetriades GF, Constantinou MC, Reinhorn AM (1993) Study of wire rope systems for seismic
protection of equipment in buildings. Eng Struct 15(5):321–334
Dimian M, Andrei P (2014) Noise-driven phenomena in hysteretic systems. Springer
Dobson S, Noori M, Hou Z, Dimentberg M, Baber T (1997) Modeling and random vibration analysis
of SDOF systems with asymmetric hysteresis. Int J Non-Linear Mech 32(4):669–680
Drucker DC (1957) A definition of stable inelastic material. BROWN UNIV PROVIDENCE RI,
Tech. rep
Duhem P (1897) Die dauernden aenderungen und die thermodynamik. i. Zeitschrift für Physikalis-
che Chemie 22(1):545–589
Eremeyev V, Morozov N (2010) The effective stiffness of a nanoporous rod. Dokl Phys 55:279–282
Fedele R, Sessa S, Valoroso N (2012) Image correlation-based identification of fracture parameters
for structural adhesives. Technische Mechanik-Europ J Eng Mech 32(2–5):195–204
Foliente GC (1995) Hysteresis modeling of wood joints and structural systems. J Struct Eng
121(6):1013–1022
Fujimura K, Der Kiureghian A (2007) Tail-equivalent linearization method for nonlinear random
vibration. Probab Eng Mech 22(1):63–76
7 A Review of the Class of Bouc-Wen Differential Models … 145

Graesser E, Cozzarelli F (1991) Shape-memory alloys as new materials for aseismic isolation. J
Eng Mech 117(11):2590–2608
Greco F, Luciano R, Serino G, Vaiana N (2018) A mixed explicit-implicit time integration approach
for nonlinear analysis of base-isolated structures. Annals Solid Struct Mech 10(1):17–29
Greco L, Cuomo M (2013) B-Spline interpolation of Kirchhoff-love space rods. Comput Methods
Appl Mech Eng 256:251–269
Hadad HA, Calabrese A, Strano S, Serino G (2017) A base isolation system for developing countries
using discarded tyres filled with elastomeric recycled materials. J Earthq Eng 21(2):246–266
Hodgdon ML (1988) Mathematical theory and calculations of magnetic hysteresis curves. IEEE
Trans Magn 24(6):3120–3122
Hossain MM, Browning R, Minkwitz R, Sue HJ (2012) Effect of asymmetric constitutive behavior
on scratch-induced deformation of polymers. Tribol Lett 47(1):113–122
Jiles D, Atherton D (1983) Ferromagnetic hysteresis. IEEE Trans Magn 19(5):2183–2185
Jiles DC, Atherton DL (1984) Theory of ferromagnetic hysteresis. J Appl Phys 55(6):2115–2120
Kikuchi M, Aiken ID (1997) An analytical hysteresis model for elastomeric seismic isolation
bearings. Earthq Eng Struct Dyn 26(2):215–231
Kreger ME, Abrams DP (1978) Measured hysteresis relationships for small-scale beam-column
joints. Tech. rep., University of Illinois Engineering Experiment Station. College of
Kwok N, Ha Q, Nguyen T, Li J, Samali B (2006) A novel hysteretic model for magnetorheological
fluid dampers and parameter identification using particle swarm optimization. Sens Actuators, A
132(2):441–451
Kwok N, Ha Q, Nguyen M, Li J, Samali B (2007) Bouc-wen model parameter identification for a
MR fluid damper using computationally efficient GA. ISA Trans 46(2):167–179
Liberatore D, Addessi D, Sangirardi M (2019) An enriched Bouc-Wen model with damage. Europ
J Mech-A/Solids 77(103):771
Lima C, Angiolilli M, Barbagallo F, Belletti B, Bergami A, Camata G, Cantagallo C, Di Domenico
M, Fiorentino G, Ghersi A, et al (2018) Nonlinear modeling approaches for existing reinforced
concrete buildings: The case study of de gasperi-battaglia school building in norcia. In: Conference
on Italian concrete days. Springer, pp 82–95
Loh CH, Mao CH, Huang JR, Pan TC (2011) System identification and damage evaluation of
degrading hysteresis of reinforced concrete frames. Earthq Eng Struct Dyn 40(6):623–640
Losanno D, Sierra IEM, Spizzuoco M, Marulanda J, Thomson P (2019a) Experimental assessment
and analytical modeling of novel fiber-reinforced isolators in unbounded configuration. Compos
Struct 212:66–82
Losanno D, Spizzuoco M, Calabrese A (2019b) Bidirectional shaking-table tests of unbonded
recycled-rubber fiber-reinforced bearings (rr-frbs). Struct Control Health Monit 26(9)
Losanno D, Palumbo F, Calabrese A, Barrasso T, Vaiana N (2021) Preliminary investigation of
aging effects on recycled rubber fiber reinforced bearings (RR-FRBs). J Earthq Eng, 1–18
Marmo F, Rosati L (2012a) Analytical integration of elasto-plastic uniaxial constitutive laws over
arbitrary sections. Int J Numer Meth Eng 91(9):990–1022
Marmo F, Rosati L (2012b) An improved flexibility-based nonlinear frame element endowed with
the fiber-free formulation. European congress on computational methods in applied sciences and
engineering (ECCOMAS 2012). Viena, Austria, pp 1–17
Marmo F, Rosati L (2013) The fiber-free approach in the evaluation of the tangent stiffness matrix
for elastoplastic uniaxial constitutive laws. Int J Numer Meth Eng 94(9):868–894
Marmo F, Serpieri R, Rosati L (2011) Ultimate strength analysis of prestressed reinforced concrete
sections under axial force and biaxial bending. Comput Struct 89(1–2):91–108
Mayergoyz ID (2003) Mathematical models of hysteresis and their applications. Academic Press
Menegotto M (1973) Method of analysis for cyclically loaded RC plane frames including changes
in geometry and non-elastic behavior of elements under combined normal force and bending. In:
Proc. of IABSE symposium on resistance and ultimate deformability of structures acted on by
well defined repeated loads, pp 15–22
146 D. Pellecchia and M. Paradiso

Nejadsadeghi N, De Angelo M, Drobnicki R, Lekszycki T, dell’Isola F, Misra A (2019) Parametric

experimentation on pantographic unit cells reveals local extremum configuration. Exp Mech
59(6):927–939
Noël JP, Kerschen G (2017) Nonlinear system identification in structural dynamics: 10 more years
of progress. Mech Syst Signal Process 83:2–35
Nuzzo I, Losanno D, Caterino N, Serino G, Rotondo LMB (2018) Experimental and analytical
characterization of steel shear links for seismic energy dissipation. Eng Struct 172:405–418
Nuzzo I, Losanno D, Caterino N (2019) Seismic design and retrofit of frame structures with hys-
teretic dampers: a simplified displacement-based procedure. Bull Earthq Eng 17(5):2787–2819
Özdemir H (1976) Nonlinear transient dynamic analysis of yielding structures[ph. d. thesis]. PhD
thesis, University of California
Park R, Paulay T (1975) Reinforced concrete structures. John Wiley & Sons
Pellecchia D, Sessa S, Vaiana N, Rosati L (2020) Comparative assessment on the rocking response
of seismically base-isolated rigid blocks. Proc Struct Integrity 29:95–102
Pideri C, Seppecher P (1997) A second gradient material resulting from the homogenization of an
heterogeneous linear elastic medium. Continuum Mech Thermodyn 9(5):241–257
Placidi L, Barchiesi E, Misra A (2018) A strain gradient variational approach to damage: a compar-
ison with damage gradient models and numerical results. Math Mech Complex Syst 6(2):77–100
Placidi L, Misra A, Barchiesi E (2019) Simulation results for damage with evolving microstructure
and growing strain gradient moduli. Continuum Mech Thermodyn 31(4):1143–1163
Ramberg W, Osgood WR (1943) Description of stress-strain curves by three parameters. Nat Adv
Committ Aeron
Sengupta P, Li B (2013) Modified Bouc-wen model for hysteresis behavior of RC beam-column
joints with limited transverse reinforcement. Eng Struct 46:392–406
Sengupta P, Li B (2014) Hysteresis behavior of reinforced concrete walls. J Struct Eng
140(7):04014,030
Serpieri R, Sessa S, Rosati L (2018) A MITC-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220
Sessa S (2010) Multiobjective non-linear random vibration analysis for performance-based earth-
quake engineering. Reliability and optimization of structural systems-proceedings of reliability
and optimization of structural systems pp 185–192
Sessa S, Valoroso N (2017) Kriging interpolation strategy for finite-element-based surrogate
responses of DCB delamination tests. In: Risk and reliability analysis: theory and applications.
Springer, pp 453–461
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through-the-thickness jacketed shells
for the elasto-plastic analysis of confined composite structures: theory and numerical assessment.
Compos B Eng 113:225–242
Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018) Evaluation of the capacity
surfaces of reinforced concrete sections: eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos B Eng 163:413–423
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force-biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54(9):1451–1469
Sessa S, Vaiana N, Paradiso M, Rosati L (2020) An inverse identification strategy for the mechan-
ical parameters of a phenomenological hysteretic constitutive model. Mech Syst Signal Process
139(106):622
Shih MH, Sung WP (2005) A model for hysteretic behavior of rhombic low yield strength steel
added damping and stiffness. Comput Struct 83(12–13):895–908
Sierra IEM, Losanno D, Strano S, Marulanda J, Thomson P (2019) Development and experimental
behavior of HDR seismic isolators for low-rise residential buildings. Eng Struct 183:894–906
7 A Review of the Class of Bouc-Wen Differential Models … 147

Sireteanu T, Giuclea M, Mitu AM, Ghita G (2012) A genetic algorithms method for fitting the
generalized bouc-wen model to experimental asymmetric hysteretic loops. J Vib Acoust 134(4)
Sivaselvan MV, Reinhorn AM (2000) Hysteretic models for deteriorating inelastic structures. J Eng
Mech 126(6):633–640
Song J, Der Kiureghian A (2006) Generalized Bouc-wen model for highly asymmetric hysteresis.
J Eng Mech 132(6):610–618
Song J, Kiureghian AD, Sackman JL (2007) Seismic interaction in electrical substation equipment
connected by non-linear rigid bus conductors. Earthq Eng Struct Dyn 36(2):167–190
Tomaževič M, Lutman M (1996) Seismic behavior of masonry walls: modeling of hysteretic rules.
J Struct Eng 122(9):1048–1054
Turco E, Giorgio I, Misra A, dell’Isola F (2017) King post truss as a motif for internal structure of
(meta) material with controlled elastic properties. Royal Soc Open Sci 4(10):171,153
Turco E, Misra A, Pawlikowski M, dell’Isola F, Hild F (2018) Enhanced Piola-Hencky discrete mod-
els for pantographic sheets with pivots without deformation energy: numerics and experiments.
Int J Solids Struct 147:94–109
Vaiana N, Spizzuoco M, Serino G (2017) Wire rope isolators for seismically base-isolated
lightweight structures: experimental characterization and mathematical modeling. Eng Struct
140:498–514
Vaiana N, Sessa S, Marmo F, Rosati L (2018) A class of uniaxial phenomenological models for sim-
ulating hysteretic phenomena in rate-independent mechanical systems and materials. Nonlinear
Dyn 93(3):1647–1669
Vaiana N, Sessa S, Marmo F, Rosati L (2019a) An accurate and computationally efficient uniaxial
phenomenological model for steel and fiber reinforced elastomeric bearings. Compos Struct
211:196–212
Vaiana N, Sessa S, Marmo F, Rosati L (2019b) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901
Vaiana N, Sessa S, Paradiso M, Marmo F, Rosati L (2019c) An efficient computational strategy
for nonlinear time history analysis of seismically base-isolated structures. In: Conference of the
Italian association of theoretical and applied mechanics. Springer, pp 1340–1353
Vaiana N, Sessa S, Paradiso M, Rosati L (2019d) Accurate and efficient modeling of the hys-
teretic behavior of sliding bearings. In: 7th international conference on computational methods
in structural dynamics and earthquake engineering (COMPDYN 2019). Crete, Greece, pp 24–26
Vaiana N, Marmo F, Sessa S, Rosati L (2020) Modeling of the hysteretic behavior of wire rope
isolators using a novel rate-independent model. Nonlinear Dynamics of Structures. Springer,
Systems and Devices, pp 309–317
Vaiana N, Capuano R, Sessa S, Marmo F, Rosati L (2021a) Nonlinear dynamic analysis of seismi-
cally base-isolated structures by a novel opensees hysteretic material model. Appl Sci 11(3):900
Vaiana N, Losanno D, Ravichandran N (2021b) A novel family of multiple springs models suitable
for biaxial rate-independent hysteretic behavior. Comput Struct 244(106):403
Vaiana N, Sessa S, Rosati L (2021c) A generalized class of uniaxial rate-independent models for sim-
ulating asymmetric mechanical hysteresis phenomena. Mech Syst Signal Process 146(106):984
Valoroso N, Fedele R (2010) Characterization of a cohesive-zone model describing damage and
de-cohesion at bonded interfaces. sensitivity analysis and mode-i parameter identification. Inter-
national Journal of Solids and Structures 47(13):1666–1677
Valoroso N, Sessa S, Lepore M, Cricri G (2013) Identification of mode-i cohesive parameters for
bonded interfaces based on DCB test. Eng Fract Mech 104:56–79
Valoroso N, Marmo F, Sessa S (2014) Limit state analysis of reinforced shear walls. Eng Struct
61:127–139
Valoroso N, Marmo F, Sessa S (2015) A novel shell element for nonlinear pushover analysis of
reinforced concrete shear walls. Bull Earthq Eng 13(8):2367–2388
Visintin A (2013) Differential models of hysteresis, vol 111. Springer Science & Business Media
148 D. Pellecchia and M. Paradiso

Wang CH, Wen Yk (1998) Reliability and redundancy of pre-Northridge low-rise steel buildings
under seismic excitation. na
Wen Y (1980) Equivalent linearization for hysteretic systems under random excitation. ASME J
Appl Mech
Wen YK (1976) Method for random vibration of hysteretic systems. J Eng Mech Div 102(2):249–
263
Xu J, Dolan JD (2009) Development of a wood-frame shear wall model in abaqus. J Struct Eng
135(8):977–984
Zamani SM, Vafai A, Aghakouchak A, Kazemi M (2012) Experimental investigation of steel frames
with single bays of symmetrical y-shaped concentric bracings. Scientia Iranica 19(2):195–210
Zhang H, Foliente GC, Yang Y, Ma F (2002) Parameter identification of inelastic structures under
dynamic loads. Earthq Eng Struct Dyn 31(5):1113–1130
Zuccaro G, Dato F, Cacace F, De Gregorio D, Sessa S (2017) Seismic collapse mechanisms analyses
and masonry structures typologies: a possible correlation. Ingegneria Sismica 34(4):121–150
Chapter 8
A Generalized Formulation of Time
Integration Methods for Nonlinear
Dynamic Analysis of Hysteretic
Mechanical Systems

Davide Pellecchia and Pasquale Cesarano

Abstract The chapter presents a generalized formulation of time integration meth-

ods that allow for the numerical solution of the nonlinear equilibrium equations
characterizing mechanical systems having hysteretic behavior. Two families of time
integration methods are derived from such a generalized formulation: the celebrated
Newmark’s family of methods and Chang’s family of explicit methods. The former is
presented since it represents one of the most employed families of conventional time
integration methods available in the literature. On the contrary, the latter is illustrated
since it is one of the most efficient families of recently developed structure-dependent
time integration methods. For each family, the formulation, the expression for the
evaluation of the unknown generalized displacement, velocity, and acceleration vec-
tors, as well as the main numerical properties are first presented. Then, some instances
as well as the implementation scheme of each family are illustrated. Finally, nonlin-
ear time history analyses are performed on a rate-independent hysteretic mechanical
system, characterized by a stiffening behavior and subjected to an external gener-
alized random force, to illustrate the numerical performance, in terms of accuracy
and computational efficiency, of some methods selected within the above-described
families.

Keywords Time integration method · Hysteresis · Nonlinear equilibrium

equations · Newmark’s time integration methods · Chang’s integration methods

D. Pellecchia
Department of Structures for Engineering and Architecture, University of Naples Federico II,
via Claudio, 21, 80124 Naples, Italy
e-mail: [email protected]
P. Cesarano (B)
CAE Technologies s.r.l. – Simulia, Centro Direzionale isola F12, 80143 Naples, Italy
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 149

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_8
150 D. Pellecchia and P. Cesarano

8.1 Introduction

Hysteretic mechanical systems typically display a complex behavior characterized by

a generalized force that is a function of the generalized displacement (velocity). In the
former (latter) case, the hysteretic behavior is referred to as rate-independent (rate-
dependent) (Dimian and Andrei 2014). There exist several hysteretic mechanical
systems that display both types of hysteretic behavior at the same time (Tsai et al.
2003; Vaiana et al. 2017; Losanno et al. 2021).
Moreover, hysteretic models can be used in order to reproduce the behavior of
more complex systems for which nonlinear analysis represents the sole reliable
strategy to estimate structural responses (Lima et al. 2018; De Angelo et al. 2019;
di Cosmo et al. 2018; Andreaus et al. 2018; Castellano et al. 2019) and their collapse
mechanisms (Zuccaro et al. 2017).
To accurately evaluate the nonlinear response of such systems, appropriate time
integration methods, having specific stability, accuracy, and computational efficiency
properties, need to be employed to numerically solve the nonlinear equilibrium equa-
tions characterizing the discrete structural model of such systems (Wilson 2002).
In the literature, a comprehensive classification of these algorithms is usually per-
formed depending on the kind of coefficients that appear in the expressions employed
for the evaluation of the unknown generalized displacement and velocity vectors at
the generic time of the analysis (Kolay and Ricles 2016).
Specifically, it is possible to distinguish between conventional time integration
methods, such as Newmark’s family of methods (Newmark 1959), and structure-
dependent time integration methods, such as Chang’s families of methods (Chang
2010, 2014). In the former, the coefficients that characterize the expressions of the
generalized displacement and velocity vectors are scalar parameters, whereas, in the
latter, the above-mentioned coefficients can be scalar parameters or matrices that may
depend on the time step Δt and on the initial properties of the mechanical system.
The formulation of such methods is typically developed in the context of linear
or nonlinear elastic systems, thus losing important numerical and implementation
details that need to be taken into consideration when hysteretic mechanical systems
are analyzed. As an example, conventional explicit methods, such as the Newmark
explicit methods (Chang 2009), as well as semi-explicit structure-dependent time
integration methods, such as the ones belonging to the family formulated by Chang
(2010), do not require iterations when elastic systems are analyzed; on the contrary,
they need to be used in conjuction with an iterative procedure when adopted for the
nonlinear dynamic analysis of hysteretic mechanical systems (Vaiana et al. 2019b;
Abd-alla et al. 2017).
In this chapter, we generalize a formulation of families of integration methods to
allow for the solution of the nonlinear equilibrium equations of hysteretic mechanical
systems displaying both rate-dependent and rate-independent hysteretic behaviors.
Specifically, using the proposed generalized formulation, that is introduced after
presenting the nonlinear equilibrium equations of a typical hysteretic mechanical
system, we derive the celebrated Newmark’s family of methods (Newmark 1959) and
8 A Generalized Formulation of Time Integration Methods … 151

the recently developed Chang’s family of explicit method (Chang 2014). The former
is presented because is the most employed family of conventional time integration
methods available in the literature, whereas the latter is illustrated since its stability,
accuracy and computational efficiency properties make it one of the most suitable
families of structure-dependent time integration methods to be employed for the
nonlinear time history analysis of hysteretic mechanical systems.
Subsequently, for each family, we present the formulation, the expressions for
evaluating the unknown generalized displacement, velocity, and acceleration vec-
tors, and the principal numerical properties. In addition, we also illustrate some
instances and an useful implementation scheme to help the reader with the computer
implementation.
Finally, we present the results of some numerical experiments conducted on a hys-
teretic mechanical system subjected to a generalized external random force to show
the accuracy and computational efficiency properties of some methods belonging to
the above-described families.

8.2 Families of Time Integration Methods

In this section, we first formulate the nonlinear equilibrium equations of a hysteretic

mechanical system that display both rate-dependent and rate-independent hysteretic
behaviors (Dimian and Andrei 2014). Subsequently, we illustrate a formulation of
families of time integration methods that we have generalized to allow for the non-
linear time history analysis of such complex mechanical systems.

8.2.1 Nonlinear Equilibrium Equations

An accurate simulation of the dynamic behavior displayed by hysteretic mechanical

systems requires the selection of an appropriate discrete structural model, consisting
of nodes connected by elements.
The response of a discrete structural model, subjected to external forces depending
on time t, is completely described by variables associated with each node, namely,
the nodal generalized displacements, velocities, and accelerations, as well as the
nodal generalized forces.
In the discrete structural model of a typical hysteretic mechanical system, the
nodal generalized forces may be classified into three different categories: elastic
forces, rate-dependent hysteretic forces, and rate-independent hysteretic forces.
The nodal generalized external forces are collected in the generalized external
force vector p(t), whereas the nodal generalized displacements, velocities, accelera-
tions, and the nodal generalized forces are collected, respectively, in the generalized
displacement u(t), velocity u̇(t), acceleration ü(t), and force f(t) vectors.
152 D. Pellecchia and P. Cesarano

The nonlinear equilibrium equations of a discrete structural model may be

obtained by adopting d’Alembert’s principle that allows one to impose the equilib-
rium between the generalized force vector and the generalized external force vector,
as follows:
f(t) = fi (t) + fr d (t) + fri (t) + fe (t) = p(t), (8.1)

where fi (t) is the generalized inertia force, fr d (t) the generalized rate-dependent
hysteretic force, fri (t) the generalized rate-independent hysteretic force, whereas
fe (t) represents the generalized elastic force.
The generalized inertia force fi (t), rate-dependent hysteretic force fr d (t), and
elastic force fe (t) vectors can be written as

fi (t) = Mü(t), (8.2a)

fr d (t) = Cu̇(t) + fr dn (t), (8.2b)
fe (t) = Ku(t) + fen (t), (8.2c)

where M, C, and K are the generalized constant mass, damping, and elastic stiffness
matrices, whereas fr dn (t) and fen (t) are the nonlinear components of the generalized
rate-dependent hysteretic force fr d (t) and elastic force fe (t) vectors, respectively.
By using Eq. (8.2), we can rewrite Eq. (8.1) as follows:

Mü(t) + Cu̇(t) + Ku(t) + fr dn (t) + fri (t) + fen (t) = p(t). (8.3)

The numerical integration of Eq. (8.3), performed by employing a suitable time

integration method, requires the following two initial conditions:

u(0) = d0 , (8.4a)
u̇(0) = v0 , (8.4b)

where d0 and v0 are, respectively, the generalized displacement and velocity vectors
of the discrete structural model defined at the beginning of the analysis.

8.2.2 Generalized Formulation of Time Integration Methods

To introduce the proposed generalized formulation of families of time integration

methods, suitable for the nonlinear time history analyses of mechanical systems with
rate-dependent and/or rate-independent hysteretic behavior, it is useful to express the
system of coupled nonlinear Ordinary Differential Equations (ODEs) of the second
order in time, given by Eq. (8.3), at the generic time t + Δt as follows:

Mü(t + Δt) + Cu̇(t + Δt) + Ku(t + Δt)

(8.5)
+ fr dn (t + Δt) + fri (t + Δt) + fen (t + Δt) = p(t + Δt),
8 A Generalized Formulation of Time Integration Methods … 153

where Δt represents the time step of a nonlinear time history analysis.

The above-mentioned generalized formulation of families of time integration
methods, required to numerically solve Eq. (8.5), is characterized by the follow-
ing set of three equations:

di+1 = di + A1 Δtvi + A2 (Δt)2 ai + A3 (Δt)2 ai+1 + qi+1 , (8.6a)

vi+1 = vi + B1 Δtai + B2 Δtai+1 + ri+1 , (8.6b)
 
ai+1 = M−1 pi+1 − Cvi+1 − Kdi+1 − (f̃r dn )i+1 − (f̃ri )i+1 − (f̃en )i+1 , (8.6c)

where di+1 (di ), vi+1 (vi ), and ai+1 (ai ) are approximate estimates of the gener-
alized displacement u(t + Δt) (u(t)), velocity u̇(t + Δt) (u̇(t)), and acceleration
ü(t + Δt) (ü(t)) vectors at the (i + 1)-th (i-th) time step, respectively; (f̃r dn )i+1 =
fr dn (vi+1 ), (f̃ri )i+1 = fri (di+1 ), (f̃en )i+1 = fen (di+1 ), and pi+1 are approximate esti-
mates of the generalized nonlinear rate-dependent hysteretic force fr dn (t + Δt), rate-
independent hysteretic force fri (t + Δt), nonlinear elastic force fen (t + Δt), and
external force p(t + Δt) vectors at the (i + 1)-th time step, respectively.
The matrices A1 , A2 , A3 , as well as B1 , B2 are coefficient matrices that define a
specific family of time integration methods, whereas qi+1 and ri+1 are load-dependent
vectors, namely, vectors that are functions of the external force vector.
According to the nature of the coefficient matrices and load-dependent vectors,
we can distinguish between (Kolay and Ricles 2016):
• families of conventional time integration methods;
• families of structure-dependent time integration methods.
In families of conventional time integration methods (Newmark 1959; Bathe
1996), all the coefficient matrices become scalar quantities, that is, A1 , A2 , A3 ,
B1 , and B2 , and both the load-dependent vectors become zero vectors, that is,
qi+1 = ri+1 = 0.
On the contrary, in families of structure-dependent time integration methods (Chen
and Ricles 2008; Chang 2010, 2014; Gui et al. 2014), only some of the coefficient
matrices become scalars; moreover, some of them are functions of the time step Δt
and of the initial properties of the discrete structural model, namely, the mass matrix
M, and the initial generalized tangent damping C0 and stiffness K0 matrices.

8.3 Conventional Time Integration Methods

Conventional time integration methods have been classified by Bathe (1996) into
two categories: explicit and implicit methods.
In explicit methods, such as the Newmark explicit method, belonging to New-
mark’s Family of Methods (NFMs) (Newmark 1959), the unknown generalized
displacement vector at time t + Δt can be determined adopting quantities evalu-
154 D. Pellecchia and P. Cesarano

ated at time steps preceding time t + Δt. Such methods, that are conditionally sta-
ble (Dahlquist 1956, 1963), do (do not) require iterations when hysteretic (elastic)
mechanical systems are analyzed.
In implicit methods, such as the Newmark average acceleration method and the
Newmark linear acceleration method, both belonging to Newmark’s family of meth-
ods (Newmark 1959), the expression of the unknown generalized displacement vector
at time t + Δt also depends on unknown quantities evaluated at time t + Δt. Such
methods, that can be conditionally or unconditionally stable (Bathe and Wilson 1972;
HUGHES 1987), always require the use of iterative procedures.
In this section, we present Newmark’s family of methods since it represents one
of the widely used families of conventional time integration methods available in the
literature (Nagarajaiah et al. 1991; Wilson 2002; Greco et al. 2018).

8.3.1 Newmark’s Family of Methods

8.3.1.1 Formulation

The family of methods developed by Newmark (1959) can be derived from Eq. (8.6)
by setting:
A1 = 1, (8.7)

and by substituting the coefficient matrices A2 , A3 , B1 , and B2 with corresponding

scalar quantities, that is, A2 , A3 , B1 , and B2 . Hence, the formulation of such a family
is

di+1 = di + Δtvi + A2 (Δt)2 ai + A3 (Δt)2 ai+1 + qi+1 , (8.8a)

vi+1 = vi + B1 Δtai + B2 Δtai+1 + ri+1 , (8.8b)
 
ai+1 = M−1 pi+1 − Cvi+1 − Kdi+1 − (f̃r dn )i+1 − (f̃ri )i+1 − (f̃en )i+1 , (8.8c)

where the scalar quantities, assumed to be constant during the complete numerical
integration procedure, are computed as follows:

A2 = (0.5 − α) , (8.9)
A3 = α, (8.10)
B1 = (1 − β) , (8.11)
B2 = β, (8.12)

in which the scalar parameters α and β define the numerical properties, namely,
stability and accuracy, of the time integration method selected within such a family.
The load-dependent vectors q and r at the (i + 1)-th time step are evaluated as
8 A Generalized Formulation of Time Integration Methods … 155

qi+1 = 0, (8.13)
ri+1 = 0. (8.14)

It is important to note that an iterative procedure needs to be implemented since the

expressions for evaluating the unknown generalized displacement di+1 and velocity
vi+1 vectors, namely, Eqs. (8.8a) and (8.8b), respectively, are both functions of the
unknown generalized acceleration vector ai+1 .

8.3.1.2 Expressions for di+1 , vi+1 , and ai+1

Adopting the above-described formulation of Newmark’s family of methods, we

derive the expressions for the evaluation of the generalized displacement di+1 , veloc-
ity vi+1 , and acceleration ai+1 vectors at the generic time step i + 1 of the analysis.
To this end, we consider a generic time interval ti ≤ t ≤ ti+1 and we assume that
the generalized external force vector pi (pi+1 ) is assigned at the beginning (end) of
such an interval. Furthermore, we also assume that the generalized displacement di ,
velocity vi , and acceleration ai vectors at time step i are known.
At this point, we first subtract the equilibrium equations at time i to the ones at time
i + 1 thus obtaining the following nonlinear equilibrium equations in incremental
form:

MΔai+1 + CΔvi+1 + KΔdi+1 + (Δf̃n )i+1 = pi+1 − Mai − Cvi − Kdi − (f̃n )i ,
(8.15)
in which the generalized nonlinear force vector (f̃n )i and the incremental generalized
nonlinear force vector (Δf̃n )i+1 are computed, respectively, as

(f̃n )i = (f̃r dn )i + (f̃ri )i + (f̃en )i , (8.16a)

(Δf̃n )i+1 = (Δf̃r dn )i+1 + (Δf̃ri )i+1 + (Δf̃en )i+1 , (8.16b)

whereas the generic incremental vector Δ(•)i+1 is given by

Δ(•)i+1 = (•)i+1 − (•)i . (8.17)

Then, combining Eqs. (8.8a) and (8.8b), we obtain the incremental generalized
velocity vector:
 
β β β
Δvi+1 = Δdi+1 − vi + 1 − Δtai , (8.18)
αΔt α 2α

whereas, from Eq. (8.8a), we derive the incremental generalized acceleration vector:

1 1 1
Δai+1 = Δdi+1 − vi − ai . (8.19)
α (Δt)2 αΔt 2α
156 D. Pellecchia and P. Cesarano

Substituing Eqs. (8.18) and (8.19) in Eq. (8.15), we get the incremental generalized
displacement vector:
Δdi+1 = K̂−1 p̂i+1 , (8.20)

where the matrix K̂, referred to as the effective generalized constant stiffness matrix,
is given by
1 β
K̂ = M+ C + K, (8.21)
α (Δt)2 αΔt

whereas the vector p̂i+1 , denominated effective generalized external force vector, is
evaluated as
   
1 1
p̂i+1 =pi+1 − (Δf̃)i+1 + M vi − 1 − ai +
αΔt 2α
     (8.22)
β β
−C 1− vi + 1 − Δtai − Kdi − (f̃)i .
α 2α

Hence, on account of Eq. (8.17), we can obtain the following expression for the
unknown generalized displacement vector di+1 :

di+1 = di + Δdi+1 , (8.23)

where Δdi+1 is given by Eq. (8.20). Furthermore, using Eq. (8.18), we get the fol-
lowing expression for the unknown generalized velocity vector vi+1 :
   
β β β
vi+1 = Δdi+1 + 1 − vi + 1 − Δtai , (8.24)
αΔt α 2α

whereas, adopting Eq. (8.19), we derive the following expression for the unknown
generalized acceleration vector ai+1 :
 
1 1 1
ai+1 = Δdi+1 − vi + 1 − ai . (8.25)
α (Δt)2 αΔt 2α

It is very important to note that an iterative procedure, such as the pseudo-force

method (Clough and Penzien 2003), is required to compute Δdi+1 from Eq. (8.20).
Indeed, the incremental generalized nonlinear force vector (Δf̃n )i+1 , defined by Eq.
(8.16) and appearing in the right side of Eq. (8.20), is a function of two unknown
quantites, namely, di+1 and vi+1 . Such a vector can be evaluated by employing
suitable phenomenological models available in the literature (Bouc 1971; Seleemah
and Constantinou 1997; Sessa et al. 2020; Vaiana et al. 2018, 2019a, b, d, 2021c,
2020, 2021a, b; Wen 1976, 1980).
A significant and appealing property of Newmark’s family of integration methods
consists in the fact that such algorithms can be used regardless of the properties of
the structural model, as long as it is statically determined. Thus, Newmark’s strategy
8 A Generalized Formulation of Time Integration Methods … 157

has become one of the most popular time step procedure in several engineering
applications and, contrarily to structure-dependent procedures, it has been often used
to analyze nonlinear shell structures subject to both monothonic loads (Valoroso et al.
2014, 2015) and cyclic excitations (Sessa et al. 2017; Serpieri et al. 2018; Sessa et al.
2019a).

8.3.1.3 Numerical Properties

The stability and the accuracy of Newmark’s family of methods depend on the values
assumed by the scalar parameters α and β.
As far as the stability is concerned, the critical time step Δtcr , that is, the maximum
value of the time step that can be adopted to avoid stability problems is different for
each natural mode of the discrete structural model evaluated at the (i + 1)-th time step
of a nonlinear time history analysis. Specifically, assuming a zero viscous damping
ratio, the value of the critical time step of the j-th mode at the (i + 1)-th time step,
( j)
that is, (Δtcr )i+1 , can be evaluated as follows (Bathe 1996):

( j)
( j) T
(Δtcr )i+1 = √ √i+1 , (8.26)
π 2 β − 2α
( j)
where Ti+1 is the natural period of the j-th mode evaluated by using the generalized
tangent stiffness matrix Ki+1 at the (i + 1)-th time step.
As regards the accuracy, the subfamily of methods for which β = 1/2 has a
second-order accuracy, does not suffer from numerical damping, and displays a
( j)
small relative period error for Δt ≤ 0.05 Ti+1 .
Generally speaking, for a fixed value of β, increasing the value of α, the critical
time step of the j-th mode at the (i + 1)-th time step increases whereas the accuracy
decreases.

8.3.2 Some Instances of the NFMs

Some celebrated instances of Newmark’s family of methods that exhibit different

stability and accuracy properties are
• Newmark explicit method, obtained for α = 0 and β = 1/2;
• Newmark linear acceleration method, obtained for α = 1/6 and β = 1/2;
• Newmark average acceleration method, obtained for α = 1/4 and β = 1/2.
Specifically, the Newmark explicit method and the Newmark linear acceleration
method are conditionally stable, whereas the Newmark average acceleration method
is unconditionally stable, as it can be easily shown by computing the critical time
step Δtcr with Eq. (8.26).
158 D. Pellecchia and P. Cesarano

8.3.3 Implementation Scheme of the NFMs

Table 8.1 summarizes the implementation scheme of Newmark’s family of methods.

Such an algorithm that adopts the Pseudo-Force Method (PFM) (Clough and Penzien
2003) as iterative procedure, is composed of two parts. In the first one, called Initial
settings, the generalized constant mass M, damping C, and stiffness K matrices are
first assembled. Then, the initial generalized displacement d0 and velocity v0 vectors
are initialized and the initial generalized acceleration vector a0 is evaluated using the
following equation:

Table 8.1 NFMs algorithm

8 A Generalized Formulation of Time Integration Methods … 159
 
a0 = M−1 p0 − Cv0 − Kd0 − (f̃r dn )0 − (f̃ri )0 − (f̃en )0 (8.27)

that has been formally obtained by setting i = −1 in Eq. (8.8c).

Subsequently, after the selection of the time step Δt and the definition of the
two scalar parameters α and β, the following six integration constants, introduced
to simply the implementation procedure, are evaluated:
 
β β β
a1 = , a2 = 1 − , a3 = 1 − Δt,
αΔt α 2α
(8.28)
1 1 1
a4 = , a5 = , a6 = 1 − .
α (Δt)2 αΔt 2α

Finally, the effective generalized constant stiffness matrix K̂ = a4 M + a1 C + K

is computed and its inverse K, that is, K = K̂−1 , appearing in Eq. (8.20), is evaluated.
It is important to note that K needs to be inverted only once during the nonlinear
time history analysis and has to be saved for the subsequent computations.
In the second part of the algorithm, called Calculations at each time step, the
generalized displacement, velocity, and acceleration vectors are computed, at each
time step of the analysis, by performing iterations.
In particular, at the first iteration of the generic time step i, namely, for j = 1,
( j)
the incremental generalized nonlinear force vector (Δf̃n )i+1 is set equal to a zero
vector. Then, after the evaluation of the effective generalized external force vector
( j) ( j)
p̂i+1 and of the incremental generalized displacement vector Δdi+1 , by using Eqs.
( j) ( j)
(8.22) and (8.20), respectively, the generalized displacement di+1 , velocity vi+1 ,
( j) ( j)
and acceleration ai+1 vectors, as well as the generalized nonlinear force (f̃n )i+1 and
( j+1)
incremental generalized nonlinear force (Δf̃n )i+1 vectors are computed by adopting
Eqs. (8.23), (8.24), (8.25), (8.16a), and (8.16b), respectively.
At this point, if the error , evaluated as the Euclidean norm of the vector
( j+1) ( j)
(Δf̃n )i+1 − (Δf̃n )i+1 , is bigger than a fixed convergence tolerance value toll, typi-
cally assumed equal to 10−8 , a further iteration is needed. In such a case, it is required
to replace j by j + 1 and to go to point (2.3).
On the contrary, if the error ε is smaller than the convergence tolerance value
toll, no further iteration is needed. Thus, after updating di+1 , vi+1 , ai+1 , (f̃n )i+1 , and
replacing i by i + 1, it is possible going to point (2.2).

8.4 Structure-Dependent Time Integration Methods

Structure-dependent time integration methods have been classified by Kolay and

Ricles (2016) into two categories: explicit and semi-explicit methods.
In explicit methods, such as the ones belonging to Chang’s Family of Explicit
Methods (CFEMs) (Chang 2014), the expressions for evaluating the unknown gen-
160 D. Pellecchia and P. Cesarano

eralized displacement and velocity vectors at time t + Δt both require known quan-
tities evaluated at the current time t of the analysis. Thus, such methods, that can be
conditionally or unconditionally stable, do not require iterations.
On the contrary, in semi-explicit methods, such as the ones belonging to Chang’s
Family of Semi-Explicit Methods (CFSEMs) (Chang 2010), the expression for the
evaluation of the unknown generalized velocity vector at time t + Δt also depends on
unknown quantities evaluated at time t + Δt. Therefore, such methods, that can be
conditionally or unconditionally stable, do (do not) require iterations when hysteretic
(elastic) mechanical systems are analyzed.
In this section, we present Chang’s family of explicit methods since it represents
one of the most suitable families of structure-dependent time integration methods for
performing nonlinear time history analyses of hysteretic mechanical systems (Vaiana
et al. 2019b).

8.4.1 Chang’s Family of Explicit Methods

8.4.1.1 Formulation

The family of explicit methods developed by Chang (2014) can be derived from Eq.
(8.6) by setting:
A1 = 1, A3 = 0, and B2 = 0. (8.29)

Hence, the formulation of such a family is

di+1 = di + Δtvi + A2 (Δt)2 ai + qi+1 , (8.30a)

vi+1 = vi + B1 Δtai + ri+1 , (8.30b)
 
−1
ai+1 = M pi+1 − Cvi+1 − Kdi+1 − (f̃r dn )i+1 − (f̃ri )i+1 − (f̃en )i+1 , (8.30c)

where the coefficient matrices A2 and B1 , assumed to be constant during the complete
numerical integration procedure, depend on the time step Δt and on the initial prop-
erties of the analyzed discrete structural model, that is, M, C0 , and K0 . Specifically,
A2 and B1 are computed as follows:
 −1
A2 = M + βΔtC0 + α(Δt)2 K0 M = S−1
0 M, (8.31)
B1 = A2 , (8.32)

in which the scalar parameters α and β define the numerical properties, namely,
stability and accuracy, of the time integration method selected within such a family.
The load-dependent vectors q and r at the (i + 1)-th time step are evaluated as
8 A Generalized Formulation of Time Integration Methods … 161

qi+1 = S−1
0 α(Δt)2 pi+1 − pi , (8.33)
ri+1 = 0. (8.34)

Such vectors allow one to eliminate the unusual amplitude growth that occurs in the
steady-state response of a high frequency mode when structure-dependent methods
are employed in forced vibration problems (Chang 2018a, b).
Considering Eqs. (8.2b) and (8.2c), the initial generalized tangent damping C0
and stiffness K0 matrices of the discrete structural model can be computed as follows:

∂fr d ∂fr dn
C0 = =C+ , (8.35)
∂ u̇ v0 ∂ u̇ v0
∂fri ∂fe ∂fri ∂fen
K0 = + = +K+ . (8.36)
∂u d0 ∂u d0 ∂u d0 ∂u d0

It is important to note that no iterative procedures need to be implemented since

the expressions for evaluating the unknown generalized displacement and velocity
vectors, namely, Eqs. (8.30a) and (8.30b) both require known quantities evaluated at
the current time of the analysis.

8.4.1.2 Expressions for di+1 , vi+1 , and ai+1

Using the above-described formulation of Chang’s family of explicit methods, we

derive the expressions for the evaluation of the generalized displacement di+1 , veloc-
ity vi+1 , and acceleration ai+1 vectors at the generic time step i + 1 of the analysis.
To this end, we consider a generic time interval ti ≤ t ≤ ti+1 and we assume that
the generalized external force vector pi (pi+1 ) is assigned at the beginning (end) of
such an interval. Furthermore, we also assume that the generalized displacement di ,
velocity vi , and acceleration ai vectors at time step i are known.
At this point, the expression for the evaluation of the unknown generalized dis-
placement vector di+1 can be obtained by substituting Eqs. (8.31) and (8.33) into
Eq. (8.30a), so that we get

di+1 = di + Δtvi + S−1
0 M(Δt)2 ai + α(Δt)2 pi+1 − pi . (8.37)

Similarly, the expression for the evaluation of the unknown generalized velocity
vector vi+1 can be derived by substituting Eqs. (8.32) and (8.34) into Eq. (8.30b),
thus obtaining:
vi+1 = vi + S−1
0 MΔtai . (8.38)

Finally, the expression for the evaluation of the unknown generalized acceleration
vector ai+1 is given by Eq. (8.30c):
162 D. Pellecchia and P. Cesarano
 
ai+1 = M−1 pi+1 − Cvi+1 − Kdi+1 − (f̃r dn )i+1 − (f̃ri )i+1 − (f̃en )i+1 . (8.39)

Such an equation can be adopted after evaluating the generalized force vectors
(f̃r dn )i+1 = fr dn (vi+1 ), (f̃ri )i+1 = fri (di+1 ), and (f̃en )i+1 = fen (di+1 ) by employing
suitable phenomenological models available in the literature (Bouc 1971; Wen 1976,
1980; Seleemah and Constantinou 1997; Vaiana et al. 2018, 2019a, b, c).

8.4.1.3 Numerical Properties

The stability and the accuracy of Chang’s family of explicit methods depend on the
values assumed by the scalar parameters α and β.
As far as the stability is concerned, the critical time step Δtcr is different for each
natural mode of the discrete structural model evaluated at the (i + 1)-th time step
of a nonlinear time history analysis. Specifically, assuming a zero viscous damping
ratio and imposing β ≥ 21 , the value of the critical time step of the j-th mode at the
( j)
(i + 1)-th time step, that is, (Δtcr )i+1 , can be evaluated as follows (Chang 2014):
⎧  
⎪
⎪ 1 1 2 ( j)
⎪
⎪ ∞ if α ≥ β+ κi+1 (8.40a)
⎪
⎪ 4 2
⎨  
( j) ( j)
(Δtcr )i+1 = T0 1 1 2 ( j)
⎪
⎪  if α < β+ κi+1 ,
⎪
⎪ 2 ( j) 4 2
⎪
⎪ 2π 1
β + 21 κi+1 − α
⎩ 4
(8.40b)
( j) ( j)
where T0 is the initial natural period of the j-th mode, whereas κi+1 is the current
degree of nonlinearity of the j-th mode at the (i + 1)-th time step:
 ( j)
2
( j) T0
κi+1 = ( j)
, (8.41)
Ti+1

( j)
in which Ti+1 is the natural period of the j-th mode evaluated by using the generalized
tangent stiffness matrix Ki+1 at the (i + 1)-th time step. It is worth being emphasized
( j)
that κi+1 turns out to be greater (smaller) than 1 in presence of (non-) stiffening
behavior.
As regards the accuracy, the subfamily of methods for which β = 1/2 has a
second-order accuracy, does not suffer from numerical damping, and displays a
( j)
small relative period error for Δt ≤ 0.05 Ti+1 .
Generally speaking, for a fixed value of β, increasing the value of α, the critical
time step of the j-th mode at the (i + 1)-th time step increases whereas the accuracy
decreases.
8 A Generalized Formulation of Time Integration Methods … 163

8.4.2 Some Instances of the CFEMs

Some instances of Chang’s family of explicit methods that exhibit good stability and
accuracy properties are
• Chang explicit method 1, obtained for α = 1/4 and β = 1/2;
• Chang explicit method 2, obtained for α = 1/3 and β = 1/2;
• Chang explicit method 3, obtained for α = 1/2 and β = 1/2.
Specifically, the Chang explicit method 1 is (un)conditionally stable for all
(non-)stiffening hysteretic mechanical systems, whereas the Chang explicit method
2 and the Chang explicit method 3 are unconditionally stable also for some stiffening
hysteretic mechanical systems, as it can be easily shown by computing the critical
time step Δtcr with Eq. (8.40).

8.4.3 Implementation Scheme of the CFEMs

Table 8.2 summarizes the implementation scheme of Chang’s family of explicit meth-
ods that is composed of two parts. In the first one, called Initial settings, the gener-
alized constant mass M, damping C, and stiffness K matrices as well as the initial
generalized tangent damping C0 and stiffness K0 matrices, obtained from Eqs. (8.35)
and (8.36), respectively, are first assembled. Then, the initial generalized displace-
ment d0 and velocity v0 vectors are initialized and the initial generalized acceleration
vector a0 is evaluated using the following equation:
 
a0 = M−1 p0 − Cv0 − Kd0 − (f̃r dn )0 − (f̃ri )0 − (f̃en )0 (8.42)

that has been formally obtained by setting i = −1 in Eq. (8.39).

Finally, after the selection of the time step Δt and the definition of the two scalar
parameters α and β, matrix S0 = M + βΔtC0 + α(Δt)2 K0 , appearing in Eqs. (8.37)
and (8.38), is computed and the inverse of matrices M and S0 , that is, M = M−1 and
S0 = S−1
0 , are evaluated. It is important to note that M and S0 need to be inverted
only once during the nonlinear time history analysis and have to be saved for the
subsequent computations.
In the second part of the algorithm, called Calculations at each time step, the
generalized displacement di+1 , velocity vi+1 , and acceleration ai+1 vectors are com-
puted, at each time step of the analysis, by using Eqs. (8.37), (8.38), and (8.39),
respectively.
164 D. Pellecchia and P. Cesarano

8.5 Numerical Experiments

In this section, the nonlinear dynamic response of a mechanical system, having

stiffening rate-independent hysteretic behavior, is simulated by solving the nonlinear
equilibrium equation by means of time integration methods introduced in the previous
sections.
Specifically, in order to numerically investigate the accuracy properties of the
recently developed structure-dependent Chang Explicit Method 3 (CEM3) and its
capability to decrease the computational burden of nonlinear time history analyses,
the numerical results and the computational times are compared with those obtained
by solving the nonlinear equilibrium equation by employing the conventional New-
mark Average Acceleration Method (NAAM).
The Chang Explicit Method 3 is selected since it has the best numerical prop-
erties among methods belonging to the CFEMs, presented in Sect. 8.4, whereas the
Newmark Average Acceleration Method is adopted since it is one of the most used
methods belonging to the NFMs, presented in Sect. 8.3.
The solution algorithms of the NFMs and of the CFEMs, summarized in Tables 8.1
and 8.2, respectively, have been programmed in MATLAB and run on a computer
having an Intel®Core™i7-4700MQ processor and a CPU at 2.40 GHz with 16 GB
of RAM.

Table 8.2 CFEMs algorithm

8 A Generalized Formulation of Time Integration Methods … 165

8.5.1 Mechanical System Properties

The discrete structural model of the analyzed hysteretic mechanical system is made
up of two nodes, namely, Node 1 and Node 2, connected by one straight element,
as shown in Fig. 8.1. Such an element can be decomposed into three subelements,
namely a Linear Elastic Subelement (LESe), with generalized force f el and general-
ized stiffness K , a Linear Rate-Dependent Hysteretic Subelement (LRDSe), having
generalized force fr dl and generalized viscous damping coefficient C, and a Rate-
Independent Hysteretic Subelement (RISe), with generalized force fri and initial
generalized tangent stiffness k.
Assuming that Node 1 is attached to a fixed support and that the element is
flexurally rigid, the only free Degree Of Freedom (DOF) of the above-described
discrete structural model is the axial displacement u of Node 2, the latter having a
lumped mass M applied along such a direction.
Specifically, the initial properties of the structural model are M = 1 Ns2 /m, C = 1
Ns/m, K = 1 N/m, and k = 40√ N/m. Thus, the initial natural period along the axial
direction, evaluated as T = 2π M/ (K + k), is 0.981 s.

Fig. 8.1 Discrete structural model of the analyzed hysteretic mechanical system
166 D. Pellecchia and P. Cesarano

8.5.2 Applied Generalized External Force

The analyses are performed for a generalized external random force p(t), that is, a
Gaussian white noise with an intensity iwn = 9 N and a time duration td = 10 s. Such
a force is applied at Node 2 along the axial direction of the element, as shown in
Fig. 8.1.

8.5.3 Hysteretic Model Parameters

The dynamic behavior of the Rate-Independent Hysteretic Subelement is simulated

adopting a novel rate-independent model; such an uniaxial model, representing a
specific instance of the general class formulated by Vaiana et al. (2018), requires
a set of only four parameters, namely, ka , kb , γ1 , and γ2 , and two internal model
parameters, namely, u 0 and f 0 , expressed as a function of ka , kb , and γ1 . Specifically,
ka > kb , ka > 0, γ1 > 0, u 0 > 0, f 0 > 0, whereas γ2 is real.
In particular, the generalized rate-independent hysteretic force fri is evaluated by
solving the following transcendental equation, when us < u j s:

ka − kb  −γ1 (us−u j s+2u 0 ) 

fri = −2γ2 u + eγ2 u − e−γ2 u − s e − e−2γ1 u 0 + kb u + f 0 s,
γ1
(8.43)
whereas, if us > u j s, it is computed as follows:

fri = −2γ2 u + eγ2 u − e−γ2 u + kb u + f 0 s, (8.44)

in which u is the generalized displacement, s is the sign of the generalized velocity,

namely, s = sign(v), and u j is the model history variable.
Specifically, the parameters adopted to perform the nonlinear analyses are ka = 40
Nm−1 , kb = 1 Nm−1 , γ1 = 30 m−1 , and γ2 = 10 Nm−1 .

8.5.4 Results of the Nonlinear Time History Analyses

The selection of an appropriate time step Δt is a crucial aspect to accurately analyze

the above-described discrete structural model. As suggested in the literature (Clough
and Penzien 2003; Bathe 1996), time history analyses of linear elastic systems can be
performed adopting Δt = T /10 or T /20. Since for nonlinear systems a smaller time
step is required to limit the detrimental effects due to factors that affect the accuracy
of the adopted time integration method, such as the period distortion, the analyses
are performed adopting a time step Δt = 0.001 s, that is smaller than T /100.
8 A Generalized Formulation of Time Integration Methods … 167

Table 8.3 NLTHAs results—random force    

d [m] v ms−1 a ms−2
tct [s] tct p [%] max min max min max min
NAAM 1.221 – 0.1839 −0.1474 0.6458 −0.6933 34.7285 −29.3396
CEM3 0.600 49.14 0.1839 −0.1476 0.6526 −0.6950 34.7456 −29.3541

The results of the Nonlinear Time History Analyses (NLTHAs) carried out on the
selected hysteretic mechanical system are shown in Table 8.3. The accuracy of the
CEM3 is very satisfactory since the maximum and minimum values of the approx-
imate estimates of the axial displacement d, velocity v, and acceleration a of the
analyzed model are numerically quite close to those predicted by using the conven-
tional NAAM.
The response of the analyzed model, simulated by applying the random force
along the axial direction, is illustrated in terms of time histories of the axial displace-
ment, velocity, and acceleration, shown, respectively, in Figs. 8.2a–c, and in terms of
axial rate-independent force-displacement hysteresis loops, illustrated in Fig. 8.2d.
Generally speaking, the comparison between the responses obtained with the CEM3
and the NAAM shows a very good agreement.
Finally, as regards the computational efficiency, Table 8.3 also shows the total
computational time tct required by the two methods as well as the total computational
time percentage tct p required by the CEM3 with respect to the NAAM, evaluated as

CEM3 tct
CEM3 tct p [%] = · 100. (8.45)
NAAM tct
It transpires that the computational burden of the CEM3 is smaller than the one
characterizing the NAAM since the former does not require iterations for each time
step of the performed nonlinear time history analyses.

8.6 Conclusion

We have presented a generalized formulation of time integration methods that take

into account the presence of generalized force vectors, namely the nonlinear rate-
dependent and rate-independent hysteretic force vectors, typical of discrete structural
models characterizing hysteretic mechanical systems.
Starting from the proposed formulation, we have derived Newmark’s family of
methods and Chang’s family of explicit method, both suitable to perform nonlinear
dynamic analysis of hysteretic mechanical systems.
To show the accuracy and the computational efficiency of two methods belonging
to the above-described families, namely, the Newmark Average Acceleration Method
(NAAM) and the Chang Explicit Method 3 (CEM3), presented in Sects. 8.3.2 and
168 D. Pellecchia and P. Cesarano

0.2 1
NAAM NAAM
CEM3 CEM3
0.1 0.5
displacement [m]

velocity [ms-1]
0 0

-0.1 -0.5

-0.2 -1
0 2 4 6 8 10 0 2 4 6 8 10
time [s] time [s]
(a) (b)
40 4
NAAM NAAM
CEM3 CEM3
20 2
acceleration [ms-2]

force [N]

0 0

-20 -2

-40 -4
0 2 4 6 8 10 -0.2 -0.1 0 0.1 0.2
time [s] displacement [m]
(c) (d)

Fig. 8.2 Discrete structural model response in terms of axial displacement (a), velocity (b), accel-
eration (c), and axial force-displacement hysteresis loops (d)

8.4.2, respectively, some nonlinear time history analyses have been performed on
a stiffening hysteretic mechanical system by modeling the rate-independent hys-
teretic force employing a novel uniaxial model developed by Vaiana et al. (2018).
Specifically, the numerical results and the computational times obtained with the
recently developed CEM3 have been compared to those obtained by employing the
conventional NAAM.
The dynamic responses of the analyzed hysteretic mechanical system, obtained
for a generalized external random force, reveal that the accuracy of the CEM3 is very
satisfactory since the numerical results closely match those predicted by the NAAM.
Furthermore, these results also reveal that the computational burden required by the
CEM3 is reduced with respect to that characterizing the NAAM; in particular, the
total computational time percentage, tct p, of CEM3 is less than 50.00%.
8 A Generalized Formulation of Time Integration Methods … 169

Future enhancements of the presented generalized formulation consist in com-

paring the computed responses with different strategies of the structural engineering
such as seismic envelopes (Sessa et al. 2015, 2018b) and limit-analysis approaches
(Sessa et al. 2018a, 2019b). Moreover, Chang’s family of integration algorithms is
particularly feasible for extensions to soil-structure interaction analyses in which the
behavior of the foundations is modeled by means of half-spaces (Marmo et al. 2016,
2018; Franciosi et al. 2019) because of its capability of accounting for the structural
behavior. Finally, the proposed formulation will be adopted for the nonlinear analy-
sis of metamaterials (De Angelo et al. 2020; dell’Isola et al. 2019, b; Nejadsadeghi
et al. 2019b, a; Turco et al. 2017, 2018, 2019; Turco and Barchiesi 2019; Turco et al.
2020).

References

Abd-alla AenN, Alshaikh F, Del Vescovo D, Spagnuolo M (2017) Plane waves and eigenfrequency
study in a transversely isotropic magneto-thermoelastic medium under the effect of a constant
angular velocity. J Thermal Stress 40(9):1079–1092
Andreaus U, Spagnuolo M, Lekszycki T, Eugster SR (2018) A Rtz approach for the static analysis
of planar pantographic structures modeled with nonlinear Euler-Bernoulli beams. Continuum
Mech Thermodyn 30(5):1103–1123
Bathe K, Wilson E (1972) Stability and accuracy analysis of direct integration methods. Earthq Eng
Struct Dyn 1(3):283–291
Bathe KJ (1996) Finite element procedures. New Yersey
Bouc R (1971) Modèle mathématique d’hystérésis. Acustica 21:16–25
Castellano A, Fraddosio A, Scacco J, Milani G, Piccioni MD (2019) Dynamic response of FRCM
reinforced masonry arches. Key Eng Mater Trans Tech Publ 817:285–292
Chang SY (2009) Accurate integration of nonlinear systems using newmark explicit method. J
Mech 25(3):289–297
Chang SY (2010) A new family of explicit methods for linear structural dynamics. Comput Struct
88(11–12):755–772
Chang SY (2014) Family of structure-dependent explicit methods for structural dynamics. J Eng
Mech 140(6):06014,005
Chang SY (2018a) Elimination of overshoot in forced vibration responses for chang explicit family
methods. J Eng Mech 144(2):04017,177
Chang SY (2018b) An unusual amplitude growth property and its remedy for structure-dependent
integration methods. Comput Methods Appl Mech Eng 330:498–521
Chen C, Ricles JM (2008) Development of direct integration algorithms for structural dynamics
using discrete control theory. J Eng Mech 134(8):676–683
Clough RW, Penzien J (2003) Dynamics of structures. Computers and Structures, berkeley. CA
di Cosmo F, Laudato M, Spagnuolo M (2018) Acoustic metamaterials based on local reso-
nances: homogenization, optimization and applications. In: Generalized models and non-classical
approaches in complex materials 1. Springer, pp 247–274
Dahlquist G (1956) Convergence and stability in the numerical integration of ordinary differential
equations. Mathematica Scandinavica pp 33–53
Dahlquist GG (1963) A special stability problem for linear multistep methods. BIT Numer Math
3(1):27–43
De Angelo M, Spagnuolo M, D’annibale F, Pfaff A, Hoschke K, Misra A, Dupuy C, Peyre P,
Dirrenberger J, Pawlikowski M, (2019) The macroscopic behavior of pantographic sheets depends
170 D. Pellecchia and P. Cesarano

mainly on their microstructure: experimental evidence and qualitative analysis of damage in

metallic specimens. Contin Mech Thermodyn 31(4):1181–1203
De Angelo M, Placidi L, Nejadsadeghi N, Misra A (2020) Non-standard Timoshenko beam model
for chiral metamaterial: identification of stiffness parameters. Mechanics Research Communica-
tions 103
dell’Isola F, Seppecher P, Alibert JJ, Lekszycki T, Grygoruk R, Pawlikowski M, Steigmann D, Gior-
gio I, Andreaus U, Turco E et al (2019) Pantographic metamaterials: an example of mathematically
driven design and of its technological challenges. Contin Mech Thermodyn 31(4):851–884
dell’Isola F, Seppecher P, Spagnuolo M, Barchiesi E, Hild F, Lekszycki T, Giorgio I, Placidi L,
Andreaus U, Cuomo M et al (2019b) Advances in pantographic structures: design, manufacturing,
models, experiments and image analyses. Contin Mech Thermodyn 31(4):1231–1282
Dimian M, Andrei P (2014) Noise-driven phenomena in hysteretic systems. Springer
Franciosi P, Spagnuolo M, Salman OU (2019) Mean green operators of deformable fiber networks
embedded in a compliant matrix and property estimates. Contin Mech Thermodyn 31(1):101–132
Greco F, Luciano R, Serino G, Vaiana N (2018) A mixed explicit-implicit time integration approach
for nonlinear analysis of base-isolated structures. Annal Solid Struct Mech 10(1):17–29
Gui Y, Wang JT, Jin F, Chen C, Zhou MX (2014) Development of a family of explicit algorithms
for structural dynamics with unconditional stability. Nonlinear Dyn 77(4):1157–1170
Hughes T (1987) The finite element method. Linear static and dynamic finite element analysis
Kolay C, Ricles JM (2016) Assessment of explicit and semi-explicit classes of model-based algo-
rithms for direct integration in structural dynamics. Int J Numer Meth Eng 107(1):49–73
Lima C, Angiolilli M, Barbagallo F, Belletti B, Bergami A, Camata G, Cantagallo C, Di Domenico
M, Fiorentino G, Ghersi A, et al (2018) Nonlinear modeling approaches for existing reinforced
concrete buildings: The case study of de gasperi-battaglia school building in norcia. In: Conference
on Italian concrete days. Springer, pp 82–95
Losanno D, Palumbo F, Calabrese A, Barrasso T, Vaiana N (2021) Preliminary investigation of
aging effects on recycled rubber fiber reinforced bearings (RR-FRBs). J Earthq Eng 1–18
Marmo F, Sessa S, Rosati L (2016) Analytical solution of the cerruti problem under linearly dis-
tributed horizontal loads over polygonal domains. J Elast 124(1):27–56
Marmo F, Sessa S, Vaiana N, De Gregorio D, Rosati L (2018) Complete solutions of three-
dimensional problems in transversely isotropic media. Contin Mech Thermodyn 32(3):775–802
Nagarajaiah S, Reinhorn AM, Constantinou MC (1991) Nonlinear dynamic analysis of 3-d-base-
isolated structures. J Struct Eng 117(7):2035–2054
Nejadsadeghi N, De Angelo M, Drobnicki R, Lekszycki T, Dell’Isola F, Misra A (2019a) Parametric
experimentation on pantographic unit cells reveals local extremum configuration. Exp Mech
59(6):927–939
Nejadsadeghi N, Placidi L, Romeo M, Misra A (2019b) Frequency band gaps in dielectric granular
metamaterials modulated by electric field. Mech Res Commun 95:96–103
Newmark NM (1959) A method of computation for structural dynamics. J Eng Mech Div 85(3):67–
94
Seleemah A, Constantinou MC (1997) Investigation of seismic response of buildings with linear and
nonlinear fluid viscous dampers. National Center for Earthquake Engineering Research Buffalo
Serpieri R, Sessa S, Rosati L (2018) A MITC-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220
Sessa S, Marmo F, Rosati L (2015) Effective use of seismic response envelopes for reinforced
concrete structures. Earthq Eng Struct Dyn 44(14):2401–2423
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through-the-thickness jacketed shells
for the elasto-plastic analysis of confined composite structures: theory and numerical assessment.
Compos B Eng 113:225–242
Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018a) Evaluation of the capacity
surfaces of reinforced concrete sections: Eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512
8 A Generalized Formulation of Time Integration Methods … 171

Sessa S, Marmo F, Vaiana N, Rosati L (2018b) A computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthq Eng 1–34
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos B Eng 163:413–423
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force-biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54(9):1451–1469
Sessa S, Vaiana N, Paradiso M, Rosati L (2020) An inverse identification strategy for the mechan-
ical parameters of a phenomenological hysteretic constitutive model. Mech Syst Signal Process
139(106):622
Tsai C, Chiang TC, Chen BJ, Lin SB (2003) An advanced analytical model for high damping rubber
bearings. Earthq Eng Struct Dyn 32(9):1373–1387
Turco E, Barchiesi E (2019) Equilibrium paths of Hencky pantographic beams in a three-point
bending problem. Math Mech Complex Syst 7(4):287–310
Turco E, Giorgio I, Misra A, Dell’Isola F (2017) King post truss as a motif for internal structure of
(meta) material with controlled elastic properties. Royal Soc Open Scie 4(10):171,153
Turco E, Misra A, Pawlikowski M, Dell’Isola F, Hild F (2018) Enhanced Piola-Hencky discrete
models for pantographic sheets with pivots without deformation energy: numerics and experi-
ments. Int J Solids Struct 147:94–109
Turco E, Misra A, Sarikaya R, Lekszycki T (2019) Quantitative analysis of deformation mechanisms
in pantographic substructures: experiments and modeling. Contin Mech Thermodyn 31(1):209–
223
Turco E, Barchiesi E, Giorgio I, Dell’Isola F (2020) A Lagrangian Hencky-type non-linear model
suitable for metamaterials design of shearable and extensible slender deformable bodies alterna-
tive to Timoshenko theory. Int J Non-Linear Mech 123(103):481
Vaiana N, Spizzuoco M, Serino G (2017) Wire rope isolators for seismically base-isolated
lightweight structures: experimental characterization and mathematical modeling. Eng Struct
140:498–514
Vaiana N, Sessa S, Marmo F, Rosati L (2018) A class of uniaxial phenomenological models for sim-
ulating hysteretic phenomena in rate-independent mechanical systems and materials. Nonlinear
Dyn 93(3):1647–1669
Vaiana N, Sessa S, Marmo F, Rosati L (2019a) An accurate and computationally efficient uniaxial
phenomenological model for steel and fiber reinforced elastomeric bearings. Compos Struct
211:196–212
Vaiana N, Sessa S, Marmo F, Rosati L (2019b) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901
Vaiana N, Sessa S, Paradiso M, Marmo F, Rosati L (2019c) An efficient computational strategy
for nonlinear time history analysis of seismically base-isolated structures. In: Conference of the
Italian Association of theoretical and applied mechanics. Springer, pp 1340–1353
Vaiana N, Sessa S, Paradiso M, Rosati L (2019d) Accurate and efficient modeling of the hysteretic
behavior of sliding bearings. 7th International conference on computational methods in structural
dynamics and earthquake engineering (COMPDYN 2019). Crete, Greece, pp 24–26
Vaiana N, Marmo F, Sessa S, Rosati L (2020) Modeling of the hysteretic behavior of wire rope
isolators using a novel rate-independent model. Nonlinear Dynamics of Structures. Springer,
Systems and Devices, pp 309–317
Vaiana N, Capuano R, Sessa S, Marmo F, Rosati L (2021a) Nonlinear dynamic analysis of seismi-
cally base-isolated structures by a novel opensees hysteretic material model. Appl Sci 11(3):900
Vaiana N, Losanno D, Ravichandran N (2021b) A novel family of multiple springs models suitable
for biaxial rate-independent hysteretic behavior. Comput Struct 244(106):403
Vaiana N, Sessa S, Rosati L (2021c) A generalized class of uniaxial rate-independent models for sim-
ulating asymmetric mechanical hysteresis phenomena. Mech Syst Signal Process 146(106):984
Valoroso N, Marmo F, Sessa S (2014) Limit state analysis of reinforced shear walls. Eng Struct
61:127–139
172 D. Pellecchia and P. Cesarano

Valoroso N, Marmo F, Sessa S (2015) A novel shell element for nonlinear pushover analysis of
reinforced concrete shear walls. Bull Earthq Eng 13(8):2367–2388
Wen YK (1976) Method for random vibration of hysteretic systems. J Eng Mech Div 102(2):249–
263
Wen Y (1980) Equivalent linearization for hysteretic systems under random excitation. J Appl
Mech-T
Wilson EL (2002) Three-dimensional static and dynamic analysis of structures. Comput Struct
Zuccaro G, Dato F, Cacace F, De Gregorio D, Sessa S (2017) Seismic collapse mechanisms analyses
and masonry structures typologies: a possible correlation. Ingegneria Sismica 34(4):121–150
Chapter 9
Quasi-Harmonic Solutions
for Transversely Isotropic
Magneto-Electro-Thermo-Elasticity:
A Symbolic Mathematics Approach

Francesco Marmo and Massimo Paradiso

Abstract The behaviour of magneto-electro-thermo-elastic materials is governed

by a set of six differential equations in which the piezoelectric, the piezomagnetic,
the magnetoelectric and thermal effects are coupled. Under the steady-state con-
dition, this system of equations is homogeneous and can be rewritten as a set of
uncoupled modified Laplace equations expressed as a function of the roots of a
characteristic polynomial associated with the original set of governing equations.
Differently from previous proposals, the presented approach employs the entire ker-
nel of the adjoint differential operator so as to preserve completeness. Finally, due to
the large number of constitutive parameters involved in the uncoupling process, two
Mathematica scripts that compute the coefficients of the characteristic polynomial
and the components of the adjoint differential operator are described in full detail.

Keywords Piezoelectricity · Piezomagnetism · Electromagnetism ·

Thermoelasticity · Laplace equation · Adjoint differential operator

9.1 Introduction

The distinctive property of magneto-electro-elastic composites is the coupling

between strain, electric polarization and magnetization. These materials offer a wide
range of applications in nanoscience and nanotechnology and can be successfully
employed for the fabrication of electronic devices such as sensors, actuators, trans-

F. Marmo (B) · M. Paradiso

Department of Structures for Engineering and Architecture, University
of Naples Federic II, via Claudio 21, 80124 Napoli, Italy
e-mail: [email protected]
M. Paradiso
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 173

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_9
174 F. Marmo and M. Paradiso

ducers and memories (Eerenstein et al. 2006; Milton et al. 2009; Wang et al. 2010;
Nejadsadeghi et al. 2019; di Cosmo et al. 2018).
The equations that govern the behaviour of this class of materials, often char-
acterized by transverse isotropy, describe coupling between the infinitesimal strain,
electric and magnetic fields, eventually accounting for the effects of temperature.
These formulas constitute a system of coupled differential equations (Bardzokas
et al. 2007) whose solution is of particular interest in practical applications.
Since the beginning of the XX century, many researchers have approached the
problem of rewriting this system of differential equation as a set of uncoupled mod-
ified Laplace equations. Accordingly, the solution can be expressed as the sum of
quasi-harmonic functions ϕi that fulfil the condition

(∂x2 + ∂ y2 + si2 ∂z2 ) ϕi = 0, (9.1)

where ∂(·) represents differentiation with respect to (·) and si are scalars evaluated as
a function of the material constitutive parameters. Their values are computed as the
roots of a characteristic equation associated with the governing system of equations.
Although these potential functions need to be evaluated for specific boundary
conditions, their determination is simpler than the direct application of the original
differential equation. Actually, many solution techniques, both analytical and numer-
ical, are available for Laplace equations, while a specific solving technique needs to
be employed for the more general governing set of coupled equations.
With reference to transversely isotropic elasticity, the first use of quasi-harmonic
potential functions dates back to the paper by Michell (1900), later followed by
the alternative solutions (Lekhniskii 1940; Elliott and Mott 1948; Shield 1951; Chen
1966; Green and Zerna 1968; Pan and Chou 1976, 1979; Lekhnitskii 1981; Fabrikant
1989; Lin et al. 1991; Ding et al. 2006; Rosati and Marmo 2014; Marmo and Rosati
2016; Marmo et al. 2016a, 2018).
These solutions employ a variable number of scalar quasi-harmonic potential
functions that range between two and eight. Questioning about the problem of com-
pleteness of the potential solutions proposed in the literature, Wang and Wang (1995)
furnished two separate expressions that employ only three quasi-harmonic functions.
However, in a recent contribution, the authors (2017) have shown a fundamental
property of the quasi-harmonic potential, which have a vector nature and can be used
to express the solution as a linear combination of arbitrary derivatives of all their
components. This property can be used to specialize the solutions that employ three
quasi-harmonic potentials to all other approaches formulated in the earlier literature,
hence showing their substantial equivalence.
More recently, the authors (2016b, 2020) have proposed an alternative approach
that exploits the entire kernel of the differential operator associated with the governing
equations, what again furnishes the same set of three quasi-harmonic potentials.
Actually, the minimum number of potential functions to handle in the solution process
is equal to number of unknown scalar fields. In case of transversely isotropic elasticity,
these amount to three components of displacement.
9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 175

Generalizations to the case of piezoelectricity have been proposed in Zikung and

Bailin (1995), Haojiang et al. (1996) and employ four quasi-harmonic potentials due
to the presence of the additional unknown scalar field represented by the electric
potential. A discussion about the completeness of these solutions is reported in Xu et
al. (2008). Solutions that account for thermal coupling, either in absence or presence
of the piezoelectric effect, are described in Podil’chuk and Sokolovskii (1994), Chen
(2000), Ashida et al. (1993).
The case of magneto-electro-elasticity, in which piezoelectric, piezomagnetic and
magnetoelectric couplings are present, is considered in Pan (2002), Wang and Shen
(2002), Ding and Jiang (2003). These solutions consider five quasi-harmonic poten-
tials, the additional one being generated by the presence of a new unknown scalar
field, i.e. the magnetic potential.
Thermal coupling is considered in Chen et al. (2004) and, as expected, includes
six quasi-harmonic potentials. However this solution is obtained by considering only
one row of the adjoint differential operator associated with the system of governing
equations. Hence, similarly to what was done by the authors in the case of transversely
isotropic linear elasticity (Marmo et al. 2020), we propose below a solution that
employs the entire kernel of the differential operator associated with the governing
equations in presence of thermal, electric, and magnetic couplings.
Accordingly, after introducing the field equations that govern the behaviour of
magneto-electro-thermo-elastic materials (Sect. 9.2), we uncouple the equations by
inverting the associated differential operator (Sect. 9.3). Due to the large number of
constitutive parameters involved in this operation, we furnish a Mathematica script
that computes the coefficients of the characteristic polynomial associated with the
magneto-electro-thermo-elastic coupling and the components of the adjoint differ-
ential operator employed to express the final solution (Sect. 9.4)

9.2 Field Equations

For a magneto-electro-thermo-elastic body, the infinitesimal stress tensor T, the elec-

tric displacement (or electric charge density displacement or electric flux density) d
and the magnetic flux b are expressed as a function of the infinitesimal strain tensor
S, the electric field e, the magnetic field h and the temperature increment T , by the
following constitutive equations (Bardzokas et al. 2007):

T =  : S − t13 e − t13 h − B T
d =  : S + De + Ah + pT (9.2)
b =  : S + Ae + Mh + lT

where , ,  and B are the elastic, piezoelectric, piezomagnetic, thermoelastic

tensors, D, A, and p are the dielectric, magnetoelectric, and pyroelectric tensors, M
and l are the magnetic permeability and pyromagnetic tensors. Although this property
176 F. Marmo and M. Paradiso

will not be used below, we recall that the thermoelastic tensor B can be expressed as
a function of the tensor of linear thermal expansion  as B =  : .
The infinitesimal strain tensor S and the electric and magnetic fields, e and h, are
evaluated as a function of the displacement field u and of the electric and magnetic
potentials, φ and ψ, as

S = sym( grad u) e = − grad φ h = − grad ψ (9.3)

Finally, the piezo-electro-magneto-thermic fields are governed by the following

steady-state equilibrium equations:

divT = 0 divd = 0 divb = 0 div(K grad T ) = K : H(T ) = 0 (9.4)

where K is the thermal conductivity tensor and H(·) is the Hessian of (·).
We assume that the material is transversely isotropic, with the plane of isotropy
parallel to the plane x − y of the reference system and we indicate by h either one
of the two direction of isotropy, i.e. x or y, and by v the direction orthogonal to the
plane of isotropy, i.e. z. Accordingly, the constitutive tensors are
⎡ ⎤
Mh 0 0 0 G hh 0 0 0 G hv
⎢ 0 λhh 0 G hh 0 0 0 0 0 ⎥
⎢ ⎥
⎢ 0 0 λhv 0 0 0 G hv 0 0 ⎥
⎢ ⎥
⎢ 0 G hh 0 λhh 0 0 0 0 0 ⎥
⎢ ⎥
=⎢
⎢ G hh 0 0 0 Mh 0 0 0 G hv ⎥
⎥ (9.5)
⎢ 0 0 0 0 0 λhv 0 G hv 0 ⎥
⎢ ⎥
⎢ 0 0 G hv 0 0 0 λhv 0 0 ⎥
⎢ ⎥
⎣ 0 0 0 0 0 G hv 0 λhv 0 ⎦
G hv 0 0 0 G hv 0 0 0 Mv
⎡ ⎤ ⎡ ⎤
0 0 Nhv 0 0 L hv
⎢ 0 0 0 ⎥ ⎢ 0 0 0 ⎥
⎢ ⎥ ⎢ ⎥
⎢ Nhv 0 0 ⎥ ⎢ L hv 0 0 ⎥
⎢ ⎥ ⎢ ⎥
⎢ 0 0 0 ⎥ ⎢ 0 0 0 ⎥
⎢ ⎥ ⎢ ⎥
=⎢
⎢ 0 0 Nhv ⎥
⎥ =⎢
⎢ 0 0 L hv ⎥
⎥ (9.6)
⎢ 0 Nhv 0 ⎥ ⎢ 0 L hv 0 ⎥
⎢ ⎥ ⎢ ⎥
⎢ E hv 0 0 ⎥ ⎢ Q hv 0 0 ⎥
⎢ ⎥ ⎢ ⎥
⎣ 0 E hv 0 ⎦ ⎣ 0 Q hv 0 ⎦
0 0 Ev 0 0 Qv
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
βh 0 0 Dh 0 0 αh 0 0 μh 0 0
B = ⎣ 0 βh 0 ⎦ D = ⎣ 0 D h 0 ⎦ A = ⎣ 0 αh 0 ⎦ M = ⎣ 0 μh 0 ⎦
0 0 βv 0 0 Dv 0 0 αv 0 0 μv
(9.7)
9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 177
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
kh 0 0 0 0
K = ⎣ 0 kh 0 ⎦ p=⎣ 0 ⎦ l=⎣0⎦ (9.8)
0 0 kv pv lv

where:
Mh − λhh
G hh = (9.9)
2
Accounting for the symmetry of the constitutive tensors, formulas (9.2) and (9.3)
are used into (9.4) to obtain the following homogeneous system of differential equa-
tions: ⎧
⎪
⎪
3
 · (u) + t13 : H(φ) + t13 : H(ψ) − B gradT = 0
⎪
⎪
⎨ 3
 · (u) − D : H(φ) − A : H(ψ) + p · gradT = 0 (9.10)
⎪
⎪  3· (u) − A : H(φ) − M : H(h) + l · gradT = 0
⎪
⎪
⎩
K : H(T ) = 0

3
where : and · respectively denote saturation of the two and three inner indices in
the product between tensors. In this respect, notice that (u) is a third order tensor
representing the Hessian of u, and is defined as
⎡ ⎤
H(u x )
(u) = ⎣ H(u y ) ⎦ (9.11)
H(u z )

Introducing the vector f = {u x , u y , u z , φ, ψ, T }t , the previous system of equa-

tions can be rewritten in compact form as

Lf = 0 (9.12)

where L is the following differential operator:

⎧ ⎫
⎪
⎪LC LE LQ LB ⎪⎪
⎨ ⎬
LtE LD L A LP
L= (9.13)
⎪
⎪ LQ
t
L A LM LL ⎪
⎪
⎩ ⎭
(0 0 0) 0 0 L K

with
⎧ ⎫
⎨ Mh ∂x2 + G hh ∂ y2 + G hv ∂z2 (λhh + G hh )∂x ∂ y (λhv + G hv )∂x ∂z ⎬
LC = (λhh + G hh )∂x ∂ y G hh ∂x2 + Mh ∂ y2 + G hv ∂z2 (λhv + G hv )∂ y ∂z
⎩ ⎭
(λhv + G hv )∂x ∂z (λhv + G hv )∂ y ∂z G hv ∂h2 + Mv ∂z2
(9.14)
178 F. Marmo and M. Paradiso
⎧ ⎫ ⎧ ⎫ ⎧ ⎫
⎨ (Nhv + E hv )∂x ∂z ⎬ ⎨ (L hv + Q hv )∂x ∂z ⎬ ⎨ −βh ∂x ⎬
L E = (Nhv + E hv )∂ y ∂z L Q = (L hv + Q hv )∂ y ∂z L B = −βh ∂ y
⎩ ⎭ ⎩ ⎭ ⎩ ⎭
Nhv ∂h2 + E v ∂z2 L hv ∂h2 + Q v ∂z2 −βv ∂z
(9.15)
L D = −Dh ∂h2 − Dv ∂z2 L A = −αh ∂h2 − αv ∂z2 L M = −μh ∂h2 − μv ∂z2 (9.16)

and

L P = pv ∂ z L L = l v ∂z L K = kh ∂h2 + kv ∂z2 ∂h2 = ∂x2 + ∂ y2 (9.17)

Remark 9.1 The elastic parameters in , i.e. see Formula (9.5), can be expressed
as a function of the Young moduli, Yh and Yv , and Poisson’s ratios, νhh and νhv , in
the plane of isotropy and along the perpendicular direction, respectively:

Yh (Yh νhv
2
− Yv ) Yv2 (νhh − 1)
Mh = M v =
(1 + νhh )[Yv (νhh − 1) + 2Yh νhv
2
] Yv (νhh − 1) + 2Yh νhv
2

Yh (Yv νhh + Yh νhv )

2
Yh Yv νhv
λhh =− λhv =
(1 + νhh )[Yv (νhh − 1) + 2Yh νhv
2
] Yv − Yv νhh − 2Yh νhv
2

Remark 9.2 The constitutive equations (9.2) can be rewritten in Voigt notation to
express the stress and strain tensors as six-component vectors, i.e.

T → t = [σx , σ y , σz , τ yz , τzx , τx y ]t S → s = [εx , ε y , εz , γ yz , γzx , γx y ]t

Accordingly, Eq. (9.2) become

t = C s − E t e − Qt h − β T d = E s + D e + A h + p T b = Q s + a e + M h + l T

where
⎡ ⎤ ⎡ ⎤
Mh λhh λhv 0 0 0 0 0 0 0 Nhv 0
⎤ ⎡
⎢ ⎥ ⎢ ⎥ βh
⎢ λhh Mh λhv 0 0 ⎥
0 E=⎣ 0 0 0 Nhv 0 0⎦
⎢ ⎥ ⎢ βh ⎥
⎢ ⎥ ⎢ ⎥
⎢ λhv λhv Mv 0 0 0 ⎥ E hv E hv E v 0 0 0 ⎢ βv ⎥
C=⎢
⎢
⎥
⎥ ⎡ ⎤ β = ⎢ ⎥
⎢ 0 ⎥
⎢ 0 0 0 G hv 0 0 ⎥ 0 0 0 0 L hv 0 ⎢ ⎥
⎢ ⎥ ⎣ 0 ⎦
⎢ 0 0 0 G hv 0 ⎥ ⎢ ⎥
⎣ 0 ⎦ Q=⎣ 0 0 0 L hv 0 0 ⎦ 0
0 0 0 0 0 G hh Q hv Q hv Q v 0 0 0


9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 179

9.3 A General Solution to the Field Equations in Terms

of Quasi-Harmonic Potentials

In this section the differential equation (9.12) is expressed in a simpler form that
involves the introduction of quasi-harmonic potentials. To this end, the differential
operator L is inverted by employing its adjoint operator L∗ so as to uncouple the
equations in (9.12). The new set of uncoupled differential equations is expressed by
a scalar differential operator |L|, representing the determinant of L.
Factorization of the differential operators is used to show that the solution can
be expressed as the sum of six is quasi-harmonic functions and that the Almansi
theorem can be used to further simplify the differential problem.
These steps are described in detail in the following subsections.

9.3.1 Inversion of the Differential Operator L

The determinant |L| of L can be evaluated by employing its Laplace expansion as a

function of the entries of the kth row. Indicating by L∗ the adjoint matrix of L, we
have
|L| = Lk̄ j L∗j k̄ (9.18)

where index notation has been adopted. Notice that, although the index k is repeated
in the previous formula, no summation is executed over k since it is a fixed index;
to remind this we adopted a bar above the index letter. Accordingly Lk̄ j and L∗j k̄ are
the jth components of the vectors Lk̄(·) and L∗(·) k̄ , i.e. the kth row and column of L
and L∗ , respectively. Similarly, it is easy to verify that

Llj¯ L∗j k̄ = 0 for l = k (9.19)

since this is the determinant of a matrix in which the lth row repeats twice. Combining
Formulas (9.18) and (9.19) for all values of k and l, one obtains the well known
property of the adjoint matrix
L L∗ = I |L| (9.20)

where I is the identity matrix.

Let us now indicate by ϕk a scalar function that fulfils the condition

|L| ϕk = 0 (9.21)

This function is used to express a solution to Eq. (9.12) as

f (k̄) = L∗(·)k̄ ϕk̄ (9.22)

180 F. Marmo and M. Paradiso

Actually, substituting (9.22) into (9.12), we have

L f (k̄) = LL∗(·)k̄ ϕk̄ = 0 (9.23)

in which (9.18), (9.19) and (9.21) have been used.

Clearly, solution (9.22) is not unique, since a different solution can be generated by
changing the value of k. Furthermore, it is easy to verify that any linear combination
of the solutions obtained by choosing all values of k is also a solution to Eq. (9.12).
Actually, we can set
 
f= c(k̄) f (k̄) = L∗(·) k̄ c(k̄) ϕk̄ = L∗ ϕ (9.24)
k̄ k̄

where c(k̄) are numerical coefficients and ϕ is the vector collecting the functions
c(k̄) ϕk̄ . It fulfils the condition
|L|ϕ = 0 (9.25)

since all components of ϕ satisfy Eq. (9.21). Hence, substituting (9.24) into (9.12)
we have
L f = LL∗ ϕ = I |L| ϕ = 0 (9.26)

in which the property (9.20) and (9.25) have been adopted.

Notice that, having expressed the solution of (9.12) by means of (9.24), the prob-
lem of computing f has been transformed into the problem of evaluating the vector
function ϕ by solving (9.25). Although the number of unknowns is unchanged, the
differential form (9.25) is advantageous since it is expressed by a scalar differential
operator.
The evaluation of the adjoint matrix L∗ and of the determinant |L| is a key aspect of
the presented solution strategy. A Mathematica script for their evaluation is reported
in Sect. 9.4.

9.3.2 Factorization of the Differential Equation |L| ϕ = 0

The differential operator |L| is computed as the determinant of L, yielding

 
|L| = L K LG p1 ∂h8 + p2 ∂h6 ∂z2 + p3 ∂h4 ∂z4 + p4 ∂h2 ∂z6 + p5 ∂z8 (9.27)

where L K is defined in (9.17)3 , LG = G hh ∂h2 + G hv ∂z2 , while p1 , …, p5 are scalar

parameters computed as a function of the constitutive parameters (9.5)–(9.9), see,
e.g. Sect. 9.4.1.
Equation (9.27) is substituted into (9.24) and the resulting equation is factorized
to obtain the equivalent equation
9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 181


6
|L|ϕ = 0 ⇔ D1 D 2 D 3 D 4 D 5 D 6 ϕ = Di ϕ = 0 (9.28)
i=1

in which the entire set of equations has been expressed by means of the six differential
operators
Di = ∂h2 + ∂z2i = ∂h2 + si2 ∂z2 i = 1, . . . , 6 (9.29)

where z i = z/si and the parameters si , i = 1, . . . , 6 are the roots of the characteristic
polynomial equation
  
G hv kv  
s −
2
s −
2
p1 s 8 − p2 s 6 + p3 s 4 − p4 s 2 + p5 = 0 (9.30)
G hh kh

The procedure for obtaining (9.28)2 is detailed below. However, written in this
form, the set of differential equations can be solved by employing the generalized
Almansi’s theorem (Almansi 1899; Wang and Xu 1990). Accordingly, the vector
function ϕ can be expressed by means of up to six vector functions i as


nr 
mi
ϕ= zmi − j ϕi (9.31)
i=1 j=1

where n r is the number of distinct roots of (9.30) and m i is the multiplicity of the ith
root. The potential functions i are quasi-harmonic since they are required to fulfil
the conditions
Di ϕ i = 0 i = 1, . . . , n r (9.32)

In order to obtain the factorization of Eq. (9.24), one first divides it by kv and G hv ,
respectively. This operation gives the first two operators D1 and D2 as equivalent
to L K and LG when the roots of the first two factors of (9.30) are employed. The
remaining operators Di , i = 2, . . . , 5, are obtained in a similar way by factorizing
the eighth degree differential operator in (9.27) as

p1 ∂h8 + p2 ∂h6 ∂z2 + p3 ∂h4 ∂z4 + p4 ∂h2 ∂z6 + p5 ∂z8

    
= p1 ∂h2 + sa2 ∂z ∂h2 + sb2 ∂z ∂h2 + sc2 ∂z ∂h2 + sd2 ∂z (9.33)

in which sa , sb , sc and sd are the roots of the third factor in equation (9.30).
Formula (9.33) can be verified by expanding the product on the right-hand side,
obtaining:
 
p1 ∂h8 + p1 sa2 + sb2 + sc2 + sd2 ∂h6 ∂z2 + p1 (sa2 sb2 + sa2 sc2 + sb2 sc2 + sa2 sd2 + sb2 sd2
 
+ sc2 sd2 )∂h4 ∂z4 + p1 sa2 sb2 sc2 + sa2 sb2 sd2 + sa2 sc2 sd2 + sb2 sc2 sd2 ∂h2 ∂z6 + p1 sa2 sb2 sc2 sd2 ∂z8
(9.34)
182 F. Marmo and M. Paradiso

Comparing the previous expression with the left-hand side of equation (9.33), we
obtain the following system of equations:
⎧  2 
⎪
⎪ p1 sa + sb2 + sc2 + sd2 − p2 = 0
⎪
⎨ p1 s 2 s 2 + s 2 s 2 + s 2 s 2 + s 2 s 2 + s 2 s 2 + s 2 s 2  − p3 = 0
⎪
a b a c b c a d b d c d
 2 2 2  (9.35)
⎪
⎪ p1 sa sb sc + sa sb sd + sa sc sd + sb sc sd − p4 = 0
2 2 2 2 2 2 2 2 2
⎪
⎪
⎩
p1 sa2 sb2 sc2 sd2 − p5 = 0

which needs to be solved to obtain sa , sb , sc and sd . Hereafter we are going to solve

(9.35) so as to derive a condition that sa has to fulfil in order to be used in the
factorization (9.33). The same procedure can be followed for sb , sc and sd .
With the objective of isolating all terms containing sa , we can rewrite (9.35)3 as
 
p1 sb2 sc2 sd2 = − p1 sa2 sb2 sc2 + sa2 sb2 sd2 + sa2 sc2 sd2 + p4 (9.36)

which can be substituted in (9.35)4 to obtain

   
− p1 sa2 sb2 sc2 + sa2 sb2 sd2 + sa2 sc2 sd2 + p4 sa2 − p5 = 0 (9.37)

that is  
− p1 sb2 sc2 + sb2 sd2 + sc2 sd2 sa4 + p4 sa2 − p5 = 0 (9.38)

Similarly, isolating all terms containing sa in (9.35)2 , we have

   
p1 sb2 sc2 + sb2 sd2 + sc2 sd2 = − p1 sa2 sb2 + sa2 sc2 + sa2 sd2 + p3 (9.39)

Substituting the previous formula into (9.37)2 we obtain

   
− − p1 sa2 sb2 + sa2 sc2 + sa2 sd2 + p3 sa4 + p4 sa2 − p5 = 0 (9.40)

which gives  
p1 sb2 + sc2 + sd2 sa6 − p3 sa4 + p4 sa2 − p5 = 0 (9.41)

Finally, from (9.35)1 we have

 
p1 sb2 + sc2 + sd2 = − p1 sa2 + p2 (9.42)

which is used in (9.40)2 to get

 
− p1 sa2 + p2 sa6 − p3 sa4 + p4 sa2 − p5 = 0 (9.43)

that is
p1 sa8 − p2 sa6 + p3 sa4 − p4 sa2 + p5 = 0 (9.44)
9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 183

The previous equation represents the condition that sa has to fulfil in order to
be used in (9.33); it is equivalent to (9.30). Similarly one can verify that the same
condition has to be fulfilled by sb , sc , and sd . In conclusion, all the parameters s(·)
used in the polynomial factorization (9.33) are the roots of the polynomial (9.30).

9.4 Automatic Evaluation of |L| and L∗ and Relevant

Coefficients

Although the evaluation of |L| and L∗ is not conceptually difficult, a series of manip-
ulations are required due to the great amount of constitutive parameters which are
involved. Furthermore, the resulting expressions are rather long and difficult to be
transcribed in a computer code for numerical applications; for their generation we
furnish below two Mathematica scripts so that the interested reader can convert the
desired expressions into C and Fortran codes by employing the functions CForm
and FortranForm available in Mathematica.

9.4.1 Evaluation of |L|

For the evluation of |L|, the differential operator L is first input

L={{Mh* x ^2+Ghh * y ^2+Ghv * z ^ 2 , ( l h h +Ghh ) * x * y , ( l h v +Ghv ) * x * z ,
( Nhv+Ehv ) * x * z , ( Lhv+Qhv ) * x * z , −bh * x } ,
{ ( l h h +Ghh ) * x * y , Ghh * x ^2+Mh* y ^2+Ghv * z ^ 2 , ( l h v +Ghv ) * y * z ,
( Nhv+Ehv ) * y * z , ( Lhv+Qhv ) * y * z , −bh * y } ,
{ ( l h v +Ghv ) * x * z , ( l h v +Ghv ) * y * z , Ghv * h ^2+Mv* z ^ 2 ,
Nhv * h ^2+ Ev * z ^ 2 , Lhv * h ^2+Qv * z ^ 2 , −bv * z } ,
{ ( Nhv+Ehv ) * x * z , ( Nhv+Ehv ) * y * z , Nhv * h ^2+ Ev * z ^ 2 , −Dh * h^2−Dv * z ^ 2 ,
−ah * h^2− av * z ^ 2 , pv * z } ,
{ ( Lhv+Qhv ) * x * z , ( Lhv+Qhv ) * y * z , Lhv * h ^2+Qv * z ^ 2 , −ah * h^2− av * z ^ 2 ,
−mh * h^2−mv * z ^ 2 , l v * z } ,
{ 0 , 0 , 0 , 0 , 0 , kh * h ^2+ kv * z ^ 2 } } ;
l h h =Mh−2* Ghh ;
P r i n t [ ' ' L = ' ' , L / / MatrixForm ]

where the differential operators ∂x , ∂ y , ∂z and ∂h2 are treated as algebraic quantities
denoted by x, y, z, and h^2, respectively. Also, within the previous code, Formula
(9.9) has been used to reduce the number of constitutive parameters in L.
The determinant of L is then computed and Formula (9.17)4 is used to simplify
its expression
DetL= Det [ L ] ;
DetL= S i m p l i f y [ C o l l e c t [ DetL , { h ^ 2 , z ^ 2 , x ^2+ y ^ 2 } , S i m p l i f y ] / / .
( x ^2+ y ^2)−> h ^ 2 ] ;
P r i n t [ ' ' | L | = ' ' , DetL ]
184 F. Marmo and M. Paradiso

Such a code generates an expression of |L| which is difficult to be handled.

However, displaying the list of factors of |L| one can verify that the last factor is a
long combination of constitutive parameters and differential operators, i.e. ∂h2 and
∂z2 . This is done by the code
F a c t O f D e t L = S i m p l i f y [ F a c t o r L i s t [ DetL ] ] ;

P r i n t [ ' ' F a c t o r s o f | L | = ' ' , F a c t O f D e t L / / TableForm ]

In order to simplify the expression of |L|, the coefficients of ∂h2 and ∂z2 are isolated
and substituted by a series of factors pi , expressed as a combination of constitutive
coefficients. This task is accomplished by the function FactorDet defined by the
following code:
F a c t o r D e t [ d e t _ , np_ , p i _ , pa_ ] : = (
FactOfDet=Simplify [ FactorList [ det ] ] ;
NFactOfDet=Dimensions [ FactOfDet ] [ [ 1 ] ] ;
MyFactOfDet = F a c t O f D e t [ [ 1 , 1 ] ] ^ F a c t O f D e t [ [ 1 , 2 ] ] *
F a c t O f D e t [ [ NFactOfDet , 1 ] ] ^ F a c t O f D e t [ [ NFactOfDet , 2 ] ] ;
Ai= C o e f f i c i e n t L i s t [ MyFactOfDet , { h ^ 2 , z ^ 2 } ] ;
NewMyFactOfDet = 0 ;
n p s =np + 1 ;
n p i =np ;
p i i =pi ;
Do [ { I f [ Not [ TrueQ [ S i m p l i f y [ Ai [ [ i , j ] ] ] == 0 ] ] ,
{ n p i = n p i +1 ,
p i i [ [ n p i ] ] = S i m p l i f y [ Ai [ [ i , j ] ] ] ,
i q =npi ,
Do [ I f [ S i m p l i f y [ p i i [ [ n p i ]] − p i i [ [ n q i ] ] ] = = 0 ,
{ i q =nqi ,
p i i [ [ npi ]]=0 ,
n p i = n p i −1 ,
Break [ ]
}
],
{ n q i , n p i − 1}
],
NewMyFactOfDet=NewMyFactOfDet+
pa [ [ i q ] ] * h ^ ( 2 * ( i −1)) * z ^ ( 2 * ( j −1))
}
]
},
{ j , D i m e n s i o n s [ Ai ] [ [ 2 ] ] } ,
{ i , D i m e n s i o n s [ Ai ] [ [ 1 ] ] }
];
NewDet=NewMyFactOfDet ;
Do [ NewDet=NewDet * F a c t O f D e t [ [ i , 1 ] ] ^ F a c t O f D e t [ [ i , 2 ] ] ,
{ i , 2 , NFactOfDet − 1 } ] ;
P r i n t [ NewDet ] ;
P r i n t [ ' ' where : ' ' ] ;
I f [ nps <= n p i , Do [ P r i n t [ ' ' p [ ' ' , i , ' ' ] = ' ' , p i i [ [ i ] ] ] , { i , nps , n p i } ] ] ;
Return [ { { NewDet } , { p i i } , { n p i } } ] ;
);
9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 185

The function FactorDet outputs an array containing the input variable det, that
is L, and the coefficients pi used therein.
The function is conceived in such a way that the list of coefficients is updated,
without repetitions, each time the function is called. Thus, at the first call of the
function, the number of coefficient, the array of coefficients and the array of their
names needs to be initialized.
np = 0 ;
p i =Array [ 0 & , 1 0 0 0 ] ;
pa =Array [ p , 1 0 0 0 ] ;

Hence the simplified expression of |L| can be obtained by typing:

Print [ ' ' Determinant of L : ' ' ] ;
FDetL= F a c t o r D e t [ DetL , np , p i , pa ] ;
DetL=FDetL [ [ 1 , 1 ] ] ;
p i =FDetL [ [ 2 , 1 ] ] ;
np=FDetL [ [ 3 , 1 ] ] ;

Finally, the expression (9.27) of |L| is output, together with the following coeffi-
cients
p1 = Mh [Dh (L 2hv + G hv μh ) − αh2 G hv − 2αh L hv Nhv + μh Nhv
2
]
p2 = −2Dh λhv L hv − 2Dh G hv λhv μh − Dh λhv μh + E hv (L hv + G hv μh )
2 2 2 2

+ Dv L 2hv Mh + Dv G hv μh Mh + Dh G hv Mh μv + Dh μh Mh Mv + αh2 (2G hv λhv

+ λ2hv − Mh Mv ) − 2αv L hv Mh Nhv + 2E v μh Mh Nhv − 2λhv μh Nhv
2

+ Mh μv Nhv
2
− 2Dh λhv L hv Q hv + Dh G hv Q 2hv + Nhv
2
Q 2hv − 2E hv Nhv (λhv μh
+ L hv Q hv ) + 2Dh L hv Mh Q v + 2αh (E hv λhv L hv − αv G hv Mh − E v L hv Mh
+ 2λhv L hv Nhv − E hv G hv Q hv + +λhv Nhv Q hv − Mh Nhv Q v )
p3 = −2E hv E v L 2hv − 2Dv λhv L 2hv − 2E hv E v G hv μh − 2E hv E v λhv μh
− 2Dv G hv λhv μh − Dv λ2hv μh − αv2 G hv Mh + E v2 μh Mh + E hv
2
G hv μv
− 2Dh G hv λhv μv − Dh λ2hv μv + Dv G hv Mh μv − αh2 G hv Mv + Dh L 2hv Mv
+ E hv
2
μh Mv + Dh G hv μh Mv + Dv μh Mh Mv + Dh Mh μv Mv − 2E v λhv μh Nhv
− 2E hv λhv μv Nhv + 2E v Mh μv Nhv + 2E hv μh Mv Nhv − 2λhv μv Nhv
2

+ μh Mv Nhv
2
− 2E hv E v L hv Q hv − 2Dv λhv L hv Q hv + 2Dh L hv Mv Q hv
+ 2E v L hv Nhv Q hv + Dv G hv Q 2hv + Dh Mv Q 2hv + 2E v Nhv Q 2hv + 2E hv
2
L hv Q v
− 2Dh λhv L hv Q v + 2Dv L hv Mh Q v + 2E hv L hv Nhv Q v − 2Dh G hv Q hv Q v
− 2Dh λhv Q hv Q v − 2E hv Nhv Q hv Q v − 2Nhv
2
Q hv Q v + Dh Mh Q 2v
+ 2αv (E hv λhv L hv − E v L hv Mh + 2λhv L hv Nhv − E hv G hv Q hv + λhv Nhv Q hv
− Mh Nhv Q v ) + 2αh {αv (2G hv λhv + λ2hv − Mh Mv ) − E hv L hv Mv
− L hv Mv Nhv − E hv Mv Q hv − Mv Nhv Q hv + E hv G hv Q v + E hv λhv Q v
+ λhv Nhv Q v + E v [G hv Q hv + λhv (L hv + Q hv ) − Mh Q v ]}
186 F. Marmo and M. Paradiso

p4 = −2Dv G hv λhv μv − Dv λ2hv μv + Dv L 2hv Mv

+ Dv G hv μh Mv + E hv
2
μv Mv + Dh G hv μv Mv + Dv Mh μv Mv + αv2 (2G hv λhv
+ λ2hv − Mh Mv ) + 2E hv μv Mv Nhv + μv Mv Nhv
2
+ 2Dv L hv Mv Q hv
+ Dv Mv Q 2hv − 2Dv λhv L hv Q v − 2Dv G hv Q hv Q v − 2Dv λhv Q hv Q v
+ E v2 (L 2hv + G hv μh + Mh μv + 2L hv Q hv + Q 2hv )
+ E hv
2
Q 2v + Dh G hv Q 2v + Dv Mh Q 2v + 2E hv Nhv Q 2v + Nhv
2
Q 2v
+ 2αv {E hv G hv Q v − αh G hv Mv − E hv L hv Mv − L hv Mv Nhv − E hv Mv Q hv
− Mv Nhv Q hv + E hv λhv Q v + λhv Nhv Q v
+ E v [G hv Q hv + λhv (L hv + Q hv )
− Mh Q v ]} − 2E v {λhv μv Nhv + αh G hv Q v
+ L hv Nhv Q v + Nhv Q hv Q v + E hv [G hv μv + λhv μv
+ (L hv + Q hv )Q v ]}
p5 = G hv [E v2 μv − αv2 Mv − 2αv E v Q v + Dv (μv Mv + Q 2v )]

9.4.2 Evaluation of L∗

For the evaluation of the L∗ , the Combinatorica package needs to be loaded

Needs [ ' ' C o m b i n a t o r i c a ' ' ]

so that the cofactor matrix of L can be computed by the Cofactor function and
simplified by adopting Formula (9.17)4 .
CofL= Table [ S i m p l i f y [ C o f a c t o r [ L , { i , j } ] ] , { i , 6 } , { j , 6 } ] ;
Do [ CofL [ [ i , j ] ] = S i m p l i f y [ C o l l e c t [ CofL [ [ i , j ] ] , { h ^ 2 , z ^ 2 , x ^2+ y ^ 2 } ,
S i m p l i f y ] / / . ( x ^2+ y ^2)−> h ^ 2 ] ,
{ i , 1 , D i m e n s i o n s [ CofL ] [ [ 1 ] ] } ,
{ j , 1 , D i m e n s i o n s [ CofL ] [ [ 2 ] ] }
];
AdjL= T r a n s p o s e [ CofL ] ;
P r i n t [ ' ' L * = ' ' , AdjL / / MatrixForm ]

Likewise |L|, each term of L∗ appears as the product of a set of factors whose
last term is characterized by having a very long expression. In order to simplify the
final expression of L∗ we employ a procedure, called FactorAdj, similar to the one
adopted for |L|.
F a c t o r A d j [ a d j _ , np_ , p i _ , pa_ ] : = (
NewAdj=Array [0& , D i m e n s i o n s [ a d j ] ] ;
n p s =np + 1 ;
n p i =np ;
p i i =pi ;
Do [ { I f [ Not [ TrueQ [ S i m p l i f y [ a d j [ [ r , s ] ] ] = = 0 ] ] ,
9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 187

{ FactOfAdj= Simplify [ FactorList [ adj [ [ r , s ] ] ] ] ;

P r i n t [ ' ' F a c t o r s of component [ ' ' , r , ' ' , ' ' , s , ' ' ] : ' ' ] ;
P r i n t [ F a c t O f A d j / / TableForm ] ;
NFactOfAdj = D i m e n s i o n s [ F a c t O f A d j ] ;
LastFactOfAdj =FactOfAdj [ [ 1 , 1 ] ] ^ FactOfAdj [ [ 1 , 2 ] ] *
F a c t O f A d j [ [ NFactOfAdj [ [ 1 ] ] , 1 ] ] ^
F a c t O f A d j [ [ NFactOfAdj [ [ 1 ] ] , 2 ] ] ;
Ai= C o e f f i c i e n t L i s t [ L a s t F a c t O f A d j , { x , y , h ^ 2 , z ^ 2 } ] ;
NewLastFactOfAdj =0;
Do [ { I f [ Not [ TrueQ [ S i m p l i f y [ Ai [ [ i , j , k , l ] ] ] = = 0 ] ] ,
{ n p i = n p i +1 ,
p i i [ [ n p i ] ] = S i m p l i f y [ Ai [ [ i , j , k , l ] ] ] ,
i q =npi ,
Do [ I f [ S i m p l i f y [ p i i [ [ n p i ]] − p i i [ [ n q i ] ] ] = = 0 ,
{ i q =nqi ,
p i i [ [ npi ]]=0 ,
n p i = n p i −1 ,
Break [ ]
}
],
{ n q i , 1 , n p i −1}
],
N e w L a s t F a c t O f A d j = N e w L a s t F a c t O f A d j + pa [ [ i q ] ] *
x ^ ( i −1) * y ^ ( j −1) *
z ^ ( 2 * ( k −1)) * h ^ ( 2 * ( l −1))
}
]
},
{ l , 1 , D i m e n s i o n s [ Ai ] [ [ 4 ] ] } ,
{k , 1 , D i m e n s i o n s [ Ai ] [ [ 3 ] ] } ,
{ j , 1 , D i m e n s i o n s [ Ai ] [ [ 2 ] ] } ,
{ i , 1 , D i m e n s i o n s [ Ai ] [ [ 1 ] ] }
];
NewAdj [ [ r , s ] ] = N e w L a s t F a c t O f A d j ;
Do [ NewAdj [ [ r , s ] ] = NewAdj [ [ r , s ] ] * F a c t O f A d j [ [ i , 1 ] ] ^
FactOfAdj [ [ i , 2 ] ] ,
{ i , 2 , NFactOfAdj [ [ 1 ] ] − 1 } ] ;
P r i n t [ ' ' Component [ ' ' , r , ' ' , ' ' , s , ' ' ] : ' ' , NewAdj [ [ r , s ] ] ] ;
}
]
},
{ s , 1 , Dimensions [ a d j ] [ [ 2 ] ] } ,
{ r , 1 , Dimensions [ a d j ] [ [ 1 ] ] }
];
P r i n t [ ' ' where : ' ' ] ;
I f [ nps <= n p i , Do [ P r i n t [ ' ' p [ ' ' , i , ' ' ] = ' ' , p i i [ [ i ] ] ] , { i , nps , n p i } ] ] ;
Return [ { { NewAdj } , { p i i } , { n p i } } ] ;
);

Thus, since the vector of coefficients pi is updated up to p5 , the simplified expres-

sion of L∗ is obtained by typing:
188 F. Marmo and M. Paradiso

Print [ ' ' d j o i n t matrix of L : ' ' ]

FAdjL= F a c t o r A d j [ AdjL , np , p i , pa ] ;
AdjL=FAdjL [ [ 1 , 1 ] ] ;
p i =FAdjL [ [ 2 , 1 ] ] ;
np=FAdjL [ [ 3 , 1 ] ] ;

which returns the terms of L∗ the additional coefficients. Their expressions are not
reported here for brevity.

9.5 Conclusion

The behaviour of magneto-electro-thermo-elastic materials is governed by a set of

differential equations, e.g. see Formula (9.10). The relevant differential operator L
is expressed as a 6 × 6 matrix of scalar operators as in Formulas (9.12)–(9.17). This
system of equations can be simplified by uncoupling the equations and express-
ing the solution as the sum of up to six quasi-harmonic functions as in Formula
(9.31). Actually, these functions are required to fulfil a series of modified Laplace
equations, see, e.g. Eq. (9.32), in which the differential operator Di is expressed by
(9.29) as a function of the roots of the characteristic polynomial (9.30) associated
with the original set of governing equations. Due to the large number of constitu-
tive parameters involved in this procedure, we have provided a Mathematica script
that computes the coefficients of the characteristic polynomial associated with the
magneto-electro-thermo-elastic coupling and the components of the adjoint differ-
ential operator employed to express the final solution. Forthcoming papers will be
devoted to show how the presented general and complete solution can be specialized
to existing solutions taken from the literature. Actually, depending on the existence of
distinct, partially, or fully coincident roots of the characteristic polynomial equation,
specific solutions already presented in the literature (Wang and Shen 2002; Ding and
Jiang 2003; Chen et al. 2004) can be obtained as special cases of the general solu-
tion derived in the paper. Interesting extensions of the proposed solution include the
effects of higher displacement gradients (Pideri and Seppecher 1997; Camar-Eddine
and Seppecher 2001; Alibert et al. 2003; Franciosi et al. 2019) and/or micropolar
elasticity (Eremeyev et al. 2012; Abd-alla et al. 2017) with applications to panto-
graphic microstructured materials (Barchiesi et al. 2018; Dell’Isola et al. 2019a, b;
De Angelo et al. 2019; Andreaus et al. 2018).

References

AenN A, Alshaikh F, Del Vescovo D, Spagnuolo M (2017) Plane waves and eigenfrequency study
in a transversely isotropic magneto-thermoelastic medium under the effect of a constant angular
velocity. J Thermal Stresses 40(9):1079–1092
Alibert JJ, Seppecher P, Dell’Isola F (2003) Truss modular beams with deformation energy depend-
ing on higher displacement gradients. Math Mech Solids 8(1):51–73
9 Quasi-Harmonic Solutions for Transversely Isotropic . . . 189

Almansi E (1899) Sull’integrazione dell’equazione differenziale δ 2n = 0. Annali di Matematica

Pura ed Applicata (1898–1922) 2(1):1–51
Andreaus U, Spagnuolo M, Lekszycki T, Eugster SR (2018) A Ritz approach for the static analysis of
planar pantographic structures modeled with nonlinear euler-bernoulli beams. Continuum Mech
Thermodyn 30(5):1103–1123
Ashida F, Noda N, Okumura I (1993) General solution technique for transient thermoelasticity of
transversely isotropic solids in cylindrical coordinates. Acta mechanica 101(1):215–230
Barchiesi E, Dell’Isola F, Laudato M, Placidi L, Seppecher P (2018) A 1d continuum model for
beams with pantographic microstructure: asymptotic micro-macro identification and numerical
results. In: Advances in mechanics of microstructured media and structures. Springer, pp 43–74
Bardzokas DI, Filshtinsky ML, Filshtinsky LA (2007) Mathematical methods in electro-magneto-
elasticity, vol 32. Springer Science & Business Media
Camar-Eddine M, Seppecher P (2001) Non-local interactions resulting from the homogenization of
a linear diffusive medium. Comptes Rendus de l’Académie des Sciences-Series I-Mathematics
332(5):485–490
Chen WT (1966) On some problems in transversely isotropic elastic materials. J Appl Mech
33(2):347–355
Chen W (2000) On the general solution for piezothermoelasticity for transverse isotropy with
application. J Appl Mech 67(4):705–711
Chen W, Lee KY, Ding H (2004) General solution for transversely isotropic magneto-electro-
thermo-elasticity and the potential theory method. Int J Eng Sci 42(13–14):1361–1379
De Angelo M, Spagnuolo M, D’annibale F, Pfaff A, Hoschke K, Misra A, Dupuy C, Peyre P,
Dirrenberger J, Pawlikowski M (2019) The macroscopic behavior of pantographic sheets depends
mainly on their microstructure: experimental evidence and qualitative analysis of damage in
metallic specimens. Continuum Mech Thermodyn 31(4):1181–1203
Dell’Isola F, Seppecher P, Alibert JJ, Lekszycki T, Grygoruk R, Pawlikowski M, Steigmann D, Gior-
gio I, Andreaus U, Turco E et al (2019a) Pantographic metamaterials: an example of mathemati-
cally driven design and of its technological challenges. Continuum Mech Thermodyn 31(4):851–
884
Dell’Isola F, Turco E, Misra A, Vangelatos Z, Grigoropoulos C, Melissinaki V, Farsari M (2019b)
Force-displacement relationship in micro-metric pantographs: experiments and numerical simu-
lations. Comptes Rendus Mécanique 347(5):397–405
di Cosmo F, Laudato M, Spagnuolo M (2018) Acoustic metamaterials based on local reso-
nances: homogenization, optimization and applications. In: Generalized models and non-classical
approaches in complex materials, vol 1. Springer, pp 247–274
Ding H, Jiang A (2003) Fundamental solutions for transversely isotropic magneto-electro-elastic
media and boundary integral formulation. Sci China Ser E: Technol Sci 46(6):607–619
Ding H, Chen W, Zhang L (2006) Elasticity of transversely isotropic materials, vol 126. Springer
Science & Business Media
Eerenstein W, Mathur N, Scott JF (2006) Multiferroic and magnetoelectric materials. Nature
442(7104):759–765
Elliott HA, Mott N (1948) Three-dimensional stress distributions in hexagonal aeolotropic crys-
tals. In: Mathematical proceedings of the cambridge philosophical society, vol 44. Cambridge
University Press, pp 522–533
Eremeyev VA, Lebedev LP, Altenbach H (2012) Foundations of micropolar mechanics. Springer
Science & Business Media
Fabrikant V (1989) Applications of potential theory in mechanics: a selection of new results. Kluwer
Academic Publishers, Dordrecht
Franciosi P, Spagnuolo M, Salman OU (2019) Mean green operators of deformable fiber networks
embedded in a compliant matrix and property estimates. Continuum Mech Thermodyn 31(1):101–
132
Green A, Zerna W (1968) Theoretical elasticity. Clarendon Press, Oxford
190 F. Marmo and M. Paradiso

Haojiang D et al (1996) General solutions for coupled equations for piezoelectric media. Int J Solids
Struct 33(16):2283–2298
Lekhniskii S (1940) Symmetrical deformation and torsion of revolution with a special kind of
anisotropy. PPM 4:43–60
Lekhnitskii S (1981) Theory of elasticity of an anisotropic body. Mir Publishers, Moscow
Lin W, Kuo C, Keer L (1991) Analysis of a transversely isotropic half space under normal and
tangential loadings. ASME J Tribol 113:335–338
Marmo F, Rosati L (2016) A general approach to the solution of Boussinesq’s problem for polyno-
mial pressures acting over polygonal domains. J Elasticity 122(1):75–112
Marmo F, Sessa S, Rosati L (2016a) Analytical solution of the Cerruti problem under linearly
distributed horizontal loads over polygonal domains. J Elasticity 124(1):27–56
Marmo F, Toraldo F, Rosati L (2016b) Analytical formulas and design charts for transversely
isotropic half-spaces subject to linearly distributed pressures. Meccanica 51(11):2909–2928
Marmo F, Toraldo F, Rosati L (2017) Transversely isotropic half-spaces subject to surface pressures.
Int J Solids Struct 104:35–49
Marmo F, Toraldo F, Rosati A, Rosati L (2018) Numerical solution of smooth and rough contact
problems. Meccanica 53(6):1415–1440
Marmo F, Sessa S, Vaiana N, De Gregorio D, Rosati L (2020) Complete solutions of three-
dimensional problems in transversely isotropic media. Continuum Mech Thermodyn 32(3):775–
802
Michell J (1900) The stress in an æolotrophic elastic solid with an infinite plane boundary. Proc
Lond Mathe Soc 1(1):247–257
Milton GW, Seppecher P, Bouchitté G (2009) Minimization variational principles for acoustics,
elastodynamics and electromagnetism in lossy inhomogeneous bodies at fixed frequency. Proc
Royal Soc A: Math Phys Eng Sci 465(2102):367–396
Nejadsadeghi N, Placidi L, Romeo M, Misra A (2019) Frequency band gaps in dielectric granular
metamaterials modulated by electric field. Mech Res Commun 95:96–103
Pan E (2002) Three-dimensional green’s functions in anisotropic magneto-electro-elastic bimate-
rials. Zeitschrift für angewandte Mathematik und Physik ZAMP 53(5):815–838
Pan YC, Chou TW (1976) Point force solution for an infinite transversely isotropic solid. J Appl
Mech 43(4):608–612
Pan YC, Chou TW (1979) Green’s function solutions for semi-infinite transversely isotropic mate-
rials. Int J Eng Sci 17(5):545–551
Pideri C, Seppecher P (1997) A second gradient material resulting from the homogenization of an
heterogeneous linear elastic medium. Continuum Mech Thermodyn 9(5):241–257
Podil’chuk YN, Sokolovskii YI (1994) Thermostress in an infinite transversally isotropic medium
with an internal elliptical crack. Int Appl Mech 30(11):834–840
Rosati L, Marmo F (2014) Closed-form expressions of the thermo-mechanical fields induced by a
uniform heat source acting over an isotropic half-space. Int J Heat Mass Transf 75:272–283
Shield R (1951) Notes on problems in hexagonal aeolotropic materials. In: Mathematical proceed-
ings of the Cambridge philosophical society, vol 47. Cambridge University Press, pp 401–409
Sp X, Gao Y, Wang W (2008) Completeness of general solutions for three-dimensional transversely
isotropic piezoelectricity. Int J Solids Struct 45(18–19):5118–5126
Wang M, Wang W (1995) Completeness and nonuniqueness of general solutions of transversely
isotropic elasticity. Int J Solids Struct 32(3–4):501–513
Wang M, Xu X (1990) A generalization of Almansi’s theorem and its application. Appl Math
Modelling 14(5):275–279
Wang X, Yp S (2002) The general solution of three-dimensional problems in magnetoelectroelastic
media. Int J Eng Sc 40(10):1069–1080
Wang Y, Hu J, Lin Y, Nan CW (2010) Multiferroic magnetoelectric composite nanostructures. NPG
Asia Materi 2(2):61–68
Zikung W, Bailin Z (1995) The general solution of three-dimensional problems in piezoelectric
media. Int J Solids Struct 32(1):105–115
Chapter 10
Mathematical Tools for the Seismic
Analysis of Reinforced Concrete
Structures: A Selected Review

Cristoforo Demartino and Sicheng Zhou

Abstract Response spectrum analysis represents the preferential strategy to ana-

lyze and design civil engineering structures subjected to seismic actions. Neverthe-
less, most structural codes were developed by following hand computation-oriented
philosophies so that their prescriptions can be hard to be implemented in finite ele-
ment frameworks and often prevent the use of innovative strategies. This contribution
presents a review of innovative tools focused on reinforced concrete framed struc-
tures aiming to establish a possible organic workflow for design procedures. Some
pivotal issues typical of such a structural typology are hereby addressed, and par-
ticularly, global torsion and capacity checks in the presence of axial force–biaxial
bending responses. This has been done by correlating innovative solutions such as
torsional spectra, seismic envelopes, and limit analysis and by presenting a numerical
procedure capable of performing capacity checks of reinforced concrete cross sec-
tions. The presented strategy aims to be a computationally efficient and exhaustive
procedure to be used within the framework of finite element analysis.

Keywords Seismic analysis · Reinforced concrete structures · Torsional

spectrum · Seismic envelopes · Limit analysis · Finite element

C. Demartino (B) · S. Zhou

Zhejiang University—University of Illinois at Urbana Champaign Institute (ZJUI),
718 East Haizhou Road, Zhejiang 314400, Haining, PR China
e-mail: [email protected]
S. Zhou
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 191

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_10
192 C. Demartino and S. Zhou

10.1 Introduction

Within the framework of earthquake engineering, response spectrum analysis repre-

sents one of the most popular techniques for structural design and it is recommended
by several international codes (Chopra 2007), including the Eurocode 8 (EN 1998–1
1998), as the preferred procedure for building design. In spite of this popularity,
standard code prescriptions are often not oriented to computer-aided computational
analysis so that their implementation in automatic procedures is burdensome and
inefficient.
In fact, some provisions were developed in order to facilitate handmade-oriented
computations in which the human designer is weak in performing large amounts of
calculations, while it is highly skilled in making qualitative and theoretical consid-
erations (Zuccaro et al. 2017). On the contrary, in computer-aided design, structural
software easily handles great computational efforts, while it is barely reliable in
making even simple qualitative choices.
Such drawbacks are even more significant for the case of reinforced concrete
structures since computational strategies capable of addressing the nonlinear and
compressive-only nature of the material are obviously more complex than those
oriented to solve linear structures (Lima et al. 2020). Such an aspect becomes pivotal
in a context aiming to assess risk and resilience of existing structures (Chiaia et al.
2019).
A first aspect concerns the global torsion behavior of buildings, consisting in rota-
tions induced by nonuniform seismic motions and structural asymmetries, which is of
the most importance in seismic design (Karimiyan et al. 2014). Standard codes often
account for torsional effects by enforcing an accidental eccentricity to be applied
by shifting the floor masses from their original location (EN 1998–1 1998; NBCC
2005; NZS 1170 2004).
Such a procedure is conceptually very simple; nevertheless, its application turns
out to be computationally burdensome because it is necessary to analyze several
structural models corresponding to different mass configurations (DeBock et al.
2014).
A further aspect concerns the fact that multicomponent responses play a significant
role in capacity check procedures of reinforced concrete structural elements. In fact,
their safety depends on the combination of more than a single response, as for the
case of biaxial shear (Selna and Lawder 1977) or axial force–biaxial bending (Chen
et al. 2001; Marmo et al. 2011) actions in concrete beams.
Such an issue is not directly addressed by the procedures recommended by stan-
dard codes. In fact, although combination of multiple responses provided by response
spectrum analysis is a well known issue since the late 1977 (Gupta and Singh
1977), code provisions often fail to account for it. Moreover, the definition of multi-
component capacity domains, in general, is addressed by conventionally extending
single-response cases (Sessa et al. 2019b).
As a matter of fact, while the scientific community has made a great effort in
investigating response spectrum analysis, common design of reinforced concrete
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 193

presents a significant gap with respect to the research advancements. In fact, recent
developments in concrete analysis and design concern crack propagation (Barchiesi
et al. 2020; Contrafatto and Cuomo 2006; Contrafatto et al. 2012, 2016), damage
(Placidi et al. 2018, 2019; De Angelo et al. 2019) and viscoelasticity (Cuomo 2017)
as well as analysis of isogeometric elements (Cuomo et al. 2014).
This contribution presents a selected review of the latest advancements concerning
analysis and design of reinforced concrete structures by response spectrum analysis
and proposes a possible synergical and comprehensive strategy. The discussed solu-
tions can be organized in a unique workflow formulated in such a way to allow for
computationally efficient procedures.
In this respect, Sect. 10.2 presents an overview of the strategies capable of account-
ing for global torsion in buildings and analyzes their main benefits and drawbacks.
These strategies are introductory to the computation of multicomponent actions in
structural elements which is addressed in Sect. 10.3.
Moreover, a review of axial force–biaxial bending capacity surfaces of reinforced
concrete beams is introduced in Sect. 10.4 in which recent advancements concerning
approximate formulations, limit analysis, and probabilistic assessment are also dis-
cussed. Then, an innovative approach to perform capacity checks in beams, based on
seismic envelopes (Menun and Der Kiureghian 2000a, b), is presented in Sect. 10.4.1.
Finally, conclusions are reported in Sect. 10.5.

10.2 Review of Strategies Accounting for Global Torsion

in Buildings

Global torsion induced by earthquake excitation is a well-known problem addressed

by several researches over years (Anagnostopoulos et al. 2015). One of the first
attempts to propose a systematic strategy was presented by Newmark in 1969
(Newmark 1969). Such an approach interpreted torsional effects as the result of a base
motion acting with different phases at different points of the foundation because of
wave passage effects and ground inhomogeneities (Chopra and de la Llera 1994a).
Nevertheless, the computed torsional spectrum was relevant to recorded response
spectra and was not determined by design prescriptions.
The main contribution of Newmark research was the definition of the accidental
eccentricity. In fact, since automatic computation was not widespread at the time,
structural analysis was performed by means of simplified models defined in terms of
center of mass and stiffness so that it was straightforward to conventionally increase
the natural eccentricity between these points. The concept of accidental eccentricity
has been furtherly investigated in years and, finally, calibrated for design purposes
(Chopra and de la Llera 1994b, c).
Despite of its intuitive simplicity, accidental eccentricity does not represent a
suitable strategy from a computational point of view since, for the case of three-
194 C. Demartino and S. Zhou

dimensional finite element analysis, it is necessary to define more than a single

structural models. For this reason, several strategies to overcome such a drawback
have been investigated over years.
In order to present a better overview of the available approaches, it is possible to
introduce three main classes.
A first family consists in performing seismic analyses by considering the sole
translational actions and to conventionally increase the structural responses to indi-
rectly account for torsional effects. In particular, multipliers of the translational
responses relevant to a single structural element are determined. They depend on
the element’s coordinates within the structural model and on a few properties of
the analyzed building (Chopra and de la Llera 1995; Goel and Chopra 1993). Such
strategies were defined for the case of flexible diaphragms (Basu and Jain 2004;
Lin et al. 2001). Nevertheless, they introduce strong hypotheses about the properties
of the structural model and cannot be used in conjunction with recent strategies to
address the design of concrete sections such as the ones discussed in Sect. 10.3.
Further typologies of computational approaches consist in adopting either recorded
or artificial global torsion response spectra and are capable of characterizing ground
motions up to six components (Falamarz-Sheikhabadi 2014).
Recent advances in seismology have been focused on recording torsional actions.
A popular strategy consists in extracting rotational motions from translational records
(Ghayamghamian and Nouri 2007) by analyzing accelerations at different stations
(Basu et al. 2013). A further approach relies on direct records by the use of torsional
measurement devices (Lee et al. 2009; Nazarov et al. 2015).
Despite of its capability in determining the actual torsional component of the
ground motions, such strategies are not capable of determining general purpose
design spectra because of the limited quantity of experimental observations.
A more effective approach in structural design consists in determining artificial
torsional spectra by processing the translational actions. In such a case, it is possible to
adopt the design response spectra of the translational analysis so that such procedures
can be applied regardless of the availability of records relevant to the building site.
A promising approach (Shakib 2004) determines the torsional spectrum as the
response of a nonlinear mechanical model, calibrated on soil and structural properties,
reproducing the behavior of the ground subject to translational actions, while a similar
one (Avilés and Suárez 2006) is focused on seismic wave reflection. Both strategies
proved to be effective although they require a sophisticated characterization of the
soil mechanics as well as an uncommon knowledge of the designer.
One of the simplest solutions to compute artificial spectra characterizing global
torsion actions is represented by the Dynamic Equivalent Rotational Spectrum
(DERS) (Sessa et al. 2018b), a recently proposed strategy based on the equivalence
of two degrees-of-freedom (DOFs) oscillators. Such an approach aims to compute
in closed form an artificial torsional spectrum equivalent to the application of the
accidental eccentricity provided by standard codes.
Its main benefit consists in the fact that the torsional spectrum is computed by ana-
lytical expressions defined in terms of the translational spectrum and of mechanical
properties of the structural model.
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 195

Table 10.1 Synoptic comparison of existing strategies for buildings sensitive to torsional effects
Method Advantages Disadvantages
Accidental eccentricity Intuitive and consolidated Burdersome for 3D structural
strategy analysis
Translational response Simple and fast approach, easy Applicable to a limited number
increment (Basu and Jain to implement of structural typologies; it
2004; Chopra and de la Llera usually requires restrictive
1995; Goel and Chopra 1993; dynamical features
Lin et al. 2001)
Recorded torsional response Computationally efficient, Lack of territorial data; it
spectra (Basu et al. 2013; provides reliable results requires complex
Ghayamghamian and Nouri computations for defining the
2007; Lee et al. 2009; Nazarov torsional spectrum. It is not
et al. 2015) permitted by some building
codes
Artificial torsional response Computationally efficient, They may require a complex
spectra (Avilés and Suárez provide conservative results. characterization of the soil and
2006; Sessa et al. 2018b; The DERS fulfills the can be not fully equivalent to
Shakib 2004) requirements of building codes the accidental eccentricity
prescriptions for multi DOFs
structures

Because of its simplicity, it will be used in the sequel to account for global tor-
sion, although the designer can choose to adopt alternative formulations or recorded
spectra; for this reason, it is summarized in Sect. 10.2.1.
As a general consideration, torsional spectra, regardless of their recorded or artifi-
cial nature, represent the best strategy to account for global torsion. In fact, response
spectrum analysis performed in the combination of modal analysis consists in super-
imposing the structural responses relevant to each spectrum of the seismic excitation.
For the case of hand computations, a further response spectrum results in an
increased computational burden, while the adoption of an increment of the natural
eccentricity is a straightforward strategy. On the contrary, when spectral analysis is
performed by automatic computations, a further component of the seismic action is
not burdensome to perform since it results in redundant calculations.
For the reader’s convenience, Table 10.1 reports a summary of the strategies
described in this sections specifying a synoptic comparison of their main advantages
and drawbacks.

10.2.1 Review of the Dynamic Equivalent Rotational

Spectrum

The analytical definition of the DERS presented in Sessa (2018b) is based on the
outcomes of a modal analysis in which structural responses are computed in terms of
196 C. Demartino and S. Zhou

modes of vibration, each one relevant to a natural pulsation. The artificial spectrum
is computed by enforcing the equivalence between two oscillators, each one having
two DOFs, modeling the response of each vibration mode.
Specifically, the response of a generic vibration mode, with pulsation ω, is mod-
eled by means of a linear oscillator presenting a translational u χ and a rotational u θ
displacements.
We also denote by ε the value of the structural eccentricity, by r the inertial radius
of gyration of the structural model about the vertical axis and by  the frequency
ratio, i.e., the ratio between the translational and rotational natural frequencies of the
structural model.
Under these assumptions, the equation of motion of the two DOFs oscillator is
ruled by the equation of motion:
⎡ ⎤
  ω2 ε    
⎢ ω
2
ü χ ⎥ uχ aχ (t)
+⎢
⎣ ω2 ε
r ⎥ =− , (10.1)
r ü θ ε2 ⎦ r u θ raθ (t)
ω  + 2
2 2
r r
where aχ is the value of the modal acceleration of the translational action, while aθ
is the value of an unknown global torsion acceleration.
Such a general oscillator can be specialized in two subcases. A coupled oscillator,
in which the value of the eccentricity is set equal to the quantity enforced by standard
codes and the global torsion acceleration is set to zero, and an uncoupled oscillator
in which there is no eccentricity and the (unknown) torsional acceleration is not null.
By enforcing equivalence between the responses of both oscillators, the torsional
acceleration aθ is the sole unknown quantity of the problem. As a consequence, if
the translational acceleration is defined by means of response spectrum, it is possible
to compute the relevant value of an artificial rotational spectrum providing the very
same response of the standard code eccentricity.
Such a problem can be solved in closed form by introducing some auxiliary
variables:

1 + 2 + ε2 /r 2 ω2 2
c = ω2 ; R= 2 + ε2 /r 2 − 1 + 4ε2 /r 2 (10.2)
2 2
2π 2π
T1 = √ ; T2 = √ (10.3)
c−R c+R

representing the two natural periods of the coupled oscillator,

2 2
c1 = ω2 ε/r ; c2 = −ω2 + c − R (10.4)

and, finally, computing the value of the artificial torsional spectrum Dθ (T /) as
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 197

(a) variable, b = 10m (b) b variable

Fig. 10.1 Global torsion response spectra

c1 c2
Dθ (T /) = 2
Dχ,1 − Dχ,2
2
− Dχ,1 Dχ,2 ρ1,2 (10.5)
r [c1 + c2 ]

where T = 2π/ω is the natural period of the considered vibration mode, Dχ,1 and
Dχ,2 are the values of the translational response spectrum relevant to periods T1 and
T2 and ρ1,2 is the modal correlation coefficient of the coupled oscillator.
The latter can be easily computed by the complete quadratic combination (Smeby
and Der Kiureghian 1985) as

8ξ 2 (T2 /T1 )3/2

ρ1,2 =  (10.6)
(1 + T2 /T1 ) (1 − T2 /T1 )2 + 4ξ 2 (T2 /T1 )

in which ξ is the damping ratio associated with the vibration mode.

We emphasize that response spectra have been reported in terms of displacements.
The corresponding pseudo-acceleration and pseudo-velocity spectra can be easily
computed as

Saθ,χ Sχ , T = Dθ (T /) ω2 ; Svθ,χ Sχ , T = Dθ (T /) ω (10.7)

Figure 10.1a and b present some examples of pseudo-velocity global torsion spec-
tra in which the accidental eccentricity has been set ε = 0.05b with b representing
√
the generic width of the structural model and the radius of gyration is r = b/ 6.
We emphasize that such artificial spectra can be quite different from recorded
ones since the DERS is not representative of any real ground motion. However,
it represents an easy procedure to account for global torsion by adopting a single
structural model.
198 C. Demartino and S. Zhou

10.3 Computation of Multicomponent Actions by Seismic

Envelopes

Response spectrum analysis computes the peak value of a structural quantity of

interest by superimposing the outcomes of different vibration modes relevant to a set
of seismic components.
It is worth being emphasized that each structural response due to a seismic action
is, in general, a quantity that varies in time and spectral analysis is not capable of
determining the instant at which each peak occurs.
More generally, if capacity checks depend on more than a single response, such
as for the case of axial force–biaxial bending capacity of reinforced concrete beams,
the peaks of different components of the response are not necessarily synchronous.
Moreover, outcomes of the spectral analysis are computed by the superposition
of the responses relevant to each seismic components which is usually defined by
means of quadratic forms. Thus, the algebraic sign of these responses is undefined.
Importance of a proper combination of different response components is a well-
established problem whose importance was highlighted in the late ’70s by Gupta
and Singh (1977). Their approach introduced a rectangular envelope, determined
by assigning to the response components all possible combinations of the algebraic
sign.
Although simple, such an approach is not very accurate, presents some draw-
backs in accounting for seismic actions with unknown input directions and does not
accounts for the correlation between the response components.
A theoretical solution overcoming these drawbacks was proposed by Menun and
Der Kiureghian (2000a, b), and later extended to perform capacity checks of rein-
forced concrete beam sections (Sessa et al. 2015). Such an approach is based on the
fact that multi-component responses computed by the response spectrum procedure
can be conveniently represented by tensor analysis (Lebedev et al. 2010).
To fix the ideas, let us consider the space determined by the components of the
structural response of interest. For the case of axial force–biaxial bending capacity
checks the response is defined by the axial force P acting on a cross section and
two bending moments M1 and M2 about the section reference axes. The Supreme
Envelope determines a boundary surface which encompasses all possible values that
the multi-component response can attain during the seismic motion.
To compute the Supreme Envelope expression, it is necessary to determine the
vibration modes of the structural models. Specifically, each mode presents three
components of the response which can be arranged in the following matrix:
⎡ ⎤
P1 . . . Pi . . . Pn
Ψ = ⎣ M11 . . . M1i . . . M1n ⎦ (10.8)
M21 . . . M2i . . . M2n

where n is the number of the considered modes.

10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 199

To account for seismic actions, it is useful to introduce two sets of n × n diagonal

matrices Dk and Γ k . Index k denotes the component of the seismic action with
k = 1, 2 corresponding to the horizontal components, k = 3 to the vertical one, and
k = 4 to the global torsion one defined in Sect. 10.2. Elements Di,i,k and Γi,i,k = γik ,
belonging to the principal diagonals of D and Γ denote, respectively, the values of
the displacement spectrum and of the participating factor of mode i with respect to
the seismic component k.
Modal superposition adopts cross-correlation coefficients of the Complete
Quadratic Combination (CQC) procedure (Wilson et al. 1981) which are arranged in
the square matrix R whose element ρi, j denotes the correlation coefficient between
modes i and j.
Denoting by fd the structural response vector produced by static forces and
by X S (α) a 3 × 3 matrix depending on modal and spectral quantities (Menun
and Der Kiureghian 2000a), the relationship determining the Supreme Envelope
is defined as follows:
X S (α) α
f (α) = fd + 0.5 (10.9)
α X S (α) α
T

where vector f (α) denotes a generic point of the envelope boundary, while α is a
unit vector used as parameter.
Matrix X S (α) is computed by combining modal responses and spectral quantities
by the equation:
 
1 1 1
X S (α) = Z1 + Z2 − Z2 P2 (α) + Z3 P3 (α) (10.10)
2 2 2

where:  4 

Z1 = Ψ Γ k Dk RDkT Γ kT ΨT (10.11)
k=1

 

2 
2
Z2 = Ψ − (−1) k + Γ l Dk RDkT Γ lT ΨT (10.12)
k=1 l=1

 2 

Z3 = Ψ (−1)k Γ 1 Dk RDkT Γ 2T + Γ 2 Dk RDkT Γ 1T Ψ T (10.13)
k=1

α T Z2 α α T Z3 α
P2 (α) = − ; P3 (α) = (10.14)
H (α) H (α)

and:  
2 2 0.5
H (α) = α T Z2 α + α T Z3 α (10.15)
200 C. Demartino and S. Zhou

Fig. 10.2 Example of a 2D

Supreme Envelope, unit
vector α, and the relevant
point of the envelope f(α)

To better understand the role of α as parameter, it is useful to introduce its physical

interpretation. Specifically, α is the unit vector normal to the hyperplane tangent to
the envelope boundary at f (α); Fig. 10.2 presents a 2D example for simplicity. In
order to obtain a large set of points belonging to the envelope, it is possible to assume
different orientations of α by making it span the whole space of the actions and obtain
all corresponding responses f(α).
It is worth being emphasized that, despite of their apparent complex form, such
relationships consist in combining quadratic forms of square matrices and these have
been obtained by combining in a suitable way spectral and modal outcomes usually
employed in classic response spectrum analysis. In this respect, all calculations are
easy to be implemented into an automatic algorithm processing the outcomes of a
finite element analysis.
Moreover, it is interesting to analyze the sum of Eq. (10.11) in which matrices Dk ,
containing the response spectrum values relevant to each mode and to each seismic
components, are multiplied by the ones containing the participation factors and the
correlation coefficients.
The presence of the global torsion response spectrum, corresponding to the fourth
component, simply determines a further sum, a fast and easy task to be performed
by a numerical algorithm. We emphasize that the classical mass-shifting strategy to
account for the accidental eccentricity would compromise the whole procedure since
different mass locations would not permit the computation of a univocal matrix Ψ
containing the modal responses of the structure. Hence, it is apparent that the adoption
of a torsional spectrum turns out to be a procedure by far easier to be implemented
than strategies dealing with mass locations.
Figure 10.3 presents the Supreme Envelope (in blue) computed by Eq. (10.9) rel-
evant to a reinforced concrete cross section compared with the corresponding rect-
angular envelope computed by the procedure (Gupta and Singh 1977). Note that, in
such a peculiar case, the latter turns out to be non-conservative with respect to the
Supreme Envelope.
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 201

Fig. 10.3 Example of the Supreme Envelope (yellow) and of the Rectangular Envelope of a rein-
forced concrete column section

Such an issue is due to the fact that, while the rectangular envelope and, in general,
the classic response spectrum strategy computes the peak responses relevant to a
single seismic input direction, the Supreme Envelope accounts for its randomness.

10.4 Capacity Checks of Reinforced Concrete Beams

A significant step in the structural design of reinforced concrete structures consists

in the determination of the capacity domain of each cross section. In fact, the seismic
actions computed by the strategies described in Sect. 10.3, in particular the Supreme
Envelope, must be fully contained within a closed surface representing the boundary
between a safe and a collapse region, namely, the Capacity Surface.
Making reference to the ultimate limit state strategy, such a boundary is relevant
to all states for which either a point of the concrete region of the cross section or one
of the reinforcement bars attains its limit state.
Standard codes, such as Eurocode 2 (EN 1992–1–1 1992), define limit states of
materials in terms of maximum normal strain εz , where z represents the axis normal
to the cross-sectional plane. Moreover, constitutive relationships of both the concrete
and reinforcement bar materials usually present a nonlinear behavior. In particular,
202 C. Demartino and S. Zhou

in Eurocode 2, steel is assumed to be elastic–perfectly plastic, while concrete can be

modeled by a Parabola–Rectangle compressive-only relationship.
Strain field is assumed to be linear over the cross section; thus, points of the
capacity surface are defined by internal actions for which the strain field attains the
limit state in at least a point of the section. In this context, the values of the axial
force and of the two bending moments relevant to a strain distributions are computed
by means of integrals of the stress over the cross section.
Such a problem, although efficiently addressed for the case of rectangular sec-
tions (Papanikolaou and Sextos 2016), in common practice needs to be performed
for arbitrarily shaped elements. Besides of the intuitive strategy of adopting a fiber
discretized description of the cross section, integrals can be computed by a large vari-
ety of different strategies such as superposition of elementary-shaped section (Silva
et al. 2009) and incremental iterative approaches (Chiorean 2010, 2013).
Interesting approaches are focused on the use of boundary integrals to get exact
solutions (Matuszak and Pluciński 2014) and proved to be more accurate than fiber-
oriented procedures both numerically (Matuszak 2017) and experimentally (Koziński
and Winnicki 2016).
One of the most promising ones is a fiber-free algorithm (Alfano et al. 2007) based
on a polygonal description of the cross section in which its boundary is expressed as
function of the coordinates of the vertices. Fixed a strain field for which the ultimate
limit state is reached, the corresponding values of the axial force and of the bending
moments are computed in closed form by applying the Gauss–Green theorem and by
adopting exact integrals of the elastic–plastic responses of steel and concrete (Marmo
et al. 2008).
Such a strategy proved to be very effective in providing exact results and it was
applied to pre-stressed concrete sections (Marmo et al. 2011) as well as to evaluate
the response of reinforced concrete elements within the framework of finite element
analysis (Marmo and Rosati 2012, 2013).
It is worth being emphasized that limit state analysis has a conventional nature;
thus, its results can be not significant of the actual physical behavior of reinforced
concrete beams. Moreover, the determination of the capacity surface is not provided
in closed form but is defined by a set of discretized points.
The first issue has been recently investigated from a probabilistic point of view.
In particular, it has been shown in Sessa (2019b) that the conventional ultimate limit
state procedure was developed for axial force–uniaxial bending actions. Hence, its
results, although conservative, turn out to be quite inaccurate for the case of biaxial
bending. The distribution of the collapse probability over the capacity surface is far
to be regular and uniform, as it would be recommendable. On the contrary, it turns
out to be over-conservative for most of its regions.
Moreover, recent investigations compared the capacity domains of reinforced
concrete sections, computed by such a fiber-free approach, and the capacity surfaces
determined by limit analysis (Sessa et al. 2018a). The latter strategy represents an
appealing tool since it is based on the well-known safe theorem and is less compu-
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 203

tationally demanding than traditional limit state analyses. It has been shown that, as
long as a sufficient ductility is ensured, limit analysis provides accurate estimations
of the capacity surface.
Use of limit analysis (Malena and Casciaro 2008) was first developed for shake-
down analyses of 3D frames (Casciaro and Garcea 2002) and has been enhanced to
account for Eurocode 8-compliant load combinations (Leonetti et al. 2015). Such a
strategy reduces the computational burden by reducing redundant constraints (Simon
and Weichert 2012; Spiliopoulos and Panagiotou 2017) and introduces the possibility
of representing the domain boundaries by approximate formulations. In particular,
capacity surface can be represented by a Minkowsky sum of ellipsoids (Yan and
Chirikjian 2015), a mathematical tool representing multidimensional surfaces as a
function of their gradient.
For the case of axial force–biaxial bending domains, the Minkowsky sum of
ellipsoids can be determined either by optimization (Bleyer and Buhan 2013) or by
a reduction approach. The latter consists in dividing the reinforced concrete section
in triangles and rectangles, whose theoretical contribution to the overall capacity can
be computed in closed form, and then by superimposing such contributions to obtain
a global capacity surface (Magisano et al. 2018).
It is worth being emphasized that the most suitable approach to determine the
capacity surface of reinforced concrete cross sections is strongly dependent on the
peculiar problem. For example, limit analysis approaches are effective for the case
of new buildings for which code prescriptions require a sufficient ductility. On the
contrary, they cannot be easily used for the case of existing buildings.
Solutions based on the exact computation of the capacity of rectangular subsec-
tions are suitable if the cross-sectional geometry fulfills some geometrical conditions,
while the fiber-free approach (Alfano et al. 2007) can address sections of arbitrary
shape but its employment can result burdensome if used in conjunction with the
Supreme Envelope described in Sect. 10.3.
A further aspect concerns the fact that, in common design, a single section can be
subject to a large set of different actions and it is convenient to determine its capacity
surface only once what is often done by means of a discretized set of points.
Referring to the cited works for the details concerning the capacity surfaces,
the following subsection proposes an approximate strategy capable of performing
capacity checks by combining actions defined in terms of the Supreme Envelope and
a capacity surface defined by means of discrete points. Such an approach aims to
propose a general procedure which permits to the designer the choice of the most
suitable capacity surface formulation, thus taking advantage of the large variety of
solutions provided by the literature.
204 C. Demartino and S. Zhou

Fig. 10.4 Example of 2D

capacity check performed by
the Supreme Envelope and a
capacity surface

10.4.1 A General Algorithm to Perform Capacity Checks

by the Supreme Envelope

Capacity checks performed by formulating the seismic action by the Supreme Enve-
lope and by adopting a capacity surface aim to determine the critical multiplier λ of
the seismic response spectra which makes the Supreme Envelope to become tangent
to the capacity surface.
For simplicity, Fig. 10.4 shows a 2D example in which the capacity surface has
been plotted in blue while the Supreme Envelope corresponding to the actual seismic
excitation has been reported in black. The Supreme Envelope amplified by the critical
load multiplier λ has been reported in red.
Note that the amplified envelope and the capacity surface present the same tangent
hyperplane at the point λf(α) with normal unit vector α.
Determination of α and λ is not straightforward and, in general, cannot be per-
formed in closed form. Numerical algorithms have been proposed to address such
an issue, although the one proposed in Menun (2004b) assumes a simplified for-
mulation of the capacity surface not suitable for reinforced concrete sections while
the numerical algorithm presented in Sessa (2015) requires the computation of the
fiber free surface (Alfano et al. 2007) and is computationally demanding if the cross
section is subject to several actions.
In order to permit efficient capacity checks regardless of the strategy adopted
for determining the capacity domain of the section, let us assume that the capacity
surface D is defined by means of a sufficiently dense set of discretized points whose
ith element is: T
di = Pi , M1,i , M2,i (10.16)

where Pi denotes the axial force while M1,i and M2,i are the two bending components.
Representation of a continue surface approximating a set of discrete point can be
conveniently performed by defining a Minkowski sum of ellipsoids which is capable
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 205

of defining n-dimensional surfaces by means of a superposition of ellipsoids (Yan

and Chirikjian 2015), as illustrated before.
The surface is defined by means of the support function, representing the distance
of the tangent hyperplane at t from the origin, which is:
0.5 0.5
π (t) = α (t)T M−1 α (t) = M−1 α (t) · α (t) (10.17)

where α is the unit vector normal to the tangent hyperplane.

Matrix M is numerically calibrated by means of a least-square minimization
algorithm so that the surface approximates the set of discretized points belonging to
the capacity domain. Hence, the vector α can be used as a parameter for computing
any point of the ellipsoids by the relationship:

M−1 α
t (α) = 0.5 (10.18)
α T M−1 α

More complex shapes can be properly approximated by the superposition of sev-

eral ellipsoids. Specifically, denoting as Mi , i = 1 . . . m, the canonic matrices of m
ellipsoids and as t0 the vector locating the domain center with respect to the origin,
the Minkowski sum is defined as:


m
Mi−1 α
t (α) = t0 + 0.5 (10.19)
i=1 α T Mi−1 α

Figure 10.5 presents an illustrative Minkowski sum (represented by the black

mesh) of three ellipsoids (represented in yellow, green and blue), approximating the
capacity domain of a RC cross section, represented by the red bullets.
It is worth being emphasized that the expression determining the approximated
domain of Eq. (10.19) is very similar to Eq. (10.9) representing the seismic Supreme
Envelope.
In this sense, the condition for which the Supreme Envelope is tangent to the
capacity domain turns out to be:

X S (α) α 
m
Mi−1 α
fd − t0 + λ 0.5 − 0.5 = 0 (10.20)
α T X S (α) α i=1 α T Mi−1 α

where λ is a multiplier of the seismic action.

For this reason, the capacity check relevant to a single cross section consists in
determining parameters λ and α fulfilling Eq. (10.20) which can be easily computed
by several numerical algorithms.
206 C. Demartino and S. Zhou

Fig. 10.5 Illustrative

3–ellipsoids Minkowski sum

The described procedure provides approximate although conservative results since

the Minkowski sum of ellipsoids is always contained within the actual capacity sur-
face. We emphasize once more that such a procedure aims to perform capacity checks
for common design purposes with limited computational burden. Hence, as long as
the capacity surface is discretized in a sufficient number of points, approximations
will be affordable. Moreover, in contrast with the capacity check procedure described
in Sessa (2015), the proposed strategy does not involve burdensome computations
provided that the capacity check is limited to numerically solve a single nonlinear
equation.

10.5 Conclusions

A review of recent developments concerning seismic analysis of reinforced concrete

structures by response spectrum has been presented. The discussed strategies repre-
sent some of the most advanced contributions for performing structural analysis and
design within the finite element framework.
To this end, a classification of the available approaches for properly modeling the
effects of global torsion in buildings has been discussed. The presented strategies
have been selected in order to ensure easy implementations in structural analysis
software, particularly in finite element frameworks, proving that the most effective
strategy consists in adopting global torsion response spectra.
Within this context, a selected state of the art concerning available formulations
of torsional spectra has been reported and a recent strategy for defining an artifi-
cial torsional spectrum has been presented. Such an approach, defined as Dynamic
Equivalent Rotational Spectrum (DERS), provides a global torsion action, as func-
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 207

tion of the mechanical properties of the structure and of the translational action,
whose response is equivalent to the effects of the accidental eccentricity prescribed
by standard codes.
The problem of multi-component structural responses, pivotal for the case of axial
force—biaxial bending actions in reinforced concrete frames, has been addressed
by means of seismic envelopes. Such a strategy takes into account the correlation
between all components of the seismic response acting on a cross section and properly
considers the randomness of the direction of propagation of earthquakes.
In order to analyze the characterization of the structural capacity of reinforced con-
crete beam elements, a comparison between different formulations of axial force—
biaxial bending capacity domains has been presented. The discussed approaches
are relevant to different theories such as the classical ultimate limit state analysis,
adopted by standard codes, and recent results obtained by the limit analysis theory
and the application of the Safe Theorem.
Finally, a numerical strategy for performing capacity checks by using seismic
envelopes has been proposed. Such a procedure needs the capacity surface to be
defined by a set of discretized points and determines an approximate capacity sur-
face defined by means of a Minkowski sum of ellipsoids. A critical multiplier of
the seismic load, defined as the scalar coefficient that makes the seismic Supreme
Envelope to be tangent to the boundary, can therefore be computed by a numerical
algorithm solving a vectorial equation.
In conclusion, the presented tools aim to represent a rational workflow for the
analysis and design of reinforced concrete beam structures. The organic use of the
discussed strategies permits an easy and computationally efficient implementation of
a unique strategy consistent with the most common procedures of the finite element
analysis.
Future challenges about the use of seismic envelopes in analyzing reinforced con-
crete structures are particularly focused on shell structural elements (Altenbach and
Eremeyev 2009; Altenbach et al. 2009, 2010). In particular, an envelope formula-
tion, based on the Mohr’s circle and capable of addressing elastic shell elements,
is already available (Menun 2004a), although it needs to be extended to nonlinear
elements in order to address complex and curved structures (Eremeyev and Zubov
2007; Marmo et al. 2019). Effective extensions to address the nonlinear behavior
of concrete should properly include reinforcements in nonlinear elements (Caggegi
et al. 2018; Valoroso et al. 2014, 2015; Salman et al. 2019) and confinement (Serpieri
et al. 2018; Sessa et al. 2017a, b, 2019a). Further developments concern the use of
explicit solutions (Marmo et al. 2016, 2020) in order to take into account the stiffness
of the foundation soil as well as the use of innovative beam models accounting for
shear and torque (Paradiso et al. 2020, 2021).
208 C. Demartino and S. Zhou

References

Alfano G, Marmo F, Rosati L (2007) An unconditionally convergent algorithm for the evaluation
of the ultimate limit state of RC sections subject to axial force and biaxial bending. Int J Numer
Methods Eng 72:924–963. https://doi.org/10.1002/nme.2033
Altenbach H, Eremeyev V (2009) On the linear theory of micropolar plates. ZAMM J Appl Math
Mech/Zeitschrift für Angewandte Mathematik und Mechanik 89:242–256. https://doi.org/10.
1002/zamm.200800207
Altenbach H, Eremeyev V, Morozov N (2009) Linear theory of shells taking into account surface
stresses. Dokl Phys 54:531–535. https://doi.org/10.1134/S1028335809120039
Altenbach J, Altenbach H, Eremeyev V (2010) On generalized cosserat-tape theories of plates
and shells: a short review and bibliography. Arch Appl Mech 80:73–92. https://doi.org/10.1007/
s00419-009-0365-3
Anagnostopoulos S, Kyrkos M, Stathopoulos K (2015) Earthquake induced torsion in buildings:
critical review and state of the art. Earthquakes and Structures 8:305–377. https://doi.org/10.
12989/eas.2015.8.2.305
Avilés J, Suárez M (2006) Natural and accidental torsion in one-storey structures on elastic foun-
dation under non-vertically incident sh-waves. Earthq Eng Struct Dyn 35(7):829–850
Barchiesi E, Hua Y, Tran C, Placidi L, Mueller W (2020) Computation of brittle fracture propagation
in strain gradient materials by the fenics library. Math Mech Solids 26(108128652095):451.
https://doi.org/10.1177/1081286520954513
Basu D, Jain S (2004) Seismic analysis of asymmetric buildings with flexible floor diaphragms. J
Struct Eng-ASCE 130. https://doi.org/10.1061/(ASCE)0733-9445(2004)130:8(1169)
Basu D, Whittaker A, Constantinou M (2013) Extracting rotational components of earthquake
ground motion using data recorded at multiple stations. Earthq Eng Struct Dyns 42. https://doi.
org/10.1002/eqe.2233
Bleyer J, Buhan P (2013) Yield surface approximation for lower and upper bound yield design of
3d composite frame structures. Comput Struct 129:86–98. https://doi.org/10.1016/j.compstruc.
2013.08.011
Caggegi C, Sciuto D, Cuomo M (2018) Experimental study on effective bond length of basalt textile
reinforced mortar strengthening system: contributions of digital image correlation. Measurement
129. https://doi.org/10.1016/j.measurement.2018.07.003
Casciaro R, Garcea G (2002) An iterative method for shakedown analysis. Comput Methods Appl
Mech Eng 191(49):5761–5792. https://doi.org/10.1016/S0045-7825(02)00496-6
Chen S, Teng J, Chan SL (2001) Design of biaxially loaded short composite columns of arbitrary
section. J Struct Eng-ASCE 127. https://doi.org/10.1061/(ASCE)0733-9445(2001)127:6(678)
Chiaia B, Barchiesi E, De Biagi V, Placidi L (2019) A novel worst-case-based structural resilience
index: definition, computation and applications to portal frame structures. Mech Res Commun
99. https://doi.org/10.1016/j.mechrescom.2019.03.007
Chiorean C (2010) Computerised interaction diagrams and moment capacity contours for compos-
ite steel-concrete cross-sections. Eng Struct 32:3734–3757. https://doi.org/10.1016/j.engstruct.
2010.08.019
Chiorean C (2013) A computer method for nonlinear inelastic analysis of 3d semi-rigid steel frame-
works. Eng Struct 57:125–152
Chopra A (2007) Elastic response spectrum: a historical note. Earthq Eng Struct Dyn 36:3–12.
https://doi.org/10.1002/eqe.609
Chopra AK, de la Llera JC (1994a) Accidental torsion in buildings due to base rotational excitation.
Earthq Eng Struct D 23:1003–1021
Chopra AK, de la Llera JC (1994b) Evaluation of code accidental-torsion provisions from building
records. J Struct Eng 120:597–616
Chopra AK, de la Llera JC (1994c) Using accidental eccentricity in code-specified static and
dynamic analyses of buildings. Earthq Eng Struct D 23:947–967
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 209

Chopra AK, de la Llera JC (1995) Estimation of accidental torsion effects for seismic design of
buildings. J Struct Eng 121:102–114
Contrafatto L, Cuomo M (2006) A framework of elastic-plastic damaging model for concrete under
multiaxial stress states. Int J Plasticity 22:2272–2300. https://doi.org/10.1016/j.ijplas.2006.03.
011
Contrafatto L, Cuomo M, Fazio F (2012) An enriched finite element for crack opening and rebar
slip in reinforced concrete members. Int J Fract 178:33–50. https://doi.org/10.1007/s10704-012-
9723-1
Contrafatto L, Cuomo M, Gazzo S (2016) A concrete homogenisation technique at meso-scale level
accounting for damaging behaviour of cement paste and aggregates. Comput Struct 173. https://
doi.org/10.1016/j.compstruc.2016.05.009
Cuomo M (2017) Forms of the dissipation function for a class of viscoplastic models. Math Mech
Complex Syst 5:217–237. https://doi.org/10.2140/memocs.2017.5.217
Cuomo M, Contrafatto L, Greco L (2014) A variational model based on isogeometric interpolation
for the analysis of cracked bodies. Int J Eng Sci 80:173–188. https://doi.org/10.1016/j.ijengsci.
2014.02.017
De Angelo M, Spagnuolo M, D’Annibale F, Pfaff A, Hoschke K, Misra A, Dupuy C, Peyre P, Dirren-
berger J, Pawlikowski M (2019) The macroscopic behavior of pantographic sheets depends mainly
on their microstructure: experimental evidence and qualitative analysis of damage in metallic
specimens. Continuum Mech Thermodyn 31. https://doi.org/10.1007/s00161-019-00757-3
DeBock D, Liel A, Haselton C, Hooper J, Henige R (2014) Importance of seismic design accidental
torsion requirements for building collapse capacity. Earthq Eng Struct Dyn 43:831–850. https://
doi.org/10.1002/eqe.2375
EN 1992-1-1 (1992) Design of concrete structures—part 1-1: general rules and rules for buildings.
Standard, CEN
EN 1998–1 (1998) Design of structures for earthquake resistance. Part I: General rules, seismic
actions and rules for buildings. Standard, CEN
Eremeyev VA, Zubov LM (2007) On constitutive inequalities in nonlinear theory of elastic shells. J.
Appl. Math. Mech./Zeitschrift für Angewandte Mathematik und Mechanik 87(2):94–101. https://
hal.archives-ouvertes.fr/hal-00835648
Falamarz-Sheikhabadi MR (2014) Simplified relations for the application of rotational components
to seismic design codes. Eng Struct 59:141–152. https://doi.org/10.1016/j.engstruct.2013.10.035
Ghayamghamian M, Nouri G (2007) On the characteristics of ground motion rotational mponents
using chiba dense array data. Earthq Eng Struct D 36:1407–1429
Goel R, Chopra AK (1993) Seismic code analysis of buildings without locating centers of rigidity.
J Struct Eng-ASCE 119:3039–3055
Gupta A, Singh M (1977) Design of column sections subjected to three components of earthuquake.
Nucl Eng Des 41:129–133
Karimiyan S, Kashan A, Karimiyan M (2014) rogressive collapse vulnerability in 6-story RC
symmetric and asymmetric buildings under earthquake loads. Earthq Struct 6:473–494
Koziáski K, Winnicki A (2016) Experimental research and analysis of load capacity and deforma-
bility of slender high strength concrete columns in biaxial bending. Eng Struct 107:47–65. https://
doi.org/10.1016/j.engstruct.2015.10.025
Lebedev L, Cloud M, Eremeyev V (2010) Tensor analysis with applications in mechanics. https://
doi.org/10.1142/7826
Lee W, Igel H, Trifunac M (2009) Recent advances in rotational seismology. Seismol Res Lett
80:479–490. https://doi.org/10.1785/gssrl.80.3.479
Leonetti L, Casciaro R, Garcea G (2015) Effective treatment of complex statical and dynamical load
combinations within shakedown analysis of 3d frames. Comput Struct 158(C):124–139, https://
doi.org/10.1016/j.compstruc.2015.06.002
Lima C, Angiolilli M, Barbagallo F, Belletti B, Bergami A, Camata G, Cantagallo C, Di Domenico
M, Fiorentino G, Ghersi A, Gregori A, Lavorato D, Luciano R, Marino E, Martinelli E, Nuti C,
Ricci P, Rosati L, Ruggieri S, Verderame G (2020) Nonlinear modeling approaches for existing
210 C. Demartino and S. Zhou

reinforced concrete buildings: the case study of De Gasperi-Battaglia School Building in Norcia,
pp 82–95
Lin WH, Chopra A, de la Llera J (2001) Accidental torsion in buildings: analysis versus earthquake
motions. J Struct Eng-ASCE 127. https://doi.org/10.1061/(ASCE)0733-9445(2001)127:5(475)
Magisano D, Liguori FS, Leonetti L, Garcea G (2018) Minkowski plasticity in 3d frames: decoupled
construction of the cross-section yield surface and efficient stress update strategy: Minkowski
plasticity in 3d frames. Int J Numer Methods Eng 116:435–464. https://doi.org/10.1002/nme.
5931
Malena M, Casciaro R (2008) Finite element shakedown analysis of reinforced concrete 3d frames.
Comput Struct 86:1176–1188
Marmo F, Rosati L (2012) Analytical integration of elasto-plastic uniaxial constitutive laws over
arbitrary sections. Int J Numer Methods Eng 91:990–1022. https://doi.org/10.1002/nme.4316
Marmo F, Rosati L (2013) The fiber-free approach in the evaluation of the tangent stiffness matrix
for elastoplastic uniaxial constitutive laws. Int J Numer Methods Eng 94:868–894. https://doi.
org/10.1002/nme.4484
Marmo F, Rosati L, Sessa S (2008) Exact integration of uniaxial elasto-plastic laws for nonlinear
structural analysis. AIP Conf Proc 1020:1219–1226
Marmo F, Serpieri R, Rosati L (2011) Ultimate strength analysis of prestressed reinforced concrete
sections under axial force and biaxial bending. Comput Struct 89:91–108. https://doi.org/10.
1016/j.compstruc.2010.08.005
Marmo F, Sessa S, Rosati L (2016) Analytical solution of the cerruti problem under linearly
distributed horizontal loads over polygonal domains. J Elasticity 124. https://doi.org/10.1007/
s10659-015-9560-3
Marmo F, Sessa S, Vaiana N, De Gregorio D, Rosati L (2020) Complete solutions of three-
dimensional problems in transversely isotropic media. Continuum Mech Thermodyn 32. https://
doi.org/10.1007/s00161-018-0733-8
Marmo F, Demartino C, Candela G, Sulpizio C, Briseghella B, Spagnuolo R, Xiao Y, Vanzi I, Rosati
L (2019) On the form of the Musmeci’s bridge over the basento river. Eng Struct 191:658–673.
https://doi.org/10.1016/j.engstruct.2019.04.069
Matuszak A (2017) Errors of stress numerical integration for cross-sections with straight and curved
boundaries. Comput Assisted Methods Eng Sci 22(2):153–176. https://cames.ippt.pan.pl/index.
php/cames/article/view/34
Matuszak A, Pluciński P (2014) Accuracy of cross-section stress numerical integration by boundary
integration formulae, pp 111–120. https://doi.org/10.1201/b16513-16
Menun C (2004a) An envelope for Mohr’s circle in seismically excited three-dimensional structures.
Earthq Eng Struct D 33:981–998
Menun C (2004b) Strategies for identifying critical seismic response combinations. Earthq Spectra
20. https://doi.org/10.1193/1.1806148
Menun C, Der Kiureghian A (2000a) Envelopes for seismic response vectors. I: Theory. J Struct
Eng-ASCE 126. https://doi.org/10.1061/(ASCE)0733-9445(2000)126:4(467)
Menun C, Der Kiureghian A (2000b) Envelopes for seismic response vectors. II: Application. J
Struct Eng-ASCE 126. https://doi.org/10.1061/(ASCE)0733-9445(2000)126:4(474)
Nazarov Y, Poznyak E, Filimonov A (2015) A brief theory and computing of seismic ground
rotations for structural analyses. Soil Dyn Earthq Eng 71:31–41. https://doi.org/10.1016/j.soildyn.
2015.01.013
NBCC (2005) National building code of Canada (NBCC). Standard, National Research Council of
Canada, Ottawa, Canada
Newmark NM (1969) Torsion in symmetrical buildings. Proc 4th World Conf Earthq Eng 3:19–32
NZS 1170, (2004) Structural design actions, Part 5: Earthquake actions. Standard, New Zealand
Papanikolaou V, Sextos A (2016) Design charts for rectangular R/C columns under biaxial bending:
a historical review toward a eurocode-2 compliant update. Eng Struct 115:196–206. https://doi.
org/10.1016/j.engstruct.2016.02.033
10 Mathematical Tools for the Seismic Analysis of Reinforced . . . 211

Paradiso M, Vaiana N, Sessa S, Marmo F, Rosati L (2020) A BEM approach to the evaluation of
warping functions in the saint Venant theory. Eng Anal Boundary Elements 113:359–371. https://
doi.org/10.1016/j.enganabound.2020.01.004
Paradiso M, Sessa S, Vaiana N, Marmo F, Rosati L (2021) Shear properties of isotropic and homo-
geneous beam-like solids having arbitrary cross sections. Int J Solids Struct 216. https://doi.org/
10.1016/j.ijsolstr.2021.01.012
Placidi L, Barchiesi E, Misra A (2018) A strain gradient variational approach to damage: a compar-
ison with damage gradient models and numerical results. Math Mech Complex Syst 6:77–100.
https://doi.org/10.2140/memocs.2018.6.77
Placidi L, Misra A, Barchiesi E (2019) Simulation results for damage with evolving microstructure
and growing strain gradient moduli. Continuum Mech Thermodyn 31. https://doi.org/10.1007/
s00161-018-0693-z
Salman O, Francoisi P, Spagnuolo M (2019) Mean green operators of deformable fiber networks
embedded in a compliant matrix and property estimates. Continuum Mech Thermodyn 31. https://
doi.org/10.1007/s00161-018-0668-0
Selna LG, Lawder JH (1977) Biaxial inelastic frame seismic behavior. Publ SP Am Concr Inst
SP-53. Am Concr Inst Annu Conv, San Francisco, Calif 53:439–461
Sessa S, Marmo F, Rosati L (2015) Effective use of seismic response envelopes for reinforced
concrete structures. Earthq Eng Struct D 44(14):2401–2423
Sessa S, Serpieri R, Rosati L (2017a) A continuum theory of through-the-thickness jacketed shells
for the elasto-plastic analysis of confined composite structures: theory and numerical assessment.
Comp Part B Eng 113. https://doi.org/10.1016/j.compositesb.2017.01.011
Sessa S, Serpieri R, Rosati L (2017b) Probabilistic assessment of historical masonry walls retrofitted
with through-the-thickness confinement devices. AIMETA 2017— Proc 23rd Conf Italian Ass
Th App Mech 3:2324–2332
Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018a) Evaluation of the capacity
surfaces of reinforced concrete sections: Eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512
Serpieri R, Sessa S, Rosati L (2018) A MITC-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Comput
Struct 191:209–220
Sessa S, Marmo F, Vaiana N, Rosati L (2018b) A computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthq Eng 25:1–34. https://
doi.org/10.1080/13632469.2018.1551161
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Comp Part B Eng 163:413–423.
https://doi.org/10.1016/j.compositesb.2019.01.018
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force-biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54. https://doi.org/10.1007/
s11012-019-00979-4
Shakib H (2004) Evaluation of dynamic eccentricity by considering soil-structure interaction: a
proposal for seismic design codes. Soil Dyn Earthq Eng 24:369–378. https://doi.org/10.1016/j.
soildyn.2004.01.003
Silva V, Barros H, Júlio E, Ferreira C (2009) Closed form ultimate strength of multi-rectangle
reinforced concrete sections under axial load and biaxial bending. Comput Concrete 6:505–521.
https://doi.org/10.12989/cac.2009.6.6.505
Simon JW, Weichert D (2012) Shakedown analysis with multidimensional loading spaces. Comput
Mech 49. https://doi.org/10.1007/s00466-011-0656-8
Smeby W, Der Kiureghian A (1985) Modal combination rules for multicomponent earthquake
excitation. Earthq Eng Struct D 13(1):1–12
Spiliopoulos K, Panagiotou K (2017) An enhanced numerical procedure for the shakedown analysis
in multidimensional loading domains. Comput Struct 193:155–171. https://doi.org/10.1016/j.
compstruc.2017.08.008
212 C. Demartino and S. Zhou

Valoroso N, Marmo F, Sessa S (2014) Limit state analysis of reinforced shear walls. Eng Struct
61:127–139. https://doi.org/10.1016/j.engstruct.2013.12.032
Valoroso N, Marmo F, Sessa S (2015) A novel shell element for nonlinear pushover analysis of
reinforced concrete shear walls. Bull Earthq Eng 13:2367–2388. https://doi.org/10.1007/s10518-
015-9724-3
Wilson E, Der Kiureghian A, Bayo E (1981) A replacement for the SRSS method in seismic analysis.
Earthq Eng Struct D 9(2):187–192
Yan Y, Chirikjian G (2015) Closed-form characterization of the Minkowski sum and difference of
two ellipsoids. Geometriae Dedicata 177:103–128
Zuccaro G, Dato F, Cacace F, Gregorio D, Sessa S (2017) Seismic collapse mechanisms analyses
and masonry structures typologies: a possible correlation. Ingegneria Sismica 34:121–149
Chapter 11
Form Finding of Shell Structures
by Using Membrane Theory

Francesco Marmo and Nicoló Vaiana

Abstract A two-step form-finding strategy for determining the heights of a shell of

a given plan-form is presented. The shell heights are determined in such a way that
the structure can equilibrate applied loads by pure membrane actions. To this end, a
set of algebraic equations are obtained by applying the finite difference method to
the differential equations derived from the membrane theory of shells. The two-step
procedure amounts to compute a distribution of projected membrane stresses that
fulfil assigned boundary conditions and to compute shell heights as a function of
the membrane stresses computed at the previous step. In case the loads are known
in advance, the shell heights are computed in closed form, while the procedure is
iterative if loads are expressed as a function of the shell heights. Two numerical
examples show the feasibility of the proposed approach.

Keywords Shell structures · Membrane theory · Form finding · Finite difference

method

11.1 Introduction

The state of stress in shell structures can be described by means of stress resultants
(or internal forces). Their components can be grouped into three categories, i.e.,
membrane (or tangential), out-of-plane shear and bending-twisting stress resultants
(Calladine 1989). When the shell structure is constituted by a uniform material,
membrane stress resultants are associated with an even distribution of stresses along
the transversal direction of the shell, while out-of-plane shear and bending-twisting

F. Marmo (B) · N. Vaiana

Department of Structures for Engineering and Architecture, University
of Naples Federic II, via Claudio 21, 80124 Napoli, Italy
e-mail: [email protected]
N. Vaiana
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 213

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_11
214 F. Marmo and N. Vaiana

internal forces cause higher values of stress near the shell mid-surface and at the
extremities of the shell thickness, respectively (Ventsel et al. 2002).
An interesting design strategy for shell structures is that of fostering suitable
employment of material by determining a structural form such that a specific design
load produces only membrane stress resultants and null out-of-plane shear and
bending-twisting internal forces. Such a design approach, nowadays referred to as
form finding of shells, has a long tradition in the history of structural engineering.
Actually, it finds practical applications for the design of one- and two- dimensional
structural elements having curved geometry and unilateral material constitution, such
as masonry arches and vaults, concrete shells, cable networks and fabric tensile struc-
tures.
Although these concepts were not mature in the classical and mediaeval ages, the
geometry of masonry and concrete arches and vaults was determined by following
simple thumb rules and adopting scale rigid block models with the goal of mitigating
and verifying the admissibility of bending internal forces (Huerta Fernández 2006).
The design of the shape of these structures started having a scientific foundation after
the introduction of the catenary principle (Block et al. 2006). Physical hanging chain
models and static graphic constructions based on the use of the funicular polygon
were employed since the Renaissance throughout the contemporary era. Famous
applications are the networks of hanging strings employed by Huerta (2006) for
designing his breathtaking architectural wonders.
In the mid-twentieth century, the catenary principle was translated to that of a
funicular membrane by, among others, Addis et al. (2013), who employed hanging
cloth models for determining the form of elegant and audacious reinforced concrete
shell roofs, and Marmo et al. (2019a), who used soap film and neoprene sheet models
to design the bearing structure of a 300 m-long bridge in Southern Italy.
During the same years, computational tools for structural analysis were being
developed Clough (2004) and, a few decades later, the same happened for numerical
form-finding methods (Bradshaw 2005).
Among the first numerical form-finding methods is the so-called grid method (Siev
and Eidelman 1964), formulated for the stress analysis cable networks composed of
a grid of rod elements. An alternative approach, based on the Dynamic Relaxation
Method (DRM), solved form-finding problems by analysing the dynamic equilib-
rium of a system of springs and masses (Day and Bunce 1970). Its computational
efficiency is reached by avoiding matrix assemblage and by employing fictitious
damping properties so as to expedite convergence to a steady-state solution.
It was the need of accuracy and efficiency required by the design of the tension
structure of the Olympic Park in Tomlow (2016) that pushed the development of more
effective computational form-finding tools. On this occasion, Argyris and coworkers
(1974) proposed a formulation based on Finite Element Method (FEM) consisting
of static analysis of elastic rods in large displacements. Meanwhile, involved in
the design of the same structure, Linkwitz and Schek (1971; 1974) formulated the
so-called Force Density Method (FDM). Their method computes the nodal coordi-
nates of a network subjected to nodal forces as a function of the ratio between axial
forces and lengths, namely force density, associated with each breach. The method
11 Form Finding of Shell Structures … 215

is extremely efficient for determining suitable structural forms since it is sufficient

to solve a system of linear algebraic equations corresponding to the equilibrium of
nodes. However, the method becomes non-linear when additional constraints are
added in order to control the distance between nodes or assign a specified value of
force in some elements.
In order to deal with the need of assigning non-intuitive quantities such as force
densities, a two-stage solving procedure that uses separately horizontal and vertical
equilibrium equations can be employed. First, the former equations are used to com-
pute a distribution of force densities as a function of the horizontal components of
nodal coordinates, while the latter are employed to evaluate nodal heights as a func-
tion of the force densities computed at the first stage of the analysis. This approach,
namely the Thrust Network Analysis (TNA), was originally proposed for the limit
analysis of masonry vaults (O’Dwyer 1999; Marmo and Rosati 2017; Marmo et al.
2018, 2019b; Fraddosio et al. 2020; Foti et al. 2016; Fraddosio et al. 2019b), and was
successfully employed for form-finding compressed shells (Block and Ochsendorf
2007; Fraddosio et al. 2019a) and used to design modern stone vaults (Rippmann
et al. 2016).
Since pioneering proposals, which were applicable only to cable networks (Greco
et al. 2014), mentioned approaches have further developed by including two-
dimensional elements. These have been employed in FEM-based approaches (Tabar-
rok and Qin 1992), as well as in FDM (Maurin and Motro 1998; Pauletti and Pimenta
2008; Marmo and Rosati 2018), TNA (Marmo and Rosati 2018) and DRM (Barnes
1999, 1988) formulations. Newer formulations combine mentioned approaches with
the tools of the isogeometric analysis (Alic and Persson 2016; Philipp et al. 2016).
Methods that employ two-dimensional elements are based on the membrane the-
ory of shells although the relationship between numerical tools and their theoretical
foundation is seldom described. This theory represents a theoretical framework in
which the height of the shell mid-surface and the membrane resultant components
are related by a unique differential equation (Pucher 1934). Several authors have used
this equation to find the following forms of compressed shells: Musmeci et al. (1977)
employed the Finite Difference Method to solve a simplified version of the basic equa-
tion; Fund (2008) developed an FEM-based solution strategy under the hypothesis
of isotropic, compressive projected membrane resultants; analytical solutions were
proposed by Olivieri et al. (2021) to develop a parametric design of compressed
shells; the procedure proposed by Xia et al. (2019) uses an isogeometric analysis
approach to design shells subjected to predefined projected membrane resultants.
The mentioned methods employ the membrane theory of shells by assigning a
predefined field of projected membrane stress resultants in the form of Airy stress
potential function. In addition to being a limitative choice, the arbitrary assignment
of the working stresses can sometimes hide the violation of trivial equilibrium condi-
tions. For instance, Xia et al. (2019) presented a series of results obtained by assigning
constant projected membrane stresses on the entire shell, even at the free edges of
the structure, where thrust has to be null in order to fulfil equilibrium at the free
boundary.
216 F. Marmo and N. Vaiana

The approach proposed below keeps the internal distribution of membrane stress
resultants unknown. It is computed by assigning only the projected membrane stress
resultants at the boundary of the covered area. These correspond to the design thrusts
that the shell is expected to transmit to its supports. Hence, the user can control these
quantities and assign them as a system of self-equilibrated actions.
After recalling the membrane theory of shells (Sect. 11.2), a form-finding algo-
rithm based on the Finite Difference Method is formulated in Sect. 11.3. Two numer-
ical examples are presented in Sect. 11.4 to illustrate the validity of the proposed
approach. Finally, conclusions and further enhancements of the proposed approach
are drawn in Sect. 11.5.

11.2 The Membrane Theory of Shells

A historic overview of the literature about the membrane theory of shells is reported
in the bibliography of the book by Flügge (2013). However, among the many propos-
als listed therein, the most adapt to solve form-finding problems is the formulation
by Pucher (1934), holding for shells of general shape. According to Pucher’s formu-
lation, the shell is represented by its mid-surface S subjected to distributed loads p,
defined as forces per unit area of the shell, and to internal actions, represented by the
membrane stress (resultants) tensor N. Equilibrium equations for the points of S are
obtained by considering all forces acting on an infinitesimal element of S, which, at
this scale, can be approximated by the plane P tangent to S at the considered point.
These equations express, in differential form, the relationship between N and the
shape of the shell mid-surface.
For the reader’s convenience, the derivation of Pucher’s equilibrium equation is
described in full detail below. However, instead of invoking the Airy stress potential,
as usual in the classical approach, the components of the membrane stress resultants
are kept explicit in the presented formulation. This allows the form-finding algorithm
to control the value of the membrane actions at the boundary of S.

11.2.1 Global and Local Reference Frames

The shell mid-surface S is defined in a Cartesian reference frame of origin O and axes
x, y and z, which have base vectors x = (1, 0, 0), y = (0, 1, 0) and z = (0, 0, 1);
see, e.g., Fig. 11.1. The equation that defines S reads

z = f (x, y) (11.1)

f being a real-valued function of class C 2 . A generic point of S has coordinates

P = (x P , y P , f (x P , y P )), while its projection on the x − y plane is indicated by
P  = (x P , y P , 0). Here and below, projections are intended to take place along z.
11 Form Finding of Shell Structures … 217

Fig. 11.1 Global and local

reference frames

The plane P, tangent to S at P, has the equation

z = f (x P , y P ) + f ,x (x P , y P )(x − x P ) + f ,y (x P , y P )(y − y P ) (11.2)

where the right-hand side corresponds to the first-order Taylor series expansion of
f at P  . Projections of the vectors x and y on P give the vectors ξ = (1, 0, f ,x )
and η = (0, 1, f ,y ), which are tangent to S at P. Their components are obtained
by transforming x and y by means of (11.2). The vectors ξ , η and ζ = ξ × η =
(− f ,x , − f ,y , 1) form the basis of a local reference frame of axes ξ , η and ζ , centred
at P. Notice that being orthogonal to ξ and η, ζ is perpendicular to P and S at P.
However, ξ and η are neither mutually orthogonal nor of unit length. The vector
ζ as well is not a unit vector. Being used to describe quantities pertaining to an
infinitesimal neighbourhood of P, the reference frame (P, ξ, η, ζ ) will be referred
to as local to distinguish it from the orthonormal reference frame (O, x, y, z) that
will be termed global since it will be used to describe an arbitrary quantity referred
to the entire shell mid-surface.
While it is convenient to consider membrane stress components in the local ref-
erence ξ , η and ζ , loads are more easily assigned in the global reference x, y and
z. Additionally, the solution of the set of equilibrium equations is more efficiently
obtained in the global reference frame. Hence, prior to writing the local equilibrium
equations, i.e., the equilibrium of the shell infinitesimal element, it is useful to obtain
the transformation formulas between infinitesimal lengths and areas defined within
the mentioned reference frames.

11.2.2 Transformation Formulas for Lengths and Areas

Infinitesimal increments along x and y have components dx = (dx, 0, 0) and

dy = (0, dy, 0), respectively. Their projection onto P can be obtained by trans-
218 F. Marmo and N. Vaiana

forming dx and dy by means of (11.2), which gives dξ = (dx, 0, f ,x dx) and

dη = (0, dy, f ,y dy), respectively. Their norms
 
dξ = |dξ | = 1 + f ,x2 dx and dη = |dη| = 1 + f ,y2 dy (11.3)

can be used to transform lengths between the local and the global reference frames.
An infinitesimal rectangular region of the x − y plane having sides dx and dy has
area dA = dx dy. Its projection on P corresponds to an infinitesimal parallelogram
of sides dξ and dη, whose area is
 
dA = |dξ × dη| = f ,x2 + f ,y2 + 1 dx dy = f ,x2 + f ,y2 + 1 dA (11.4)

This formula transforms areas between considered reference frames.

Within an infinitesimal neighbourhood of P the shell mid-surface S and its tangent
plane P are indistinguishable. For this reason, transformation formulas for infinites-
imal lengths and areas can equivalently be employed for quantities defined either on
P or S. In particular, this correspondence will be used hereafter for computing the
resultants of distributed loads and membrane stresses.

11.2.3 Distributed Loads and Stress Components

Loads applied to the shell mid-surface are modelled as forces per unit area of the
shell and are indicated as p = ( px , p y , pz ). The resultant of p over an infinitesimal
region of S having area dA is

pd A = p f ,x2 + f ,y2 + 1 dxdy = q dA (11.5)

where 
q=p f ,x2 + f ,y2 + 1 (11.6)

takes the meaning of a force per unit projected area and is referred to as projected
load.
All components of the membrane stress tensor N lie within the plane P. In Fig.
11.2 is represented an infinitesimal neighbourhood I of P of quadrilateral shape with
the opposite sides parallel to ξ and η. These sides have length dξ and dη, respectively,
and are subjected to the corresponding components of the membrane stress tensor
N, i.e., the membrane stresses denoted as Nξ , Nη and Nξ η = Nηξ . Accordingly, N is
represented by the matrix  
Nξ Nξ η
N= (11.7)
Nξ η Nη
11 Form Finding of Shell Structures … 219

Fig. 11.2 Load and

membrane stress resultants

in the local reference.

The component Nξ , for instance, acting along ξ , on the side of length dη, has
resultant Nξ dη. Employing formulas (11.3)2 , its norm is computed as

Nξ dη = Nξ 1 + f ,y2 dy (11.8)

in the global reference frame.

The direction of such a stress resultant is represented by the vector ξ /|ξ |, which
has a unitary norm in the global reference. Hence, according to (11.3), its components
in the global reference frame are given by
⎛  ⎞
1 + f ,y2 1 + f ,y2
Nξ dη ξ /|ξ | = ⎝  Nξ , 0,  Nξ f ,x ⎠ dy = N x , 0, N x f ,x dy
1 + f ,x2 1 + f ,x2
(11.9)
where 
1 + f ,y2
Nx =  Nξ (11.10)
1 + f ,x2

assumes the meaning of a projected membrane stress component for obvious reasons
and represents the first element of the matrix associated with N in the global reference.
The force Nξ dη is expected to have a variation δξ (Nξ dη) when moving to the
opposite side of I; see, e.g., Fig. 11.2. Likewise Nξ dη, such a variation has direction
ξ /|ξ |, so that its components are

δξ (Nξ dη ξ /|ξ |) = N x , 0, N x f ,x dy ,ξ
dξ = N x , 0, N x f ,x
dx dy ,x
(11.11)
where formula (11.3)1 and the chain rule have been applied to express (·),ξ dξ =
(·),x dx. Thus, the components of δξ (Nξ dη) in the global reference frame are finally
computed as
220 F. Marmo and N. Vaiana

N x , 0, N x f ,x dy ,x
dx = (N x,x , 0, N x,x f ,x + N x f ,x x ) dxdy (11.12)

The same approach can be followed for computing the resultants and the variations
of the remaining components of N. Such resultants have global components

Nξ η dη η/|η| = (0, N x y , N x y f ,y ) dy
Nηξ dξ ξ /|ξ | = (N yx , 0, N yx f ,x ) dx (11.13)
Nη dξ η/|η| = (0, N y , N y f ,y ) dx

where the additional projected membrane stresses


1 + f ,x2
Ny =  Nη and N x y = N yx = Nξ η = Nηξ (11.14)
1 + f ,y2

have been introduced. The corresponding stress variations are

δξ (Nξ η dη η/|η|) = (0, N x y,x , N x y,x f ,y + N x y f ,x y ) dxdy

δη (Nηξ dξ ξ /|ξ |) = (N yx,y , 0, N yx,y f ,x + N yx f ,x y ) dxdy (11.15)
δη (Nη dξ η/|η|) = (0, N y,y , N y,y f ,y + N y f ,yy ) dxdy

11.2.4 Equilibrium

Equilibrium equations of I can be obtained by considering all loads, membrane stress

components and relevant variations computed above. Hence, equilibrium along x
yields

qx dxdy − N x dy + N x dy + N x,x dxdy − N yx dx + N yx dx + N yx,y dxdy = 0

(11.16)
Cancelling out opposite addends, employing (11.14)2 to substitute N yx with N x y ,
and dividing by dxdy, one obtains

N x,x + N x y,y + qx = 0 (11.17)

By the same procedure, one can obtain the equilibrium equations along y

N y,y + N x y,x + q y = 0 (11.18)

and z
11 Form Finding of Shell Structures … 221

N x,x f ,x + N x f ,x x + N x y,x f ,y + N x y,y f ,x + 2N x y f ,x y + N y,y f ,y + N y f ,yy + qz = 0

(11.19)
where f ,yx = f ,x y by Shwartz’ theorem.
The previous three equations are combined to derive a unique differential equation
that represents the complete equilibrium of an infinitesimal element of the shell
mid-surface. The classical approach for combining previous equations employs the
analogy between formulas (11.17) and (11.18) and the equilibrium equations of a
plane stress problem. Accordingly, the horizontal equilibrium equations can be solved
by introducing the Airy stress function . Thus, the membrane stress components

N x = ,yy − qx dx N y = ,x x − q y dy N x y = −,x y (11.20)

solve Eqs. (11.17) and (11.18). Use of previous positions into (11.19) yields

,yy f ,x x − 2,x y f ,x y + ,x x f ,yy = qx f ,x + qx dx f ,x x + q y f ,y + q y dy f ,yy − qz

(11.21)
which is the definitive equilibrium equation of the membrane theory of shells.
The previous equation replaces the set of three differential equations (11.17)–
(11.19). For this reason, it is easier to handle when the stress field is unknown.
Actually, all membrane stress components are here substituted by a unique unknown
stress function . Instead, in form-finding analyses, it is preferable to control working
membrane stresses. Hence, it is convenient to consider the original set of differential
equations, which is explicitly expressed as a function of membrane stress compo-
nents.
However, Eq. (11.19) can be rewritten in a simpler form by obtaining N x y,x and
N x y,y from (11.17) and (11.18),

N x y,y = −N x,x − qx N x y,x = −N y,y − q y (11.22)

and then substituting these two derivatives into (11.19) to get

N x f ,x x + 2N x y f ,x y + N y f ,yy − qx f ,x − q y f ,y + qz = 0 (11.23)

which replaces (11.19).

Accordingly, Eqs. (11.17), (11.18) and (11.23) can be used to determine the height
f of a shell mid-surface in such a way that applied loads can be equilibrated by
pure membrane stresses. The problem can be solved in two phases: first, a suitable
distribution of projected membrane stresses is assigned, then the height of the shell
mid-surface is computed as a function of the assigned membrane stresses. In more
complicated situations, projected values of loads are a function of the shell height.
In this case, the previous two phases are iterated.
222 F. Marmo and N. Vaiana

11.3 Form-Finding Algorithm

The proposed form-finding algorithm employs Eqs. (11.17) and (11.18) to define an
equilibrated distribution of projected membrane stresses N x , N y and N x y . However,
these equations are not sufficient to determine a unique distribution of projected
membrane stresses; hence, an additional equation is required. To this end, the pro-
posed methodology makes use of the analogy between Eqs. (11.17) and (11.18) and
the equilibrium equations of a plane stress problem. Accordingly, the compatibility
equation associated with isotropic plane stress problems (Timoshenko and Goodier
1951) is considered. It reads

N x,x x + N x,yy + N y,x x + N y,yy = −(1 − ν)(qx,x + q y,y ) (11.24)

and, together with (11.17) and (11.18), it is used to define a suitable distribution of
projected membrane stresses.
Once projected membrane stresses are defined, the form-finding analysis proceeds
with the evaluation of the value assumed by function f . It is computed from Eq.
(11.23) and relevant boundary conditions.
In more general cases, projected values of loads are not given in advance. This
is the case, for example, in which distributed loads represent the shell self-weight.
In this case, the value of projected loads can be computed only after the height of
shell mid-surface is determined. For this reason, an iterative procedure is required. It
amounts to computing the projected membrane stresses N and the shell mid-surface
height f at each iteration for a tentative distribution of projected loads. Then the
corresponding value of projected loads is updated according to the last estimate of
f . The procedure is iterated until two successive estimates of f differ by less than a
given tolerance.
The phases of the form-finding analysis described above are solved by specific
algorithms that will be illustrated below in full detail. However, these algorithms
amount to solve one or more differential equations corresponding to the projected
equilibrium equation of the membrane theory of shells and the compatibility equation
of isotropic plane stress. Hence, prior to describing each algorithm, the discretization
of such differential equations by the Finite Difference Method is recalled hereafter.

11.3.1 Discretization of the Equilibrium Equations

by the Finite Difference Method

The differential equations (11.17), (11.18), (11.23) and (11.24) are converted into a
set of algebraic equations by employing the finite difference method. Accordingly,
these equations are written for a discrete set of points positioned at the nodes of a
rectangular grid defined on the domain of integration, which lies on the x − y plane.
11 Form Finding of Shell Structures … 223

Fig. 11.3 Grid of nodes on

the x − y plane

For simplicity, grid spacing is assumed constant and equal to x and  y along the
two directions of the reference frame; see, e.g., Fig. 11.3.
The grid is composed of I × J nodes, each having coordinates (xi , y j ) and iden-
tified by the pair of indices (i, j). The value assumed by a generic quantity F at the
node (i, j) is indicated as F (i, j) = F(xi , y j ). Hence, Eqs. (11.17), (11.18), (11.23)
and (11.24) are rewritten for a generic node as
(i, j)
N x,x + N x(i,y,yj) + qx(i, j) = 0 (11.25)

(i, j)
N y,y + N x(i,y,xj) + q y(i, j) = 0 (11.26)

N x(i, j) f ,x(i,x j) + 2N x(i,y j) f ,x(i,y j) + N y(i, j) f ,yy

(i, j)
− qx(i, j) f ,x(i, j) − q y(i, j) f ,y(i, j) + qz(i, j) = 0
(11.27)
and
(i, j) (i, j) (i, j) (i, j) (i, j) (i, j)
N x,x x + N x,yy + N y,x x + N y,yy = −(1 − ν)(q x,x + q y,y ) (11.28)

The differential nature of these equations can be eliminated by considering the

second-order Taylor series expansion of all differentiated quantities. Actually, a
generic function F can be approximated as

(x − xi )2 (i, j)
F(x, y) ≈ F (i, j) + (x − xi )F,x(i, j) + (y − y j )F,y(i, j) + F,x x +
2 (11.29)
(y − y j )2 (i, j)
+ F,yy + (x − xi )(y − y j )F,x(i,y j)
2
This formula can be used to rewrite the derivatives of F at the generic node (i, j) as
a function of the values assumed by F at the 8 nodes that surround the considered
node; see, e.g., Fig. 11.4.
224 F. Marmo and N. Vaiana

Fig. 11.4 Node (i, j) and

surrounding 8 nodes

Setting x = xi + x and y = y j , the previous formula is used to compute an

approximate estimate of F(xi + x , y j ) = F (i+1, j) , that is,

2x (i, j)
F (i+1, j) ≈ F (i, j) + x F,x(i, j) + F (11.30)
2 ,x x
while, setting x = xi − x and y = y j , one has

2x (i, j)
F (i−1, j) ≈ F (i, j) − x F,x(i, j) + F (11.31)
2 ,x x
Subtracting (11.30) from (11.31), one obtains

F (i+1, j) − F (i−1, j)
F (i+1, j) − F (i−1, j) ≈ 2x F,x(i, j) ⇔ F,x(i, j) ≈ (11.32)
2x

while the sum of the same equations yields

(i, j) (i, j) F (i+1, j) + F (i−1, j) − 2 F (i, j)

F (i+1, j) + F (i−1, j) ≈ 2 F (i, j) + 2x F,x x ⇔ F,x x ≈
2x
(11.33)

Similarly, using (11.29) to approximate F(xi , y j +  y ) = F (i, j+1) and F(xi , y j

−  y ) = F (i, j−1) , one has

(i, j) 2y (i, j) (i, j) 2y (i, j)

F (i, j+1) ≈ F (i, j) +  y F,y + F,yy , F (i, j−1) ≈ F (i, j) −  y F,y + F,yy
2 2
(11.34)
11 Form Finding of Shell Structures … 225

which can be subtracted or added to get

F (i, j+1) − F (i, j−1) F (i, j+1) + F (i, j−1) − 2 F (i, j)

F,y(i, j) ≈ , (i, j)
F,yy ≈ (11.35)
2 y 2y

(i, j)
For estimating the value of mixed derivative F,x y , formula (11.29) is used to
compute the four values F(xi ± x , y j ±  y ) = F (i±1, j±1) as

2x (i, j) 2y

F (i+1, j+1) ≈F (i, j) + x F,x(i, j) +  y F,y(i, j) + F + x  y F,x(i,y j) + (i, j)
F,yy
2 ,x x 2
2 2y
F (i−1, j−1) ≈F (i, j) − x F,x(i, j) −  y F,y(i, j) + x F,x(i,x j) + x  y F,x(i,y j) + (i, j)
F,yy
2 2
 2 2y
F (i−1, j+1) ≈F (i, j) − x F,x(i, j) +  y F,y(i, j) + x F,x(i,x j) − x  y F,x(i,y j) + (i, j)
F,yy
2 2
 2 2y
F (i+1, j−1) ≈F (i, j) + x F,x(i, j) −  y F,y(i, j) + x F,x(i,x j) − x  y F,x(i,y j) +
F (i, j)
2 2 ,yy
(11.36)
Subtraction of the last two equations from the sum of the first two furnishes

F (i+1, j+1) + F (i−1, j−1) − F (i−1, j+1) − F (i+1, j−1)

F,x(i,y j) ≈ (11.37)
4x  y

Finally, employing formulas (11.32), (11.33), (11.35) and (11.37), one can rewrite
the differential equations (11.25), (11.26), (11.27) and (11.28) in the form of algebraic
equations as follows:

1 1
N x(i+1, j) − N x(i−1, j) + N x(i,y j+1) − N x(i,y j−1) + qx(i, j) = 0 (11.38)
2x 2 y

1 1
N y(i, j+1) − N y(i, j−1) + N x(i+1,
y
j)
− N x(i−1,
y
j)
+ q y(i, j) = 0 (11.39)
2 y 2x

Nx
(i, j)   N (i, j)  
f (i+1, j) + f (i−1, j) − 2 f (i, j) + f (i, j+1) + f (i, j−1) − 2 f (i, j) +
y
2x y 2

(i, j)  
Nx y
+ f (i+1, j+1) + f (i−1, j−1) − f (i−1, j+1) − f (i+1, j−1) +
2x  y
(i, j)   q (i, j)  
qx (i, j)
f (i+1, j) − f (i−1, j) − f (i, j+1) − f (i, j−1) + qz
y
− =0
2x 2 y
(11.40)
226 F. Marmo and N. Vaiana

and
1
N x(i+1, j) + N y(i+1, j) + N x(i−1, j) + N y(i−1, j) − 2N x(i, j) − 2N y(i, j) +
2x
1
N x(i, j+1) + N y(i, j+1) + N x(i, j−1) + N y(i, j−1) − 2N x(i, j) − 2N y(i, j) +
2y
1 + ν (i+1, j) 1 + ν (i, j+1)
qx − qx(i−1, j) + qy − q y(i, j−1) = 0
2x 2 y
(11.41)

11.3.2 Assigning the Distribution of Projected Membrane

Stresses

The distribution of projected membrane stresses is assigned by solving a system of

three differential equations corresponding to the horizontal equilibrium (11.17) and
(11.18) and the compatibility Eq. (11.24). In order to control thrusts at supports and to
impose null stresses on the free edges of the shell, projected tractions Nn are assigned
on the entire boundary of the region covered by the shell, n being the outward unit
vector associated with the boundary. In particular, it is set as Nn = 0 on the free
boundary, while Nn can be set arbitrarily on the supported boundary. These values
constitute the boundary conditions holding for the mentioned differential equations.
Following the Finite Difference Method, this set of differential equations is trans-
formed into the corresponding algebraic counterpart (11.38), (11.39) and (11.41).
The corresponding set of equations is

Dh = −qh (11.42)

where D is a matrix of coefficients, h is the vector collecting sets of triplets

(N x , N y , N x y ) associated with each node, while qh collects known terms. The vec-
tor h K is split into h K , which collects the Known values of projected membrane
stress, i.e., those assigned through the boundary conditions, and by hU collecting
the Unknown ones. The vector qh and the coefficient matrix D are partitioned corre-
spondingly, so that the previous equation is rewritten as

DUU hU + DU K h K = −qhU
(11.43)
D K U hU + D K K h K = −qh K

The first equation is used to compute the unknown values of projected membrane
stresses as
−1
hU = −DUU (qhU + DU K h K ) (11.44)
11 Form Finding of Shell Structures … 227

Since the second equation is not used, there is no need to assemble equations corre-
sponding to the boundary nodes. Hence, no special treatment is required for equations
relevant to these nodes.

11.3.3 Evaluation of the Shell Mid-Surface Height

After the projected membrane stress components are assigned, the height f of the
shell mid-surface needs to be computed. This is done by solving the set of algebraic
equations corresponding to the vertical equilibrium of each node of the discretized
domain of integration, which is represented by Eq. (11.40). In order to account for
suitable boundary conditions, the nodes of the boundary are grouped into two sets,
respectively, corresponding to i) the nodes where the shell mid-surface has unknown
height, and ii) those where f is given.
This first set of nodes correspond to the free boundary of the shell so that, by
equilibrium, Nn = 0 at these nodes. Accordingly, the vertical equilibrium equation
can be simplified for these nodes. In particular, one has N x = N x y = 0 at nodes
pertaining to the free boundary of outward normal ±x. Similarly, one has N y =
N x y = 0 on the free boundary of outward normal ±y.
For these nodes, a reformulation of the vertical equilibrium equation is also
required since it involves the values that f attains at nodes positioned outside the
domain of integration. To this end, the approximate evaluation of the first derivatives,
given by formulas (11.32)2 and (11.35)1 , are substituted by those obtained by trun-
cating the Taylor series expansions (11.30), (11.31) and (11.34) to the linear term.
Accordingly, for the nodes pertaining to the free boundaries of outward normal +x
and −x, one has

F (i, j) − F (i−1, j) F (i+1, j) − F (i, j)

F,x(i, j) ≈ F,x(i, j) ≈ (11.45)
x x

respectively, while the expressions

F (i, j) − F (i, j−1) F (i, j+1) − F (i, j)

F,y(i, j) ≈ F,y(i, j) ≈ (11.46)
y y

are used, respectively, for the nodes of the free boundaries of outward normal +y
and −y.
Hence, for the nodes of the free boundaries of normal +x, −x, +y and −y, Eq.
(11.40) modifies to
228 F. Marmo and N. Vaiana

(i, j)
Ny
f (i, j+1) + f (i, j−1) − 2 f (i, j) +
2y
(i, j) (i, j)
(11.47)
qx (i, j) (i−1, j) qy (i, j+1) (i, j−1)
− f − f − f − f + qz(i, j) =0
x 2 y

(i, j)
Ny
f (i, j+1) + f (i, j−1) − 2 f (i, j) +
2y
(i, j) (i, j)
(11.48)
qx (i, j) (i−1, j) qy (i, j+1) (i, j−1)
− f − f − f − f + qz(i, j) =0
x 2 y

(i, j)
Nx
f (i+1, j) + f (i−1, j) − 2 f (i, j) +
2x
(i, j) (i, j)
(11.49)
qx (i+1, j) (i−1, j) qy (i, j) (i, j−1)
− f − f − f − f + qz(i, j) =0
2x y

and
(i, j)
Nx
f (i+1, j) + f (i−1, j) − 2 f (i, j) +
2x
(i, j) (i, j)
(11.50)
qx (i+1, j) (i−1, j) qy (i, j+1) (i, j)
− f − f − f − f + qz(i, j) =0
2x y

respectively.
Similarly, at the corner nodes, if pertaining to the free boundary, Eq. (11.40) is
substituted by
(i, j) (i, j)
qx qy
− f (i, j) − f (i−1, j) − f (i, j) − f (i, j−1) + qz(i, j) = 0 (11.51)
x y

for the node positioned at the corner between the boundaries of outward normals +x
and +y, while
(i, j) (i, j)
qx qy
− f (i, j) − f (i−1, j) − f (i, j+1) − f (i, j) + qz(i, j) = 0 (11.52)
x y

(i, j) (i, j)
qx qy
− f (i+1, j) − f (i, j) − f (i, j) − f (i, j−1) + qz(i, j) = 0 (11.53)
x y

(i, j) (i, j)
qx qy
− f (i+1, j) − f (i, j) − f (i, j+1) − f (i, j) + qz(i, j) = 0 (11.54)
x y
11 Form Finding of Shell Structures … 229

are used for the corners (+x, −y), (−x, +y) and (−x, −y), respectively. From
these last four equations, it is clear that, in the case of purely vertical loads, corner
nodes cannot pertain to the free boundary.
Conversely, the vertical equilibrium equations of nodes pertaining to the con-
strained boundary, i.e., where f is given, are not even assembled since they are not
used to solve the form-finding problem. Actually, the full set of equations, comprising
those relevant to both internal and boundary nodes, is symbolically written as

Cf = −qz (11.55)

where C is a matrix of coefficients, while the vectors f and qz collect the nodal values
(i, j)
f (i, j) and qz , respectively.
Vectors f and qz are split into the sub-vectors fU and f K , qzU and qz K by selecting
the elements corresponding to the nodes where f is Unknown, i.e., internal nodes
and those pertaining to the free boundary of the shell, and those where f is Known.
The coefficient matrix C is partitioned into four sub-matrices accordingly. Hence,
the system of vertical equilibrium equations of nodes is rewritten as

CUU fU + CU K f K = −qzU
(11.56)
C K U fU + C K K f K = −qz K

The unknown values of f are computed by solving the first equation as

−1
fU = −CUU (qzU + CU K f K ) (11.57)

while the second equation is not used. Consequently, only the rows of C corre-
sponding to the nodes where f is unknown are needed. Accordingly, the vertical
equilibrium equations of nodes where f is known are not assembled.

11.3.4 Iterative Procedure for Assigning Projected Loads

In many form-finding analyses, the value of projected loads depends on the shell
form. Practical examples are those in which the load represents the shell self-weight
or the weight of a coating distributed on the shell. In such cases, it is convenient to
assign the load as a force per unit area of the shell mid-surface, namely the quantity
p in formula (11.5). Hence, the evaluation of the corresponding projection can be
done only after the height of the shell is evaluated.
Actually, if the load p keeps its direction independently from the form of the shell,
it is sufficient to employ formula (11.5) to evaluate the corresponding projection q.
Of course, this formula can be applied, only after that the shell mid-surface height
f has been computed for each node of the model, which in turn depend on the
assumed values of the projected load. Hence, an iterative procedure is implemented
as described hereafter.
230 F. Marmo and N. Vaiana

At the generic kth iteration, a tentative value for f is estimated for each node of
(i, j)
the model, namely f k . At first iteration, the procedure is initiated by assigning
(i, j)
f 1 = 0 at all nodes.
These values are then used to evaluate nodal values of projected loads by means
of the numerical counterpart of formula (11.5), which is obtained by substituting the
derivatives of f by formulas (11.32) and (11.35)1 , according to the Finite Difference
Method. This gives the kth estimate of the projected loads

 2  2
 (i+1, j) (i−1, j) (i, j+1) (i, j−1)
 fk − fk fk − fk
= p(i, j) 
(i, j)
qk + +1 (11.58)
42x 42y

Recall that the previous formula needs to be suitably corrected for boundary nodes,
where formulas (11.45) and (11.46) are used in place of (11.32) and (11.35)1 for the
evaluation of the numerical derivatives of f .
Current values of projected loads are then used to assign projected membrane
(i, j)
stresses and evaluate a new estimate of nodal heights, that is, f k+1 . These values are
then used at the next iteration to repeat the same sequence of operations.
Convergence is reached when the relative difference between successive values
of applied projected loads is lower than a given tolerance tol, that is,
   
 (i, j) (i, j)   (i, j) 
qk+1 − qk  ≤ tol qk  ∀ i, j (11.59)

11.4 Numerical Examples

In order to show the feasibility of the presented form-finding algorithm, we consider

the case of two fully compressed shells, both covering a rectangular area of sides L x ×
L y = 4 m × 6 m. A grid of 40 × 60 equally spaced nodes is assumed for discretizing
the shell plan-form. The origin of the reference frame is positioned at the centre of
the covered area so that the rectangle is defined by L x /2 ≤ x ≤ L x /2, L y /2 ≤ y ≤
L y /2.
Shells have a constant thickness t = 0.2 m and their material has weight per unit
volume d = 24 k N /m 3 . Accordingly, assuming that the shell is subjected to its own
weight, the vertical load applied to the shell’s surface amounts to pz = −d × t =
−4.8 k N /m 2 .
The difference between the two shells is at their boundaries. The first is supported
on three sides, while the boundary of outward normal +y is free. The second is
supported only at corners, while central parts of the four sides are free.
11 Form Finding of Shell Structures … 231

11.4.1 Shell with One Free Side

A distribution of boundary thrusts that satisfies the global equilibrium of the shell
is assigned. In particular, null thrusts are assigned to the nodes of the free boundary
while compressive thrusts are assigned to the sides of outward normal ±x and −y,
as follows:    
A Ly B 4y 2
Side −x: N x = y+ , Nx y = − 1− 2
3 2 2 Ly
   
A Ly B 4y 2
Side +x: N x = y+ , Nx y = 1− 2
3 2 2 Ly
 2
Ly 4x
Side −y: N y = B 1 − 2 , Nx y = 0
Lx Lx
Side +y: N y = 0, N x y = 0
Numerical parameters A and B have been chosen so as to have a semi-negative
definite (compressive) projected membrane resultant tensor N on the entire shell
plan-form. This assignment is done iteratively, by computing the projected membrane
resultants by means of formula (11.44) for some given values of A and B and then
correcting these tentative values in order to have compressive principal projected
stress resultants on the entire domain. This condition is fulfilled when A = 2B =
2 pz . The corresponding principal values and directions of the projected membrane
stresses are represented in Fig. 11.5. Segments drawn in this figure are parallel
to the principal directions of N, and lengths are proportional to the corresponding
principal values. Red and blue colours are used for negative and positive principal
values, respectively, and, as shown, no tensile principal stress resultants have been
computed.
This distribution of projected stress resultants is then used to compute nodal
heights according to the iterative procedure described in Sect. 11.3.4. Assuming a
tolerance tol = 10−3 , convergence is reached after 6 iterations and the corresponding
shell form is reported in Fig. 11.6, where an isometry and two side views of the
solution are drawn.

11.4.2 Shell Supported at Corners

A shell supported at corners has been selected as a second example. Only a small
region of each side of the covered area is supported, that is, a portion  = 0.25 m
wide, next to each corner. Boundary horizontal thrusts are assigned as a function of
the distance δ between the considered node and corners. Accordingly, for the nodes
near the south-west corner, the following thrusts are applied:
SW corner (δ ≤ ), Side −x: N y = −1.25 C( − δ), N x y = −C( − δ)
SW corner (δ ≤ ), Side −y: N x = −C( − δ), N x y = −C( − δ)
232 F. Marmo and N. Vaiana

Fig. 11.5 Principal values and directions of the projected stress resultants for the shell with one
free side

Fig. 11.6 Isometry and side views of the form-found shell with one free side
11 Form Finding of Shell Structures … 233

Fig. 11.7 Principal values and directions of the projected stress resultants for the shell supported
at corners

Similar boundary conditions, modified in sign in order to guarantee compressive

thrusts, are assigned at nodes pertaining to the portion of the boundary near the other
corners of the covered area. The numerical parameter C is assigned in such a way that
the maximum height of the form-found shell is slightly below 1.5 m. Accordingly, C
has been assumed equal to C = 90 pz . The corresponding distribution of projected
stress resultants is reported in Fig. 11.7.
The computed distribution of N generates the nodal heights depicted in Fig. 11.8
when the iterative procedure described in Sect. 11.3.4 is employed. This form has
been obtained after 6 iterations by setting a tolerance tol = 10−3 .
234 F. Marmo and N. Vaiana

Fig. 11.8 Isometric and side views of the form-found shell supported at corners

11.5 Conclusion

A form-finding algorithm based on the membrane theory of shells has been presented.
The numerical implementation of the equilibrium equations is obtained by the Finite
Difference Method. In order to avoid any predefined distribution of projected mem-
brane stresses, an additional equation corresponding to the compatibility condition
of plane stress problems has been employed. The proposed approach employs an
iterative procedure in order to adjust the initially unknown distribution of projected
load components. Two numerical examples show the feasibility of the presented
method which is capable to find the form of compressed shells having a rectangular
plan.
Further enhancements of the presented approach will regard the following issues:
(i) The compatibility equation used by the presented form-finding procedure is
relevant to plane stress elasticity problems defined on the horizontal plane, while
the stress state is plane on the shell mid-surface. Hence, the feasibility of such an
equation needs to be further investigated. In this regards, it is useful to adopt the
approach followed in (Flügge and Geyling 1957) to derive a more suitable equation
that will substitute the one employed in the presented approach.
(ii) A more general optimization of the shell form can be achieved by introducing
an additional unknown function representing the shell thickness. Consequently, a
modification of the presented iterative procedure is required since distributed loads
representing the shell self-weight need to be modified accordingly.
(iii) The employment of the Finite Difference Method to discretize the basic
differential equations restricts the field of application of the presented method to
shells having a rectangular plan. In this regard, improved implementations of the
Finite Difference Method have been presented in the literature; see, e.g., Liszka and
Orkisz (1980), which expand the field of applicability of the proposed approach.
11 Form Finding of Shell Structures … 235

(iv) The presented procedure can be used to define the form of the shell mid-
surface, but proper algorithms are required for a comprehensive design of this kind of
structures. In particular, for reinforced concrete shells, proper modelling techniques
are required to account for the complex (Altenbach and Eremeyev 2009; Altenbach
et al. 2010; Tran et al. 2020; De Angelo et al. 2019; Andreaus et al. 2018) and the
non-linear (Contrafatto and Cuomo 2006; Valoroso et al. 2014, 2015; Placidi et al.
2018) behaviour of materials, transverse interaction (Sessa et al. 2017; Serpieri et al.
2018; Sessa et al. 2019; Abd-alla et al. 2017), surface stresses (Altenbach et al.
2009) and finite deformations (Eremeyev and Zubov 2007; Barchiesi et al. 2020a, b;
Franciosi et al. 2019).

References

AenN Abd-alla, Alshaikh F, Del Vescovo D, Spagnuolo M (2017) Plane waves and eigenfrequency
study in a transversely isotropic magneto-thermoelastic medium under the effect of a constant
angular velocity. J Therm Stress 40(9):1079–1092
Addis B et al (2013) Toys that save millions: a history of using physical models in the design of
building structures. SESOC J 26(2):28
Alic V, Persson K (2016) Form finding with dynamic relaxation and isogeometric membrane ele-
ments. Comput Methods Appl Mech Eng 300:734–747
Altenbach H, Eremeyev VA (2009) On the linear theory of micropolar plates. ZAMM-J Appl Math
Mech/Zeitschrift für Angewandte Mathematik und Mechanik 89(4):242–256
Altenbach H, Eremeyev V, Morozov N (2009) Linear theory of shells taking into account surface
stresses. Doklady Phys 54:531–535
Altenbach J, Altenbach H, Eremeyev VA (2010) On generalized cosserat-type theories of plates
and shells: a short review and bibliography. Archive Appl Mech 80(1):73–92
Andreaus U, Spagnuolo M, Lekszycki T, Eugster SR (2018) A ritz approach for the static analysis of
planar pantographic structures modeled with nonlinear euler-bernoulli beams. Continuum Mech
Thermodyn 30(5):1103–1123
Argyris JH, Angelopoulos T, Bichat B (1974) A general method for the shape finding of lightweight
tension structures. Comput Methods Appl Mech Eng 3(1):135–149
Barchiesi E, dell’Isola F, Hild F, Seppecher P (2020a) Two-dimensional continua capable of large
elastic extension in two independent directions: asymptotic homogenization, numerical simula-
tions and experimental evidence. Mech Res Commun 103(103):466
Barchiesi E, Eugster SR, Dell’isola F, Hild F (2020b) Large in-plane elastic deformations of bi-
pantographic fabrics: asymptotic homogenization and experimental validation. Math Mech Solids
25(3):739–767
Barnes MR (1988) Form-finding and analysis of prestressed nets and membranes. Comput Struct
30(3):685–695
Barnes MR (1999) Form finding and analysis of tension structures by dynamic relaxation. Int J
Space Struct 14(2):89–104
Block P, Ochsendorf J (2007) Thrust network analysis: a new methodology for three-dimensional
equilibrium. J Int Assoc Shell Spatial Struct 48(3):167–173
Block P, DeJong M, Ochsendorf J (2006) As hangs the flexible line: equilibrium of masonry arches.
Nexus Netw J 8(2):13–24
Bradshaw RR (2005) History of the analysis of cable net structures. In: Structures congress 2005:
metropolis and beyond, pp 1–11
Calladine CR (1989) Theory of shell structures. Cambridge University Press
236 F. Marmo and N. Vaiana

Clough RW (2004) Early history of the finite element method from the view point of a pioneer. Int
J Numer Methods Eng 60(1):283–287
Contrafatto L, Cuomo M (2006) A framework of elastic-plastic damaging model for concrete under
multiaxial stress states. Int J Plastic 22(12):2272–2300
Day A, Bunce J (1970) Analysis of cable networks by dynamic relaxation. Civil Eng Public Works
Rev 4:383–386
De Angelo M, Spagnuolo M, D’annibale F, Pfaff A, Hoschke K, Misra A, Dupuy C, Peyre P,
Dirrenberger J, Pawlikowski M, (2019) The macroscopic behavior of pantographic sheets depends
mainly on their microstructure: experimental evidence and qualitative analysis of damage in
metallic specimens. Continuum Mech Thermodyn 31(4):1181–1203
Eremeyev V, Zubov L (2007) On constitutive inequalities in nonlinear theory of elastic shells.
ZAMM-J Appl Math Mech/Zeitschrift für Angewandte Mathematik und Mechanik: Applied
Mathematics and Mechanics 87(2):94–101
Flügge W (2013) Stresses in shells. Springer
Flügge W, Geyling F (1957) A general theory of deformations of membrane shells. Int Assoc Bridg
Struct Eng 17:23
Foti P, Fraddosio A, Lepore N, Piccioni MD (2016) Three-dimensional lower-bound analysis of
masonry structures. In: Proceeding of the third international conference on structures and archi-
tecture, Guimaraes, Portugal
Fraddosio A, Lepore N, Piccioni MD (2019a) Further refinement of the corbelling theory for the
equilibrium analysis of corbelled domes. Curved Layer Struct 6(1):30–40
Fraddosio A, Lepore N, Piccioni MD (2019b) Lower bound limit analysis of masonry vaults under
general load conditions. In: Structural analysis of historical constructions, Springer, pp 1090–
1098
Fraddosio A, Lepore N, Piccioni MD (2020) Thrust surface method: An innovative approach for
the three-dimensional lower bound limit analysis of masonry vaults. Eng Struct 202(109):846
Franciosi P, Spagnuolo M, Salman OU (2019) Mean green operators of deformable fiber networks
embedded in a compliant matrix and property estimates. Continuum Mech Thermodyn 31(1):101–
132
Fund AI (2008) Form-finding structures. PhD thesis, Massachusetts Institute of Technology
Greco L, Impollonia N, Cuomo M (2014) A procedure for the static analysis of cable structures
following elastic catenary theory. Int J Solids Struct 51(7–8):1521–1533
Huerta S (2006) Structural design in the work of Gaudi. Architec Sci Rev 49(4):324–339
Huerta Fernández S (2006) Geometry and equilibrium: the gothic theory of structural design. Struct
Eng 84(2):23–28
Linkwitz K, Schek HJ (1971) Einige bemerkungen zur berechnung von vorgespannten seilnetzkon-
struktionen. Ingenieur-Archiv 40(3):145–158
Liszka T, Orkisz J (1980) The finite difference method at arbitrary irregular grids and its application
in applied mechanics. Comput Struct 11(1–2):83–95
Marmo F, Rosati L (2017) Reformulation and extension of the thrust network analysis. Comput
Struct 182:104–118
Marmo F, Rosati L (2018) Form finding of compressed shells by the thrust membrane analysis. In:
Proceedings of IASS annual symposia, international association for shell and spatial structures
(IASS), vol 2018, pp 1–6
Marmo F, Masi D, Rosati L (2018) Thrust network analysis of masonry helical staircases. Int J
Architec Heritage 12(5):828–848
Marmo F, Demartino C, Candela G, Sulpizio C, Briseghella B, Spagnuolo R, Xiao Y, Vanzi I, Rosati
L (2019a) On the form of the musmeci’s bridge over the basento river. Eng Struct 191:658–673
Marmo F, Ruggieri N, Toraldo F, Rosati L (2019b) Historical study and static assessment of an
innovative vaulting technique of the 19th century. Int J Archit Heritage 13(6):799–819
Maurin B, Motro R (1998) The surface stress density method as a form-finding tool for tensile
membranes. Eng Struct 20(8):712–719
Musmeci S, et al (1977) Ponte sul basento a patenza. L’industria Italiana del Cemento
11 Form Finding of Shell Structures … 237

O’Dwyer D (1999) Funicular analysis of masonry vaults. Comput Struct 73(1–5):187–197

Olivieri C, Angelillo M, Gesualdo A, Iannuzzo A, Fortunato A (2021) Parametric design of purely
compressed shells. Mech Mater 155(103):782
Pauletti RM, Pimenta PM (2008) The natural force density method for the shape finding of taut
structures. Comput Methods Appl Mech Eng 197(49–50):4419–4428
Philipp B, Breitenberger M, D’Auria I, Wüchner R, Bletzinger KU (2016) Integrated design and
analysis of structural membranes using the isogeometric b-rep analysis. Comput Method Appl
Mech Eng 303:312–340
Placidi L, Misra A, Barchiesi E (2018) Two-dimensional strain gradient damage modeling: a vari-
ational approach. Zeitschrift für angewandte Mathematik und Physik 69(3):1–19
Pucher A (1934) Über den spannungszustand in gekrümmten flächen. Beton und Eisen 33:298
Rippmann M, Van Mele T, Popescu M, Augustynowicz E, Méndez Echenagucia T, Calvo Barentin
CJ, Frick U, Block P (2016) The armadillo vault: Computational design and digital fabrication
of a freeform stone shell. Adv Architec Geometry 2016:344–363
Schek HJ (1974) The force density method for form finding and computation of general networks.
Comput Methods Appl Mech Eng 3(1):115–134
Serpieri R, Sessa S, Rosati L (2018) A mitc-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through-the-thickness jacketed shells for
the elasto-plastic analysis of confined composite structures: Theory and numerical assessment.
Compos Part B Eng 113:225–242
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos Part B Eng 163:413–423
Siev A, Eidelman J (1964) Stress analysis of prestressed suspended roofs. J Struct Div 90(4):103–
121
Tabarrok B, Qin Z (1992) Nonlinear analysis of tension structures. Comput Struct 45(5–6):973–984
Timoshenko S, Goodier J (1951) Theory of elasticity, 1951. New York 412:108
Tomlow J (2016) Designing and constructing the Olympic roof (Munich 1972). Int J Space Struct
31(1):62–73
Tran CA, Gołaszewski M, Barchiesi E (2020) Symmetric-in-plane compression of polyamide pan-
tographic fabrics-modelling, experiments and numerical exploration. Symmetry 12(5):693
Valoroso N, Marmo F, Sessa S (2014) Limit state analysis of reinforced shear walls. Eng Sruct
61:127–139
Valoroso N, Marmo F, Sessa S (2015) A novel shell element for nonlinear pushover analysis of
reinforced concrete shear walls. Bull Earthq Eng 13(8):2367–2388
Ventsel E, Krauthammer T, Carrera E (2002) Thin plates and shells: theory, analysis, and applica-
tions. Appl Mech Rev 55(4):B72–B73
Xia Y, Mantzaflaris A, Jüttler B, Pan H, Hu P, Wang W (2019) Design of self-supporting surfaces
with isogeometric analysis. Comput Method Appl Mech Eng 353:328–347
Chapter 12
Influence of Non-structural Components
on Equivalent Linearization of Buildings

Salvatore Sessa and Luciano Rosati

Abstract Influence of non-structural elements on the seismic response of nonlin-

ear systems is hereby investigated. In particular, tail-equivalent linearization has
been adopted because of its capability of determining the statistics of stochastic
response processes in order to characterize secondary excitation. A random vibra-
tion analysis determined a parameter range, characterizing the dynamic properties of
non-structural components, for which the linearized system is not influenced by the
presence of such devices. Numerical results show that tail equivalent linearization is
an appealing strategy for the characterization of secondary seismic excitations.

Keywords Equivalent linearization · Seismic analysis · Nonlinear systems ·

Tail-equivalent linearization · Stochastic response · Vibration analysis

12.1 Introduction

Seismic events recently occurred worldwide have drawn attention on the influence
of non-structural devices on the seismic response of nonlinear buildings. To this end,
such secondary devices have been classified in several typologies and classes (FEMA
1994) in order to evaluate secondary seismic actions by means of dynamic filters.
Moreover, such a topic is of particular importance in assessing structural resilience
(Chiaia et al. 2019) as well as for the case of seismic protection of cultural heritage
artifacts exposed in museums (Calio and Marletta 2003) or of archaeological vestiges.
In the latter cases, several elements, such as balustrades, roofs, or ruins, can be not
part of the main structure although they be an essential part to be preserved by proper
analyses adopting secondary excitations. At the same time, presence of such devices

S. Sessa (B) · L. Rosati

University of Naples Federico II, via Claudio 21, Napoli, Italy
e-mail: [email protected]
L. Rosati
e-mail: [email protected]

© Springer Nature Switzerland AG 2021 239

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_12
240 S. Sessa and L. Rosati

may strongly influence the dynamic response of the main structure and, subsequently,
the secondary excitation.
Given the importance of properly account for nonlinearity of structures subject to
dynamic loads, several approaches have been proposed in the last decades including
variational strategies (Barchiesi and Khakalo 2019) and Lagrangian methods (Turco
et al. 2019, 2020; Andreaus et al. 2018).
Within the context of equivalent linearization approaches, the Tail-Equivalent
Linearization Method, proposed by Fujimura and Der Kiureghian (2007), is a partic-
ularly effective strategy which defines a linearized system by means of a collection
of response functions in the time or frequency domain (Garrè and Der Kiureghian
2010).
Compared to traditional linearization strategies, such an approach proved to be
more accurate (Kiureghian and Fujimura 2009) and is capable to address mul-
ticomponent (Broccardo and Der Kiureghian 2016) and multisupport (Wang and
Der Kiureghian 2016) seismic actions as well as multi-objective responses (Sessa
2010). More in general, a representation of structural responses in frequency and
time domain is a very powerful technique which can be also used to characterize
even more complex problems (Alibert et al. 2003; Barchiesi et al. 2018; Dell’Isola
et al. 2019).
Since the dynamic response of a nonlinear system can be significantly adulterated
by non-structural devices (Badillo-Almaraz et al. 2007; Milton et al. 2009; Villaverde
1997), the definition of the linearized system should properly account for their pres-
ence within the mechanical system. Unfortunately, common procedures of structural
design are not sufficiently detailed to model non-structural parts which are merely
introduced as additional masses and loads. On the contrary, a proper computation
of the linearized system and of secondary excitations must properly account for the
dynamic behavior of secondary devices.
The present research investigates the influence of non-structural devices on the
tail-equivalent linearized system of nonlinear buildings by introducing a parametrized
secondary oscillator in the structural system.
Specifically, a brief review of the TELM is reported in Sect. 12.2 while Sect. 12.3
presents the computation of the linearized system and Sect. 12.3.1 discusses the
response sensitivity with respect to the mass and stiffness properties of the secondary
devices.
Moreover, statistics of the random process representing the peak response is com-
puted in Sect. 12.3.2 by means of first excursion probability in order to determine
boundary intervals of the properties of non-structural devices for which their influ-
ence on the structural response is negligible.
Numerical results prove that such an interval includes several typologies of sec-
ondary devices so that they can be neglected in performing tail equivalent analysis.
12 Influence of Non-structural Components on Equivalent Linearization of Buildings 241

12.2 Brief Review of Tail-Equivalent Linearization

Conceptually, the Tail-Equivalent Linearization Method (TELM) determines a lin-

earized system by equating the tail probability of its response with the first-order
approximation of the tail probability relevant to the nonlinear structure for a fixed
threshold.
Let F be a discretized representation of a stochastic excitation:


n
F(t) = si (t)u i = s(t)u (12.1)
i=1

where si (t) are deterministic functions depending on the covariance of the seismic
random process and  u i are n random variables defined in the standard normal space.
Denoted by X tˆ the response of the nonlinear structure, its tail probability rele-
vant to a threshold of interest x  and to an instant-in-time tˆ is defined by
  
Pr X tˆ ≤ x 

and its first-order approximation is associated with the solution u (or performance
point) of the optimization problem:
     
u tˆ, x  = arg min u x  − X tˆ = 0 (12.2)

which can be computed by the First-Order Reliability Method (FORM) (Ditlevsen

and Madsen 1996). 
The dynamic response X l tˆ of a linear oscillator can be computed by means of
a convolution integral:
 tˆ 
 n

X l tˆ = h(τ ) si (τ )u i dτ = a tˆ · u (12.3)
0 i=1

where h(τ ) represents the Impulse Response Function (IRF) of a linear system.
Enforcing an equivalence condition between the linear and nonlinear
  response at
the performance point permits to compute a deterministic vector a j tˆ as
 
 x u x  , tˆ
a tˆ =     (12.4)
u x  , tˆ u x  , tˆ

whose components are associated with the IRF of the linear system:

  n
   
a j tˆ = h tˆ − τ j si τ j t (12.5)
j=1
242 S. Sessa and L. Rosati
 
Therefore, the impulse response function ĥ t, tˆ, x  of a linearized system asso-
ciated with the threshold x  and at time tˆ can be computed by inverting Equation
(12.5) which represents a linear system of equations.
It is worth being emphasized that the explicit dependency of the IRF on the
threshold and time is due to the fact that, for different values of x  and tˆ, the nonlinear
structure provides a distinct response and subsequently it is associated with a different
linearized system.
A Tail-Equivalent Linearized System can be defined by computing the IRFs asso-
ciated with an array of thresholds x  k . Such functions proved to be insensitive to
amplitude and frequency content of the base excitation (Fujimura and Der Kiureghian
2007).
In order to use the TELS within the context of random vibration analysis, it is
useful to define the Frequency Response Functions (FRFs) as the Furier transforms
of the IRFs:  ∞
 
  
H f, tˆ, x k = ĥ τ, tˆ, x  k exp (−iωτ ) dτ (12.6)
0

where ω = 2π f denotes the pulsation and f is the frequency.

The stationary response X  of the linearized
 system can be characterized by its
power spectral density  X X f, tˆ, x  k computed as
     2
 X X f, tˆ, x  k =  F F f, tˆ, x  k  H f, tˆ, x  k  (12.7)

where  F F is the power spectral density of the base excitation.

12.3 Influence of Secondary Devices on TELS

In order to investigate the influence of non-structural devices on the response of the

global structural system, a case study consisting of a 4-storey building is defined. It
has been analyzed in OpenSees, a finite-element object-oriented framework (Maz-
zoni et al. 2006), capable of performing TELM analyses.
According to Fujimura and Der Kiureghian (2007), the system has been modeled
by defining the interaction between adjacent floors by means of a Bouc-Wen nonlinear
relationship (Baber and Noori 1985). Specifically, the inter-story shear force R(t) is
defined by
R(t) = αk0 υ(t) + (1 − α) k0 z(t) (12.8)

where υ(t) is the inter-story drift, k0 is the initial stiffness, α is the yield ratio, and
z(t) is a nonlinear inter-story drift whose first time-derivative is
 
ż(t) = υ̇(t) A0 − βsign (υ̇(t)) |z(t)|n−1 z(t) + γ |z(t)|n (12.9)
12 Influence of Non-structural Components on Equivalent Linearization of Buildings 243

(a) Unperturbed system (b) Parameterized system

Fig. 12.1 Structural schemes of the unperturbed and parameterized structural systems

Each floor has mass m 0 = 3 · 105 kg; constitutive parameters have been assumed
α = 0.05, A0 = 2, n = 1. Stiffness k0 is computed so that a linear oscillator with mass
m 0 has natural frequency of 5 Hz and finally γ = β = 1/(2υ y ) where υ y = 0.0182 m
is the equivalent yield inter-story drift.
Figure 12.1a shows the model representing the main structure where the structural
response of interest X (t) is the horizontal displacement at the top floor. Note that in
such a model no secondary device has been introduced; thus, it will be referred to as
unperturbed system.
The perturbed system, shown in Fig. 12.1b, is defined by introducing in the unper-
turbed model a linear oscillator representing the non-structural devices. Such a
Parametrized Oscillator (PO) is constrained at the top floor and is defined by means
of mass m s and frequency f s so that its stiffness turns out to be ks = (2π f s )2 m s .
Tail-Equivalent linearization is performed for both the structural models in order to
compare the linearized systems associated with the perturbed and unperturbed struc-
ture. To this end, a set of parameters characterizing the secondary device have been
adopted. In particular, the oscillator mass m s spans within the interval [0.01m 0 , m 0 ]
while its natural frequency belongs to the interval [0.5, 30.0] Hz plus the further
limit case of f s = ∞.
Linearized systems have been determined by means of the frequency response
functions HU ( f, x  ) and H P ( f, x  , m s , f s ) relevant to the unperturbed and pa-
rametrized structure, respectively. For the latter case, dependency of the FRFs on m s
and f s has been explicitly reported.

12.3.1 Frequency Content Comparison

A first comparison between the parametrized and unperturbed system can be made
in terms of frequency response functions in order to investigate the influence of the
PO on the frequency content of the system.
244 S. Sessa and L. Rosati

Fig. 12.2 TELS frequency response functions depending on the threshold x 

Fig. 12.3 TELS frequency response functions at threshold x  = 45.45 mm depending on the PO
mass m s

Figure 12.2a reports the FRFs of the unperturbed system associated with thresh-
olds x  . The dominant structural frequency occurs at about 2.5 Hz; moreover, both the
peak amplitude and the dominant frequency of the FRFs are inversely proportional
to the threshold value.
Frequency response functions relevant to the parametrized system with f s = ∞
and m s /m 0 = 1 are reported in Fig. 12.3b. Such a model represents a limit case since
the mass of the secondary device is equal to the floor mass and the PO is infinitely
stiff meaning that its mass is perfectly constrained to the top floor. In particular,
such modeling assumptions correspond to a structural design which introduces the
secondary devices in the model merely as masses and loads.
Functions reported in Fig. 12.2b show that the peaks are significantly higher that
the corresponding ones of the unperturbed model and occur at lower frequencies
12 Influence of Non-structural Components on Equivalent Linearization of Buildings 245

Fig. 12.4 TELS frequency response functions at threshold x  = 45.45 mm depending on the PO
frequency f s

meaning that the model is stiffer. As a consequence, a random vibration analysis

would compute very different response statistics for the considered models.
In general, such a limit case is not frequent in common practice; hence, it is
essential to investigate the dependency of the structural response with respect to
the mass m s of the secondary device. To this end, Fig. 12.3a and b report the FRFs
at threshold x  = 45.45 mm associated with frequencies f s = 2 Hz and f s = 5 Hz,
respectively.
In both the cases, amplitude of the FRFs associated with low mass rations remains
almost unaltered meaning that the influence of the secondary device is limited. More-
over, higher values of m s turn out to significantly influence the frequency content of
the model since the FRFs present a further peak nearby the natural frequency of the
PO.
Such an issue is significant for the case of f s = 2 Hz; in particular, for m s /m 0 >
0.1 FRFs present multi-mode behavior and peaks significantly shifted from their
unperturbed location. On the contrary, influence of the PO on the functions relevant
to f s = 5 Hz, reported in Fig. 12.3b, turns out to be less significant.
The different sensitivity of such FRFs is due to the fact that the value f s = 2 Hz
is very close to the dominant frequency of the main structure so that the system
presents resonance phenomena. On the contrary, the value f s = 5 Hz implies that
the PO behaves as a sort of mass damper reducing the frequency of the system.
Frequency response functions reported in Fig. 12.4a and b, relevant to sec-
ondary oscillators with varying natural frequency and mass ratios m s /m 0 = 0.1 and
m s /m 0 = 1, respectively, confirm the qualitative behavior deduced by comparing
Fig. 12.3a and b.
In particular, Fig. 12.4a shows that the effects of oscillators with limited mass
on the structural response are significant only if the dominant frequencies of the
secondary oscillator and of the main structure are similar. Specifically, for f s = 2 Hz
246 S. Sessa and L. Rosati

and f s = 3 Hz, resonance is triggered while for different values of f s the curves
present limited differences from the unperturbed case.
A further aspect can be observed by analyzing curves relevant to m s /m 0 = 1
reported in Fig. 12.4b. Structural frequency response is strongly adulterated by the
presence of a significant mass. For f s ≥ 3 Hz, although peaks move to the lower
frequencies and increase in amplitude, the shape of the function remains almost
unaltered. Such an issue is due to the fact that a stiff secondary oscillator results
in a mere increment of the structural mass and is even more visible for frequencies
of f s = 5 Hz and f s = 15 Hz which progressively tend to the response relevant to
f s = ∞.
On the contrary, for low values of f s , the response is very close to the unperturbed
one although it has a further peak at f = f s . Hence, the secondary oscillator induces a
narrow band response uncorrelated to the one of the main structure. This is confirmed
by observing curves relevant to f s = 0.5 Hz and f s = 1.5 Hz.

12.3.2 First Excursion Probability Comparison

In order to investigate the influence of the non-structural components on the seismic

vulnerability of the main structure, it is convenient to determine the First Excursion
Probability (FEP) over a time interval. Such a quantity is defined as the probability
 the response of interest X (t) crosses a fixed bound x during the time interval
that
0, t f p and is defined as
      
FX x, t f p = Pr max |X (x, τ )| ; ∀τ ∈ 0, t f p ≥ x (12.10)

where a double-bound ±x is assumed.

Referring to the original paper (Kiureghian and Fujimura 2009) for the details,
TELM takes advantage of the linearization by determining the FEP relevant to each
threshold by means of Vanmarcke’s extreme peak distribution:
⎡  √ ⎤

 +  1 − exp − 2δ x
1.2
  x2
FX x, t f p = 1 − exp − exp ⎣−ν X 0 t f p   ⎦
2 exp 0.5x 2 − 1
(12.11)
where ν X denotes the mean up-crossing rate and δ is the regularity factor. Denoting
by  X X (ω) the power spectral density of the structural response X (t), δ and ν X can
be computed as
   
 ∞
λ21   1 λ2 x2
λm = 2ωm  X X (ω) dω; δ = 1− ; νX x +, t f p = exp −
0 λ0 λ2 2π λ0 2λ0
(12.12)
in which λm is the mth spectral moment, x + denotes the up-crossing of threshold x,
and ω = 2π f is the pulsation.
12 Influence of Non-structural Components on Equivalent Linearization of Buildings 247

Statistics of the response relevant to the unperturbed and parametrized systems

have been computed for a stationary excitation consisting of a filtered white noise
with intensity σ0 = 0.0448 m/s2 filtered by a linear oscillator with natural frequency
of f F F = 1.5 Hz and damping ζ F F = 0.3.
Moreover, a comparison between the response statistics is performed by means
of the rate  and its envelope ˆ defined as:
    
 
  FXp x  , t f p , f s , m s − FXu x  , t f p 
 t f p , x , fs , m s =     (12.13)
 FXu x  , t f p 
   
ˆ t f p , x  = max  t f p , x  , f s , m s | ∀ f s ∈ [0.5, 30] H z, ∀ (m s /m 0 ) ∈ [0.01, 1]
 (12.14)
  p 
where FXu x, t f p and FX x, t f p , f s , m s are the FEP of the unperturbed and
parametrized system, respectively.
Colormaps of FEP-rates relevant to six secondary oscillators and depending by
threshold x  and time interval t f p are reported in Fig. 12.5a–f. Note
 that white regions
are associated with negligible error values  t f p , x  , f s , m s ≤ 10−4 .
It is worth being emphasized that FEP-rate is higher for short time intervals and
high thresholds. Specifically, for x  < 0.055 m the first excursion probabilities of the
unperturbed and parametrized systems tend to coincide if sufficiently lasting time
intervals are adopted.
Such a phenomenon is due to the transient phase of the extreme peak process. In
fact, because of stationarity, at high thresholds (i.e., x  > 0.055 m) the rates present
a constant trend independently by the assumed t f p . This physically means that both
systems reached stationarity with different response peak values.
Furthermore, comparison between Fig. 12.5a and f proves that the FEP-ratio is
more sensitive to the mass m s of the non- structural components than to the PO
frequency f s .
In order to provide a complete  overview of the investigated phenomenon, the
FEP-rate envelope  ˆ t f p , x  , defined by Eq. (12.14), is reported in Fig. 12.6a and
b by means of a 3D surface and of a colormap, respectively.
A very interesting issue consists in the fact that, for all the considered values of
m s and f s , the first passage probabilities are sensitive to the threshold rather that to
the time interval. In particular, for sufficiently lasting intervals with t f p ≥ 15 s and
at thresholds x  < 0.055 m, the extreme peak processes of the unperturbed and of
the parametrized structure converge to the same tail probability.
248 S. Sessa and L. Rosati

(a) fs = 0.5 Hz; ms /m0 = 0.01 (b) fs = 0.5 Hz; ms /m0 = 0.1

(c) fs = 2 Hz; ms /m0 = 0.01 (d) fs = 2 Hz; ms /m0 = 0.1

(e) fs = 5 Hz; ms /m0 = 0.01 (f) fs = 5 Hz; ms /m0 = 0.1

Fig. 12.5 First excursion

 probability rate  as function of the time interval t f p and threshold x 
(In white regions  t f p , x  , f s , m s ≤ 10−4 )
12 Influence of Non-structural Components on Equivalent Linearization of Buildings 249

(a) 3D surface (b) Colormap

Fig. 12.6 First excursion probability

 rate ˆ as function of the time interval t f p and
 envelope 
threshold x  (In white regions 
ˆ t f p , x  ≤ 10−4 )

12.4 Conclusion

Influence of non-structural components on the response of a structural system is

investigated and discussed. To this end, the Tail-Equivalent Linearization Method,
providing a linearized equivalent system by means of response functions, has been
adopted.
A comparison between the frequency content of an unperturbed structure, in
which the presence of non-structural devices is neglected, and of a parameterized
model, where a secondary, parametrized oscillator is introduced, proved that struc-
tural responses are strongly influenced by resonance phenomena.
Despite of the fact that non-structural components dramatically adulterate the
response in time of the model, their influence on the first excursion probability turns
out to be more limited.
In particular, for sufficiently lasting time intervals and moderate thresholds, the
peak response of the parametrized model tends to coincide with the one of the unper-
turbed structure independently of the dynamical properties of the secondary oscilla-
tor.
Such an aspect becomes pivotal if the linearized system aims to be used for
determining secondary excitations. In fact, in such a case, the selected thresholds
cannot overcome the values for which the main structure becomes irremediably
damaged: it could be meaningless to characterize a seismic excitation to be used for
non-structural components if the main structure collapsed.
In this sense, the presented results are very appealing and open a new perspective
on the possible applications of Tail-Equivalent linearization.
Future research directions will be focused on the characterization of secondary
excitation processes. Within the framework of protection of cultural heritage arti-
facts and archaeological vestiges, such investigations may include isolation of statues
(Vaiana et al. 2017) by introducing a biaxial characterization of the nonlinear model
250 S. Sessa and L. Rosati

(Vaiana et al. 2021) and appropriate integration procedures (Greco et al. 2018).
Moreover, significant improvements consist in extending tail-equivalent lineariza-
tion to more complex structural models including shell elements (Altenbach et al.
2009, 2010; Cuomo et al. 2014) damage (Contrafatto and Cuomo 2006; Franciosi
et al. 2019), viscoelasticity (Cuomo 2017), metamaterials (De Angelo et al. 2019;
di Cosmo et al. 2018; Abd-alla et al. 2017), and reinforced concrete (Contrafatto
et al. 2012, 2016).

Acknowledgements The present research was supported by the University of Naples Federico II
and the Compagnia di San Paolo, which are gratefully acknowledged by the authors, as part of
the Research Project MuRA— Multi-risk assessment and structural protection of archaeological
vestiges in volcanic scenarios, FRA grants, CUP E69C21000250005.

References

AenN Abd-alla, Alshaikh F, Del Vescovo D, Spagnuolo M (2017) Plane waves and eigenfrequency
study in a transversely isotropic magneto-thermoelastic medium under the effect of a constant
angular velocity. J Therm Stress 40(9):1079–1092
Alibert JJ, Seppecher P, Dell’Isola F (2003) Truss modular beams with deformation energy depend-
ing on higher displacement gradients. Math Mech Solids 8(1):51–73
Altenbach H, Eremeyev V, Morozov N (2009) Linear theory of shells taking into account surface
stresses. Doklady Phys 54:531–535
Altenbach J, Altenbach H, Eremeyev VA (2010) On generalized cosserat-type theories of plates
and shells: a short review and bibliography. Arch Appl Mech 80(1):73–92
Andreaus U, Spagnuolo M, Lekszycki T, Eugster SR (2018) A Ritz approach for the static analysis
of planar pantographic structures modeled with nonlinear Euler-Bernoulli beams. Contin Mech
Thermodyn 30(5):1103–1123
Baber TT, Noori MN (1985) Random vibration of degrading, pinching systems. J Eng Mech
111(8):1010–1026
Badillo-Almaraz H, Whittaker AS, Reinhorn AM (2007) Seismic fragility of suspended ceiling
systems. Earthq Spectra 23(1):21–40
Barchiesi E, Khakalo S (2019) Variational asymptotic homogenization of beam-like square lattice
structures. Math Mech Solids 24(10):3295–3318
Barchiesi E, Dell’Isola F, Laudato M, Placidi L, Seppecher P (2018) A 1d continuum model for
beams with pantographic microstructure: asymptotic micro-macro identification and numerical
results. In: Advances in mechanics of microstructured media and structures. Springer, pp 43–74
Broccardo M, Der Kiureghian A (2016) Multicomponent nonlinear stochastic dynamic analysis by
tail-equivalent linearization. J Eng Mech 142(3):04015,100
Calio I, Marletta M (2003) Passive control of the seismic rocking response of art objects. Eng Struct
25(8):1009–1018
Chiaia B, Barchiesi E, De Biagi V, Placidi L (2019) A novel structural resilience index: definition
and applications to frame structures. Mech Res Commun 99:52–57
Contrafatto L, Cuomo M (2006) A framework of elastic-plastic damaging model for concrete under
multiaxial stress states. Int J Plass 22(12):2272–2300
Contrafatto L, Cuomo M, Fazio F (2012) An enriched finite element for crack opening and rebar
slip in reinforced concrete members. Int J Fract 178(1–2):33–50
Contrafatto L, Cuomo M, Gazzo S (2016) A concrete homogenisation technique at meso-scale level
accounting for damaging behaviour of cement paste and aggregates. Comput Struct 173:1–18
12 Influence of Non-structural Components on Equivalent Linearization of Buildings 251

di Cosmo F, Laudato M, Spagnuolo M (2018) Acoustic metamaterials based on local reso-

nances: homogenization, optimization and applications. In: Generalized models and non-classical
approaches in complex materials, 1. Springer, pp 247–274
Cuomo M (2017) Forms of the dissipation function for a class of viscoplastic models. Math Mech
Complex Syst 5(3):217–237
Cuomo M, Contrafatto L, Greco L (2014) A variational model based on isogeometric interpolation
for the analysis of cracked bodies. Int J Eng Sci 80:173–188
De Angelo M, Spagnuolo M, D’annibale F, Pfaff A, Hoschke K, Misra A, Dupuy C, Peyre P,
Dirrenberger J, Pawlikowski M, (2019) The macroscopic behavior of pantographic sheets depends
mainly on their microstructure: experimental evidence and qualitative analysis of damage in
metallic specimens. Contin Mech Thermodyn 31(4):1181–1203
Dell’Isola F, Turco E, Misra A, Vangelatos Z, Grigoropoulos C, Melissinaki V, Farsari M (2019)
Force-displacement relationship in micro-metric pantographs: Experiments and numerical sim-
ulations. Comptes Rendus Mécanique 347(5):397–405
Ditlevsen O, Madsen HO (1996) Structural reliability methods, vol 178. Wiley, New York
FEMA (1994) Fema e-74, reducing the risks of nonstructural earthquake damage-a practical guide
Franciosi P, Spagnuolo M, Salman OU (2019) Mean green operators of deformable fiber networks
embedded in a compliant matrix and property estimates. Contin Mech Thermodyn 31(1):101–132
Fujimura K, Der Kiureghian A (2007) Tail-equivalent linearization method for nonlinear random
vibration. Prob Eng Mech 22(1):63–76
Garrè L, Der Kiureghian A (2010) Tail-equivalent linearization method in frequency domain and
application to marine structures. Marine Struc 23(3):322–338
Greco F, Luciano R, Serino G, Vaiana N (2018) A mixed explicit-implicit time integration approach
for nonlinear analysis of base-isolated structures. Annals Solid Struct Mech 10(1):17–29
Kiureghian AD, Fujimura K (2009) Nonlinear stochastic dynamic analysis for performance-based
earthquake engineering. Earthq Eng Struct Dyn 38(5):719–738
Mazzoni S, McKenna F, Scott MH, Fenves GL et al (2006) Opensees command language manual.
Pacific Earthq Eng Res (PEER) Center 264
Milton GW, Seppecher P, Bouchitté G (2009) Minimization variational principles for acoustics,
elastodynamics and electromagnetism in lossy inhomogeneous bodies at fixed frequency. Proc R
Soc A Math Phys Eng Sci 465(2102):367–396
Sessa S (2010) Multiobjective non-linear random vibration analysis for performance-based earth-
quake engineering. In: Proceedings of reliability and optimization of structural systems, pp 185–
192
Turco E, Dell’Isola F, Misra A (2019) A nonlinear Lagrangian particle model for grains assemblies
including grain relative rotations. Int J Num Anal Methods Geomech 43(5):1051–1079
Turco E, Barchiesi E, Giorgio I, Dell’Isola F (2020) A Lagrangian hencky-type non-linear model
suitable for metamaterials design of shearable and extensible slender deformable bodies alterna-
tive to timoshenko theory. Int J Non-Linear Mech 123(103):481
Vaiana N, Spizzuoco M, Serino G (2017) Wire rope isolators for seismically base-isolated
lightweight structures: experimental characterization and mathematical modeling. Eng Struct
140:498–514
Vaiana N, Losanno D, Ravichandran N (2021) A novel family of multiple springs models suitable
for biaxial rate-independent hysteretic behavior. Comput Struct 244(106):403
Villaverde R (1997) Seismic design of secondary structures: state of the art. J Struct Eng
123(8):1011–1019
Wang Z, Der Kiureghian A (2016) Tail-equivalent linearization of inelastic multisupport struc-
tures subjected to spatially varying stochastic ground motion. Journal of Engineering Mechanics
142(8):04016,053
Chapter 13
Do We Really Need Pantographic
Structures?

Mario Spagnuolo and Emilio Barchiesi

Abstract This chapter attempts to provide a comprehensive answer to the challeng-

ing question: do we really need pantographic structures? This question may arise
spontaneously given the recent proliferation of contributions on this type of meta-
material in the literature. A reasoned answer to this question may be crucial for the
future development and orientation of research concerning this metamaterial. More
generally, we show the context from which the studies that led to the development of
pantographic structures originated and observe how an excessively orthodox view of
Continuum Mechanics may prevent interesting developments. Within the framework
of generalised theories and second-gradient models, pantographic structures assume
an important role.

Keywords Pantographic structures · Metamaterials · Continuum mechanics ·

Second gradient theory · Microstructure · Variational principles

13.1 Introduction

Recently, the literature has seen a proliferation of scientific contributions concerning

so-called pantographic structures (dell’Isola et al. 2016; Boutin et al. 2017; Eremeyev
et al. 2018, 2019; Giorgio et al. 2019; Rahali et al. 2015; Andreaus et al. 2018;
Barchiesi et al. 2019a; Scerrato and Giorgio 2019). The mechanical properties of such
objects are reported to be exotic and very high-performance. In this short chapter, we

M. Spagnuolo (B)
Dipartimento di Ingegneria Civile, Ambientale e Architettura (DICAAR), Universitá degli
Studi di Cagliari, Cagliari, Italy
e-mail: [email protected]
M. Spagnuolo · E. Barchiesi
International Research Center for the Mathematics and Mechanics of Complex Systems,
University of L’Aquila, L’Aquila, Italy
E. Barchiesi
École Nationale d’Ingénieurs de Brest, ENIB, UMR CNRS 6027, IRDL, 29200 Brest, France

© Springer Nature Switzerland AG 2021 253

F. Marmo et al. (eds.), Mathematical Applications in Continuum
and Structural Mechanics, Advanced Structured Materials 127,
https://doi.org/10.1007/978-3-030-42707-8_13
254 M. Spagnuolo and E. Barchiesi

would like to attempt to answer the insistent question that is asked by some scholars
when topics related to pantographic structures are presented at scientific congresses:
do we really need pantographic structures?
In order to answer this question, we will need to retrace in an organic manner the
path that led to the formulation of the theory governing these objects. We will have
to give a clear and agreeable definition of what a metamaterial is Barchiesi et al.
(2019b), Abali and Yang (2019), Carcaterra et al. (2015), dell’Isola and Steigmann
(2020), Eugster et al. (2019), Giorgio et al. (2020), Yang et al. (2018). In the end, we
will have to show, as in any worthwhile theory, what is the real theoretical necessity
behind the existence of such devices.
As we will see specifically, the environment in which pantographic structures
have been developed has to be researched back to the split that originated in the
nineteenth century, when the school of Continuum Mechanics related to Cauchy,
Navier and Poisson prevailed over that of Lagrangian, D’Alembertian and Piolan
inspiration (Lagrange 1853, 1806; dell’Isola et al. 2014, 2019, 2015a).
In order to be able to address this study in a consistent and self-contained manner,
it is necessary to specify, at least in brief, what a pantographic structure is. A pan-
tographic structure consists of a planar grid made up of two families of continuous
fibres oriented orthogonally and interconnected by hinges located at the intersections.
From a purely theoretical point of view, the mechanical behaviour of pantographic
structures is treated in the formal context of higher gradient continua, i.e. continua
whose deformation energies depend on higher gradients of the displacement field, as
opposed to the Cauchy continuum in which the deformation energy is only a function
of the first gradient of displacement.

13.2 Metamaterials Are (Natural) Materials on Demand

In its formulation due mainly to Lagrange, Continuum Mechanics studies how the
equilibrium shapes of a continuum body are modified by external interactions. A
given body is assumed to consist, at each of its material points, of a specific material.
The actual shape of such a body is mathematically modelled by means of a place-
ment function and, in its elastic deformation range, by the corresponding deformation
energy density, which objectively depends on the placement gradient. Further con-
stitutive functions and kinematic descriptors can be introduced for the modelling of
damage (Cuomo 2019; Misra and Poorsolhjouy 2020; Placidi 2015; Placidi et al.
2018; Spagnuolo et al. 2017) and plastic phenomena (Altenbach and Eremeyev
2014; Bertram 2015). The adoption of the above mathematical context is impor-
tant in order to generalise the concepts of Continuum Mechanics to non-standard
materials generally referred to as artificial or non-natural. Such a change of perspec-
tive, in fact, makes the expression “natural material” completely meaningless. In
fact, from a purely modelling point of view, we can only speak of materials with a
simple microstructure and materials with a complex microstructure. In the context
of Continuum Mechanics based on the definition of the deformation energy density
13 Do We Really Need Pantographic Structures? 255

associated with a given material, every material that can exist is natural by definition
and what changes is only its mechanical behaviour, which is questioned if it can be
described by certain constitutive functions.
To answer the question that underlies this chapter, we are interested in placing
pantographic structures, or pantographic material as it is often referred to, in the
proper context of generalised theories in Continuum Mechanics. For this reason,
we are interested in discussing the positioning of theories concerning metamaterials
in Continuum Mechanics. To this end, the definition given of a metamaterial is of
fundamental importance. A definition that is often found in the literature and that,
for what we have said so far, seems paradoxical consists in defining the theory of
metamaterials as the theory of those materials that are not natural: but, we have
already pointed out, all materials are by definition natural. Another possibility also
often found in the literature (Seppecher et al. 2011) is to define metamaterials as those
materials whose mechanical behaviour is “exotic”. This definition, however, is also
easily attacked. Indeed, it is necessary to specify what we mean by exotic mechanical
behaviour. It seems reasonable to argue that exotic mechanical behaviour is a type of
behaviour that has not yet been experimentally observed. Of course, what is exotic
at one point in time may become standard at another. For example, Lamé, Navier,
Cauchy, Poisson, all regarded a material with a negative Poisson’s ratio as very exotic,
and some scholars of their time even believed that such a material was not physical
and could not exist. However, auxetic metamaterials do exist and play an important
role in modern engineering (Evans 1991; Evans and Alderson 2000).
The cited approach to the theory of metamaterials produces several misunder-
standings that can lead the scientific investigation to results completely detached
from reality. Indeed, claiming that a given mechanical behaviour is standard as
opposed to another implies that one is confusing a mathematical model for a material
with the physical material itself, and that one is implicitly assuming that particular
assumptions accepted to describe particular phenomena are universally valid in every
physical situation. This attitude, as well as being unscientific, does not lead to any
advancement in scientific research and is, therefore, to be avoided. The confusion
between model and modelled object has led over time to conceptual statements
that are completely cacophonous. The example we want to give here consists of
the assumption that “materials described by second-gradient models do not exist
because the materials used in engineering do not show their properties and the stan-
dard theoretical framework does not envisage them”. Those who agree with the
above statement are clearly confusing first gradient models (model belonging to a
symbolic-mathematical description) with materials existing in nature (physical object
describable by a model). Exaggerating this non-scientific attitude to its extreme con-
sequences, one may come to believe that, without having a theory to describe it, one
cannot use a material even though he has it in his hands, and may even believe that
a certain material does not exist. This is maybe the main reason for which we need
a metamaterial as the pantographic one: its existence demonstrates by itself that the
first gradient continua do not model every existing material.
In view of what has been said so far, we can now try to give an operational
definition of what a metamaterial is. A metamaterial is a material on demand: we
256 M. Spagnuolo and E. Barchiesi

establish a priori the mechanical behaviour we wish to observe in such a material, and
only secondarily we search for a microstructure that, following a homogenisation
procedure, can present the required mechanical behaviour at a macroscopic level.

13.3 Second Gradient Theories

As we have observed, in the self-proclaimed standard school of Continuum Mechan-

ics there is no place for theories other than those studied in the tradition of Caucy,
Navier and Poisson. Gabrio Piola introduced in 1848 a generalised continuum model
by means of deformation energies dependent on the nth gradient of the placement
(Piola 1846). However, this type of model did not fit into the orthodox formulation
of Continuum Mechanics à la Cauchy. Paradigmatic, in this context, is the unani-
mous agreement by Cauchy and his acolytes of the so-called “Cauchy postulate”,
which asserts that the contact forces, within continua, can only be forces per unit
area depending only on the normal to the Cauchy cuts.
In Cauchy’s version of Continuum Mechanics some ad hoc restrictions are
included, among which is the fact that the deformation energy of a continuum
medium can only depend on the first gradient of the displacement field. A priori,
nothing would prevent a dependence on higher order gradients, but the simplest
choice, coherent with the phenomenology shown by Cauchy’s continuum model, is
to restrict to the first gradient of the displacement. Piola, on the other hand, introduces,
for a mere rational demand, the higher gradients of displacement in the calculation of
the deformation energy, arguing for characterising those microstructures for which
the homogenised models must be of this more general type. In Alibert et al. (2003),
Seppecher et al. (2011), dell’Isola and Seppecher (1995, 1997), Pideri and Seppecher
(1997), Seppecher (1989, 2000) it is shown that models in which the second gra-
dient of displacement takes on a non-negligible role, at the macroscopic level, are
obtained by homogenisation from a microstructure, or architecture, at a lower scale
in a continuum medium where high stiffness contrasts are present.
From what we have said, it seems clear that in order to be able to evaluate and
observe experimentally effects that can be assimilated to a description by means of the
second gradient of the displacement field, it is necessary to have a technology capable
of producing a microstructured material (Spagnuolo et al. 2019; Altenbach and Ere-
meyev 2014; Eremeyev et al. 2012) and, above all, a material whose microstructure
shows the appropriate highly contrasted stiffness fields, so that, at the macroscopic
level, the terms used by Piola appear in the deformation energy. In the following we
will shortly show how Paul Germain demonstrates that the presence of a microstruc-
ture can determine, at the macroscopic level, the necessity of using a second-gradient
model (Germain 1973, 2020; Epstein and Smelser 2020).
The technological capacity of an era can also block its scientific development. As
long as technology does not reach a sufficient level to test the results introduced in the
new theories, the new theories will remain blocked, ignored and, certainly, unusable.
The absurdity of the contemporary situation consists in the fact that regardless of the
13 Do We Really Need Pantographic Structures? 257

technological ability to produce materials whose behaviour is described by Piola’s

theory (and cannot be described within the framework of Cauchy’s models), there
are actually still scholars who insist on denying its usefulness.
Stressing a concept we have already mentioned before, this last remark partly
answers the question we asked at the beginning of this chapter: do we really need
pantographic structures? From a purely theoretical point of view, the answer is clearly
affirmative, because pantographic structures have been expressly designed as a mate-
rial that can be described fundamentally in terms of a second-gradient theory: if we
can fabricate a material that can be described by means of a given theory and not
by means of the theory generally adopted in the description of materials, then this
theory becomes indispensable and it is absolutely inconceivable that it should not be
applicable.

13.4 Microstructure in Continuum Mechanics

At present, with recently developed and improved techniques, the fabrication of

materials with complex microstructures is not as implausible as before. As we have
specified above, due to the advancements in the field of additive manufacturing,
it is now possible to produce microstructured metamaterials exhibiting mechanical
properties that cannot be described in the context of Cauchy’s Continuum Mechanics.
Here we also want to address an issue concerning the terminology adopted in the
field of metamaterials. Some scholars claim that the term microstructure cannot be
used because it refers to a scale of micrometres, whereas it is more appropriate to
refer to architecture. However, in our opinion, this associates the metamaterial with an
artefact connotation that is not of fundamental interest in the theory. More specifically,
the relevant point is to understand what determines the mechanical properties of
the so-called standard or natural materials: clearly the difference between a cubic
material and a tetragonal one, just to give an example, is in the geometry according
to which the “particles” that compose it are arranged, i.e. in its microstructure. Now
some people refer to architecture in the case of metamaterials because the present
technological capabilities do not yet allow manufacture on scales comparable to
those currently existing in nature. But the idea is the same as that underlying the
differences between the mechanical behaviour of cubic and tetragonal materials: the
mechanical behaviour of a metamaterial is determined by the geometric arrangement
of its elementary constituents. This is why, in our opinion, distinguishing architecture
from microstructure is only misleading.
As we have seen above, Piola was the first to study continuum models in which
the deformation energy depends on higher gradients of the displacement field. Later,
in the twentieth century, this kind of models was also studied and reformulated by
various authors. Among others, we mention in particular two pioneering studies pre-
sented by Mindlin (1965) and by Germain (1973, 2020); Epstein and Smelser (2020).
A very interesting aspect that emerges from these studies, and of course from others,
is that they show how the existence of the microstructure in some cases could induce
258 M. Spagnuolo and E. Barchiesi

higher order terms in the equilibrium equations of the material under consideration.
Unlike classical homogenisation techniques, equations containing terms dependent
on second or higher order derivatives of the displacement are obtained in this case,
thus introducing higher gradient theories in a logical way.
Germain shows in general that, by applying the Principle of Virtual Power (anal-
ogous to the Principle of Virtual Work), the classical equations of the Continuum
Mechanics are easily obtained. When considering a microstructured continuum, these
equations have terms that depend on the second gradient of the displacement field.
In this context, it is crucial to assign the right kinematics. Therefore, in the case
of a usual continuum, this is considered to consist of a continuum distribution of
particles geometrically represented by a material point and its velocity components.
When considering the microstructure, from a macroscopic point of view each parti-
cle is still represented by a material point, but its kinematics must be defined more
precisely. The main feature of the method explained by Germain is that, having
assigned the required kinematics, the associated continuum theory can be deduced
immediately through the Principle of Virtual Work. He shows that the kinematics
due to the presence of the microstructure generates a second-gradient continuum at
the macroscopic level. We refer to the original work by Germain for the technical
details (Germain 1973, 2020; Epstein and Smelser 2020).
This can be considered the starting point in the study of pantographic structures.
A certain microstructure is chosen to get a second-gradient continuum as simple as
possible and then homogenisation techniques are used to determine this appropriate
macroscopic continuum model.

13.4.1 The Synthesis Problem

The real mathematical challenge facing us today, therefore, is to design metamate-

rials that can be described within the framework of a generalised theory (Maugin
2011; Altenbach et al. 2010; Altenbach and Eremeyev 2010, 2013; Altenbach et al.
2013; Auffray 2015; dell’Isola et al. 2009; Eremeyev and dell’Isola 2018). Thus, the
fundamental problem in the theory of metamaterials consists in the problem of the
synthesis of microstructures that produce a certain desired macro-behaviour (Rahali
et al. 2015; Placidi et al. 2020; Khakalo and Niiranen 2020; Abdoul-Anziz et al.
2019). As we have seen briefly, in this context, the most complex problem to address
from a mathematical point of view is to link microstructures and macro-behaviours.
This is done in terms of mathematical procedures called homogenisation: starting
from the elementary constituents and the basic cell of the chosen microstructure, one
must link this microstructure to the given macroscopic theory (appropriate action
functions and consequent stationarity conditions) chosen a priori, and this, in order
to be of some general use, must be approached in an algorithmic manner, thus allow-
ing generalisation.
The basic ideas in the field of metamaterial synthesis can be deduced by analogy
from the theory of analogue circuit synthesis. In this theory, it can be shown that any
13 Do We Really Need Pantographic Structures? 259

passive linear element can be synthesised algorithmically using inductors, capacitors,

resistors and transformers (Bloch 1944, 1945; Kron 1945; Mablekos and Weidmann
1968). The main challenge is to conjecture that this method can also be applied to the
synthesis of non-linear (and multiphysical) mechanical systems (Spagnuolo 2020;
Spagnuolo and Scerrato 2020).

13.5 Why We Really Need Pantographic Structures

Up to now we have tried to exhibit the background needed to answer a question often
repeated as a result of the enormous development of literature on pantographic struc-
tures: do we really need pantographic structures? For what purpose do they serve?
There are many arguments advanced on this topic which are commonly referred
against second and higher gradient models: some scholars observe that the extremely
formal mathematical investigation required to formulate the theory underlying this
type of metamaterial is unnecessarily over-discussed and studied, and that there are
no practical applications of the object studied; others point out that, on the contrary,
from a theoretical point of view it is not necessary to introduce second-gradient
theories, because they are useless mathematical complications, but that the classical
Cauchy theory on its own is able to explain and represent all the phenomenology
observed in Continuum Mechanics; finally, some argue that there are many other
metamaterials much more interesting than the pantographic one. We limit ourselves
to observe that one of the most cited works in the field of metamaterials (Bertoldi
et al. 2010), the work of researchers belonging to the Harvard intelligentsia, seems
to us to be rather weak and basically a patchwork of experimental results obtained
for a microstructure that produces a material with auxetic mechanical behaviour.
But it is not our purpose to comment on the null and void scientific contribution of
the various power groups of the moment. Instead, we are interested in precisely defin-
ing the reasons why pantographic structures are actually worth studying. Basically,
we can divide the positive scientific contributions resulting from the development of
this type of metamaterials into four areas: theoretical, practical, methodological and
multiphysical.

13.5.1 The Existence of Pantographic Metamaterial

Motivates the Need of Second Gradient Theories

From a theoretical and methodological point of view, the importance of pantographic

structures seems indisputable. In fact, consider the fundamental objection that has
historically been made to second-gradient theories: they are not necessary, since the
materials can be very well described within the framework of the classical Cauchy
first gradient theory, possibly by adding ad hoc corrections. If, then, we are able
260 M. Spagnuolo and E. Barchiesi

to show a material not describable by means of a first gradient theory, but only by
invoking a second-gradient one, then this is enough to motivate the necessity of such
theories. We want, accordingly, to postulate the existence of such a material by writ-
ing its governing equations, i.e. its deformation energy, which should be dependent of
the second gradient of displacement. We subsequently ask which microstructure can
produce this macroscopic deformation energy after homogenisation. This approach
is methodologically the reverse of that used in the majority of works on metama-
terials, including the aforementioned (Bertoldi et al. 2010). We therefore ask what
characteristics a microstructure must have in order to produce a second-gradient
continuum at the macroscopic level.
Methodologically we start from the following observation, which is certainly
superficial, but definitely indicative to try to define a microstructure suitable to pro-
duce a second-gradient mechanical behaviour: if we consider an Euler-Bernoulli
beam, even a linear one, its deformation energy can be separated into two compo-
nents, elongation and bending. The elongation energy depends on the first derivative
of the longitudinal displacement, while the bending energy depends on the second
derivative of the transverse displacement. In a sense, we can say that the Euler-
Bernoulli beam is first gradient in extension and second gradient in bending. We
want to produce a second-gradient material at extension. We can therefore conceive
an assembly of beams such that a macroscopic extension action corresponds to a
bending action from a microscopic point of view. In other words, by extending or
compressing the metamaterial we are flexing the fibres that make up its microstruc-
ture. In this way, the deformation energy of the metamaterial subjected to extensional
load should correspond to a “microscopic” bending energy and consequently to a
second-gradient energy. This is how the pantographic microstructure has been orig-
inated.

13.5.2 A Mechanical Diode

From a practical point of view, we would like to highlight just one aspect that seems
extremely promising with respect to the possible applications of the pantographic
metamaterial. Consider that, due to the peculiar microstructure chosen, from the
point of view of mechanical behaviour, we observe a phenomenology similar to that
presented by the diode in the analysis of electrical circuits. The diode is a circuit
element that exhibits an extremely particular voltage-current response. In various
applications of interest, the voltage-current behaviour of an ideal diode, under static
conditions, can be approximated by a linear piecewise function. In this approxi-
mation, the current can be considered to be zero if the voltage between anode and
cathode is less than or equal to a certain threshold value Vγ ; if, on the contrary,
the voltage is higher, the diode can be approximated to a voltage generator, whose
current is imposed by the circuit to which it is subordinated. In the field of mechan-
ics, a response formally identical to that exhibited by the diode in electrical circuits
is shown by the pantographic metamaterial. If the hinges connecting the fibres of
13 Do We Really Need Pantographic Structures? 261

the two families are perfect, and therefore no deformation energy is associated with
them, then considering a bias-extension test of a pantographic structure the following
response will be observed in terms of a force-displacement measurement: up to a
certain imposed displacement a very low value of the reaction force will be mea-
sured; after a threshold value of the displacement, the force will begin to increase
considerably. The shape of the observed curve is reminiscent of the voltage-current
diagram of the diode. In this sense, one can refer to the pantographic metamaterial
as a kind of mechanical diode.
This mechanical behaviour can be useful when one wants to insert a mechani-
cal element into a structure which does not function directly as a spring, but only
exhibits linear elastic behaviour after a certain elongation threshold value (at least
for deformations not too far beyond the threshold value).
The explanation for this unusual mechanical behaviour lies in the deformation
of the microstructure: in a bias-extension test, in a first phase the predominantly
observed deformation corresponds to the bending of the fibres clamped at the ends
of the pantographic structure (whereas the unclamped fibres are simply free to rotate).
This bending is concentrated in very precise areas of the pantographic structure and
gives rise, from a theoretical point of view, to the bending term modelled by means
of the second gradient of placement. It is common experience that a beam is easier to
bend than to stretch, so it is easy to agree that the bending energy of the pantographic
structure (beam assembly) is certainly lower than the elongation energy. However,
since the extension test is conducted along a direction in which the fibres are biased,
then the first mechanism to occur corresponds to the bending of the fibres, whereas
their elongation only begins when the rotating fibres touch (which corresponds to
the threshold value above which the measured force begins to increase significantly).
One can refer to Spagnuolo (2020), Spagnuolo and Scerrato (2020) for more details.

13.5.3 An Iterative Algorithm for Synthesising Metamaterials

From a methodological-theoretical point of view, we want to underline an important

aspect recalled in Alibert et al. (2003), Seppecher et al. (2011). Once the panto-
graphic microstructure has been obtained, which at the macroscopic level produces
the required second-gradient behaviour by means of a suitable homogenisation, one
can think of generalising this procedure to the production of metamaterials described
by energies dependent on higher displacement gradients. In Alibert et al. (2003),
Seppecher et al. (2011) it is shown how, by using a Warren bridge microstructure
in association with a pantographic microstructure, it is possible to obtain a third
gradient-in-bending material. With an iterative procedure it is suggested that one can
proceed from (2n + 1)-th gradient-in-bending materials to (2n + 2)-th gradient-in-
extension materials. This problem, only mentioned in Alibert et al. (2003), Seppecher
et al. (2011), turns out to be of utmost importance in the synthesis of new metama-
terials.
262 M. Spagnuolo and E. Barchiesi

13.6 Conclusion

In this chapter, we have briefly discussed some of the fundamental motivations behind
the development of pantographic metamaterial. We have established that studies of
this metamaterial are strongly motivated from several points of view: from a theoreti-
cal point of view, it demonstrates the necessity of the introduction of second-gradient
models; from a practical point of view, this metamaterial possesses extremely pecu-
liar characteristics and, therefore, is worthy of detailed study; from a methodological
point of view, it offers a useful example in the field of procedures for the synthesis
of new metamaterials.
These reasons that we have listed and discussed lead us to believe that panto-
graphic structures may provide a new class of materials worth studying in depth. In
fact, there are several indications that the phenomenology of this metamaterial is very
variegated. Here we limit ourselves to a few cases of extreme interest: the observation
of Poynting reversal effects for torsion tests conducted on pantographic structures
(Misra et al. 2018; Auger et al. 2020); the observation of wave phenomena in the case
of very dense mesh structures of sufficiently large length (dell’Isola et al. 2015b); the
study of the distribution of displacement and velocity fields for high-frequency vibra-
tory phenomena (Laudato et al. 2018; Barchiesi et al. 2018; Laudato and Barchiesi
2019; Laudato et al. 2020); the extremely non-standard phenomenology in the case
of three-point bending tests (Yildizdag et al. 2020).
The enormous amount of phenomenology and experimental observations obtained
on pantographic structures require the development of precise methods of analysis,
such as those based on Digital Image Correlation, already applied to the case of
this metamaterial with good results (Hild et al. 2020; Barchiesi et al. 2020b, a),
and of numerical implementation of theoretical models in order to conduct precise
model validation. This also requires the development of numerical methods capable
of simulating the cases of interest (Cazzani and Atluri 1993; Cazzani and Lovadina
1997; Cazzani et al. 2016c, b, a, 2020; Cuomo et al. 2014; Greco and Cuomo 2013,
2014; Maurin et al. 2019; Capobianco et al. 2018; Turco et al. 2017, 2018, 2019b, a;
Turco and Barchiesi 2019; Schulte et al. 2020; Hesch et al. 2017; Capobianco and
Eugster 2018; Eugster and Glocker 2013; Huang et al. 2020, 2021; Barchiesi et al.
2020c; Yang et al. 2019; Jafarzadeh et al. 2020; Namnabat et al. 2020; Rahbar et al.
2020). The methods and reasoning developped for pantographic structures may also
be interesting for other applications, such as civil engineering (Vaiana et al. 2021,
2019; Serpieri et al. 2018; Sessa et al. 2019a, 2017, 2018a, 2019b, 2018b, 2015;
Cricri et al. 2015; Greco et al. 2018; Perricone et al. 2020; Marmo et al. 2018a, b,
2019; Vaiana et al. 2017; Paradiso et al. 2019).

Acknowledgements Il presente articolo é stato realizzato nell’mbito del progetto di ricerca finanzi-
ato con le risorse del P.O.R. SARDEGNA F.S.E. 2014–2020—Asse III “Istruzione e Formazione,
Obiettivo Tematico: 10, Obiettivo Specifico: 10.5, Azione dell’ccordo di Partenariato:10.5.12”
Avviso di chiamata per il finanziamento di Progetti di ricerca—Anno 2017
13 Do We Really Need Pantographic Structures? 263

References

Abali BE, Yang H (2019) Parameter determination of metamaterials in generalized mechanics as a

result of computational homogenization. In: Indeitsev D, Krivtsov A (eds) Advanced problems in
mechanics. APM 2019. Lecture Notes in Mechanical Engineering. Springer, Cham, Swizerland,
pp 22–31
Abdoul-Anziz H, Seppecher P, Bellis C (2019) Homogenization of frame lattices leading to sec-
ond gradient models coupling classical strain and strain-gradient terms. Math Mech Solids
24(12):3976–3999
Alibert JJ, Seppecher P, dell’Isola F (2003) Truss modular beams with deformation energy depending
on higher displacement gradients. Math Mech Solids 8(1):51–73
Altenbach H, Eremeyev V (2014) Strain rate tensors and constitutive equations of inelastic microp-
olar materials. Int J Plas 63:3–17
Altenbach H, Eremeyev VA (2010) On the theories of plates based on the Cosserat approach. In:
Mechanics of generalized continua. Springer, pp 27–35
Altenbach H, Eremeyev VA (2013) Cosserat media. In: Eremeyev VA, Altenbach H (ed) Generalized
continua from the theory to engineering applications. Springer, pp 65–130
Altenbach H, Bîrsan M, Eremeyev VA (2013) Cosserat-type rods. In: Generalized continua from
the theory to engineering applications. Springer, pp 179–248
Altenbach J, Altenbach H, Eremeyev VA (2010) On generalized Cosserat-type theories of plates
and shells: a short review and bibliography. Arch Appl Mech 80(1):73–92
Andreaus U, Spagnuolo M, Lekszycki T, Eugster SR (2018) A Ritz approach for the static analysis
of planar pantographic structures modeled with nonlinear Euler-Bernoulli beams. Contin Mech
Thermodyn 30(5):1103–1123
Auffray N (2015) On the algebraic structure of isotropic generalized elasticity theories. Math Mech
Solids 20(5):565–581
Auger P, Lavigne T, Smaniotto B, Spagnuolo M, dell’Isola F, Hild F (2020) Poynting effects in panto-
graphic metamaterial captured via multiscale DVC. J Strain Anal Eng Design 0309324720976625
Barchiesi E, Laudato M, Di Cosmo F (2018) Wave dispersion in non-linear pantographic beams.
Mechan Res Commun 94:128–132
Barchiesi E, Ganzosch G, Liebold C, Placidi L, Grygoruk R, Müller WH (2019a) Out-of-plane
buckling of pantographic fabrics in displacement-controlled shear tests: experimental results and
model validation. Contin Mech Thermodyn 31(1):33–45
Barchiesi E, Spagnuolo M, Placidi L (2019b) Mechanical metamaterials: a state of the art. Math
Mech Solids 24(1):212–234
Barchiesi E, dell’Isola F, Hild F, Seppecher P (2020a) Two-dimensional continua capable of large
elastic extension in two independent directions: asymptotic homogenization, numerical simula-
tions and experimental evidence. Mech Res Commun 103(103):466
Barchiesi E, Eugster SR, dell’Isola F, Hild F (2020b) Large in-plane elastic deformations of bi-
pantographic fabrics: asymptotic homogenization and experimental validation. Math Mech Solids
25(3):739–767
Barchiesi E, Yang H, Tran C, Placidi L, Müller WH (2020c) Computation of brittle fracture prop-
agation in strain gradient materials by the fenics library. Math Mech Solids 1081286520954513
Bertoldi K, Reis PM, Willshaw S, Mullin T (2010) Negative Poisson’s ratio behavior induced by
an elastic instability. Adv mater 22(3):361–366
Bertram A (2015) Finite gradient elasticity and plasticity: a constitutive mechanical framework.
Contin Mech Thermodyn 27(6):1039–1058
Bloch A (1944) XXXVIII. A new approach to the dynamics of systems with gyroscopic coupling
terms. London, Edinburgh, and Dublin Philosoph Mag J Sci 35(244):315–334
Bloch A (1945) Electromechanical analogies and their use for the analysis of mechanical and
electromechanical systems. J Instit Electric Eng Part I Gener 92(52):157–169
Boutin C, Giorgio I, Placidi L et al (2017) Linear pantographic sheets: asymptotic micro-macro
models identification. Math Mech Complex Syst 5(2):127–162
264 M. Spagnuolo and E. Barchiesi

Capobianco G, Eugster S (2018) Time finite element based moreau-type integrators. Int J Numer
Methods Engi 114(3):215–231
Capobianco G, Eugster SR, Winandy T (2018) Modeling planar pantographic sheets using a non-
linear Euler–Bernoulli beam element based on B-spline functions. PAMM 18(1):e201800,220
Carcaterra A, dell’Isola F, Esposito R, Pulvirenti M (2015) Macroscopic description of microscopi-
cally strongly inhomogenous systems: A mathematical basis for the synthesis of higher gradients
metamaterials. Arch Rational Mech Anal 218(3):1239–1262
Cazzani A, Atluri S (1993) Four-noded mixed finite elements, using unsymmetric stresses, for linear
analysis of membranes. Comput Mech 11(4):229–251
Cazzani A, Lovadina C (1997) On some mixed finite element methods for plane membrane prob-
lems. Comput Mech 20(6):560–572
Cazzani A, Malagù M, Turco E (2016a) Isogeometric analysis of plane-curved beams. Math Mech
Solids 21(5):562–577
Cazzani A, Malagù M, Turco E, Stochino F (2016b) Constitutive models for strongly curved beams
in the frame of isogeometric analysis. Math Mech Solids 21(2):182–209
Cazzani A, Stochino F, Turco E (2016c) An analytical assessment of finite element and isogeometric
analyses of the whole spectrum of Timoshenko beams. ZAMM J Appl Math Mech/Zeitschrift
für Angewandte Mathematik und Mechanik 96(10):1220–1244
Cazzani A, Serra M, Stochino F, Turco E (2020) A refined assumed strain finite element model for
statics and dynamics of laminated plates. Contin Mech Thermodyn 32(3):665–692
Cricrì G, Perrella M, Sessa S, Valoroso N (2015) A novel fixture for measuring mode iii toughness
of bonded assemblies. Eng Fract Mech 138:1–18
Cuomo M (2019) Continuum damage model for strain gradient materials with applications to 1d
examples. Contin Mech Thermodyn 31(4):969–987
Cuomo M, Contrafatto L, Greco L (2014) A variational model based on isogeometric interpolation
for the analysis of cracked bodies. Int J Eng Sci 80:173–188
dell’Isola F, Seppecher P (1995) The relationship between edge contact forces, double forces and
interstitial working allowed by the principle of virtual power. Comptes rendus de l’Académie des
sciences Série II, Mécanique, physique, chimie, astronomie 321(8):303–308
dell’Isola F, Seppecher P (1997) Edge contact forces and quasi-balanced power. Meccanica
32(1):33–52
dell’Isola F, Steigmann DJ (2020) Discrete and continuum models for complex metamaterials.
Cambridge University Press, Cambridge. https://doi.org/10.1017/9781316104262
dell’Isola F, Sciarra G, Vidoli S (2009) Generalized Hooke’s law for isotropic second gradient
materials. Proc R Soc A Math Phys Eng Sci 465(2107):2177–2196
dell’Isola F, Maier G, Perego U, et al (2014) The complete works of Gabrio Piola: Volume I
(Advanced structured materials, vol 38). Springer, Berlin
dell’Isola F, Andreaus U, Placidi L (2015a) At the origins and in the vanguard of peridynamics, non-
local and higher-gradient continuum mechanics: an underestimated and still topical contribution
of Gabrio Piola. Math Mech Solids 20(8):887–928
dell’Isola F, Giorgio I, Andreaus U (2015b) Elastic pantographic 2d lattices: a numerical analysis
on the static response and wave propagation. Proc Estonian Acad Sci 64(3):219
dell’Isola F, Giorgio I, Pawlikowski M, Rizzi NL (2016) Large deformations of planar extensible
beams and pantographic lattices: heuristic homogenization, experimental and numerical examples
of equilibrium. Proc R Soc A Math Phys Eng Sci 472(2185):20150,790
dell’Isola F, Maier G, Perego U et al (2019) The complete works of Gabrio Piola, vol II. Springer,
Cham
Epstein M, Smelser R (2020) An appreciation and discussion of Paul Germain’s “The method of
virtual power in the mechanics of continuous media, I: second-gradient theory". Math Mech
Complex Syst 8(2):191–199
13 Do We Really Need Pantographic Structures? 265

Eremeyev VA, dell’Isola F (2018) A note on reduced strain gradient elasticity. In: Altenbach H.,
Pouget J., Rousseau M., Collet B., Michelitsch T. (eds) Generalized models and non-classical
approaches in complex materials 1. Advanced structured materials, vol 89. Springer, Cham, pp
301–310
Eremeyev VA, Lebedev LP, Altenbach H (2012) Foundations of micropolar mechanics. Springer,
Heidelberg
Eremeyev VA, dell’Isola F, Boutin C, Steigmann D (2018) Linear pantographic sheets: existence
and uniqueness of weak solutions. J Elast 132(2):175–196
Eremeyev VA, Alzahrani FS, Cazzani A, dell’Isola F, Hayat T, Turco E, Konopińska-Zmysłowska
V (2019) On existence and uniqueness of weak solutions for linear pantographic beam lattices
models. Contin Mech Thermodyn 31(6):1843–1861
Eugster S, dell’Isola F, Steigmann D (2019) Continuum theory for mechanical metamaterials with
a cubic lattice substructure. Math Mech Complex Syst 7(1):75–98
Eugster SR, Glocker C (2013) Constraints in structural and rigid body mechanics: a frictional
contact problem. Annals Solid Struct Mech 5(1–2):1–13
Evans KE (1991) Auxetic polymers: a new range of materials. Endeavour 15(4):170–174
Evans KE, Alderson A (2000) Auxetic materials: functional materials and structures from lateral
thinking!. Adv Mater 12(9):617–628
Germain P (1973) The method of virtual power in continuum mechanics. Part 2: Microstructure.
SIAM J Appl Math 25(3):556–575
Germain P (2020) The method of virtual power in the mechanics of continuous media, I: second-
gradient theory. Math Mechan Complex Syst 8(2):153–190
Giorgio I, Rizzi NL, Andreaus U, Steigmann DJ (2019) A two-dimensional continuum model of
pantographic sheets moving in a 3D space and accounting for the offset and relative rotations of
the fibers. Math Mech Complex Syst 7(4):311–325
Giorgio I, Ciallella A, Scerrato D (2020) A study about the impact of the topological arrangement
of fibers on fiber-reinforced composites: some guidelines aiming at the development of new
ultra-stiff and ultra-soft metamaterials. Int J Solids Struct 203:73–83
Greco F, Luciano R, Serino G, Vaiana N (2018) A mixed explicit-implicit time integration approach
for nonlinear analysis of base-isolated structures. Annals Solid Struct Mech 10(1):17–29
Greco L, Cuomo M (2013) B-Spline interpolation of Kirchhoff-Love space rods. Comput Methods
Appl Mech Eng 256:251–269
Greco L, Cuomo M (2014) An implicit G1 multi patch B-spline interpolation for Kirchhoff-Love
space rod. Comput Methods Appl Mech Eng 269:173–197
Hesch C, Schuß S, Dittmann M, Eugster S, Favino M, Krause R (2017) Variational space-time
elements for large-scale systems. Comput Methods Appl Mech Eng 326:541–572
Hild F, Misra A, dell’Isola F (2020) Multiscale DIC applied to pantographic structures. Exper Mech
Huang J, Boisse P, Hamila N, Zhu Y (2020) Simulation of wrinkling during bending of composite
reinforcement laminates. Materials 13(10):2374
Huang J, Boisse P, Hamila N, Gnaba I, Soulat D, Wang P (2021) Experimental and numerical
analysis of textile composite draping on a square box, influence of the weave pattern. Compos
Struct 113844
Jafarzadeh H, Farrahi GH, Levitas VI, Javanbakht M (2020) Phase field theory for fracture at large
strains including surface stresses. arXiv preprint arXiv:201113324
Khakalo S, Niiranen J (2020) Anisotropic strain gradient thermoelasticity for cellular structures:
Plate models, homogenization and isogeometric analysis. J Mech Phys Solids 134(103):728
Kron G (1945) Numerical solution of ordinary and partial differential equations by means of equiv-
alent circuits. J Appl Phys 16(3):172–186
Lagrange JL (1806) Traité de la résolution des équations numériques de tous les degrés: avec des
notes sur plusieurs points de la théorie des équations algébriques. chez Courcier
Lagrange JL (1853) Mécanique analytique, vol 1. Mallet-Bachelier, Paris
266 M. Spagnuolo and E. Barchiesi

Laudato M, Barchiesi E (2019) Non-linear dynamics of pantographic fabrics: modelling and numer-
ical study. In: Wave dynamics, mechanics and physics of microstructured metamaterials. Springer,
pp 241–254
Laudato M, Manzari L, Barchiesi E, Di Cosmo F, Göransson P (2018) First experimental observation
of the dynamical behavior of a pantographic metamaterial. Mech Res Commun 94:125–127
Laudato M, Manzari L, Scerrato D, Göransson P, Giorgio I (2020) Spectral properties of 2d panto-
graphic metamaterial: experimental results. Mech Res Commun 109(103):613
Mablekos E, Weidmann H (1968) Realization of negative impedance inverter using Kron’s mesh
primitive-branch approach. J Franklin Instit 286(1):59–75
Marmo F, Masi D, Rosati L (2018a) Thrust network analysis of masonry helical staircases. Int J
Architec Heritage 12(5):828–848
Marmo F, Ruggieri N, Toraldo F, Rosati L (2018b) Historical study and static assessment of an
innovative vaulting technique of the 19th century. Int J Architec Heritage
Marmo F, Demartino C, Candela G, Sulpizio C, Briseghella B, Spagnuolo R, Xiao Y, Vanzi I, Rosati
L (2019) On the form of the musmeci’s bridge over the basento river. Eng Struct 191:658–673
Maugin GA (2011) A historical perspective of generalized continuum mechanics. In: Altenbach H,
Maugin G, Erofeev V (eds) Mechanics of generalized continua. Advanced structured materials,
vol 7. Springer, Berlin, Heidelberg, pp 3–19
Maurin F, Greco F, Desmet W (2019) Isogeometric analysis for nonlinear planar pantographic
lattice: discrete and continuum models. Contin Mech Thermodyn 31(4):1051–1064
Mindlin RD (1965) Second gradient of strain and surface-tension in linear elasticity. Int J Solids
Struct 1(4):417–438
Misra A, Poorsolhjouy P (2020) Granular micromechanics model for damage and plasticity of
cementitious materials based upon thermomechanics. Math Mech Solids 25(10):1778–1803
Misra A, Lekszycki T, Giorgio I, Ganzosch G, Müller WH, Dell’Isola F (2018) Pantographic
metamaterials show atypical poynting effect reversal. Mech Res Commun 89:6–10
Namnabat MS, Barzegar A, Barchiesi E, Javanbakht M (2020) Nonlinear buckling analysis of
double-layered graphene nanoribbons based on molecular mechanics. Carbon Lett 1–16
Paradiso M, Marmo F, Rosati L (2019) Consistent derivation of a beam model from the saint
Venant’s solid model. Int J Solids Struct 159:90–110
Perricone V, Grun T, Marmo F, Langella C, Carnevali MDC (2020) Constructional design of echi-
noid endoskeleton: main structural components and their potential for biomimetic applications.
Bioinspiration Biomimetics
Pideri C, Seppecher P (1997) A second gradient material resulting from the homogenization of an
heterogeneous linear elastic medium. Contin Mech Thermodyn 9(5):241–257
Piola G (1846) Intorno alle equazioni fondamentali del movimento di corpi qualsivogliono, con-
siderati secondo la naturale loro forma e costituzione: memoria, vol 10282. BD Camera
Placidi L (2015) A variational approach for a nonlinear 1-dimensional second gradient continuum
damage model. Contin Mech Thermodyn 27(4–5):623–638
Placidi L, Barchiesi E, Misra A (2018) A strain gradient variational approach to damage: a compar-
ison with damage gradient models and numerical results. Math Mech Complex Syst 6(2):77–100
Placidi L, dell’Isola F, Barchiesi E (2020) Heuristic homogenization of Euler and pantographic
beams. In: Picu C, Ganghoffer JF (eds) Mechanics of fibrous materials and applications. CISM
International Centre for Mechanical Sciences (Courses and Lectures), vol 596. Springer, Cham,
pp 123–155
Rahali Y, Giorgio I, Ganghoffer J, dell’Isola F (2015) Homogenization à la Piola produces second
gradient continuum models for linear pantographic lattices. Int J Eng Sci 97:148–172
Rahbar H, Javanbakht M, Ziaei-Rad S, Reali A, Jafarzadeh H (2020) Finite element analysis of
coupled phase-field and thermoelasticity equations at large strains for martensitic phase trans-
formations based on implicit and explicit time discretization schemes. Mech Adv Mater Struct
1–17
Scerrato D, Giorgio I (2019) Equilibrium of two-dimensional cycloidal pantographic metamaterials
in three-dimensional deformations. Symmetry 11(12):1523
13 Do We Really Need Pantographic Structures? 267

Schulte J, Dittmann M, Eugster S, Hesch S, Reinicke T, Dell’Isola F, Hesch C (2020) Isogeometric

analysis of fiber reinforced composites using kirchhoff-love shell elements. Comput Methods
Appl Mech Eng 362(112):845
Seppecher P (1989) Etude des conditions aux limites en théorie du second gradient: cas de la
capillarité. Comptes rendus de l’Académie des sciences Série 2, Mécanique, Physique, Chimie,
Sciences de l’univers, Sciences de la Terre 309(6):497–502
Seppecher P (2000) Second-gradient theory: application to Cahn-Hilliard fluids. In: In: Maugin
G.A., Drouot R., Sidoroff F. (eds) Continuum thermomechanics. Solid mechanics and its appli-
cations, vol 76. Springer, Dordrecht, pp 379–388
Seppecher P, Alibert JJ, dell’Isola F (2011) Linear elastic trusses leading to continua with exotic
mechanical interactions. J Phys Conf Series 319(1):012,018
Serpieri R, Sessa S, Rosati L (2018) A mitc-based procedure for the numerical integration of
a continuum elastic-plastic theory of through-the-thickness-jacketed shell structures. Compos
Struct 191:209–220
Sessa S, Marmo F, Rosati L (2015) Effective use of seismic response envelopes for reinforced
concrete structures. Earthq Eng Struct Dyn 44(14):2401–2423
Sessa S, Serpieri R, Rosati L (2017) A continuum theory of through-the-thickness jacketed shells for
the elasto-plastic analysis of confined composite structures: Theory and numerical assessment.
Compos Part B Eng 113:225–242
Sessa S, Marmo F, Rosati L, Leonetti L, Garcea G, Casciaro R (2018a) Evaluation of the capacity
surfaces of reinforced concrete sections: Eurocode versus a plasticity-based approach. Meccanica
53(6):1493–1512
Sessa S, Marmo F, Vaiana N, Rosati L (2018b) A computational strategy for eurocode 8-compliant
analyses of reinforced concrete structures by seismic envelopes. J Earthq Eng 1–34
Sessa S, Marmo F, Vaiana N, De Gregorio D, Rosati L (2019a) Strength hierarchy provisions for
transverse confinement systems of shell structural elements. Compos Part B Eng 163:413–423
Sessa S, Marmo F, Vaiana N, Rosati L (2019b) Probabilistic assessment of axial force-biaxial
bending capacity domains of reinforced concrete sections. Meccanica 54(9):1451–1469
Spagnuolo M (2020) Circuit analogies in the search for new metamaterials: phenomenology of
a mechanical diode. In: Altenbach H, Eremeyev V, Pavlov I, Porubov A (eds) Nonlinear wave
dynamics of materials and structures. Advanced structured materials, vol 122. Springer, Cham,
pp 411–422
Spagnuolo M, Scerrato D (2020) The mechanical diode: on the tracks of James Maxwell employing
mechanical–electrical analogies in the design of metamaterials. In: Abali B, Giorgio I (eds)
Developments and novel approaches in biomechanics and metamaterials. Advanced structured
materials, vol 132. Springer, Cham, pp 459–469
Spagnuolo M, Barcz K, Pfaff A, dell’Isola F, Franciosi P (2017) Qualitative pivot damage analysis in
aluminum printed pantographic sheets: numerics and experiments. Mech Res Commun 83:47–52
Spagnuolo M, Peyre P, Dupuy C (2019) Phenomenological aspects of quasi-perfect pivots in metallic
pantographic structures. Mech Res Commun 101(103):415
Turco E, Barchiesi E (2019) Equilibrium paths of Hencky pantographic beams in a three-point
bending problem. Math Mech Complex Syst 7(4):287–310
Turco E, Giorgio I, Misra A, Dell’Isola F (2017) King post truss as a motif for internal structure of
(meta) material with controlled elastic properties. R Soc Open Sci 4(10):171,153
Turco E, Misra A, Pawlikowski M, dell’Isola F, Hild F (2018) Enhanced Piola-Hencky discrete mod-
els for pantographic sheets with pivots without deformation energy: numerics and experiments.
Int J Solids Struct 147:94–109
Turco E, Dell’Isola F, Misra A (2019a) A nonlinear Lagrangian particle model for grains assemblies
including grain relative rotations. Int J Numer Anal Methods Geomech 43(5):1051–1079
Turco E, Misra A, Sarikaya R, Lekszycki T (2019b) Quantitative analysis of deformation mech-
anisms in pantographic substructures: experiments and modeling. Contin Mech Thermodyn
31(1):209–223
268 M. Spagnuolo and E. Barchiesi

Vaiana N, Spizzuoco M, Serino G (2017) Wire rope isolators for seismically base-isolated
lightweight structures: experimental characterization and mathematical modeling. Eng Struct
140:498–514
Vaiana N, Sessa S, Marmo F, Rosati L (2019) Nonlinear dynamic analysis of hysteretic mechanical
systems by combining a novel rate-independent model and an explicit time integration method.
Nonlinear Dyn 98(4):2879–2901
Vaiana N, Capuano R, Sessa S, Marmo F, Rosati L (2021) Nonlinear dynamic analysis of seismically
base-isolated structures by a novel opensees hysteretic material model. Appl Sci 11(3):900
Yang H, Ganzosch G, Giorgio I, Abali BE (2018) Material characterization and computations of a
polymeric metamaterial with a pantographic substructure. Zeitschrift für angewandte Mathematik
und Physik 69(4):105
Yang H, Abali BE, Timofeev D, Müller WH (2019) Determination of metamaterial parameters by
means of a homogenization approach based on asymptotic analysis. Contin Mech Thermodyn
1–20
Yildizdag ME, Barchiesi E, dell’Isola F (2020) Three-point bending test of pantographic blocks:
numerical and experimental investigation. Math Mech Solids 25(10):1965–1978