Frank U.

Rückert
Michael Sauer
Tuomo Liimatainen
Dirk Hübner

Digital Twin
Development
An Introduction to Simcenter Amesim
Digital Twin Development
Frank U. Rückert · Michael Sauer ·
Tuomo Liimatainen · Dirk Hübner

Digital Twin Development

An Introduction to Simcenter Amesim
Frank U. Rückert Michael Sauer
Fluid Energy Machines Fluid Machinery and Measurement
University of Applied Sciences Saarbrücken University of Applied Sciences Saarbrücken
Saarbrücken, Saarland, Germany Saarbrücken, Saarland, Germany

Tuomo Liimatainen Dirk Hübner

Mechanical Engineering Lightweight Construction
LAB University of Applied Sciences University of Applied Sciences Saarbrücken
Lappeenranta, Etelä-Suomi, Finland Saarbrücken, Saarland, Germany

ISBN 978-3-031-25691-2 ISBN 978-3-031-25692-9 (eBook)

https://doi.org/10.1007/978-3-031-25692-9

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Switzerland AG 2023
This work is subject to copyright. All rights are solely and exclusively licensed 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
Preface

This book is dedicated to our families who gave us time together for writing
it. We worked on it from winter semester of 2020 until summer of 2022 at the
University of Applied Sciences Saarbrücken in Germany. Additionally, a seminar on
this topic was conducted at the LAB University of Applied Sciences in Lappeen-
ranta, Finland, during the summer session in order to get feedback from inter-
national English-speaking students. To assess in-course progress, students were
asked to provide feedback, suggest outcome improvements, and avoid any possible
misunderstandings.
The authors have many years of experience in creating sophisticated simulation
models for complex, technical solutions in automotive industry and power plant
research. In order to improve further development in the field of renewable energies,
we added examples of this subject area. For us, it was important that the students
could easily start building their own digital twins for complex technical systems.
It was important that students explore the various parts of the model or initially
develop uncomplicated systems to examine the behavior in a user-friendly environ-
ment thereby learning a step-by-step approach. The graphical visualization of the
innovation helped to analyze and solve technical problems much faster.
The software tool, Simcenter Amesim, was provided free of charge by Siemens
for study purposes, increasing time efficiency and minimizing challenges through
modeling. Networking and collaboration during the development process facilitated
the exchange of information effortlessly. As model-based engineering continues to
grow, it is important to start educating young ones early on.
Although our book was written for students of natural sciences, engineering, or
information technologies, the language is deliberately kept easy to understand for
anyone who is interested in the development of digital twins for physical problems
and/or wants to start modeling. Especially for students, the early possibility of a free,
easy-to-understand simulation environment can enrich physics lessons at school and
contribute significantly to the understanding of technical relationships of the models.
Our book takes into account the feedback given to us by the students and will help
especially novices to find alternative ways to use the simulation tools. We expect that

v
vi Preface

the book will help to make the subject easy to understand and that the reader will find it
educational and enjoyable when developing new technical solutions and innovations.

Saarbrücken, Germany Frank U. Rückert

October 2022
Acknowledgements

It is anticipated that interested readers will have as much fun reading this book and
playing with the models as we had during writing the chapters and compiling the
simulation examples. Our aim is that students will have fun while modeling easy
solutions without struggling because of antiquated and unnecessary mathematical
doctrine and quibbles.
We would like to thank our long-time industrial companions Stephan Wursthorn,
Christian Steinbrecher, Heiko Roth, Gerhard Sünderhauf, Christoph Magel, Thilo
Klingel, Martin Katz, and Sibel Yilmaz from Robert Bosch GmbH. We also want to
thank Rüdiger Thieman and Rolf Hartge for their support. During many discussions,
they taught us a lot, while we spent hours working together with Simcenter Amesim
on technical problems and innovations, we drank a cup or two of tea or coffee.
We thank our faithful and hardworking assistants and students at the University of
Applied Sciences Saarbrücken, namely Philipp Spindler, Benjamin Allweyer, Sebas-
tian Grün, Emile Degro, Xenia Hülsmann, Tarek Khiar, Daniel Lehser-Pfeffermann,
Alexander Hamman, Johannes König, Yannick Planta, Badre Ait Amghar, Tim
Dennemärker, Barbara Kaiser, Nils Kleiber, Jan Molter, Christian Boiger and Tim
Breuer, who created many simulation models and examples for this work. Tim Breuer
passed away during the Corona pandemic, and we will always remember him and
his service fondly. I want to thank the students who participated in our lectures
about Digital Twins during Summer School 2021 and 2022 at LAB University of
Applied Sciences. We would like to thank the staff of the International Office in Saar-
brücken and Lappeenranta, namely Doris Kollmann, for the financial support within
the Erasmus+ program and the European Union, we hope this will be possible again
in the near future. We want to thank Michael Sandoval for reading the manuscript. His
positive feedback was an enrichment. We would also like to thank Claudio Santarelli
and Helge Tielbörger from Siemens AG for their kind support, help, and good advice.
Last but not least, a big thank you to our friends at Lake Saimaa in Lappeenranta,
Finland, who were such good hosts during wartime. Thanks to Tuomo for giving me
the chance for holding my lecture at LAB. It means a lot to me that you gave my
mother a moose antler. Finally, we would like to thank Michael Sauer for repairing
the boat.

vii
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Motivation from History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 What Is a Digital Twin? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 How to Create a Twin? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Adding Physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5 Analysis of Digital Twins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2 Mathematics, Signals and Control Library . . . . . . . . . . . . . . . . . . . . . . 13
2.1 The First Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 How to Build a Simple Calculator . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3 The Mechanical Twin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 What Is Mechanics? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 The Model of a Bouncing Ball . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 The Mechanical Rocker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4 How a Row on a Pulley Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.4.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.5 Damper of a Driving Car . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

ix
x Contents

3.5.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.5.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.6 How to Use 3D Geometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4 The Thermal Twin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1 Examination of Heat Transfer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Heat Transfer in Electric Generators . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Site Selection for a Solar Collector . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.3.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5 The Hydraulic Twin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.1 What Is Hydraulics? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.2 Two Fuel Oil Tanks and a Pump . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
5.2.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.2.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.2.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 How Does a Hydraulic Jack Work? . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5.3.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6 The Pneumatic Twin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.1 Pneumatic, Fluid Flow, and Turbulence . . . . . . . . . . . . . . . . . . . . . . 71
6.2 Safety Valve for a Biogas Tank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.2.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
6.2.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.2.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.3 Ventilation System of a Building . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.3.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
6.3.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.3.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
Contents xi

Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7 The Electric Twin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.1 Permanent Electric Motor with Load . . . . . . . . . . . . . . . . . . . . . . . . 81
7.1.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.1.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.1.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7.2 How an Asynchronous Motor Works? . . . . . . . . . . . . . . . . . . . . . . . 85
7.2.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7.2.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7.2.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
7.3 Electric Generator with Resistor and Heat Transfer . . . . . . . . . . . . 88
7.3.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.3.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 89
7.3.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8 Analysis of Complex Technical Systems . . . . . . . . . . . . . . . . . . . . . . . . . 95
8.1 How Does the Liquid Piston Compressor Work? . . . . . . . . . . . . . . 95
8.2 Design and Function of a Liquid Piston Compressor . . . . . . . . . . . 95
8.3 Liquid Piston Compressor with One Cylinder for One Stroke . . . 96
8.3.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
8.3.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.3.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.4 Liquid Piston Compressor with One Cylinder and Multiple
Strokes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.4.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.4.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 103
8.4.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
8.5 Liquid Piston Compressor with Two Cylinder and Multiple
Strokes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
8.5.1 Simulation Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
8.5.2 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8.5.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
9 Digital Twins and Artificial Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . 111
9.1 Neural Networks in Nature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
9.2 Neural Networks and Digital Twins . . . . . . . . . . . . . . . . . . . . . . . . . 111
9.3 The Artificial Frog Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
9.3.1 Submodels and Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 115
xii Contents

9.3.2 Optiflow Neural Network (ONN) . . . . . . . . . . . . . . . . . . . . . 117

9.3.3 Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
10.1 Overview of This Textbook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
10.2 What You Can Take Away . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
10.3 Teaching Methods and Gamification . . . . . . . . . . . . . . . . . . . . . . . . 123
10.4 Our Outlook for the Future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
10.5 Important: Disclaimer for Our Work . . . . . . . . . . . . . . . . . . . . . . . . 124
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Chapter 1
Introduction

1.1 Motivation from History

What could early humans have done to plan ahead for their problem-solving to save
resources and energy? One first step could be to draw a situation in order to become
familiar with different aspects of the problem. Figure 1.1 shows such a situation. It’s
a picture painted after early petroglyphs located at Bryce Canyon National Park in
Utah (USA). It shows a hunter on his horse and a deer [1]. Similar illustrations can be
found all over the world. If you look closely at the scene, you can almost imagine the
arrow moving. The purpose of problem visualization was used to describe and solve
a specific situation. Additionally, the picture could be used to teach younger hunters
with less experience. The petroglyph also depicts technical equipment. Drawing on
a wall can be considered as an early kind of simulation or analysis of a specific
situation. Today, the first thing students are advised to do is to draw a sketch of a
technical device. Visualizing a problem at least helps to analyze and solve it. We
believe that the petroglyph scene was for teaching or self-reflection. Knowing what
to do is as important today as it was in the past.
What we don’t see in Fig. 1.1 is that there are connections between the different
objects. They are not painted, but in fact, there is a story and a connection between
the different symbols and people. Two larger persons with antlers on their heads
could also be identified as supervisors evaluating the situation and the quality of
the hunting scene. We want to demonstrate the idea behind simulation models for
problem-solving processes and how we can create artificial twins for any kind of
problem. The core problem is that today’s technical solutions are becoming more
and more complicated. More details are included, leading to unexpected behavior.
Simulation models can help understand your work, but—beware—they can also
make it worse because they make the system more complex. Too much ingenuity
can very quickly lead to failure. An engineer has to understand the aim and purpose
of his simulation to answer the question why. After that, answer to question how
is often more trivial. At the end of this book, in the Conclusions chapter, we will

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 1

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_1
2 1 Introduction

Fig. 1.1 Sketch after a scene from petroglyphs in Bryce Canyon National Park, USA [1]

discuss further methods for teaching and introduce an additional new idea on how to
teach complex technical issues very quickly by introducing a role play method.
During creating the so-called digital twins, we wanted to introduce the simula-
tion tool Simcenter Amesim, which can be used to simulate the behavior of systems.
The software we use is from Siemens PLM Software [2]. A free-of-charge demo
version can be installed without much effort on Microsoft Windows or Linux sys-
tems. Working on Apple operation systems, one can use the program via a virtual
private network (VPN) connection. Of course, this is less convenient. The software
tool contains a ready-made library, including smaller models to simulate pneumatic,
hydraulic, thermal, electrical, and mechanical behavior. It additionally includes a
theoretical library for signals, mathematics, and artificial intelligence, e.g. artificial
neural networks (ANN) or reduced order modeling (ROM).
We will explain the basic functions and how to start modeling with this software
tool. It is considered as a multi-domain software, meaning that all the different
physical domains can be linked with each other or with the signal library. The software
offers the possibility of simulating systems without the need for computer-aided
design (CAD) geometry, but 3D geometries can be added and controlled by the
models if this is needed. These models can also be printed with a 3D printer. In
former times, small models made of wood or clay were built to study the behavior
of systems. Today, our students are fascinated by 3D printing, which plays a similar
role. The driver is similar to that of the petroglyphs, and is why we use it for a clearer
understanding.
The usable symbols for the different libraries in Simcenter Amesim are partly
standard symbols defined and therefore easy to recognize. The software allows it to
interact with other programs such as Excel or MATLAB/Simulink. One big advantage
is, if compared, for example, with Simulink. Many industrial models are already
finished and can be used as reference cases from the onset.
1.2 What Is a Digital Twin? 3

In order to bring the basics closer in the study, it was necessary to plan a lecture
with the contents, e.g. the artificial model construction and the practical application
of the used software. As previously described, the fascination lies not only in the
manual creation of the simulation model but also in the possibilities that arise when
one prototype has to be constructed. With a digital twin, technical, physical, or
biological components can be tested long before they are used. The digital twin must
reproduce certain properties of its realistic model to allow its functions and behavior
to be scrutinized.
The earlier in the development process the engineer has the opportunity to make
quantitative statements about the function of a component in the system, the more
effective a later design or development process can be. The digital twin is the image
of its real counterpart and represents the first step in the development process. In our
daily work experience, there is a difference between creating and using a numerical
model. Even though the creation of the twin requires an expert who must know the
process to be mapped down to the smallest detail, the user of the digital twin does
not need to understand everything in detail in order to work with the twin. We use
the example of the light switch. You do not have to understand how it works to use
it. The gap between creation and use should be bridged by this book. Possibly this
was also one idea behind the petroglyph scene in Fig. 1.1.
The relevance of this topic in today’s world is obvious. Also in the most diverse
areas of product development. Today, the topic of sustainability also encompasses
the behavior of a product over its lifetime cycle. This means that a simulation model
should not only make a statement about how a technical product works and which
functionality in the various phases of its life but also how large an effect really is and
what its consequences are is only known to a very limited extent. Today, for example,
a very small amount of a powder containing anthrax virus can start a war. Quantities
of harmless gas that are only in the ppm range (ppm = parts per million) can bring an
entire industry to a standstill. On the other hand, sometimes small amounts of a virus
are enough to have a very large effect. In such a case, the complexity of the overall
system is crucial. When we build systems more complex, they get more susceptible
to small quantities.

1.2 What Is a Digital Twin?

When creating the model or digital twin from reality, it should be noted that the
original cannot be reproduced in its entirety. Rather, certain properties are to be
reproduced in isolation, i.e. in digital form as in Fig. 1.2. What is meant by this?
We can explain it by the example of a human body or better with a prometheus, as
described in [3].
In the case of this human being, like in Fig. 1.2, it can be possible to digitally
reproduce only the blood vessels and thus create a kind of hydraulic twin. Such a
twin would not provide any information about the bone structure. To represent the
statics of the bones of the human being, one would have to create a mechanical twin.
4 1 Introduction

Fig. 1.2 Digital twin is the counterpart to the real twin, but with limited capabilities

The original can therefore have different digital twin brothers, which only partially
reflect the physical or geometric properties of their real counterpart.
A thermal twin could, for example, allow conclusions to be drawn about whether
the subject has a fever or how quickly he cools down in winter, but not whether
there is a fracture in the thigh. The pneumatic twin could, for example, provide
conclusions about lung function or movements of other compressible fluids inside
the body. For the human nervous system, there is a so-called signal database for
neural networks. To control the digital twin, the nervous system is needed, just like
in real life. This shows how important the idea of artificial intelligence (AI) is. To
explain in detail how to work with it in Amesim, we will first start in the following
chapter by rebuilding a simple calculator with the signal database. At the end, we
explain what a neural network is and how it can be used to control our twin.
As can be seen in Fig. 1.3 the digital twin is a virtual replica consisting of data and
algorithms of a real object or product. This replica can be a model, a simulation, or
even an algorithm that emulates the behavior of real objects. this offers the advantages
of testing and simulating systems and products in advance. Parameters of the virtual
image are set and tested, the results can be output and parameters can be adjusted
quickly. The goal is to simplify planning and thus save time and costs.
Application types of a digital twin can be distinguished depending on the product
life phase. In the planning phase, the digital product twin is the preliminary product
a computer-aided design (CAD) or 3D model. The model helps to examine the
behavior of the product in different phases to make forecasts. A virtual model can be
quickly adapted and also respond to customer requests. The digital production twin
is used to forecast efficiency and quality in manufacturing. Furthermore, the virtual
variant can be used to test deliveries and the availability of materials in advance.
1.3 How to Create a Twin? 5

Fig. 1.3 Different use cases can lead to different digital twins

All digital twins can be divided into several archetypes. Van der Valk et al. divided
them into five different categories or so-called archetypes to compare the degree of
evolution [4]. In Table 1.1, all archetypes are presented.

1.3 How to Create a Twin?

As described above, we provide the reader with the opportunity for an easy intro-
duction to the topic. As shown in Fig. 1.4, various technical systems are, e.g. simple
fluid energy machines, which can initially be understood as a system model. Fig. 1.4
illustrates that components from different disciplines can be combined and thus built
up into complex, cyber-physical systems. Thus, the technical system can then include
thermal, pneumatic, mechanical as well as hydraulic or electric components. Once
the system has been set up, it is possible to study the behavior of, for example, a fluid
energy machine over a period of time. The goal is to be able to perform a virtual
investigation in an isolated space with it.
6 1 Introduction

Table 1.1 Definitions of the archetypes of digital twins [4]

Archetype Description
Basic Digital Twin This digital twin provides an human-machine
interface (HMI). It therefore extends the digital twin
by a single feature and is therefore considered
fundamental
Enriched Digital Twin Based on basic archetype, this digital twin enriches
its database by preprocessed data from
supplementary systems. Many objects in this cluster
describe the possibility of semi-manual data
collection
Autonomous Control Twin This twin is an advancement from enriched digital
twin. It offers autonomous control, but at the same
time, it contains an HMI for the option to intervene.
As direct communication with another virtual or
physical machine, this archetype needs at least
interoperability via a translator interface
Enhanced Autonomous Control Twin The digital twin offers autonomous control over a
physical asset while integrating external,
downstream data processing systems. The
interoperability needs to be secured at least over a
translator as well. On the downside, this archetype
does not offer an HMI
Exhaustive Twin A digital twin with exhaustive data acquisition
options, data processing, and control over a physical
asset. This archetype provides the user with all
options. The twin is able to work and control
autonomously. At each point, humans have the
ability to intervene or to enrich the database and,
hence, provide a semi-manual data linkage. This
archetype demands a fully interoperable data linkage
to downstream systems as well as to the physical
asset itself

As already described, no programming skills are required for the creation of the
twins, and we will not dwell on mathematical subtleties. Mathematics should be sub-
ordinated to the development process as an auxiliary science and not pursued as an
end in itself. We want to get results as quickly as possible. The creation of the phys-
ical model is facilitated in the platform by numerous well-documented examples. A
very detailed illustrated documentation of the theoretical basics is also available. The
company Siemens decided to make its development tool Simcenter Amesim available
free of charge. It gives the possibility to create digital twins as preliminary stage of
the design of a new product with little effort. The program can be used by university
staff and students free of charge. Integration into the teaching process and promo-
tion of young academics is thus facilitated. A license must be purchased as soon as
Amesim is used commercially.
1.3 How to Create a Twin? 7

Fig. 1.4 Example of combined models: propeller, mechanical friction, solar collector, and hydraulic
pump

Important: You can test the program before you buy it, find a free student
version of Simcenter Amesim at the following link:
https://www.plm.automation.siemens.com/plmapp/education/simcenter/
en_us/free-software/student

Paths for software can change, and books are usually more persistent than websites
on the Internet. Therefore, the specified path may no longer be correct. In this case,
try to find the described software using a common search engine. The free-of-charge
version from Siemens has some restrictions such as limitation of the library. With
this student version, the tasks were designed for a lecture plan of this textbook. Only
the available libraries are explained in more detail below and all examples could be
done with the free version.
The user interface of Simcenter Amesim illustrated in Fig. 1.5 is mostly operated
graphically with the mouse. Here, the graphical program interface is illustrated with
the digital twin of a bouncing ball. The program offers a tidy interface, which is
initially divided into a workbench, as well as different library windows. On the
workbench, different items can be dragged from the libraries with the mouse and
connected with each other in the so-called sketch mode. The time period over which
the digital twin is to be observed in a virtual laboratory is then defined.
8 1 Introduction

Fig. 1.5 Graphical user interface of Simcenter Amesim offers a clear overview

1.4 Adding Physics

All subsequent chapters of this book are structured in the same way for better clarity
and how to work with it. The chapters are always divided, first into the section Simu-
lation model where the technical system has to be sketched by the students with the
mouse on the workbench. This first step should be followed by a section on the text
page Submodels and Parameters where the exact physical quantities for the simula-
tion model must be specified. Here, you need to go into detail. Finally, there is always
a section presenting the simulation results and a reference to further work proposals,
where modeling, simulation, monitoring and mirroring are essential [5]. To create
the model or simulation network, Simcenter Amesim provides an integrated simula-
tion platform for cyber-physical problems and multidisciplinary system simulation
at different stages of innovation and life cycle and could be divided into autonomous,
federated, and digital twin maturity models.

Important: Please do not focus too much on detailed physical quantities and
variables at the beginning. The overview of the problem is more important.
Parameters and boundary conditions can be changed later.

We have several categories of libraries available to share below. Modeling with

Simcenter Amesim takes place in four steps. First, a closed model must be created
using the components from the existing libraries. For this purpose, it is also important
to define used fluids or media. In the Submodel mode, the physical model is selected,
1.4 Adding Physics 9

Table 1.2 Definitions of different modes in Simcenter Amesim

Mode Description
Sketch mode Drawing the model on the worksheet
Submodel mode Selection of the different submodels from the libraries
Parameter mode Detailed definition of the physical parameters
Simulation mode Run and review the simulation and graph the results

which is assigned to each component. The third step is processed in the Parameter
mode as given in Table 1.2.
Between the step from Submodel mode to Parameter mode, a compilation of the
model takes place. The individual mathematical models become an overall model
description. In Parameter mode, the required parameters for the physics of the com-
ponents are set. Here, you have to take care about the quantity of the models. For the
simulation, appropriate settings can be made such as the time, resolution, or whether
the calculations should be dynamic or stable behavior. After that, the simulation run
can be started in the Simulation mode and the results can be analyzed.
To add physics for the technical problem we have to choose the right physical
library. Each library has its own color. The signal library is usually illustrated in red
color. We want to take the chance to introduce the functions of Amesim in detail.
• Signals and Control library The signal library contains different mathematical
functions and is used to control the other libraries. Basic mathematical functions
are available. We will explain in the next chapter how to use them. This library
contains artificial intelligence models.
• Mechanical library This category contains linear and rotary motion elements.
It is often used to complement other library categories to simulate mechanical
ratios. Components such as masses, springs, dampers, friction elements, and also
transmission ratios can be found in this section.
• Hydraulic library This section contains general hydraulic components such as
pumps, pipes, and valves for the simulation of the dynamic behavior of fluids.
• Pneumatic library It contains components that are used to model compressed
air-based systems. Typical components include pressure and flow sources, valves,
or gas properties.
• Thermal library With the thermal library, heat states and interactions can be sim-
ulated. The thermal components such as convection, conduction, or even radiation
elements enable calculations of heat flow and losses.
• Electric library In this library are electrical components with which systems can
be reproduced, especially for automotive electronics. Components are resistors,
inductors, capacitors, batteries, and also already prefabricated motors.
The applications of the presented program are described in more detail in further sec-
tions. The color of the items can also be changed by right-clicking it with the mouse.
But at the beginning, we should leave it in the default color for better understanding.
10 1 Introduction

1.5 Analysis of Digital Twins

Various classical mathematical methods can be used for the analysis of digital twins.
A 3D geometry helps to understand the large amount of data through visualization. An
example of such a digital twin coupled with a 3D geometry is illustrated in Fig. 1.6.
The figure shows the model of a transporter chassis driving over an obstacle. For
this purpose, these hydraulic and mechanical models must be further developed and
certain skills must be trained by algorithms. The algorithms can be used to control
the 3D models in a virtual space, e.g. the transporter has a drive unit and a hydraulic
damping system. It has the task of driving over a stone sidewalk which produces a
lot of data. It should be mentioned that the amount of data can be high, but not so
high that we should call it Big Data [6]. The figure shows how the chassis drives
over the stone, and its functionality can be tested under real conditions.
Due to its hydraulic suspension and the pneumatic model of the tires, the trans-
porter succeeds in doing so and is able to drive independently from one side of the
slab to the other. The physics of the hydraulic damping system is realistically repro-
duced according to physical principles. Using the data for training certain skills such
as structural thinking is covered by the terms machine learning or deep learning.
These are two sub-fields of artificial intelligence (AI). This would allow the chassis
model to learn how best to drive over the obstacle. While machine learning is the
generic term for allowing an algorithm to learn further by feeding it data, deep learn-
ing is a method of machine learning that builds on artificial neural networks. We will

Fig. 1.6 3D geometry and damping of a transporter chassis can be tested under realistic conditions
References 11

describe this in detail in the last chapter of this book. Machine learning, by feeding
structured data, can create an algorithm that can analyze new information based on
the training. Deep learning can handle large unstructured data but requires powerful
computers to do so. Artificial intelligence technologies offer a wide range from weak
AIs that are used to solve limited tasks to strong AIs that emulate human intelligence
[7]. Their use is intended to be a support for humans in the industrial, or also in
the domestic area, to deal with information overload and to work autonomously.
In the production process, for example, there is potential for this in research and
development, in-process optimization, or also for maintenance systems for machine
maintenance [7].
To create an accurate virtual image and thus minimize subsequent production
errors, it is advantageous to enter as much data as possible. Digital twins could assist
to generate this data. To deal with a large amount of data, data processing techniques
for Big Data will become necessary [6]. It is used to present the information so that
it is available quickly and can be used effectively by companies. It must be able to
quickly capture and collect the data before it can be processed further.

References

1. Hiser, D.: Records of the Environmental Protection Agency, The Environmental Protection
Agency’s Program to Photographically Document Subjects of Environmental Concern, 1972–
1977. https://catalog.archives.gov/id/545671
2. Siemens AG: Simcenter Amesim. https://www.plm.automation.siemens.com/
3. Shelley Wollstonecraft, M.: Frankenstein, or, The Modern Prometheus. Oxford University Press,
New York (1818)
4. Hendrik van der Valk, H., Hasse, H., Möller, F., Otto, B.: Archetypes of digital twins. Bus. Inf.
Syst. Eng. 64(3), 375–391 (2022)
5. Kim, Y.-W.: Digital Twin maturity model, WEB 3D 2020 industrial use cases workshop on
digital twin visualization. https://doi.org/10.13140/RG.2.2.28750.48967 (2020)
6. Weisberg Business Consulting GmbH: Einfach erklärt: Was ist Big Data? https://weissenberg-
solutions.de/einfach-erklaert-was-ist-big-data/
7. Smith, S.E.: Neuromuscular junction. In: Zaimis, E. (ed.) Handbook of Experimental Pharma-
cology, vol. 42, p. 593. Springer, Heidelberg (1976)
Chapter 2
Mathematics, Signals and Control
Library

2.1 The First Steps

Before we get into the interesting, physical issues, let’s briefly revisit the topic of
signals and mathematics [1]. This rather dry topic is necessary because we will come
across the use of so-called signals again and again in the further course of modeling.
To show how Simcenter Amesim works, we will first create a simple calculator. After
starting Amesim, the first thing that opens is the program’s interface. The first step
to create a model is to drag single items from the model library to the drawing layer
of the program shown in Fig. 2.1.
In this book, all examples and exercises were created under the operating system
Microsoft Windows. But there is also the possibility to run Amesim under the operating
system Linux. This can be advantageous if you want to couple the program with other
simulation tools or if you want to automate the models. Here, however, we want to
operate Amesim exclusively via the graphical user interface (GUI). We start with
the first example in the following chapter, to perform a simple calculation like:

1+2=3 (2.1)

We want to build up the model for this mathematical calculation by trying to recreate
a calculator with the signal library (red library).

2.2 How to Build a Simple Calculator

We start with the model for a simple calculator. To do this, we use the mouse to drag
the item for a constant from the Signal library onto the drawing area. Make sure
that you are in Sketch mode. Otherwise, you will not be able to add anything to the
drawing area shown in Fig. 2.2.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 13

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_2
14 2 Mathematics, Signals, and Control Library

Fig. 2.1 Graphical user interface of Simcenter Amesim after startup—no model is added yet

We can activate this item for the constant with the mouse and then, as is common
in Windows, copy it with the key combination Ctrl + C, and paste it again with the
key combination Ctrl + P. With the key combination Ctrl + R, we can rotate the
items. Further shortcuts will be given by a menu if you press the right mouse button.
The equation 1 + 2 = 3 is to be solved with the calculator; therefore, one must drag
a sum item with the mouse onto the drawing area from the library shown in Fig. 2.3.
After the two constants and the sum item have been copied or dragged on the
workbench, they must be connected with the mouse as shown in Fig. 2.4.
The items must be connected to each other as shown in Fig. 2.5. To do this, you
have to move the mouse pointer to the so-called ports of the items until they are
displayed in green. Now, you can connect the ports of the items with each other.
When all ports of a model are properly connected, the color of the items changes.
The model is then no longer highlighted in dark color and can be used. If the con-
necting does not work, there is usually a thinking error in the design. This can be
checked if you compare the units of the two sides of the connections.
2.2 How to Build a Simple Calculator 15

Fig. 2.2 Example: Creating a simple calculator; Step 1: Creating a constant

Fig. 2.3 Example: Creating a simple calculator; Step 2: Copying constants and adding a sum
function
16 2 Mathematics, Signals, and Control Library

Fig. 2.4 Example: Creating a simple calculator; Step 3: Connection of the two constants and the
sum function

Fig. 2.5 Connecting an item

for constant value with the
item for summation
2.2 How to Build a Simple Calculator 17

Fig. 2.6 Example: Creating a simple calculator; Step 4: Closing the model with a sink item

Important: Always make sure that the elements are connected correctly and
highlighting has been changed. It is also possible that the elements cannot be
connected if a physical model is illogical, e.g. if the units do not fit. In this
case, you should question if your model is logical.

In general, you can connect the elements from the signal library with all other
libraries. A sink from the signal library, as shown in Fig. 2.6, must therefore always
be compatible with the other libraries and can then be used accordingly if you cannot
find a suitable connection or continuation of the model for a particular simulation
model in a physical library.
Once the model is complete for addition, no item will be darkened. All items
should then be displayed in normal color. Now, you can switch from Sketch mode to
Submodel mode as shown in Fig. 2.7.
Switching to the Submodel mode is possible only if the model has been drawn
correctly before. In order to enter individual values in the graphical model, it is then
necessary to switch further to the parameter mode as shown in Fig. 2.8.
In the mode parameter it is now possible to enter the values to solve our equation
1 + 2 = 3. You can see exactly where to enter the values in Figs. 2.9 and 2.10.
Additionally, in Figs. 2.11 and 2.12, you can see how to switch to simulation mode
and then read the result of the calculation.
18 2 Mathematics, Signals, and Control Library

Fig. 2.7 Example: Creating a simple calculator; Step 5: Change from Sketch mode to Submodel
mode

Fig. 2.8 Example: Creating a simple calculator; Step 6: Change from Submodel mode to Parameter
mode
2.2 How to Build a Simple Calculator 19

Fig. 2.9 Example: Creating a simple calculator; Step 7: Entering a value for the first constant

Fig. 2.10 Example: Creating a simple calculator; Step 8: Entering a value for the second constant
20 2 Mathematics, Signals, and Control Library

Fig. 2.11 Example: Creating a simple calculator; Step 9: Switching to Simulation mode and starting
the simulation

Fig. 2.12 Example: Creating a simple calculator; Step 10: The simulation is finished and the result
can be read off
2.2 How to Build a Simple Calculator 21

Problems

For better understanding, you can change the signal model now and try out any other
calculation [2].

2.1 Sum up three or four different constants to one result.

2.2 Use another operator like, e.g. multiplication or division.

2.3 Copy your model with Ctrl+C, paste it somewhere on the same workspace, and
use two different models in one workspace.

2.4 Plot a diagram of your results. Try to plot two different results from different
models in one diagram.

References

1. Günther, M., Velten, K.: Mathematische Modellbildung und Simulation, Wiley-VCH (2014)
2. Bossel, H.: Modellbildung und Simulation. Springer Vieweg, Wiesbaden (1992). https://link.
springer.com/book/10.1007/978-3-322-83658-8
Chapter 3
The Mechanical Twin

3.1 What Is Mechanics?

The word mechanics is derived from the ancient Greek word mechané. In physics,
mechanics is usually understood as the classical mechanics of moving bodies and
particles. It deals with the motion of bodies and the forces acting on them. It is needed
for the planning of machines or buildings. In engineering, this includes static and
dynamic mechanics. Let us solve a easy physical problem first. We want to examine
a mechanical system and have chosen the example of a bouncing ball. The necessary
mechanical library can be found on the right side of the interface in the library tree.
The colour of the mechanical library is green.

3.2 The Model of a Bouncing Ball

The idea of what exactly happens when a ball falls to the ground and bounces up
again is relatively easy to understand and will not be explained again here. We want
to start right away within the symbolism and workflow of Amesim. Therefore we
should create the bouncing ball model directly in the Sketch mode. To do this, we
first open a so-called simulation model for the mechanical process and create our
twin as shown in Fig. 3.1. The sketch already contains all the essential functionali-
ties. When creating the simulation model, however, please do not yet worry too much
about how long, large or heavy a body is and exactly what forces are acting. First,
draw the model as you imagine it. The details will then be clarified afterwards. After
that all other quantities, such as the weight of the ball or the acting forces, are entered
in section Submodels and Parameters.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 23

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_3
24 3 The Mechanical Twin

Fig. 3.1 Mechanical

simulation model for a
falling ball that is bouncing
on the ground

Important: Focus on a correct sketch of the model in the beginning and try
if it can be compiled and the simulation runs. For this, estimate the physical
parameters and boundary conditions first. After that step you can refer the
correct quantities and add them. This will help you to keep track.

3.2.1 Simulation Model

So for, it doesn’t really matter how heavy or how big the ball is or from which height
it hits the ground. Let’s start with the sketch of the functional model. We represent
our sphere with a so-called mass symbol [MAS001]. We drag it with the mouse onto
the empty workspace in Sketch mode.
Now it gets a bit more complicated. We will need a symbol for the height from
which the ball falls to the ground, or the better say the gap between the ball and the
ground. The impact of the elastic ball should be damped when it hits the ground. To
represent these effect with our twin, we choose an elastic contact symbol from the
mechanical library. In addition to the mass item, that represents the ball, we drag a
symbol for the elastic contact [LSTP00A] and the ground [V001] onto the drawing
area. You can see how the finished simulation model should look like in Fig. 3.1.
Please also make sure that all symbols are properly connected and no longer
have a dark background. Only then does the model work. When all the symbols are
connected, your first bouncing ball twin is ready. Now comes the final touch and we
need to look a little closer. Switch to the parameter mode and enter the parameters
for our twin.
3.2 The Model of a Bouncing Ball 25

Fig. 3.2 Press button to set run parameters—specify simulation times and time step size

3.2.2 Submodels and Parameters

Once the graphical creation of the simulation model is complete, the physical quan-
tities for the different Submodels can be selected on the Submodel tab. At first, not
much changes in the presentation. The current parameters can be specified under the
Parameters tab.
Important is, that the parameters are constant quantities, which do not change
during calculation, but have a significant influence on the result. The user must know
these parameters himself or observe them in reality and specify them to the model.
One determines them for example by measuring or weighing the bodies.
For the mass representing our ball, the weight (here: 1 [kg]) and the angle with
which the ball is dropped down must be entered. In our example the ball should fall
vertically downwards with an degree of 90◦ . For the gap between the ball and the
ground, the height of 1 [m] as well as the contact stiffness of the ball at impact (here:
it is best to enter a very large value) as well as the damping at impact (here: 150
[N/(m/s)]) must be entered. The ball will not be able to penetrate the ground with
this damping model, but will bounce off the ground. In the sections Submodels and
parameters we have always summarized all parameters in a table (see: Table 3.1).
This is the input to Amesim.
In Table 3.1 the name of the model or submodel is also given under the Item
column. Under this name, you can also search for the corresponding item using the
search function in the parameter window. If all parameters are correct, the simulation
can be started as shown in Fig. 3.2. After clicking with the mouse on the button to set
the Run parameters, you first have to enter how long the simulation should run. In
our case you start at 0 [s] and calculate up to the simulation time 4 [s]. Additionally,
a reasonable step size is selected (here: 0.01 [s]). This specifies when results are to
be written for each time step. The simulation times are given in Table 3.2.
26 3 The Mechanical Twin

Table 3.1 Parameters for the bouncing ball simulation model

Item Parameter
MAS001 Mass = 1 [kg]
Inclination = 90 [degree]
LSTP00A Gap or clearance with both displacements zero = 1000 [mm]
Contact stiffness = 1e+06 [N/m]
Contact damping = 150 [N/(m/s)]
Penetration for full damping = 0.001 [mm]
V001 Linear displacement = 0 [m]

Table 3.2 Simulation time Simulation settings

for the bouncing ball
simulation model Start time = 0 [s]
Final time = 4 [s]
Print interval = 0.01 [s]

Fig. 3.3 Successfully completed the simulation for the digital twin of a falling ball

When the progress bar has risen to 100% as shown in Fig. 3.3, the simulation
was completed successfully. Now the results can be examined. If a simulation does
not run successfully, or if it takes a very long time to generate a result you have to
change the simulation parameters. There is the possibility to choose the simulation
time (Final time) a bit lower and you could also sets the Print interval to a smaller
time step size.
This allows you to check which possible error could be present in the simulation
model. Mostly the error is caused by physical values that do not make sense, e.g. the
mass of the ball is much too high or the distance is too large. If you select the time
steps too large, it is possible that the simulation cannot be carried out successfully. It
is then said that the mathematical model does not converge. But also if you choose the
time steps too small it is possible that the simulation cannot be executed successfully.
3.2 The Model of a Bouncing Ball 27

Fig. 3.4 Generating diagrams: visualisation of the result by dragging the variable on the work
space

This is because the Amesim program writes results to the hard disk for each time
step. If too much data is written, problems with disk space may occur.
Next, let us take a look at the most important simulation results that result from
the simulation for our ball model. We want to analyse them. Drag the desired result
variables over the workspace to create a diagram of the value as shown in Fig. 3.4.
After that, we will give some suggestions for further developments or further inves-
tigations on our digital twin. One thing should be kept in mind when creating the
model. If something would not work in reality, e.g. if the mass of the ball or the
distance to the ground would be much too high, then this can also lead to an abort
and an error message. That’s why simulations should only be done with halfway
plausible parameters.

3.2.3 Results and Analysis

The program stores a lot of different physical quantities in the background during
the simulation run. This could lead to a lot of data if the model is big and time steps
are low. However, in this section we will only look at the really interesting quantities
and discuss them. We will show you how what this means.
To be able to draw and map simulation results in a diagram, one must click on the
result variable after the successful simulation, then drag the name of the variable as
shown in Fig. 3.4 from the variables window to the workspace with the mouse held
28 3 The Mechanical Twin

Fig. 3.5 Size of gap is a measure of the distance between the ball and the ground

down. This will generate a diagram of this result variable over the time course of the
simulation.

Important: Drag a result variable with the mouse from the menu on the right
side to the workbench to plot a diagram in an x,y-chart.
What results are of particular importance to us in the bouncing ball? At the
moment, there are essentially two quantities in this model. We are interested in
how the height of the ball changes and how it bounces back from the ground. We can
best visualize this by looking at the gap or clearance between the ball and the ground
(see: Fig. 3.5). How should we now read and interpret the diagram in Fig. 3.5? Sup-
pose we drop the ball from a height of 1.0 [m] to the ground. After about one second
it hits the ground for the first time at a height of 0.0 [m]. It then jumps up again but
does not reach the full height. After several jumps, the distance between the ground
and the ball becomes smaller and smaller. Your first simulation for the bouncing ball
was successful. After this first trial we look at more advanced problems and work
suggestions for this example.
Important: If you want to display two result variables in a single diagram,
you should simply drag the second variable onto the diagram window in the
same way as the first one. This way the second value will also be displayed
in the same diagram.
To get to know the digital twin of your bouncing ball problem a bit better, let
us work on a few more advanced work suggestions. In addition, let’s take a look at
the speed of the bouncing ball. For this we drag the variable Velocity at port 1 from
3.3 The Mechanical Rocker 29

the variables field onto the workbench to display this variable in a diagram as well.
When the ball is released the velocity is still small, towards the end it gets bigger
and bigger until the ball hits the ground and is bounced back.

Problems

3.1 Investigate after what time the ball first touches the ground if you set the distance
between the ball and the ground at the beginning from the height 1.0, 2.0 and 4 [m].

3.2 How does the trajectory of the ball change when you increase the weight of the
ball from 1.0 to 1.5 [kg] and to 2.0 [kg]?

3.3 How does the trajectory of the ball change if you decrease the contact damping
force from 150 to 50 [N/(m/s)]?

3.3 The Mechanical Rocker

Now let’s consider another use case that we can recreate with the mechanical library.
We want to create a twin for a mechanical rocker. On the right side, we want the bar
of the rocker to be pushed down. On the left side of the seesaw, we want a spring
that is attached between the left lever arm and the bottom. Let’s start right away and
build the model of the rocker as shown in Fig. 3.6.

3.3.1 Simulation Model

Our simulation model for the digital twin of the mechanical rocker initially consists
of a lever arm on the left side to which the spring is attached. The model lever2
[LML012] is used as the rocker. The length of the left arm is 1 [m]. On the right side,
a force pushes the rocker down by a distance of 1 [m]. The right lever arm of the
seesaw should also be 1 [m] long. For the time being, it does not matter how great
the friction is at the joint of the rocker. Moreover, this digital twin always assumes
that we are on Earth and that the gravity of our planet is acting.
30 3 The Mechanical Twin

Fig. 3.6 Simulation model

for a mechanical rocker with
lever arms of equal length

Once you have created the simulation model for the mechanical rocker with its
associated spring, you can enter the specific values, such as spring force, length of
the lever arms, or the distance by which the right lever arm is pushed down. We will
make these specifications in the next section Submodel and parameters.

3.3.2 Submodels and Parameters

We want to try to push down the rocker arm on the right side in the model. How can
we teach this operation to the twin? To do this, we selected a ramp function from the
signal library (red library) in Fig. 3.7 [RAMP0]. For this item, we specify the value
for the slope of the ramp function as 1. This means that within one second the right
arm is pushed down by one meter. On the left side of the seesaw we attach the spring.
In the next step, we have selected an element that shows us how far down to push
the right arm of the rocker. In the element for the seesaw we specify how long the
two arms of should be. For both sides of the seesaw we should choose lever arms that
are 1 [m] long. The values from Table 3.3 must always be entered in the parameter
mode.
The simulation time is set to 1 [s] according to the given Table 3.4. In this cal-
culation we have to take care that the simulation time is not too long. This is for a
very simple reason. If we choose the simulation time too long, then the arms of the
rocker on the right side will be pushed down too far, which would lead to nonsensical
behavior.
3.3 The Mechanical Rocker 31

Fig. 3.7 Ramp function that allows us to push the seesaw down on the right side

Table 3.3 Parameters for the mechanical rocker simulation model

Item Parameter
RAMP0 Slope = 1 []
XVLC01 Time constant for derivative of displacement = 0.0001 [s]
LML012 Distance port 1 to pivot = 1 [m]
Distance port 2 to pivot = 1 [m]
SPR000A Spring stiffness rate = 1000 [N/m]
Spring force with both displacements zero = 0 [N]

Table 3.4 Simulation time Simulation settings

for the simulation model of
the mechanical rocker Start time = 0 [s]
Final time = 1 [s]
Print interval = 0.01 [s]

3.3.3 Results and Analysis

We want to evaluate two different events. Of course, with the rocker, we already
know what needs to happen. It is always good to check the results for plausibility
and to critically question if everything is correct. First, in Fig. 3.8 we have a look how
the movement of the lever arms changes. The left arm of the beam moves exactly
32 3 The Mechanical Twin

Fig. 3.8 Height by which left lever is lifted up and the right lever arm is pushed down

one meter upwards in a time span of 1 [s], the right lever of the seesaw moves one
meter downwards. This is a reasonable behavior.
Important: Always check your models for plausibility and logic before you
make the models too complicated. Check the dimensions and the units of
your results.
The fact that the right bar is pushed down is shown by the negative value. If the
left bar had a greater length, it would move further up than the right bar would move
down. So we see that both lever arms are moved by 1 [m] in the time 1 [s].
We can also look at something else in Fig. 3.9. That is, by what angle the beam of
the rocker is rotated around its axis. This can be evaluated by the size angular lever
position. Here, the seesaw rotates by about 57◦ . The example is relatively simple
now, but we can do the following suggested work to make it a little more interesting
and explore our mechanical problem a little further. In the Problems part you find
suggestions for further work.

Problems

Why don’t they do the following investigations with your rocker model?

3.4 How do the results change when we press the seesaw for more than one second?
At what point do the results become nonsensical and why?
3.4 How a Row on a Pulley Works 33

Fig. 3.9 Angle in degrees by which the beam of the rocker rotates around its axis

3.5 Check how the height to which the weight is pulled changes when the mass
hanging at the bottom of the rope changes. Does the length of the rope also change
during the process?

3.6 What happens if we extend the right arm of the bar, e.g. to 3 [m]? Interpret your
new results.

3.7 What changes if we increase the spring force on the left arm of the seesaw?
Interpret these results as well.

3.4 How a Row on a Pulley Works

We will build a mechanical twin for a pulley. With the rope we will lift a weight.
The rope is to run over a pulley. We will also investigate whether the pulley runs
without friction or not. When modeling, you will notice that the length of the lower
end of the rope also has an influence on the result. This is because this part of the
rope can still stretch when it is pulled. Please always remember this when you want
to practice bungee jumping.

3.4.1 Simulation Model

Our simulation model in Fig. 3.10 for the mechanical twin of the rope on a pulley
consists initially on the left side of a force F with the unit Newton [N]. This can be
34 3 The Mechanical Twin

Fig. 3.10 Simulation model

for a rope to lift a weight
running over a pulley

used to pull on the rope. Then the rope runs over the pulley and at the hanging end
of the rope there is the weight. A pulley never runs completely without friction, so
you have to add a model for the friction (rotary load) to the pulley.
Again, it doesn’t matter how big the friction is, the force on the left side or the
weight hanging on the bottom of the rope is. You should first understand the basic
structure of the simulation model for the problem. Then you assemble it and test it
for function. Only after that they should enter the exact parameters.
In the following step, one can then enter the concrete values, e.g. the weight or
length of the rope. In addition, the gravity of the earth is still specified with the apple
symbol. We discuss later, if this makes sense. If you want, you can remove the apple
and check what happens. In the following section Submodel and parameters we enter
the parameters of the model.

3.4.2 Submodels and Parameters

Let’s take a closer look at the parameters again. We are pulling on the left side of the
rope with a force of 50 [N]. Since we are pulling on the rope to the left, we need to
set the force to −50 [N], otherwise we would push the rope, which makes no sense.
So, the value is given with a negative sign. In addition, we must now also specify
the size of the pulley we want to use and we must also choose the length of the rope.
The weight of the body we want to lift is also crucial, and we must specify that it
should weigh 1 [kg]. All other values are given in Table 3.5.

Important: Gravity has a great influence on many mechanical models, but

also on other physical phenomena. Thus, the applications may be different in
space or on other planets.
3.4 How a Row on a Pulley Works 35

Table 3.5 Parameters for the simple rope hoist simulation model
Item Parameter
CONS00 Constant value = −50 [N]
RSHE002A Diameter of the sheave = 500 [mm]
Roping angle = 90 [degree]
MECROPE0 Stiffness of unit length of rope = 1e+06 [N/m]
Viscous friction of unit length of rope = 1000 [N/m/s]
Initial length = 10 [m]
RL00A Moment of inertia = 1 [kgm**2]
Coefficient of viscous friction = 1 [Nm/(rev/min)]
(All other values are set to 0)
MAS001 Mass = 1 [kg]
GRAV0 Constant gravity value = 9,80665 [m/s/s]

Table 3.6 Simulation time Simulation settings

for the simulation model
simple wire rope hoist and Start time = 0 [s]
pulley Final time = 4 [s]
Print interval = 0.01 [s]

The symbol with the apple [GROV0] indicates the magnitude of gravity. On the
earth the gravitational acceleration of 9,80665 [m/s2 ] acts almost everywhere. If we
were on another planet like Mars or Venus, e.g. we would have to use a different
value for gravity. If we were on our neighboring planet Mars, we would not be able
to use this digital twin because the gravity there is 3.711 [m/s2 ]. On the moon, the
gravity is 1.62 [m/s2 ].
We can’t change the gravity acceleration arbitrarily in the individual items, and
the apple icon doesn’t affect all models the same. Unfortunately, however, you can
always paste the icon on the desktop, and that’s what we did, because the apple icon
looks very nice. But inside the submodels, unfortunately, the gravitational constant
of the earth is still used in some places. This should be adapted once in following
versions of the program Amesim. Then digital twins can also be created for other
planets.
Here we want to do the calculation for 4.0 [s]. Please make sure that the simulation
time in Table 3.6 is not too long, otherwise there is a risk that you will pull the whole
rope through the pulley. Our model is not designed for this and no useful results
would come out.
36 3 The Mechanical Twin

Fig. 3.11 Height by which the weight is lifted at the bottom of the rope

3.4.3 Results and Analysis

What result can we evaluate now? It definitely makes sense to first look at how many
meters the weight at the bottom of the rope is lifted when pulling with a force of 50
[N]. Again, please pay attention to the sign. It is negative because the rope is pulled
away from the pulley.
Take care, when specifying boundary conditions such as forces, velocities or mass
flows, it is always important in which direction the force or the flows act. According
to the direction you have to select the sign. If you do not know which sign to use.
Just test if it has to be a plus or a minus. If we look at the displacement port variable,
we see in Fig. 3.11 that the weight can be pulled up by about one meter in 4 [s]. This
fact is still relatively easy to understand and we could have derived it with simple
physical considerations.
Just take a closer look at Fig. 3.12. This is because our digital twin can predict
even more. Since the rope is pulled up very quickly over the pulley, there is a brief
lengthening of the rope at the beginning of the pulling process. It is first tensioned,
so to speak, and thus stretched. You can see this in the following figure from the brief
change in length. As soon as the mass is pulled up evenly, the change in length of
the rope decreases again.
This process is much more complex and cannot be determined so easily with
paper and pencil without a simulation model of the technical system. However, the
additional lengthening of the rope can be quite important in engineering or in daily
life. Just think of bungee jumping.
3.5 Damper of a Driving Car 37

Fig. 3.12 Change in length of the rope due to acceleration at the start

Problems

Now let’s look at what else we can use our digital twin of the rope running over a
pulley for. They can investigate the following work suggestions on their own using
their simulation model:
3.8 What would happen if they double the force with which they pull on the rope.
How fast will the mass be lifted then?
3.9 Check how the height to which the weight is pulled changes when the mass
hanging at the bottom of the rope changes. Does the length of the rope also change
during the process?
3.10 Try to create a simulation model where you run the cable over two different
pulleys. What does this do for you?
3.11 Just remove the apple symbol and see if the model still runs.

3.5 Damper of a Driving Car

A damper or shock absorber holds the wheel of a car and is a mechanical, hydraulic
or pneumatic device designed to absorb and damp shock impulses from the road. It
does this by converting the kinetic energy of the shock into another form of energy
which is then dissipated [1]. We want to show you in this example how engineers
work on a problem and try to solve it step by step.
38 3 The Mechanical Twin

Fig. 3.13 Model for one

wheel of a car with a
mechanical damper and road
profile

3.5.1 Simulation Model

Pneumatic and hydraulic damper or shock absorbers are used in conjunction with
cushions and springs. An modern automobile shock absorber contains spring-loaded
check valves and orifices to control the flow of oil through an internal piston [1]. In
this example we concentrate on the simpler mechanical damper, because pneumatic
and hydraulic library will be introduced later. The simulation model in Fig. 3.13 for
the damper consists only of one wheel for the the whole car, this is done to make
the model simpler. After sketching the model of the damper, one can define the
parameters.

3.5.2 Submodels and Parameters

We have mentioned that the model of the car is simplified, we modeled only one
wheel of the car and the total mass. We have modeled the car with only one mass
element and estimate the weight of the vehicle to be 1500 [kg]. The second mass
model is for the wheel. In Table 3.7 we give the full list of parameters for the model.
3.5 Damper of a Driving Car 39

Table 3.7 Parameters for the shock absorber of a driving car

Item Parameter
MAS002 (of the car)
Mass = 1500 [kg]
Inclination = 0 [degree]
SPR000A Spring rate = 100000 [N/mm]
Spring force with both displacements zero = 0 [N]
DAM0000 Damping rate = 1000 [N/(m/s)]
MAS002 (of the wheel)
Mass = 35 [kg]
Inclination = 0 [degree]
SD0000A Spring rate = 100000 [N/m]
Damper rating = 1000 [N/(m/s)]
XVLC01 Time constant for derivative of
Displacement = 0.0001 [s]
UD00 Constant gravity value = 9,80665 [m/s/s]
Number of stages = 2
Cyclic = no
Time at which duty cycle starts = 0 [s]
Output at start of stage 1 = 0 null
Output at end of stage 1 = 0 null
Duration of stage 1 = 2 [s]
Output at start of stage 2 = 0.08 null
Output at end of stage 2 = 0.08 null
Duration of stage 2 = 8 [s]

Table 3.8 Simulation time Simulation settings

for shock absorber of a
driving car Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]

We took a simulation time of 10.0 [s], which is of course rather short, so we only
want to model the point of time where the wheel hits the edge of the sidewalk. The
simulation time and time step is defined in Table 3.8.

3.5.3 Results and Analysis

Before we have a deeper look at the results of the car movement and damper behavior,
we want to check if the simulation is running and the model works. In Fig. 3.14
40 3 The Mechanical Twin

Fig. 3.14 Height of the road profile, e.g. step of a sidewalk

the signal of the submodel [UD00] is shown and gives us an impression, that this
function was defined in the right way. Of course, you can make the road profile more
sophisticated.
Now let us take a closer look at the more interesting results. The damping of the
wheel and bouncing of the damper after step is shown in Fig. 3.15. You can see, that
the pneumatic tire which is modeled by a spring [SD0000A] reacts with an oscillation
of the weight on the step of the sidewalk.
The movement and oscillation of the whole cabin of the car is shown in Fig. 3.16
and gives us an impression how the second damping acts on the car. It can be seen
that here the amplitude of the oscillation is even higher. This is probably not very
convenient and should be avoided in an normal car.
At the end of this section you can work on some problems that we have defined
for you. This is a normal and good approach and shows you how engineers work.
First we created a simple model that works and after this introduction to the problem
we will try to improve our car model more and more until we end up with a good car
behavior.
3.5 Damper of a Driving Car 41

Fig. 3.15 Damping of the wheel and bouncing of the damper after step

Fig. 3.16 Oscillation of the car cabin in the moving car after step of the sidewalk

Problems

We have already discussed some problems. Now you can investigate the following
problems for your damper with your simulation model.
3.12 Try to make the damping of the spring system softer to reduce the oscillation
in the cars cabin.
42 3 The Mechanical Twin

3.13 Change the size of the step and examine, how the damping changes.

3.14 We have modeled only one wheel as an example, try to build a digital twin of
a car with four wheels. Consider how to model a correct road profile for four wheels.

3.15 Work through the chapters on hydraulics and pneumatics and also try to build
a hydraulic or pneumatic shock absorber.

3.16 Try to generate a more complicated road profile and chose longer simulation
times.

3.6 How to Use 3D Geometries

To get a more vivid idea of the mechanical twin and to be able to have a discourse
during the innovation process, the results can also be visualized three dimensional
in Simcenter Amesim. An example, how to visualise such an model in 3D is shown
in Fig. 3.17. Here we use a model of a spring to visualise the working behavior. The
visualization can be done with Animation dialog, which can be opened by clicking
on the Animation button in the menu bar of the main graphical user interface (GUI).
Based on every digital twin, 3D geometries can be generated for better under-
standing of dynamic simulation results and test their behavior and interaction. We
refer to this also as computer aided engineering (CAE) or CAx, which includes also
computer aided design (CAD). The designer is given the opportunity to perform
real-time tests on the digital twin, thus applying theory directly to the design and
questions how well it works. Theoretical knowledge is deepened through such virtual
dry runs and various application scenarios can be run through independently. In the
following chapter examine the digital twin of a chassis with a hydraulic damping
system when driving over an obstacle. A complete vehicle in a landscape can also be
visualized, as in Fig. 3.18. The 3D geometry of the car can be created by the designer,
e.g. this could also be done with the tool Blender [2] or any other alternative CAD
tool. How the vehicle behaves, how the damper system works or which forces act on
the chassis depends on the spring model of the damper an is then again calculated
by the Amesim main program making it possible to test the virtual geometry of the
chassis under real conditions and to investigate in advance how it works. With this
model, e.g. driving cycles of the whole car can be simulated, as shown in Fig. 3.18.
The behavior of different bumpers can be examined. The results are much more
understandable, if the whole car maneuver can be visualised.
So, simple 3D geometries can be added within the Animations dialog. If an 1D
model is already running, e.g. the spring of the cars damper system, all the other
object of the scene can be added with the Animations dialog. Standard objects like
boxes, spheres, splines or springs can be added by pressing the Add menu in the
menu bar. In addition, 3D models created with Blender can also be imported. To
connect the 1D model from the main GUI, the result values of the simulation run can
3.6 How to Use 3D Geometries 43

Fig. 3.17 Animation screen for 3D visualization and animation of the model with the animation
dialog

Fig. 3.18 Results of a driving cycle with 3D geometry of a car and realistic driving behavior
44 3 The Mechanical Twin

Fig. 3.19 Connecting a 1D spring model with a 3D spring model in the animations dialog

be dragged to geometric values of the 3D model. This can be seen in Fig. 3.19 and
is done simply by dragging the result value on the animation window.
After this step the simulations result data can be used to move and animate the 3D
model by pressing the play button in the menu bar of the animation dialog. As shown
in Fig. 3.20 this will lead to an animation scene of the deformed damper spring.
Furthermore, in conjunction with artificial intelligence (AI) as shown in a later
chapter of this book, the program can learn to find its way over the obstacle as
efficiently as possible. In the remainder of this book we will refrain from using 3D
geometry as this is beyond the scope of this book, but we have further books on this
topic planned.
Today there is a really good new smartphone app called poly-cam [3]. It is a 3D
capture application for high-quality 3D models from photos with mobile device, and
rapidly generate scans of spaces with the LiDAR sensor. LiDAR is an acronym of
light detection and ranging or laser imaging, detection, and ranging [4]. This process
is sometimes referred to as 3D laser scanning, a special combination of 3D scanning
and laser scanning. Graphical models can be scanned from real objects very easily
and you can use them to create animated presentations of results.
After scanning the 3D models can be edited with the free of charge tool Blender
[2] and it is possible to export them in different file formats, e.g. GLTF or STL. GLTF
(GL Transmission Format) is an open-source and royalty-free 3D file that supports
References 45

Fig. 3.20 3D animation of the spring at different simulation time steps

static models, animation, and moving scenes. GLTF is used in games, native web
applications, and 3D ads. STL is a file format native to the stereo-lithography CAD
software created by 3D Systems. STL has several acronyms such as Standard Triangle
Language or Standard Tessellation Language. This file format is supported by many
other software packages. It is commonly used for rapid prototyping, 3D printing,
and computer-aided manufacturing. STL files describe only the surface geometry
of a three-dimensional object without any representation of color, texture, or other
common CAD model attributes [5].

References

1. Bauer, H. (ed).: Automotive Handbook, 4th edn, Robert Bosch GmbH, ISBN 0-8376-0333-1,
p 584 (1996)
2. https://www.blender.org
3. https://poly.cam/
4. Travis, S.: Introduction to Laser Science and Engineering. CRC Press, Taylor (2019)
5. https://en.wikipedia.org/wiki/STL_(file_format)
Chapter 4
The Thermal Twin

4.1 Examination of Heat Transfer

After the mechanical library in the previous chapter, we now want to turn to a com-
pletely different problem. For example, we don’t know how much a seesaw heats
up when exposed to sunlight, or what amount of heat is generated by the friction
between the rope and the pulley. Therefore, we have chosen the so-called thermal
library to create a thermal simulation model. The color of the thermal library is dark
brown.
There are three different heat transfer processes. First, we speak of heat conduc-
tion inside a body or between two neighboring bodies. This process is also called
conduction. The second mechanism is called convection. In this process, heat is trans-
ported by a gas or liquid. The third main heat transfer process is radiation. There are
further heat transfer processes, but we will not go too deep into this topic.

Important: There are different transport mechanisms for heat. In most cases,
heat transport occurs from a higher temperature to a lower temperature. Not
in the opposite direction.

Let us consider practical examples of heat conduction in an aluminium heat sink

and a solar collector. The temperature of a body results due to heat transfer with
its environment or due to transformation processes and chemical reactions inside
it. In heat transport, one essentially considers heat conduction, convection as well
as heat transport by radiation. It is important to what mass a body has. This allows
us to calculate how much heat can be stored and what temperature it assumes. All
individual processes can be described by different digital twins. At the end, further
working suggestions are given.
Also crucial to understanding the philosophy of Simcenter Amesim is that the
thermal library can be connected to the mechanical library or any other library in one

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 47

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_4
48 4 The Thermal Twin

model. For example, one can calculate how much heat is generated by friction from
a brake disc during the braking process of a vehicle to a couple of different complex
twins. For example, you can link certain items of the mechanical library with the
thermal library. However, it is only possible to connect items if this makes physical
sense. If two items cannot be connected, then this almost always has a reason.
Now let’s look at our thermal model for the heat sink of a generator. The design and
creation of heat sinks are very important for many technical applications. For exam-
ple, almost all electronic components such as electric motors, generators, batteries
or accumulators need to be cooled during operation.

4.2 Heat Transfer in Electric Generators

When dealing with the temperature of bodies, we should first realize that the tem-
perature of a body (in [◦ C]) is usually the effect of a previous heat or energy transfer
in joules [J] per unit time. In the case of an electrical generator, eddy currents cause
a heat input. An attempt is usually made to use heat sinks to keep the temperature of
the material low and to dissipate heat to the environment. The material from which
the generator and heat sink are made plays a role here. A heat transfer process is
given the unit watt [W].

4.2.1 Simulation Model

For the thermal simulation model, we need to specify for the metal bodies what
material they are made of. Here we want to model the heat conduction between the
two metals iron and aluminium. The necessary material parameters are already stored
in Simcenter Amesim. In addition, we want to consider the so-called convection, i.e.
the cooling of the metals by surrounding air, as well as the heat radiation.

Important: If we cool a heat sink with a fluid like air or water it is always
important how fast the fluid moves past the body and if turbulence occurs.
This is also characterized by the so-called Nusselt Number [1].

In our twin, we have to use a special element [THCD00] for the heat conduction
between the metal bodies. We composed our digital twin from a total of four different
metal bodies [THC00]. To tell the model which material each item in Fig. 4.1 is made
of, we have to use a so-called solid type index.
A heat source of 50.0 [W] in [THHS0] is attached to the right side of the iron block
and to the aluminium block, respectively. At the top is the aluminium block where
heat is dissipated by convection and radiation. Here, too, it is possible to change the
air speed in order to influence the heat transport.
4.2 Heat Transfer in Electric Generators 49

Fig. 4.1 Simulation model for heat conduction, the heat is transferred between a body made of
iron and a body made of aluminium

4.2.2 Submodels and Parameters

For the model parameters in Table 4.1, note that in each case the two aluminium
bodies have an equal mass, and the same weight was also used for the two iron
bodies. Here, the solid type index = 1 is set for aluminium and the solid type index =
2 for iron. Other materials would also be possible. Be careful to use the right solid
type index for the components.
It has to be noted, that for the estimation of the contact surface for [THCD00] the
volumes of the [THC000] blocks have to be taken into account. The simulation time
in Table 4.2 is chosen to be very long for this example, because such heat conduction
processes can take quite a long time power source is low and weight is relatively
high. Therefore, we adjust the time step size accordingly to higher time step sizes.

4.2.3 Results and Analysis

If we want to look at the results and estimate how heat from the source is conducted
first through the iron and then through the aluminium elements, we look at the time
period of 1 h (= 3600 [s]) in Fig. 4.2. We have applied 50.0 [W] of heating power on
the right side which is due to eddy currents, e.g. from an electric generator. The mass
elements are numbered consecutively and after 1 hour the temperature has risen up
50 4 The Thermal Twin

Table 4.1 Parameters for the simulation model of an aluminium heat sink
Item Parameter
THHS0 Heat flow rate at port 1 = 50 [W]
THSD00 Solid type index = 1 [-]
Material definition = pure aluminium (Al) [-]
THSD00 Solid type index = 2 [-]
Material definition = pure iron (Fe) [-]
THC00 Solid type index = 1 [-]
Mass of material = 5 [kg]
THCD00 Contact surface = 1000 [mm**2]
Thermal contact conductance = 1000 [W/mm**2/degC]
THGCV0 Inclination angle = 90 [degree]
Width = 100 [mm]
Length = 150 [mm]
Velocity of the fluid = 4 [m/s]

Table 4.2 Simulation time Simulation settings

for the model of an
aluminium heatsink Start time = 0 [s]
Final time = 3600 [s]
Print interval = 0.1 [s]

to approx. 80.0 [◦ C] for the elements that are directly applied to the 50.0 [W] heat
source.
The temperatures inside the metal blocks in Fig. 4.3 result from the correspond-
ing heat fluxes between the heads and the surroundings. Towards the left side, the
temperatures of both the iron and aluminium blocks only increased to about 50.0
[◦ C] each. The explanation for this is that on the left side, cooling of the two bodies
takes place by convection. During convection, heat is transferred from the metal to
the ambient air. The level of air velocity plays a decisive role in this process.
Heat conduction between bodies is given in units of watts [W]. We see that about
30.0 [W] of heat flows between both aluminium and the iron. The convection that
takes place on the left side is about 15 [W] after 1 hour. Note that this is only the case
when cooling air is flowing past the heat sink at a velocity of 4 [m/s]. Fig. 4.4 is also
interesting. It shows how the heat is conducted from the iron core to the aluminium
heat sink. We assume here that the heat conduction is quite good. Let us have a look
at the heat flows in a diagram.
Near the 50.0 [W] source on the right side, more heat is conducted than on the
left side, where cooling by the air takes place. This can be explained by the fact that
we are already at a lower temperature level there and there are no longer such large
amounts of heat stored inside the material. Now let us have a look at a few more
work proposals or problems that we can investigate with this thermal twin.
4.2 Heat Transfer in Electric Generators 51

Fig. 4.2 Average temperatures inside the iron and the aluminium blocks

Fig. 4.3 Heat conduction within the bodies and convection at the boundaries
52 4 The Thermal Twin

Fig. 4.4 Heat conduction from the iron block to the aluminium heat sink

Problems

Using your digital twin for the heat sink, investigate the following work suggestions
once:
4.1 Change the materials and use copper instead of aluminium, for example. How
does this change the temperatures of the bodies?
4.2 Investigate what happens if you increase the masses of the iron bodies from 5
to 10 and 15 [kg].
4.3 Set the heat source on the right side from 50 to 100 and 150 [W]. Observe what
happens to the temperatures of the metal bodies.
4.4 Try to cool the iron body further by increasing the air velocity on the right side
to 20 [m/s].
4.5 Have a look at the result output for convection. There you will find different
comparison numbers, like the Nusselt number or the Reynolds number. What do
these numbers stand for?

4.3 Site Selection for a Solar Collector

We will now look at heat transfer by radiation. In a solar or solar collector, radiation
from the sun strikes a collector body and heats it. Through subsequent conduction
4.3 Site Selection for a Solar Collector 53

and convection processes, this energy can be further distributed. However, it is also
crucial how much solar energy falls on the collector on a given day and where it is
located.

4.3.1 Simulation Model

To design the solar collector in Fig. 4.5, it is important to know where it is located. The
location plays a crucial role in its function. A collector located near Johannesburg in
South Africa has much more solar energy available than a collector located in Paris
or Helsinki. Furthermore, the time of day at which the solar radiation falls on the
collector is of decisive importance. In most places, the radiation is strongest around
noon. In Simcenter Amesim, it is possible to set these environmental conditions
using a custom model. This makes it possible to design the collector exactly for its
particular position. We could also create the collector from single items. But this is
not absolutely necessary. There is already a ready-made model that we can use.
The solar collector should still be connected with elements made of aluminium,
which should symbolize the design and structure of the collector support and its heat
exchangers. Through the contact, it can give its energy to these elements and thus
will heat them up.

Fig. 4.5 Simulation model for a solar collector and the heat conduction in the adjacent material
54 4 The Thermal Twin

4.3.2 Submodels and Parameters

In the simulation with the parameters from Table 4.3, all components are to be made
of aluminium. We also want to choose a very long period of 30.000 [s] simulation
time for this simulation. During this time, the collector heats up slowly. We select
Paris as the location for our twin number two.
Decisive for the absorption of the amount of heat is the area of the collector and
at what angle the collector is inclined to the horizon. We can specify many different
cities all over the world to indicate the position. However, it is also possible to
enter the GPS coordinates of a location. GPS stands for global positioning system.
A system by which signals are sent from satellites used to show the position on
the globe. This makes it possible to calculate yields that a solar collector field can
produce. The simulation times are given in Table 4.4.

Table 4.3 Parameters for the simulation model of a solar collector in Paris
Item Parameter
THSD00 Material definition = pure aluminium (Al)
THAMBCOND0 City name = Paris
Year = 2007
Month = December
Day = 16
Hour = 8
Minute = 00
THGCV0 Inclination angle = 90 [degree]
Width = 100 [mm]
Length = 150 [mm]
Velocity of the fluid = 5 [m/s]
THC000 Solid type index = 1 [-]
Mass of material = 50 [kg] and 20 [kg]
THR03 Equivalent emission factor wall/gas = 1 []
Exchange area = 100 [m**2]
Temperature of the gas = 20 [degC]
THRSOL002 Solar radiation setting mode = using ambient conditions
Exchange area = 1 [m**2]
Adsorption factor = 0.7 []
Equivalent emission factor gas/surface = 0.9 []
Surface inlination = 45 []
4.3 Site Selection for a Solar Collector 55

Table 4.4 Simulation time Simulation settings

for the simulation model of a
solar collector Start time = 0 [s]
Final time = 30000 [s]
Print interval = 1 [s]

Fig. 4.6 Angle with which the sun is above the solar collector

4.3.3 Results and Analysis

We have already described that when simulating the solar collector, it is of crucial
importance where exactly it is located and on which day the investigation is carried
out. We can use the twin to determine the exact solar path as shown in Fig. 4.6 for
any given day over the collector.
In our example, the morning of Dec. 16, 2007, was considered. These are, together
with 06/24/2011, the best days of the century or millennium. On the morning of this
day, the solar radiation heated up the collector and, as shown in Fig. 4.7, we can
increase the temperature in the components with the amount of energy absorbed
as a result. We can also see in Fig. 4.7 that air cooling of the left component will
not have a high impact on the temperature profile at the low air velocities. The
radiant power absorbed by the collector can now be accurately determined in Fig. 4.8.
Unfortunately, this lacks information on the extent to which the energy has been
suppressed by moisture in the atmosphere like fog, clouds, or rain. Dense clouds and
rain in the atmosphere have of course a very large influence on the solar radiation.
56 4 The Thermal Twin

Fig. 4.7 Temperature profile of the aluminium components of the solar collector

Fig. 4.8 Radiant power impinging on the surface of the solar collector
Reference 57

Problems

With the solar collector model is completed, you can now perform various tests and
modifications:

4.6 Once, change the weight of the left aluminium bracket to half its value. What
happens? How does the temperature change when they increase the wind speed from
5.0 to 8.0 [m/s]?

4.7 How does the energy input to the collector change, if we move the collector to
Johannesburg or Helsinki? Select some alternative cities for simulation.

4.8 Consider a model setup where they can use the collector to heat a medium such
as air or water.

Reference

1. Incropera, F.P., DeWitt, D.P.: Fundamentals of Heat and Mass Transfer, 4th edn. Wiley (1996)
Chapter 5
The Hydraulic Twin

5.1 What Is Hydraulics?

The word hydraulics comes from the ancient Greek word hydro for water and aulos
for pipe. We refer to a technology in which fluids are transported or used to transmit
power or energy. However, it is not just about water or oil. Many other fluids in such
a system can also be considered. When transporting liquids, the so-called pressure
loss plays a special role [1]. Any diversion, bend, or fitting will cause a pressure loss
within the pipeline. In the case of liquids, this usually has to be overcome by a pump.
Otherwise no transport takes place and the liquid cannot flow. Systems for the use
of heat or kinetic energy can also be considered as hydraulic systems, e.g. heating
systems or hydroelectric power plants. We want to use the hydraulic library to create
them. The hydraulic library (see: Fig. 5.1) of Amesim is located on the right side of
the workbench in the Library Tree.

5.2 Two Fuel Oil Tanks and a Pump

In the application in Fig. 5.1 we have two fuel oil tanks. The models for these tanks
[TK10] can be found in the hydraulic library. We connect them by a line with a pump
[PU001] in between. With this pump we can increase the pressure and transport
the oil into the second tank. You can also connect the hydraulic library with other
libraries of Simcenter Amesim. For example, you can connect the pump shaft to the
mechanical library to account for friction. Or you can couple the pump casing with
the heat library to calculate the heat dissipation at the casing [2].

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 59

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_5
60 5 The Hydraulic Twin

Fig. 5.1 Simulation model

for two fuel oil tanks with
one pump

5.2.1 Simulation Model

There is one major difficulty in creating networks of pipes. If one tank is on a higher
position compared to the other, you have to define the slope within the line connecting
the tanks and you will get a pressure difference between these two tanks. However,
since this is too complex for the first example, we only want to connect the oil tanks
with normal pipes without any slope. The resulting pressures in the pipelines will
then only result from the filling level inside the tanks. We will show how to create
inclined pipelines with a slope later.
When all the items are connected, the digital twin for our hydraulic system is
ready. Now we need to look a little closer and define the submodels and set the
parameters according to Table 5.1 for our model. In this model, we have used the
simplest model for a motor [PM001] without any friction. Of course, we can use
much more complicated models from mechanical library here.

5.2.2 Submodels and Parameters

Both tanks [TK10] should be the same size and contain the same amount of oil at the
beginning of the simulation. We define the contained amount by specifying the fill
level of the tank. It matters in the model whether we place the connections at the top
or at the bottom of the tank. We also need to define what material properties [FP04]
of the fuel oil because the fluid properties have a major impact on the flow behavior.
Please make sure that each tank contains the same fluid by choosing the same index
of hydraulic fluid for every component.
5.2 Two Fuel Oil Tanks and a Pump 61

Table 5.1 Parameters for the simulation model for two connected heating oil tanks
Item Parameter
TK10 Index of hydraulic fluid = 0 []
Initial height of liquid tank = 0.25 [m]
Tank area = 0.5 [m**2]
Minimum height alarm level = 0.1 [m]
Maximum height alarm level = 1.0 [m]
PU001 Index of hydraulic fluid = 0 []
Pump displacement = 100 [cc/rev]
Typical speed of pump = 1000 [rev/min]
PM000 Shaft speed = 1500 [rev/min]
FP04 Index of hydraulic fluid = 0 []
Temperature = 40 [degC]
Density = 850 [kg/m**3]
Bulk modulus = 17000 [bar]
Absolute viscosity = 51 [cP]

Table 5.2 Simulation time Simulation settings

for the simulation model for
two connected heating oil Start time = 0 [s]
tanks Final time = 10 [s]
Print interval = 0.01 [s]

You start the simulation again at 0 [s] according to Table 5.2 and let it run until a
calculation time of 10 [s]. When the progress bar has risen to 100 %, the simulation
has been completed successfully.
If a simulation does not run successfully or it takes a long time to generate a
result, there is a possibility to make the simulation time smaller. This allows to better
check which possible errors are present in the simulation model. Mostly it is due to
non-physical parameters.

Important: Always pay attention to what liquid you expect, e.g. whether it is
oil, water, or another liquid. Change the liquid properties, density and viscos-
ity, with the droplet symbol [FP04]. The index for the fluid to be used must
also be set there. This index can be redefined for each hydraulic component.
Several different fluids are possible within one larger system.

If you look at the fluid properties of the liquid in this experiment, you will notice
that it is not water but an oil, e.g. hydraulic oil, heating oil or diesel fuel, due to its
low density. Here the index of the hydraulic fluid = 0 was selected.
62 5 The Hydraulic Twin

Fig. 5.2 Filling level of the different oil tanks as a result of the analysis

5.2.3 Results and Analysis

During the simulation run all important quantities are saved and you can evaluate
the interesting and decisive quantities when it is finished [3]. To check if the results
make sense you can first have a look at a quantity that is easy to understand. The
level inside the two tanks shown in Fig. 5.2 is the first thing we should look at and
analyze. The level of the tanks is the simplest quantity and we can estimate whether
changing this value is realistic or not.
We want to show the filling level from the upper tank and the filling level from
the lower tank in the same diagram and can now compare the results very well. Since
both tanks are exactly the same size, the level in the upper tank has increased by
exactly the amount of oil that is missing in the lower tank.
The next thing is, we want to evaluate the effect of the pressure increase behind
the pump. As the level in the upper tank in Fig. 5.2 continues to increase, the pressure
in the line between the tanks also increases. We have to apply an increasingly higher
pressure with the pump to move the fuel oil to the second tank. However, we have
specified in our model in Fig. 5.3 that the speed of the pump should be kept constant.
Therefore, an ever-increasing torque must occur at the pump shaft during the pumping
process. This expected result is also shown in the graph in Fig. 5.3.
5.2 Two Fuel Oil Tanks and a Pump 63

Fig. 5.3 Torque of the pump shaft and pressure behind the hydraulic oil pump

Problems

Let’s look at some problems we can explore with the digital twin of the two connected
tanks.

5.1 What happens to the simulation when the bottom tank is empty? Investigate
how long you can run the simulation. Discuss the result.

5.2 How do the results change when you increase the footprint of the lower tank
from 0.5 [m] to twice the value of 1.0 [m]?

5.3 Operate the experimental setup with water instead of fuel oil. What do you have
to change? Compare the results.

5.4 Investigate what happens when the speed of the pump is doubled.

5.5 Try to build a network of tanks with a slope of the pipes between them.
64 5 The Hydraulic Twin

5.3 How Does a Hydraulic Jack Work?

Just like in the previous example with the two tanks, we will have to deal with liquids.
Let’s take a closer look at the example of the hydraulic jack [HJ001]. It basically
consists of a hydraulic piston. A hydraulic jack can be used to lift up cars or other
vehicles to change the wheels.
Most liquids such as water or oil, unlike gases such as air, have the property
that they are incompressible. This means that they cannot be compressed. Density
changes are therefore usually not very large. This property can be exploited very
well in lifting platforms, jacks, or an excavator bucket of a digger. Volume and mass
flows are almost always constant and can be easily balanced.

5.3.1 Simulation Model

Our model for the jack in Fig. 5.4 consists of an item for the force [FORC]. This
force is coupled with the step signal from the red library [STEP0]. It then acts on the
model of a so-called hydraulic jack [HJ000]. The model is already quite complete
and complex. The equations behind the model of the hydraulic jack can be found in
the help function of Amesim. You can open this help dialog by right-clicking on the
item with the mouse. All equations are very well documented and sketches for the
technical models are offered for better understanding.

Fig. 5.4 Simulation model of a hydraulic jack connected with a pressure compensation tank
5.3 How Does a Hydraulic Jack Work? 65

Important: For liquid lines, the slope is very important because the liquid
can only flow downwards if there is no pressure difference. If the liquid has
to flow upwards, we need additional pressure, e.g. from a pump. Therefore,
when creating the line, you must pay attention where the starting point (1)
and the end point (2) are located. Specify the angle of the line between these
starting and end points. Negative angles are also possible or the pipes can
also be connected in reverse.

A pressure accumulator is connected to the bottom of the piston via a throttle

[HYDORF0]. In this case, the pressure accumulator [HA001] is relatively small at
0.5 [L]. The pressure inside the accumulator is 50 [bar]. An overflow for leakage is
also connected to the hydraulic piston. This is not just a simple connection, but a real
line. This model [HL0002] is created automatically by Amesim you will not find it
in the library tree. With this type of line, it is important whether the fluid can flow
from top to bottom. You must therefore specify the slope of the line.

5.3.2 Submodels and Parameters

Carefully enter the parameters according to Table 5.3. The values must be re-entered
in Parameter mode in the corresponding field. Also make sure that the index of the
hydraulic fluid is set correctly.
In order to improve the jack’s function, an additional restriction should be placed
between the pressure vessel and the hydraulic piston to limit the flow and reduce
pressure pulsations. Set the simulation time back to 10 [s] according to Table 5.4.

5.3.3 Results and Analysis

The simulation runs for exactly 10 [s] according to the specification given in Table
5.4. After the first second, the jack is pushed down from the top with a force of 100
[N]. As can be seen in Fig. 5.5, this happens suddenly. The signal occurs within a
very short time window. Due to its viscosity, the hydraulic oil cannot react as quickly
as possible. This is noticeable in an oscillation of the oil pressure.
We see in Fig. 5.6 that the piston does move downwards in accordance with the
force signal. However, it responds with an oscillation at the end of the force applica-
tion due to the damping of the oil volume inside the piston. The oscillation continues
through the pipeline. The components “phone” each other.
In the accumulator tank, we can see an increase in pressure from 50 to over 52
[bar]. When the force decreases, the piston can then release the pressure from the
accumulator tank. However, in Fig. 5.7, we also see a clear pressure oscillation in the
accumulator despite the throttle. We can now perform some experiments with the
66 5 The Hydraulic Twin

Table 5.3 Parameters for the jack simulation model

Item Parameter
STEP0 Value after step = 100 [zero]
Step time = 1 [s]
FP04 Index of hydraulic fluid = 0
Temperature = 40 [degC]
Density = 850 [kg/m**3]
Bulk modulus = 17000 [bar]
Absolute viscosity = 51 [cP]
HJ000 Pressure at port 1 = 40 [bar]
Pressure at port 2 = 0 [bar]
Rod velocity = 0 [m/s]
Rod displacement = 0.3 [m]
Index of hydraulic fluid = 0
Piston diameter = 25 [mm]
Rod diameter = 12 [mm]
Length of stroke = 0.3 [m]
Dead volume at port 1 end = 50 [cm**3]
Dead volume at port 2 end = 50 [cm**3]
Total mass being moved = 250 [kg]
Angle rod makes with horizontal = 90 [degree]
Coulomb friction force = 0 [N]
Stiction force = 0 [N]
Viscous friction coefficient = 0 [N/(m/s)]
Leakage coefficient = 0 [L/min/bar]
HYDROF0 Index of hydraulic fluid = 0
Number of parallel orifices = 1 []
Orifice geometry = circular []
Diameter = 3 [mm]
HL0002 Index of hydraulic fluid = 0
Section type = circular
Diameter = 25 [mm]
Length = 1 [m]
Relative roughness = 1e-05 []
HA001 Pressure at port 1 = 40 [bar]
Index of hydraulic fluid = 0 []
Adiabatic initialization = 1 []
Gas precharge pressure = 40 [bar]
Accumulator volume = 0.5 [L]
Polytropic index = 1.4 [zero]
TK000 Tank pressure = 0 [bar]
5.3 How Does a Hydraulic Jack Work? 67

Table 5.4 Simulation time Simulation settings

for the simulation model of a
hydraulic jack Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]

Fig. 5.5 Force on the jack—after one second the force is increased from 0 to 100 [N]

Fig. 5.6 Downwards movement of the hydraulic piston in the jack

68 5 The Hydraulic Twin

Fig. 5.7 Pressure oscillation—pressure in pressure accumulator increases from 50 to 52 [bar]

digital twin of the jack in Fig. 5.4. We can investigate the following work suggestions
for this hydraulic model on our own.

Problems

What can we investigate with our model if we want to lift up a car? Have a look at
the following problems.

5.6 How do the results change when the moving mass increases or decreases? Use
masses with 150, 250, and 500 [kg].

5.7 What happens if we increase the accumulator? Change it once to 50 [L]. How
does this affect the pressure oscillation?

5.8 How does the pressure oscillation change if you increase or decrease the throttle
cross section?

5.9 Change the material values of the hydraulic oil. What would happen if you ran
the jack with water? Does that make sense technically?
References 69

References

1. Idel’cik, I.E.: Handbook of Hydraulic Resistance, 3rd edn. Begell House Inc., (1996)
2. Miller, D.S.: Internal Flow Systems, 2nd edn. Amazon Technology (1989)
3. Bode, B.: Verfahren zur Extrapolation wichtiger Stoffeigenschaften von Flüssigkeiten unter
hohem Druck. Tribol. Schmierungstech. 37(4), 197–202 (1990)
Chapter 6
The Pneumatic Twin

6.1 Pneumatic, Fluid Flow, and Turbulence

Gases such as air have the property of being compressible under common ambient
conditions. This means that they can be compressed by a cylinder. Simplified, we
can say that gases obey the ideal gas law. Density changes can be very large in gases.
Also, the mathematical models and solution methods are usually more elaborate for
compressible media than for incompressible fluids. Mass flows are also constant here,
but volume flows need not always be constant. Therefore, the same technical appli-
cations are often not possible for gases as for liquid fluids. And solution instabilities
will be different.

Important: The numerical stability of a library is also influenced by its

physics. All fluid media must be divided into compressible or incompressible
media and it must be clear before selecting a library whether the problem is
hydraulic or pneumatic. Use the index of the fluid to define the properties.
The description of turbulence is weak in Amesim. Use a computational fluid
dynamics (CFD) software tool for detailed description of turbulence.

The word pneumatic comes from the ancient Greek pneuma and means breath or
wind. In creating the digital twin, we examine how gases can be transported inside
pipe systems. When using the pneumatic library, you should keep in mind that gases
are usually compressible. This can lead to higher computational effort and a worse
convergence of the mathematical equation system than in incompressible problems.
It is also difficult to calculate complicated phenomena such as vortexes or turbulence.
To model turbulence, one should use another simulation tool such as ANSYS CFX or
ANSYS Fluent [1].

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 71

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_6
72 6 The Pneumatic Twin

6.2 Safety Valve for a Biogas Tank

Biogas consists largely of a mixture of methane and propane gas with small amounts
of air. We want to consider this biogas here simplified only as methane gas. Via a
pipeline of 10 [m] the gas is to be led into a tank with a volume of 10 [L].
Normally this is possible without any problems. To prevent over pressure inside
the tank and in the pipeline an additional safety valve has to be installed. This safety
valve is to prevent the wall of the tank or the pipeline from bursting if the pressure
inside rises too high. Often spring is used to seal the valve from the environment.
If the gas pressure in the pipeline rises above a certain level, the valve opens and
releases a certain amount of the gas into the environment.

6.2.1 Simulation Model

Let us build a simulation model like the one shown in Fig. 6.1 in which methane
[PNGD00] is filled into the biogas tank through a pipe. The pressure in the tank
is supposed to rise above 3 [bar]. A safety valve [PNCV01] is installed in front of
the tank [PNCH022] to protect it. The biogas tank should have a volume of 10 [L].
As soon as the pressure rises above a value of 1.8 [bar], the valve opens. All other
parameters are given in Table 6.1.

Fig. 6.1 Pneumatic simulation model for the safety valve of a biogas tank
6.2 Safety Valve for a Biogas Tank 73

Table 6.1 Parameters for the simulation model biogas tank and safety valve
Item Parameter
PNGD00 Temperature at port 1 = 293.15 [K]
Fluid definition = air
PNCS001 Index of hydraulic fluid = 0 []
Pressure at port 1 = 1 [barA]
PNGD00 Gas type index = 2 []
Fluid definition = methane (CH4)
PNCS001 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 3 [barA]
PNCV001 Gas type index = 1 []
Orifice area = 5 [mm**2]
Check valve cracking pressure = 1.8 [bar]
Hysteresis for opening/closing = 0 [bar]
PNCH022 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
Gas type index = 2 []
Volume = 10 [L]
Thermal exchange coefficient = 500 [J/m**2/K/s]
Thermal exchange area = 0.1 [m**2]
Exchange temperature = 293.15 [K]
PNL000R Gas type index = 2 []
Diameter of pipe = 10 [mm]
Pipe length = 10 [m]
Relative roughness = 1e-05 [zero]

Table 6.2 Simulation time Simulation settings

for the simulation model
biogas tank and safety valve Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.1 [s]

6.2.2 Submodels and Parameters

The simulation times for the biogas tank with the safety valve are given in Table 6.2.
We set the simulation time to a high value to see the full reaction of the pressure
inside.
74 6 The Pneumatic Twin

Fig. 6.2 Pressure in the biogas tank and mass flow of the gas through the safety valve

6.2.3 Results and Analysis

In the simulation results in Fig. 6.2, two curves are now shown in the diagram. On
the one hand, the pressure in front of the tank and, on the other hand, the mass flow
of the methane gas can escape through the safety valve.
After a successful simulation, click on the result variable to draw and map the
simulation results in a diagram in Fig. 6.2. It can be seen that the pressure increases
up to the limit of the gas valve. Then the valve opens and the methane is released into
the environment for safety reasons. Thereupon, the curve for the pressure flattens
out.

Problems

Have a look at a few work proposals that we can explore with our models: To do
this, you should find the appropriate model parameters in the simulation model and
modify them.

6.1 Lower the pressure of the methane gas at the inlet. Observe what happens.

6.2 Increase and decrease the volume of the tank and observe how this affects the
pressure curve in the tank.
6.3 Ventilation System of a Building 75

6.3 Change the opening behavior of the safety valve by decreasing the pressure at
which the valve opens. What happens to the pressure in the tank?

6.4 How do the results change if you use propane instead of methane as biogas?
What changes if you simulate with air?

6.3 Ventilation System of a Building

The design of the ventilation system for a building is not as trivial as it first seems.
However, compared to a system of water pipes, we encounter them relatively often.
In large department stores or furniture stores, you can see them on the ceiling. Even
though air flows have only small pressure differences compared to liquid flows,
differences of even a few pascals [Pa] can lead to high velocity gradients.
Anyone who has observed ventilation pipes with an expert’s eye may have noticed
that the pipes still have a larger diameter at the beginning. At the end, where the air
exits the vents, the pipes become thinner and thinner. This is because at the begin-
ning you still have very high volume flows, but these decrease with each additional
branching of the pipe.
In addition, another difference compared to liquids should be pointed out. Gas
flows are usually compressible flows. Gases can be compressed during movement or
at rest. They thereby reduce their volume but increase their density. This relationship
must be taken into account during the calculation.

6.3.1 Simulation Model

We want to build a simulation model for a building where we introduce air from a
main duct into a ventilation system and then split it to other ducts. The pipe branches
and different lengths result in different pressure losses.
In turn, the volume flow in the other parts of the pipe network depends on the
pressure drop in a pipe. The whole model is shown in Fig. 6.3. The parameters and
submodel are listed in Table 6.3 (see Table 6.4). The simulation time is given in Table
6.5. To make this example a bit more interesting and challenging, the reader should
try to find the ramp for the step signal [UD00] by himself. In Fig. 6.4, the mass flow
of the compressor can be seen and the duration of the steps of the signal [UD00] can
be calculated from this diagram.
76 6 The Pneumatic Twin

Fig. 6.3 Pneumatic simulation model of a ventilation system for buildings

Table 6.3 Parameters for the simulation model of a ventilation system (part 1)
Item Parameter
PNGD00 Gas type index = 1 []
Fluid definition = air
PNCS001 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
PNCP00 Gas type index = 1 []
Compressor displacement = 1000 [cc/rev]
Polytropic constant = 1.35 [null]
PNPC1 Gas type index = 1 []
Diameter of pipe = 8 [cm] or 15 [cm]
Pipe length = 5 [m]
Relative roughness = 1e-05 [null]
PNBP001 Gas type index = 1 []
Hydraulic diameter = 8 [cm]
Curvature radius = 10 [cm]
Center angle = 60 [degree]
Relative roughness = 1e-05 [zero]
6.3 Ventilation System of a Building 77

6.3.2 Submodels and Parameters

Table 6.4 Parameters for the simulation model of a ventilation system (part 2)
Item Parameter
TPTE001 Gas type index = 1 []
Diameter at port 1 = 15 [cm]
Diameter at ports 2 and 3 = 20 [cm]
Friction factor in the main branch = 0.1 [null]
Friction factor side branch = 1.2 [null]
PN3P000 Gas type index = 1 []
Pressure drop coefficients = Idelchik
Side branch diameter (port 1) = 15 [cm]
Straight passage diameter (port 2 and 3) = 15 [cm]
Angle between side branch and straight passage = 90 [degree]
Critical Reynolds number = 5000 [zero]
Time constant = 1e-06 [s]
Transition accuracy = 0.9 [null]
PNL0000 Gas type index = 1 []
Model = polytropic
Diameter of pipe = 10 [cm], 15 [cm] and 20 [cm]
Pipe length = 1 [m], 2 [m], 3 [m] and 20 [m]
Polytropic constant = 1.35 [zero]

Table 6.5 Simulation time Simulation settings

for the simulation model of a
ventilation system Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.1 [s]

6.3.3 Results and Analysis

To generate the airflow, we use a compressor. Actually, a blower should be used for
this purpose. However, modeling a blower in Simcenter Amesim is quite complex
and only feasible with additional efficiency tables. Therefore, we want to do without
it in the model in Fig. 6.4. The compressor is driven by a motor. Different speeds are
to be investigated.
78 6 The Pneumatic Twin

Fig. 6.4 Mass flow rate of the air at the compressor at different speeds

Fig. 6.5 Mass flow of the air at all three outlets of the ventilation system

In Fig. 6.4, we look at the mass flows of air at the end of the three branches of
the ventilation system. There is a different mass flow in each pipe. Depending on
the length and diameter of the pipes, but also due to diversions, there are different
pressure drops. This further complicates the behavior of the pipe system.
Reference 79

You can see the air distribution at the three outlets of the duct system in Fig. 6.5.
Interpret what this means for the ventilation system. Is it desirable to have a strong
uneven distribution of air flows? How does this affect the indoor climate inside the
building? Also work on the following suggestions.

Problems

We want to look at following problems and solve them with the model of the venti-
lation system:

6.5 Try changing the speed distribution on the compressor. How does this affect the
airflow?

6.6 Does the air distribution stay the same if they keep increasing the speed of the
compressor?

6.7 Investigate how the air distribution changes when they make individual pipes
thinner.

6.8 Each of the ventilation pipes should have a slider. You can open and close it.
Sliders result in changed pressure drops at the outlets, on which in turn the volume
flow depends. How can they create a uniform air distribution?

6.9 Create additional flow channels and try to adjust them so that the air is distributed
as evenly as possible at the outlets.

Reference

1. https://www.ansys.com/
Chapter 7
The Electric Twin

7.1 Permanent Electric Motor with Load

The electrical motor model shown in Fig. 7.1 is a permanent magnet motor. It can
be found in larger fluid energy machines, such as for driving ship propellers or com-
pressors. The model can work either as a motor or a generator. The model of the
machine or generator is [EMDPMDC01] which can be seen as reversible.

Important: The electric current in the rotor is needed to produce torque.

It is obtained by electromagnetic induction from the magnetic field of
the stator winding [1]. This produces waste heat inside the armature.
An induction motor can be made without electrical connections to the
rotor. An induction rotor can be either wound or squirrel-cage type [1].

7.1.1 Simulation Model

Let us build the simulation model as shown in Fig. 7.1, in which the motor is intro-
duced and connected to the electric current. The current is given by a function from
the signal library. The signal step model [UD00] is a duty cycle submodel with a
dimensionless output. The user may specify three stages giving a start value, an end
value, and the duration for each stage. Linear interpolation is used to determine the
output. Thus constant sections, ramps, and steps may be constructed with the signal
library. For the signal library, a designer dialog is given to define the specific appli-
cation to create the piecewise linear function. It is available by double clicking on
the icon in parameter mode and in simulation mode. The item is used to generate
piecewise linear signals like ramps, steps, squares, saw tooth, or trapezoidal signals.
Using the electrical motor convention, the rotor relative speed can be measured
by sensors for torque T and rotary speed ω as an output speed at the mechanical

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 81

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_7
82 7 The Electric Twin

Fig. 7.1 Permanent electric DC motor with voltage source and mechanical load

Fig. 7.2 Stepwise voltage increase as an input to the permanent electric motor

port and the electromagnetic torque can be measured with T as an output torque at
the mechanical port. When ω and T are of same sign, the machine is working as a
motor and the output mechanical power P is positive. When ω and T are of opposite
sign, the machine is working as a generator and the output mechanical power P is
negative. This formula can be used for further examinations.
The diagram in Fig. 7.2 shows a stepwise voltage increase and the resulting rotor
speed of this example. You can choose whether the machine is acting as a motor or
generator by looking at the signs. If the signs of the speed (n) and the torque (M) are
different, the result of the mechanical power is negative (P) and thus the machine
7.1 Permanent Electric Motor with Load 83

Table 7.1 Parameters for the simulation model of the permanent motor with load
Item Parameter
EMDPMDC01 Reference temperature = 25 [degC]
Armature winding resistance = 0.6 [Ohm]
Corrective coefficient armature winding = 0 [1/K]
Armature winding inductance = 0.012 [H]
Electromotive force and torque constant = 1.8 [V*s/rad]
Corrective coefficient electromotive force = 0.0 [1/K]
UD00 Number of stages = 3 []
Cyclic = no
Time at which duty cycle starts = 0 [s]
Output at start of stage 1 = 48 []
Output at end of stage 1 = 48 []
Duration of stage 1 = 3 [s]
Output at start of stage 2 = 100 []
Output at end of stage 2 = 100 []
Duration of stage 2 = 3 [s]
Output at start of stage 3 = 200 []
Output at end of stage 3 = 200 []
Duration of stage 3 = 3 [s]
THTS1 Temperature at port 1 = 20 [degC]
RL02 Moment of inertia = 0.005 [kg m**2]

is operated as a generator. As can be seen, the signs are different in this example
and the machine is regarded as a generator. The following formula gives the ratio of
speed and torque:
P = 2π · M · n (7.1)

With the variables P for power in [W], M torque [Nm], and n speed [1/s] . The
electromotive force and torque coefficient and the armature winding resistance are
corrected with the temperature. The variables to solve this equation could be taken
from the sensors in the model (see Table 7.1).

7.1.2 Submodels and Parameters

To improve the function of the electric model, set the simulation time to 9.0 [s]
according to Table 7.2. The armature current can then be computed inverting the
generalized Ohm’s law equation for the rotor phase. For details concerning the cal-
culation of power, energy, and activity in this submodel, please refer to the definitions
specific to this library in the help menu.
84 7 The Electric Twin

Table 7.2 Simulation time Simulation settings

for the simulation model for
permanent motor with load Start time = 0 [s]
Final time = 9 [s]
Print interval = 0.001 [s]

The simulation times for the twin of the permanent electrical motor with load are
given in Table 7.2.

7.1.3 Results and Analysis

When looking at the simulation results of speed and torque in Fig. 7.3, two curves
are now shown in the diagram. This is always helpful if you want to explain how the
values influence each other. After successful simulation, you can click on the result
variable and drag it into an existing diagram like in Fig. 7.3.
On one hand, the relative motor speed due to the different currents is shown. After
each step, an oscillation of the rotor speed can be seen, even if the electromagnetic
torque remains more or less constant. It reacts to the oscillations in the generator’s
magnetic field.

Fig. 7.3 The torque [rev/min] and speed [Nm] of the electric permanent DC motor
7.2 How an Asynchronous Motor Works? 85

Problems

For our model of the electric motor we can make several working suggestions. Take
a look at a few work suggestions that we can explore with our twin: To do this, they
should find and modify the appropriate model parameters in the simulation model.

7.1 Change the armature winding resistance to 1.0 [Ohm] and to 2.4 [Ohm]. Explain
what happens to the results.

7.2 Change the temperature of the [THTS1] from 20 [degC] to 40 [degC]. What
happens?

7.3 Couple the motor’s armature with an additional convective cooling element
from the thermal library. Test different materials, such as iron, cast iron, aluminium,
etc.

7.4 Change the load’s moment of inertia. Explain what is happening.

7.2 How an Asynchronous Motor Works?

A three-phase, asynchronous motor consists of two basic components, stator and

rotor. The stator is the fixed part of the motor. It has a stator core and field windings
often made of metal wires. Inside the stators of three-phase asynchronous motors,
there is a separate winding for each phase. It is operating with three-phase alternating
current. An asynchronous motor model setup in Fig. 7.4 shows a three-phase motor
operated with all three phases of alternating current. These asynchronous motors are

Fig. 7.4 Asynchronous

squirrel-cage induction
machine with three phase
and mechanical load
86 7 The Electric Twin

often used in industry for pumps, tools, or even electric vehicles. Friction at mechan-
ical load lowers the efficiency of the motor. Heat conduction plays an important role
for its efficiency.

7.2.1 Simulation Model

Let us build a simulation model of the electric motor where we can change the three
phases of the electric current stepwise by a sinus signal function. The sinus functions
can be modeled by [SIN0]. Figure 7.4 shows the input voltage of the asynchronous
motor. The phases are each shifted by 120 [degree]. The motor is a squirrel-cage
induction machine in a rotor frame. The simulation model [EMDSCIM01] is used
to model linear Squirrel-Cage Induction Machine (SCIM) and it is used to rotate a
load.
In the [EMDSCIM01], the reference angle used for transformation is the rotor
position. In the rotor frame, the model quantities vary with the slip frequency at
steady-state operating points. The reference angle used for transformation is the
rotor flux linkage position. In the rotor flux linkage frame, the model quantities are
constant at steady-state operating points.

7.2.2 Submodels and Parameters

The important parameters for the submodels of the three-phase motor setup are
compiled in Table 7.3. Two submodels [EMDSCIM01] and [EMDSCIM02] return
the same results. The model [EMDSCIM02] produces usually faster simulations but
it may raise numerical problem when the currents tend to zero.
To improve the function, an additional constriction should be placed, set the
simulation time to 0.04 [s] according to Table 7.4. Use [EMDSCIM01] when the
best robustness is required. Note also that [EMDSCIM01] internal variables in the
frame are not consistent with the internal variables of the control elements.

7.2.3 Results and Analysis

In Fig. 7.5, you can see the phase shift of the three different currents. The machine
windings are balanced, and two different connections are possible. The enumeration
winding connection enables to choose between star and delta connection for the
equivalent windings A, B, and C.
The three-phase electrical ports are associated on the icon with the notations I, II,
and III. The star and delta connections for the windings A, B, and C as well as the
winding voltages and currents are defined within this convention.
7.2 How an Asynchronous Motor Works? 87

Table 7.3 Parameters for the simulation for the electric three-phase motor
Item Parameter
EMDSCIM01 Winding connection = star [connectType]
Number of pole pairs = 1 []
Reference temperature = 25 [degC]
Stator winding resistance = 1 [Ohm]
Corrective coefficient on stator resistance = 0 [1/K]
Stator cyclic inductance = 0.25 [H]
Rotor time constant = 0.153 [s]
Corrective coefficient on rotor time = 0 [1/K]
Dispersion coefficient = 0.066 []
RL02 Moment of inertia = 0.005 [kgm**2]
THTS1 Temperature at port 1 = 20 [degC]
W000 Fixed angle = 0 [degree]
SIN0-1 Sine wave frequency = 50 [Hz]
Mean level = 0 []
Sine wave amplitude = 230 [s]
Phase shift = 0 [degree]
SIN0-2 Sine wave frequency = 50 [Hz]
Mean level = 0 []
Sine wave amplitude = 230 [s]
Phase shift = −120 [degree]
SIN0-3 Sine wave frequency = 50 [Hz]
Mean level = 0 []
Sine wave amplitude = 230 [s]
Phase shift = +120 [degree]

Table 7.4 Simulation time Simulation settings

for the simulation model of
the electric three-phase motor Start time = 0 [s]
Final time = 0.04 [s]
Print interval = 0.0001 [s]

Problems

Let’s look at some problems that we can now investigate with our electric motor twin.
To do this, you should modify the corresponding model parameters in the simulation
model.

7.5 Change frequency of the three sinusoidal signal functions up to 100 [Hz].

7.6 Try a different stator winding resistance like 2 or 3 [Ohm].

88 7 The Electric Twin

Fig. 7.5 Phase-shifted voltage curves of the asynchronous motor with three phase and mechanical
load

7.7 Change the rotor time constant to 0.165 [s]. What happens?

7.8 Change the load of the model and change the limit temperature of the armature
of the motor. Give an interpretation of your results.

7.3 Electric Generator with Resistor and Heat Transfer

The permanent magnet motor or generator can be defined as a motor which includes
a permanent magnet pole, and therefore it is called Permanent Magnet DC Motor
(PMDC). The motor includes an armature core, commutator, and armature winding.
There are usually two different types of windings in a conventional DC motor. A
solenoid can be used to make the flux working within the air gap in its place of the
field winding. This type of machine is always reversible, so it can work either as a
motor or as a generator. The rotor structure is similar to the straight DC motor.

7.3.1 Simulation Model

Let us build a motor simulation model in combination with the thermal library that
was already described before and combine it with a heat transfer problem. Figure 7.6
shows the input voltage permanent magnet DC motor.
7.3 Electric Generator with Resistor and Heat Transfer 89

Fig. 7.6 Electric permanent magnet DC motor (PMDC) in combination with the thermal library

The model setup in Fig. 7.4 shows a generator operated by an mechanical drive
[OMEGC0]. The main function of field winding is to produce the functioning mag-
netic flux within the air gap as well as wound on the stator of the motor while
armature winding can be wound on the rotor. Inactive carbon brushes are pushed
on the commutator like in conventional DC motor. Using the motor convention, the
rotor relative speed can be measured with W as an output speed at the mechanical
port and the electromagnetic torque can be measured with T as an output torque at
the mechanical port. When W and T are of same sign, the machine is working as a
motor and the output mechanical power P is positive. When W and T are of opposite
sign, the machine is working as a generator and the output mechanical power P is
negative.

7.3.2 Submodels and Parameters

The important parameters for the submodels of the generator setup are compiled in
Table 7.5. There are two submodels, namely, [EMDPMDC01] and [THC000]. The
model [EMDPMDC01] produces a heat source to heat up the mass model [THC000].
To improve the function, an additional constriction should be placed, set the
simulation time to 0.04 [s] according to Table 7.6. Use [EMDSCIM01] when the
best robustness is required. Note also that [EMDSCIM01] internal variables in the
frame are not consistent with the internal variables of the control elements.
90 7 The Electric Twin

Table 7.5 Parameters for the simulation for the electric three-phase motor
Item Parameter
EMDPMDC01 Reference temperature = 25 [degC]
Armature winding resistance at reference temperature = 0.6 [Ohm]
Corrective coefficient on armature winding resistance = 0 [1/K]
Armature winding inductance = 0.012 [H]
Electromotive force and torque constant at reference temperature =
1.8 [V*s/rad]
Corrective coefficient on electromotive force and torque constant = 0
[1/K]
EBR02 Type of reference resistor value = resistor value directly entered by
user
Resistance = 1 [Ohm]
THSD00 Solid type index = 1
Material definition = pure aluminium (Al)
THC000 Solid type index = 1
Mass of material = 0.5 [kg]
UD00 Number of stages = 3 []
Cyclic = no
Time at which duty cycle starts = 0 [s]
Output at start of stage 1 = 48 []
Output at end of stage 1 = 48 []
Duration of stage 1 = 3 [s]
Output at start of stage 2 = 100 []
Output at end of stage 2 = 100 []
Duration of stage 2 = 3 [s]
Output at start of stage 3 = 200 []
Output at end of stage 3 = 200 []
Duration of stage 3 = 3 [s]

Table 7.6 Simulation time Simulation settings

for the simulation model of
the electric three-phase motor Start time = 0 [s]
Final time = 10.0 [s]
Print interval = 0.01 [s]

7.3.3 Results and Analysis

To examine the simulation model of the generator, we check the torque at the shaft
of the generator firstly to see if mechanical load is provided. In Fig. 7.7, we can see
that all boundary conditions are set correctly.
7.3 Electric Generator with Resistor and Heat Transfer 91

Fig. 7.7 Checking the boundary conditions of the generator’s torque based on mechanical drive

In Fig. 7.8, the electric power is presented which can be calculated by the P
formula. The mechanical load will be converted into electric capacity in the generator
and the direct current machine will not only generate electric energy, but also heat.

P =U·I (7.2)

The thermal losses are identified with Joule’s losses in the machine which can be
estimated by the changing temperature in the generator’s housing, as can be seen in
Fig. 7.9. Joule heating, also known as resistive, resistance, or Ohmic heating, is the
process by which the passage of an electric current through a conductor produces
heat.
The temperature change in the housing is influenced by the mass of the housing
[THC000] and the material. Here in Fig. 7.9 we used aluminium, but alternative
materials like iron or cats iron can also be used. We can see how the temperature
changes inside the generator housing.
Cooling of housing for electric generators as well as the construction and design
of fins for heat conductors is an important topic. It is widely used for different
applications like wind turbines, power plant, and other fluid energy machines. Here
we see a good example of the connection between the thermal library and the electric
library.
92 7 The Electric Twin

Fig. 7.8 Changes of the current and voltage steps based of the production of the generator

Fig. 7.9 Joule’s losses produce a heat source in the housing of the generator

Problems

We want to have a look at some problems that we now can explore our generator. To
do this, you should modify the simulation model.

7.9 Change the thermal mass that is connected to the housing to 5 [kg] and to 10
[kg] and compare the results. Use cast iron as alternative housing material. Discuss
the results.
Reference 93

7.10 Change the armature winding resistance at reference temperature to 2 and 6

[Ohm]. What happens?

7.11 How long does it take to heat up the housing up to 350 [K]. Do you see any
influence on the current?

7.12 Try to build a convective cooling unit with the thermal library to cool down
the electric housing.

Reference

1. IEC 60050 Rotation Machinery – General, IEV ref. 411-31-10: Induction Machine – an asyn-
chronous machine of which only one winding is energized. Publication date: 1990-10. Section
411-31
Chapter 8
Analysis of Complex Technical Systems

8.1 How Does the Liquid Piston Compressor Work?

The storage of energy, e.g. from wind or water turbines, is one of the main objectives
in the future development of stable power grids [1, 2]. In this chapter, a model for a
liquid piston compressor (LPC) is presented. This LPC can be used to store mechan-
ical energy produced by a wind or water turbine [3]. Based on a short explanation of
the structure and function of an LPC, it is shown how such a system can be modeled
from existing hydraulic, pneumatic, and mechanical components. In sections that
build on one another, a basic model is first developed and parameterized. Subse-
quently, the model is successively extended, e.g. second piston, control system, etc.
and simulation results are presented and discussed in diagrams after each section.
Finally, a parameter study on the overall model is presented.

Important: In Amesim all elements of the different libraries can be connected

and combined to create more complex models. They can be combined with
each other or with the signal library. In the student version only some basic
libraries are included.

8.2 Design and Function of a Liquid Piston Compressor

In one of the simplest designs of a compressor with a liquid piston, the liquid piston
acts as a direct hydraulic-pneumatic transformer. This possibility of compressing
air in the application of an energy storage system was presented, among others,
in [4]. Figure 8.1 schematically shows the circuit diagram of such a system with a
hydraulic-pneumatic transformer.
In this process, a cylinder filled with gas is filled with a liquid, e.g. hydraulic oil
or water, by a hydraulic pump (the hydraulic valve was controlled for this purpose—
switching position 2). The liquid flowing into the cylinder compresses the gas in the

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 95

F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_8
96 8 Analysis of Complex Technical Systems

Fig. 8.1 System with technical components of a liquid piston compressor (LPC) [3]

cylinder and the pressure in the system increases. When the pressure prevailing in
the accumulator is reached, the check valve “2” opens and releases the gas flow to
the accumulator. When the cylinder is almost completely filled with the liquid, the
hydraulic valve switches again (switching position 1) and the liquid in the cylinder
can run back into the liquid tank. The vacuum thus created ensures that the cylinder
is filled with gas from the environment via check valve “1”. The process can then
run again. We want to present three variants of the LPC in detail.

8.3 Liquid Piston Compressor with One Cylinder for One

Stroke

8.3.1 Simulation Model

For modeling purposes, the existing components from the areas of hydraulics, pneu-
matic, and mechanics are to be used first. For this purpose, the real system is replaced
by the combination of a spring-loaded single-acting hydraulic cylinder in combina-
tion with a spring-loaded single-acting pneumatic cylinder. To couple the two cylin-
ders, a mass is interposed, since coupling of both cylinders is not possible without
an inertial mass (natural frequency of the system). The model for a liquid piston
Compressor with One Cylinder for One Stroke can be seen in Fig. 8.2.
The role of the relief valve [RV010] is to limit the upstream pressure within a
hydraulic circuit and thus protect hydraulic components from over pressure. This
component is also known as pressure limiting valve, maximum-pressure valve, or
safety valve [5].
8.3 Liquid Piston Compressor with One Cylinder for One Stroke 97

Fig. 8.2 Digital twin for liquid piston compressor with one cylinder for multiple strokes

8.3.2 Submodels and Parameters

To improve the function of the model, set the simulation time to 10 [s] according to
Tables 8.1 and 8.2 in this submodel, refer to the definitions specific to this library in
the help menu. The fluid [FP04] inside the hydraulic cylinder is an hydraulic oil.
The existing model must now be parameterized accordingly. In the following only
those parameters are shown, which deviate from the basic settings. The simulation
times for the simulation model are given in Table 8.3.

8.3.3 Results and Analysis

To check, if the twin of the first liquid piston compressor is working, we plot the
movement of the mass element between the pneumatic and the hydraulic chamber.
The result is presented in Fig. 8.3, and we can see that the hydraulic pressure-relief
valve [RV010] is working and switches and the mass reaches its maximum position
at about 4.5 [s].
From the results, the pressure curves inside the hydraulic cylinder, the pneumatic
cylinder, and the pressure accumulator can be evaluated, among other things. The
98 8 Analysis of Complex Technical Systems

Table 8.1 Parameters for simulation model of the liquid piston compressor with one cylinder for
one stroke (part 1)
Item Parameter
PNCH022 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
Gas type index = 1
Volume = 0.1 [L]
Thermal exchange coefficient = 0 [J/m**2/K/s]
Thermal exchange area = 0.1 [m**2]
External temperature = 293.15 [K]
PNCV010 Gas type index = 1
Check valve cracking pressure = 0.2 [bar]
Check valve mass flow rate pressure gradient = 10 [g/s/bar]
Valve hysteresis = 0 [bar]
PNGD00 Gas type index = 1
Fluid definition = air
Properties definition = semi perfect
PNAS001 Temperature at port 1 = 293.15 [K]
PNJ002 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
Gas type index = 1
Model = with thermal exchange
Use initial displacement = yes []
Displacement of piston = 0.5 [m]
Piston diameter = 25 [mm]
Rod diameter = 12 [mm]
Length of stroke = 0.5 [m]
Dead volume at port 1 end = 50 [cm**3]
Thermal exchange coefficient = 0 [J/m**2/K/s]
External temperature = 293.15 [K]
Viscous friction coefficient = 0 [N/(m/s)]
Leakage coefficient = 0 [g/s/barA]
Spring preload = 0 [N]
Spring rate = 1 [N/m]
Pressure in rod chamber = 1.013 [barA]
Spring rate at endstops = 100000 [N/mm]
Damping coefficient at endstops = 100000 [N/(m/s)]
Deformation on endstops at which damping rate = 0.001 [mm]
8.3 Liquid Piston Compressor with One Cylinder for One Stroke 99

Table 8.2 Parameters for simulation model of the liquid piston compressor with one cylinder for
one stroke (part 2)
Item Parameter
RV010 Index of hydraulic fluid = 0
Relief valve cracking pressure = 5 [bar]
Relief valve flow rate pressure gradient = 500 [L/min/bar]
Valve hysteresis = 0 [bar]
HJ0023 Pressure at port 1 = 1.013 [bar]
Index of hydraulic fluid = 0
Use initial displacement = yes
Displacement of piston = 0 [m]
Piston diameter = 25 [mm]
Rod diameter = 12 [mm]
Length of stroke = 0.5 [m]
Dead volume at port 1 end = 50 [cm**3]
Viscous friction coefficient = 0 [N/(m/s)]
Leakage coefficient = 0 [L/min/bar]
Spring preload = 0 [N]
Spring rate = 1 [N/m]
Pressure in rod chamber = 1 [bar]
Spring rate at endstops = 100000 [N/m]
Damping coefficient on endstops = 10 [N/(m/s)]
Deformation on endstops at which damping rate is fully effective = 0.001 [mm]
PU001 Index of hydraulic fluid = 0
Pump displacement = 2 [cc/rev]
Typical pump speed = 1500 [rev/min]
PM000 Shaft speed = 1500 [rev/min]
FP04 Type of fluid properties = elementary
Index of hydraulic fluid = 0
Temperature = 40 [degC]
Name of fluid = unnamed fluid
Density = 850 [kg/m**3]
Bulk modulus = 17000 [bar]
Absolute viscosity = 51 [cP]
MECMAS21 Velocity at port 1 = 0 [m/s]
Displacement at port 1 = 0 [m]
Use friction = no
Endstop type = none
Mass = 0.1 [kg]
Inclination (+90 port 1 lowest, −90 port 1 highest) = 0 [degree]
TK000 Tank pressure = 0 [bar]
100 8 Analysis of Complex Technical Systems

Table 8.3 Simulation time Simulation settings

for the simulation model of
the liquid piston compressor Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]

Fig. 8.3 Displacement of the mass element between hydraulic and pneumatic pistons

pressure curves in the hydraulic cylinder and the pneumatic cylinder are shown in
Fig. 8.4.
It can be clearly seen that the compression process takes place exactly once. When
the mass reaches its maximum position, the cylinders are not able to move further
on. So the hydraulic pump increases the pressure in the hydraulic system very fast.
At a pressure of 5 [bar] the pressure-relief valve opens and the hydraulic oil flows
through this valve directly back to the tank. The pressure curves of the pressure
accumulator are shown in Fig. 8.5. First, the pressure in the pneumatic cylinder is
increased by the extension of the hydraulic cylinder, which corresponds to a retraction
of the pneumatic cylinder, until the accumulator pressure is reached. A slight pressure
fluctuation in the fluid [FP04] of the hydraulic cylinder can be seen.
The valve [RV010] is normally closed. When the pressure drop across the valve
exceeds the relief valve cracking pressure (typically a spring force), the valve opens
and let the fluid flow across so that the pressure drop gets regulated to the cracking
pressure [5]. Subsequently, the pressure in Fig. 8.5 remains almost constant and the air
present is forced into the pressure accumulator. In the process, the pressure increases
both in the pressure accumulator and in the pneumatic and hydraulic cylinders. When
the hydraulic cylinder is completely extended (end position), the hydraulic pressure
8.3 Liquid Piston Compressor with One Cylinder for One Stroke 101

Fig. 8.4 Diagram of pressure curve in the hydraulic, pneumatic cylinder

Fig. 8.5 Diagram of pressure curve in the pressure accumulator chamber

increases like mentioned before until the opening pressure of the pressure-relief valve
is reached and the process is finished.
102 8 Analysis of Complex Technical Systems

8.4 Liquid Piston Compressor with One Cylinder and

Multiple Strokes

8.4.1 Simulation Model

In order to be able to run through the compression process several times, a control
system for the movement of the piston must be implemented. For this purpose, a
2/3-way hydraulic valve [HSV23_02], a displacement sensor [MECDS0B], and a
“THRESHOLD” are to be added to the model, as can be seen in Fig. 8.6. The sensor
gives a trigger signal for the lift and controls the valve.
In Fig. 8.6, we can see the digital twin for liquid piston compressor with one cylin-
der for multiple strokes with components from the Amesim model. We improved
the model from the further section and added a two-position three-port hydraulic
servo-valve [HSV23_02] to the model. The existing model must now be parame-
terized accordingly. In the following only the parameters of the added components
are shown, which differ from the basic settings to avoid repetition. Only additional
parameters will be given in Table 8.4.

Fig. 8.6 Digital twin for liquid piston compressor with one cylinder for one stroke
8.4 Liquid Piston Compressor with One Cylinder and Multiple Strokes 103

Table 8.4 Additional parameters for liquid piston compressor with one cylinder for multiple strokes

Item Parameter
MECDSS0B Sign convention: positive from = 0
Offset to be subtracted from displacement = 0 [m]
Gain for signal output = 1 [1/m]
TRIG0 Initial output value = low
High input threshold value = 0.499 null
Low input threshold value = 0.001 null
High output value = 0 null
Low output value = 1 null
HSV23_02 Index of hydraulic fluid = 0
Valve rated current = 1 [mA]
Valve natural frequency = 80 [Hz]
Valve damping ratio = 0.8 null
Ports P to A characteristic flow rate at maximum opening = 100 [L/min]
Ports P to A corresponding pressure drop = 0.01 [bar]
Ports P to A critical flow number (laminar/turbulent) = 100 null
Ports A to T flow rate at maximum opening = 100 [L/min]
Ports A to T corresponding pressure drop = 0.01 [bar]
Ports A to T critical flow number (laminar/turbulent) = 1000 null

Table 8.5 Simulation time Simulation settings

for the liquid piston
compressor with multiple Start time = 0 [s]
strokes Final time = 10 [s]
Print interval = 0.01 [s]

8.4.2 Submodels and Parameters

Most parameters for the digital twin of the liquid piston compressor with one cylinder
and multiple strokes stay the same as in the previous model in Fig. 8.6 for the system
with one stroke. Therefore, we will make Table 8.4 for the parameters a bit shorter,
and add only the parameters for the additional submodels and deviating parameters.
Time steps and simulation time for the liquid piston compressor with multiple
stroke are given in Table 8.5.

8.4.3 Results and Analysis

The travels of the hydraulic and pneumatic cylinders show that they both work in
opposite directions. The signal from the displacement sensor is used to switch over
104 8 Analysis of Complex Technical Systems

Fig. 8.7 Displacement of the mass element between hydraulic and pneumatic pistons with multiple
strokes

the hydraulic valve via the “THRESHOLD” when the hydraulic cylinder is almost
completely extended. In Fig. 8.7, we can see that movement of the mass occurs two
times according to the trigger of the signal [TRIG0] and the displacement of the mass
element between hydraulic and pneumatic pistons with multiple strokes.
In Fig. 8.9, the pressure in the pneumatic and hydraulic cylinder is shown. The
hydraulic cylinder after the first stroke retracts again due to the residual pressure in
the pneumatic cylinder and the spring force. The pneumatic cylinder draws in new
air.
When the hydraulic cylinder is almost completely retracted, the process is
repeated. With each conveying process, the pressure in the pressure accumulator
increases step by step. It can be seen that pressure fluctuations inside the hydraulic
fluid are rather high. In technical applications, this behavior should normally be
avoided (see Fig. 8.10).
8.5 Liquid Piston Compressor with Two Cylinder and Multiple Strokes 105

Fig. 8.8 Digital twin for liquid piston compressor with two cylinder and multiple strokes

8.5 Liquid Piston Compressor with Two Cylinder and

Multiple Strokes

8.5.1 Simulation Model

In order to prevent the system from not compressing air during the emptying of
the hydraulic cylinder, i.e. not working, so to speak, the system can be extended by
another fluid piston. This configuration is shown in Fig. 8.8. The two fluid pistons
then work alternately, each using the hydraulic oil of the cylinder that is just emptying
to compress the air. For this purpose, the model must be appropriately planed with
two cylinders.
Now, we added a two-position four-port hydraulic valve [HSV24_02] to the model
to switch between the two different hydraulic cylinders. This valve is controlled also
by the trigger signal [TRIG0].
106 8 Analysis of Complex Technical Systems

Fig. 8.9 Pressure curves in hydraulic, pneumatic cylinder with multiple stroke and high oscillations

Fig. 8.10 Pressure curves in the pressure accumulator chamber with multiple strokes
8.5 Liquid Piston Compressor with Two Cylinder and Multiple Strokes 107

Table 8.6 Additional parameters for liquid piston compressor with one cylinder for multiple strokes

Item Parameter
HSV24_02 Index of hydraulic fluid = 0
Valve rated current = 1 [mA]
Valve natural frequency = 80 [Hz]
Valve damping ratio = 0.8 null
Ports P to A characteristic flow rate at maximum opening = 100 [L/min]
Ports P to A corresponding pressure drop = 0.1 [bar]
Ports P to A critical flow number (laminar/turbulent) = 1000 null
Ports B to T characteristic flow rate at maximum opening = 100 [L/min]
Ports B to T corresponding pressure drop = 0.1 [bar]
Ports B to T critical flow number (laminar/turbulent) = 1000 null
Ports P to B characteristic flow rate at maximum opening = 100 [L/min]
Ports P to B corresponding pressure drop = 0.1 [bar]
Ports P to B critical flow number (laminar/turbulent) = 1000 null
Ports A to T characteristic flow rate at max. opening = 100 [L/min]
Ports A to T corresponding pressure drop = 0.1 [bar]
Ports A to T critical flow number (laminar/turbulent) = 1000 null

8.5.2 Submodels and Parameters

Additional parameters and data for the new two-position four-port hydraulic servo-
valve [HSV24_02] were sampled in Table 8.6. All other parameters are similar to
the model in the previous section.

8.5.3 Results and Analysis

Finally, this model is to be extended according to the basic scheme in Fig. 8.11.
For this purpose, it was recorded how the cylinders move in both pistons. It can be
seen clearly in the diagram that the movement of the right piston is inverse to the
movement of the left piston. When the right piston moves up, the left piston moves
down. This is completely plausible. Also, the motion is in the same range because
both pistons are of the same size. It could be investigated how the movement of the
pistons changes if we choose different volumes.

Important: Comparing different variants of the same technical application is

an effective method for selecting the best solution. Even though the absolute
results may be somewhat different when compared with measurements. The
decision for a favored variant might be possible.
108 8 Analysis of Complex Technical Systems

Fig. 8.11 Three displacements of the mass elements of the two pistons of the LPC

Figure 8.12 shows the pressure curves of the pneumatic actuator on the left and
the pressure curve of the pneumatic actuator on the right. It should be mentioned

Fig. 8.12 Pressure curve in hydraulic, pneumatic cylinder with multiple strokes
Problems 109

Fig. 8.13 Pressure curve in the pressure accumulator chamber with multiple strokes

that we have not visualized the pressure variations in the hydraulic chamber as in
the previous section, but they are comparable. The pressure in the chamber of the
pressure accumulator is shown in Fig. 8.13. It can be seen that there are many more
stages forming the stepwise pressure rise in the pressure accumulator chamber.
We want to analyze our complex technical system and have a look at some prob-
lems to improve the model. You can modify the appropriate parameters in the model
and try out what’s happening.

Problems

8.1 Change the size of the pressure accumulator from 0.1 to 10 [L]. What happens?

8.2 Change the trigger signal of the sensor and try to reduce the fluctuations inside
the hydraulic chamber.

8.3 Add an additional third hydraulic and pneumatic actuator and repeat the simu-
lations.

8.4 Examine the temperature of the gas inside the pneumatic accumulator. Is it
adiabatic? How can you change the model to reduce the temperature inside the
accumulator?

8.5 Think about a new component to store the heat energy of the gas in the accu-
mulator.
110 8 Analysis of Complex Technical Systems

8.6 Add a hydraulic chamber in the back flow area between the 2/4-way hydraulic
valve and the pump.

References

1. Staudacher, T., von Roon, S., Vogler, G.: Energy storage - status, perspectives and eco-
nomic viability; study summary report; client boarfeleven; performed by Forschungsstelle für
Energiewirtschaft e.V., 03.2009
2. German Federal Ministry for Economic Affairs and Energy (The energy of the future,
2021): 8th Monitoring Report on the Energy Transition – Reporting years 2018
and 2019, Access: 2021-05-09. https://www.bmwi.de/Redaktion/DE/Publikationen/Energie/
achter-monitoring-bericht-energie-der-zukunft.html
3. Grün, S., Hübner, D.H., Ortwig, H.H., Rückert, F.U.: Investigation and optimization of a three
stage inter-cooled piston compressor for an energy storage system with pressurized air. In: Pro-
ceedings of the Interdisciplinary Conference on Mechanics, Computers and Electrics (ICMECE
2022), 6–7 October 2022, Barcelona, Spain
4. Grün, S., Hübner, D.H., D., Molter, J.: Investigation and simulation based optimization of an
energy storage system with pressurized air. Arch. Thermodyn. 42(4), 183–200 (2021). https://
doi.org/10.24425/ather.2021.139658. https://journals.pan.pl/Content/122213/PDF/art11_corr.
pdf
5. Siemens Digital Industries Software: Simcenter Amesim Student Edition (2020)
Chapter 9
Digital Twins and Artificial Intelligence

9.1 Neural Networks in Nature

We want to give an example how one can use an artificial neural network for the
physical libraries described before. Neural networks can be used widely in different
fields of business administration, economics, science, and technology. They were
built similar to biological neural networks of animals. In the following, we will see
that our brain or the brain of any other animal in a certain form performs its services
with the help of matrix calculation. For this some basics have been formulated by
Spitzer [1].

Important: Artificial intelligence (AI) can be used to reproduce results from

digital twins much faster, but the quality of the results could be weaker. We can
also use AI to control the twins or make results of the calculations available
faster and with less effort. AI can be used in many different areas of science.

9.2 Neural Networks and Digital Twins

Sensory cells in the eye, ear, nose, skin, etc. react specifically to physical or chemical
stimuli and produce output by discharging a binary electrical signal. The so-called
action potential. We can describe this in mathematics with the coded 0 and 1. This
signal is transmitted via the axon of the sensory cell and usually reaches a large
number of other neurons. The nerve cells, which further process the corresponding
input. In our brain, each neuron receives impulses from about 10.000 other neurons
in this way. An input cell is connected to another nerve cell and, given sufficient
stimulus, sends an electrical impulse along the connecting nerve fiber.
The AI can also collaborate with complex networks of digital twins and predict
new results. Often the predictions are much faster than with a complex system of
digital twins, but the quality can be weaker. To produce data for such neural networks,

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 111
F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_9
112 9 Digital Twins and Artificial Intelligence

Fig. 9.1 Artificial neural networks are used to collaborate with digital twins and predict results

you also need, for example, the pneumatic or thermal library as shown in Fig. 9.1.
However, the current does not flow directly into a nerve cell, but into a synapse, which
usually docks with so-called dendrites or with the cell body of N. The synapse then
generates more or less neurotransmitter molecules, e.g. glutamate, depending on its
own strength. The synapse then generates more or less neurotransmitter molecules
like glutamate depending on its own strength [1].
The synapse then generates more or less neurotransmitter molecules, depending
on its own strength, which pass over to the nerve cell. The synapse then generates
an impulse, which passes over to the nerve cell, depending on its own strength. The
synapse then generates an impulse when nerve cell receives an impulse. The nerve
cell in turn generates an impulse when a sufficient number of transmitter molecules
arrive. An important role is played by the fact that very many nerve fibers end and
therefore receive transmitter molecules from many synapses simultaneously.
This behavior has to be transferred to a mathematical formulation. It depends
on the synapse whether it emits inhibitory or excitation transmitter molecules. The
synapse thus multiplies the incoming signal (0 or 1) by its own strength, which can
be modeled as a number in the interval [−1, 1]. The neuron sums all the values
obtained in this way and fires when the summed value exceeds a critical threshold
characteristic of the neuron. Creation of an artificial neural network (ANN) is simple
in Simcenter Amesim, as shown in Fig. 9.2. A ready-to-use model is already included.
The [DYNNNFF01] submodel allows evaluating a restricted feed-forward neural
network model having one hidden layer.
9.2 Neural Networks and Digital Twins 113

Fig. 9.2 Creation of an artificial neural network with Simcenter Amesim based on signal library

Replacing some Simcenter Amesim physical components by a feed-forward arti-

ficial neural network might be a source of undesired algebraic loops since all the
dynamics of the original model have been removed. Indeed, such a neural network
establishes a direct algebraic relationship between each output and all the inputs. If
necessary, algebraic loops can be broken by placing some first-order lags or some
discrete delays on some of the neural network’s outputs. Beware that time constants
need to be carefully set, depending on the dynamics of the system from which a
surrogate model is being created.
The icon of [DYNNNFF01] submodel is a dynamic icon. When placing it on the
sketch, the user must indicate the number of inputs and the number of the outputs of
the neural network model. In addition, the number of neurons used for the training
process under MATLAB or an external program which could be provided on our
homepage needs to be provided [4]. A backpropagating neural network is a network
of neurons, the simplest unit of this technique. A neuron receives inputs, changes
its internal state (activation state), and produces outputs in accordance with inputs
and activation states. The connection of these elementary units forms the network,
connecting the outputs of certain neurons with the inputs of other ones [3].
To start a first test with the neural network data for learning is needed. In Fig. 9.3,
we connect our first neural network to input data and output data. We have to make
sure that we have two columns of input data for the network and one column of
output data as defined in Fig. 9.2. We will come back to this setup later. Firstly, we
want to explain how a neural network works and how learning can train the brain.
114 9 Digital Twins and Artificial Intelligence

Fig. 9.3 Adding data elements to the artificial neural network based on signal library

9.3 The Artificial Frog Model

Ever since Alessandro Volta used a frog to demonstrate the influence of electric
current, frogs have been widely used for scientific experiments. How a frog processes
what it sees? We will discuss a simple frog example for a neural network. Think
about a frog whose eye consists of three photo receptors which is shown in Fig. 9.4.
Thus, inputs of the network are propagated to the network’s output through the
neurons, following weighted connections. When the direction of propagated signals
goes always from inputs to outputs, i.e. signals are never propagated backwards,
the network is called a feed-forward neural network. The weights and activation
functions are modified during the training process in order to get values that provide
the best outputs. To train a neural network, a learning algorithm is used to minimize
a related cost function. For example, mean least squares can be used between the
desired output and the predicted one [3].
There are three events that are important for the frog. The frog sees a stork, a fly,
or just the blue sky. Each time, the photoreceptors are excited in a characteristic way.
The photoreceptors are now connected to three neurons that initiate an action when
they are sufficiently excited.
9.3 The Artificial Frog Model 115

Fig. 9.4 Photoreceptors of a

frog are now connected to
three neurons that initiate an
action when they are
sufficiently excited [2]

9.3.1 Submodels and Parameters

Let us consider a simple test case for the frog model [2]. We imagine a simple living
creature, e.g. like a frog in a pond. How can a frog processes what it sees? If the
stork is detected like excitation in pattern A, flight is initiated excitation of nerve
cell 1. For the frog the fly is the prey. If the fly is sighted (B), the frog sticks out its
tongue to catch the fly. This means excitation of nerve cell 3. If the frog sees the blue
sky (C), it activates digestion. Excitation of nerve cell 2. For this event, data can be
generated. In Fig. 9.5, we give you an example data set with 10 hidden neurons for
this case.
We now graphically represent the connection between the three visual cells and
the three nerve cells, the connecting nerve fibers and the synapses. In each of the
synapses, their strength is plotted as a value between -1 and 1. Each of the three
nerve cells fires when the sum of the transmitters arriving via the synapses exceeds
the measure 0.8 in an interval from 0 to 1 (see Fig. 9.6).
The model [DYNNNFF01] can be compared with the [DYNPSM02]. For now,
this process needs to be done using the MathWorks Neural Network toolbox for
MATLAB [5]. It is also a submodel that allows creating global approximations, and
it is especially suitable in presence of non-smooth or highly non-linear models. The
input file listed in Table 9.1 has to be generated by MATLAB or by our in-house
programmed optiflow neural network tool.
The only drawback of this submodel is that it currently has a dependency to the
commercial code MATLAB which is rather slow for creating the artificial neural net-
work, but not for evaluating it. Note that, just as [DYNPSM02], [DYNNNFF01] only
uses standard Amesim XY tables, which is particularly interesting to create a surro-
gate model that can be efficiently exported as a Simulink blackbox, as a Functional
116 9 Digital Twins and Artificial Intelligence

Fig. 9.5 Definition file frog_example.net for the frog with training and test data

Table 9.1 Data files written by optiflow neural network tool for Simcenter Amesim
Data file Description
B1.data Bias file for the connection of the input layer with hidden layer
B2.data Bias file for the connection of the hidden layer with output layer
W1.data Weighting file for the connections between input and hidden layers
W2.data Weighting file for the connections between hidden and output layers
Scaling_in.data File with the input scaling matrix
Scaling_out.data File with the output scaling matrix

Mock-up Unit (FMU), or executed on a real-time target. Examples for the XY-tables
needed are given in Table 9.1.

Important: Artificial intelligence and artificial neural networks are currently

a hot topic and under very fast development. Therefore, the information in this
book might be outdated already. To prevent this we developed a standalone
tool to calculate the weights of the neural network.

The icon of [DYNNNFF01] submodel is a dynamic icon. When placing it on

the sketch, the user must indicate the number of inputs and the number of the out-
9.3 The Artificial Frog Model 117

Fig. 9.6 Enter the input data for the experience of the frog’s life and hunting behavior

Table 9.2 Ports for the neural network model

Port Description
Port 1 Used to supply inputs to the neural network model
Port 2 Provides the responses estimated by the neural network model

puts of the neural network model. In addition, the number of neurons used for the
training process under MATLAB needs to be provided or you can use our in-house
programmed code [3]. For our frog model, we should define three inputs and outputs,
and 10 hidden neurons. The item component comprises two ports that are listed in
Table 9.2.

9.3.2 Optiflow Neural Network (ONN)

To try out simple neural network on your own and to predict test data results, we
offer you the optiflow neural network (ONN) tool. The tool can be downloaded under
[4]. It is easy to use and free of charge. The user interface to the tool can be seen in
Figs. 9.7 and 9.8. After starting the tool one can open a net-file to define the neural
network. Doing this is quite simple. An example file is given in Fig. 9.5. For our frog
we have to define the number of input cells (N inputs = 3), the number of hidden
neurons (N inputs = 10), and the number of output neurons (N out puts = 3).
118 9 Digital Twins and Artificial Intelligence

Fig. 9.7 Software tool optiflow neural network (ONN) can be used to create weights

Under the topic Connectcalls you define how to connect the sensor cells and neu-
rons against each other. The line (1 3 4 13) means that all three input cells have one
connection to the hidden layer (4 to 13). And all hidden neurons (4 to 13) have a con-
nection to the output cells (14 to 16). After loading the example file frog_example.net
you can start the training of your frog brain.

Important: Note that some basic knowledge of artificial neural networks is

beneficial before using this component [3]. We provide our own in-house
programmed tool optiflow neural network (ONN).
The tool and an example parameter set with data files for the frog model can
be downloaded at https://www.optiflow.htwsaar.de/.

As can be seen in Fig. 9.7 you just have to press the button Start Training. The tool
will start with the training for a certain number of epochs. The number of epochs
can be defined in the GUI together with the random seed of our random number
generator. This seed is used to ensure a reproducible simulation run.
9.3 The Artificial Frog Model 119

Fig. 9.8 Testing your neural network of the frog’s example

9.3.3 Results and Analysis

After finishing the training of your frog brain you can test the neural network. As
can be seen in Fig. 9.8, the neural network gives quite good results for prediction
of the frog’s behavior. So we can learn that 10 hidden neurons might be pretty well
to reproduce the frog’s behavior. The data files that are described in Fig. 9.5 can be
written with the menu topic File/Save Weights. Weights from further simulation runs
can be loaded and used for prediction.

Problems

We want to have a look at some problems that we now can explore with an artificial
neural network. To do this, you try the following things with our froggy.
9.1 Create a data set for the learning behavior of the frog. It should contain 100
entries of training data. Verify the results with your own test data.
9.2 Train your neural network with the frog information data set. And try to predict
the behavior.
120 9 Digital Twins and Artificial Intelligence

9.3 Visualize the test cases in a diagram and compare input and output data. Discuss
the result.

9.4 Try to create a mesh with 20 hidden layers and compare the results of both
meshes.

References

1. Spitzer, M.: Lernen – Gehirnforschung und die Schule des Lebens. Elsevier (2007)
2. Mathematik für Wirtschaftsingenieure; Skript, Hochschule für Wirtschaft und Technik des
Saarlandes (htw saar) (2020)
3. LMS Imagine.Lab Amesim Users Guide, Siemens Industry Software S.A.S (2016)
4. https://www.optiflow.htwsaar.de/
5. https://www.mathworks.com/help/matlab/ref/rand.html
Chapter 10
Conclusions

10.1 Overview of This Textbook

At the beginning of this book, we introduce the handling of the program Simcenter
Amesim. It was chosen because it is easy to learn and no deep mathematical skills or
programming knowledge is necessary. It can be used by students free of charge. We
started with the creation of a simple pocket calculator, with which they can perform
first mathematical calculations. In the individual chapters, digital twins from differ-
ent, technical fields are presented and developed. The authors have chosen to divide
them into mechanical, thermal, hydraulic, pneumatic, or electric twins. For exam-
ple, simple models of ventilation and tank systems or a solar collector are modeled.
The combination of digital twins with artificial intelligence is also presented. The
combination of digital twins with artificial intelligence is also presented. In the last
chapter, it is shown how a simple artificial neural network of the behavior of a frog is
modeled and how this can be used to control the digital twins. In the last chapter, it
is shown how a simple artificial neural network of the behavior of a frog is modeled
and how this can be used to control the digital twins.
A complex technical twin in which all the individual libraries have been combined
into one overall model of a new innovation, the liquid piston compressor (LPC).
This invention could be used, after further development, to store mechanical energy
generated by wind or water turbines. The book closes with a description how to
connect the digital twins with artificial intelligence, e.g. neural networks to control
the twins and learn from the data.
To carry out the exercises and solve the problems, the appropriate equipment is
very important. Each student must have the opportunity to work independently on a
computer. A desire was expressed to provide a second screen for each student. This
will allow one screen to be used to work with the program and the other to follow
the lecture. Switching back and forth between the program and the online meeting
on only one screen makes it difficult to follow the lecture, the second screen would
make this easier. Also it has to be mentioned that the program has a really good help
dialog with many examples from industry.

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 121
F. U. Rückert et al., Digital Twin Development,
https://doi.org/10.1007/978-3-031-25692-9_10
122 10 Conclusions

So far, exercises are given in the form of tasks from lesson to lesson, without
handing out a proper task template. What is desired is a written accompaniment to
the respective tasks, which have a real reference for better understanding, as well as
a textbook in which some processing steps can be looked up. Another suggestion
for improvement is to set aside time at the beginning of each lecture hour to discuss
final papers and clarify difficulties.

10.2 What You Can Take Away

The first run-through of the lecture with the program Simcenter Amesim proved
to be still very expandable, since both lecturers and students were new to using it.
Suggestions for improvement are to provide better equipment for students, a textbook,
as well as written task descriptions and to discuss the tasks at the beginning of each
lecture.
Through student evaluation, the critical points can be revised and eliminated for
future lectures. Thus there is a continuous learning process to improve the handling of
the system. If the program Amesim can be taught in an understandable way, this serves
very well as a basis for similar modeling and simulation programs. Furthermore, the
students are introduced to the practical experience and overlapping topics with other
lectures through the handling. They thus gain a deeper understanding in topics that
are otherwise discussed purely theoretically. But what outcome can we take from
this book?

• You will have noticed that in this essential we have tried to introduce the program
at the beginning and only briefly present the examples later on. We don’t want to
try to document all examples completely, but rather to encourage you to create
your own simulation model.
• Our goal is to make the reader want to create digital twins of technical devices
and machines himself. The aim is to train the eye for the essentials. You start with
simple models and roughly estimated volumes, distances, forces, or weights. In
the course of the work one goes then ever further into the detail.
• The interested reader, student, engineer, or designer should be able to create and test
first detailed, quantitative estimations and calculations for his new product before
he starts with the complex work of designing and constructing with common CAD
tools. Further, the simulation engineer can use the twin to search for boundary
conditions for more complex simulations.
• Digital twins may be a new phenomenon. However, drawings, sketches, and math-
ematical and physical equations have been created for engineering products in
advance of a design for a much longer time.
• The combination of digital twins together with artificial intelligence is an important
step. You can learn how to train an artificial neural network with a data set and
how to use this trained network to predict further results and compare them with
data for testing.
10.3 Teaching Methods and Gamification 123

10.3 Teaching Methods and Gamification

The book was written during winter semester 2020 and summer semester 2022 at
the University of Applied Sciences Saarbrücken (Germany). During this time, all
lectures had to be held as online classes with Microsoft Teams due to Corona crisis
regulations. Students could install the program Amesim at home and use it for their
homework.
In addition, all exercises were tested in face-to-face summer school together with
students from LAB University in Lappeenranta (Finland). We tested our textbook
in two semesters with students from LAB University (Finland) in such a way that
we first formed teams of two who sketched the examples one by one and solved the
given problems. During these exercises they learned how to improve the models and
at the end of the seminar they had to create their own complex digital twin together
with a 3D visualization.
The summer school lasts 1 week, during which time not all examples can be
worked on. Students had to create each twin from the illustrations in the book by
themselves using their own computers. During lecture, it was helpful to form collab-
orating teams of two or three to help each other. Initial uncertainties and difficulties
hopefully subsided during the review sessions. At the end of each seminar, all team
members created their own digital twins to a new, previously unknown problem and
had to present them to the audience to pass the exam.
Discussing and presenting the twin in front of the other ones always was a helpful
and interesting situation and often resulted in new ideas and further improvements
of the models.
Showing animations of the digital twin in 3D environment gave a good overview
and was always fun at the end of each presentation, e.g. they had to build a robot and
its 3D model was moving between cones to an exit. Here all the different libraries of
Amesim could be tested. To improve the teaching method, students were asked for
their feedback.
Further improvements of the teaching method with gamification were done in a
1-week workshop at Universidade de Aveiro (Portugal) in November 2022. Within
the framework of the teaching project e-CLOSE financed by the European Union
(EU), we have developed this form of seminar into a more self-motivating method
[1]. We combined the tasks from the problem section of each twin with the innova-
tive teaching method of role-playing. A role-play is a form of experiential learning
described by Russell and Shepherd in 2010 [2]. Students take on assigned roles of a
fictive character and act out those role through a scripted play. The method should
be carried out by two teams with four persons in each team. The two teams have to
work against each other. Each member of the group should take on a specific role,
like an actor on stage.
124 10 Conclusions

Important: Communication and new teaching methods are important. When

working with the models, users and engineers should learn how to create
the real applications. But always question how realistic a result is. Rely on
experience, never decide based on math alone, or you will end up failing.

Then the teams create a technical scenario with the digital twins and changed the
parameters and constraints to produce different results for the same model. Based
on the results, roles and rules need to be clearly defined in two scripts. As with the
hunting scene from the petroglyphs in Bryce Canyon, the moderator with the antlers
handed the two alternative scripts with the different results to the two teams and gave
them the task of discussing them and trying to convince the other side. Teams could
act out a specific discussion in this safe environment.
Our idea was that this could be a dynamic learning experience for actors to gain
knowledge about physical behavior by immersing them in simulated real-world prob-
lems. But also in industry, such an artificial dispute in a safe, moderated environment
can help teams learn very quickly and focus on a technical problem. Adrenalin is
provided free of charge and no additional coffee is needed.

10.4 Our Outlook for the Future

For further work, it would be very interesting to combine the digital twins on one
hand with 3D visualization to show the behavior in a much more practical way and to
develop more 3D models for the physical twins. The second most important step has
to be the combination of the digital twins with artificial intelligence or to depict the
physical models to neural networks. However, it must also be said that artificial neural
networks are definitely not the end of the line in artificial intelligence. Further work
will focus much more on reduced-order models (ROM) that lead to more sophisti-
cated predictions and better represent the digital twins. We also want to improve the
gamification teaching method role-play and create scripts with controversy, technical
tasks.

10.5 Important: Disclaimer for Our Work

The information contained in this book has been obtained solely from personal
experience. The information provided may not be correct, complete, or accurate.
The authors assume no responsibility for losses or other related liabilities and make
no claim as to the accuracy of the information contained in this book resulting from
its use in any manner, or as to the infringement of any patent rights which may result.
Likewise, the authors and the publisher do not warrant that the procedures and
software tools described are free of third-party intellectual property rights. Simcen-
ter Amesim is a registered trademark of Siemens Industry Software NV. The tools,
References 125

Blender, ANSYS CFX, ANSYS Chemkin, ANSYS Fluent, openFOAM, and MATLAB,
are also registered trademarks. The reproduction of common names, trade names,
product designations, etc. in this work does not entitle the user to assume that such
names are or would be considered free within the meaning of trademark protection
legislation, even without special identification. Further information can be found on
the Siemens PLM Software homepage [3].

References

1. https://www.ua.pt/en/news/9/69847
2. Russell, C., Shepherd, J.: Online role-play environments for higher education. Br. J. Educ.
Technol., Special Issue: Learning Objects in Progress 41, 6 (2010)
3. https://www.plm.automation.siemens.com/global/en/