NTIS #PB2019-######

SSC-475

GUIDELINES FOR EVALUATION OF

MARINE FINITE ELEMENT
ANALYSES
Editors: Dr. E.D. Wang, Mr. J.S. Bone, Dr. M. Ma,
and Mr. A. Dinovitzer

This document has been approved

for public release and sale; its
distribution is unlimited.

SHIP STRUCTURE COMMITTEE

2019
 SHIP STRUCTURE COMMITTEE
RDML Richard Timme RADM Lorin Selby
U. S. Coast Guard Assistant Commandant Chief Engineer and Deputy Commander
for Prevention Policy (CG-5P) For Naval Systems Engineering (SEA05)
Co-Chair, Ship Structure Committee Co-Chair, Ship Structure Committee

Mr. Jeffrey Lantz Mr. Derek Novak

Director, Commercial Regulations and Standards (CG-5PS) Senior Vice President, Engineering & Technology
U.S. Coast Guard American Bureau of Shipping

Mr. H. Paul Cojeen Dr. John MacKay

Society of Naval Architects and Marine Engineers Head, Warship Performance, DGSTCO
Defence Research & Development Canada - Atlantic

Mr. Kevin Kohlmann Mr. Luc Tremblay

Director, Office of Safety Executive Director, Domestic Vessel Regulatory Oversight
Maritime Administration and Boating Safety, Transport Canada

Mr. Albert Curry Mr. Eric Duncan

Deputy Assistant Commandant for Engineering and Group Director, Ship Integrity and Performance Engineering
Logistics (CG-4D) (SEA 05P)
U.S. Coast Guard Naval Sea Systems Command

Mr. Neil Lichtenstein Dr. Thomas Fu

Deputy Director N7x, Engineering Directorate Director, Ship Systems and Engineering Research Division
Military Sealift Command Office of Naval Research

SHIP STRUCTURE SUB-COMMITTEE

UNITED STATES COAST GUARD (CVE) AMERICAN BUREAU OF SHIPPING
CAPT Robert Compher Mr. Daniel LaMere
Mr. Jaideep Sirkar Ms. Christina Wang
Mr. Charles Rawson

SOCIETY OF NAVAL ARCHITECTS AND MARINE DEFENCE RESEARCH & DEVELOPMENT CANADA
ENGINEERS ATLANTIC
Mr. Frederick Ashcroft Dr. Malcolm Smith
Dr. Roger Basu Mr. Cameron Munro
Dr. Robert Sielski
Dr. Paul Miller

MARITIME ADMINISTRATION TRANSPORT CANADA

Mr. Todd Hiller Ms. Veronique Bérubé
Mr. Todd Ripley Mr. Bashir Ahmed Golam
Ms. Tayyebe Seif

UNITED STATES COAST GUARD (FLEET) NAVSEA/ NSWCCD

CAPT George Lesher Mr. David Qualley
Mr. Martin Hecker Mr. Dean Schleicher
Mr. Timothy McAllister Dr. Pradeep Sensharma
Mr. Debu Ghosh Mr. Daniel Bruchman

MILITARY SEALIFT COMMAND OFFICE OF NAVAL RESEARCH

Ms. Jeannette Viernes Dr. Paul Hess
 PROJECT TECHNICAL COMMITTEE
The Ship Structure Committee greatly appreciates the contributions of the individuals that volunteered
their time to participate on the Project Technical Committee, listed below, and thanks them heartily. They
were the subject matter expert representatives of the Ship Structure Committee to the contractor,
performing technical oversight during contracting, advising the contractor in cognizant matters pertaining
to the contract of which the agencies were aware, and performing technical peer review of the work in
progress and upon completion.

Chair:

Dr. Paul Miller, US Coast Guard Academy

Members:

Dr. Roger Basu, Roger Basu & Associates Inc.

Mr. Jason Cordell, US Coast Guard
Mr. Paul Lara, US Navy
Dr. Miguel Núñez-Sánchez, Universidad Politécnica de Madrid, Spain
Mr. Earl Powers, US Coast Guard
Mr. Charles Rawson, US Coast Guard
Dr. Mahmud Sazidy, Defence Research and Development Canada
Dr. Pradeep Sensharma, US Naval Sea Systems Command
Dr. Robert Sielski
LT Braden Rostad, US Coast Guard

Ship Structure Committee Executive Director:

LT Braden Rostad, US Coast Guard

1. Report No. 2. Government Accession No. 3. Recipient's Catalog No.
NTIS #PB2019-######
SSC-475
4. Title and Subtitle 5. Report Date
GUIDELINES FOR EVALUATION OF MARINE FINITE December 2019
ELEMENT ANALYSES 6. Performing Organization Code

7. Author(s) 8. Performing Organization Report No.

SR-1481
Dr. E.D. Wang, Mr. J.S. Bone, .Dr. M. Ma, and Mr. A. Dinovitzer
9. Performing Organization Name and Address 10. Work Unit No. (TRAIS)
BMT Designes and Planners Inc.MIL Systems
2900 South Quincy Street, Suite 210 11. Contract or Grant No.
Arlington, VA, 22206, United States

12. Sponsoring Agency Name and Address 13. Type of Report and Period Covered
Ship Structure Committee Final
COMMANDANT (CG-ENG-2/SSC)
ATTN (ADMIN ASST/SHIP STRUCTURE COMMITTEE) 14. Sponsoring Agency Code
CG-5P
US COAST GUARD
2703 MARTIN LUTHER KING JR. AVE SE MAILSTOP 7509
WASHINGTON DC 20593-7509
15. Supplementary Notes
Sponsored by the Ship Structure Committee. The research completed by the above author for the Ship Structure
Committee was reviewed by the Project Technical Committee for satisfactory completion of the objectives outlined
in the Statement of Work developed and approved for funding by the Principal Members of the Ship Structure
Committee.
16. Abstract
Commercial and open source finite element analysis (FEA) programs can easily be used to model structures
and generate impressive looking results even when fundamental mistakes are introduced by engineers with
little previous design experience or with improper modeling techniques. This can result in inadequate
structures from the point of view of strength, fatigue, vibration, and other design or analysis criteria. Some
structural failures have demonstrated that, if not appropriately used, FEA may mislead the designer with
erroneous results. The original SSC-387 Guideline for Evaluation of Finite Elements and Results published in
1996 addressed this concern. The use of finite element analysis (FEA) techniques in ship design and analysis
has grown since the original SSC-387 Guideline for Evaluation of Finite Elements and Results was
published in 1996. This guide is an update to SSC-387 and includes current best practices for FEA
application to ship structures and advanced analysis topic discussion and sample applications for the following:
Impact and Plasticity, Fracture and Fatigue, Whole Ship Analysis, and Frequency Response Vibration Analysis.
This document structure follows the original document structure. This document provides, in checklists and
discussions, support for the review of FEA models and output to ensure that the analysis is prepared
appropriately for the intended situation. The document is no substitute for a solid education, enhanced by the
experience of the impact of modeling choices on results. The document is to be construed as a guideline to
assist the analyst and reviewer in determining deficiencies or identifying good practice in an FEA; it is not a
substitute for technical qualifications.

17. Key Words 18. Distribution Statement

Distribution Unlimited
Finite Element Method, Finite Element Analysis, FEA, Available from:
Ship Structure, Structural Analysis, Quality
National Technical Information Service
Assessment, Impact Analysis, Collision Analysis, Springfield, VA 22161
Fracture Analysis, Damage Analysis, Fatigue Analysis,
Whole Ship Analysis, Vibration Analysis
19. Security Classlf. (of this report) 20. Security Classification (of this page) 21. No. of Pages 22. Price
Unclassified Unclassified 404
Contents

PROJECT OVERVIEW ................................................................................................... 1

1/1 EXECUTIVE SUMMARY .......................................................................................... 1

1/2 INTRODUCTION ...................................................................................................... 1

1/2.1 Background .......................................................................................................... 1

1/2.2 Scope ................................................................................................................... 2
1/2.3 Overview of Report .............................................................................................. 2
1/2.4 About the Guidelines ........................................................................................... 4
1/2.5 Using the Guidelines ............................................................................................ 4
1/2.6 The Guidelines as Quality Procedures ................................................................ 4
1/2.7 Where to Get Further Information ........................................................................ 5

ASSESSMENT METHODOLOGY FOR FINITE ELEMENT ANALYSIS ...................... 2-1

GUIDELINES FOR ASSESSING FINITE ELEMENT MODELS AND RESULTS ........... 3-1

3/1 PRELIMINARY CHECKS ...................................................................................... 3-1

3/1.1 Documentation Requirements .......................................................................... 3-1

3/1.2 Job Specification Requirements ....................................................................... 3-4
3/1.3 Finite Element Software Requirements ............................................................ 3-5
3/1.4 Personnel Qualification Requirements ............................................................. 3-8

3/2 ENGINEERING MODEL CHECKS ..................................................................... 3-10

3/2.1 Analysis Type ................................................................................................. 3-10

3/2.2 Analysis Geometry .......................................................................................... 3-16
3/2.3 Material Properties .......................................................................................... 3-25
3/2.4 Loads and Boundary Conditions .................................................................... 3-31
3/2.5 Impact/Collision Analysis ................................................................................ 3-32
3/2.6 Fatigue and Fracture Analysis ...................................................................... 3-33
3/2.7 Whole Ship Analysis ....................................................................................... 3-40
3/2.8 Frequency Response Vibration Analyses ....................................................... 3-53

3/3 FINITE ELEMENT MODEL CHECKS .................................................................. 3-55

3/3.1 CAD Importing ................................................................................................ 3-55

3/3.2 Element Types ................................................................................................ 3-57

i
 3/3.3 Mesh Design ................................................................................................... 3-62
3/3.4 Substructures and Submodeling ..................................................................... 3-84
3/3.5 Loads and Boundary Conditions ..................................................................... 3-90
3/3.6 Analysis Control and Solution Options ........................................................... 3-99

3/4 FINITE ELEMENT RESULTS CHECKS ............................................................ 3-114

3/4.1 General Solution Checks .............................................................................. 3-114

3/4.2 Postprocessing Methods .............................................................................. 3-116
3/4.3 Displacement Results ................................................................................... 3-117
3/4.4 Force Results ................................................................................................ 3-118
3/4.5 Stress Results ............................................................................................... 3-118
3/4.6 Strain Results ............................................................................................... 3-122
3/4.7 Energy Results ............................................................................................. 3-123
3/4.8 Fracture Results ........................................................................................... 3-123
3/4.9 Fatigue Results ............................................................................................. 3-124
3/4.10 Vibration Results ....................................................................................... 3-124

3/5 CONCLUSIONS CHECK ................................................................................... 3-125

3/5.1 FEA Results and Acceptance Criteria .......................................................... 3-125

3/5.2 Load Assessment ......................................................................................... 3-126
3/5.3 Strength/Resistance Assessment ................................................................. 3-126
3/5.4 Accuracy Assessment .................................................................................. 3-126
3/5.5 Overall Assessment ...................................................................................... 3-127

BENCHMARK PROBLEMS FOR ASSESSING FEA SOFTWARE .............................. 4-1

4/1 INTRODUCTION ................................................................................................... 4-1

4/2 THE BENCHMARK PROBLEMS ........................................................................... 4-3

4/2.1 BM-l Reinforced Deck Opening ........................................................................ 4-3

4/2.2 BM-2 Stiffened Panel ........................................................................................ 4-4
4/2.3 BM-3 Vibration isolation System ....................................................................... 4-5
4/2.4 BM-4 Mast Structure ......................................................................................... 4-6
4/2.5 BM-5 Bracket Connection Detail. ...................................................................... 4-7

4/3 THE BENCHMARK TEST FEA PROGRAMS ........................................................ 4-8

4/4 APPLICATION OF BENCHMARKS FOR ASSESSING FEA SOFTWARE ............ 4-9

ii
 ADVANCED ANALYSIS SAMPLE APPLICATIONS .................................................... 5-1

5/1 IMPACT AND PLASTICITY ................................................................................... 5-1

5/1.1 Introduction ....................................................................................................... 5-1

5/1.2 Engineering Model ............................................................................................ 5-1
5/1.3 FE Models and Simulation Setups .................................................................... 5-4
5/1.4 Analysis Control and Solution Options ............................................................. 5-5
5/1.5 Results .............................................................................................................. 5-6
5/1.6 Discussions and Comments ............................................................................. 5-8

5/2 FRACTURE AND FATIGUE ................................................................................ 5-11

5/2.1 Introduction ..................................................................................................... 5-11

5/2.2 Engineering model .......................................................................................... 5-12
5/2.3 FE models and simulation setups ................................................................... 5-16
5/2.4 Analysis control and solution options .............................................................. 5-18
5/2.5 Results ............................................................................................................ 5-19
5/2.6 Discussions and Comments ........................................................................... 5-25

5/3 WHOLE SHIP ANALYSIS.................................................................................... 5-26

5/3.1 Sample Application Description ...................................................................... 5-26

5/3.2 Finite Element Model Development ................................................................ 5-26
5/3.3 FE Model Mass Properties and Hydrostatic Loading ...................................... 5-27
5/3.4 Boundary Conditions ...................................................................................... 5-30
5/3.5 Hull Girder Design Waves .............................................................................. 5-31
5/3.6 Limit State or Failure Criteria .......................................................................... 5-36
5/3.7 Finite Element Analysis and Results .............................................................. 5-37
5/3.8 Sample Showing the Significance of Equilibrium for Whole Ship Analysis ..... 5-39

5/4 FREQUENCY RESPONSE VIBRATION ANALYSIS .......................................... 5-43

5/4.1 Sample Application Description ...................................................................... 5-43

5/4.2 Introduction ..................................................................................................... 5-43
5/4.3 Method ............................................................................................................ 5-43
5/4.4 Given and Assumed Parameters .................................................................... 5-44
5/4.5 Results ............................................................................................................ 5-45
5/4.6 Conclusions .................................................................................................... 5-47
5/4.7 Appendix A: GEOMETRY PLOTS .................................................................. 5-48
iii
 5/4.8 Appendix B: LOADING AND BOUNDARY CONDITION PLOTS ................... 5-53
5/4.9 Appendix C: DISPLACEMENT PLOTS .......................................................... 5-56
5/4.10 Appendix D: VELOCITY PLOTS ................................................................. 5-60

CONCLUSIONS AND RECOMMENDATIONS ............................................................ 6-1

APPENDIX A EVALUATION FORMS FOR ASSESSMENT OF MODELS AND RESULTS ........ 2

A/1.1 Documentation Requirements ............................................................................. 4

A/1.2 Job Specification Requirements .......................................................................... 6
A/1.3 Finite Element Analysis Software Requirements ................................................. 7
A/1.4 Personnel Qualification Requirements ................................................................ 9

A/2 ENGINEERING MODEL CHECKS ................................................................................. 10

A/2.1 Analysis Type .................................................................................................... 10

A/2.2 Analysis Geometry ............................................................................................. 12
A/2.3 Material Properties ............................................................................................. 16
A/2.4 Loads and Boundary Conditions ........................................................................ 18
A/2.5 Impact and Plasticity .......................................................................................... 20
A/2.6 Fatigue and Fracture Analysis ........................................................................... 21
A/2.7 Whole Ship ........................................................................................................ 23
A/2.8 Frequency Response Vibration Analysis ........................................................... 24

A/3 FINITE ELEMENT MODEL CHECKS ............................................................................. 26

A/3.1 CAD Importing ................................................................................................... 26

A/3.2 Element Types .................................................................................................. 27
A/3.3 Mesh Design ...................................................................................................... 32
A/3.4 Substructures and Submodeling ........................................................................ 34
A/3.5 Loads and Boundary Conditions........................................................................ 36
A/3.6 Analysis Controls and Solution Options ............................................................. 37

A/4 FINITE ELEMENT RESULTS CHECKS .......................................................................... 38

A/4.1 General Solution Checks ................................................................................... 38

A/4.2 Post Processing Methods ................................................................................ 39
A/4.3 Displacement Results ........................................................................................ 40
A/4.4 Force Results ..................................................................................................... 41
A/4.5 Stress Results .................................................................................................... 42

iv
 A/4.6 Strain Results .................................................................................................... 43
A/4.7 Energy Results .................................................................................................. 45
A/4.8 Fracture Results ................................................................................................ 46
A/4.9 Fatigue Results .................................................................................................. 47
A/4.10 Vibration Results ............................................................................................ 48

A/5 CONCLUSION CHECKS .............................................................................................. 49

A/5.1 FEA Results and Acceptance Criteria ............................................................... 49

A/5.2 Load Assessment .............................................................................................. 50
A/5.3 Strength / Resistance Assessment .................................................................... 51
A/5.4 Accuracy Assessment ....................................................................................... 52
A/5.5 Overall Assessment .......................................................................................... 53

APPENDIX B EXAMPLE APPLICATION OF ASSESSMENT METHODOLOGY ......................... 1

ANNEX B-1 FINITE ELEMENT ANALYSIS OF ARCTIC TANKER WEB FRAME .......................................... 4

ANNEX B-2 COMPANY AND PERSONNEL QUALIFICATIONS .............................................................. 41

ANNEX B-3 FEA RESULTS VERIFICATION...................................................................................... 42

APPENDIX C EXAMPLES OF VARIATIONS IN FEA MODELING PRACTICES AND RESULTS

...................................................................................................................................................... 1

C1.0 STIFFENED PANEL .................................................................................................................. 3

C2.O MULTIPLE DECK OPENINGS.................................................................................................. 19

C3.0 MAST 26

APPENDIX D SHIP STRUCTURE BENCHMARK PROBLEMS FOR ASSESSING FEA

SOFTWARE .................................................................................................................................. 1

REFERENCES ............................................................................................................................. 1

v
 LIST OF ILLUSTRATIONS
Figure 3/2-1 Analysis Type Flow Chart .................................................................................... 3-12
Figure 3/2-2 An example of topology optimization [10] ............................................................ 3-16
Figure 3/2-3 Examples of Simple Models that can Indicate Extent of Structure to be Modelled .. 3-
19
Figure 3/2-4 Large Changes in Stiffness to Indicate the Extent of Model ................................ 3-20
Figure 3/2-5 Illustration of Saint-Venant's Principle ................................................................. 3-21
Figure 3/2-6 Mass Distribution Required for Accurate Determination of Natural Frequencies 3-23
Figure 3/2-7 Selection of Dynamic DOFs .............................................................................. 3-24
Figure 3/2-8 Modeling Rotational Inertia .................................................................................. 3-25
Figure 3/2-9 Comparison of engineering and true stress and strain curves. Reproduced from
Fig.8.4 in Ref. [24] .................................................................................................................... 3-27
Figure 3/2-10 Discrete Stress-Strain Curves ........................................................................... 3-28
Figure 3/2-11 Stress-Strain Comparison of Non-Calibrated FEA Results with Experimental
Results ..................................................................................................................................... 3-30
Figure 3/2-12 Neuber Correction from the Elastic Stress State to the Elasto-Plastic Stress State
................................................................................................................................................. 3-31
Figure 3/2-13 Fatigue Analysis Procedures ............................................................................. 3-34
Figure 3/2-14 Cyclic Load Parameters .................................................................................... 3-35
Figure 3/2-15 Crack Growth Rate vs. Stress Intensity Range ................................................. 3-36
Figure 3/2-16 Differentiation of Continuum, Fracture, and Damage Mechanics ...................... 3-37
Figure 3/2-17 Basic Modes of Crack Extension: (a) Mode-I: Opening mode; (b) Mode-II: in-plane
shear mode; (c) Mode-III: out-of-plane shear mode. ............................................................... 3-38
Figure 3/2-18 Size Effect of Plastic Zone in the Crack Tip Stress Field : (a) SSY (b) LSY with an
intermediate amount of plasticity (c) LSY with the pronounced amount of plasticity ............... 3-39
Figure 3/2-19 Modelling of Stiffened Plate ............................................................................... 3-42
Figure 3/2-20 A Nominal Frigate Full Load Stillwater Hydrostatic Pressure Distribution ......... 3-43
Figure 3/2-21 A Nominal Frigate Vertical Bending Moment and Shear Force Distribution ...... 3-44
Figure 3/2-22 Mode 7 eigenmode ............................................................................................ 3-44
Figure 3/2-23 Mode 8 eigenmode ............................................................................................ 3-45
Figure 3/2-24 Typical full-ship boundary constraints ............................................................... 3-45
Figure 3/2-25 Extreme Load Analysis Procedure .................................................................... 3-49
Figure 3/2-26 Typical buckling failure modes of stiffened panel .............................................. 3-51
Figure 3/2-27 Frigate Evaluation Panels .................................................................................. 3-52
Figure 3/2-28 Longitudinal stress distribution comparison between FEA and simple beam theory
................................................................................................................................................. 3-53

vi
Figure 3/3-1 Not Connected Surfaces Reproduced from Ref. [15] .......................................... 3-56
Figure 3/3-2 Elements for Transition Zones [91] ...................................................................... 3-58
Figure 3/3-3 Typical Lattice Structure ...................................................................................... 3-63
Figure 3/3-4 Transitions from Coarse to Fine Meshes ............................................................. 3-65
Figure 3/3-5 Transitions Using Triangular Elements ................................................................ 3-65
Figure 3/3-6 Six Element Transitions ....................................................................................... 3-66
Figure 3/3-7 Mesh Transitions ................................................................................................. 3-66
Figure 3/3-8 Connecting Elements with Different Nodal Degrees of Freedom ........................ 3-67
Figure 3/3-9 Modeling In-Plane Rotational Stiffness of Membrane Elements.......................... 3-67
Figure 3/3-10 Meshes for the Analysis of Blunting and Local Crack Opening Criteria ............ 3-70
Figure 3/3-11 Nodes of Quadrilateral Element Collapse Coincident to Model Crack Tip [47] . 3-70
Figure 3/3-12 Crack Tip Elements ........................................................................................... 3-71
Figure 3/3-13 Crack Tip Opening Displacement (CTOD) At the Intersection of 90° vertex with
Crack Faces ............................................................................................................................. 3-71
Figure 3/3-14 Blunt Crack Face of A Single-Edge Notched Tension (SENT) Specimen with
Spider-Web Meshes [117] ....................................................................................................... 3-72
Figure 3/3-15 J-integral vs. Load Line Displacement for a C(T) Specimen [118] .................... 3-73
Figure 3/3-16 Spring Elements [123] ....................................................................................... 3-74
Figure 3/3-17 Cohesive Element [96] ...................................................................................... 3-74
Figure 3/3-18 Mesh Transition Example from DNVGL-CG-0129 [133] .................................... 3-75
Figure 3/3-19 Two Levels of Meshed Models for Stress Concentration Analysis 50 x 50 mm (b) t
x t refinement [134] .................................................................................................................. 3-76
Figure 3/3-20 Stress Distributions of Two Levels of Meshed Models for Stiffener and Lug Details
(a) 50 x 50 mm (b) t x t refinement [134] .................................................................................. 3-76
Figure 3/3-21 Coarse Mesh Stiffener Lumping ........................................................................ 3-77
Figure 3/3-22 Assessing Accuracy of Higher Modes ............................................................... 3-78
Figure 3/3-23 Aspect Ratio of Plane Elements ........................................................................ 3-79
Figure 3/3-24 Element Shape Limitations ................................................................................ 3-80
Figure 3/3-25 Reviewing Mesh Quality (a) Contour Plot (for example, red color indicates
element fails to meet criteria) (b) Highlighted and Selected Fail Elements for Subsequent Fixing
[15] ........................................................................................................................................... 3-84
Figure 3/3-26 Schematic Illustration of The Static Condensation Process .............................. 3-87
Figure 3/3-27 Two-Stage Analysis ........................................................................................... 3-88
Figure 3/3-28 Use Superelement Technique to Represent a Satellite [15] .............................. 3-90
Figure 3/3-29 Minimum Support Conditions for Models ........................................................... 3-92
Figure 3/3-30 Different Types of Symmetry ............................................................................. 3-93

vii
Figure 3/3-31 Coupled DOF: Nodes 1, 2 and 3 Coupled in the y-Direction and About the y Axis
................................................................................................................................................. 3-94
Figure 3/3-32 Constraint Equation ........................................................................................... 3-95
Figure 3/3-33 Definition of Nodal Force ................................................................................... 3-96
Figure 3/3-34 Examples of How to Applied Nodal Forced ....................................................... 3-96
Figure 3/3-35 Definition of Nodal Temperature ........................................................................ 3-97
Figure 3/3-36 Definition of Face Pressure ............................................................................... 3-97
Figure 3/3-37 Definition of Edge Pressure ............................................................................... 3-98
Figure 3/3-38 Definition of Beam Temperature ........................................................................ 3-98
Figure 3/3-39: Implicit and Explicit Solver Applications Chart ................................................ 3-101
Figure 3/3-40 Implicit and Explicit CPU Requirements .......................................................... 3-101
Figure 3/3-41 Three Feature Points in Load vs. Displacement Curves Reproduced from DNV GL
C208 ...................................................................................................................................... 3-103
Figure 3/3-42 Test of Hourglass Control [28] ......................................................................... 3-107
Figure 3/3-43 Relationship between Simulation Time and Time Step Size With and Without
Mass Scaling Features .......................................................................................................... 3-109
Figure 3/3-44 SN Curve with Different Survival Certainty ...................................................... 3-111
Figure 3/3-45 Mean Stress Correction Methods in Haigh Diagram ....................................... 3-112
Figure 3/3-46 Different SN Curve Approaches ...................................................................... 3-112
Figure 3/3-47 Morrow and SWT Methods for Nonlinear Fatigue Analysis ............................. 3-113
Figure 3/3-48 EN Curve Approach ......................................................................................... 3-113
Figure 3/4-1 Distribution of Element Stresses ....................................................................... 3-121
Figure 3/4-2 Stress Contours in Coarse and Fine Meshes .................................................... 3-122
Figure 3/4-3 Schematic Failure Assessment Diagram (FAD) for Fracture Resistance Analysis . 3-
124
Figure 4/1-1 Summary of Ship Structure FEA Benchmark Problems ........................................ 4-3
Figure 4/2-1 Benchmark Problem BM-1 Reinforced Deck Opening .......................................... 4-4
Figure 4/2-2 Benchmark Problem BM-2: Stiffened Panel .......................................................... 4-5
Figure 4/2-3 Benchmark Problem BM-3: Vibration Isolation System ......................................... 4-6
Figure 4/2-4 Benchmark Problem BM-4 : Mast Structure .......................................................... 4-7
Figure 4/2-5 Benchmark Problem BM-5 : Bracket Detail ........................................................... 4-8
Figure 5/1-1 FE Models of Ship Bow Under Frontal Collision .................................................... 5-1
Figure 5/1-2 Schematic Stress and Strain Relationship With Yield Plateau and Strain Hardening
Effects ........................................................................................................................................ 5-3
Figure 5/1-3 True Stress and True Strain Curves of S235 and S355 Grade Steel for LS-DYNA
Input ........................................................................................................................................... 5-3

viii
Figure 5/1-4 Schematic Illustration of Boundary Conditions and Applied Load ......................... 5-4
Figure 5/1-5 Results of Stress and Strain at the Indentation of 2.0 m ....................................... 5-6
Figure 5/1-6 Force-Indentation Curves of the Bow Structures against the Design Curve from
DNVGL-RP-C204 ....................................................................................................................... 5-7
Figure 5/1-7 Flattening of the Bulb Tip under Frontal Impact of Rigid Wall ............................... 5-7
Figure 5/1-8 Local Failure Mechanism of the Bulb Tip: Folding and Buckling ........................... 5-8
Figure 5/1-9 (a) Testing Results of Failure Mode; (b) Numerical Results of Failure Modes and
von Mises Stress Distributions for Four Different Mesh Sizes (i.e., Length/Thickness = 1, 2, 5,
and 10) ..................................................................................................................................... 5-11
Figure 5/2-1 Cracking Occurring on a Horizontal Plane at Floor Stiffener to Longitudinal
Connection ............................................................................................................................... 5-12
Figure 5/2-2 Locations of Cracking Incidents in the Double Bottom ........................................ 5-12
Figure 5/2-3 FlawCheck GUI® and Fatigue Analysis Module .................................................. 5-13
Figure 5/2-4 FE Model of a Typical Hold of a Vessel ............................................................... 5-14
Figure 5/2-5 Local FE Model with Finer Meshing for Capturing Variations of Stress Gradients
Close to Hot Spots ................................................................................................................... 5-14
Figure 5/2-6 SN Curve of Ballast Steel for Corrosive Environment with Slope = 3 and Intercept
Point = 11.705 .......................................................................................................................... 5-15
Figure 5/2-7 Illustrations of the pressure loads of (a) Ballast and (b) Cargo loaded cases ..... 5-16
Figure 5/2-8 Loading and Fixation Conditions of Three Load Cases ....................................... 5-16
Figure 5/2-9 Hot Spot Stress Extrapolation in FE Models when Weld Toe (Blue-Color Shading)
is not Modeled, Since the Use of Shell Elements .................................................................... 5-18
Figure 5/2-10 Possible Four Load Paths to Two Hot Spots at the Lower Connection ............. 5-19
Figure 5/2-11 Displacement Results of Global Model Due to Longitudinal Bending (Load Case –
1) .............................................................................................................................................. 5-19
Figure 5/2-12 Maximum Principal Stress Due to Longitudinal Bending (Load Case – 1) ........ 5-20
Figure 5/2-13 Maximum Principal Stress of Local Sub Model Due to Ballast Load Conditions
(Load Case – 2) ....................................................................................................................... 5-20
Figure 5/2-14 Maximum Principal Stress of Local Sub Model Due to Cargo Load Conditions
(Load Case – 3) ....................................................................................................................... 5-20
Figure 5/2-15 The stress values of hot spots for four load paths from FEA and corresponding
stress ranges at upper and lower connections ........................................................................ 5-22
Figure 5/2-16 Defining SN Transfer Functions for Load Case – 1 (bending) in FlawCheck® . 5-23
Figure 5/2-17 Setup of Loading Spectrum in FlawCheck® ...................................................... 5-24
Figure 5/2-18 Fatigue Life Under the Moment Loading, where the Accumulated Damage Plots
against Time ............................................................................................................................ 5-24
Figure 5/2-19 Direction of Crack Growth along the Load Path C at the Upper Connection ..... 5-25
Figure 5/2-20 Fatigue Life Estimation Using the Fracture-Mechanics-Based Fatigue Analysis

ix
Method, Where the Flaw Length Plots Against Time ............................................................... 5-26
Figure 5/3-1 A Notional Frigate ................................................................................................ 5-27
Figure 5/3-2 A Notional Frigate Meshing Approach ................................................................. 5-27
Figure 5/3-3 Liquid Tank Definition of the Notional Frigate ...................................................... 5-28
Figure 5/3-4 Frigate “Wettable” Elements ................................................................................ 5-28
Figure 5/3-5 Frigate Full Load Tank Definition ......................................................................... 5-29
Figure 5/3-6 Frigate Lightship Weight Distribution ................................................................... 5-30
Figure 5/3-7 Typical Full-ship Boundary Constraints ............................................................... 5-31
Figure 5/3-8 Quasi-static Pressure Under Sagging and Hogging Waves ................................ 5-31
Figure 5/3-9 Frigate Model Evaluation Panels ......................................................................... 5-32
Figure 5/3-10 Example of Generating an Extreme Vertical Bending Moment Design Wave ... 5-33
Figure 5/3-11 Hydrodynamic Pressure and Vertical Bending Moment of the Linear Design
Waves ...................................................................................................................................... 5-33
Figure 5/3-12 Example of Setting Up a Weakly-Nonlinear Time Domain Simulation .............. 5-34
Figure 5/3-13 Example of Vertical Bending Moment Time History at Mid-ship ........................ 5-34
Figure 5/3-14 Maximum Vertical Bending Moments in this Time Domain Simulation.............. 5-35
Figure 5/3-15 Pressure Distribution when the Maximum Vertical Bending Moments Occur .... 5-35
Figure 5/3-16 Frigate Evaluation Panels .................................................................................. 5-36
Figure 5/3-17 ABS HSNC Buckling and Ultimate Strength Limit States .................................. 5-37
Figure 5/3-18 Allowable Stress Under Time Domain Sagging Condition ................................ 5-38
Figure 5/3-19 All Limit States Under Time Domain Sagging Condition ................................... 5-38
Figure 5/3-20 Weight Distribution Comparison ........................................................................ 5-39
Figure 5/3-21 Buoyancy Distribution Comparison ................................................................... 5-39
Figure 5/3-22 Bending Moment and Shear Force Distribution before Applying Restraining
Forces ...................................................................................................................................... 5-40
Figure 5/3-23 Bending Moment Distribution after Applying Restraining Forces ...................... 5-40
Figure 5/3-24 Deflection and Stress Distribution ..................................................................... 5-40
Figure 5/3-25 Buoyancy Distribution Comparison with MAESTRO Auto-Balancing ................ 5-41
Figure 5/3-26 Bending Moment and Shear Force Distribution after Hydrostatic Balance ....... 5-41
Figure 5/3-27 Stress Distribution after Hydrostatic Balance .................................................... 5-41
Figure 5/3-28 Bending Moment and Shear Force Distribution after Inertia Relief ................... 5-42
Figure 5/3-29 Stress Distribution after Inertia Relief ................................................................ 5-42
Figure 5/3-30 Comparison of Bending Moment Distribution .................................................... 5-42
Figure 5/4-1 Bulkhead to Bulkhead Model Image 1 ................................................................. 5-49
Figure 5/4-2 Bulkhead to Bulkhead Model Image 2 ................................................................. 5-49

x
Figure 5/4-3 Bulkhead to Bulkhead Model Image 3 ................................................................. 5-50
Figure 5/4-4 Bulkhead to Bulkhead Model Image 4 ................................................................. 5-50
Figure 5/4-5 Tank Top Model Image 1 ..................................................................................... 5-51
Figure 5/4-6 Tank Top Model Image 2 ..................................................................................... 5-51
Figure 5/4-7 Tank Top Model Image 3 ..................................................................................... 5-52
Figure 5/4-8 Tank Top Model Image 4 ..................................................................................... 5-52
Figure 5/4-9 Main Engine Forcing Loading Condition Image 1 ................................................ 5-54
Figure 5/4-10 Main Engine Forcing Loading Condition Image 2 .............................................. 5-54
Figure 5/4-11 Gear Box Forcing Loading Condition Image ..................................................... 5-55
Figure 5/4-12 Boundary Conditions Image .............................................................................. 5-55
Figure 5/4-13 Main Engine Maximum Displacement at 15 Hz ................................................. 5-57
Figure 5/4-14 Main Engine Maximum Displacement at 20 Hz ................................................. 5-57
Figure 5/4-15 Gear Box Maximum Displacement at 15 Hz ...................................................... 5-58
Figure 5/4-16 Gear Box Maximum Displacement at 20 Hz ...................................................... 5-58
Figure 5/4-17 Gear Box Maximum Displacement at 25 Hz ...................................................... 5-59
Figure 5/4-18 Main Engine Maximum Velocity at 15 Hz .......................................................... 5-61
Figure 5/4-19 Main Engine Maximum Velocity at 20 Hz .......................................................... 5-61
Figure 5/4-20 Gear Box Maximum Velocity at 15 Hz ............................................................... 5-62
Figure 5/4-21 Gear Box Maximum Velocity at 20 Hz ............................................................... 5-62
Figure 5/4-22 Gear Box Maximum Velocity at 25 Hz ............................................................... 5-63
Figure A/0-1 Overall Evaluation Methodology Checklist ............................................................... 3

xi
 LIST OF TABLES
Table 3/1-1 Minimum Recommended Experience Levels (adapted from NAFEMS, 1990) . 3-9
Table 3/2-1 FEA Analysis Types .............................................................................................. 3-10
Table 3/2-2 Example Metallic Material Properties ................................................................... 3-25
Table 3/2-3 Typical Fracture Mechanics Parameters .............................................................. 3-40
Table 3/3-1 General Problems with Average Element Size of 50mm ...................................... 3-81
Table 3/3-2 Vibration Problems with Average Element Size of 50mm .................................... 3-82
Table 3/3-3 Fatigue Problems with Average Element Size of 50mm ....................................... 3-82
Table 3/3-4 Plasticity and Crash Problems with Average Element Size of 50mm ................... 3-83
Table 3/3-5 Plasticity and Crash Problems with Average Element Size of 100 mm ................ 3-83
Table 3/4-1 Stresses Represented by Element Type ............................................................ 3-120
Table 5/1-1 Crash problem with 50mm element size ................................................................. 5-5
Table 5/2-1 Fatigue Problem with 200mm Element Size ......................................................... 5-17
Table 5/2-2 Fatigue Problem with 25mm Element Size ........................................................... 5-17
Table 5/2-3 Loading Spectrum Under Longitudinal Bending Moments in a Year .................... 5-23
Table 5/3-1 Design Bending Moments ..................................................................................... 5-35
Table 5/4-1 ABS Machinery Vibration Limits [14] .................................................................... 5-45
Table 5/4-2 Bulkhead to Bulkhead Modal Frequencies ........................................................... 5-45
Table 5/4-3 Bulkhead to Bulkhead Model Vibration Results: Port and Starboard Power Trains in
Sync ......................................................................................................................................... 5-46
Table 5/4-4 Bulkhead to Bulkhead Model Vibration Results: Port Power Train Only............... 5-47
Table 5/4-5 Tank Top Model Vibration Results ........................................................................ 5-47

xii
 ACKNOWLEDGMENTS
The authors gratefully acknowledge the discussions with Dr. Martin Storheim at Entail AS and his
help in providing FE models for the impact analysis. The authors also wish to thank Canarctic
Shipping Limited, and in particular, Mr. John McCallum, for permission to use the Arctic tanker
example presented in Appendix B.
The authors also wish to thank Altair Software for providing free trials to aid in the preparation of
this updated guide.

LIMITATIONS
The following document is intended to support the development and review of finite element
modeling through a suggested review methodology. The methodology presented has been
developed based on the experience and expertise of the authors. The use of this document does
not replace education or experience and should be treated as good practice guidance but not a
guarantee of acceptable performance for all design or analysis scenarios.

xiii
 PROJECT OVERVIEW
1/1 EXECUTIVE SUMMARY
Commercial and open source finite element analysis (FEA) programs can easily be used to model
structures and generate impressive looking results even when fundamental mistakes are introduced
by engineers with little previous design experience or with improper modeling techniques. This can
result in inadequate structures from the point of view of strength, fatigue, vibration, and other design
or analysis criteria. Some structural failures have demonstrated that, if not used properly, FEA may
mislead the designer with erroneous results. The original SSC-387 Guideline for Evaluation of Finite
Elements and Results published in 1996 addressed this concern. The use of finite element analysis
(FEA) techniques in ship design and analysis has grown since the original SSC-387 Guideline for
Evaluation of Finite Elements and Results was published in 1996. This guide is an update to
SSC-387 and includes current best practices for FEA application to ship structures and advanced
analysis topic discussion and sample applications for the following:
 Impact and Plasticity
 Fracture and Fatigue
 Whole Ship Analysis
 Frequency Response Vibration Analysis
This document structure follows the original document structure. This document provides, in
checklists and discussions, support for the review of FEA models and output to ensure that the
analysis is prepared appropriately for the intended situation. The document is no substitute for a
solid education, enhanced by the experience of the impact of modeling choices on results. The
document is to be construed as a guideline to assist the analyst and reviewer in determining
deficiencies or identifying good practice in an FEA; it is not a substitute for technical qualifications.
1/2 INTRODUCTION
1/2.1 Background
Finite element analysis (FEA) is a common structural analysis method for advanced problems. Great
strides have been made in theoretical and computational aspects of FEA. This has been
accompanied by phenomenal advances in computer technology, both in hardware and software,
together with a rapid reduction in the cost of this technology. A consequence of this is a dramatic
increase in the affordability of, and accessibility to, finite element technology. In marine industries,
the use of this technique is widespread in ship structure design, reliability analysis, and performance
evaluation.
Finite element analysis is a powerful and flexible engineering analysis tool that allows the analyst
considerable freedom in designing the finite element model, exercising it, and interpreting the
results. Key components of this process include the selection of the computer program (e.g., OS –
operating system and FEA software package), the determination of the loads and boundary
conditions, development of the mathematical model, choice of elements, and the design of the
mesh. The analyst needs to make numerous experience-, scenario- and modeling tool-based
decisions during this process. Results from FEAs for the same structure performed by different
individuals or organizations may differ significantly as a result of differences in the assumptions and
modeling procedures employed.
An unsatisfactory analysis is not always obvious, and the consequences may not manifest
themselves until the vessel is in service. Design changes and any structural modifications required
at this stage are generally much more expensive to implement than would be the case if the
deficiency was discovered earlier.

1-1
A particular difficulty is faced by those who have the responsibility for assessing and approving
FEAs. The individual concerned may not be an expert in FEA, or familiar with the software package
used, and would face a dilemma when coming to judge the acceptability, or otherwise, of the results
of the FEA. This may require the evaluator to incur further cost and time in the attempt to ensure
satisfactory FEA results.
In response to the difficulty faced by those who evaluate FEAs, a systematic and practical
methodology was required to support the assessment of the validity of the FEA results efficiently.
That need resulted in the original SSC-387 Guideline for Evaluation of Finite Elements and Results.
SSC-387 included in support of this methodology a selection of finite element models that illustrated
good modeling practice. Besides, SSC-387 included benchmark tests to allow the validation of new
FEA software packages or packages that have undergone significant modification.
Since SSC-387 was issued in 1996, the use of finite element analysis has advanced in terms of:
 Tools available (e.g., automated meshing, interaction with drafting/solid modeling tools)
 Materials considered (e.g., steel, aluminum, plastic, composites, non-linear (post-yielding)
behavior)
 Load conditions (e.g., fluid-structure interaction, collision, blast simulation)
 Analysis types (e.g., implicit versus explicit (time-domain) modeling)
 Element formulation (non-linear, hybrid, and contact elements)
 Structural geometries (crack-tip elements, connection, and weldments; contact/sliding
component fit-up)
 Larger models being able to be analyzed more rapidly than in the past.
1/2.2 Scope
This guideline provides a systematic and practical methodology to support the rapid assessment of
the validity of FEA results for ship structures. The methodology is an updated version of SSC-387,
which includes current FEA best practice, and can be used for the following types of ship structure
analysis problems:
 Linear static
 Non-linear (plastic) static
 Impact/Collision
 Fracture
 Fatigue
 Whole ship
 Natural frequency
 Forced vibration
The emphasis is on the structural assembly level, rather than on local details, such as weldments.
Only FEA of structures composed of isotropic materials are addressed, thus excluding fiber
reinforced plastics and wood. Despite these limitations, the guidelines apply to most ship structure
FEAs.
1/2.3 Overview of Report
The report is structured in six parts, four appendices, and a reference section as follows:

1-2
Part 1: Project Overview
This part introduces the document and provides the background for the methodologies developed for
assessing FEAs and FEA software, which are described in subsequent Parts.
Part 2: Assessment Methodology for Finite Element Analysis
This part presents a systematic methodology for assessing FEAs. Appendix A contains forms that
can be used for the evaluation process. Appendix B presents an example of an FEA and its
evaluation.
Part 3: Guidelines for Assessing Finite Element Models and Results
This part provides guidance in support of the methodology presented in PART 2. It is a
comprehensive description of good FEA practices. As an aid to the assessment of FEA models and
results, some FEAs, typical of ship structures, are presented in Appendix C. These examples are
designed to illustrate the influence on the results of varying specific model parameters.
Part 4: Benchmark Problems for Assessing FEA Software
The assessment methodology described in PART 2 includes a requirement that suitable FEA
software is used. In support of the assessment, new or significantly-modified FEA should be
evaluated regarding its suitability for ship structure FEA. The benchmark problems and results
presented in PART 4 are for this purpose. The benchmark problems are presented in Appendix D.
Part 5: Advanced Analysis Sample Applications
This part presents sample applications for the following analysis types:
 Impact and Plasticity
 Fracture and Fatigue
 Whole Ship
 Frequency Response Vibration
Part 6: Conclusions and Recommendations
This part summarizes observations and insights gained in the course of this project regarding the
process of evaluating finite element models and results and FEA software. Also presented is a
summary of where effort should be directed to further improve the methodologies in response to
likely future trends in finite element technology.
Appendices
The following appendices are included:
 Appendix A Evaluation Forms for Assessment of Finite Element Models and Results
 Appendix B Example Application of Assessment Methodology
 Appendix C Examples of Variations in FEA Modeling Practices and Results
 Appendix D Ship Structure Benchmarks for Assessing FEA Software
References
The literature considered in preparing this guide, and that may be of interest to readers, is
summarized in this section. The literature list includes books, reports, scientific papers, tutorials, and
help documents used in this best practice. The list of references is also an excellent resource to help
readers find more technical descriptions and solutions in the use of advanced FEA.

1-3
1/2.4 About the Guidelines
The purpose of the guidelines presented in this document is to provide support for evaluating finite
element models and results, and also FEA software.
There are many attributes to any FEA, and it is difficult to assess quality unless the FEA has been
comprehensively documented, and a systematic assessment methodology is applied. This volume
presents such a methodology.
The methodology is presented in three levels:
1. Level 1 comprises a checklist of attributes of the FEA that need to be evaluated as part of
the assessment process.
2. Level 2 comprises a more detailed breakdown of the checklist provided under Level 1. Level
1 can be regarded as a summary of the Level 2 assessment.
3. Level 3 contains guidelines on acceptable finite element modeling practices. The guidelines
are cross-referenced with the Level 2 checklists. During the assessment process, the
evaluator may, if required, refer to Level 3 guidelines for advice.
For simple FEAs, an experienced evaluator can probably perform the assessment without referring
to Level 2 checklists. The methodology is structured to allow the evaluator to apply the methodology
at the appropriate level of detail. The reader is referred to Figure A/0-1 in Appendix A for summary of
the methodology.
In addition to presenting an assessment methodology and supporting material, this report presents
benchmark problems for assessing the quality of the FEA software and its suitability for ship
structural analysis.
1/2.5 Using the Guidelines
The primary audience for these guidelines is the evaluators of FEAs. The guidelines assume that the
evaluator is trained in ship structural analysis and design, but is not necessarily expert in FEA.
Ideally, the guidelines would be provided to the analysts as part of the job specifications (such as a
statement of work or statement of requirements). Level 1 and 2 guidelines could then be viewed as
acceptance criteria for the work. The documentation requirements listed in the guidelines could then
be used to stipulate the documentation required.
The methodology can be used for conducting reviews that could then be used to provide
intermediate and final approvals. For this purpose, each of the five areas of an FEA shown in Figure
A/0-1 would be treated as a phase in the project. Reviews could be held at the end of each phase, or
less frequently for smaller projects. Depending on the outcome of the review, approval to proceed to
the next stage could be given, or, in the case of severe deficiencies, rework would be required.
Most FEAs are iterative in character. This applies particularly to analyses performed in support of
design tasks. The iterative nature also applies to certain aspects of the analysis itself. Some
modeling decisions can only be validated during the evaluation of the results. To facilitate this, this
document is presented as a step-by-step QA process, and therefore, can accommodate iterations
where necessary.
The document may be considered as guidance and a memory aid supporting the review process.
With this understanding, the analyst may use the guide to self-check or plan their work, thus
developing or documenting their design or analysis process.
1/2.6 The Guidelines as Quality Procedures
The guidelines presented in this document incorporate several elements of a quality system as it
pertains to FEA and, as such, could be incorporated in an organization's quality system for FEA.

1-4
The requirements for such a system have been developed under the direction of the National
Agency for Finite Element Methods and Standards (NAFEMS) Quality Assurance Working Group.
These requirements in [1] are intended as a supplement to ISO (International Organization for
Standardization) 9001.
1/2.7 Where to Get Further Information
While the information provided in the guidelines is self-contained, there may be circumstances when
more detailed information is required.
There are many texts that describe FEA and theory. The reader is referred to a comprehensive
bibliography of books and monographs on finite element technology in [2]. Besides these texts, there
are several publications more suited for engineering office use. These include the following
guidelines and application-oriented texts that the reader may wish to consult:
 MORRIS, A, A Practical Gude to Reliable Finite Element Modelling, John Wiley & Sons,
2008 [3].
 MACDONALD, B.J., Practical Stress Analysis with Finite Elements (2nd edition), Glasnevin
Publishing, 2011 [4].
 BRAUER, J.R., What Every Engineering Should Know About Finite Element Analysis (2nd
edition), Marcel Dekker, Inc., New York, 1993 [5].
 MEYER, C. (Ed.), Finite Element Idealization for Linear Elastic Static and Dynamic Analysis
of Structures in Engineering Practice, American Society of Civil Engineers, New York, 1987
[6].
 NAFEMS, Guidelines to Finite Element Practice, National Agency for Finite element Methods
and Standards, National Engineering Laboratory, East Kilbride, Glasgow UK, August 1984
[7].
 STEELE, J. E., Applied Finite Element Modeling, Marcel Dekker, Inc., New York, 1989 [8].
 DNV-GL, DNVGL-CG—0127 Finite Element Analysis, 2015 [9].

1-5
 ASSESSMENT METHODOLOGY FOR FINITE ELEMENT ANALYSIS
The methodology developed for evaluating finite element analyses of ship structures is provided in
Appendix A. The evaluation is carried out at two levels conducted in parallel. The highest level
(Level 1) addresses general aspects of the finite element analysis (FEA) broken down into five (5)
main areas:
1. Preliminary Checks
2. Engineering Model Checks
3. Finite Element Model Checks
4. Finite Element Results Checks
5. Conclusions Checks
These five groups are identified in each of the five (5) main boxes shown in Figure A/0-1. The
Preliminary Checks cover background and objectives for the analysis. The Engineering Model
Checks cover inputs and assumptions; the Finite Element Model Checks cover pre-processing; the
Finite Element Results Checks cover post-processing for the analysis. The Conclusion Checks cover
evaluations, recommendations, and reports for the analysis.
Evaluation of each of these five general aspects, in turn, requires that certain related detailed (Level
2) aspects be checked. The Level 2 aspects to be checked are listed within the main boxes and are
presented in detail in separate tables in Appendix A that form the core of the evaluation process.
The Level 2 tables contain many detailed questions regarding specific aspects of the FEA.
The methodology is intended to be used as follows. The evaluator will begin by assembling the
analysis documentation and perhaps computer files of the finite element (FE) model and results. The
evaluation then begins with the Preliminary Checks contained in Figure A/0-1 No 1 Preliminary
Checks. The first of the Preliminary Checks involves assessment of the contents of the analysis
documentation (Section A/1.1). To perform this assessment, the evaluator refers to the table entitled
"A/1.1 Documentation Requirements", which is a systematic list of checks of the documentation for
information that is essential for the FEA evaluation. The table also directs the evaluator to Section
3/1.1 of the guideline should further explanation or guidance be necessary.
 Checkmark (√)
If an item is contained in the documentation, the evaluator should place a checkmark (√) in
the corresponding box under the "Result" column.
 Cross mark (X) or “NA” or “?”
If an item is not included with the documentation, the evaluator may enter a cross mark (X) in
the result box, or “NA” (for Not Applicable), or “?” (for Further Information Required).
After checking off each item in the table, the evaluator is asked to answer Question 1.1 at the bottom
of the page based on the evaluators' assessment of each item listed in the table. The evaluator
should place the answer in the "Result" box to the right of the question, and then if acceptable check
off the corresponding circle in Figure A/0-1. The table in Section A/1.1 also includes spaces for the
evaluator to enter comments regarding specific and overall aspects of the document contents. At the
end of the evaluation process, these comments will provide the evaluator with reminders of specific
aspects of the FEA that were good, bad, or not explained well. The evaluator may refer to these
comments to seek further explanation or clarification from the analyst performing the work before
deciding on the final acceptability of the FEA.
Having completed the first of the preliminary checks, the evaluator then proceeds to the second set
of checks entitled "A/1.2 Job Specification Requirements". In a manner similar to the previous

2-1
checks, the evaluator will refer to the table in Section A/1.2 and perform checks 1.2.1 to 1.2.7, which
are aimed at verifying that the analysis covers the main requirements and objectives of the job
specification (or contract, or statement of work, etc.). Based on the results of these checks, the
evaluator should answer Question 1.2 and then if acceptable check off the corresponding circle in
Figure A/0-1. This procedure is repeated for the other Preliminary Checks (i.e., A/1.3 Finite Element
Analysis Software Requirements, and A/1.4 Personnel Qualification Requirements).
Having answered all of the Level 2 questions for Prelliminary Checks and checked off the
appropriate circles in Figure A/0-1, the evaluator should then move on to Engineering Model Checks.
If the subject analysis does not meet these requirements then the evaluator may choose to terminate
the evaluation.
The evaluation process continues as described above for each of the five main areas identified in
Figure A/0-1. It should be noted, however, that the process is designed to proceed in the order
provided. The evaluator should not move on to the next of the five major sections until the previous
sections are satisfactorily completed.
Ideally, at the start of the job, the contractor would be given the assessment methodology as part of
the job specification in order to encourage self-checking and ensure that the data provided by the
contractor to the customer is complete.
In this document, consideration is given to both in-house FE analysts or contractors. In both cases,
the quality of the FEA is highly dependent on the work processes, training, and tools employed.
The Appendix A forms can be adopted and converted into a spreadsheet format for if desired.

2-2
 GUIDELINES FOR ASSESSING FINITE ELEMENT MODELS AND RESULTS
The guidelines recommended below are structured to match the Assessment Methodology
described in PART 2. Therefore, the guidelines are grouped under the same five sections:
1. Preliminary Checks
2. Engineering Model Checks
3. Finite Element Model Checks
4. Finite Element Results Checks
5. Conclusions Checks
3/1 PRELIMINARY CHECKS
This section describes the checks that need to be undertaken to ensure that the finite element
analysis (FEA) satisfies certain basic requirements. The first requirement before evaluating an FEA
is to ensure that there is sufficient documentation provided with the analysis. This step should
ensure the analysis addresses the objectives, scope, and requirements of the work specification. It is
necessary to establish that the tools the analyst uses in the FEA are adequate and appropriate to
the analysis; this applies particularly to the software used. Finally, the analyst should be
appropriately trained and should have sufficient experience.
3/1.1 Documentation Requirements
Proper documentation is an essential part of any FEA. The documentation submitted should be
sufficient to allow a thorough evaluation of the FEA. The complete documentation package, which
can be defined as that required by an independent party to reproduce the analysis, should be
available and submitted if required by the evaluator. The complete documentation would typically
include information from each of the five major groups being evaluated. The key information needed
to evaluate the FEA analysis is as follows. Note that the items identified with a letter are included in
the Section A/1.1 Documentation Requirements checklist.
Preliminary Checks
Job Specifications
a) Scope and objective of analysis.
 The rationale for using FEA (e.g., types of solving problems and
approaches in setup)
 2D or 3D problems
 Static (or quasi-static) or Dynamic (with inertial effects)
 Approaches in modeling
 List of colors, if requested (for each property, each component, each
material, and each thickness)
Reference documents
b) Reference codes, manuals, and/or standards required.
c) Physical problem documentation references.
 drawings
 CAD models

3-1
  geometric specifications and tolerances (e.g., 2D drawings/ sketches
or 3D models) of the subject structure (with assembly if applicable)
 Identification of Initial imperfections (e.g., geometry imperfections,
thickness reduction, crack profiles for fatigue or fracture analysis)
 gauging reports
FE Software Requirements
d) FEA software used.
 Finite element software, including its version and release date
 Pre- and post-processing software
Personnel Qualifications
e) Personnel qualifications.
Engineering Model Checks
Analysis type and assumptions
f) Analysis type(s).
Geometry Assumptions
g) Description of physical problem.
h) Description of engineering model.
i) Description of the FEA model.
 Contacts (e.g., linear or nonlinear)
 Loading (per load case section)
 Boundary condition (per load case section)
j) Plots of full FEA model and local details.
k) System of units.
l) Coordinate axis systems.
Material Properties
m) Material properties.
 Metallic (e.g., steel, aluminum, and magnesium)
 Non-metallic (e.g., polymer and rubber)
 Composite (e.g., GFRC, CFRC, foam, and piezoelectric)
 Analytical, semi-analytical, or empirical constitutive relationship
 With or without damage model (e.g., element erosion)
 High cycle or low cycle fatigue (e.g., SN or EN)
 Environmental assist corrosion-resistant (e.g., hydrogen diffusion)
 User-defined material models (e.g., UMAT or VUMAT in Abaqus)
Stiffness and Mass Properties
n) Stiffness and mass properties.

3-2
 Loads and Boundary Conditions
o) Loads and boundary conditions.
 Loading history if applicable
Finite Element Model Checks
Element Types
p) Element type(s).
 Element formulations (e.g., 1D, 2D or 3D, full integration or reduced
integration)
 The thickness for shell elements (with or without corrosion deduction)
 User-defined element formulation
Mesh Design
q) Meshing idealizations/assumptions/representations/simplifications.
 General structures (i.e., shell plates, deck plates, bulkhead plates,
stringers, transverse webs)
 Connectors (e.g., spring, beam, and weldment)
 Girders
 Pillars
 Stiffeners
 Openings
r) Meshing criteria for 2D and 3D elements.
 Standard mesh size with maximum and minimum element sizes
 Finer mesh size with maximum and minimum element sizes
 Aspect ratios
 Warpage
 Skew
 Jacobian
 Maximum and minimum angle for quadratic and triangular elements
 Taper
 Percentage of triangular elements
 Mesh sensitivity study (if needed)
 Mesh and CAD overlay review
FE loads and boundary conditions
s) FE loads and boundary conditions.
Solution Options and Procedures
t) Solution options and procedures.
Finite Element Results Checks

3-3
 u) Results.
 Critical state(s) and corresponding location(s)
 Maximum values (e.g., displacement, force, energy, stress, strain, or
strain energy)
 Animations of interested states (e.g., deflections in modal or buckling
analysis)
 Validation (e.g., comparison with analytical solutions)
Conclusion Checks
Results and Acceptance Criteria
v) Comparison of results with acceptance criteria.
Accuracy Assessment
w) Accuracy assessment.
Overall Assessment
x) Conclusions and recommendations for amendments.
The input and output data should be presented in graphical, tabulated, or textual form depending on
what is the most convenient for evaluation purposes.
The documentation requirements listed in A/1.1 Documentation Requirements are the minimum
required. In general, any additional information considered necessary for a complete evaluation
should also be provided.
Plots should be annotated appropriately to show the location of the subject structure in the ship (e.g.,
frame numbers, deck numbers, etc.), axes to orient the model, location of equipment supported by
the structure, and the position of major structural features that define boundaries (e.g., bulkheads).
All symbols used in the plots should be defined either on the plots or in the body of the report.
3/1.2 Job Specification Requirements
The purpose of this check is to ensure that the analysis has been undertaken according to the
requirements of the job specification. This can be done only if the documentation provided
addresses every requirement of the job specification. It is not possible to list all such requirements,
but at least the following items should be addressed:
 Definition of the problem
 Scope and objectives of the analysis
 All relevant documentation such as drawings, sketches, CAD, and reports to completely
define the subject structure and loading
 Any previous analyses, service experience and experimental data related to the subject
structure
 Acceptance criteria (e.g., allowable stress in an analysis in support of a design)
It is expected that the analyst has carefully read the job specifications and followed it as closely as
possible. Deviations from the specifications, if any, should be identified and justified. All reference
documents should be identified.
If the job specification does not explicitly call for an FEA, then the analyst should explain the
rationale for using FEA in preference to another method of structural analysis, or preference to
experiments. It is also expected that the analyst is aware of any previous related studies and their

3-4
outcome.
The selection of FEA as the preferred method of structural analysis will depend on many features of
the engineering problem. Features of the problem that should be discussed include, but are not
limited to, the following:
 purpose of analysis;
 complexity of the structural form;
 redundancy of structural system;
 assessment of expected accuracy;
 accuracy of known input variables such as loads, material properties, etc.; and
 suitability, or otherwise, of hand calculation methods.
3/1.3 Finite Element Software Requirements
There are many finite element software systems available, i.e., commercial and open-source codes,
which include pre-/post-processing software and solvers. Most are intended for general purpose
FEAs (e.g., Ansys, Abaqus, LS-DYNA, and MSC.Nastran/Marc). There are some, however, that are
known to be designed specifically for ship structure applications. Ship structure FEA is, to a certain
extent, specialized in nature, and there are some efficiencies that can be gained from using a ship
structure-specific application. It is not, however, necessary to employ ship structure-specific
software, and a multitude of ship structure structural problems are analyzed using the general-
purpose FEA software.
It is necessary to ensure that the software has been verified and validated. This may be less of a
concern than when this guide was initially published. However, the FEA evaluator should still confirm
that a verified and validated FEA software package is being used. This may be more of a concern for
advanced analysis undertakings, such as fracture or impact analysis.
Commercial finite element analysis systems are large and complex. Developing and maintaining
such systems require systematic methods to be applied to the design and development of the code,
the testing, the verification and validation of the code, and the configuration management of the
software system. Reputable software vendors rely on quality systems to ensure that the relevant
processes that comprise the development and maintenance of the software are appropriately
controlled. The evaluation of FEA software should include an assessment of the vendor's quality
system.
In the past decades, several open-source FEA codes developed by reputable engineering
companies (e.g., Code-Aster by EDF – Électricité de France) are being used and associated with
verification and validation modules.
There are several ways in which FEA software can be validated. The methods for validating FEA
software include:
 independent analysis
in general, compared with analytical solutions or verified data
 experimental results
Note that when it is impossible to compare the numerical output with experimental data. It is
possible to validate the FEA software by validating relatively smaller-scale coupon tests.
Additional attention needs to be considered, such as distributions of flaws or imperfections
(e.g., the weakest link theory).
 service experience

3-5
 Especially for those time-dependent problems, the service experience is extremely valuable
to validate whether the numerical results are converged or reliable. For example, the fatigue
lives of weldment are in the same scales or comparable.
 cross platforms (codes) analysis
Results from well verified and validated FEA codes can be used as the benchmark. For
example, the open-source code WARP3D specialized in macro- and mesoscale plasticity
and fracture mechanics was validated by comparing with numerical results from Abaqus.
Many finite element software vendors publish verification examples. Generally, verification examples
are based on problems with closed-form solutions. The analytical results are compared with those
obtained by exercising the finite element code. While a comprehensive set of satisfactory verification
examples is convincing evidence of high-quality code, it does not constitute proof. Verification
examples based on problems based on closed-form solutions are necessarily simple, and the finite
element models are generally not too demanding on the software. It is necessary, therefore, to
employ additional methods to validate the software.
An additional validation method is to use benchmark problems that, while simple, are more
representative of typical structure. In contrast to the type of verification example mentioned above,
benchmark problems can be designed to use combinations of element types, element shapes that
vary from the ideal, complex boundary conditions, multiple load cases, etc. to test the software.
These problems are more closely related to how the software will be used in practice.
Closed-form solutions are generally not available for benchmark problems. However, results from
other well-established FEA software could be regarded as an example of an independent analysis. If
results from several other FEA software systems are consistent, or where any differences can be
rationalized, then these results can be regarded as benchmarks. Any significant differences between
benchmark results and those obtained from the candidate FEA software system would be an
indication of unsatisfactory performance.
Depending on the size of the organization and the volume of FEA work, it may be useful to maintain
a register of FEA software validated based on satisfactory performance using the methods outlined
above. Similar logs or case-specific validations should be documented, where possible, supporting
advanced analysis modeling scenarios. Alternatively, this function could be performed by a body
representative of the industry, such as a professional society.
In the absence of such an arrangement at present, benchmark problems typical of ship structures
have been formulated, and the results documented in Part 4 of this report. These benchmark
problems could be used to evaluate candidate FEA software. If there is documented evidence
(based on previous applications of the software to ship structural analysis problems) that the
software can perform the required analysis, this requirement may be waived at the discretion of the
evaluator.
Successful performance of the candidate FEA software on the benchmark problems is a necessary,
but not sufficient, condition for approving the software. The software should also satisfy
requirements outlined in the opening paragraphs of this section, particularly regarding requirements
for the vendor's quality system.
In recent years, there is a trend that industries started making parametric studies by working with
FEA codes, aiming to obtain the optimal designs (such as stiffness, weight, strength, fatigue life,
and/or critical stress/strain). The FEA solvers adopted in the parametric studies should also be
validated by the means mentioned above.
To date, there are a lot of FEA codes developed for solving engineering problems. To the author's
knowledge (perhaps not an exhaustive listing) the widely used commercial and open-source FEA
codes for general purpose applications as well as parametric analysis codes are listed as follows:
1. List of selected commercial FEA codes

3-6
 a. Pre- and Post-processing software
i. Altair series (e.g., HyperMesh, HyperView, and HyperCrash)
ii. ANSYS series
iii. BETA ANSA
iv. LSTC LS-Pre/Post
v. MSC series (e.g., Patran and Apex)
vi. Siemens series (e.g., FEMAP)
b. Solvers
i. ADINA
ii. Altair series (e.g., Optistruct, Radioss, and partner alliance packages)
iii. ANSYS series (e.g., Mechanical APDL, Workbench, Fluent, HFSS)
iv. LSTC. LS-DYNA
v. MAESTRO
vi. nCode
vii. MSC series (e.g., Nastran and MARC)
viii. Siemens series (e.g., NX. Nastran)
ix. SIMULIA. ABAQUS
x. Strand7
2. List of selected open-source FEA codes
a. Pre- and Post-processing software
i. Cubit / Trelis
ii. EDF Salome
iii. Gmsh
iv. ParaView
b. Solvers
i. EDF Code_Aster
ii. Impact
iii. Warp3D
3. Parametric analysis codes
a. Altair HyperStudy
b. SIMULIA Tosca
c. VR&D Genesis
Recently, the vendors of commercial FEA codes start adopting the benchmark examples provided
by the not-for-profit companies (e.g., NAFEMS) to demonstrate the accuracy of their codes [10]. As
to the end-user, it is encouraged to independently examine the accuracy of the FE code by verifying
the results of benchmark examples.

3-7
3/1.3.1 Reasons for Using A Particular FEA Software Package
It is recognized that the analyst will prefer to use FEA software packages that are readily available
and that the analyst has experience with. However, the analyst should assess the suitability of the
selected FEA software for the analysis under consideration. The items that should be discussed
include the following:
 availability of required element types (e.g., general types, special types, or user-defined
types)
 availability of required material types (e.g., general types, special types, or user-defined
types)
 availability of required material behavior (e.g., yielding, plasticity, crack, creep, fracture,
and fatigue)
 availability of required load types (e.g., forced-displacement, concentrated force,
distributed force/pressure, gravity, cyclic loading, thermal, electric, fluid and
multidiscipline)
 availability of required boundary condition types (e.g., slide, pin, clamp, and inertial
relieve)
 availability of required contact types (e.g., linear and nonlinear)
 capability of the software to perform the required analysis
 preprocessing and postprocessing capabilities
 support from vendors (e.g., documentation and technical supports)
 Other concerns
o Geometry effects (e.g., initial imperfection, stress stiffening, 2nd order load
effects)
o Temperature effects (e.g., material degradation, thermal expansion)
o Nonlinear load effects (e.g., follower loads)
Note that vendors of FEA software may be able to provide the following [11].:
 recommended hardware specifications
 compatible hardware vendors/ manufacturers
 training resources or programs for analysts
 Technical direction and troubleshooting support
 correct driver versions
3/1.4 Personnel Qualification Requirements
The personnel performing and checking the analysis must meet minimum training and experience
requirements. The following aspects of personnel background will need assessment:
 formal academic or professional qualifications
 engineering expertise in design and analysis of ship structures
 relevant experience in the modeling and analysis of design problems using the finite element
method

3-8
  familiarity with, and appreciation of, the limitations of the software employed
Personnel are seperated into two categories: analysts and checkers. An analyst is a person who
undertakes the FEA. A checker performs independent checks of the analyst's work and certifies the
quality of the work. In some instances, there may be a lead analyst that provides direction and
oversight to less seasoned analysts. In this instance, the judgment as the appropriateness of the
lead analyst to act as the checker should be considered, and the qualifications of the less seasoned
analyst may be considered to be augmented by the supervision provided by the lead analyst.
Neither these guidelines or the routines built into software and documentation provided for those
routines can serve as a substitute fro knowledge of the principles involved. Ideally, an analysis
should be backed by analysis of the results of experimentation on a similar structure.
3/1.4.1 Academic and Professional Qualifications
Ideally, the analyst and the checker should be qualified to first degree level in engineering and have
taken at least one full course in FEA. Professional Engineer (or equivalent) status is good practice
for the checker and desirable for the analyst. In addition, different FEA vendors and professional
engineer committees are now providing professional training in helping understand both theoretical
and practical aspects.
3/1.4.2 Training and Experience
The analyst and checker should have received training in the application of the finite element
method. Either of the following is acceptable, in principle, as training:
 Training provided by various courses offered by educational establishments and software
vendors. These courses are only acceptable if they are application-oriented. It is preferable
that these courses also cover the theoretical background that could allow analysts and
checkers to understand the limit of finite element methods.
 In-house formal or informal training provided by a supervisor capable of satisfying the
requirements of a checker. The content of the training should be at least comparable to an
equivalent period of the application-oriented training program. The training course(s) should
be documented. In addition, it is recommended that the group has an internal best FEA
practices (company-specific) as a reference that is continually updated.
The analyst or checker must be familiar with the design requirements, codes of practice, analysis,
and design standards relating to ship structures. The checker must have, and the analyst should
preferably have, experience with analyses of comparable size and complexity as the analysis under
assessment.
The checker should be an experienced analyst with substantial experience in the application of the
finite element method. This experience should include working as an analyst on finite element
analyses that are comparable in complexity to the analysis the checker will be verifying. The
documentation should include a brief outline of previous experiences.
The experience requirements for analysts recommended by NAFEMS [1] are summarized in Table
3/1-1. The experience required of the analyst depends on the criticality of the analysis. The criticality
category depends on the consequences of the failure of the structure being analyzed. In addition, it
should be noted that the requirements listed are just minimums. Finding more experienced analysts
is encouraged.
Table 3/1-1 Minimum Recommended Experience Levels (adapted from NAFEMS, 1990)

3-9
Analysis Category Engineering Experience FE Modeling and Problem Solving
Design & Analysis FE Experience After Relevant Jobs Performed
Experience Formal Training for Each
Analysis Type
1. Vital: endanger human 5 years 6 months 2 x Category 1 under
life, or property or the supervision or
environment on a scale of 5 x Category 2 properly
a public disaster assessed
2. Important: Category 1 2 years 2 months 1 x Category 1 or 2 under
problem however analysis supervision or
is not an exclusive part of 3 x Category 3 properly
the integrity demonstration assessed
3. Advisory: All analysis 1 year 1 month Prescribed Benchmarks
other than the ones 1
covered in Categories 1
and 2
1 For example, see PART 4 of this report for benchmark problems
3/2 ENGINEERING MODEL CHECKS
The checks recommended in this section are generic in nature, and they form part of any
engineering analysis. The engineering model is a simplified representation of the physical problem,
and hence it is crucial that this modeling process is undertaken correctly since the finite element
analysis (FEA) cannot improve on a poor engineering model. The aspects covered in this section
include:
 Analysis Type
 Analysis Geometry
 Material Properties
 Loads and Boundary Conditions
The discussion here is restricted to an understanding of the physical problem as well as the
corresponding impacts of different simplifications. Translating these aspects into a finite element
model, in a format recognized by the software program, is covered in Section 3/3.
3/2.1 Analysis Type
The available FEA analysis types are as summarized in Table 3/2-1, and the conditions for the
analysis types are presented in a flowchart (see Figure 3/2-1 Analysis Type Flow Chart)
The analysis targets for each analysis type have also been briefly summarized in Table 3/2-1. Note
that the analysis targets are not limited to those listed in the table, and it is up to analysts to define
the analysis type and choose the proper FE solvers. It is also recommended to carefully review the
job specifications, and such that the analysis types and targets can be appropriately defined prior to
the pre-processing of FEA works.
Table 3/2-1 FEA Analysis Types

Analysis Types Analysis Target(s)

1 Linear static analysis Stiffness, etc.

Strength, remaining strength, permanent
2 Nonlinear analysis deformation, limit load, material/structural
failure, pre-load, etc.

3-10
 Analysis Types Analysis Target(s)
Buckling analysis
3 Critical load, post-buckling behavior, etc.
(linear or nonlinear)
Impact analysis Energy dissipation, critical stroke, failure
4
(quasi-static and dynamic) modes, crash resistance, etc.
Stress intensity factor, stress concentration,
5 Fracture/damage analysis
fracture driving force, crack propagation, etc.
Fatigue analysis
6 Fatigue life, remaining life, etc.
(high cycle and low cycle)
7 Whole Ship Analysis Hull girder stresses
Vibration Analysis
(frequency/transient/random-spectrum Eigenmode, modal/frequency response
8
response) vibration, etc.
(linear or nonlinear)
Thermal analysis
9 Temperature, temperature gradient, flux, etc.
(conduction, convection, radiation)
Optimization / DOE
10 Weight reduction, strength, fatigue life, etc.
(linear or nonlinear)
Multi-physics, electromagnetics, welding and/or Antenna design and placement, EMC,
11
co-simulations Structure-thermal-fluid interaction, etc.
An engineering model is a simplification and idealization of an actual physical structure or
component. The analyst should include, as a minimum, discussion of the following analysis type
topics in the body of the report:
 purpose of analysis (e.g., design, optimizations, failure investigation, integrity
assessment, validation & verification, design of experiment, etc.)
 whether the problem is static or dynamic
 whether the problem is linear elastic, nonlinear elastic, or nonlinear elasto-plastic
 whether the problem is time-independent or time-dependent
 appropriateness of linear elastic analysis
 appropriateness of nonlinear elastic analysis
 appropriateness of nonlinear elasto-plastic analysis
 assumptions of material models (i.e., constitutive relationship), as well as damage
models if applicable, for nonlinear problems
The loadings for all physical problems will be either static, dynamic, or a combination. Some
dynamic loads can be treated as quasi-static, and this should be done where possible, and it is
desirable to minimize the required computational power. However, special care should be taken
before treating any inertia force related loads as quasi-static.
The analyst will need to consider the frequency range over which there is significant energy in the
forcing function when a dynamic analysis is required. This will determine the number of modes to be
extracted. Note that different algorithms may develop modal values with certain differences.

3-11
 Figure 3/2-1 Analysis Type Flow Chart

The analyst can use the flow chart shown in Figure 3/2-1 to determine which type of analyses or set
of simulation methods should be used for a given physical problem. The flowchart is a two-way
lookup diagram:
(1) From top to bottom: if limit information is available for the physical problem, the
“YES”/”NO” logic helps the analyst understand which simulation method should be used.
For example, if a ship structure is under a blast load, which is generally an intermediate
or high-speed scenario, the explicit solver should be used for the simulation. Note that it
becomes a time-dependent problem where the applied loads and boundary conditions
should be representative of the physical observations.
(2) From bottom to top: the analyst may directly find the analysis type, if available. The
flow chart helps understand which set of simulation methods should be used in the
analysis. For example, the high cycle fatigue analysis is one kind of linear static
analysis, which means it is time-independent, no inertia effect, and no nonlinear
phenomenon considered.
The remainder of this section provides brief descriptions for a select number of Table 3/2-1 analysis
types.
3/2.1.2 Linear static analysis
A linear analysis has a straight-line relationship between force and deflection. This analysis type is
suitable for evaluating structures before yielding (i.e., stress-strain relationship within the linear
region) where buckling is not a concern. The calculation does not consider failures in nature (i.e.,
material models), and it is up to the analyst to conclude whether the structure meets the criteria
(e.g., yield strength or ultimate tensile strength).
Static analysis has loads that are independent of time. Note that in order for this to be the case, the

3-12
sum of the forces and moments must necessarily be in equilibrium, so they do not change with
respect to time.
3/2.1.3 Nonlinear analysis
A nonlinear analysis does not have a straight-line relationship between force and deflection. This
may be the result of one or more of the following nonlinearities:
 A material nonlinearity.
 A geometric nonlinearity.
 A contact or boundary condition nonlinearity
Material nonlinearity is caused by nonlinear stress and strain relationship (i.e., a nonlinear material
model). There are several types of nonlinear stress and strain relationships, including nonlinear
elastic, elasto-plastic, visco-elastic, and visco-plastic. For general metallic materials, a nonlinear
analysis is needed when the stress or strain state exceeds the yielding point. Note that a low cycle
fatigue analysis is an example of a material nonlinear analysis. Some non-metallic materials (e.g.,
rubber, plastic, and composite materials) may also show a nonlinear stress-strain relationship.
Material nonlinearity may also be observed at a typical creep loading condition (i.e., elevated
temperatures, external loads, and over a certain period).
A geometric nonlinearity indicates that the slope of force and deflection (F = Kd) curve is not a
straight line, i.e., K is dependent on d. For example, the value of K changes suddenly after reaching
the bifurcation point for a buckling problem. Note that for geometric nonlinear analysis, small strain
formulations are sufficient to capture the deformation/rotation and stress/strain states. It should be
noted that the force direction may change and follow the instantaneous local coordinate (i.e.,
deformed state) – also known as the follower force.
In an analysis that includes contact, the contact stiffness may need to be defined to capture the
contact and separation, where the contact stiffness is a function of displacement.
It should be noted that in the FEA codes, the analyst needs to define or enable the formulations
used for solving nonlinear problems, such as large or small deformation and finite or small strain
formulations.
Buckling analysis is a special form of nonlinear analysis. Buckling can occur when one or more of
the following are true: (1) compressive load (or equivalent compressive load from bending), (2) high
height-over-width ratios (e.g., slender beams or sheet metal), (3) bending stiffness is not comparable
with axial stiffness in terms of scales, i.e., K bending << K axial, (4) large lateral deformation, and (5)
geometric imperfection.
3/2.1.4 Impact analysis
The impact analysis can be either quasi-static (relative time-independent) and dynamic (time-
dependent). Note that, in general, it is recommended to use the implicit and explicit time-integration
methods to carry out quasi-static and dynamic impact analyses, respectively. The differences
between the two integration techniques will be discussed in the following section (3/3.6.2.2). The
current practice focuses on the time-dependent impact (crash/collision) analysis. The impact load is
characterized by the kinetic energy, which is a function of the mass of the ship and the collision
speed.
3/2.1.5 Fracture/damage analysis
Fracture and damage analyses are special FEA analysis types that investigate what will happen
locally in the vicinity of high-stress concentration areas. In such an analysis, high-stress
concentrations appear close to the tip of flaws/pores/voids/ cracks, and the magnitude of stress
depends on the tip geometry and the shape of the structure. The fracture analysis, in general, is
carried out to evaluate the structural integrity. The results may include critical load values, critical

3-13
crack sizes, fracture toughness, and remaining fatigue life. A damage analysis, in general, is carried
out to predict the material failure as well as the failure mode by evaluating the crack formation and
propagation in a macro- or mesoscale without considering the failure mechanism of materials in the
micro-scale.
In a linear fracture analysis, the stress intensity factor (K) and crack driving force (e.g., G – energy
release ratio and CTOD – crack-tip opening displacement) is measured. In a nonlinear fracture
analysis, the crack driving force (e.g., J-integral and CTOD) and crack-tip constraint (e.g., T-stress,
Q-index, Ai-index) are measured. Note that a nonlinear fracture analysis includes both small-scale
and large-scale yielding conditions. Therefore, formulations and solving options should be chosen
with extra care.
In a damage analysis, different damage models have been developed for both metallic and non-
metallic materials. In general, the damage criteria can be based upon the maximum principal stress,
maximum principal strain, critical strain energy, cohesive strength, triaxiality, and so on. The FEA
input parameters need to be calibrated from testing, and sometimes user-defined material models
are needed. Note that extra cares are needed to carry out such kinds of special analyses.
3/2.1.6 Fatigue analysis
In a fatigue analysis, there are two different types of analyses: high cycle fatigue analysis and low
cycle fatigue analysis. A high cycle fatigue analysis typically uses a stress-life approach with an SN
curve as the input. A low cycle fatigue analysis typically uses a strain-life approach with an EN curve
as the input. In general, a high cycle fatigue analysis applies greater than 1000 loading-unloading
cycles to a subject structure, to initiate a crack or exhaust the fatigue life. In design codes (e.g.,
[12,13]), the fatigue life can be evaluated from fracture based power law (e.g., Paris equation) and a
damage accumulative rule (e.g., Miner’s summation).
3/2.1.7 Vibration analysis
Vibration analysis has loads that are dependent on time. Dynamic analysis can be either linear or
nonlinear. Dynamic analysis in FEA typically takes one of five forms:
1. a modal analysis (also known as a natural frequency of free vibration analysis),
2. a frequency response analysis (also known as a forced vibration analysis),
3. a spectral analysis (also known as a random vibration analysis),
4. a transient response analysis
5. a shock analysis (i.e., dynamic design analysis method [DDAM])
A modal analysis determines the modes and the natural frequencies of a system. A frequency
response analysis determines the system response based on a single frequency input. The
response values can include displacement, velocity, and acceleration at any given node in an FEA
model. This type of analysis is discussed more fully in Section 3/2.8. A spectral analysis determines
the statistical likelihood of a given response (displacement, velocity, and/or acceleration) based on
an applied input spectrum. The input spectrum for such an analysis is a formula with frequency as
the independent variable and energy density as the dependent variable. The three foregoing
analyses are determined in the frequency domain. A transient analysis is a time-domain analysis
that determines the response at a given time. Lastly, a shock analysis is a specialized vibration
analysis typically used for naval vessels.
Ship structure vibration can cause structural fatigue, malfunction of machinery and equipment,
passenger discomfort, and/or crew habitability issues. It is therefore vital for ship structural
engineers to analyze ship structural vibration during the design phase and be able to correct it if
necessary.
Ship structure vibration issues can be addressed by making a change to one or more of the

3-14
three primary vibration variables. These are the excitation force, the structural stiffness, and the
damping. Reducing or modifying the frequency of the excitation force can reduce or modify the
vibration response. Increasing or decreasing the structural stiffness can move the natural
frequency of the structure away from the resonant frequency. Increasing damping can reduce
vibration response. None of these, however, are easy to modify in service; therefore, early
detection of any vibration issues is extremely important to any given design.
The primary sources of vibration excitation on a conventional propulsion ship are the main
engine and the propellers. The propeller induced excitations are the result of two phenomena.
First, the hull wake effect that causes alternating thrust to be transmitted through the propeller
shaft. Second, the hull-propeller clearance effect that causes alternating pressure pulses on the
hull. These excitation forces can be difficult to obtain. The main engine excitation force
information should be provided by the engine manufacturer. They should have information that
provides both the amplitudes of the forcing and the axis in which the forcing acts. See [14] for
more information regarding the main engine, hull wake, and hull-propeller clearance excitation
forces.
Ship structural vibration responses can be grouped into two groups: whole-ship hull-girder
responses and local responses. The hull-girder responses are the lowest natural frequencies of
a vessel. They are the result of the bending and twisting of the hull girder. Large ships have hull
girder frequencies of less than 5 Hz. Small ships can, however, have higher hull girder
frequencies. Local responses tend to have higher frequencies and are the result of mode
shapes specific to stiffened panels, stanchions, and other ship structures. As an approximate
guide, the following may be used for the first few modes of the below structures:
1. Hull Girder 1- 5 Hz
2. Main Mast 5 - 10 Hz
3. Superstructure 10 - 20 Hz
4. Typical Stiffened Plate Decks 10 - 40 Hz
Modal analysis is relatively simple to perform using FEA. These are used quite often in ship
structures problems such as determining the natural frequency of mast. Frequency response
analyses and spectral analyses are more valuable because they determine the actual
displacements, velocities, and accelerations of a structure based on a given cyclical force.
These, however, require premium FEA software and expertise. They are used less often at
present than modal analysis but are growing in use as the FEA software tools develop. They
are a valuable part of today’s ship design process. Transient analysis requires more
computational power than is practically available for most ship structure problems and is largely
not used in practice today in most ship design processes.
3/2.1.8 Thermal analysis
Thermal analysis is used to evaluate the temperature and fluxes of structures under thermal loading.
The temperature shows the amount of thermal energy, and the flux shows the flow of thermal
energy. In heat transfer analyses, there are various types of mechanisms, for example: (1) thermal
conduction, (2) thermal convection, and (3) thermal radiation. A thermal conduction analysis deals
with the thermal energy exchange by molecular motions [15]. A thermal convection analysis deals
with the thermal exchange between solid and fluids surrounded. A thermal radiation deals with
electromagnetic radiation emitted from a material.
The thermal analysis may be coupled with structural analysis to evaluate the stress or strain
responses of the structure in light of the thermal loading.
It is also worth noting that in thermal analysis, heat transfer may be steady or unsteady, and may be
linear or non-linear. Thermal radiation, as an example, is a highly non-linear process (4th power of
temperature).

3-15
3/2.1.9 Optimization and Design of Experiments (DOE)
The optimization analysis can be categorized into two different types: concept design optimization
and revision (or tuning) optimization. Concept design optimization includes topology, topography,
and free size optimization. Revision or tuning optimization includes size and shape optimization (see
Figure 3/2-2). Before carrying out such optimization analyses, the design variables and
corresponding targets should be predefined, including associated upper and lower bounds. For
example, in a size optimization analysis where the thickness of a metal panel is the design target,
the analyst would input a starting thickness and allowable thickness ranges.

Figure 3/2-2 An example of topology optimization [10]

3/2.2 Analysis Geometry

3/2.2.1 General
Ship structures are usually large and complex in nature and can typically only be analyzed after
idealization of the structure due to computing power limitations. It is, therefore, highly recommended
to carry out FE analyses by conveying only the essential parts of the structure.
One of the first questions to arise during the planning phase of an FEA is “how much of the structure
needs to be modeled to yield answers of the required accuracy.” This is best approached by
considering what the influence on the results of interest is by extending or reducing the extent of the
model. If the influence is negligible, then the extent of the model can be established in advance.
However, performing such an exercise on complex structures through intuition alone is difficult and
these decisions can be aided by a discussion of the purpose of the analysis. In such cases, the
analyst should ask the following questions before starting the FE modeling.
 “What is the purpose of this analysis?”
 Is it being done for the design or assessment of the existing structure?
 Is the purpose to analyze for stress, fatigue, vibration, dynamic response, or other purposes?
The answer to those questions will influence the appropriateness of the modeling performed. It will
also help the analyst determine which structure needs to be considered in the analysis.
Several simplifying assumptions are made in the idealization process. In order to do this
successfully, it is necessary to have a reasonable qualitative understanding of the expected
response. This will allow the reduction of the complex response of the actual structure to its
essentials. The elements that need to be considered in this idealization process are the character of
loading, the primary loading paths, and the parts of the structure that participate in the response.
Consideration of the likely load paths, deformed shapes/patterns, failure modes, or critical areas will
help establish the extent of the structure that should be modeled, and what boundary conditions
might be appropriate. Once the main structural actions are identified, it is possible to apply simplified
structural models to guide the analyst in deciding the extent of the structure to be modeled.

3-16
Figure 3/2-3 illustrates the concept with simple examples. The following general principles should be
borne in mind when using this approach:
Drastic changes in stiffness are potential regions to end the model. Figure 3/2-4 presents an

3-17
example in which the left-hand side of a beam is supported by a stiff structure. The bending stiffness
of beams is proportional to (I/L3), where I and L are the second moment of area about the neutral
axis and the span/length of the beam, respectively. In this example, a difference in stiffness of, say,
two orders of magnitude would be sufficient to justify the modeling approach shown in the figure.
This general approach can be adapted for other more complex structures. It should be noted that
different beam elements can be assigned with different cross-section types as well as I values.
Identification of load paths, deformed shapes/patterns, failure modes, or critical areas is a good
indicator of which parts of the structure are best to model.

3-18
Figure 3/2-3 Examples of Simple Models that can Indicate Extent of Structure to be
Modelled

3-19
 Figure 3/2-4 Large Changes in Stiffness to Indicate the Extent of Model

The actual extent of the finite element model depends on a tradeoff between the resources available
for the analysis and the general requirements that all significant portions of the structure be
modeled.
Most real structures are discontinuous and irregular at a local level. For example, it is likely that
there will be brackets attached to the structure, openings, access holes, etc. The explicit modeling of
these features is not practicable, and not necessary if only the global response is of interest, or such
discontinuity has little impact on the response of interest area.
All structures are three-dimensional (with thickness). Depending on the configuration, it is often
possible to reduce the number of dimensions to be considered. One basic application of this
reduction is the typical reduction of a ship structure from a 3D solid model to a 3D model that uses
2D elements that have a mathematical means of representing their thickness, including the stresses
across the thickness.
If a free or forced vibration analysis is being performed, additional cares should be exercised in
determining the extent of the model. The PART 5 ADVANCED ANALYSIS SAMPLE
APPLICATIONS frequency response application Section 5/4.5 Force shows that the extent of the
model has a significant impact on the results. Specifically, a main engine forced vibration analysis
may require the ship structure to be modeled from engine room bulkhead to engine room bulkhead
or beyond depending on the ship's structural arrangement. Just modeling the main engine
foundation down to the tank top would result in a non-conservatively stiff FEA model. Note also that
in the current engineering practice, the substructure technique (e.g., super-element method [16]) is
being widely adopted for such kinds of noise, vibration, harshness, and/or fatigue analyses.
If the FEA is concerned primarily with local effects, the concepts underlying Saint-Venant's principle
can be helpful in establishing the extent of the model [17,18]. Essentially, this principle states that
the replacement of a load (which could be caused by a restraint) by a different, but statically
equivalent, load causes changes in stress distribution only in regions close to the change. Figure

3-20
3/2-5 illustrates the principle. Note that such kind of time-independent equivalent load technique can
be achieved by using different load types or the combinations of them (e.g., force, displacement,
pre-stress, pre-strain, thermal, etc.) [19–21].

Figure 3/2-5 Illustration of Saint-Venant's Principle

The analyst should describe and justify the extent of the model. The justification statements should
include a discussion of:
 all significant structural actions captured by the model.
 including and excluding parts of the structure
 taking advantage of symmetry, antisymmetry, axisymmetry, plane-stress, or plane-strain
 requirements to accurately predict stresses, strains, forces, deflections, and/or other
parameters of interest.
 regions of the structure of particular interest
 whether the Saint-Venant's principle is satisfied
 obvious changes in structural stiffness that may suggest a model boundary
 very local application of the load to a large uniform structure
 for large models, can top-down analysis be used?
 for large models, can substructure (or super element) technique be adopted?
 whether the structure can be modeled with line elements (1D), area elements (2D), or
volume elements (3D) or a combination of different element types (e.g., hybrid or user-
defined)
 whether the formulation of the element is 1st- or high-order, or special type (e.g., crack-tip
element to present singularity or cohesive element)
3/2.2.2 Mass and Added Mass
Certain problems in ship structures require that the interaction between the structure and the fluid be
considered. The comments made here are limited to cases in which fluid displacements are small.
The most common example is the vibration of plated structures adjacent to the fluid.
For vibrations of the plated structure adjacent to fluid, the practice is to account for the presence of
the fluid by adding masses to the structure to represent the fluid. This mass is usually termed "added
mass" and represents the part of the mass of fluid that the structure has to accelerate during
vibrations. There are several sources for data on added mass appropriate to plate vibrations (see

3-21
ISSC, 1991- Report 11.2 for typical sources).
Hull girder vibrations can be treated similarly. Reference [1] provides guidance on approximate
methods for computing added mass for the hull girder.
The use of added masses to account for fluid-structure effects is generally quite approximate. More
rigorous methods require the finite element modeling of the surrounding fluid. Many general-purpose
FEA systems include fluid elements that allow certain types of acoustics, sloshing, and fluid-
structure analysis problems to be solved. This is a specialist area. For guidance, the reader is
referred to finite element texts and the user manuals of the FEA system to be used in the analysis.
See also Reference [2] for additional guidance on the application of added mass to vibration
analyses.
3/2.2.3 Shock Analysis Mass Modeling Reduction
Shock analysis, that would be otherwise too large, require a reduction in the number of dynamic
degrees of freedom (DOFs). This can be done by changing the way the mass is modeled.
Certain techniques, such as Subspace Iteration, implicitly reduce the size of the problem. The
degree of reduction depends on the number of modes that need to be extracted. The reduction
process can also be accomplished more directly by a procedure known as condensation, and
perhaps the best known such technique is Guyan reduction. While the condensation process is
generally detrimental to accuracy, the loss of accuracy need not be significant if the appropriate
guidelines are followed.
There are two alternative methods for mathematically modeling mass. The simpler of the two
methods is the lumped mass method in which concentrated mass is located at nodes. The value of
the mass represents the mass of the surrounding structure and equipment. This approach yields
mass matrices that are diagonal. Rotational inertias may also be modeled in this fashion or can be
condensed out. Rotational inertias are often ignored when this method is used. The alternative
approach is called the consistent mass method. This is a theoretically rigorous method that results in
a mass matrix with off-diagonal terms. The presence of these off-diagonal terms in the mass matrix
is responsible for making dynamic analysis using consistent mass matrices more computationally
demanding than when using lumped mass matrices. For large models, there does not appear to be
much difference between the two methods in terms of the accuracy attained, at least for lower
frequencies.
Whatever the technique may be for calculating natural frequencies and modes, the mass distribution
needs to be accurately modeled.
Natural frequencies and modes are calculated for one of the following reasons:
1. to compare natural frequencies and modes of a structure with the frequency/ies of some
source of vibration
2. as the first stage in the calculation of structural response.
In either case, it is necessary to anticipate the results to some extent. In the first case, the natural
frequencies calculated must bracket the frequency of the vibration source. In the second case, the
spectrum of the forcing function, for example, harmonic forces from the propellers or impulse loads
from underwater shock, will suggest the range of natural frequencies of the structure that need to be
calculated.
The higher the vibration mode, the more detailed the mass distribution needs to be. The general
principle is illustrated in Figure 3/2-6. In the actual structure, the mass is distributed over the length.
Hence, a reasonable number of lumped masses are required to represent the distributed mass. For
higher modes, a more detailed representation of mass is required because the mode shape is more
complex. In the example shown in the figure, essentially, a single mass is being used to represent
the dynamics of one lobe of the third vibration mode. This is in contrast to the five masses used to

3-22
represent the dynamics of the single lobe in the first mode.

Figure 3/2-6 Mass Distribution Required for Accurate Determination of Natural

Frequencies

Once the frequency range of interest is decided upon, the mode shape for the highest
frequency in this range needs to be estimated. This will indicate the number of dynamic DOFs
required to yield accurate results. Predicting a mode shape in advance is usually very difficult
unless the structure is relatively simple. Therefore, it may be necessary to follow an iterative
process in which the mass distribution is refined at each iteration.
Certain algorithms require any problem size reduction to be undertaken by the analyst. In this
case, the analyst selects the number of dynamic DOFs to be used in the analysis. The selection
of the dynamic dot's to be used in the dynamic analysis requires considerable skill except for
the simplest structures. The selection of dynamic DOFs can be automated. The principle
underlying the Guyan reduction process provides a guide on how this should be done if done
manually. The most important dynamic DOFs are those that have the largest mass-to-stiffness
ratio. This is because such masses are responsible for most of the vibration energy at lower
modes. The concept underlying the selection of dynamic DOFs is shown in Figure 3/2-7.
Viewing a plot of the mode shapes will allow an assessment to be made of the reasonableness
of the selection of dynamic DOFs.

3-23
 Figure 3/2-7 Selection of Dynamic DOFs

For most structural dynamics problems, translational masses are sufficient to define the
problem. However, when components and equipment with large dimensions are being modeled,
it is prudent to model their rotational inertia. If a single mass element is being used to model the
component, then three rotational inertias should be input in addition to translational mass data.
Alternatively, several masses can be input that approximately simulates the mass distribution.
The procedures are summarized in Figure 3-2.6.
A summary of guidelines to be followed in selected in dynamic DOFs is given below:
1. The number of dynamic DOFs should be at least three times the highest mode required. For
example, if thirty modes are required, at least ninety dynamic degrees of freedom should be
specified.
2. Dynamic DOFs should be located in regions where the highest modal deflections are
anticipated.
3. Dynamic DOFs should be located where the highest mass-to-stiffness ratios occur on the
structure.
4. If a dynamic response computation is to be eventually performed, dynamic DOFs should be
located at points where forces are to be applied.
5. For slender structures, such as masts, only translation dynamic DOFs need to be selected.
6. For stiffened plate structures, only dynamic DOFs at right angles to the plane of the structure
need to be selected.
7. Enough dynamic DOFs should be retained such that the modeled mass does not differ from
the actual mass by more than 10%.

3-24
 Figure 3/2-8 Modeling Rotational Inertia

3/2.3 Material Properties

The most common materials used in the construction of ships are metallic. Non-metallic materials
are also used, for example, glass-fiber or carbon-fiber-reinforced plastic (GRFP or CRFP)
composites, and other kinds of cellular materials (e.g., wood). The scope of these guidelines is
confined to metallic materials, mainly working in the elasto-plastic range. Note that the properties of
certain materials are loading rate-dependent, and this may need to be considered.
While Poisson's ratio for steel is not very sensitive to the increase of temperature, Young's modulus
and yield strength do reduce significantly when the temperature starts to get above a few hundred
degrees Centigrade [22]. Nuclear air blast explosions can cause thermal effects of sufficient
magnitude to influence the values of Young's modulus and yield strength. High strain rates may
increase the value of the yield and ultimate strengths of certain materials. For example, steel and
aluminum show a strong rate-dependency, but magnesium can be regarded as rate-independent.
However, these strain rates have to be very high to have a significant effect. Examples, where
structures may be subject to high strain rates, include structural response to collision, armor-
piercing, underwater explosions, and nuclear air blast. As a general guide, the effects of strain rate
should be considered for strain rates over 0.1 s-1.
3/2.3.1 Metallic Materials
The common inputs for metallic materials are yield strength, Young’s modulus, Poisson’s ratio, shear
modulus, and density. For advanced analyses, more detailed material properties are required.
Example metallic material properties are listed in Table 3/2-2.
Table 3/2-2 Example Metallic Material Properties

Material Yield Young’s Poisson’s Density Heat Thermal

(reference strength modulus ratio [kg/m3] Capacity conductivity
grade) [MPa] [GPa] [-] [J/(kg°C)] [W/m·K]

Steel 235 210 0.3 7.8*103 0.66 25

(S235)
Aluminum 145 70 0.33 2.7*103 0.90 154
(6061-T4)
Magnesium 130 45 0.35 1.8*103 1.00 61
(AM60)

3-25
Note that these Table 3/2-2 values are reference values and strongly dependent on the grades as
well as alloy compositions. It is highly recommended to get verified material properties from
specifications or material vendors/suppliers for each FEA analysis.
3/2.3.2 Composite Materials
Modeling the behavior of composite materials is more complex than modeling isotropic materials
such as steel. Composite materials are anisotropic and cannot always be regarded as a continuum.
In cases where the global response is of interest, it may be reasonable to model composite materials
using an anisotropic continuum model. More local analysis requires explicit modeling of the material.
Most general-purpose FEA software systems include the capability to compute the elastic properties
of composite materials. This is done by defining the individual layers that comprise the composite.
Alternatively, it is often possible to input the constitutive matrices that define the relationship
between generalized forces and moments to generalized strains and curvatures.
The failure modes of composite materials are also more complex than those that typically apply to
isotropic materials. To check the adequacy of a structure made from composite materials, it is
necessary to define the failure criteria that must be applied. Whereas with isotropic materials, a
single failure criterion (e.g., yield stress) is typically applied, with composite materials failure criteria
are generally different for different directions and can be applied to strains, stresses, and
combinations of stresses and strains.
There are other modeling issues that are particular to composite materials. Depending on the design
of the composite, it may not be possible to apply symmetry conditions even when the loading and
the overall geometry are symmetrical about one or more axes.
3/2.3.3 Stress and Strain Definitions
In general, the stress-strain relationship can be obtained by carrying out standard tests (e.g., uniaxial
tensile test). For example, ASTM-E8 [23] is one of the most widely adopted standard test methods to
obtain the stress-strain relationship of metallic materials. In the standard, the geometric
configurations of standard coupons (e.g., cross-section area – A0, gauge length – L), as well as the
specifications of testing rigs, are well defined, aiming to obtain a set of reliable data.
The stress and strain from standard tests are engineering stress (ENG) and engineering strain
(ENG), respectively:
ENG = F/A0
ENG = L/L
Where F, A0, L, and L are measured force from load cells, initial cross-section area, length
increment, and initial gauge length. A typical stress-strain curve is shown in Figure 3/2-9.

3-26
 Figure 3/2-9 Comparison of engineering and true stress and strain curves.
Reproduced from Fig.8.4 in Ref. [24]

The FEA software input is often required to be true stress (T) and true strain (T) calculated from
updated geometry:
T = F/A = ENG(1 + ENG)
T = ln(ENG + 1)
where A is the instantaneous cross-section area. From the engineering stress and engineering strain
shown in Figure 3/2-9, the true stress and true strain are generated. It is clearly seen that there are
pronounced differences between the two sets of curves. Note that nowadays, several commercial
codes have the feature to automatically convert the engineering stress and engineering strain input
to true stress and true strain, but additional care in the FEA setup is needed for end-users.
It should be noted that both engineering stress-strain and true stress-strain are widely mentioned
and presented with definitions. However, what happens behind the scene in the solver – “solver
Blackbox” involving other stress and strain definitions, e.g., Cauchy stress, 1st Piola Kirchhoff stress,
2nd Piola Kirchhoff stress, Green-Lagrange strain, or Euler-Almansi strain [25]. These stress and
strain definitions are adopted for solving different problems, such as small deformation/strain (Green
strain) and finite/large deformation/strain (Almansi strain). In addition, the limitations in formulations
on the use of elements should be recorded and evaluated.
3/2.3.4 Material Models Stress-Strain Relationships
For metallic materials, the time-independent elasto-plastic material models are generally adopted.
The following are several key components included in the material model:
3/2.3.4.1 Yielding Criteria
The stress-strain state when the material yields and generate plastic strain. von Mises criterion [24]
is commonly used for metals. Some may also use the Tresca criterion [24], but extra care is needed
in these cases.
3/2.3.4.2 Stress-Strain Curves
There are generally two types of stress-strain curves: discrete curves and continuum curves.
Discrete curves are widely used in solving general engineering problems, which include (a) rigid
ideal plastic, (b) ideal plastic, (c) bi-linear, and (d) multi-linear curves (see Figure 3/2-10). Continuum
curves are used for solving special engineering problems (e.g., fracture, collision, and explosive).
Several widely used continuum stress-strain formulations are Ramberg-Osgood, Power-law, and

3-27
Johnson-Cook [26,27].

Figure 3/2-10 Discrete Stress-Strain Curves

3/2.3.4.3 Hardening Laws

Three common types of hardening laws are:
1. Isotropic hardening model (yield surface gradually expanded after each loading-reverse
loading cycle)
2. Kinematic hardening model (yield surface translated after each loading-reverse loading
cycle)
3. A combination of both isotropic and kinematic hardening models.
3/2.3.4.4 Flow Rules
In plasticity theory, deformation theory (i.e., equivalent to nonlinear elastic) and incremental plastic
theory (J2 theory) are the two rules to define the response of plastic strain with the variation of
stress. The J2 theory is commonly adopted in the FEA simulation. Note that in solving certain
fracture mechanics related problems, deformation theory is still used to calculate the crack driving
force (J-integral).
3/2.3.4.5 Rate and Temperature Effects
The material properties (such as yield strength and ultimate tensile strength) of some metallic
materials depend on strain rate 𝜀̇ (at 𝜀̇ > 0.1 s-1), e.g., steel and aluminum. A typical strain rate effect
includes an increase of yield strength as well as the raise of the stress-strain curve after yielding and
maybe a reduction of ductility (ultimate tensile strength and strain).
The Cowper-Symonds and Johnson-Cook models [28] are the models used frequently to simulate
the strain rate effects:
 Cowper-Symonds model
1
𝜀̇ 𝑝
𝜎 = 𝜎𝑠𝑡𝑎𝑡𝑖𝑐 (1 + ( ) )
𝐶

where C and p are user-defined input constants. This model is essentially based on a scaling
term to predict the dynamic stress.
 Johnson-Cook model
𝑛
𝜎 = (𝐴 + 𝐵(𝜀𝑝 ) ) (1 + 𝐶𝑙𝑛(𝜀̇ ∗ )) (1 − (𝑇 ∗ )𝑚 )

where A, B, C, n, and m are user-defined input constants; 𝜀𝑝 is the equivalent plastic strain,

3-28
 ∗
𝜀̇ ∗ (= 𝜀̇𝑝 /𝜀̇0 ) is the normalized strain rate based on 𝜀̇0 = 1𝑠 −1 , and 𝑇 ∗ (= (𝑇 − 𝑇𝑟𝑒𝑓 ) )
reflects the temperature impact.
3/2.3.4.6 Failure and Damage Models
In general, a “failure” refers to either material failure or structural failure (e.g., buckling). The
explanations of this sub-section will be restricted to material failure. A metallic material failure may
include the following forms:
 Damage (e.g., pores/voids)
 Microcracks (e.g., coalescence)
 Fracture (e.g., separation)
 Reaching a high-stress level over-yielding (e.g., > UTS)
Failures can be predicted from the stress-strain state from the following hypotheses (but not limited
to):
 Maximum principal stress hypothesis – In this hypothesis, failure occurs when the principal
stress reaches a critical value.
 Maximum principal strain hypothesis – In this hypothesis, failure occurs when the principal
strain reaches a critical value.
 Maximum strain energy hypothesis - In this hypothesis, failure occurs when the total
deformation energy per unit volume reaches a critical value.
Failures can also be predicted from ductile damage models, such as the Lemaitre damage model
[29,30] and the Gurson damage model [31].
3/2.3.4.7 Material Property Calibration
The basis of selecting a material model is to represent the material behavior (linear and/or nonlinear)
under certain loading conditions. The material model should be carefully selected after adequately
calibrating against experimental or empirical data such that the numerical output can represent the
structural behavior accurately.
The stress-strain relationship is one such material property that requires calibration against
experimental results. The below Figure 3/2-11 shows the difference in the stress-strain relationship
between a non-calibrated FEA test and an experimental test of an ASTM E8 dog bone coupon.
Calibration of the FE model parameters will result in better agreement between the model and
physical trials, thus demonstrating that the model is capable of simulating reality.
Recently, several analytical methods have been adopted to calibrate the stress-strain relationship
and to replace the traditional experiment requirement. These include the least-square curve fitting,
parametric optimization, and design of experiments (DoE) [32]. These methods can reduce the
calibration time and increase the calibration quality.

3-29
Figure 3/2-11 Stress-Strain Comparison of Non-Calibrated FEA Results with Experimental
Results

3/2.3.5 Neuber Correction Method

The Neuber correction method has become popular in carrying out a fatigue analysis. The method is
one of the scaling methods that can capture the nonlinear stress-strain relationship after the yielding
point based on the linear elastic analysis. This method includes three steps:
1. Calculate the stress and strain values in the hot area(s), i.e., high stress or strain, from the
linear elastic analysis;
2. Determine the strain energy of interest areas (Area-1 in Figure 3/2-12);
3. Estimate the stress and strain pair from the nonlinear stress-strain curve that has the same
strain energy (Area-2 in Figure 3/2-12), i.e., Area-1 = Area-2.
The Neuber correction is depicted in Figure 3/2-12. This is a straightforward procedure to apply the
correction, as long as the stress-strain curve is provided. More details and free calculation software
are available in [33–39]. It should be noted that the Neuber correction method is only valid under
relatively low loading levels. A rule of thumb when using this method in a durability analysis is that it
is applicable if the load level is less than 1/3 of the limit load (full yield level). One of the advantages
of adopting this method is a significant amount of calculation time can be saved in the analysis.

3-30
Figure 3/2-12 Neuber Correction from the Elastic Stress State to the Elasto-Plastic Stress
State

3/2.4 Loads and Boundary Conditions

All loads that need to be considered should be described. The description should include a brief
discussion of the accuracy level of the load.
Loads (compiled by Giannotti & Associates, 1984) typically applied in ship structural analyses
include the following:
1. Hull Girder Loads consist of wave-induced and still water loads on the hull girder. This
load should be considered for longitudinal structure in the main hull, and for the
interaction of a long continuous deckhouse (superstructure).
2. Hydrostatic Loads are pressure loads due to fluids. The pressure could be either
internal or external. Examples of hydrostatic loads are external pressure of the bottom
and sides of shell plating, and internal pressure in tanks and on watertight bulkheads.
3. Hydrodynamic Loads consist of liquid sloshing in tanks, shipping of green water on the
weather deck and impacting on the house front, and wave slap on all exposed structure
and equipment above the waterline, etc.
4. Live Loads consist of uniform deck loading, concentrated loads such as forklift or
aircraft landing and parking loads, support reactions from stanchions and equipment,
cargo container reactions, etc.
5. Dead Loads consist of the weight of the structure.
6. Ship Motion loads consist of inertial forces that act on the entire ship and are important
design loads for masts and topside foundations, such as topside cargo attachments.
The effect of ship motion loads on the hull girder is to produce vertical and horizontal
bending moments and torsion. A lengthy analysis is required to determine these values
for a particular ship and service characteristics.

3-31
 7. Shock Loads consist of displacements, velocities, and accelerations in all three
directions. This load is important for naval ships in the design of vital equipment and
their foundations, and ship structure in the vicinity of these foundations.
8. Missile and Gun Blast Loads consist of transient pressure and thermal load for all
structures within the blast impingement area; usually, a static equivalent pressure is
used.
9. Nuclear Overpressure consists of the transient traveling pressure waves from a nearby
nuclear air blast; this is an important consideration in the analysis of deckhouses
(superstructures).
10. Vibratory Loads consists of cyclic loading from rotating machinery, especially from
propellers, low-frequency full girder response from slamming and springing can also be
significant.
11. Thermal Loads are caused by heat inputs
a. from solar radiation
b. exhaust impingement from stack gases
c. operation of machinery, especially combustion engines (important to deckhouses
and exhaust ducting), diesel generator foundations, and condenser foundations
12. Environment loads consist of wind, snow, and ice loads.
13. Impact loads consist of displacement or velocity in all three directions.
A description of the boundary conditions applied to the model, and the reasons for the approach
adopted should be described. The description should include, but not be limited to, a discussion
of:
 model symmetry, antisymmetry, and axisymmetry
 material property changes at the boundary
 stiffness changes at the boundary
 assessment of influence on results of assumptions made concerning boundary conditions
3/2.5 Impact/Collision Analysis
An impact or collision problem is a highly nonlinear problem, which is generally simulated by using
the explicit integration method (see 3/3.6.2). The impact problem is generally characterized by
kinetic energy. The kinetic energy is governed by the impact speed and the mass of the ship
(including hydrodynamic added mass).
If the ship structure collides with a relatively rigid object, kinetic energy is dissipated as strain energy
in the ship structure. If the ship structure collides with a deformable object (e.g., ship to ship, ship to
offshore structures, ship to ice, etc.) then the kinetic energy will be dissipated as strain energy into
both the ship and the object, and there may be some remaining kinetic energy if both the ship and
object are floating.
The impact problem involves large deformation and large plastic strain, and, therefore, significant
structural damage may occur. The strain energy dissipation should be estimated from the force-
displacement relationship. The structural damage can be captured by using different techniques.
These include element erosions, cohesive elements, or other element formulations associated with
material models with damage criteria.
In general, there are three different levels of impact analysis [40].
1. Local cross-section

3-32
 2. Component and sub-structure
3. Total system
The analyst needs to decide which level of engineering model should be adopted for the study. Note
that the interaction between three levels of energy dissipation should be considered and verified.
It should be noted that the elastic energy may contribute significantly on a global level, which may be
driven by total mass, impact speed, structural flexibility, material properties, and other factors.
In the FEA software, there are generally two means to define the impact speed: velocity and
displacement. Assigning a given velocity is a widely used method for impact analysis. The use of
displacement, however, is also convenient for sub-structure and local cross-section analyses (a
special type of submodeling analysis). This works especially well using the initial global model
displacement outputs at the boundary areas of submodels.
3/2.6 Fatigue and Fracture Analysis
3/2.6.1 Fatigue Analysis
About 90% of service failures are caused by fatigue that leads to a fracture, and sometimes the
failure locations predicted from the static or dynamic analysis are different from the test and field
observations [41,42]. One of the reasons for this is that the failure is caused by fatigue damage.
Fatigue is defined as progressive and localized damage that occurs when a structure is subjected to
cyclic loading [43]. There is a need to keep in mind that a fatigue failure is a probabilistic event, and
even a good structural design may not be fatigue-free. A reasonable structural design against fatigue
should involve both analysis and testing validation.
The fatigue analysis needs three key inputs that the analyst must review during the engineering
model check: the material properties, the geometric profiles, and the loading history. Figure 3/2-13
shows these within a schematic representation of the standard procedure for fatigue analysis. The
fatigue related material properties include the stress-cycle (SN) curves, strain-cycle (EN) curves, and
cyclic stress-strain curves. Note that the SN or EN curve should be consistent with the cyclic loading
conditions, such as R ratios, R = σmin/σmax, where σmin and σmax are maximum and minimum stresses
(see Figure 3/2-14). It is the most common to test at an R ratio of 0.1. The loading history includes
information such as the amplitude-frequency and loading path.

3-33
 Figure 3/2-13 Fatigue Analysis Procedures

Note that in Figure 3/2-13, some factors that may affect fatigue life are not presented. These other
factors include, but are not limited to, surface finish and treatment conditions, notch presence, and
residual stress presence. The analyst needs to collect this information by reviewing the engineering
model.
3/2.6.2 Basics of Fatigue Mechanics
In general, there are four different types of fatigue analysis:
1. Stress life fatigue analysis: This is also known as the high cycle fatigue (HCF). The data of
stress vs. cycle numbers (SN curve) is used as FE input.
2. Strain life fatigue analysis: This is also known as low cycle fatigue (LCF). The data of
strain vs. cycle numbers (EN curve) is used as FE input.
3. Fracture mechanics fatigue analysis: Either linear and nonlinear fracture mechanics
theories are used to estimate the crack growth rate as a function of crack driving forces
based on data from stress or strain life calculations.
4. Vibration fatigue analysis: This analysis can capture the resonance effect. Transient
cyclical loadings and frequency-domain power spectral density functions are the possible FE
input.
3/2.6.2.1 Cyclic Loading Definition
The cyclic loading can be characterized by using different methods such as a unit cycle associated
with repeating numbers, a whole cyclic history, or a simplified loading history from Rainflow counts
[44,45]. A unit cycle is one representative segment of the whole cyclic load history that can be
regenerated by repeating the unit cycle, see Figure 3/2-14.

3-34
 Figure 3/2-14 Cyclic Load Parameters

The unit cycle can be defined by the following parameters:

1. Mean stress, σmean
𝜎𝑚𝑎𝑥 + 𝜎𝑚𝑖𝑛
𝜎𝑚𝑒𝑎𝑛 =
2
2. Stress range, σrange
𝜎𝑟𝑎𝑛𝑔𝑒 = 𝜎𝑚𝑎𝑥 − 𝜎𝑚𝑖𝑛
3. Stress amplitude, σamplitude
𝜎𝑚𝑎𝑥 − 𝜎𝑚𝑖𝑛
𝜎𝑎𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 =
2
4. Stress ratio, R
𝜎𝑚𝑖𝑛
𝑅=
𝜎𝑚𝑎𝑥
Since the stress amplitude is half of the stress range, the unit cycle load can be explicitly defined just
by either the mean stress and the stress range, or the mean stress and the stress amplitude.
3/2.6.2.2 Fracture Mechanics Based Fatigue Analysis
The fracture mechanics-based fatigue analysis generally uses Paris’ law to determine the crack
growth rate from the stress intensity factor. A typical relationship between the crack growth rate and
stress intensity range is shown in Figure 3/2-15.

3-35
 Figure 3/2-15 Crack Growth Rate vs. Stress Intensity Range

Paris’ law is defined as follows:

𝑑𝑎
= 𝐶(Δ𝐾)𝑚
𝑑𝑁
where a is the crack length, N is the number of cycles, da/dN is the fatigue crack growth rate, and
K is the stress intensity factor range. The K is the stress intensity range and calculated from the
following equation.
Δ𝐾 = 𝐾𝑚𝑎𝑥 − 𝐾𝑚𝑖𝑛
where Kmax and Kmin are the maximum and minimum stress intensity factor for each load cycle,
respectively. The C and m terms are non-dimensional curve fitting factors to fit the power-law
relationship curve to the experimental test results. C and m are generally determined by fitting
regime B in the figure.
3/2.6.3 Fracture Analysis
In the presence of geometric features with sharp radii such as notches, flaws, and /or cracks, the
conventional approach of continuum mechanics would give erroneous answers. The fracture
mechanics-based analysis is carried out, which can describe the behavior of solids and structures
with geometrical discontinuity in a meso- or macro-scale. If required, a damage mechanics-based
analysis can capture the behavior in the meso- or micro-scale. Figure 3/2-16 shows the differences
among the classical continuum mechanics, fracture mechanics, and damage mechanics. In the
figure, their applicable conditions are schematically shown:
1. In the continuum mechanics problem, the material is homogenous.
2. In the fracture mechanics, the material is homogenous but with cracks or flaws.
3. In the damage mechanics, the material is homogenous, but the effects of pores, voids, and
cracks are considered.

3-36
Note that the current FEA practice mainly focuses on the fracture mechanics, and a limit amount of
damage mechanics related to material models is discussed

Figure 3/2-16 Differentiation of Continuum, Fracture, and Damage Mechanics

To perform a fracture analysis, an analyst needs to have a thorough understanding of basic local
crack/flaw geometric profiles, fracture mechanics material models, how to apply loads, and other
related information. Also, the fracture and damage parameters available have limits and are
dependent on conditions such as test temperature and test duration. An analysis can only get an
accurate evaluation with proper inputs. For example, if the J-integral (J) is of interest, small-
deformation and small-strain formulations are sufficient to describe the crack-tip field. However, if the
crack-tip opening displacement (CTOD) or crack-tip opening angle (CTOA) is of interest, the large-
deformation and large-strain formulations have to be adopted to capture the crack blunting. Note the
parameter J is a global fracture mechanics parameter, and the parameter CTOD/CTOA is a local
fracture mechanics parameter.
It should be noted that the fracture analysis and fracture mechanics-based fatigue analysis
(presented in the above two sections) are not widely used for designing the ship structures. These
analysis types are more relavent to structural integrity assessment or maintenance. For illustation
purposes, the fracture and fatigue example in Section 5/2 presents the approach of how to use these
methods to find the root cause of a significant crack observed in the bottom tank. It may also help
estimate the remaining service life when repairs have been undertaken.
3/2.6.4 Basics of Fracture Mechanics
In fracture mechanics, there are three deformed modes, and each type is shown in Figure 3/2-17.

3-37
 (a) Mode-I (b) Mode-II (c) Mode-III
Fig. XXX. Three basis modes of crack extensions:
Figure 3/2-17 Basic Modes of Crack Extension: (a) Mode-I: (a)Opening
Mode-I: Opening
mode; (b) Mode-II: in-
mode;
plane (b) Mode-II:
shear mode;in-plane shear mode;
(c) Mode-III: (c) Mode-III:
out-of-plane out-of-plane
shear mode.
shear mode.
 Mode-I is the opening mode
 Mode-II is the in-plane shear mode (also known as a sliding mode)
 Mode-III is the out-of-plane shear mode (also known as tearing mode)
Fracture mechanics can be divided into linear elastic fracture mechanics (LEFM) and elasto-plastic
fracture mechanics (EPFM). In general, LEFM is used for solving problems with brittle/elastic
materials, and EPFM theories are used for solving problems with ductile materials. If the loading
level is low enough, LEFM can also provide a reasonable approximation of physical reality for ductile
materials.
3/2.6.4.1 Linear Elastic Fracture Mechanics
In LEFM, the fracture is under K-control [46], and the crack-tip field can be accurately described by
using two fracture mechanics parameters,: K and G. The parameter K is the stress intensity factor
that is developed from the first term in the linear elastic crack-tip stress equation (Eq. 2.41 in [47]).
The stress intensity factor characterizes the stress, strain, and displacement field in the vicinity of the
crack tip. The parameter G is the Griffith energy release rate proposed by Irwin [48]. The energy
release rate quantifies the net change in potential energy caused by the increase of crack
extensions. Due to their natures, K is a local parameter, and G is a global parameter.
There are mathematical relationships between K and G that depend on whether the problems are
linear elastic or not, and on what the mode of crack extension is. For problems that are linear elastic
materials and are of the Mode I type, K and G are related by the following relationship:
1. Plane strain condition: G =𝐾𝐼2 /E’, where E’ = E/(1-v2)
2. Plane stress condition: G =𝐾𝐼2 /E
where E is the Young’s Modulus, v is the Poisson’s ratio, and KI is the Mode-I (opening) stress
intensity factor. The relationship between K and G in other failure modes (e.g., Mode-II or Mode-III)
can be found in the literature [47].
The analytical and/or empirical solutions of K have been widely investigated in the past several
decades, and the solutions have been reported in literature (e.g. Tada et al. [49]) and fitness-for-
service (FFS) standards (e.g. BS-7910 [12], API-579 [13], ASME Boiler & Pressure Vessel Codes

3-38
[50]).
Note that the parameter K can also be used for predicting the fatigue crack growth using Paris’s law
mentioned in the previous section (2.8.2).
3/2.6.4.2 Elasto-Plastic Fracture Mechanics
The EPFM can be divided into two different categories based on the yielding conditions in the vicinity
of the crack tip:
 small-scale yielding (SSY)
 large-scale yielding (LSY).
In some references, the full-scale yielding (FSY) condition is discussed separately from LSY. Here,
FSY is regarded as a subset of LSY. SSY, LSY, and FSY also refer to the global loading levels that
develop the crack tip with limit plastic zones, pronounced plastic zones, and fully plastic zones,
respectively. The details of these definitions are discussed in Ref. [47]. In general, the fracture in
both SSY and LSY problems of EPFM is controlled by J-integral (J), i.e., J-control. However, it
should be noted that just as there are limits to K-controlled LEFM, J-controlled EPFM may become
suspect when excessive plasticity or significant crack growth is presented (see Figure 3/2-18(c)).

Figure 3/2-18 Size Effect of Plastic Zone in the Crack Tip Stress Field : (a) SSY (b) LSY
with an intermediate amount of plasticity (c) LSY with the pronounced amount of
plasticity

Figure 3/2-18 schematically presents the size effect of the plastic zone in the crack tip stress field. In
SSY and LSY with an intermediate amount of plasticity (Figure 3/2-18 (a) and (b)), the crack tip
stress field is under J-control and can be described using HRR singularity. The HRR singularity
defines the stress variations in the plastic zone of the crack tip, which is proposed by Hutchinson,
Rice, and Rosengren [51,52].
Figure 3/2-18 (c) schematically shows a pronounced crack blunting of crack surfaces and plastic
zones under large scale yield. The one-parameter fracture mechanics theory breaks down in the
presence of such extensive plasticity, and the fracture toughness may become dependent on the
size and geometry of the test specimen. Note that there are strict conditions applied in measuring K
and J resistant curves per standards (e.g. ASTM E1820 [53], ASTM E399 [54], ASTM E1290 [55],

3-39
BS 7448 [56], and ISO 12737 [57]), aiming to minimize those impacts.
In the past decade, the two or multiple parameters fracture mechanics were proposed to describe
the crack tip stress field with extensive plasticity, e.g., T-stress (Williams [58]), J-Q (O’Dowd and
Shih [59,60]), and J-A2 (Chao et al. [61,62]). Table 3/2-3 summarizes the fracture mechanics
parameters being widely used in industry and adopted in FFS handbooks and failure access
diagram (FAD) methods (e.g., BS 7910-2015 [12]). Note that the multiple parameter fracture
mechanics recently presented in the scientific publications need further validations. Therefore, it is
recommended to use either J-T or J-Q in the analysis, which has been validated by the industry (per
BS7910 [12] or R6 [63]).
Table 3/2-3 Typical Fracture Mechanics Parameters

LEFM EPFM EPFM

List of parameters Two or multiple
One parameter One parameter
parameters
T-Q
Local stress-based K (stress intensity factor) T (T-stress)
T-A2
J-Q
Far-field energy G (Griffith energy
J (J-integral) J-A2
based release rate)
J-Q-Tz / J-A2-Tz
Local geometric CTOD CTOD CTOD-Q
profile based CTOA CTOA CTOD-A2

3/2.7 Whole Ship Analysis

3/2.7.1 Introduction
The objective of a whole ship finite element analysis is to obtain a reliable description of the overall
hull girder stiffness and to assess the global stresses and deformations of all primary hull members
for specified load cases resulting from realistic loading conditions including the wave-induced forces
and moments. Generally, the focus of the whole ship finite element analysis is not to judge local
stresses due to stiffener or plate bending, but rather to assess the stiffness and strength of hull
girder at the global level. A whole ship type model may be used to develop realistic boundary
conditions for a local model or loading that is investigating a more localized behavior.
The whole ship analysis is generally used for design and scantling development. In certain cases, for
example, commercial ships being designed under classification society rules and guidance, the class
rules will specify the extent of the whole ship finite element model, or possibly an extent covering
several midbody cargo holds. The class rules will also likely include guidance on key topics such as
mesh density, boundary conditions, loading conditions, and evaluation of results. Other whole ship
applications may be suited to using a finite element model that includes the entire ship structure,
bow to stern, full beam, and hull plus superstructure. Whole ship analysis can be conducted using
loading and structural design guidance or criteria from safety authorities such as class societies or
from naval design authorities. Whole ship FEA is essentially a larger, more complete version of a
standard FE model, and as such, solving it can be accomplished using many commercial finite
element modeling and analysis codes such as Nastran, ANSYS, or MAESTRO.
Disclaimer
The authors of this section are most familiar with the capabilities and limitations of MAESTRO
and have therefore used it for many illustrations. This does not constitute an endorsement of
that computer code over any other. An important part of this report is justifying the particular
computer code used for the finite element analysis, and any code should be acceptable if it
meets those guidelines.

3-40
3/2.7.2 Whole Ship Finite Element Model Development
The modeling necessary for the whole ship structural analysis is that the structural model should
provide results suitable for performing buckling, yield, fatigue and vibration assessment of the
relevant parts of the vessel as required by safety authorities or owner’s requirements. This is done in
whole ship FEA by using a 3D model of the whole ship, supported by one or more levels of sub-
models. The full-ship should be modeled, including all relevant structural elements as dictated by the
requirements of the analysis from safety authorities or the design team itself. For large primary
longitudinal girders and deep transverse frames, the web plate is best modeled by quadrilateral
plate/shell elements and the flange plate can be modeled as beam elements or plate/shell elements.
Stiffeners are often modeled by beam elements. Stiffened panels and grillages may be modeled as
an assembly of plate-shell elements and beam elements. Structures not contributing to the global
strength, and that has no influence on stresses in the evaluation area of the vessel may be omitted.
For ships with a relatively small deckhouse, such as tankers, the deckhouse may be omitted from
the model. For ships with a relatively large superstructure, such as naval ships, the decision of
whether including superstructure in a finite element model depends on the purpose of the analysis,
specific criteria, and the design requirements. A set of general guidelines for finite element modeling
of ship hulls is given by safety and design authorities such as classification societies and naval
design authorities [9,64,65]. The whole ship FE model and analysis can also be used to identify
locations where local refined FE mesh models within the full-ship model should be used to provide
more detailed evaluation to resolve high-stress levels or other design criteria or structural
performance issues.
3/2.7.2.1 Hybrid beam vs. Offset beam
Plates are often reinforced by attached beams (stiffeners, frame, and girders) to one side of the plate
(Figure 3/2-19). Beam nodes and plate nodes do not coincide but are typically separated by a
distance that is small in comparison with other dimensions. Common finite element programs offer
two formulations to address the problem.
1. “Hybrid” beam element (designation after Hughes [66], shown in Figure 3/2-19Figure 3/2-): In
this formulation, the beam axial stiffness is determined by the cross-sectional area of the
beam element, while the bending and shear stiffness is governed by the combined section of
the beam and the effective part of the plate, be. The plate is additionally idealized by plate
elements, normally taking only loads within their plate. The beam node is assumed to
coincide with the plate node.
2. “Offset” or “Eccentric” beam (shown in Figure 3/2-): In this formulation, the beam nodes,
which locate at the beam neutral axis, are connected to the plate nodes by rigid links. The
displacements of the beam nodes can be derived from the displacements of the plate nodes.
Thus the plate mid-surface becomes the reference plane of the assembled structure.
Stiffness matrices of beam elements are transformed.

3-41
 Figure 3/2-19 Modelling of Stiffened Plate

Both types of formulation are sufficient to evaluate hull girder primary stresses. However, they suffer
from disadvantages when evaluating hull girder secondary stresses. For hybrid beams, the effective
breadth of the plate, be, has to be specified by the user as data input to consider it in the sectional
properties, which becomes a problem to review and validate because there are different formulations
on how a plate effective breadth is defined. Furthermore, the hull girder global stiffness is slightly
affected by the assumption of shifting the beam neutral axis. For ships with stiffeners below the
deck, an increased moment of inertia and section modulus is obtained. For the offset beam
formulation, the data input is easy, and their global behavior is better than that of the hybrid beams
due to the correct location of their cross-sectional area, but their stiffness is underestimated when
the assembly is subjected large shear forces [66,67]. The error may become nontrivial in a coarse
mesh model where a stiffener between frames is represented by a single beam element. The error in
displacements and stresses decreases with increasing mesh refinement because the rigid links at
the additional nodes distribute the shear forces further between stiffener and plate [67]. In the early
days of full ship model finite element analysis, the mesh density was quite coarse. The hybrid beams
gave better results than offset beams. As computational power grew, the results using offset beam
formulation got better with the increasing mesh density. It should be noted that although the offset
beam has become industry standard in the full ship finite element analysis [65,68] because of its
simplicity, the hybrid beam concept is still used in the limit state beam-column buckling analysis.
3/2.7.3 FE Model Mass Properties and Hydrostatic Loading
Ship design is a multi-disciplinary engineering process, and the whole ship analysis must be
carefully integrated with the overall ship design. The full-ship FE model and analysis must reflect
accurate mass properties in order to generate accurate finite element analysis results. As a floating
structure, the overall ship mass properties will affect both hydrostatic and hydrodynamic loading. The
whole ship FEA will require hydrostatic loading and may also require hydrodynamic loading, as
discussed in Section 3/2.7.6. By the time a ship design reaches the structural finite element analysis
stage, the full-ship total weight distribution at 21 stations is often available from a weight report. In
addition, other weight items, such as significant equipment weights, miscellaneous or distributed
lightship weights, tank liquid weights, and cargo weights, may also be defined. The structural self-
weight can be automatically calculated using the finite element geometry and the associated
material density. For liquid tank loads, it is recommended to apply hydrostatic/hydrodynamic
pressure on the tank boundary. Note that for boundaries shared by two tanks, the pressure direction
and the element normal need to be carefully synchronized. Some FEA software[69], simplify this

3-42
process by grouping tank surfaces within the FE model and providing tank fill inputs in a format that
enables using traditional ship loading data such as stability model input.
The applicable significant weight items are usually modeled as nodal mass loads. Many of these
nodal masses are connected to the surrounding structure using rigid spline elements (RBE2/REB3)
in order to connect a single node at the item’s center of gravity to the expected footprint of the weight
item. This permits a more accurate distribution of forces when the load case includes accelerations.
Cargo masses can be distributed to a group of elements or nodes. For lightship weight, it is
recommended to smear the mass throughout the model. For general finite element packages, this is
often done by adjusting material densities. For ship specific finite element packages [69], the
program may include built-in routines to simplify this process.
To simulate hydrostatic buoyancy in a finite element model, the elements likely in contact with
seawater are grouped as “wettable” elements. The pressure side of these “wettable” elements is
also set to ensure the hydrostatic pressure is applied on the appropriate side of the elements. The
hydrostatic buoyancy forces are calculated by user-defined equations. Due to the nature of the finite
element faceted surfaces, partially immersed shell elements, and the model simplification (such as
omitting propeller and rudder), the buoyancy forces calculated using a hydrostatic tool (such as GHS
or NAPA [70]) are going to be different from the forces evaluated in the finite element model.
Consequently, to get equilibrium in a finite element model, the user needs to either use inertia relief
or adjust heel, trim, and height of the waterplane free surface to re-balance the finite element model.
Applying a waterplane free surface directly from another hydrostatic analysis tool on a finite element
model usually will result in imbalanced forces and moments and produce corresponding FEA errors.
An example is given in Section 5/3.8.
For ship specific finite element packages, a hydrostatic analysis is often built-in and will
automatically adjust the waterplane free surface to reach equilibrium using a finite element model.
The hydrostatic pressure is then computed and applied to the wetted elements after the model is in
hydrostatic balance or equilibrium. Figure 3/2-20 shows a nominal frigate stillwater hydrostatic
pressure distribution under the full load condition. To check whether a model is in equilibrium, it is
often useful to examine whether the hull girder longitudinal bending moment distribution and the
shear force distribution are in closure (or 0), as shown in Figure 3/2-21. The closure of the vertical
bending moments at ends is equivalent to ∑ 𝑀𝑦 = 0, and the closure of the vertical shear force at
ends is equivalent to ∑ 𝐹𝑧 = 0.

Figure 3/2-20 A Nominal Frigate Full Load Stillwater Hydrostatic Pressure Distribution

3-43
 (A)Bending Moment Distribution (B)Shear Force Distribution
Figure 3/2-21 A Nominal Frigate Vertical Bending Moment and Shear Force Distribution

3/2.7.4 Modal Analysis Check of the FE Model

While it may not be necessary to perform checks in free-free modal analysis for a pure static model,
it is recommended the check be performed if time and budgets allow. The model should be run
unconstrained with 7~12 modes. The first six modes should be rigid body modes, and the
frequencies are virtually zero. If the next couple of modes are not hull girder global elastic modes, it
often indicates
 Some of the elements may not be connected properly
 Certain structure stiffness is not properly modeled
 Some nodal mass is too large
In the following free-free modal analysis example, the first six modes represent rigid body motions of
the model. The first elastic mode (mode 7) did not display hull girder bending mode. By reviewing
the internal structure, it was found that a few beam elements were not connected correctly. After
fixing those modeling errors, mode 7 displayed a mode shape as expected. Figure 3/2-22 showed
the eigenmode before and after fixing element connection modeling errors. Figure 3/2-23 showed
the mode shape of mode 8. The local deflection at the sonar dome indicated either the stiffeners
were missing in the area or large nodal masses were incorrectly placed at those nodes.

Before fixing element connection modeling

After fixing element connection modeling errors
errors
Figure 3/2-22 Mode 7 eigenmode

3-44
 Figure 3/2-23 Mode 8 eigenmode

For large models, it is recommended that checks are performed as the model is being built. While
this may not always be possible, checking as-you-go is significantly easier than assembling an entire
large model and then debugging.
3/2.7.5 Boundary Conditions
Because ships are floating structures, the constraints are needed to remove ships’ rigid body
motions for static analysis. Applying appropriate boundary conditions is important and can be
complex. If the boundary conditions are not properly applied, high restraint forces and other factors
can easily render the analysis inaccurate. Constraints should be minimally applied to the structure
resulting in a minimal number of locations where boundary conditions are enforced.
3/2.7.5.1 Traditional Rigid Body Motion Constraint
For sagging and hogging load cases, it is suggested that rigid body motion constraints are to be
applied at the bow and stern locations, at the height of the closest continuous deck near the vertical
location of the neutral axis, as shown in Figure 3/2-24. These locations should be evaluated and
updated accordingly to minimize reaction forces and artificial stresses around constraints. For other
load cases, these constraints may also be applicable, as long as the constraint reaction forces are
small. If a constraint reaction force is greater than 1% of the whole ship weight, consider changing a
constraint location. A multihull vessel may require a different set of rigid body motion constraints
depending on the vessel proportions and number of hulls. These would need to be selected by the
analyst based on the given situation. In such cases, the constraints chosen should seek to follow
the same principle of ensuring that the constraint reaction forces are small.

Figure 3/2-24 Typical full-ship boundary constraints

3-45
Alternatively, a remote displacement can be used to restrain the model against rigid body motion if
the correct settings are available. A remote displacement creates a master node and slave nodes
that are connected by rigid beam elements. Some FEA software allows the user to set the behavior
of the slave nodes to “deformable,” which allows the shape of the slave nodes to change relative to
each other. This type of remote displacement, with a master node roughly located on the hull girder
neutral axis at midship and the slave nodes located along the length of the vessel on a rigid line,
may better distribute any non-real residual forces along the length of the vessel in a way that does
not impact the results. Such a rigid line may be the main deck to side shell intersection or a hard
chine on a box-shaped vessel (such as a dry dock). This approach should be executed with caution,
however, and a good understanding of how the deformable remote displacement functions within the
given software. The analyst should also pay just as much attention to minimizing the residual forces
and moments in this approach as in the three-point constraint approach.
3/2.7.5.2 Inertia Relief Constraint
The technique of inertia relief has been a well-known approach for the analysis of unsupported
systems such as air vehicles in flight, automobiles in motion, or satellites in space [71]. The sum of
forces and moments are calculated and applied to achieve an equilibrium state in inertia relief
analysis. In inertia relief calculation, the unconstrained structure or system is assumed to be in a
state of static equilibrium. Acceleration is computed to counterbalance the applied loads. A set of
translational and rotational accelerations provide distributed body forces over the structure in such a
way that the sum of applied forces and the sum of moments are zero. Since rigid body motions are
restrained, conventional static analysis can be performed. The inertia relief feature is available in
most general finite element packages. It should be noted that the accelerations calculated to balance
the applied loads should be small. If one of the acceleration components becomes relatively large, it
often indicates the model is not properly loaded, and a review of loads and the model is required.
3/2.7.6 Hull Girder Design Waves
The prediction of the behavior of the ship in waves represents a key aspect in the quantification of
both global and local loads acting on the ship. The solution of the seakeeping problem yields the
loads directly generated by external pressures and also provides ship motions and accelerations.
The latter is directly connected to the quantification of inertial loads and provide inputs for the
evaluation of other types of loads, such as slamming and sloshing.
A traditional analysis for the evaluation of wave-induced loads is represented by a quasi-static
design wave approach. The ship is statically positioned on a wave of given characteristics in a
condition of equilibrium between weight and static buoyancy. The scheme is analogous to the one
described for still water loads, with the difference that the waterline upper boundary of the immersed
part of the hull is no longer a plane, but it is a curved surface. By definition, this procedure neglects
all types of dynamic effects. One of the static design wave approaches is based on investigations
during the 1930s and 1940s that determined that the trochoidal profile appropriately represents sea
waves and that 0.61√𝐿𝐵𝑃 in meters (1.1√𝐿𝐵𝑃 in feet) represents the steepest stable wave height to
length ratio for ships under 150 meters in length [72]. The profile or form of the wave is chosen as
“trochoidal,” with the crest/trough of the standard wave centered amidships. Using the prescribed
wave profiles, the hydrostatic pressure to the wettable surfaces can be calculated and applied to the
corresponding finite elements. Note that the wave profile needs to be placed properly on the finite
element model to ensure equilibrium. For ship specific finite element packages [69], the sinkage,
trim, and heel of the model can be automatically adjusted into a quasi-static equilibrium between the
defined weight and the buoyancy.
The second method for applying hull girder design bending conditions is to prescribe a hull girder
bending moment distribution, then to assign artificial nodal forces in the FE model to generate the
prescribed bending moment distribution. The hull girder bending moment distributions are usually
given in the form of empirical equations by safety authorities. For U.S. Naval ships, SPECTRA [73] is
often used to generate hull girder bending moment distributions. Sikora [74,75] published significant

3-46
papers and reports detailing a reliability-based design method for primary seaway loadings on
surface ships. Typically, the lifetime operational profile of the ship is considered. The objective is to
form, on a reliability basis, estimates of global, primary hull girder loads. These include vertical and
lateral bending and longitudinal torsion, in addition to criteria for lifetime fatigue strength. Low-
frequency, wave-induced bending, and high-frequency, impact-induced whipping are also included.
Sikora et al. [74]incorporated empirical data and analysis into a computer program called SPECTRA
[73]. SPECTRA computes lifetime exceedance, histogram, and reliability information of hull girder
bending moments for a specified hull section. One way to apply SPECTRA lifetime extreme bending
moments to a 3D finite element model is to assume the longitudinal distribution of the dynamic
vertical bending moment as follows,

𝑀𝑚𝑖𝑑 2𝜋𝑥
𝑀𝑥 = (1 − cos )
2 𝐿
where Mmid is the dynamic wave bending moment value at midship. To simulate the bending moment
on a 3D finite element model, a large set of vertical nodal forces are placed on the model such that
the resulting hull girder longitudinal bending moment distribution matches the prescribed bending
moment distribution.
Another method to derive hull girder design dynamic wave loads is using potential flow based
hydrodynamic seakeeping codes. Most seakeeping tools are capable of exporting hull girder loads,
such as bending moments and shear forces. For tools with only sectional forces and moments
available, a method based on quadratic programming can be used to transfer the sectional loads to
3D finite element models [76]. For tools with panel pressure available as an output, such as VERES
(strip theory) [77], WAMIT [78], HydroStar [79] and PRECAL [80], it is desirable to map panel
pressure loads from a hydrodynamic model to the corresponding 3D finite element model to get a
more realistic structural response. However, because meshes for hydrodynamic analyses are often
much coarser than the corresponding finite element model meshes, the structural model often
becomes imbalanced after the panel pressure is mapped from the hydrodynamic mesh to the
structural mesh.
Although various interpolation methods have been proposed and used in design practice,
transferring panel pressure loads from a seakeeping model to a finite element model while
maintaining equilibrium still remains challenging. Often, the “inertia relief” method [81] has to be
used for the final adjustment to balance the structural model. The “inertia relief” technique is very
powerful and can correct any imbalanced model. However, there are two shortcomings to this
approach. First, the additional inertial forces cause a change in the hull girder response (such as
bending moment). Second, the change of the accelerations has to be relatively small to ensure the
fictitious inertial forces do not significantly distort the original structural response. This often requires
visual inspection and engineering judgments.
Malenica et al. [82] proposed a method that mapped the panel source strength instead of the panel
pressure from a hydrodynamic mesh to the structural mesh and then formulated the equations of
motion using the structural mesh. This concept can be adopted to achieve a balanced structural
model without using “inertia relief” (Ma et al. [83] ), Some ship / marine-specific FEA software
provides the ship designer with an integrated frequency-domain/time-domain computational tool to
predict the motions and wave loads of floating structures. When this tool is integrated within an FEA
toolset, the hydrodynamic loading process learning curve is greatly reduced, and the issue of
mapping data between a seakeeping code and an FEA program is eliminated. This approach takes
advantage of the existing structural mesh and defined loads to formulate the equations of motion,
and results in perfect equilibrium for the structural model, so no inertia relief or artificial loads are
required to balance the model. Bending moments, shear forces, and torsional moments are all
automatically in closure and will not result in the distorted FE results associated with the excessive
inertial balancing of a whole ship FE model.

3-47
Regardless of the toolset or computer codes used, the procedure must first compute the response of
ship motions, hull girder loads, and panel pressures for incident waves of unit amplitude based on
wave frequency, ship speed, and heading. The response amplitude per unit wave amplitude is often
referred to as the response amplitude operator (RAO). RAOs are effectively transfer functions which
give the proportion of wave amplitude “transferred” by the ship system into ship response. The short-
term and long term most probable extreme wave loads can be generated by the: hull girder
dominant load parameters (DLP) RAOs, operational profile, wave scatter diagram, wave spectra,
and the return period. In principle, the base assumptions of the linear wave theory are valid only for
small wave excitations, small motion responses, and low speed of the ship. In practice, the field of
successful applications extends far beyond the limits suggested by the preservation of realism in the
base assumptions; the method is used extensively to study extreme loads and high-speed vessels.
The typical hull girder dominant load parameters (DLPs), which represent the hull girder loading
conditions that are used as a set of design conditions for the hull structure, are:
 Vertical bending moment
 Vertical shear force at ¼ and ¾ of the vessel length
 Vertical acceleration at bow
 Roll motion
 Relative vertical velocity at the bow
 Longitudinal torsional moment
 Horizontal bending moment
 Horizontal shear force
The corresponding unit wave for each DLP is then scaled up to the extreme value for an equivalent
design wave load based on extreme ocean wave theory [84]. For each of these DLP load effects, an
equivalent regular wave, defined by its wave height, wavelength, heading angle and position along
the ship length, is determined so that the maximum response for the selected load effect is equal to
its value given by the rules or design criteria for the probability of exceedance considered. The
amplitude of the other effects is obtained from a ship motion analysis assuming the ship to be
positioned on the equivalent regular wave. The superposition of the dynamic design wave load and
the still water load is used for whole ship structural analysis. This procedure, illustrated in Figure
3/2-25, can be fully integrated into some software toolset [69].

3-48
 Figure 3/2-25 Extreme Load Analysis Procedure

Due to the level of detail in whole ship finite element models and the computational power required
to solve the source strength for each hydrodynamic panel, limiting the number of wetted panels in
the potential flow computation is an important consideration. Most potential flow codes have a
specified upper limit, typically of 3,000 to 5,000 panels. In the case of some software tools [3], three
different approaches can be identified to map the hydrodynamic mesh to the finite element mesh.
The first method is to compute the loads on a one-to-one ratio between the finite element and
hydrodynamic meshes. This method gives more accurate results but is computationally intensive,
and it is also limited in computational time, not the number of panels. The second method is to use
the limit state evaluation patches as hydrodynamic panels if they are available. The third method is
to leverage a NURBS surface model and map the surface mesh to the “wettable” finite element
elements. The number of hydrodynamic panels can be further reduced by using panel merging
algorithm, which automatically merges adjacent evaluation patches into larger panels while
maintaining a user-specified aspect ratio. Using the evaluation panel hydrodynamic mesh
discretization means that each panel is made up of several finite elements.
3/2.7.6.1 Loads Due to Nonlinear Seakeeping Analysis
The ABS Guide for Dynamic Loading Approach [81] Subsection 6/1 requires that a non-linear
seakeeping analysis is to be performed to establish the instantaneous design loads at a specific time
when each Dominant Load parameter (DLP) reaches its maximum. The frequency-domain extreme
load analysis results are used to identify the expected magnitude of each DLP’s most probable
extreme value and the ‘critical’ speed, heading, and sea state where it occurs. Time-domain
simulations associated with each DLP’s critical speed, heading, and sea state are run to capture the
nonlinear effects for each DLP and create what is considered to be a more physics-based
representation of the ship’s inertial conditions and associated hydrodynamic loading. A number of
separately seeded (independent) time-domain runs are computed for each DLP in order to create a
sampling of maximum conditions to select from. Each time-domain simulation uses a unique speed,
heading, and sea state combination. The panel pressures, point forces, accelerations, and hull girder
loads are typically recorded at every 0.5 seconds, and the time-step associated with the maximum

3-49
DLP value will be identified for use as the design load case condition.
A load scale factor can be determined using the ratio of the peak time-domain hull girder loads
(DLPs) and the frequency domain hull girder loads. The panel pressures, point forces, and
accelerations at the peak response time-step of each time-domain simulation are scaled such that
the magnitude of the DLP is equal to that predicted from the frequency domain analysis. If the peak
of the time domain run exceeds the frequency domain extreme value, the peak of the time domain
run is used for that particular DLP. To get an improved physics-based simulation of ship motions and
hydrodynamic loads while preserving a process that is practical in a design context, a weakly-
nonlinear time-domain simulation in a sea state is often used. The term weakly nonlinear refers to
the use of potential flow analysis to predict hydrodynamic loading in which the radiation and
diffraction forces are computed on the ship’s stillwater line in order to remain computationally
practical, while the Froude-Krylov and hydrostatic forces are computed using the actual wave profile
on the ship’s hull associated with each time step in the simulation.
3/2.7.7 Limit State or Failure Criteria
To assess the strength of whole ship primary structures, a set of acceptance criteria is often used.
The acceptance criteria include not only stress limits but also buckling rules and other limit states.
While the stress limit can be readily evaluated by extracting element stresses directly from finite
element analysis, evaluating structural component buckling is more complicated because a buckling
event occurs at the stiffened panel level but not at the finite element level. Currently, the use of an
empirical or semi-analytical approach for buckling assessment is an attractive strategy due to its
effectiveness in terms of engineering accuracy and related computational time. This aspect becomes
even more important when dealing with highly nonlinear analyses and in the context of optimization
procedures, in which repeated analyses are required.
The primary modes of overall failure for a stiffened panel under predominately compressive loads
can be categorized into the following six types [85]:
 Mode I: overall collapse after overall buckling
 Mode II: collapse of the plating between stiffeners without the failure of stiffeners
 Mode III: beam-column type collapse of a stiffener with attached plating
 Mode IV: local buckling of stiffener web (after buckling collapse of attached plating)
 Mode V: lateral-torsional buckling (tripping) of a stiffener
 Mode VI: gross yielding

Mode I Mode II

3-50
 Mode IV
Mode III

Mode V
Figure 3/2-26 Typical buckling failure modes of stiffened panel

Class societies and safety authorities have published acceptance rules for the above buckling failure
modes, shown in Figure 3/2-26. While the rule-based strength criteria equations, also called “limit
states.”, are relatively simple, they are not integrated into general FEA programs. It often becomes
the responsibility of the engineering analyst to use internal organizational processes (e.g.,
spreadsheets and macros) to first extract structural response results (i.e., deformations/stresses)
and to then perform external strength calculations. Typically, when using whole ship FEA for hull
structure design, the engineering analyst post-processes these results within the context of the
global FEA model, which can also be difficult, time-consuming, and error-prone. In some finite
element programs [69], the limit state analysis is automated.The user can select from so-called first
principles engineering algorithms including limit states as defined in Hughes’ “Ship Structural
Design” [66], and ULSAP as defined by Hughes and Paik in “Ship Structural Analysis and Design,
Ultimate Limit State Design of Steel-Plated Structures” [85], or from implementations of U.S. Navy or
ABS criteria including HDBK-MIL-519 [72], DDS 100-4, ABS-DLA [81] and ABS’s High-Speed Naval
Craft rules (HSNC) [86].
This limit state evaluation automatically defines evaluation panels, which are a collection of finite
elements within the whole ship model used to define a stiffened panel or the true span of a beam,
and compute the panel’s and beam’s loads and load effects from the finite element analysis results.
Using this automation, the code evaluates the entire global structure, at the level of limit state
evaluation panels or structural entities, for every finite element analysis load case. Figure 3/2-27
shows two types of evaluation panels automatically defined for the entire ship’s structure, one for
stiffeners and plates, and one for stiffened panels.Each colored group of elements represents a
single evaluation patch, which is evaluated against each of the selected limit states for all load cases
defined.

3-51
 Figure 3/2-27 Frigate Evaluation Panels

3/2.7.8 Finite Element Analysis and Results

Before performing all analyses of various load cases on the full-ship model, obvious modeling errors,
such as materials, properties, units, weights, warped elements, element connections, and free
edges, should be checked and corrected. In addition, results under simple loads need to be
examined qualitatively to see if they “look right.” For example, for a global full-ship analysis, it is
often prudent to check and review the stresses and deflections of still water, and quasi-static
sagging and hogging load conditions. For these load cases, the stresses obtained from the finite
element analysis and from traditional beam theory should have a reasonable correlation. A stress
distribution comparison of finite element analysis and a simple beam analysis under hogging and
sagging waves is shown in Figure 3/2-28. This figure illustrates that the longitudinal hull girder stress
distribution of the finite element analysis tracks well with the simple beam analysis, but presents
more realistic stress behavior such as shear lags and corner stress concentrations. It should also be
noted that the stresses calculated using simple beam theory depended on the structural longitudinal
effectiveness assigned by the user. In this example, some of the deckhouse structure members and
the structures near the opening should be assigned as longitudinally non-effective and do not make
any contribution to the section modulus.

3-52
 FEA (Hogging wave) Simple beam theory (hogging wave moment)

FEA (Sagging wave) Simple beam theory (sagging wave moment)

Figure 3/2-28 Longitudinal stress distribution comparison between FEA and simple beam
theory

For full-ship global finite element analysis, longitudinal strength under vertical bending moment and
vertical shear forces is the first important strength consideration. The hull strength under other load
conditions, including horizontal bending, horizontal shear, and longitudinal torsion, should also be
considered. For multi-hull models, load conditions of transverse bending, transverse shear,
transverse torsion, squeezing, and prying need to be evaluated. Whole ship FEA typically includes
these major hull girder load cases, and FE results will be generated and reviewed for these
conditions. The reactions of structural components of the ship hull to external loads are usually
measured by stresses and deflections. Stress is related to gradients of the field quantity, and
gradients in a given element depend on field quantities at nodes attached to that element only.
Stress contours are discontinuous across element boundaries. Strong discontinuities indicate too
coarse mesh density or mesh transition, whereas practically continuous stress contours suggest
unnecessarily fine discretization. For each load case, the element stresses extracted from the FE
model are used to check against design acceptance criteria of yield, buckling, and fatigue, as
discussed in the previous section. Excessive deflection may also limit the structural effectiveness of
a member, even though material failure does not occur, if that deflection results in misalignment or
other geometric displacement of vital components of the ship's machinery, navigational equipment,
etc., thus rendering the system ineffective.
Finally, it should be noted that finite element analysis is only the mathematical approximation of a
real-world system. Modeling decisions are influenced by what information is sought, what accuracy
is required, the modeling and computational expense of FEA, and its capabilities and limitations.
Also, initial modeling decisions are provisional. It is likely that the results of the first FEA will suggest
refinements of the model and loads.
3/2.8 Frequency Response Vibration Analyses

3-53
3/2.8.1 Basics
A frequency response analysis requires less computational power because the FEA equations
can be solved in the frequency domain once the frequency is assumed to be constant. A
frequency response analysis determines the steady-state response of a structure that is loaded
at a single frequency. This type of analysis ignores any initial transient motion and assumes the
structure is linear.
A frequency response analysis determines the frequency response at a single frequency
independent of all other frequencies. The results are how the structure would respond at that
single frequency. A frequency response analysis, however, also typically produced results for a
series of frequencies. The results for each frequency are independent of the other frequency
results.
FEA programs often include post-processing tools that can convert the frequency domain
results into time-domain results for visualization purposes. This feature, however, is not
required because the frequency domain results contain all the necessary response information
in terms of displacement, velocity, and acceleration.
If an engineering model has more than one load and the loads occur at the same frequencies,
then these can be evaluated in the same frequency response vibration analysis. The loads also
can have different phase relationships with one another.
If an engineering model has more than one load, and the loads occur at different frequencies,
then multiple frequency response vibration analyses will need to be run. The maximum
response values can then be summed together to get the total maximum response because the
structure is assumed to behave linearly (as previously stated).
The interested reader may also refer to [87] for an excellent primer on frequency response
analyses.
3/2.8.2 Modeling Requirements
The model should be large enough to model all ship structural vibrations of concern. If the hull
girder vibration is a concern, then the whole ship should be modeled. If the superstructure fore
and aft vibration is a concern, then the superstructure should be modeled as well. If the analysis
is concerned with only local vibrations, then considerable care should be exercised in deciding
where the boundaries of the model should be. Reference [88] recommends that if a local
structure is being evaluated, then the model should extend one major transverse bulkhead
forward and aft from the area of interest. The 5/3 FREQUENCY RESPONSE VIBRATION
ANALYSIS sample application is a local analysis. It includes a comparison of modeling the full
engine room bulkhead to bulkhead versus just modeling the structure above the tank top within
the engine room.
The model should include all relevant masses. This will certainly include the structural weight.
Equipment masses should be modeled separately in the FEA, where they may have an impact
on the results. The remainder of the loading condition mass should also be included.
ABS [14] recommends that added mass be accounted for and provides some different options.
See Section 3/2.4 for an additional discussion on the inclusion of the fluid mass surrounding the
structure.
ABS [14] recommends that damping be accounted for in the model using a damping coefficient
of 1.5% of the critical damping and that this is applicable to the entire range of typical propeller
and main engine frequencies.
Recent use of ANSYS for a local frequency response vibration analysis did not require any
need to reduce the model size using the dynamic degrees of freedom reduction methods
discussed in Section 3/2.5. These techniques may be needed for large whole-ship models

3-54
depending on the mesh density.
3/2.8.3 Frequency Range Requirements
Reference [88] recommends that a frequency response analysis be formed for a range of
frequencies +7 and -3% of the excitation frequencies. This is recommended for “relatively high
mesh density models” and that “it may be beneficial to increase this range in some cases, such
as when a lower density mesh is used.” This is recommended, “because the prediction at any
single frequency is likely to be erroneous.” The worst-case result of this range should be
reported in the conclusion of analysis to be conservative.
3/3 FINITE ELEMENT MODEL CHECKS
The subject of this section is the checks that should be performed to ensure that the physical
problem is appropriately translated into the finite element model. Guidance is provided on various
aspects of a finite element model, such as appropriateness of the element types used, the density of
finite element mesh used for plated structures, substructuring, and submodeling used to optimize the
problem size, loads and boundary conditions application, and the solution setup process. There is
also a short subsection on graphical checks using the software's pre and post-processors to
scrutinize the finite element model and results.
Since access to the software is essential to perform many of these checks, it is the responsibility of
the analyst to ensure that these checks are performed. However, documentation, in the form of plots
and graphs, should be available for audit.
Several examples illustrating finite element modeling practices are presented in Appendix C. The
purpose of these examples is to show the effect of varying certain finite element modeling
parameters on the results. The main modeling parameters addressed in this appendix are element
type and mesh density.
3/3.1 CAD Importing
3/3.1.1 Geometry Check and Cleanup
The very first step of pre-processing a model is to deal with CAD geometry. The pre-processing
starts with the import of CAD data. The data can be in different formats (e.g., CATIA v4/v5, STEP,
UG, IGES / IGS, Parasolid, Pro-E, STL, ACIS, DXF, JT, etc.). Note that it is highly recommended to
have the CAD model in a format that is compatible with the FEA code being used. However, it is
common that issues or errors may occur when importing CAD data. Some of the “issues” are for FE
meshing but not for CAD designing, and it is up to the analyst to clean up these geometric issues for
meshing [15].
Several CAD issues that need to be cleaned up are listed as follows:
 Not stitched surfaces, i.e., gap(s) between surfaces
 Overlapped surfaces
 Broken surfaces
 Misaligned surfaces
 Redundant surfaces
 Not connected surfaces, i.e., penetrated each other without connecting (see Figure 3/3-1)
 Small surfaces not possible to mesh with a reasonable quality
 Other detail will result in meshing with very poor-quality elements
A more detailed list of issues can be found from Question 3.1.3 of PART 2.

3-55
 Figure 3/3-1 Not Connected Surfaces Reproduced from Ref. [15]

All the geometric issues and errors in the imported CAD model need to be resolved and cleaned up
by carrying out topology repair. Reference [15] recommends geometric cleanup be performed based
on the following strategy:
1. Define the global element size from the model scale
2. Determine the geometric cleanup tolerance (e.g., 15~20% of global element size) from the
global element size
3. Review the topology of CAD data to find the issues to be fixed
4. Clean up CAD models
a. Find and delete duplicate surfaces
b. Stitch surfaces with free edge pairs (e.g., equivalence to combine)
c. Toggle or untoggled issue edges
d. Fill missing surfaces
Many FEA software vendor user manuals and tutorials provide more detail steps and explanations of
how to find and fix geometric issues.
3/3.1.2 Geometry Defeaturing
Once accomplishing the geometry cleanup, the analyst needs to make a further geometric
assumption to eliminate unnecessary CAD information. This is the second step of pre-processing
and generally known as “geometry defeaturing.” In this step, the geometric details (e.g., fillets or
corners with a small radius and tiny holes) that may not contribute to the overall performance of the
component can be simplified or removed. The geometric profiles recommended to be defeatured are
listed in Question 3.1.3 of PART 2. Reference [15] recommends geometric defeaturing be performed
according to one of the following three algorithms:
 Preserve boundaries between components
 Recognize and preserve major feature edges
 Recognize and suppress construction edges
If the 2D mesh type is adopted, it is necessary to extract the mid-surface from the 3D geometry data.
The details of choosing a mesh type and mesh design will be discussed in the next section. The

3-56
selected variables for the mid-surface extraction are listed as follows:
 Algorithm for topology extraction
 Maximum thin solid thickness to width ratio
 Maximum thick solid thickness
 Minimum feature angle between the solid’s edge and its faces
 Thickness transition as well as the transition increment (e.g., ± 0.5mm)
After geometry cleanup and geometry defeaturing, the analyst needs to carry out another iteration of
the CAD data reviewing. This process will result in the model being free from topology violations,
especially in the area of interest.
3/3.2 Element Types
To some extent, all finite element types are specialized and can only simulate a limited number of
types of responses. An important step in the finite element modeling procedure is choosing the
appropriate element type(s). The elements best suited to the particular problem should be selected
while being aware of the limitations of the element type. A good guide to the suitability of an element
type is their performance in other similar situations. Also, it is rare to use only one type of element to
solve engineering problems. Typically multiple element types are used in combination (e.g., 1D, 2D,
3D, and others).
Element performance is generally problem-dependent. An element or mesh that works well in one
situation may not work as well in another situation. An understanding is required of how various
elements behave in different situations. The physics of the problem should be understood well
enough to make an intelligent choice of the element type. As a reasonable guideline, Cook [89]
considers elements of intermediate complexity work well for many problems. According to this
reference, the use of a large number of simple elements or a small number of very complex
elements should be avoided. The selection of element type is mainly dependent on four factors: (1)
geometry size and shape, (2) type of analysis, (3) structure actions to be modeled, and (4) time
allocated for the project per references [15,90].
3/3.2.1 Geometry Size and Shape
In each FE simulation, a model must be defined in 3D properly with discrete meshing (e.g., nodes
and elements). The global geometry can be divided into representative local geometries (in 1D, 2D,
and 3D), and then the element type for each part can be selected accordingly.
 For a geometry with one dimension significantly larger than the other two dimensions (e.g., X
>>> Y, Z), it is recommended to use 1D elements. For example, beam-like structures, long
shaft, rods, columns, bolt joints, fasteners, or connectors. The 1D elements can include 2-
node, 3-node, and 4-node elements.
 For a geometry with two dimensions significantly larger than the other one dimension, it is
recommended to use 2D elements. The 2D elements, in general, include the following
kinematic assumptions:
o Plane stress
o Plane strain
o Axisymmetric
o Generalized plane strain
o 3D plane stress (membrane)
The typical 2D element (i.e., SHELL) may have the following formulations:

3-57
 o 3-node triangle (1st order)
o 6-node triangle (2nd order)
o 7-node triangle (2nd order with additional center integration point)
o 4-node quadrilateral (1st order)
o 8-node quadrilateral (2nd order)
o 9-node quadrilateral (2nd order with additional center integration point)
 For a geometry with comparable sizes in three dimensions, the “3D element” should be
adopted. Some typical 3D elements are listed as follows:
o Hex / Brick elements
 8-node (1st order)
 20-node (2nd order)
 21-node (2nd order with additional body-center integration point)
 27-node (2nd order with an additional body- and surface-center integration
points)
o Tetra elements
 4-node (1st order)
 10-node (2nd order)
 11-node (2nd order with additional body-center integration point)
o Penta / Wedge elements
 6-node (1st order)
 15-node (2nd order)
o Pyramid elements
 5-node (1st order)
 13-node (2nd order)
 14-node (2nd order with additional body-center integration point)
o Transition elements

Figure 3/3-2 Elements for Transition Zones [91]

3-58
Elements with different numbers of nodes along each geometric feature line are used for transition
zones (see Figure 3/3-2)
The choice of elements in current FEA codes, in general, is not limited to the above-mentioned
types, and there are other types available for meeting special requirements. For example, pipe
elements, spring/damper/mass elements, alignment elements, cohesive elements, and connector
elements.
3/3.2.1.1 Truss/Beam Elements
Truss elements are the simplest in form. The only physical property required is the cross-sectional
area. Beam elements, on the other hand, are considerably more complex. The various sectional
properties needed to define beam elements are discussed in the following paragraphs.
The basic sectional properties required to define beam elements are cross-sectional area, shear
areas in two orthogonal directions normal to the longitudinal axis of the element, torsional constant,
and the second moments of the area about two orthogonal axes. The axes are usually chosen to
coincide with any axes of symmetry that may exist. While this definition of beam properties is
complete for the vast majority of cases, there are circumstances in which additional factors need to
be considered.
The torsional stiffness is based on the torsional constant alone, and therefore no account is taken of
warping effects. Warping is most relevant for open sections. The error introduced by ignoring
warping is, fortunately, usually not serious because of the circumstances in which open sections are
generally used in structures. However, in situations where the main structural force acting on an
open-sectioned beam is torsion, this shortcoming should be considered in calculating rotations and
torsional stresses. Structures modeled using standard beam elements in most general-purpose FEA
software would yield incorrect results. Some FEA software does offer beam elements that account
for warping effects.
Shear flexibility is vital for deep short beams. Ignoring shear effects for this configuration would
result in an overestimate of flexural stiffness.
3/3.2.1.2 Shell Elements
The input data required for plate and shell members is thickness. Most finite element computer
programs can accommodate the non-uniform thickness and have the facility to input different
thicknesses at each node.
Type of Analysis
Linear stress field elements are the most commonly used. Almost all finite element analysis (FEA)
software have families of elements that include elements with linear stress capabilities. For many
portions of structures, a mesh of linear stress elements can provide a good description of the stress
and strain state. In areas of discontinuities, high thermal gradients, fatigue/fracture studies, or
nonlinear material problems, where there is an interest in evaluating more than just a linear stress
state, linear elements associated with a relatively fine mesh can sometimes produce reasonable
results. However, it is recommended to adopt elements with quadratic and higher-order stress fields,
which employ cubic or higher-order displacement functions. Note that for certain engineering
problems, only high order elements can capture the variations of stress or strain gradients, e.g.,
stress singularity of a crack tip [47,92]. These elements have either more nodes per element and/or
more degrees of freedom per node. These can be more expensive in terms of computational effort to
form the element stiffness matrices.
Complex structures (e.g., ship deck structure with openings) require relatively fine meshes to model
the geometrical discontinuities adequately. According to Kardestuncer [93], higher-order elements
are also practical when modeling areas of the high-stress gradient with a relatively coarse mesh.
Note that reference [93] points out that the quadratic or higher-order fit may over or underestimate
the stresses at free surfaces. The order of the stress function must match the gradient properly. The

3-59
behavior of a linear stress element is easy to visualize, which is one reason for their popularity and
simplicity. Another limitation higher-order elements suffer is the limited availability of companion
elements (i.e., meshing compatibility). Lower order element families have a complete range of
elements, and therefore it is easier to use these element types when it is necessary to mix different
elements (e.g., plates and beams).
Several recommendations [15], on the selection of elements for different analysis types, are
summarized as follows.
 Structural and fatigue analysis: Quadrilateral and Hexahedral elements are preferred over
Triangular, Tetrahedral, and Pentahedral.
 Crash and nonlinear analysis: Quadrilateral and Hexahedral are preferred over Triangular
and Tetrahedral, and priority to mesh flow lines (including topology features).
 Dynamic analysis: 2D elements are preferred over 3D. Shell elements can capture the
modes with fewer nodes and elements.
 Fracture and damage analysis: Quadrilateral and hexahedral elements are preferred over
triangular and tetrahedral elements, and high order elements are preferred over low order
elements.
 Mold flow analysis: Triangular elements are preferred over quadrilateral elements.

3/3.2.2 Structural Action to be Modelled

When a finite element model of a structure is being planned, it is necessary to have a clear concept
of the main structural actions. Each element type has limitations and is designed to model a single or
limited number of structural actions.
Before modeling a structural problem, it is useful to have a general idea of the expected behavior of
the structure. This knowledge serves as a useful guide in several modeling decisions that need to be
made in building the model. In an ideal situation, the first model will yield adequate results. However,
the first model is seldom adequate. Hence, one or more revisions will usually be necessary.
In triangulated framed structures, if the members are relatively slender, then the main action is axial
with limited bending action. In this case, the use of truss elements would be justified, and the use of
beam elements may introduce an unnecessary complication. In some instances, a mixed approach
may be appropriate. Consider a lattice mast, as shown in Figure 3/3-3. The main legs, which are
continuous, should perhaps be modeled using beam elements, whereas the bracing members would
be better modeled using truss elements.
Similarly, deck structure in ships that are subject primarily to in-plane loadsare better modeled using
plate/shell elements. In some cases, transverse shear effects may be significant. Certain element
formulations do not account for shear. Some FEA software provides plate bending elements in which
the ability to model transverse shear is optional and has to be selected by the analyst.
If through-thickness stresses are considered to be important, then the use of solid elements is
prudent.
3/3.2.3 Time Allotted for Project
In general, the selection of elements, mesh designs, and reasonable mesh quality is highly
recommended for FEA simulations. However, the project sometimes needs to be accomplished
under a very tight schedule. For such situations, it is recommended to perform the simulation work
with reasonable assumptions and approaches. For example, the following approaches may help
analysts to accomplish the project with reasonable accuracy and also in a timely manner.
 Use automatic or batch meshing tools embedded in the FEA code.

3-60
  Perform preliminary studies by adopting tetrahedral elements that are generally meshed
using the automatic meshing function with quality control criteria.
 Mesh the critical or interesting parts with finer meshing and high-order elements if the
assembly of several components is involved. Mesh the rest of the parts with either coarse
mesh or representative 1D elements.
3/3.2.4 Additional Concerns
From references [15,42,90,94–97], there are some additional useful comments and concerns
regarding selections of element types. Most of them can be categorized as empirical engineering
experiences (i.e., Rules of Thumb). They are:
 If possible, do NOT overly use triangular elements or mix triangular and quadrilateral
elements, especially close to areas of interest. The use of triangular elements can lead to a
stiffer structure compared with reality.
 If the model is meshed with either triangular or quadrilateral elements, carry out the mesh
sensitivity study based on one element type, i.e., coarse vs. refined meshed model for the
shell elements.
 After preliminary or linear FEA studies that generally are based on tetrahedral elements,
compare results from shell elements with that from tetrahedral elements. For example, the
comparisons of modal analysis results in terms of eigenvalues and hot spots can properly
indicate the qualities of mesh design and mesh sensitivity.
 Whenever possible, control the element size consistently in the global model or at least in
the areas of interest.
 It is recommended to add a layer of ultra-thin shell element on top of the 3D elements in
areas of discontinuity for mesh transitions and fatigue analyses. This procedure will capture
the high stress and strain states on the skin by eliminating solid element smoothing to its free
surface.
 In an assembly model, it is acceptable to have a combination of different mesh types (1D,
2D, and 3D). However, the analyst should first understand all the assumptions and
implications of the problem.
3/3.2.5 Miscellaneous One-Dimensional Element Comments
From the references of NASTRAN [95] and OPTISTRUCT [98], the 1D element types are
summarized as follows. It should be noted that other FEA codes (e.g., LS-DYNA) have different 1D
element types, and the analyst should choose the 1D element type after checking the corresponding
references. In general, there are two types of 1D elements: rigid-body elements and physical
elements.
Rigid-body elements have the following characteristics:
 They do not need material properties and cross-section profiles.
 One master node drives multiple slave nodes, i.e., RBE2 [95]. It is used as a shortcut to
model another body or control multiple nodes through one node.
 One master node transfers distributed loading (e.g., force or displacement) to multiple slave
nodes, i.e., RBE3 [95]. This type of element can simplify the modeling process and mimic the
distributed loading.
Physical elements have the following characteristics:
 They need material properties and cross-section profiles.

3-61
  Several widely used 1D physical elements are listed as follows:
o Spring elements (i.e., CELAS1 [95])
o Bushing elements (i.e., CBUSH [95])
o Rod elements (i.e., CROD [95])
o Bar elements (i.e., CBAR [95])
o Beam elements (i.e., CBEAM [95])
o Pipe elements (i.e., CPIPE [95])
By properly using the 1D element, the simulation time can be reduced significantly. However, extra
care is needed in model development, and the analyst needs to fully understand the geometries,
meshing approaches, and structural responses.
3/3.3 Mesh Design
Mesh design, the discretization of a structure into a number of finite elements, is one of the most
critical tasks in finite element modeling. The following parameters need to be considered in
designing the layout of elements: (1) mesh size & mesh density, (2) mesh transitions, (3) stiffness
ratio of adjacent elements, (4) joint modeling, (5) advanced analysis mesh parameters, (6) element
shape limitations, (7) mesh quality control, and (8) numbering. This subsection provides tips on
these aspects of mesh design.
It should be noted that most of the FEA codes nowadays are using two different algorithms in
designing mesh: geometry-based and FE-based. The former type of mesh design (geometry-based)
is associated with the CAD geometry, that means the mesh design will be updated (sometimes
automatically) as the input FE model geometry is modified. Some of the FEA codes have the feature
to control or lock the mesh design when the CAD geometry is updated. The latter type of mesh
design (FE-based) is independent of the geometry. Note that some of the state-of-the-art techniques
can generate geometry in a mesh-based format, for example, laser scanning, 3D printing, and DIC
(digital image correlation). It is recommended to use the geometry-based means to design the
meshed model.

3-62
 Figure 3/3-3 Typical Lattice Structure

3/3.3.1 Mesh Size & Mesh Density

The accuracy of the FEA results in local areas within an FEA model is dependent on the finite
element size. An engineer may create an FE model with a finer mesh (smaller element size) to
generate highly accurate local results, but this will require more time for modeling and computing.
Alternatively, an engineer may use his engineering judgment to create an FE model with coarser
mesh (larger element size) that excludes details that are judged to be unnecessary. For example,
lightening holes and cutouts are usually not modeled in a global full-ship analysis. These coarse
mesh models may lead to less accurate results in the areas of these holes, but the engineering
decision saves time on modeling, computing, and postprocessing. In generating FEA models, the
foremost problem is to choose appropriate element size based on the analyst’s understanding of
what features are important or unimportant such that the created models will yield accurate FEA
results for the specified analysis while saving as much modeling, computing and postprocessing
time as possible.
As a general rule, a finer mesh is required in areas of high stress or strain gradient. It is possible, of
course, to use a fine mesh over the whole model, but this cost more in terms of computational power
and time. This is undesirable on three counts: (1) economy, (2) the greater potential for manipulation
errors, and (3) accumulated errors caused by an overly fine-meshed model. Hence, the meshes of
variable density are usually used. Care is required in transitioning of mesh density. Abrupt
transitioning introduces errors of a numerical nature.
The size of the mesh depends upon several factors, such as the type of analysis. The factors are
briefly summarized as follows:
 Previous experience - At first, it is upon the analyst to decide the right element size. The
analyst’s experience plays an important role in striking a balance between economy and
adequate mesh size. Analysis of similar structures under similar loading conditions in the
past can help in the identification of stress concentrations and regions of rapid changes in
stress patterns. In addition, the experience from experimental observations may also help in
detecting sensitive areas. This issue is extremely important in the case of novel designs.
 Analysis type - For the linear elastic problems, it is relatively easier to solve the FE model
with a large number of elements and nodes (i.e., large degree of freedom – DOF). However,
for nonlinear problems, it is more difficult to solve and requires pronounced iterations to
obtain converged results. In addition, it may depend upon the type of FEA solver (e.g.,
implicit vs. explicit) and formulations (e.g., small strain vs. finite strain) adopted.
 Hardware and software limits - One of the major bottlenecks in numerical simulations is the
hardware and/or software, e.g., CPU/GPU architecture, CPU/GPU clock speed, and I/O
bandwidth. The analyst should understand the upper bound of DOF numbers that the
hardware and software can handle properly. In addition, the analyst should also consider the
available time (i.e., cycle time and total project time), which can be impacted by the DOF
numbers.
 Interested FEA results (e.g., stress, strain, fracture driving force, etc.) - As the interested
FEA output may vary in different analysis types, it is recommended to choose the mesh size
based upon the minimum requirement. For example, if the deflection of a beam or column is
of interest, it is recommended to use 1D element instead of 2D or 3D elements; if the
eigenvalues of a structure are of interest, it is suggested to adopt high order 2D or 3D
elements with fewer DOF numbers instead of 1st order 2D or 3D elements with more DOF
numbers. However, the analyst should know that for projects with special requirements (e.g.,
fracture and damage), the mesh size is strongly dependent on the problem investigated and
sometimes has a physical meaning (e.g., comparable with grain sizes in the micro- or
mesoscale). [31,47,99–103]

3-63
  Sensitivity of results - Compared with the global response (e.g., displacement/ deflection),
the local response (e.g., stress and strain states) can be extremely sensitive to the mesh
size and quality, especially around the hot spots. The analyst should use his/her engineering
sense to predict the high stress/strain areas that may be prone to trouble and refine the
mesh. Some typical geometric features with high stress/strain concentrations are listed as
follows:
o Profiles with sharp radius, for example:
 Holes
 Threats
 Fillets
 Chamfers
 Flaws/defects or 2nd phase particle
o Contact surfaces
o Loading or fixing areas/points
o Sudden cross-section changes along the load paths
o Abrupt material changes
o Crack or damage paths
The mesh density directly depends upon the mesh size. The basic rule is that the mesh is refined
most in the regions of hot spots with steep stress or strain gradients. Therefore, if such regions can
be identified during mesh design, the probability of developing an economical mesh with sufficient
refinement is high.
In cases where the experience of a particular configuration is lacking and where it is difficult to
anticipate the nature of the stress gradients, an iterative approach is necessary. Where stress or
strain values show a sharp variation between adjacent elements, the mesh should be refined, and
the analysis rerun. As a general practice, it is recommended to refine the meshes around the hot
spots (e.g., areas with high stress/strain gradient) and resolve again after the preliminary run. This
process may take several iterations until the difference of stress/strain values meets the
convergence criteria (e.g., 5 – 10% in Ref. [15]).
3/3.3.1.1 Meshing Application
Once the element size and element density are defined, the analyst needs to mesh the geometry
using the pre-processing tools of the selected FEA software. Essentially, there are three different
means to mesh a model: (1) automatic meshing with predefined parameters, (2) mapped meshing
with more predefined parameters and pre-trimmed partitions, and (3) manual meshing section by
section.
Automatic meshing requires minimum user involvement but generates more elements and nodes
that the other meshing means. A clean CAD geometry is needed to carry out automatic meshing
accurately.
Mapped meshing requires more time and user involvement compared with automatic meshing, but
mesh quality can be significantly improved, especially around the areas of interest. This technique
also needs clean CAD geometry as the basic input.
Manual meshing requires the most time and user involvement but allows the analysis of full control
over the mesh design and quality. This technique can be independent of geometry and results in the
least elements and nodes.
The analyst has to specify all the parameters (e.g., minimum hole diameter, minimum fillet radius,

3-64
average and minimum element length, quality parameters, etc.). The selected parameters are
presented in 3/3.3.7, and the recommended values are shown in Table 3/3-1 General Problems with
Average Element Size of 50mm Table 3/3-1, Table 3/3-2, Table 3/3-3, Table 3/3-4, and Table 3/3-5.
The software can generate meshing automatically after several iterations by fulfilling all or most of
the specified instructions. However, the analyst still needs to review the mesh quality and
revise/refine areas that violate the criteria.
3/3.3.2 Mesh Transitions
If the mesh is graded, rather than uniform, as is usually the case, the grading should be done in a
way that minimizes the difference in size between adjacent elements. Figure 3/3-4 presents several
examples of transitions using quadrilateral elements. These examples attempt to keep within the
guidelines for element distortion discussed in Section 3/3.3.6.

Figure 3/3-4 Transitions from Coarse to Fine Meshes

Another way of viewing good transitioning practice is to minimize large differences in stiffness
between adjacent elements. A useful measure of stiffness is the ratio E/Ve, where E and Ve
represent the elastic modulus and the element volume, respectively. As a working rule, the ratios of
E/Ve for adjacent elements should not change by more than a factor of two [104].
Sometimes transitions are more easily achieved using triangular elements. Transitions of this type
are illustrated in Figure 3/3-5. Most FEA programs will allow two nodes of a quadrilateral element to
be defined as a single node in order to collapse the element to a triangular shape.

Figure 3/3-5 Transitions Using Triangular Elements

Figure 3/3-6 presents a summary of using either triangular or quadrilateral elements in designing
mesh transitions [15]. These six recommended types can be used as a basis to design mesh without

3-65
violating the element shape limitation criteria. The mesh transitions of 3D elements can be designed
in a similar fashion.
1 to 3 using triangular elements 2 to 4 using quadrilateral elements

1 to 3 using quadrilateral elements 1 to 2 using triangular elements

2 to 4 using triangular elements 1 to 2 using quadrilateral elements

Figure 3/3-6 Six Element Transitions

In modern FEA codes, most analysts rely on preprocessors to develop the finite element mesh. In
general, automatic mesh generators yield adequate meshes. However, in very demanding
configurations, the mesh generator may produce a poor mesh. In such situations, the mesh should
be manually improved to meet the guidelines. Note that some preprocessors have the feature that
can highlight the poor mesh areas.
In regular rectangular meshes, there are two basic types of transition. One is the change in element
density in the direction of the stress gradient. The second is transverse transitioning, which is used
between areas with different element sizes and densities across a transverse plane, as shown in
Figure 3/3-7.

Figure 3/3-7 Mesh Transitions

Many rules of thumb for transitioning of elements are based on element strain energy and strain-
energy density calculations. The ideal finite element model should have a mesh with constant strain
energy in each element. To achieve constant strain energy of elements, the volumes must be
relatively small in regions of high stress or strain and large in regions of low stress or strain.
Transverse transition regions should be used only in areas of low-stress gradient and never near
regions of maximum stress or deflection.
Improper connections between elements of different types can cause errors. Solid element types, for
example, have only translational nodal degrees of freedom. If solid elements are interconnected with
beam or plate/shell type elements, which have rotational degrees of freedom, in addition to

3-66
translational ones, care must be taken to allow for the transfer of moments if that is what is intended.
If this is the case, then it is best accomplished with linear constraints or multipoint constraints. In
case the program does not offer such options, the beam (or plate) can be artificially extended
through the solid elements. Figure 3/3-8 illustrates the problem and a solution for the sample
problem.

Figure 3/3-8 Connecting Elements with Different Nodal Degrees of Freedom

Most flat plate/shell element formulations do not have a shape function for the rotational degree of
freedom about a normal to the surface of the element. Hence, in- plane rotational stiffness is not
modeled. Some programs provide a nominal rotational stiffness to prevent free rotation at the node.
Other programs use certain formulations to improve this aspect of performance but at the cost of the
presence of spurious modes. The user should be aware of the possible limitations in the program
that is being used when modeling situations in which moments are to be transferred into the plane of
assemblages of flat plate/shell elements. The problem and one possible solution are illustrated in
Figure 3/3-9.

Figure 3/3-9 Modeling In-Plane Rotational Stiffness of Membrane Elements

3-67
3/3.3.3 Stiffness Ratio of Adjacent Structure
In modeling complex structural assemblies, there is a possibility of constructing models where
adjacent structural elements have very different stiffnesses. These types of stiffness combinations
can cause ill-conditioning of the equilibrium equations, which can severely degrade results. The
transitioning guidance given above avoids this problem in models that use two or three-dimensional
elements. For truss and frame structures, a different approach is required. To prevent large
numerical errors in these cases, stiffness ratios of the order of 104 and more between members
making up a model should be avoided. This is admittedly a conservative number. More realistic
guidance can be obtained by undertaking tests.
The problem of stiffness mismatch is most severe in structures where a relatively rigid portion of the
structure is supported on the flexible structure. In such cases, the deflections in the rigid portion are
due more to rigid-body movement rather than elastic distortion. In these cases, it is suggested that
the stiff portion be treated explicitly as a rigid body using rigid links, rigid regions, constraints, or
combinations of these approaches.
3/3.3.4 Joint Modeling
Modeling of joints (e.g., fasteners and welds) in FEA simulation is a debatable procedure. In general,
joints can exhibit highly nonlinear behaviors. Different design codes recommend different means to
mimic the joint in the model. The analyst needs to decide which modeling approach should be
adopted for different types of joints. The following is a discussion of a few key topics in joint
modeling.
3/3.3.4.1 Mechanical joints
The mechanical joints are one of the most widely used in assemblies. They include fasteners such
as bolts and rivets. In FEA modeling, mechanical joints are assumed to carry the full shear load
between the surfaces as long as it is under the peak load. In some FEA codes, the element type for
mechanical joints may also include other parameters such as tensile or shear failure strength. Note
that it is needed for the analyst to understand the basic engineering principles and verify the input
variables manually. The tensile or shear loads can be included in the FEA output by adopting a
cross-sectional evaluation feature [28,105].
3/3.3.4.2 Welds and adhesive joints
Welds are widely used to connect sheet metal components (e.g., spot or seam welds).
Rigid or beam elements can be used to model the spot welds. The former is more straightforward in
modeling but may cause the singularity (ill condition) of the stiffness matrix, due to the lack of drilling
degrees of freedom. The latter naturally eliminates the singularity and can record the variations of
forces by assigning length, material properties, and cross-section profiles.
Rigid, beam, or shell elements can be used to model seam welds. If the rigid or beam elements are
used, there is no degree of freedom between neighboring welds. The number of rigid or beam
elements needs to be calibrated based upon the test data with the same geometric profiles and
weldment properties. If the shell elements are used, extra cares are needed to make the modeling
meet meshing quality criteria. Note that sometimes it is difficult to capture the peak stress caused by
irregular geometry of the actual welds. The nominal values can be adopted to compare with the
testing data for validations. If the sharp notch is presented, it is recommended to carry out the
structural integrity analysis associated with complicated meshed welds.
3/3.3.4.3 Pre-loaded joint
In general, joints are pre-loaded either from the mechanical assembly or residual stress/strain. There
is a need to properly account the pre-load in a joint. In a common way, it requires a detail solid
model in the joint area to represent the joint. However, in the recent development of FEA codes,
there are novel element types that allow the analyst to adopt that can consider pre-loads from a

3-68
thermal shrink or bolt pretension on a segment or cross-section of the fastener [28]. The magnitude
of thermal or pretension can be adjusted to achieve the desired pre-load after the preliminary run. It
should be noted that in some nonlinear analysis or fatigue analysis, the detail solid model is required
aiming to capture the high stress or strain gradient.
3/3.3.5 Advanced Analysis Mesh Parameters
3/3.3.5.1 Impact/Collision Problems
In the studies of impact or collision problems, large deformations and/or failures may be presented,
and therefore the mesh design should be adequately detailed to capture them. In general, the
aspects of the mesh design for these problems are consistent with those described in the previous
sections. However, extra cares are needed for specific problems. For example,
1. For ductility evaluations, several elements should be modeled in the yield zone to capture
the strain gradient [106];
2. For stability evaluations, minimum 3 to 6 elements per expected half wave are required to
capture the buckling modes [106];
3. For the application of the automatic mesh refinement, the mesh design and element type
should be compatible with the algorithm of mesh refinement;
4. For failure analyses, the mesh design should be refined along with the expected failure or
crack paths, and the element type should be compatible with the material models associated
with damage formulations;
For the mesh refinement study, the analyst needs to verify that the mesh design is adequate for
representing all the failure modes except the above-mentioned aspects. In Ref. [40,106,107], it is
recommended to rerun the analysis with half the element size and check the difference of results.
The identical results indicate a reasonable convergence of the mesh design.
3/3.3.5.2 Fracture Analysis
Different from the above-mentioned engineering problems, FEA studies on fracture and damage
mechanics use unconventional continuum mechanics with singularities and discontinuities included.
In these problems where cracks or flaws are included, it is required to have a relatively more
profound engineering knowledge and analyst intervention regarding the mesh design.
This section reviews some techniques in constructing the mesh for crack-related analysis. However,
it should be noted that such kind of analysis sometimes applies user-defined materials and user-
defined elements. It is challenging to cover all the aspects that the analyst may face, and it is
recommended to consult the published literature that covers numerous theoretical knowledge and
practical aspects with respect to mesh design for crack problems [47,90,96,99,101,108].
In general, there are two different sets of modeling techniques regarding the presence of crack
propagating, i.e., stationary cracks and propagating cracks. When modeling the crack faces (a pair),
the nodes along each crack face are usually matched or coincide in terms of coordinates, see Figure
3/3-10(a). Note that some FEA preprocessors may merge these coincide nodes automatically, and
one technique to avoid such a limit is to model a small gap between faces, see Figure 3/3-10(b).
Figure 3/3-14 schematically shows an example of the half single-edge notched tension (SENT)
specimen under tension, where two crack faces with a gap of 25 micros are modeled [109]. It should
be noted that if the FEA output is related to crack mouth or crack tip opening indexes, the analyst
should use the relative values instead of the absolute values.

3-69
 Figure 3/3-10 Meshes for the Analysis of Blunting and Local Crack Opening Criteria

In the stationary crack problems, the crack driving forces (e.g., stress intensity factor – K and energy
release rate – G, J, CTOA, CTOD) are of interest for the subsequent engineering analysis, e.g.,
fatigue life and structural integrity assessment [12,63,110–116]. When modeling the vicinity of the
crack tip, it is recommended to use either quadrilateral or brick elements for 2D and 3D problems,
respectively, [47]. At the crack-tip, the quadrilateral or brick elements are usually collapsed and
degraded to triangle or wedge, respectively (see 3-color highlighted circles in Figure 3/3-11).

Figure 3/3-11 Nodes of Quadrilateral Element Collapse Coincident to Model Crack Tip [47]

3-70
 Figure 3/3-12 Crack Tip Elements

In elastic problems, the high order element should be adopted, and the nodes of the crack front are
moved to the ¼ points (25% of element length from the crack tip), as shown in Figure 3/3-12(a).
Such modification of elements can mimic the singularity strain field in the vicinity of the crack tip (i.e.,
f(1/sqrt(r))) and enhance the numerical accuracy. On the other hand, if plastic zones are formed in
the vicinity of the crack tip, the singularity of the strain field becomes f(1/r), and there is no need to
adjust the location of mid-side nodes (see Figure 3/3-12(b)). Note that in [47], the author pointed out
that the side benefit of plastic singular element design is that it can assist in capturing the crack-tip
opening displacement by computing the intersection of 90 vertices with crack faces. Figure 3/3-13
schematically shows the CTOD calculation.

Figure 3/3-13 Crack Tip Opening Displacement (CTOD) At the Intersection of 90° vertex
with Crack Faces

The most efficient mesh design for the crack tip region is the “Spider Web” configuration (see Figure
3/3-14 – SENT), which consists of concentric rings to quadrilateral elements focusing toward to the
crack tip. Because the stress and strain field of crack tip regions vary drastically, it is recommended
to use the finest mesh close to the crack tip and gradually decrease the mesh density towards the
boundary of the spider-web. Note that the analyst should carry out a sensitivity study on the mesh
design and verify the FEA output by comparing it with the benchmark examples.

3-71
Figure 3/3-14 Blunt Crack Face of A Single-Edge Notched Tension (SENT) Specimen with
Spider-Web Meshes [117]

In general, the variable of elastic problems (e.g., K and G) can be accurately captured based upon
relatively coarser mesh design. It is mainly because of the use of advanced numerical approaches,
such as contour integral and virtual crack extension methods, which is a domain-based approach
and eliminates the requirements of solving the local crack-tip field. On the other hand, the variables
of elastic-plastic problems (e.g., J and CTOD) can be mesh depended, and extra care is needed.
First, the mesh design must be sufficiently refined in the vicinity of the crack tip, and proper strain
formulations (i.e., small or finite/large strain) should be adopted, aiming to capture the stress, strain,
and deformation with the presence of the large or full yielding. The analyst needs to have strong
engineering knowledge regarding the theoretical definitions of the variables investigated and
understand the limits of FEA codes adopted. For example:
1. The J-integral is essentially a far-field term to capture the energy release rate. The J values
should be estimated from rings far from the first ring, and those from far-field rings are also
independent of strain formulations adopted. Figure 3/3-15 shows the variations of J results
as the increase of loading, where J is evaluated from different rings and different strain
formulations [118]. In the figure, J values from ring No.20 are almost identical to the
benchmark values (e solid line) calculated from the ASTM standard [53].
2. CTOD is strongly depended on the mesh design and strain formulations because of its
definitions (e.g., 90° intersection and GKSS – CTOD5). Aiming to capture the deformation of
crack faces, such as crack blunting, the finite/large strain formulations have to be employed.

3-72
 Figure 3/3-15 J-integral vs. Load Line Displacement for a C(T) Specimen [118]

In the propagating crack problems, the crack path(s) and failure mechanism(s) are examined. Note
that as discussed in Section 3/2.6, there are different variables to quantify the critical value of crack
initiation and propagation. If the critical value is reached or exceeded, a crack starts to grow. There
are generally three different techniques to simulate the crack growth, and the mesh should be
designed accordingly [119]. These techniques are:
 Node release method based on crack driving forces (e.g., Jc, Gc, CTODc, and CTOAc)
[101,120–122]. Figure 3/3-16 schematically shows this concept and mesh design
[68,123,124].
 Cohesive element method [125–127]. Figure 3/3-17 shows the mesh design [96,101,128].
 Constitutive equations based on damage mechanics concepts [129–131].
Note that in such simulations, user-defined elements, and user-defined materials may be employed.
It is recommended to design the mesh as well as to control the mesh quality by the following
guideline in the FEA code vendor documentation.

3-73
 Figure 3/3-16 Spring Elements [123]

Figure 3/3-17 Cohesive Element [96]

3/3.3.5.3 Fatigue Analysis

In the fatigue analysis of a ship structure, there are three levels of FE models:
1. Global model
2. Partial ship model
3. Local structure model
The mesh design of the local structure model is discussed in the remainder of this section.

3-74
The local structure models are used for the investigation of hot spots with high stress or strain
gradients, e.g., weldment and free plate edges. The typical geometric details considered in the
standards [9,132–135] are as follows:
1. Bracket toe and flange terminations of stiffeners
2. Slots and lugs in the web frames at the intersection with stiffeners
3. Bracket and flange terminations of girder systems
4. Panel knuckles
It is recommended to evaluate the hot spots based on the following rules:
1. Refined meshing with an element size of t × t or 10 × 10 mm, where t refers to plate
thickness
2. Minimum 10 elements in all directions from the refined (hot spot) zone to the coarse zone
3. Mesh transition from refined zone to the coarse zone should be done gradually per Section
3.3.4.
A typical example of mesh transition is shown in Figure 3/3-18 [132]. In the figure, an FE model of
longitudinal hatch coaming end bracket to deck plating in the hot zone is presented, where the hot
spot is designed with t × t mesh to capture the high-stress gradient.

Figure 3/3-18 Mesh Transition Example from DNVGL-CG-0129 [133]

If the stress concentration factors are required in the fatigue life calculation per [134], it is
recommended to carry out two levels of analyses:
1. A 50 × 50 mm meshed model to evaluate the semi-nominal stress.
2. A t × t meshed model to capture the stress gradient at the hot spot.
Note that the purpose of evaluating the semi-nominal stress (1) is to determine the local geometric
stress flow and effect of cut-outs, web frame-toes, and tripping brackets [134].
An example of a meshed model is shown in Figure 3/3-19 [134]. In Figure 3/3-19 (b), the hot spots
are modeled with the mesh density of t × t and web-mesh with four elements in all directions at all
relevant hot spots. Figure 3/3-20 shows the stress distributions of the two levels of the meshed
model. From Figure 3/3-20, it is clearly seen that the stress variations in the hot spots can only be

3-75
accurately captured by using the t × t model, where the fine mesh design with 4 layers of elements is
used.

Figure 3/3-19 Two Levels of Meshed Models for Stress Concentration Analysis 50 x 50
mm (b) t x t refinement [134]

Figure 3/3-20 Stress Distributions of Two Levels of Meshed Models

for Stiffener and Lug Details (a) 50 x 50 mm (b) t x t refinement [134]

The analyst may also adopt the submodeling technique to accomplish such kind of multi-level fatigue
analyses. Note that it is an approximate solution by assuming that the displacement values from the
global model are sufficiently accurate and can be used for the local model.
3/3.3.5.4 Whole Ship
The level of discretization of the global ship model depends on the complexity and size of the model.
Ship structure is generally modeled by assembling structural elements such as stiffened panels and
stiffeners. A stiffened plate is composed of a plate and its stiffeners, considered as secondary
stiffeners. Traditionally in whole-ship mesh design, the nomenclature of the “coarse mesh model”
and “finer mesh model” is separated by how a stiffened panel is modeled, as given in [136]. A coarse
mesh model referred to an orthotropic plate, equivalent plate, or lumped stiffener model, where
stiffeners are integrated into a plate element. A finer mesh model is a model where stiffeners are

3-76
explicitly modeled as beam elements at their exact geometric location. It is noted that some
documents (for example, NSWCCD’s “Modeling Guidelines Full-ship FEM Development” [137] and
ABS’s Rules for Building and Classing Marine Vessels [64]) referred to the traditional “finer mesh
model” as coarse mesh model.
In a whole-ship coarse mesh model, the most common and easiest way to represent the equivalent
plate element is by lumping the plate internal stiffeners to plate element edges. In the lumping
process, shell plates are modeled as plate elements, and all secondary stiffeners are modeled as
beam elements. The stiffeners located inside the plate element are put at the edges of the plate
element to reduce the number of nodes, as illustrated in [136] and shown in Figure 3/3-21.

Figure 3/3-21 Coarse Mesh Stiffener Lumping

Also, in a whole-ship coarse mesh model, each stiffener causes an increase in the cross-sectional
properties of the lumped beam elements. Since the stiffeners’ cross-sectional area and direction are
preserved, the actual hull girder cross-sectional properties are well represented. Therefore, the hull
girder global response, especially under sagging and hogging conditions, should be close to the
response of a finer mesh model. However, it is difficult to post-process the limit states of plates and
stiffeners for a lumped model because many of the plate-stiffener geometry properties, such as the
space between the stiffeners, are lost during the lumping process. In addition, accurately counting
the number of stiffeners in a plate element and properly splitting the stiffeners to the edges of the
plate element are also tedious and error-prone. Finally, it is difficult to review and check the lumped
model graphically. Hughes [66] suggested the orthotropic plate technique, where the stiffeners are
blended with the plating so that the plate has different stress-strain properties in two directions.
Since the stiffened panel is represented as one orthotropic plate, the approach does not allow
analyzing plate and stiffener separately, and the obtained normal stress is the average of the plate
and the stiffener. Avi et al. [138] proposed a three-layer laminate element to model an equivalent
stiffened panel, where the first layer represents the plate, the second layer the stiffener web, and the
third the stiffener flange. Both the laminate equivalent plate and orthotropic plate method address
most deficiencies of the lumped stiffener method and are more accurate. However, these methods
are not readily available in general finite element packages such as Nastran and ANSYS.
A whole-ship finer mesh model is a model where stiffeners are explicitly modeled as beam elements
at their exact geometric location. Transverse frames and longitudinal girders are generally modeled
as beams, but in the case of larger beams such as web frames or longitudinal girders in the way of
propulsion foundations, they might be modeled with their webs as quad elements and their flanges
as either quads or flat bar beams. The longitudinal mesh is generally 2-4 elements between
transverse frames. The finer mesh model approach is well suited to today’s general tendency in both

3-77
ship design offices and shipyards, which is to develop a model for the unified design process and to
share the model information for numerous purposes. A finer mesh model provides higher fidelity for
a full-ship structural analysis. However, it takes more resources to model, analyze, and post-
process.
3/3.3.5.5 Vibration
The recommended model mesh size is dependent on the type of analysis being performed and
the size of the ship. Hull-Girder vibration analysis on a large ship can have a mesh size of 3 to 4
stiffener spacings per ABS in [14]. However, a hull girder vibration analysis on a much smaller
54 ft catamaran is recommended to have a mesh size of 4 to 8 elements between each stiffener
per SSC in [139].
A local model should have a finer mesh. As a general rule, the element size should be
sufficiently small to capture the mode shapes expected within the structure.
The curvatures in structural mode shapes in higher modes are more severe than at lower modes,
and several nodes (and associated masses) are required to represent the kinetic energy accurately
at higher modes. The mesh density of a vibration analysis may require some iteration as natural
frequencies are determined, and then the mesh density refined to make sure the structural response
is being modeled accurately. Figure 3/3-22 illustrates the difference between a good mesh density
and a poor mesh density for a given mode shape.

Figure 3/3-22 Assessing Accuracy of Higher Modes

3/3.3.6 Element Shape Limitations

The following is a discussion of element shape aspect ratio, skewness, and warpage.
3/3.3.6.1 Aspect Ratio
The element aspect ratio in a 2D element or a face of a 3D element is the ratio between the longest
and shortest element dimensions, as shown in Figure 3/3-23.

3-78
 Figure 3/3-23 Aspect Ratio of Plane Elements

A crude rule of thumb that can be used is to limit the aspect ratio of the membrane and bending
elements to three for good stress results, and to five for good displacement results. The ideal shape
for quadrilateral elements is square and equilateral for triangular elements. Hence, the use of ideally
shaped elements is particularly desirable in areas of high-stress gradients. In general, higher-order
elements are less sensitive to departures from the ideal aspect ratio than lower-order elements. This
observation also applies to solid elements.
Since an element's sensitivity to aspect ratio is dependent upon both element formulation and the
nature of the problem, general tests and problem-dependent checks may be justified in cases where
element performance is not well known.
3/3.3.6.2 Skewness
Generally, the performance of elements degrades as they become more skewed. Skewing is defined
as the deviation of vertex angles (𝛽) from 90° for quadrilateral shaped elements, and from 60° for
triangularly shaped elements, as shown in Figure 3/3-24. For quadrilateral elements, angles greater
than 135° and smaller than 45° are not recommended, i.e., 45° ≤ 𝛽 ≤ 135°. The limiting range
recommended for triangular elements is 45° and 90°, i.e., 45° ≤ 𝛽 ≤ 90°. Skewed quadrilateral
elements shaped more like parallelograms generally perform better than more irregularly shaped
ones. The check of skewing angels on all faces of a 3D element is performed in the same fashion.
3/3.3.6.3 Warpage
When element nodes are not in the same plane, the element is warped, as shown in Figure 3/3-24.
This is undesirable, and the degree to which this impairs the performance of plate elements depends
on the element formulation. The warpage in a 2D element can be calculated by splitting a
quadrilateral element into two triangular elements and evaluating the angle between two planes of
triangular elements, where the maximum angle refers to the warpage of an element. For 3D
elements, the warpage value on all faces of an element is evaluated in the same fashion. The best
guidance in regard to limiting levels of warping is contained in the particular FEA program's user
manual. The use of triangular elements is an option where the curvature of the structure is high.

3-79
 Figure 3/3-24 Element Shape Limitations

3/3.3.7 Mesh Quality Control

The mesh quality is one of the critical subjects in FEA simulations, which directly affects the results
of the calculations. For example, a model with poor element quality may need unexpected iterations
and time to obtain converged results. This is because, as discussed in Section 3/3.6.2.3, the time-
step is directly calculated from the minimum length of elements in an explicit time-integration based
analysis. Although there are typically discussions about this in FEA training and lectures, it is
typically not fully understood by junior engineers. That is mainly because the element quality is a
complex topic, and in nature, it is a ‘relative’ approximation that needs analyst’s engineering
judgments. Therefore, it is recommended to have in-house validated mesh quality criteria that can
meet the requirements of different analysis problems. The analyst should always check the mesh
quality in the meshing process. The key aspects related to the mesh quality and the corresponding
definitions are summarized as follows:
1. Penalty value - A weight factor to evaluate whether the element meets the criteria.
2. Max / Min length - Maximum and minimum dimension of the element. Note that the minimum
length of the element is crucial to determine the time-step using the explicit solver.
3. Max / Min Angle Quadrilateral / Triangular - Individual maximum and minimum angles of the
element.
4. Aspect Ratio - See Section 3/3.3.6.
5. Warpage Angle - See Section 3/3.3.6.
6. Skewness -See Section 3/3.3.6.
7. Jacobian Ratio - A scale factor to describe the transformation of the coordinate system.
8. Distortion - Evaluate the element topology from determining of Jacobian Ratio and areas in
local/global coordinates.
9. Stretch - A ratio to describe the maximum topology dimension in terms of the minimum
topology dimension.
10. Chordal Deviation - Determine how well curvatures have been modeled.
11. % of trials - Percentage of triangular elements in the mesh design.
12. Taper - Describe the angular distortion of a quadrilateral shell element.
13. Tetra collapse - The normalized height of a tetrahedron element.
14. Volumetric skew (for tetrahedron elements) - The normalized difference of volumes between
actual and ideal tetrahedron elements for a given spherical domain passing all corner nodes.

3-80
In modern FEA codes, the pre-processor may include some or all of the above-mentioned aspects.
In addition, the pre-processor may also provide a series of values for each aspect and divided them
into four or five different commentary categories, e.g. (1) Ideal, (2) Good, (3) Warn, (4) Fail, and (5)
Worst. A set of criteria examples for different analysis problem has been presented by Altair
HyperMesh [15] and shown as follows:
1. For general problems with an average element size of 50 mm, the recommended criteria are
shown in Table 3/3-1.
2. For vibration problems with an average element size of 50 mm, the recommended criteria
are shown in Table 3/3-2.
3. For fatigue problems with an average element size of 50 mm, the recommended criteria are
shown in Table 3/3-3.
4. For plasticity and crash problems with an average element size of 50 and 100 mm, the
recommended criteria are shown in Table 3/3-4 and Table 3/3-5, respectively.
Table 3/3-1 General Problems with Average Element Size of 50mm

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 50 41.65 33.2875 24.9375 22.85
2 max length 50 66.0625 82.125 87.5 125
3 aspect ratio 1 2 4 5 10
4 warpage 0 5 15 20 40
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 4 6 8 12
13 taper 0 0.2 0.5 0.6 0.9

3-81
 Table 3/3-2 Vibration Problems with Average Element Size of 50mm

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 50 38.335 26.665 15 12.085
2 max length 50 62 73 90 113
3 aspect ratio 1 2 4 5 10
4 warpage 0 4.5 13.5 18 36
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 4 6 8 12
13 taper 0 0.2 0.5 0.6 0.9

Table 3/3-3 Fatigue Problems with Average Element Size of 50mm

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 50 38.5 26.7 15 12.1
2 max length 50 65 79 100 128.6
3 aspect ratio 1 2 4 5 10
4 warpage 0 3.75 11.25 15 30
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 1.5 2.25 3 4.5
13 taper 0 0.2 0.5 0.6 0.9

3-82
 Table 3/3-4 Plasticity and Crash Problems with Average Element Size of 50mm

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 50 43.33 36.66 30 28.33
2 max length 50 62 73.2 90 112.9
3 aspect ratio 1 2 4 5 10
4 warpage 0 5 15 20 40
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 3 4.5 6 9
13 taper 0 0.2 0.5 0.6 0.9

Table 3/3-5 Plasticity and Crash Problems with Average Element Size of 100 mm

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 100 80 60 40 35
2 max length 100 124 146 180 226
3 aspect ratio 1 2 4 5 10
4 warpage 0 5 15 20 40
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 3 4.5 6 9
13 taper 0 0.2 0.5 0.6 0.9
In the above five tables, five columns represent the boundary of ideal, good, warn, fail, and worst,
respectively.
1. The ideal set of mesh quality is the optimal value of an element that can achieve.
2. The good set of mesh quality is slightly worse than the ideal set, but it can still progenerate
reliable FEA results.
3. The warn set is the intermediate level of mesh quality and should be regarded as the
baseline. Extra cares are required from the analyst, especially close to the area of interest.

3-83
 4. The fail set indicates that the current mesh design and quality are unacceptable for the
analysis. There is a need to fix these elements before performing any analysis.
5. The worst set denotes the elements that fail the criteria by a large margin. They have to be
fixed immediately.
It should be noted that the criteria presented in Table 3/3-1, Table 3/3-2, Table 3/3-3, Table 3/3-4,
and Table 3/3-5 can be used as a reference to prepare criteria for analysis problems with other
element sizes. Except the key aspects mentioned above, there are still other factors may drive the
mesh quality, such as the all the geometric issues (mentioned in Section 3/3.1), the presence of
duplicate elements / nodes, inconsistent element normal directions, the presence of temporary / free
nodes, the boundary conditions (e.g. contact pair faces, loading node(s), and fixture node(s)). In
addition, it is also recommended to review the mass as well as the center of gravity of the FEA
model.
The modern FEA pre-processors may provide advanced features to visually assist analysts in
checking the mesh qualities. For example, different contour colors may represent a different level of
mesh qualities, e.g., red depicting worst [15] (see Figure 3/3-25).

Figure 3/3-25 Reviewing Mesh Quality

(a) Contour Plot (for example, red color indicates element fails to meet criteria)
(b) Highlighted and Selected Fail Elements for Subsequent Fixing [15]

3/3.3.8 Numbering
After preprocessing, the FE model should be carefully rechecked before generating the solver input
files. Sometimes, due to the frequent import and export operations (especially coping with included
files or super element), the ID numbers (e.g., nodes, elements, components, materials, properties,
contacts, boundary conditions, and loadings) can be very large, and they might be conflicted or over
the digit limit. It is recommended that a project has the numbering rules, and all the analysts use the
same numbering rules in the pre- and post-processing.
3/3.4 Substructures and Submodeling
3/3.4.1 Substructuring
The primary reason for using substructuring (also known as super element technique) is to reduce
computational effort in the solution process. Substructuring can result in an over 50% decrease in
computational time. Substructuring can also reduce the risk of human errors in an iterative design
process. Substructuring can also allow a project team to keep certain confidential component
information secret from the larger team. This can be done because modern FE codes allow the
transfer of boundary matrices (e.g., mass, stiffness, damping, and loads) without showing any
geometric information [140]. However, these foregoing benefits have to be weighed against the cost
of certain other computations that substructuring requires, which a normal analysis would not entail.
Irons and Ahmed [141] identify three circumstances in which substructuring might be attractive:
1. The same substructure is used repeatedly in the structure.
2. A relatively small portion of a structure may behave nonlinearly.

3-84
 3. In a major design effort, different teams may be developing different parts of the structure.
The use of substructuring would allow substructures of different versions of parts of the
structure to be analyzed together. This feature could be very useful during the exploratory
and concept design phases of large structures.
The use of substructuring in the FEA of ships is only likely to be attractive for models involving a
substantial portion of the ship. If a general-purpose FEA system is used, it is essential to have an
understanding of the substructuring technique. Even in the case of design-oriented FEA programs, it
is useful to have an appreciation of the technique.
The ease with which substructuring can be undertaken depends on the features available in the FEA
code being used. This section will be confined to a broad description of the steps necessary to
undertake successful FEA using substructuring, guidelines in using substructuring techniques, and
structural configurations where such techniques might be considered.
The basic steps in FEA using substructuring are:
1. Review of the global model and identification of portions of the structure that repeat. Sketch
of the global model indicating substructure boundaries. Design of mesh in substructures and
determination of boundary nodes.
2. Enter input data. Undertake condensation of substructures and develop substructure
stiffness and load matrices.
3. Generation of global stiffness matrix, which, in general, will require combining the reduced
substructure matrices with portions of the structure not modeled as substructures. At this
point, all the elements of the system equilibrium equations are available.
4. Solve the system equilibrium equations. This run will only yield displacements at
substructure boundaries and portions of the model that were modeled in the usual way.
5. The displacements from the global model can be back substituted into the substructure
equations, as described below, to yield displacements and stresses within the substructures.
This will be repeated for each substructure since, in general, the boundary displacements for
identical substructure models will be different.
The following guidelines for substructure analysis are adapted from Steele [142]:
1. Substructures can be generated from individual finite elements, from other substructures, or
both.
2. Master nodes to be retained must be identified and specified as input when the stiffness
matrices for substructures are calculated. Master nodes include boundary nodes and nodes
subject to loads.
3. Nodes on substructure boundaries that will be used to connect the substructure to the rest of
the global model must be retained as master nodes.
4. Nodes constrained in substructures when substructure stiffness matrices are calculated will
be constrained in subsequent stages of the analysis. These constrained nodes cannot be
released in later stages. However, master nodes can be restrained during the analysis of the
global model.
5. For a substructure to be cost-effective, it should be used at least three times (i.e., replicated
twice).
The following paragraphs contain a description of static condensation, which is a technique
fundamental to substructuring. Also discussed is the two-stage analysis technique, which has found
favor with many analysts. This is followed by a summary of recommendations.

3-85
3/3.4.2 Static Condensation
In the condensation technique, the number of DOF in a portion of the structure is reduced by
condensing out the internal DOF that the remaining active ones being on the boundary. The process
is illustrated in Figure 3/3-26. This substructure can be regarded as a special type of finite element,
and, indeed, is sometimes referred to as a superelement. The mathematics of the process is
relatively simple. The equilibrium equations of the substructure with all its DOFs intact is partitioned
as follows:

Equation 3/3-1

in which the subscripts r and c refer to DOFs to be retained and condensed out respectively. An
expression for c can be extracted from the lower partition, which can then be substituted in the
upper partition to yield:
Equation 3/3-2

or in more compact form:

Equation 3/3-3

Where
Equation 3/3-4

and
Equation 3/3-5

The equilibrium equations given by Equation ( Equation 3/3-4) can be

solved in the usual way. If required, displacements internal to the substructure can be recovered by
static condensation of Equation (Equation 3/3-1) using the Gaussian reduction procedure. Static
condensation amounts to eliminating selected variables using the Gaussian reduction procedure. It
is important to note that no approximation is involved in this process. The condensed-out DOFs are
often called slave DOFs, and the retained DOFs are called master DOFs.

3-86
 Figure 3/3-26 Schematic Illustration of The Static Condensation Process

3/3.4.3 Two-Stage Analysis

In cases where local mesh refinement is required, a two-stage analysis may be justified (see [142]
for practical aspects of two-stage analyses). The first stage of this technique involves the analysis of
a coarsely meshed global model. The local area of particular interest is remeshed using a finer mesh
and reanalyzed using prescribed displacements at the boundary of the refined model as boundary
conditions. The prescribed displacements are taken from the global analysis. The process is
illustrated in Figure 3/3-27. The applied loading, i.e., stresses from the global analysis translated into
pressure loading for the refined model, can also be used as boundary conditions. However, the use
of displacements as boundary conditions is a more common practice since it eliminates the need to
provide additional restraints for sufficiently supporting the model.

3-87
 Figure 3/3-27 Two-Stage Analysis

Design-oriented FEA programs, which model the whole or a substantial part of a ship, suit this
technique. The displacements from a model developed employing such programs can be used as
prescribed boundary conditions for a local fine-mesh model.
In general, there will be several nodes on the boundary of the refined mesh model that are not
modeled in the global model. Therefore, prescribed displacement values are only available for
boundary nodes that exist in the global model. The practice is to assume a linear variation in
displacement, interpolated from the displacements from the global model, for intermediate nodes.
This observation is suggestive of where the appropriate position for the boundary might be. Ideally,
boundaries should be placed in areas where gradients in displacement are small. A comparison of
undeflected and deflected plots of the global model will yield this information.
A finer finite element model is generally more flexible than its coarser equivalent. Hence, there will
be a tendency to underpredict the stresses in the refined model when using displacements
generated in the global model. It is possible to correct approximately for this tendency using a
procedure described by Cook et al. [89]. The procedure requires the computation of the nodal loads
produced by the prescribed boundary displacements. The nodal loads for the local area in the global
model are given by:

in which Kg, δg, and Fg are the stiffness matrix, displacements, and calculated forces pertaining to the
degrees of freedom associated with the nodes on the boundary of the local area. The corresponding
expression for the refined model is:

3-88
The subscript "r" refers to the refined model. Note that only the nodes common to both, the local
area in the global model and refined model, are included in the above expressions.
Once the forces for both cases have been derived, the vector norms for these quantities are
calculated. The norm is a measure of the "size" of vector or the size of the nodal loads. There are
many types of norms, but for present purposes, the following version is recommended:

where Fi refers to the value of the nodal load and n is the number of degrees of freedom on the
boundary that is common to both the local area of the global model and the refined model. The ratio
of the norms for both the cases is calculated to yield a factor as follows:

This factor, which usually exceeds unity, when applied to all stress results from the refined model,
approximately corrects for the overstiffness of the global model results.
The convenience with which this technique can be applied will depend on the FEA software being
used.
3/3.4.4 Superelement Generations
The FEA codes Nastran (e.g., MSC and NX) and Optistruct have three different methods to generate
superelements [95,140]:
1. Static condensation superelement - This method reduces the elastic FE model and load
vectors (e.g., points and pressure) to the interface degrees of freedom. Note that from this
method, the stiffness matrix is accurate, but that of the mass matrix is obtained based on an
estimated approach [140].
2. Dynamic reduction superelement - This method reduces the elastic FE model to the interface
degrees of freedom and a set of normal modes, which are generated from static modes and
normal modes analyses.
3. Component dynamic superelement - This method is efficient for models under iterative
loading
The analyst should plot elements to schematically show the boundaries and size of FE models.
Figure 3/3-28 shows a satellite model represented by using the superelement technique [140],
where the RED and BLUE beams are plot elements to show the connectivity between components.

3-89
 Figure 3/3-28 Use Superelement Technique to Represent a Satellite [15]

3/3.5 Loads and Boundary Conditions

The task of selecting appropriate boundary conditions for the model is often challenging. Generally,
the support condition assumed for the degree of freedom concerned is idealized as completely rigid
or completely free. In reality, the support condition is usually somewhere in between, and in some
special cases, there is no fixture (e.g., satellite with the inertia relief condition).
Several techniques are used to minimize the impact on the analysis of the assumptions made in
boundary conditions. The most popular is to develop models large enough such that the area of
interest is sufficiently remote from the boundary. It is also the practice to make conservative
assumptions so that the results will represent upper bound solutions.
The best guide for determining the extent of structure to model and determining the locations for
boundaries are natural structural restraints or rigid or stiff supports such as major structural
bulkheads, vertical pillars, and columns, or other structural components such as deep fabricated
beams and girders.
It is possible to simulate various types of symmetry, antisymmetry, and axisymmetry by applying the
appropriate boundary conditions. These and other topics related to boundary conditions are
discussed in greater detail below.
3/3.5.1 Minimum Support Conditions
For certain models, it is necessary to provide minimum support for the structure. A good example of
this is hull girder modeling in which the structure is, in reality, supported by the pressure distribution
on the hull. In FEA modeling, a structure with self-equilibrating forces, without any supports, is not
admissible except inertia relief cases. Without proper support or with the presence of rigid body
motion, the equilibrium equations would be singular and, therefore, not solvable.
Models in a plane have three DOFs, and hence need to have two translations and a rotation
constrained. Care is needed to avoid the possibility of rigid body motion. These principles are
illustrated in Figure 3/3-29. Models in three- dimensional space need three translations and three
rotations constrained, i.e., 6 DOFs. Examples to illustrate the minimum support conditions required
are provided in Figure 3/3-29.

3-90
3/3.5.2 Boundary Conditions for Simulating Symmetry
Many structures have one or more planes of symmetry. It is possible to take advantage of this in
FEA, and model just one portion of the structure. Through various devices, it is possible to analyze
structures with a plane of symmetry but subjects to nonsymmetric loads. Such approaches are used
to reduce modeling and computational effort.
In engineering applications, the most commonly encountered types of symmetry are reflective
symmetry, rotational symmetry, and inversion symmetry, as shown in Figure 3/3-30.
In engineering problems, the characterization of symmetry requires not only geometrical symmetry,
but also symmetry with respect to other factors (e.g., material properties, loadings, fixtures, and
contacts (if applicable).
When only part of a symmetric structure is modeled, the symmetric or antisymmetric boundary
conditions must be applied at artificial boundaries introduced because of symmetry. If the y-z plane
is the plane of symmetry, and Ux, Uy, Uz, and Rx, Ry, Rz are assumed as the x, y and z
components of displacement and rotation respectively, the following boundary conditions have to be
applied to the nodes on the plane of symmetry or antisymmetry:
Ux = Ry = Rz = 0 - for symmetry
Rx = Uy = Uz = 0 - for antisymmetry
In the case of symmetry, the points lying in a plane of symmetry can suffer no translation out of the
plane and no rotation about the in-plane axes. For antisymmetry, the complementary set of degrees
of freedom is constrained.
The above discussion has been devoted exclusively to static problems, but free vibration problems
(eigenvalue problems) can also exploit symmetry. The calculation of all-natural frequencies and
mode shapes of a symmetric structure would require one modal analysis for each unique
combination of symmetric and antisymmetric boundary conditions. Note that when only symmetric
boundary conditions are applied to the plane of symmetry, antisymmetric frequencies and mode
shapes are not calculated.
The conditions for static problems discussed above apply equally to linear (time- dependent)
analysis. In addition, if the load is not symmetric or antisymmetric, it will be necessary to decompose
the load into symmetric and antisymmetric components and run the problem twice for each case and
combine the results.

3-91
Figure 3/3-29 Minimum Support Conditions for Models

3-92
 Figure 3/3-30 Different Types of Symmetry

3/3.5.3 Constraints
Constraints are enforced relationships between the DOFs of several nodes. There are many
situations in which constraints can be useful modeling devices. Various types are discussed below
and illustrated using simple examples. The circumstances in which they may be applied, and
limitations in their application, are also discussed.

3-93
The simplest form of constraint is when certain DOFs of different nodes are coupled. The coupling
can be used to enforce symmetry and to release forces and moments. A simple example is
presented in Figure 3/3-31. During analysis, if the independent node is displaced in the y-direction
and/or rotates about the y-axis, the dependent nodes are automatically displaced by the same
magnitude in the same directions.

Node 1 is independent
Figure 3/3-31 Coupled DOF: Nodes 1, 2 and 3 Coupled in the y-Direction and About the y
Axis

Releases can be introduced conveniently using coupling. For example, a pin can be introduced at
mid-span in a continuous beam by coupling translational degrees of freedom of two coincident
nodes. In certain circumstances, the coupling can introduce apparent violations of equilibrium.
A more powerful and general method for introducing constraints is by using constraint equations: A
constraint equation is a linear equation that relates the displacement or rotational DOFs of nodes.
These are sometimes referred to as multi-point constraints (MPC). Constraint equations may be
used for many purposes, such as coupling of nodes by rigid members, rectifying small geometric
discrepancies, and coupling adjacent nodes representing locally offset supports and attachments.
Rigid regions in structure may be defined using constraint equations. Figure 3/3-32 illustrates the
use of constraint equations using the example shown in Figure 3/3-31. In this case, the equation
ensures that there is no relative movement between Nodes 1 and 2 in the x-direction.
Note that in certain analyses (e.g., crash problems), there is no specific constraint defined, and the
calculation is based on the equilibrium between kinetic energies (external) and strain energies
(internal).

3-94
 Figure 3/3-32 Constraint Equation

3/3.5.4 Loads - General

Loading in finite element modeling may be applied in a variety of ways. Typical structural loads are
forces, pressure load, gravity, body forces, and temperatures applied at nodes and on elements of
the model. The load can be applied to:
1. nodes (e.g., nodal forces and body forces)
2. element edges or faces (e.g., distributed line loads, pressure, contact)
3. entire model (e.g., gravity loads and remote loads)
Generally, the load types and methods of its application to the model are specific to a particular FEA
software package. However, descriptions of typical load types are provided in the following
paragraphs.
3/3.5.5 Loads - Nodal Force and Prescribed Displacement
A nodal force is the combination of forces applied to the six nodal DOFs. A nodal force consists of:
1. force magnitudes in X, Y and Z directions
2. moment magnitudes about X, Y, and Z axes (for structural elements)
Nodal forces are usually applied in Nodal Coordinate System, as shown in Figure 3/3-33.

3-95
 Figure 3/3-33 Definition of Nodal Force

Applied nodal loads must be compatible with the element type used. For example, a model
consisting of only solid elements has no rotational degrees of freedom. Any nodal moment loads
would have to be applied in such a case as a force couple with the forces acting at different nodes.
Also, forced or prescribed nonzero displacement may be input directly to nodes as a load case. This
displacement should be prescribed with precision because small changes can cause large
differences in stress response.
In fracture mechanics related problems, the load needs to be applied at the boundaries of the body.
For example, a single-edge notched bend (SENB) specimen is loaded in a setup of three-point
bending, where the force is applied at mid-span and restrained at two ends. In the three-point
bending setup, the mid-span of the specimen is loaded by using a rigid pin. However, in the FEA
setup, it is not recommended to mimic the setup by only applying load to a single node (see Figure
3/3-34(a)). That will cause local stress and strain concentration, and the element connected to the
applied node will yield almost immediately. It is recommended to apply a locally concentrated load
through a finite domain (3 nodes in Figure 3/3-34(b)) or to apply the load by using a rigid pin
associated with friction or frictionless contacts [47].

Figure 3/3-34 Examples of How to Applied Nodal Forced

3/3.5.6 Loads - Nodal Temperature

A nodal temperature is a single temperature value or pair of values applied to a node, as illustrated
in Figure 3/3-35. A pair of values may represent the shell top and bottom surface temperatures.
Some programs allow the specification of a pair of values representing the shell mid-plane

3-96
temperature and a gradient.

Figure 3/3-35 Definition of Nodal Temperature

3/3.5.7 Loads - Face Pressure

A face pressure is a single pressure value applied to selected faces of elements, as shown in Figure
3/3-36. The units of pressure value are force per unit area. The pressure is applied to each selected
element face across the entire face and acts in a direction perpendicular to the face. Some FEA
programs allow the user to specify pressure at nodal points. A variation of pressure over an element
surface can thus be defined. Constant pressure is then a special case corresponding to all element
nodes having the same pressure.

Figure 3/3-36 Definition of Face Pressure

3/3.5.8 Loads - Edge Loads

An edge load is the combination of the forces and moments that can be applied to the edge of an
element, as shown in Figure 3/3-37. The types of edge loading depend on the type of element. An
edge load can be applied to beam elements as:
1. axial force

3-97
 2. shear force
3. torque
4. bending moment.
Uniformly distributed loads on beam elements can be handled exactly, and no further subdivision of
the beam element is required to improve the representation of the load.
For membrane elements, edge loads can be applied as in-plane forces, and for plate bending
elements both in-plane, and out-of-plane forces can be applied along with bending moments.

Figure 3/3-37 Definition of Edge Pressure

3/3.5.9 Loads – Thermal

A beam temperature is the temperature at the centroid of the beam's cross-section and is applied as
temperature, Y-axis gradient, or Z-axis gradient in degrees, as shown in Figure 3/3-38.
Most programs allow for input of thermal loading directly on elements. Others permit, in addition,
specified nodal temperature and temperature-dependent material properties.

Figure 3/3-38 Definition of Beam Temperature

3/3.5.10 Gravity and Acceleration

Inertial loads are generated as a result of the body accelerating. A special case is the self-weight of
a structure, or body, which is generated by the acceleration due to gravity. Inertial loads are

3-98
generated as a result of one or more of the following:
1. translational acceleration
2. angular velocity
3. angular acceleration
FEA software systems treat weight data in different ways. It is important, therefore, particularly for
dynamics problems, to be aware of the way in which the system treats mass and gravitational
forces.
3/3.6 Analysis Control and Solution Options
The analyst must select the type of solver, the solver solution method, and the solver solution
method settings (analysis controls/settings). These are discussed in turn in the sections that follow.
In the analysis control, the parameters specify the overall aspects for the analyzed FE model, e.g.,
degree-of-freedom, number of time steps and type of time functions, displacement and strain
formulations, convergence tolerance, printout types. It should be noted that different analysis types
may require different settings of analysis control. The analysis types have been presented in Section
3/2.1.
Aiming to run the jobs accurately and efficiently, it is highly recommended to use and control
arguments of solution options instead of the default values. All the arguments of solution options for
a given FEA code are generally available in the help documents.
This topic needs (1) lots of experience and understanding of what we want to accomplish, (2) a good
understanding of analysis capabilities of the software adopted, and (3) extra cares in setting up
models. This is a big topic, and it is recommended to read through the FEA and continuum
mechanics books [25,28,99,143–151] for additional information.
3/3.6.1 Types of Solvers
In the FEA codes, the analyst may have several types of equation solvers in the nonlinear analysis,
such as spare solver, 3D-iterative solver, and non-symmetric spare solver. In general, the spare
solver is the most effective, the 3D-iterative solver is effective for solving models with high-order 3D
elements or contact, and the non-symmetric spare solver is suitable for solving the model with Mohr-
Columb or Drucker-Prager material models.
Note that in some FEA solvers (e.g., NASTRAN), there is an option to allow the job to continue
execution, although the stiffness matrix is not positive definite. The forced accomplished solution
may produce misleading results, and the analyst should review the model errors. In general, the
non-positive definite stiffness matrix indicates an insufficient restraint condition.
3/3.6.2 Solver Solution Methods
3/3.6.2.1 Selection of Implicit or Explicit (or Co-simulation) Methods
There are mainly two different types of solution methods in solving equations of FEA simulations: the
implicit method and the explicit method. The difference, essentially, between the implicit and explicit
method is the incremental nature in nonlinear analysis. For a nonlinear analysis (e.g., geometric
and/or material), it is required to gradually apply the load (or displacement) with an increment. After
each increment, the global stiffness matrix needs to be updated due to the change of geometry
and/or material. In both the implicit and explicit schemes, the stiffness matrix is updated in the
incremental procedure, but the primary difference between the two methods is the implicit method
has an additional step to enforce equilibrium of the internal loads with external loads by using the
Newton-Raphson methods. This additional step makes the implicit method unconditional stable. In
general, both methods can be used to solve the same nonlinear problem. However, as to the
question of which method to use (implicit or explicit?), the answer is that – it depends. The following
sections describe the difference between the two methods and help the analyst make decisions on

3-99
which technique to use.
3/3.6.2.2 Integration Methods: Implicit vs. Explicit
Both implicit and explicit methods can be used to solve the partial differential equations (FEA). The
general form of the equilibrium equation is as follows:
Equation 3/3-6

𝑭𝒊𝒏𝒆𝒓𝒕𝒊𝒂 (𝒕) + 𝑭𝒅𝒂𝒎𝒑𝒊𝒏𝒈 (𝒕) + 𝑭𝒊𝒏𝒕𝒆𝒓𝒏𝒂𝒍 (𝒕) = 𝑭𝒆𝒙𝒕𝒆𝒓𝒏𝒂𝒍 (𝒕)

Equation 3/3-7

𝑀𝑢̈ (𝑡) + 𝐶𝑢̇ (𝑡) + 𝐾𝑢(𝑡) = 𝐹𝑒𝑥𝑡𝑒𝑟𝑛𝑎𝑙 (𝑡)

where M, C, and K are mass, damping, and stiffness matrixes, respectively; u and F are
displacement and force matrixes.
Both explicit and implicit solvers can be adopted in solving equations Equation 3/3-6 and Equation
3/3-7. However, these two different solvers have their pros and cons regarding the equilibrium
conditions and computational cost. For example, the explicit solvers are more computationally
economical for solving dynamic problems (time-dependent) because the solution scheme does not
require matrix inversion (unlike implicit solvers). However, the explicit solver solution is not
unconditionally stable (unlike implicit solvers), and it is stable only if the time step is sufficiently
small.
The time step calculation equations from element size or nodal mass are as follows:
Equation 3/3-8

𝒍 𝝆
Δ𝒕 = = 𝒍√
𝒄 𝑬
Equation 3/3-9

𝟐𝒎
Δ𝒕 = √
𝒌

where l, c, 𝜌, and E are the dimension of the smallest element, speed of sound, the density of the
material, and modulus; m and k are nodal mass and equivalent nodal stiffness, respectively. In
practical applications, we generally add a scaling factor of 0.9 in Equation 3/3-9 to make sure that
the time-step estimated is sufficiently small.
Note that interfaces (contact problems) also control the time step during the simulation because they
are essentially stiff spring elements. In addition, it should be noted that the mesh quality is another
key factor that controls the time steps, i.e., a model with higher mesh quality (especially by
controlling smallest element sizes properly) has a reasonable time step.
It is interesting to know that nowadays, some FEA preprocessors have the ability to evaluate the
time steps automatically.
Figure 3/3-39 schematically shows where implicit and explicit solvers fit in simulating an engineering
problem. For example, a short-time transient problem (e.g., impact or explosion response analysis)
is preferable to adopt explicit schemes, but a long-time transient problem is preferable to adopt
implicit schemes. Figure 3/3-40 summarizes the cost of using implicit or explicit solvers in terms of
the complexity of the solution. From two figures, the analyst may choose which solver to use in
solving the engineering problem. A discussion of solution controls settings to improve accuracies is
found in the following Section 3/3.6.2.3.

3-100
 Figure 3/3-39: Implicit and Explicit Solver Applications Chart

Figure 3/3-40 Implicit and Explicit CPU Requirements

3/3.6.2.3 Explicit Solution Controls

In the explicit analysis, the number of calculating steps depends on the time step determined based
on Equation 3/3-8 and Equation 3/3-9. For a problem with a relatively longer time duration, analysts
may want to save the calculation time and reduce the computational cost by (1) accelerating the
event (to reduce the total duration time), and/or (2) adding mass to increase density 𝜌 – (mass
scaling to increase the time step). Note that although these methods are useful and can significantly
reduce the calculation time, extra care is needed. Several recommended actions are listed as
follows.
 Effect of inertia force is less pronounced (such as crack driving force J-integral calculation in
quasi-static problems [28]);
 The increase of velocity and/or mass must be documented;
 If automatic mass scaling is enabled, the increase of mass at each time must be
documented;

3-101
  The ratio of kinetic energy (increased) to the deformation energy should not exceed 1% for
the quasi-static problem or 5% for the dynamic problem [28].
 It is recommended to use ‘double precision’ in the simulation to control the error
accumulation and keep the stability of the solution.
3/3.6.2.4 Implicit Solution Controls
In the implicit analysis, the Newton Raphson or its derivatives methods are adopted to calculate the
stiffness matrix for each iteration. For dynamic implicit analysis, the time integration procedures
(e.g., Newmark family of methods and the -method) are used with Newton methods. As a rule of
thumb, the minimum number of time steps (tstep_dynamic_implicit) should be no less than 10, i.e.,
tstep_dynamic_implicit ≥ 10. The controlling parameters can be tuned based on the experience as
well as the preliminary runs. For the -method [28,106,152,153], there are three parameters , ,
and , and the method is unconditionally stable if having the following relationship:
Equation 3/3-10
𝟏
𝜷= (𝟏 − 𝜶)𝟐
𝟒
Equation 3/3-11
𝟏 𝟏
𝜸= − 𝜶 𝒇𝒐𝒓 − ≤ 𝜶 ≤ 𝟎
𝟐 𝟑
when  is less than 0, it gives a numerical damping. Such an approach can help reduce the noise of
results from high-frequency modal analyses. The current FEA software may provide some
recommended values of numerical damping that are unconditionally stable.
 Without numerical damping
 = 0,  = 0.25, and  = 0.5
 With numerical damping
 = − 0.05,  = 0.2756, and  = 0.55
 = − 0.1,  = 0.3025, and  = 0.6
Note that some codes may provide different recommended numbers of numerical damping
parameters, e.g.,  = 0.38, and  = 0.6 [152].
In the static implicit analysis, the dynamic effects are less pronounced, and the equilibrium equation
becomes
Equation 3/3-12

𝑲𝒖(𝒕) = 𝑭𝒆𝒙𝒕𝒆𝒓𝒏𝒂𝒍 (𝒕)

In such a static or quasi-static analysis, the load or displacement (forced-displacement) is applied
gradually with a user-defined increment (i.e., time steps). It should be noted that in the implicit
method, the time steps generally have no physical meaning, whereas, in the explicit method, the
time steps should be defined from the physical observations. Similar to the experimental tests, there
are two kinds of control algorithms, i.e., displacement-control and force-control. The selection of
control algorithm should be based on the expected responses as well as the interested results (e.g.,
post-buckling). There can be three feature points (i.e., limit point, bifurcation point, and turning point)
regarding the force-displacement relationship (see Figure 3/3-41).

3-102
 Figure 3/3-41 Three Feature Points in Load vs. Displacement Curves
Reproduced from DNV GL C208

Three feature points correspond to three different ways to choose the control algorithm:
 Force-control algorithm: The load and displacement relationship can be captured up to the
limit or bifurcation point, i.e., the tangent stiffness is greater than 0 or no decrease of force.
 Displacement algorithm: The load and displacement relationship can be captured up to the
turning point, i.e., no decrease of displacement.
 Displacement algorithm associated with the Arc-Length method: The full force-
displacement relationship with the presence of snap-back (turning point) can be captured.
In the implicit analysis, the use of the Newton Raphson method (with user-defined tolerance) tends
to be more accurate sometimes with bigger increment steps. This method is robust and can handle
highly nonlinear problems such as cyclic loading, snap-through, and snap back as well as the proper
control algorithm is used. However, one of the major drawbacks of this method is that the global
stiffness matrix needs to be updated and reconstructed in each iteration, which can be costly in
terms of time and computational power. An alternative method – Modified Newton Raphson method
can be adopted to decrease the cost.
This type of analysis tends to be more accurate and can take somewhat bigger increment steps.
Also, this type of analysis can handle problems better, such as cyclic loading, snap-through, and
snap back so long as sophisticated control methods such as arc length control or generalized
displacement control are used. One drawback of the method is that during the Newton-Raphson
iterations, one must update and reconstruct the stiffness matrix for each iteration. This can be
computationally costly. (As a result, there are other techniques that try to avoid this cost by using
Modified Newton-Raphson methods.) If done correctly, the Newton-Raphson iterations will have a
quadratic rate of convergence, which is very desirable. The Newton-family methods have a quadratic
convergence if done correctly.
3/3.6.2.5 Memory Management
In solving large problems, the usage of memory can often become an issue, especially when using
the implicit method-based solver. In most FEA codes, the implicit solver has two different ways to
manage and allocate the memory: in-core (physical memory) and out-of-core (virtual memory on the
hard drive). Noted that it is more efficient to run simulations in the in-core mode than out-of-core
mode if there is sufficient memory. If there is not sufficient memory, the simulation will be run in the
out-of-core mode. The following three parameters are generally needed in the setup.
1. Maximum memory for solution: This parameter specifies the maximum memory to be
used for the solution. It mainly includes the memory for allocating the model data and solving
the equations. The rule of thumb to determine the maximum memory (MAX) is 80% of the
physical memory (RAM), i.e., MAX = 0.80 × RAM.
2. Memory for storing model data: As above-mentioned, this parameter is part of the
maximum memory and specifies the amount of memory storing stiffness and/or mass matrix
and element information.

3-103
 3. Automatic memory allocation: In modern FEA codes, there is an option to allocate the
maximum memory automatically that may provide optimal value for the analyst, e.g.,
90%RAM in ABAQUS (User Manual 3.4.1) [96] and 85%RAM in ADINA [128].
Note that the above-described contents are reasonable for SMP (shared memory parallel) based
processing and that for DMP (distributed memory parallel) based processing will be different. In
addition, the memory management is directly related to the number of processors (or cores) used in
a run, as well as the precision of solvers (i.e., single or double precision). Note that the setup for 16-
bit and 32-bit computers is not considered here.
3/3.6.2.6 Batch Mode
Most of the FEA codes on the workstation are run in an interactive mode. However, sometimes, it is
more convenient and efficient to run the simulation jobs in the batch mode without user input. The
analyst needs to write the batch command to run jobs per FEA references. Nowadays, there may be
a job manager and third-party software with GUI (graphic user interface) available for the analyst to
submit and run jobs in the batch mode, for example, LS-RUN for running LS-DYNA jobs in Windows
OS. Several key input parameters are listed as follows:
 Input and output
o Input and output file directory
o Input and output file name(s)
Note: some solvers have strict file naming rules.
 Solver
o Solver directory
o Solver version and precision
 Hardware
o Number of CPU(s) and/or GPU(s)
o Number of thread(s)
o Size of RAM
o Local workstation of the computer cluster
 Other arguments
o Check run
o Kill job
o Stop job
o Create Information file(s)
o Create restart file(s)
3/3.6.2.7 Nonlinear Solution Convergence Monitor
For the nonlinear analysis, there is a need to monitor the convergence of simulation results at each
time step. At a typical time step, it may need several iterations to achieve an equilibrium state where
all the key indexes (e.g., force, energy, and contact) meet the criteria/tolerances. It should be noted
that there are exceptions when special algorithms are adopted in the calculations, and a higher
tolerance value (i.e., easier to converge) may be needed. The history of solution convergence is
generally recorded in the result output file. Note that the recommended parameters (for example,
max/min iterations per time step, convergence tolerance, and algorithms) are available in software
manuals and benchmark examples. Some ship-structure-focused publications [154–158] also
provide the best approaches and examples for convergence practice.
3/3.6.3 Static Analysis
Static analysis is used to determine the displacements, stresses, strains, and forces in structures
due to loads that do not induce significant inertia and damping effects. The loads and the structure's
response are assumed to vary slowly, if at all, with respect to time. The primary application of FEA in
ship structures is in support of design, and this usually involves static analyses. These may range
from global models encompassing the whole ship, to very detailed local models. Apart from FEA

3-104
performed in support of design, static analysis is also used in the investigation of certain types of
structural failures. Static FEA analyses can be categorized into two groups: linear static and
nonlinear static.
3/3.6.3.1 Linear Static Analysis
In general, the linear static analysis is the easiest one to obtain converged simulation results.
Therefore, before simulating complex problems, an approximated linear static analysis is carried out
to check the degree of nonlinear behaviors in terms of stress levels above the yield strength. As
indicated in Section 3/2.1.2, this type of analysis is related to stiffness (e.g., Young’s modulus and
Poisson’s ratio), and there is no need to define material density. Note however that if the effects of
mass are required (e.g., inertia relief), the material density needs to be considered.
One of the key solution options is to define the time step quantity for each load step. The time step
quantity determines how many increments the calculation needs to reach the predefined loading. For
example, some FEA codes assume a one-time step for a load step in the linear static analysis.
Another key solution option is to define the load steps/sequences if there are multiple loads included.
Note that the linear static analysis mentioned here refers to general analysis and excludes special
problems such as high cycle fatigue.
3/3.6.3.2 Nonlinear Static Analysis
The implicit integration method is generally used for solving nonlinear static analysis because such
types of analyses can be regarded as being time-independent or infinitely slow. The time
independence results in the partial differential equations being stable (unconditionally stable). This is
discussed further in Section 3/3.6.2. The key non-linear solution options are the following:
1. Load steps or load sequences: A loading history may consist of several load steps (e.g.,
loading, unloading, reverse loading, and unloading). The solution of nonlinear static analysis
is always history-dependent. Therefore, the load steps need to reflect the actual scenario
defined in the problem description.
2. Time steps or time increment: The time step or time increment describes how many
increments it takes to reach the predefined load at a given load step. Note that the number of
time steps defined in the implicit method does not have physical meaning. For example, if
“time step = 10”, it means that the loading level increases 10% at each time step. By
breaking the simulation into a number of steps, it may help find the approximate equilibrium
configuration at the end of each time step, and the results of interest can be exported to
output database files.
3. Iterations for each time step: The number of iterations is trials or attempts at finding the
equilibrium solution in an increment. If the iteration process diverges, the iteration process
can be terminated or recalculated with a smaller increment size.
4. Finite or large strain formulations: If the problem involves large geometric nonlinearity (for
example, crack-tip opening displacement calculations), the large/finite strain formulations
need to be adopted in the calculation. Note that the analyst should examine whether the
formulations of shell and material are compatible with the finite strain formulations.
5. Newton Raphson: Newton Raphson method and its derivatives are one of the most widely
used numerical methods to find the solution of static equilibrium equations. Most of the FEA
software provides two options: (i) Full Newton Raphson method and (ii) Modified Newton
Raphson method. The major differences between the two methods are that the modified
Newton Raphson method uses the stiffness matrix from the first iteration to find the
converged solutions, whereas the full Newton Raphson method uses the updated stiffness
matrix after each iteration. For highly nonlinear problems, the Newton Raphson method can
also use the line search method, which can assist in obtaining the converged solution. Note
that the analyst may also use other methods to solve the equations, such as the load-

3-105
 displacement incrementation method and Broyden-Fletcher-Goldfarb-Shanno (BFGS)
method.
6. Others: Except the above-mentioned five major aspects, the analyst may also need to
determine other aspects, such as direct or iterative equation solver, symmetric or
unsymmetric matrix storage, full newton, or quasi newton solution technique.
It is crucially important to choose a proper time step. In some FEA codes (e.g., ANSYS, ABAQUS,
LS-DYNA, ADINA, Optistruct/NASTRAN), the automatic time-stepping method is available and can
be adopted to obtain the converged solution with fewer time steps during the equilibrium iteration. It
is a reliable and suitable method for a wide range of nonlinear static problems. The time step
estimated from this method is a dynamic value that keeps updating based upon the size of the time
step where convergence was achieved previously. For example, if the original time step (t) cannot
attain convergence, the approach reduces the time step (e.g., 0.5t); if the new time step cannot
attain convergence, the approach reduced the time step again (e.g. 0.25t). The time step will be
automatically revised until the convergence is achieved, or the run will be terminated if the number of
trials reaches the limit.
3/3.6.4 Buckling Analysis
Depending on the structural element, the estimate of buckling load can be very sensitive to the
inevitable presence of discontinuities, imperfection, residual stresses, and material nonlinearity. The
application of FEA techniques to solving buckling problems should be approached with caution. The
results can be very sensitive to assumptions made in regard to deviations from the ideal, more so
than is typical for linear static analysis. The usual practice in design situations is to adapt classical
solutions to the problem.
There are two different types of analyses for evaluating the buckling resistance of a ship structure,
i.e., linear and nonlinear buckling analyses. The buckling resistance measured from the nonlinear
analysis is generally higher than that from the linear analysis. However, the difference is dependent
on the actual model.
The buckling behavior may be observed on a global ship structure or the local element(s), e.g.,
stiffened/unstiffened panels, girder webs, bulkheads, etc. The presence of elastic buckling of a local
element may not be critical due to the redundancy characteristic.
3/3.6.4.1 Linear Buckling Analysis
For the linear buckling analysis, the analysis control is similar to a linear static analysis (see Section
3/3.6.3.1), because it is essentially an eigenmode analysis. The number of eigenmodes should be
defined as well as the type of solving algorithms (e.g., Lanczos, AMSES, or AMLS). Note that if the
effect of gravity is considered, the pre-stress state needs to be included in the analysis. Since this is
a linear elastic analysis, the stress level should be less than the yield strength. The first (and
sometimes second) eigenmode is the dominant one and should be used for buckling resistance
calculations.
3/3.6.4.2 Nonlinear Buckling Analysis
The nonlinear buckling analysis includes nonlinear factors, such as imperfections, misalignments,
discontinuities, residual stresses, material, and geometric nonlinearities. Most FEA codes have
approaches with built-in or third-party functions to represent the profiles of imperfections,
misalignments, discontinuities, and residual stresses in the analysis control (e.g., a geometric
perturbation in LS-DYNA [28]).
The analysis controls of material and geometric nonlinearity are similar to those of nonlinear static
analysis (see Section 3/3.6.3.2). The nonlinear stress-strain relationship is adopted to reflect the
material nonlinearity. The large deformation formulations need to be activated to capture the
geometric nonlinearity. Note that there are two different ways to control the load, i.e., force control or

3-106
displacement control, and the differences between the two methods have been presented in Section
3/3.6.2.4. If the force control is employed in the analysis, the magnitude of the load must be higher
than the limit load (i.e., failure load).
Note that in Section 5.4 of DNVGL-RP-C208 (2016 edition) [106], a hybrid method for determining
the buckling resistance is proposed, which is based on both nonlinear buckling analysis and the
calibration against testing results or empirical formations.
3/3.6.5 Impact Analysis
The existing explicit solvers for solving impact (or crash) analyses are historically from the same
open-source research code – DYNA3D developed by Lawrence Livermore National Laboratory
[159]. Therefore, the analysis controls (e.g., control decks/cards) are quite similar among those
solvers.
As discussed in Section 3/3.6.2, the explicit solver is NOT unconditional stable, and extra care is
required to obtain reliable results. Except material models and mesh designs/qualities discussed in
Sections 3/3.2 and 3/3.3 of this part, respectively, the following parameters and criteria in the FEA
settings [28,91,105,160,161] are also predominant factors in the analysis control:
1. Timestep
a. critical time step Δ𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙
b. scale factor, e.g., 0.9 in the time step calculation Δ𝑡 = 0.9Δ𝑡𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙
2. Hourglass
The hourglass describes a weird type of deformation mode that shows deformation in the
“hourglass” shape without generating any stress when using reduced integration elements.
An example of the hourglass effect is shown in Figure 3/3-42 [105,162]. In general, there are
several hourglass formulations available in the FEA codes.
A rule of thumb in monitoring the hourglass is that the hourglass energy should be no greater
than 10% of the internal energy.

Figure 3/3-42 Test of Hourglass Control [28]

3. Result output time step

It includes two sets of time steps: plotting results (e.g., d3plot files in LS-DYNA) and
exporting information of solving equations (e.g., ASCII text files where energy, contact force,
stress, strain, SPC force, status, etc.).
The time step of plotting results (Tplot) defines the number of interest frames from the initial
time to the termination time, and it is recommended to have at least 10 frames/time steps.
For example, if the termination time is 100 ms, and the time step is 10 ms, the plotting result
files will be generated at each 10 ms increment. In total, there will be eleven (11) result files,

3-107
 including that at the initial time step.
The time step of exporting information of solving equations (TI/O) refers to the number of
calculation information recorded, which are critical when reviewing the simulation results. It is
recommended to store at least 10 sets of calculation information between each time step of
plotting results, e.g., TI/O = 0.1×Tplot.
4. Initial contact penetration
The non-zero initial contact penetration will induce calculation errors. The analyst can either
review the penetration in the pre-processor or run a simulation with a finite time step (e.g., 1
ms). In general, the initial penetration will be highlighted in the pre-processor (e.g., ANSYS
and HyperMesh [15]). From the trial run, the ASCII text files will indicate whether there is a
non-zero initial contact force or contact energy, which indicates an initial contact penetration.
5. Contact models and compatibilities
The modern FEA codes have become smarter and more robust in terms of definition
contacts, e.g., automatic single/multiple surface contacts in LS-DYNA [28]. However, it is still
recommended to review the contact models to check whether the definition (e.g., linear or
nonlinear formulations) is compatible with physical problems.
6. Compatibility between element formulations and material models
Similar to the usage of implicit solver, the analyst needs to review the compatibility between
element formulations and material models and understand the overall limits of the element
and material adopted. The solver theory and keyword reference manuals generally include
the checklist of compatibilities. In addition, it is recommended to have a verification run
before the formal run.
7. Mass scaling
As mentioned in Section 3, the analyst has to check the weight of the FE model after
assigning thickness and properties to meshes. However, because of small-time step size
from the usage of small-size or fine elements, the analyst needs to activate the mass scaling
feature, which can increase the time-step size in each cycle and reduce the cost of
computation power. This feature is extremely useful in solving quasi-static or low-speed
problems, where the inertia force (kinetic energy) does not play a predominant role. The
analyst may also increase the mass locally on several critical elements (e.g., the smallest
element size) such that the time step increases per Equation 3/3-9. Figure 3/3-43 shows the
time step size with and without applying the mass scaling technique.
There are several ways to activate the mass scaling feature [163–165], for example:
1. Increase the mass density of the component(s);
2. Add mass automatically to elements invoking small time steps (with predefined
criteria);
3. Similar to (2) but added mass with the time-step scaling factor (generally default
value is 0.9).
It is up to the analyst to make an engineering judgment on gauging the effects of mass
scaling on simulation results. As a rule of thumb, a simulation with a mass increase of less
than 5% can be regarded as a reliable and effective run. Again, it should be noted that a
pronounced increase in mass can cause overly severe penetration problems and increase
the dynamic effects significantly. The penetrations and kinetic energy have to be monitored
closely by the analyst after each run.

3-108
 Figure 3/3-43 Relationship between Simulation Time and Time Step Size With and
Without Mass Scaling Features

3/3.6.6 Fracture Analysis

3/3.6.6.1 Fracture Analysis
In general, the fracture toughness indexes (e.g., KI, II, III, GI, II, III, JI, II, III, CTOD, CTOA, C*, etc.) are of
interest for the fracture analysis. The analyst needs to choose the method for calculating these
fracture toughness indexes. For example:
1. The linear contour method is good for calculating J and K (linear elastic condition) the 2D
problems with the stationary crack and the propagating crack. The method is compatible with
both blunt and straight crack front profiles.
2. The virtual crack extension (VCE) method with nodal virtual shifts (NVS) method can be
used for both 2D and 3D problems with the stationary crack and the propagating crack. The
VCE-NVS method is compatible with both blunt and straight crack front profiles.
3. The VCE with station virtual shifts (SVS) method is not applicable for 2D problems but 3D
problems. The VCE-SVS method is valid for both mapped mesh and free-form mesh near
the crack front. In addition, this method can deal with more than one crack in the analysis
where linear contour and VCE-NVS methods cannot.
As discussed in Section 3/2.6.4 Basics of Fracture Mechanics, the extra cares are needed in
determining the number of contours for J calculations, and which one should be used for J value
output. The following bases should be noted:
1. If the small deformation and small strain formulations are adopted, the J values are
essentially contour-independent, except the 1st contour, and the last one touches the model
boundary.
2. If the large/finite deformation and large/finite strain formulations are adopted, the J values
are strongly contour-independent. It is recommended to use the value that represents the
far-field works.
The general analysis control is the same as those presented in Section 3/3.6.1. Most of the
conventional constitutive material models (e.g., stress-strain relationship) can be adopted in this

3-109
analysis, and it is up to the analyst to verify the compatibly between the material model and solvers
per the user manual of adopted FEA solver.
Note that the averaged J values should NOT be used, which has no physical meaning. If the CTOD
values are of interest, the analyst needs to understand which type of CTOD definition is adopted and
how to interpret it from FEA results.
3/3.6.6.2 Fracture Analysis with Damage Criteria
Similar to Section 3/3.6.6.1, general analysis control is the same as those presented in Section
3/3.6.1. If the damage model adopted is not stress- and/or strain-indexes based, it is up to the
analyst to check the compatibility between the damage model and FEA setups (e.g., element types
and strain formulations). The micromechanical parameters in the damage models require a hybrid
methodology to be determined and calibrated from both testing and numerical simulations.
If the Gurson model is adopted to detect the local damage, the user-defined material model is
required, which is generally not provided by the FEA software.
If the cohesive model is adopted, it is recommended to choose the multi-polynomial law (i.e., four
parameters) where the model can fulfill the following requirements [118]:
1. The initial stiffness of the cohesive element can be varied;
2. The initial thickness of the cohesive element has a minimum or no impact on the results;
3. A region can be defined, where the traction is kept constant;
4. The curve of the cohesive law is steadily differentiable and reason for numerical
implementations.
3/3.6.7 Fatigue Analysis
Fatigue analysis includes two types of problems: linear elastic FEA based and nonlinear elastic-
plastic FEA based. The major differences between them in the analysis control are the loading levels
and material models. The loading levels are higher or comparable to the limit load, where the
corresponding stress and strain relationship is beyond or close to the proportional limits (i.e., beyond
yield strength). If both types of fatigue damage cases included, the damage from linear fatigue
analysis should be added to that from nonlinear fatigue analysis, and the total damage should be the
summation of both, i.e.
Damage Total = Damage Linear + Damage Nonlinear
The following parameters affect the fatigue life and should be considered in the analysis control.
1. Cyclic loading (e.g., type, stress range, mean stress, stress amplitude, R ratio)
2. Cyclic stress and strain relationship (e.g., uniaxial or multiaxial, base metal or weldment,
yield strength, ultimate tensile strength, and plastic strain level)
3. Stress/strain vs. number of cyclic loading at interest R ratios (e.g., SN or EN)
4. Thickness effect
5. Temperature effect
6. Surface finish (e.g., polishing, grounding, machining, hot rolling, and forging)
7. Surface treatment (e.g., nitriding, shot peening, and cold rolling)
8. Residual stress effects
9. Notch presence (e.g., fracture mechanics-based fatigue analysis)
10. Safety factors (e.g., Dang Van Criterion)

3-110
Note that the FEA code may have selected parameters mentioned above, and it is up to the analyst
to review the calculation results and determine whether there should be a subsequent correction by
considering the impacts of other factors.
If the rainflow cycle counting is adopted, it is recommended to define a threshold value (also known
as gate value) for load peaks simplification.
If the shell element is used, the stress values from nodal corners and skin layers under tension
should be used for fatigue life calculations.
The certainty of survival of SN or EN curves should be defined or recorded because the data curve
is a statistical approximation of experimental results. For example, if the value of survival certainty is
defined to be 90%, the SN or EN curves offset downward such that 90% of the experimental data
lays above the offset curves (see Figure 3/3-44) [166].

Figure 3/3-44 SN Curve with Different Survival Certainty

It is recommended that the analyst has a sense of the potential fatigue failure mechanisms (e.g.,
uniaxial or multiaxial) as well as the failure modes (e.g., Mode-1, 2, 3, or Mix-mode).
3/3.6.7.1 Linear Fatigue Analysis
Except for the parameters presented in the above section, the mean stress correction approaches
should also be determined in the analysis. The three widely used mean stress correction
approaches in FEA analysis control are the Goodman method [167], the Gerber method, and
Soderberg method [168]. The Goodman method (a bi-linear line plotted in Haigh diagram, see
Figure 3/3-45) is recommended for analyzing brittle materials, and the Gerber method (a polynomial
curve, see Figure 3/3-45) is recommended for analyzing ductile materials. Note that the Goodman,
Gerber, SWT, and Morrow correction methods may have alternative or modified formulations
presented in FEA codes.

3-111
 Figure 3/3-45 Mean Stress Correction Methods in Haigh Diagram

Three approaches of discrete SN curves are generally available in FEA codes, and the SN curves
should be defined accordingly based upon the availability of parameters or experimental data.
1. One segment SN curve in log-log scale with fatigue life defined (Figure 3/3-46(a)).
2. One segment SN curve in log-log scale with the cycle limit of endurance defined (Figure
3/3-46(a)).
3. Two segment SN curve in log-log scale (Figure 3/3-46(b))
In Figure 3/3-46, the abbreviations SR1, B1, NC1, B2, and FL refer to fatigue strength coefficient,
first fatigue strength exponent, the cyclic limit of endurance / transition point, second fatigue strength
exponent, fatigue limit, respectively [166].

Figure 3/3-46 Different SN Curve Approaches

3-112
3/3.6.7.2 Nonlinear Fatigue Analysis
The two widely used mean stress correction approaches in FEA analysis control are the Morrow
method and Smith-Watson-Topper (SWT) method. The Morrow method (the red-line curve in Figure
3/3-47) is recommended for analyzing brittle materials, and the SWT method (the green-line curve
Figure 3/3-47) is recommended for analyzing ductile materials.

Figure 3/3-47 Morrow and SWT Methods for Nonlinear Fatigue Analysis

The settings of material nonlinearity analysis are similar to that of nonlinear static analysis. However,
the fatigue life with the presence of plastic strain is dependent on the loading magnitudes and
sequences, and extra care required in defining load steps.
A typical discrete EN curve in the FEA code is shown in Figure 3/3-48. The abbreviations SF, B, EF,
and C refer to fatigue strength coefficient, fatigue strength exponent, fatigue ductility coefficient, and
fatigue ductility exponent, respectively [166].

Figure 3/3-48 EN Curve Approach

3-113
3/3.6.8 Vibration Analysis
Vibration analyses in ship structures are usually performed for the following reasons:
1. To ensure that the natural frequencies of sensitive structures and components do not
coincide with those of the hull girder or with the forcing frequencies associated with
propellers and other mechanical sources of vibration energy.
2. In preparation for dynamic response computations.
Several quasi-static design procedures have been developed for design against dynamic load
conditions. For some of these procedures, for example, the Design Response Spectrum Method
used for shock analysis, it is often necessary to compute several tens of natural frequencies of the
subject structure or component. In complex structures, such as a large ship’s mast, the natural
frequencies and modes can usually only be calculated using FEA.
As an alternative to quasi-static procedures, a more rigorous dynamic response calculation may be
used. Two methods are available: direct integration of the equations of motion or the superimposition
of modal responses. For nonlinear behavior, such as that associated with large deflections and/or
plasticity, only the former is appropriate.
Transient dynamic response analysis is used primarily for computing response to suddenly applied
loads and/or short-duration loads. Examples include forces due to collisions, wave slamming, and
shock, and blast. In these cases, the loading is very uncertain. Various procedures have been
developed to compute loads from these types of loading. For example, procedures are available to
model the shock forces generated as a result of underwater explosions. The procedure models the
underwater explosion, the pressure-induced on the hull, and finally, the transmission of the dynamic
forces through the hull structure to the structure or component in question. Many transient dynamic
problems involve fluid-structure interaction phenomena where the structural response affects the
loading on the structure. Sometimes it is possible to treat such phenomena very approximately by
adding a certain amount of fluid mass to the elements adjacent to the fluid.
3/4 FINITE ELEMENT RESULTS CHECKS
The results obtained from a finite element analysis (FEA) should always be verified, and their validity
established. To make sure that the results are devoid of any errors in modeling or analysis, it is
necessary to perform the checks outlined in this section. These checks ensure that the FEA results
are calculated, processed, and presented consistently with the analysis requirements.
3/4.1 General Solution Checks
Many of the following checks can be performed using the graphical display features available with
most FEA software systems. Where such features are not available, these checks will have to be
performed by examining the modeling results output.
3/4.1.1 Errors & Warnings
Well established finite element software systems generally have several built-in checks to identify
poor modeling and analysis practices. A warning or an error message is issued when built-in criteria
are violated. The correct practice is to resolve any such messages and take the appropriate remedial
action. If the warning/error message is not applicable to the analysis, proper justification should be
provided.
An example could be a warning message for the angle between adjacent edges in a quadrilateral
shell element. The generally recommended range is between 45°and 135°. If this rule is not
followed, valid justification could be that the element in consideration is located well away from the
area of interest.

3-114
3/4.1.2 Mass and Center of Gravity
It is good practice to verify the mass of the model and the location of the model's center of gravity of
the model. Several programs provide the mass without the need for a full analysis. If this option is
unavailable, the analysis could be run with a 1G loading (with no other applied loads).
3/4.1.3 Self-Consistency
The results should be checked for 'self-consistency.' For example, displacements at fixed supports
should indeed have zero displacements, and any symmetries in the model should be reflected in the
stress and deflection results.
3/4.1.4 Static and Dynamic Balance
This is a fundamental check. The applied loads should be compared with the reactions. The check
should include moments, inertial forces, residual forces/stresses, pre-tension, or other types of
source loadings where appropriate. This check ensures that the applied loads and reactions are in
balance and ensures that the user-specified loading definitions are appropriately interpreted by the
program. When the applied loads and reactions are not in balance, this is an indication of a severe
error.
Checking the forces and reactions also ensures that the results are actually for the intended load. In
the case of pressure loads, due to possible discrepancies in arriving at nodal forces from pressures,
the actual load level could be different from that intended.
In addition, the energy indexes (e.g., kinetic energy, hourglass energy, etc.) should be checked in
both static and dynamic analysis. This check ensures that all the energy indexes are on the right
scale and ensures that the loading and boundary conditions are properly defined in the model. For
example, the energy indexes related to contacts (i.e., slave energy, master energy, friction energy)
can be useful indicators to evaluate whether the contact is defined correctly.
3/4.1.5 Defaults
All FEA software packages have built-in defaults. For certain input parameters, default values or
options are assumed if a value has not been input, or if an option has not been selected. Hence,
checks should be performed to ensure that where defaults have been used, they are consistent with
the assumptions of the analysis.
3/4.1.6 Checklist
The following is a list of checks to ensure the quality of the FEA. The checklist covers both pre-run
and post-run checks. A reference checklist is shown as follows. The analyst can adjust the content
of the list based on the different load cases. For example, in the 2D analysis, “Plot thickness of shell
elements by color” is an efficient way to review whether the thickness is appropriately assigned.
1. Pre-Run Checks - Graphical:
a. Extremities of the model - global dimensions OK
b. Free edges - look for element connectivity
c. Shrunken elements - no missing elements
d. Duplicate nodes
e. Duplicate elements
f. Size of adjacent elements - avoid ill-conditioning
g. Mesh density
h. Mesh transitions

3-115
 i. Plot material properties by color
j. Plot physical properties by color
k. Loads applied to correct elements
l. Direction of loads correct
m. Boundary conditions applied to correct nodes
In addition to the graphical pre-run checks, it is good practice for the analyst to complete
numerical -pre-run checks that may identify common issues such as key geometric
dimensions, magnitudes of loads and material properties, as well as, verification of the
consistency of the units of measure for each of the numeric values.
2. Post-Run Checks:
a. Static or dynamic balance
b. Comparison
i. classical results
ii. simple finite element model
c. Numerical accuracy
i. residuals
ii. stiffness ratio
d. Model response sensibility, e.g., Does the deformed shape make sense? (see
3/4.4.2)
e. Numeric response sensibility, e.g., Are the stresses produced by the model in the
realm of possibility?
3/4.2 Postprocessing Methods
Methods used for postprocessing of derived quantities from an FEA should be explained. The
derived quantities include parameters such as stresses, design margins, factors of safety, etc.
The need and justification for applying correction factors for FEA results should be explained. The
need for applying correction factors may arise due to the necessity to compare FEA results with
design codes. Note that sometimes, the same FEA results can be interpreted and plotted differently
in different post-processors. That is mainly because the default setting for the result visualization is
different, e.g., presentation of averaged, non-average, nodal, or elemental values.
3/4.2.1 Result Accuracy
Since FEA is an approximate technique, the accuracy of simulation results is always dependent on
the engineering approach applied and solver adopted. Where the analytical solution is not available,
the analyst can use experimental data or verified numerical data as the benchmark to check the
accuracy of the results. There are two major aspects to check:
1. computational accuracy
a. strain energy including nominal and residual
b. reaction forces and moments
c. convergence test including mesh sensitivity study
d. average and unaverage stress difference
2. correlation with experimental testing, analytical solution, or verified numerical data

3-116
 a. strain gauge – stress, strain, and displacement comparison
Note: The gauge “sensor” element is available is some FEA codes
b. natural frequency comparison
c. dynamic response comparison including modal or deformed shape comparison
d. temperature and pressure distribution comparison
In general, FEA results that have not greater than 10 to 15% difference from the benchmark results
are a reasonable correlation and accurate calculation. This is according to automotive standards. If
the deviation is more than 15%, one or several of the following input may not be correctly set up.
1. Geometric idealization of the structure
2. Idealization of the applied loading
3. Boundary conditions and/or contact definitions
4. Material properties and/models
5. Presence of residual stress or strain
6. Local geometric effects (e.g., weldment)
7. Local force effects (e.g., bolt with pre-tension)
It should be noted that there is a chance that the benchmark value may also have errors. In such a
case, it is recommended that the analyst review the results together with the project technical
supervisors or senior engineers.
3/4.2.2 View and Interpret Results
The displacement and animation for deformation should always be the first to view, and then any
other output can be viewed. It is recommended [15] that the analyst should visually imagine how the
object would deform for the given loading condition before seeing the result. If an excessive
displacement or illogical movement of the object is observed, it may indicate errors in the FEA setup.
The exaggerated displacement or deformation can be plotted by controlling the scale factor in the
post-processor. It is recommended to set a scale factor a scale higher than the default value (= 1),
e.g., scale factor = 10.
To animate the result is another useful visualization technique. It can help in interpreting results from
both static and dynamic analyses. The animated motions mimic the structural deformation due to the
applied loads and constraint, and also provide insight into the overall structural response.
3/4.3 Displacement Results
In the design of ship structures, the primary result parameter of interest is stress. Most design
criteria are expressed as allowable stresses. Although deflection criteria are not as numerous as
stress criteria in design codes and standards, they can be just as critical. Stiffness requirements for
various components of navigation and combat systems are often quite onerous. Stiffness
requirements are often related to dynamic requirements in which the coincidence of equipment
operating frequencies and those of the equipment-support structure system is to be avoided. As
noted elsewhere, modeling for dynamic analysis is considerably more difficult than modeling for
static analysis. This is particularly true for higher modes of vibration.
In interpreting displacements, it is essential to have an understanding of the accuracy of the FEA,
how they vary for different response parameters, and the influence on the accuracy of modeling
decisions made earlier.
In general, displacements are more accurately determined by FEA than stress.
The methods used for plotting the displacements of framed structures and certain plated structures

3-117
in many FEA software packages may understate the actual accuracy. Beams are often plotted as
straight lines. In reality, the displacement function for beam elements is a cubic polynomial. The
same observation applies to plate bending elements.
In general, displacements in structures composed of beam and truss elements are accurately
predicted within the limitations of the engineering model. In terms of the finite element model
doubling the number of beam elements in, say, a grillage will not improve the accuracy of the result.
The response of two and three-dimensional structures is much more complex, and hence, in
general, displacement results are sensitive to mesh refinement level. Therefore, interpreting
displacement results in plated and solid models require more care. Gross errors are generally
uncovered by the application of intuition and knowledge of previous analyses and physical
experiments. More subtle errors are more difficult to uncover.
3/4.4 Force Results
As mentioned in Section 3/4.1.4, the check of force and moment balance helps in estimating the
numerical accuracy. If the critical or limit load is the design criteria, the summation of force results
should be interpreted and reviewed. Note that there are different approaches to interpreting the force
results. For example, if distributed loads or fixation are applied, the beam elements (e.g., RBE-2 and
RBE-3) can be employed to automatically output the force summation. Note that RBE stands for
Rigid Body Elements. Note also that in general, RBE-2 and RBE-3 elements are used for
transferring displacement and force loadings, respectively. For example, the RBE-3 element should
be used for the multi-points constraint problems. The information of force results is also available in
the output text file.
If the global or local stiffness is of interest, the force results should be extracted at the same location
where the displacement results are extracted.
If the boundary condition nonlinearity (e.g., contacts and following force) is defined in the model,
extra care is needed to check the global or local balance of forces and/or moments. For example, if
the frictional contact is defined, the frictional forces on the master side (e.g., point or surface) should
be equilibrium to the slave side.
As above mentioned in 3/2.3.5, the force results also indicate the level of plastic deformations in a
Neuber correction-based fatigue analysis.
In a fracture analysis, the force results can be used to estimate the J values based on empirical
equations in standards, e.g., [53]. Note that the J values from force results can be used as a
benchmark value to review the accuracies of J values from methods such as contour integral or
virtual crack extension.
3/4.5 Stress Results
As noted earlier, stresses are more difficult to predict accurately than displacements. Limitations in
the finite element method are such that stresses are not normally continuous across boundaries
between elements. For ease of interpretation of results, most FEA software averages stress in some
fashion before presenting the results.
These results are presented attractively as stress contours in color plots, and the underlying
discontinuous nature of the stresses may be obscured as a result of averaging processes, thus
engendering a false sense of confidence in the results.
These problems can be compounded by misunderstandings in regard to the type of stress being
plotted.
The analyst should review the magnitudes of the stresses presented to confirm that they are in the
realm of possibility (e.g., are the stresses significantly higher than the material yield stress). Small
areas of high-stress remote form the area of interest may be developed due to low mesh refinement
or geometric simplifications and may be justified by the analyst. If the magnitude of the stresses can

3-118
not be appropriately described, they may be an indication of modeling data or geometry
inconsistencies.
Stress contours provide a good qualitative indication of the adequacy of the density of the mesh.
Smoothly changing contours usually indicates that the mesh is suitably fine. Alternatively, stresses in
adjacent elements can be compared. It is difficult to give firm qualitative guidance since the accuracy
required depends on the nature of the analysis. A change in stress of more than + /- 20% would be
regarded as unsatisfactory for design purposes. A discontinuity or abrupt change in the stress
pattern across the elements, especially in the vicinity of the maximum stress, indicates that there is a
need for local mesh refinement. In a fatigue analysis, the submodeling technique can be adopted to
capture the high-stress gradient with finer meshing.
In addition to the default option of average stress (generally known as “simple average”), the current
post-processor may also offer other plotting options, such as nodal value, elemental value, corner
value, centroidal value, gauss point value, unaverage value, etc. The stress values plotted from the
nodal, corner, and unaverage options are usually higher than those from default average and
elemental values. If the analysis is being performed to meet a specific regulatory body rule set, then
that regulatory body rule set should be reviewed to confirm see if average stress results (or some
other stress results) are to be reviewed.
In a linear analysis or a unit cycle fatigue analysis, the analyst has to check whether the maximum
stress level is higher than the yield strength.
In a fatigue analysis, it is recommended to use corner stress values rather than the nodal stress
values.
In a fracture analysis, the calculation of J integral is based on the stress values from gauss
integration points. In the literature, the nodal stress of crack front is generally used for calculating the
crack-tip stress.
If the element erosion or damage failure algorithm is activated in the simulation, it is a case by case
scenario. For example, if the maximum principal strain is the criteria index, the strain values of shell
elements are calculated from integration points for both in-plane and through-thickness in each
element.
Appendix E of DNVGL-CG-0129 (2018) standard [88,132] proposed an approach on how to interpret
stress values close to the hot spot for the ship fatigue analysis.
Note that the stress index to check should be made on a case-by-case basis. However, in general,
von Mises stress or equivalent stress should be reviewed or checked for linear and nonlinear
analysis, except special FEAs (e.g., fatigue, fracture, damage, and vibration). In addition, the check
of stress results (1st and 2nd principal stresses, and von Mises stress) sometimes can be used to
understand how load transfers in a local or global structure. That may assist the analyst in revising
the structural design.
3/4.5.1 Stress Components
The unknowns solved for in FEA are displacements (translations and rotations). These
displacements are then used to calculate strains in the element, and hence the stresses. For some
element types, intermediate steps are involved. The nature of inter-element stress discontinuities
depends on the element type concerned.
In one-dimensional elements such as truss and beam elements, there are no discontinuities
because the displacement functions are sufficiently detailed. For example, the standard beam
element is based on cubic displacement and hence can represent linear variations of bending
moment.
Two and three-dimensional lower-order elements generally have discontinuities in the stress field at
element boundaries unless they are in a constant stress field. For plane and solid elements, stresses

3-119
depend on displacement derivatives, and on curvature for plate bending elements.
The stress state at a point is defined by several stress components depending on the element type.
These are summarized in Table 3/4-1.
Table 3/4-1 Stresses Represented by Element Type

The state of stress in plated and solid structures is generally quite complex and has to be combined
in some way for design situations. Many failure theories have been developed wherein "failure" is
said to have occurred when some equivalent stress exceeds the yield stress. The equivalent stress
combines all the stresses acting at a point in the material. The most popular of these is the von
Mises stress which is given by:

The use of the equivalent stress for checking the critical buckling stress is not appropriate. For
buckling checks, normal stress (σx, σy) and shear stress (τxy), as appropriate, should be used.
Generally, normal stresses will not be uniform across the panel. Where this is the case, it will be
necessary to approximate the stress by a linear distribution for which there are standard buckling
formulae. In some cases, the stress state may be biaxial and/or there may be significant shear
stresses. To check these situations, it is usual to calculate the ratios of actual stress and critical
stress for individual stress states and combine the effects using interaction formulae.
In a high cycle fatigue analysis, the maximum principal stress is usually of interest, and the key input
for durability life calculations. Note that it depends on the failure modes to determine which principal
stress index to use in the calculation. For example, if the stress ratio R = 0.1 (only tension), the 1st
principal stress (i.e., maximum tensile stress) is used in a fatigue life estimation. On the other hand if
the failure mode is caused by compression, the 3rd principal stress (i.e. maximum compressive
stress) is used.
In a fracture analysis, all the stress components are used for stress intensity factor or toughness
calculations. In a stress-based damage analysis, the ultimate tensile strength, together with the
maximum principal stress, is used for estimating the critical state.
In a structural optimization analysis (e.g., topology and shape optimizations), the stress components
or scalar derivatives (von Mises stress) can be used as the constraint condition in the objective
functions.
The von Mises stresses should be reported for ductile materials, and the maximum principal
stresses should be reported for brittle materials. Extra care is needed in the nonlinear analysis,
where the true and engineering stresses are used. Some FEA software does allow the use of
engineering stress and strain as input, but some software does not. The definitions of true and
engineering stress are presented in Section 3/2.3.3.

3-120
3/4.5.2 Average and Peak Stresses
Except for the one-dimensional elements, each stress component for each element meeting at a
node will be different. In FEA programs, various techniques have been developed to average
stresses. The stresses in four adjacent membrane elements may look something like the distribution
depicted in Figure 3/4-1.

Figure 3/4-1 Distribution of Element Stresses

Stresses can be calculated at any point in the element. It has been shown, however, that depending
on the element formulation, there are optimal points for computing stresses. In general, stresses are
least accurate at corners, more accurate at mid sides, and most accurate at certain interior points.
For two and three-dimensional elements based on the isoperimetric formulation (by far the most
popular), these interior points are the so-called Gauss points (integration points). One popular
method is to extrapolate the stresses calculated at the Gauss points to the nodes using a more
suitable formula than the actual interpolation functions such as least squares. However, in some
FEA software, the values at the Gauss points are copied to the nearest node without extrapolation,
unless otherwise instructed. There are yet other methods for estimating nodal stresses.
Once the nodal stresses have been calculated for all elements contributing to the node, they can be
averaged to yield average nodal stress. This will be done for all appropriate stress components.
Averaged nodal stresses are much more reliable than element nodal stresses, although the extent of
the stress discontinuity at the nodes should decrease with mesh refinement.
The different methods used by FEA software systems for extrapolating Gauss point stresses to the
nodes is perhaps the main reason analyses of the identical problem, using different systems, can
yield identical displacement results yet differing stress results. One technique used to overcome this
problem is to employ dummy line elements in critical regions of the structure. In this technique, a
dummy truss element is included in the model in the area of interest. An example of such a situation
is the placement of such an element at the edge of an opening. The stress results from the truss
element are directly calculated and are not dependent on extrapolation. The area of the truss

3-121
element should be small enough to have a negligible influence on response. An area of t 2/100,
where t is the thickness of the plate, is a reasonable upper bound. The use of such elements in the
interior of plated structure, or indeed any structure, should be undertaken with caution. Line
elements will yield only normal stresses in the direction of the axis of the element. In general line
elements will not be aligned with the direction of principal stress.
The current popularity of producing smoothed stress fields in stress plots have hidden dangers. It
hides large disparities in stress in adjacent elements. Large disparities indicate too coarse a mesh. A
more revealing plotting technique is stress contours.
These should be smooth and not jagged. It is evident from Figure 3/4-2 that the contours in the
coarse mesh are not smooth. This might be regarded as an unacceptably coarse mesh. An even
more revealing method with modern postprocessing systems is stress isoband / isosurface plots.
These plots will show a "checkerboard" type of distribution for unacceptable stress distributions.
The stress results from an FEA undertaken in support of design are often plotted in terms of von
Mises stresses, although principal stresses and component stresses are. also sometimes plotted.
There are two potential pitfalls that should be guarded against in interpreting stresses:
1. At nodes on boundaries between membrane elements of different thickness stresses, of
course, cannot be simply averaged. A check should be made to ensure that the software
does not perform averaging blindly in such a configuration.
2. Care should be taken in interpreting stresses at nodes where two-dimensional elements are
not in the same plane. Clearly, simple averaging is not appropriate.

Figure 3/4-2 Stress Contours in Coarse and Fine Meshes

3/4.6 Strain Results

The strain results are usually viewed along with the stress results in a nonlinear analysis, especially
for ductile materials. They can be plotted and interpreted in the same fashion as stress results. The
strain levels can indicate the ductility and degree of deformation. Similar to the stress indexes, some
FEA software allows the use of the engineering strain as input, and some only allow us to use the
true strain. The calculated results are generally from the true strain in the nonlinear analysis.
In a linear analysis, if the stress level is greater than the yield strength, either a Neuber correction
analysis or an elasto-plastic nonlinear analysis should be carried out to evaluate the degree of
plastic deformations. The stress and strain curves from testing or analytical material models (e.g.,
Ramberg-Osgood) can be used in the simulation.
In the low cycle fatigue analysis, the strain results of both in-plane and through-thickness are
examined, assuming 2D shell elements are employed.

3-122
In the analysis has an excessive plastic deformation (e.g., forming and crash), the strain results may
become unreliable once the strain level exceeds the ultimate tensile strain. That is mainly because
the predominant deformation mechanism changes from the grain dislocation to the void nucleation
and void coalescence. It is recommended to use the material model with damage criteria if there is a
need to capture the local failure modes. Note that the strain results are the key input to calculate the
damage parameters (e.g., Johson-Cook model and Gurson model).
In “strain-based design,” the strain results at the critical fracture toughness level are used in the
structural integrity analysis[169,170]. Note that the strain-based design method allows the structure
from high strength steel to sustain a certain level of plastic deformation. It is different from the stress-
design concept and can reduce the conservatism in the structural design.
The following four-strain indexes are generally viewed in the post-process and used for assessment:
1. Yield strain - The strain value at the yielding point.
2. Ultimate tensile strain - The strain value at the necking point.
3. Equivalent strain - The strain value calculated from von Mises stress. Note that in the explicit
solver, the equivalent strain plotted at a given time step is usually a cumulated value [28].
4. Maximum principal strain - The strain vector (i.e., value and direction) for a given coordinate
or a component of the strain tensor.
Note that if material models with multiple damage criteria are used in the simulation, the strain
results may be interpreted based on other strain concepts (e.g., Green strain or Almansi strain). For
some advanced damage models, additional parameters (e.g., Lode angle [171]) work along with the
strain results.
3/4.7 Energy Results
The strain energy is usually viewed in an analysis. For structural design, the strain energy results
can indicate the local stiffness or ability of a structure to sustain the load. For example, in a linear
analysis (e.g., stiffness and vibration), a region with high strain energy results infers that the
structure has a relatively lower stiffness but also indicates the subject high strain energy area is the
predominant load paths. Therefore, in a structural optimization analysis, the strain energy results are
always used as one of the key constraint factors. Not that the design philosophy in a crash-resistant
design is to balance compliance and stiffness. This focus is different from the linear analysis, which
is stiffness focused.
In addition to the strain energy, the following (but not limited to) energy indexes are also viewed in
the nonlinear analysis:
 Internal energy
 Kinetic energy
 Contact or sliding energy
 Hourglass energy
 System damping energy
 Rigidwall energy
The internal energy not only includes the energies for 2D and 3D elements, but also 1D elements,
e.g., spring, damper, and joint. Note that if element erosion or similar element deleting functions are
used, there can be an energy imbalance.
3/4.8 Fracture Results
Nowadays, many FEA software can automatically calculate the stress intensity factor (K), T-stress

3-123
(T), J integral (J), and C* integral (C*) results and export the results to text files. However, other
fracture parameters and related damage parameters, e.g., CTOD and CTOA, require “labor-
intensive” postprocessing. That is mainly because the parameter either have different definitions
(e.g., CTOD) or are under development (including user-defined element and/or materials).
The fracture results from FEA analyses are often used as input for subsequent structural integrity
analyses. For example, K is used for Pairs’ law-based fatigue life calculation; K or J is used for
calculating the fracture ratio (Kr) in the failure assessment diagram (FAD) [12,172–177]. Figure 3/4-3
schematically shows a FAD, where Lr is the load ratio, and the cut-off line refers to the local plastic
collapse state. The fracture results located in the “green” region are safe and acceptable. Note that
in the fitness-for-service design codes [12,13,63,114,178–181], the different levels (or options) of
assessment adopt different fracture results in the fracture resistance analysis.

Figure 3/4-3 Schematic Failure Assessment Diagram (FAD) for Fracture Resistance
Analysis

3/4.9 Fatigue Results

Several FEA software has embedded or optional add-ons for directly calculating fatigue life. These
features may require either a third-party software (e.g., nCode) or manual post-processing for the
fatigue life calculation. As above mentioned in sections 3/4.5 and 3/4.6, the stress or strain results
are usually viewed before calculating the fatigue life. The calculation of fatigue results is relatively
straightforward.
It should be noted that in the case of non-homogeneous materials, other types of stress indexes may
be used for the fatigue life calculation, including:
 Absolute or signed maximum shear stress
 Absolute or signed von Mises stress
 Absolute or signed maximum principal stress
3/4.10 Vibration Results
3/4.10.1 Free Vibration Analysis Results
As discussed in Section 3/3.3.5.4, it is important that the mesh density is selected in accordance
with the mode shapes that are evaluated. A course mesh density can result in accuracy issues when
higher-order mode shapes are being evaluated.
The general observations regarding the sensibility of displacements and continuity of displacement

3-124
should be considered as the first check of analysis results.
3/4.10.2 Frequency Response Vibration Analysis Results
There may be a significant error in Frequency Responses Vibration Analyses results. The FE
prediction accuracy [88] in forced vibration analysis can be generalized as follows:
 Prediction of resonance frequencies for local ship structures to within 5-10% or less.
 Prediction of local response magnitudes to known input forces and machinery sources
within a factor of 2-3, with results often being less than a factor of 2.
The above cautions should be taken into consideration when evaluating the results of this type
of analysis.
3/5 CONCLUSIONS CHECK
This section deals with the final phase, conclusions and recommendations, of a finite element
analysis (FEA). It is necessary to perform these checks to ensure that the loading, strength, and
acceptance criteria are considered in arriving at the conclusions. This is a critical aspect of a finite
element analysis since engineering decisions will typically be based on recommendations contained
in this section. The following sections are grouped into five subsections dealing with various aspects
of FEA conclusions.
3/5.1 FEA Results and Acceptance Criteria
A statement confirming that all analysis procedure quality assessment checks have been executed
satisfactorily should be included.
Finite element analysis is an approximate solution technique, and, in spite of the careful effort, the
results can only be approximations of the real solution. Therefore, the FEA results should always be
validated using an alternative method/s. The alternative method includes comparison with
experimental data, approximate analytical models, textbook and handbook cases, preceding
numerical analyses of similar problems, numerical analysis of a related but simpler problem, and
results for the same problem predicted by a different program (which could be based on a different
numerical method). Many closed-form solutions of structures with simple geometry are available in
handbooks and manuals, which could provide a suitable means for comparison. Numerical analysis
using FEA of similar but simpler models could also be used for comparison. An example could be
the use of a grillage model to check the results of a finite element model of the typical deck structure.
Despite the remarks made in the previous paragraph, the results from alternative solution methods
should also be treated cautiously. Analytical models incorporate idealizations, mistakes may be
made in the calculations, textbooks and handbooks may contain errors, numerical solutions are
subject to errors in coding and in data preparation, and experiments may be improperly performed,
and the results misinterpreted. Therefore, when the FEA results do not compare well with alternative
methods, the possible reasons should be investigated.
The results should be presented so that they can be easily compared with the design/acceptance
criteria. Finite element analysis results are identified based on node numbers and element numbers.
These should be translated into the actual physical problem. For example, in a lattice mast, the
members that do not meet the safety requirements should be highlighted on a figure of the model for
easy identification.
When the FEA results do not meet the acceptance criteria, possible reasons should be explored and
documented. In the case of large deviations, further justification regarding the validity of the FEA
results should be provided.
The results should be assessed based on the knowledge of the physical problem. For analyses of
high importance, an independent assessment should always be done by a qualified and experienced
person.

3-125
3/5.2 Load Assessment
In case of discrepancies in the results, the loading applied to the model should be reviewed as part
of the investigation into the source of the problem. The appropriateness of the types of loads, load
cases, magnitudes, directions, load combinations, load factors, boundary conditions, etc., should be
reviewed.
The loads applied to a finite element model are approximations of the actual loads. The analyst
should provide a general description of the method used to approximate the actual loads. If the load
distribution is simplified to more regular or uniform distribution, this should be justified to ensure that
the simplified load distribution closely approximates the actual distribution in magnitude and
direction. For example, if concentrated forces, at nodes, are used to approximate a pressure
distribution, the calculations used in assigning the values of nodal forces should be explained. When
concentrated forces are used to duplicate pressure, it is important that the load is applied such that
the resultant acts through the center of pressure.
Details on load factors used in the analysis should also be provided. The information on whether the
loads are based upon serviceability limit states or ultimate limit states should also be provided.
Finally, an assessment of the accuracy of the applied loads should be used in describing the results
from the analysis.
3/5.3 Strength/Resistance Assessment
In design situations using traditional methods, the practice is to apply a nominal design load to the
structure and compare the computed stress with some allowable stress. The latter is usually some
fraction of the yield stress or the theoretical buckling stress.
In the modeling process, several assumptions are made, which may, or may not be, conservative.
An assessment of conservatism, or otherwise, should be made particularly in regard to the
underlying assumptions implicit in the design criteria that are being applied. Often design criteria
have evolved with design methods based on hand calculation. Different design criteria may be
appropriate if FEA is used to compute stresses. This factor should be included as part of the
strength/resistance assessment.
In making an assessment of the strength/resistance of the structure based on the results of an FEA,
appropriate allowances should also be made for factors that were not accounted for in the analysis.
Some of these factors include geometric and material imperfections, misalignments, manufacturing
tolerance, initial strains, and corrosion.
The design criteria being applied may implicitly include an allowance for some, or all, of these
factors.
3/5.4 Accuracy Assessment
In assessing the accuracy of FEA results, factors to be considered include the level of detail and
complexity modeled, type of behavior modeled, mesh refinements, etc. In deciding the level of detail,
the analyst would necessarily have omitted some elements of the structure. The effect of these on
the results should be assessed. The limitations of the element type/s used should also be assessed
with respect to its capacity to model the required behavior.
For example, the element type used might model only the membrane actions when both membrane
and bending behavior are significant.
The joints and connections between members might not be properly detailed in the model, making
the model behave in a significantly different way. The effect of the mesh density used on the results
should also be assessed. Simple parametric studies on smaller models may sometimes be
necessary to assess the accuracy of the mesh used in the model.
Performing checks on the numerical accuracy of an FEA is difficult. Generally, reliance is placed on

3-126
a combination of following good modeling practices and on parameters output by the FEA program.
Common parameter outputs include the ratio of the largest and smallest stiffness found in the
stiffness matrix and the so-called residuals.
Unfortunately, satisfactory values for these parameters are necessary, but not sufficient, conditions
for satisfactory numerical performance.
The acceptability, or otherwise, of the ratio of the largest to smallest stiffness depends on the
computer hardware and software, and it is suggested that the guidance provided by the warning and
error messages issued by the FEA program are heeded.
The frequency response vibration analysis range requirements discussed in Section 3/2.8.3 and
accuracy issues discussed in Section 3/4.10 should be taken into consideration.
3/5.5 Overall Assessment
All of the above-described factors should be used in conducting an overall assessment of the FEA.
The results of this overall assessment should be included as part of the documentation. Deviations,
if any, from the actual response should be justified.
Recommendations, if any, for future FEA should be clearly stated. If there is an anticipated
continuation for the project at a later date, information on all computer files, documentation, etc.
should be documented.

3-127
 BENCHMARK PROBLEMS FOR ASSESSING FEA SOFTWARE
4/1 INTRODUCTION
The assessment methodology presented in PART 2 includes a requirement that suitable FEA
software be used. The determination of the suitability of a particular FEA code should involve,
among other things, an assessment of its capability to analyze the types of problems that will be
applied. This part describes the development and application of a series of standard benchmark test
problems that can be used to assess the suitability of new, or significantly modified, FEA software for
ship structure analysis.
As a means of qualifying FEA software, the benchmarks represent a category of tests between that
of large-scale validation efforts and that of smaller-scale verification problems. The actual structural
behavior of even the simplest component depends on such a large number of variables of varying
complexity that isolating the response modeled by FEA codes is extremely difficult. As such, large
scale validation of FEA software is typically very complex and expensive, often requiring comparison
of FEA predictions with physical test results. Although such validation testing may be a requirement
for certain critical structure applications, it is not a practical approach for assessing FEA software on
a routine basis.
Most FEA software developers perform verification tests as part of their internal quality assurance
procedures. For example, the verification test set for the ANSYS FEA program consists of over 5500
test cases at revision 5.1. Some software developers publish and / or make available a subset of the
tests in the form of examples or verification manuals. Other developers include "textbook" verification
examples in their marketing media. Verification problems of this sort are usually simple and small-
scale in character and typically have closed-form theoretical solutions. They are generally designed
to test a very specific aspect of the FEA code, such as the numerical performance of a certain type
of element in a certain geometry, loading condition, and type of analysis. However, the verification
problems rarely resemble "real life" engineering problems involving irregular geometries with large
numbers of element types, in various shapes and sizes, combined with several load types and
boundary conditions. Thus, while verification problems of the type described above are a necessary
step in verifying and validating FEA software, they are not sufficient on their own.
The benchmark problems presented here are intended to represent the next step in ensuring that
the candidate FEA software is appropriate for the FEA of linear elastic ship structure. The
benchmarks are summarized in Figure 4/1-1 and cover a range of typical problems and
requirements encountered in "real life" ship structure FEAs. The problems involve simple
configurations of a number of representative ship structures but are detailed enough to retain the key
characteristics of the structural assembly or detail. The problems typically require that several types
of elements, materials, and loads be used in combination. An attempt has been made to design the
benchmarks such that, collectively, all key features that determine the quality of FEA packages are
addressed. The benchmark problems are described in 4/2 with complete details given in Appendix
D.
The benchmarks are designed to exercise the FEA software rigorously without making the
evaluation process overly demanding. The problem size has been limited to a maximum of 200
nodes to ensure that the process of benchmarking new and modified software is not onerous. The
200-node limit should also allow, in some cases, for the user to test demonstration or evaluation
versions of FEA software. Such versions are usually based on the "full" versions of the FEA code,
but typically have limits on the number of nodes and elements that can be modeled. These are
usually available from the FEA software developer at a small nominal fee to allow testing and
evaluation prior to making a larger financial commitment.
The benchmarks do not have closed-form theoretical solutions. Instead, the results from analyzing
the benchmark problems using three well known FEA software programs are used to establish the

4-1
reference benchmark results. The three programs used were ANSYS, MSC/ NASTRAN, and
ALGOR and are described in 4/3. The presentation and discussion of the benchmark results are
included in Appendix D.
Care has been taken to ensure that the test models for the benchmark problems are sufficiently
detailed or refined that the results approach a converged solution. Element formulations, stress
averaging / extrapolation algorithms, and other aspects of FEA software performance tend to be
optimized for ideal configurations. Testing different FEA software of an ideal configuration (e.g., a
rectangular plate with uniform rectangular elements) will tend to give virtually identical results.
However, once the FEA model deviates from an ideal configuration, as is the case for the
benchmarks, differences in the results manifest themselves. In these circumstances, the rate of
convergence of results from different FEA programs may differ. Ensuring that the results obtained by
the test models are near a converged solution should minimize any discrepancies that can be
attributed to poor mesh design of the benchmark test models.
New or significantly modified, FEA software can be evaluated by exercising the software with the
benchmark problems and comparing the results obtained with the reference benchmark results. The
process by which this should be accomplished is presented in 4/4.
WARNING
The benchmark problems and associated FEA models presented in this document are
intended for the express purpose of evaluating FEA software for ship structural analysis
applications. While attempts have been made to ensure that the FEA models follow good
modeling practices, they should not necessarily be regarded as appropriate for any purpose
other than that for which they are intended.

4-2
 Figure 4/1-1 Summary of Ship Structure FEA Benchmark Problems

4/2 THE BENCHMARK PROBLEMS

The ship structure FEA benchmarks include the following problems:
1. Reinforced Deck Opening
2. Stiffened Panel
3. Vibration Isolation System
4. Mast
5. Bracket Detail
Figure 4/1-1 summarizes the main modeling and analysis features that the benchmarks are intended
to test. The following sections provide a summary description of the benchmark test problems.
Complete details of the benchmark problems are presented in Appendix D.
4/2.1 BM-l Reinforced Deck Opening
Openings and penetrations are among the most commonly encountered sources of high-stress
levels in surface ship structures. In most cases, the openings are reinforced by coamings or insert
plates to attenuate the resultant stress concentrations. FEA may be required to evaluate the stress
levels and the effectiveness of the reinforcement technique. This benchmark tests the capability of

4-3
FEA packages to analyze this category of a ship structure problem and is shown in Figure 4/2-1. The
benchmark tests the FEA program's capability to analyze a plane stress concentration problem using
either 4-node or 8-node shell elements. However, it goes beyond the classical hole-in-a- plate
problem by including two plate thicknesses for the deck and the reinforcement insert plate, and by
including stiffeners in the plane of the deck.

Figure 4/2-1 Benchmark Problem BM-1 Reinforced Deck Opening

4/2.2 BM-2 Stiffened Panel

Stiffened panels are the most common structural component in ships. This benchmark tests the
capability of FEA packages to analyze this type of structure using the various plate and stiffener
element modeling techniques. These include :
 4-node shell elements for plate and in-plane beam elements for stiffeners
 4-node shell elements for plate and off-set beam elements for stiffeners
 4-node shell elements for plate and stiffeners
 8-node shell elements for plate and stiffeners
Both static and modal analyses are conducted for each model. The static analysis involves surface
pressure loading causing out-of-plane panel bending under symmetric boundary conditions (i.e.,
quarter model). The modal analysis tests the capability of the program for calculating natural
frequencies and mode shapes under symmetric and antisymmetric boundary conditions.

4-4
 Figure 4/2-2 Benchmark Problem BM-2: Stiffened Panel

4/2.3 BM-3 Vibration isolation System

Vibration isolation systems are often required for ship equipment and machinery. FEA analyses may
be used to optimize the isolation system and ensure that vibration and shock design criteria are
achieved. This benchmark considers a 12 degree of freedom system consisting of a generator that is
mounted and isolated on a raft structure, which is, in turn, isolated from the foundation structure. The
problem is summarized in Figure 4/2-3. Some of the key testing features include of this benchmark
include:
 Modal analysis;
 Point mass including rotational inertia terms (to the model generator)
 Spring elements with stiffness in three directions; and
 "Rigid" beam elements connecting generator mass and isolator springs to the raft.

4-5
 Figure 4/2-3 Benchmark Problem BM-3: Vibration Isolation System

4/2.4 BM-4 Mast Structure

Mast structures on ships must be designed to withstand environmental loads (wind and ship
motions). Masts on naval ships usually have additional requirements for resisting shock and blast
loading. The mast benchmark problem is summarized in Figure 4/2-4, and the key modeling and
testing features include :
 Beam elements (with axial and bending stiffness) for main legs and polemast
 Axial line elements (spar, truss, rod) for braces
 Point mass elements for equipment "payloads"
 Inertial loading in three directions combined with nodal force loading
 Two materials (steel and aluminum)
 Modal analysis
While the benchmark problem is that of a lattice mast structure, it can be used to assess the FEA
program's capabilities for modeling similar frame or truss-like structures such as booms and derricks,

4-6
especially where beam and spar elements are used in combinations.

Figure 4/2-4 Benchmark Problem BM-4 : Mast Structure

4/2.5 BM-5 Bracket Connection Detail.

Welded connection details on ships are subject to fatigue loading. Poorly designed or constructed
details can lead to premature fatigue failure. Finite element methods are frequently used to calculate
fatigue stresses and to aid in the development of improved detail geometry and configurations. This
benchmark problem is summarized in Figure 4/2-5. Some of the key modeling and testing features
of this benchmark include:
 3·0 geometry with shell elements of varying thicknesses;
 Axial line elements for bulkheads, deck, and flange of the bracket; Transition from coarse to
fine mesh at the bracket weld;
 Prescribed nonzero nodal displacement boundary conditions.
The latter feature was included since, in many cases, the boundary conditions for a detail FEA are
obtained from displacements and loads derived from a global FEA.
This particular bracket detail problem is complicated by the existence of a stress singularity at the
end corner or toe of the bracket. In a linear elastic analysis, the stress at this point is theoretically
infinite. Refining the finite element mesh gives
progressively higher stresses, which are meaningless. One method which is commonly used to get

4-7
around this problem is to use the so-called "hot spot" stress. In calculating the hot spot stress no
account is taken of the weld geometry, and in an idealized finite element representation (ignoring the
weld) the stress is equal to the value at about one plate thickness from the corner (Chalmers, 1993).

Figure 4/2-5 Benchmark Problem BM-5 : Bracket Detail

4/3 THE BENCHMARK TEST FEA PROGRAMS

As previously mentioned, the benchmark problems do not have readily obtainable theoretical
solutions. Instead, the results from analyzing the benchmark problems using three well known FEA
software programs are used to establish the reference benchmark results. The three programs used
were ANSYS, MSC / NASTRAN, and ALGOR.
The ANSYS FEA program is developed and marketed by ANSYS Inc. of Houston, PA. ANSYS is a
mature, general-purpose FEA program that has been commercially available on various computer
platforms since 1970. It includes extensive analysis capabilities, a large, comprehensive library of
elements, and extensive pre- and post-processing capabilities. The ANSYS Version 5.1 program
was run on a DEC 3000 workstation for the benchmark test cases.
The MSC / NASTRAN FEA program is developed and marketed by The MacNeal- Schwendler
Corporation, Los Angeles, CA. Traditionally it has been most widely used by the aerospace industry,
having evolved from the National Aeronautics and Space Administration (NASA). MSC / NASTRAN
is a very comprehensive and mature FEA program that has been commercially available for several
decades. It is, to some extent, regarded, along with ANSYS, as the industry standard. MSC /
NASTRAN for Windows 1.0 on an IBM 486 PC was used for the benchmarks.
The ALGOR FEA program is developed and marketed by ALGOR Inc., Pittsburgh, PA. It was one of
the first FEA programs to be developed, especially for the personal computer, and has become one
of the most popular FEA programs for PC applications. The program features a relatively wide range
of modeling and analysis capabilities.

4-8
4/4 APPLICATION OF BENCHMARKS FOR ASSESSING FEA SOFTWARE
The intended application of the benchmarks is to provide a methodology for assessing FEA
software. This assessment consists of modeling and analyzing the benchmark problem with the FEA
software and comparing the results with those obtained by the reference FEA programs as
presented in Appendix D. The data files for the benchmark problems in ANSYS, NASTRAN, and
ALGOR formats may be obtained by contacting the Ship Structure Committee.
As was discovered in the benchmark results of the three reference FEA programs, there are liable to
be differences between the results obtained by different FEA software packages. The differences
may arise from a multitude of factors ranging from the numerical accuracy of the hardware and
software platforms to different element formulations, solution algorithms, and results presentation
techniques to actual errors or limitations in the FEA software. The question that arises is how much
variation or deviation from the reference results is acceptable.
The authors suggest the following approach be used to judge the acceptability or otherwise of the
benchmark results for any FEA software:
1. Result differences less than 2% with respect to the reference FEA software results for
displacements, reaction forces, and lower mode natural frequencies are considered
acceptable. The 2% limit is generally within what would normally be the required engineering
accuracy for these types of problems.
2. Result differences between 2% and 5% are probably acceptable for beam and plate element
stress results and higher mode natural frequencies. However, the user should endeavor to
ensure that there are plausible explanations when differences get much past 2%. This may
involve further testing of the problem by, for example, refining the FEA mesh or switching the
analysis options to I from the defaults used by the FEA program.
3. Result differences greater than 5 % should be considered abnormal and require an
explanation. If a reason cannot be found, the developer of the FEA software should be
contacted and requested to investigate the difference. Where no explanation exists, the FEA
software should probably be viewed as suspect for the particular type of analysis covered by
the benchmark problem.
Particular attention should be paid to ensure that the proper loads and boundary conditions have
been applied and that the stress contours, deformed shape or mode shapes (depending on what is
applicable) are consistent with the reference results. The user should also be sure of the default
analysis assumptions and solution techniques used by the software. These can be especially
important for problems where transverse shear effects need to be considered, or when performing
modal analyses. The user should also be aware of how the FEA software extrapolates and I or
averages plate element stress results at nodes.
The benchmarks are necessary but by no means a complete method of validating an FEA program.
The benchmarks primarily check that a particular FEA code will perform and produce results that are
consistent with the three reference FEA codes. However, it is strongly recommended that users of
new or significantly modified FEA software become fully aware of all features and limitations of that
program for the particular applications involved. This should include testing the software on
simplified versions of the main problems of interest in order to build confidence in the modeling
approach, choice of elements, mesh densities, etc. as discussed in Part 3, Section 1.3.

4-9
 ADVANCED ANALYSIS SAMPLE APPLICATIONS
5/1 IMPACT AND PLASTICITY
5/1.1 Introduction
The FE model of a ship bow structure under frontal collision (see Figure 5/1-1) is described in this
section. The objective of this study is to characterize the collision resistance of the bow structure and
evaluate whether the ship structure has sufficient resistance to collision accidents. The nonlinear
finite element analysis is conducted to assess the structural strength during the event. Note that the
examples presented in this section were initially developed by researchers (e.g., Dr. Martin
Storheim) from NTNU – the Norwegian University of Science of Technology and now adopted by
DNVGL-RP-C208 [106].

Figure 5/1-1 FE Models of Ship Bow Under Frontal Collision

5/1.2 Engineering Model

5/1.2.1 FEA Software
LS-DYNA® shared memory parallel (SMP) version R8.0.0 with double precision was employed for
calculating the structure response. Altair® HyperMesh and HyperView are used for pre- and post-
processing, respectively.
5/1.2.2 System Units
In this study, the unit system for this model is defined as follows:
1. Length millimeter [mm]
2. Mass metric tons [t]

5-1
 3. Time second [s]
4. Force Newton [N]
5. Pressure megapascal [MPa] or [N/mm2]
6. Stress megapascal [MPa] or [N/mm2]
7. Strain no unit [%]
8. Velocity millimeter per second [mm/s]
9. Energy millijoule [mmJ]
5/1.2.3 Geometry assumptions
As shown in Figure 5/1-1, only bow structure is modeled and the rest of the ship structure is not
included. Note that this is a partial ship structural analysis, and the geometry assumptions of
stiffeners, supporting members, and openings follow the rules of DNVGL-CG-0127 [9].
5/1.2.4 Material properties
The bow structure is built of the S235 and S355 grade steel. The stress-strain relationship follows
the Hollomon-type power-law relationship as follows:
𝜎 = 𝜎𝑦𝑖𝑒𝑙𝑑 𝑖𝑓 𝜀𝑝 ≤ 𝜀𝑝,𝑦𝑖𝑒𝑙𝑑−2
{ 𝑛 Eq. (5.1)
𝜎 = 𝐾(𝜀𝑝,𝑒𝑓𝑓 + 𝜀𝑝 ) 𝑖𝑓 𝜀𝑝 > 𝜀𝑝,𝑦𝑖𝑒𝑙𝑑−2
where n is the strain hardening exponent; K is the strength index or strength coefficient;  and
with subscripts are the strain and stress indexes. To account for the presence of the strain
plateau, the work hardening is delayed until the plastic strain level reaches the plateau strain
𝜀𝑝,𝑦𝑖𝑒𝑙𝑑−2 . The effective plastic strain in the above equation is defined as
𝜎𝑦𝑖𝑒𝑙𝑑 1/𝑛
𝜀𝑝,𝑒𝑓𝑓 = 𝜀𝑝,𝑦𝑖𝑒𝑙𝑑−1 − 𝜀𝑝,𝑦𝑖𝑒𝑙𝑑−2 = ( ) − 𝜀𝑝,𝑦𝑖𝑒𝑙𝑑−2 Eq. (5.2)
𝐾

where 𝜀𝑝,𝑦𝑖𝑒𝑙𝑑−1 is the strain at the initial yield point. Figure 5/1-2 schematically shows definitions of
stress and strain indexes in the equation mentioned above. The strain hardening exponent n is
defined as:
𝑛 = 𝑙𝑛(1 + 𝜀𝑈𝑇𝑆 ) Eq. (5.3)
where UTS is the ultimate tensile strain.
The engineering stress-strain data are converted to true stress-strain data by using the following
equation.

𝜀𝑡𝑟𝑢𝑒 = 𝑙𝑛(1 + 𝜀𝑒𝑛𝑔 ) Eq. (5.4)

𝜎𝑡𝑟𝑢𝑒 = 𝜎𝑒𝑛𝑔 𝑒𝑥𝑝(𝜀𝑡𝑟𝑢𝑒 ) Eq. (5.5)
The parameters defining true stress-strain data are as follows:
1. Young’s modulus E [MPa] 210,000
2. Poisson’s ratio v [-] 0.3
3. Proportional stress prop [MPa] 285.8 (S235), 384 (S355)
4. Yield strength yield [MPa] 318.9 (S235), 428.4 (S355)
5. Yield strength – 2 yield,2 [MPa] 328.6 (S235), 439.3 (S355)
6. Yield strain – 1 yield-1 [%] 0. 4%

5-2
 7. Yield strain – 2 yield-2 [%] 2% (S235), 1.5% (S355)
8. Strength index K [MPa] 700 (S235), 900 (S355)
9. Ultimate tensile strength UTS [%] 2% (S235), 1.5% (S355)
10. Strain hardening exponent n [-] 0.166
The density of steel is assumed to be 7.85×10-9 t/mm3 in the study.

Figure 5/1-2 Schematic Stress and Strain Relationship

With Yield Plateau and Strain Hardening Effects

The true stress and true strain curves of S235 and S355 for FE input are shown in Figure 5/1-3. Two
curves are defined using the LS-DYNA command card – *DEFINE_CURVE _TITLE. The material
card number 24 (*MAT_024 – *MAT_PIECEWISE_LINEAR_ PLASTICITY_TITLE) is employed
associated with two defined stress and strain curves.

Figure 5/1-3 True Stress and True Strain Curves of S235 and S355 Grade Steel
for LS-DYNA Input

5-3
5/1.2.5 Loading and Boundary Conditions
The bow structure is assumed to be fixed and impacted by a moving rigid-wall. Figure 5/1-4
schematically shows the boundary condition and applied load. All degrees of freedom of the bow
structure boundaries are constrained. The rigid wall is moving toward the bow structure at a speed of
4 m/s along the x-direction, where the other five degrees of freedom are constrained. The initial
velocity is defined by using command card *BOUNDARY_PRESCRIBED_MOTION_RIGID.
The penalty contact algorithm is used to define the contact among components. Two surface-to-
surface contact pairs (*CONTACT_AUTOMATIC_SURFACE_TO_SUR-FACE_ID) are defined: (1)
forecastle to rigid wall, and (2) bulb to a rigid wall. One single surface contact
(*CONTACT_AUTOMATIC_SINGLE_SURFACE_ID) is defined for all bow structures. The friction
effect of contact is not considered, i.e., static and dynamic friction coefficient (fstatic and fdynamic) = 0.

Figure 5/1-4 Schematic Illustration of Boundary Conditions and Applied Load

5/1.3 FE Models and Simulation Setups

5/1.3.1 Element
The bow structure is modeling by using 2D shell elements. The key parameters are summarized as
follows:
 The Belytschko-Tsay reduced integration co-rotational formulation is adopted, i.e., ELFORM
= 2.
 Five integration points through-thickness are defined to capture the variations of stress and
strain gradients, i.e., NIP = 5.
 The shear correction factor is about 5/6, i.e., SHRF = 0.833.

5-4
5/1.3.2 Mesh design and quality control
The mesh was designed and tuned up by following DNVGL-CG-0127 [9], and the average element
size is about 50 mm. The mesh of the FE model is revised based on the criteria shown in Table
5/1-1. Note the criteria are developed based on the description and discussion in PART 3. The mesh
quality of the revised model meets all the criteria listed in Table 5/1-1.
Table 5/1-1 Crash problem with 50mm element size

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 50 43.33 36.66 30 28.33
2 max length 50 62 73.2 90 112.9
3 aspect ratio 1 2 4 5 10
4 warpage 0 5 15 20 40
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 3 4.5 6 9
13 taper 0 0.2 0.5 0.6 0.9

The bow structure FE model is constructed by using 336,261 nodes and 344,354 elements. The
numbers of quad and tria elements are 335,146 and 9,208, respectively. It means there are 2.67% of
tria elements in the model, which is less than the 6% “Fail” criterion in Table 5/1-1.
5/1.3.3 Include files
The FE models are decomposed into several key files:
 Bow structure
 Rigid wall
 Solution and output control
 System control (e.g., number of CPUs, size of memory, type of solver)
A master file is created to call all these key files using including file function – *INCLUDE.
5/1.4 Analysis Control and Solution Options
Several advanced features are also activated to control the simulation accuracy. For example:
 Hourglass control uses the stiffness-based method with Flanagan-Belytschko integration
formulations, i.e., IHQ = 4 and QH = 0.03.
 Warping stiffness is captured by using Belytschko-Wong-Chiang formulations with full
projection, i.e., BWC = 1, and PROJ = 1.
 Critical warpage angle of shell element is 20 degrees, i.e., WRPANG = 20.
 Mass scaling is applied to the critical element to reducing computational time, i.e., dt2ms = -
3.0×10-6.

5-5
The following results are exported to ASCII text files at every 0.001s:
 Global data – *DATABASE_GLSTAT
 Material energy – *DATABASE_MATSUM
 Resultant interface forces – *DATABASE_RCFORC
 Sliding interface energy – *DATABASE_SLEOUT
 SPC fixation reaction force – *DATABASE_SPCFORC
 Boundary condition forces and energy – *DATABASE_BNDOUT
 Rigid body data – *DATABASE_RBDOUT
The binary results (*.d3plot files) of the entire model and contact interface are generated at every
0.1s, i.e., *DATABASE_BINARY_D3PLOT and *DATABASE_BIN-ARY_INTFOR.
5/1.5 Results
The results of von Mises stress, first principal stress, and effective plastic strain at the indentation of
2.0 m are shown in Figure 5/1-5. Figure 5/1-5 (a) shows that the bow structure has a local plastic
deformation in the forecastle and bulb. It should be noted that the inertia effect is not considered in
this study, which may induce different stress-strain distributions and failure mechanisms.

Figure 5/1-5 Results of Stress and Strain at the Indentation of 2.0 m

5/1.5.1 Impact Resistance

The impact force and indentation curves of the bow structure are shown in Figure 5/1-6. Note that
the impact forces of bulb and forecastle to rigidwall are captured from the contact pairs. The design
curve for a bow structure under the frontal impact is from standard DNVGL-RP-C204 [107] is also
presented in Figure 5/1-6. It is seen that the resistance of the forecastle is similar to the design
target at the indentation within 0.8m and less than the design target at the indentation between 0.8m
and 2.0m. Compared with the forecastle, the bulb adsorbs the predominant impact energies and is
the main structure to resist impact loads.

5-6
It should be noted that DNVGL-RP-C204 [107] does not consider the bow structure with a bulb. The
current bow structure with both forecastle and bulb meets the design requirement of impact
resistance.

Figure 5/1-6 Force-Indentation Curves of the Bow Structures against the Design Curve
from DNVGL-RP-C204

5/1.5.2 Failure mechanism

By assuming a flat rigid wall, the forecastle structure first contacts the rigid wall and then the bulb
structure after time step of 0.01 s. The bulb tip flattens under the frontal impact shown in Figure
5/1-7. The predominant failure mechanisms are the local buckling of stringers and the plastic folding
of structures. Figure 5/1-8 shows the local failure mechanisms of the bulb at two different time steps.
The similar failure mechanisms are also observed in the forecastle.

Figure 5/1-7 Flattening of the Bulb Tip under Frontal Impact of Rigid Wall

5-7
 Figure 5/1-8 Local Failure Mechanism of the Bulb Tip: Folding and Buckling

5/1.6 Discussions and Comments

The present FE model of the bow structure has been tuned up and can produce results with
reasonable accuracies. In [40], a similar model is calculated by employing ABAQUS®. The results of
the impact load indentations and stress/strain distributions from ABAQUS® are identical to those
reported in this document.
This study was carried out based on simplified approaches of boundary conditions and material
models. For example, the boundary condition of the present model is that a moving rigid wall
impacts against a fixed bow structure. This setup does not consider the inertia effect of the ship
structure, which can be different from the real collision cases.
In general, complete material behavior up to fracture should be modeled to capture the response of
impact resistance and local failure mechanisms. That includes the stress-strain relation, strain rate
effect, damage criteria, and, in special cases, thermal response. As discussed in PART 3, different
types of damage criteria have been developed, e.g., strain-state dependence of fracture. The
following sections discuss the impact of these factors on the simulation results.
Note that extra care is needed in preparing FE models if using advanced material models. For
example, caution needs to be made concerning the material failure models that are inherent in finite
element codes. Codes such as LS-DYNA have many different material response models, often
including "failure." Code users often do not understand the question of what models to use for their
problems, and how to obtain required parameters. The analyst needs to understand the
assumptions and limitations built into these failure models. Before applying them for a practical
analysis, their validity needs to be demonstrated through analysis of the results of experiments
conducted with the same materials at the same rate of loading as in the analysis.
5/1.6.1 Work hardening effect
The work hardening controls where the strain localizes. Nowadays, a great number of plastic
hardening models have been developed and implemented in the FEA software. Three material
models with isotropic hardening are generally used in the ship structure design: (1) bilinear model
per DNVGL-RP-C208 [106], (2) power-law model without a yield plateau, and (3) power-law model
with a yield plateau (see Eq. 5.1). The hardening model is implemented based on the experimental

5-8
data. It should be noted that since the stress-strain data from testing can include nonuniformity in
results or test to test or variability, the yield strength of materials for ship structures is determined
from the specified minimum value. That is typically at a fifth percentile of the probability distribution
of testing data.
In the literature (e.g.,[154,155,157,158,182–186]), the effect of the work hardening has been
systematically investigated in the damage extent in impact events. Paik [154,155] reported that the
shape of stress-strain curves directly affects the simulation results, such as the local failure
mechanism. Later, Hogstrom [157,182,183] carried out a parametric study on the work hardening
effects associated with the damage model in impact events. It is found that not only the load-
indentation relationship is directly dependent on the work hardening, but also damage or failure
locations. Storheim [158,184–186] also investigated the effect of the work hardening variance in a
full-scale ship collision scenario. Both upper- and low-bound stress-strain relationship was
considered for a given strain hardening value. The upper bound strength is 1.15 times higher than
the lower bound one. From the load indentation curves, the work hardening value has relatively low
influences on the bow structure, but it has pronounced influences on the ship's side. The difference
in impact resistance is mainly because of the governing failure mechanism. The governing
deformation (e.g., buckling, and plastic folding) continuously develops and evenly spreads in the
bow structure, whereas the governing deformation only happens when the failure propagates to
stiffeners or vertical web frames in the ship side.
5/1.6.2 Strain rate effects
The strain rate is the velocity of deformation and defined in an explicit form – time derivate of strain:
𝜀̇ = 𝜀 ⁄𝑡 Eq. (5.6)
Most of the steel materials show dependency on strain rate, which affects the yield strength, ultimate
tensile strength, strain hardening, elongation limit, and fracture limit. In general, the yield strength,
ultimate tensile strength, and strain hardening increase with the increase of strain rates, thus the
flow stress increases to resist further deformation. Once the necking starts, the strain rate increases
with the presence of necking localization. On the other hand, the elongation limit (also known as
ductility) decreases as the increase of strain rate. The fracture limit does not show a clear correlation
with the strain rate based on the experimental results. That can be because the fracture failure
mechanism is complicated and sometimes also driven by other factors (e.g., level of triaxiality and
Lode angles).
The strain rate effect is usually considered in the impact, explosion, or ballistic scenarios. Storheim
and Amdahl [186] evaluated ship structures under different impact speed and pointed out two crucial
points of strain rates in the ship collision:
1. The local instantaneous strain rate can be significant even at a low impact speed.
2. The strain rate is strongly mesh-size-dependent. For example, at the same impact speed – 4
m/s, the strain rate for an average mesh size of 100 mm can range from 5 – 15 s-1, but the
strain rate for a mesh size of 2 mm can be over 400 s-1.
Note that the high strain rates often have a short time duration and usually show in the local necking
zone.
Two material models are widely adopted in the FEA simulations: (1) Cowper-Symonds [187] and
Johnson-Cook [188]. The Cowper-Symonds model defines the dynamic stress is proportional to the
static stress with a scaling factor – known as dynamic hardening factor:
𝜀̇ 1/𝑝
𝜎𝑑𝑦𝑛𝑎𝑚𝑖𝑐 = [1 + ( ) ] 𝜎𝑠𝑡𝑎𝑡𝑖𝑐 Eq. (5.7)
𝐶

where 𝜀̇ is the strain rate; C and p are parameters calibrated from experimental data. The standard
DNVGL-RP-C208 [106] recommended calibrating the dynamic hardening factor to the maximum

5-9
expected stress and strain. If there are no experimental data, parameters C = 4000 s-1 and p = 5 for
the common offshore steel.
The Johnson-Cook model defines stress as a function of three factors: strain hardening, strain rate,
and temperature.
𝑛 𝑚
𝜎 = [𝐾(𝜀𝑝,𝑒𝑓𝑓 + 𝜀𝑝 ) ](1 + 𝐷 ln 𝜀)(1 − 𝑇 ∗ ) Eq. (5.8)
where K and n are the same as those in Eq. 5.8; D and m are coefficients; T* is the temperature
parameter relative to a reference temperature (e.g., room temperature). This model is usually
adopted for ballistic simulations where the thermal load is coupling with mechanical loading.
5/1.6.3 Damage criteria
As mentioned above, the present study does not consider the material damage, and the stress or
strain values are calculated over the elongation or fracture limit. The collision resistance of the ship
structure can be overestimated without accounting for structural failure or damage. There are
different damage criteria developed (see PART 3), e.g., stress-based, strain-based, and energy-
based. The strain-based damage criteria are usually used to capture the structural failure in the
impact event.
The maximum principal strain is usually used as the strain-based damage criteria, which assumes
the predominant damage mechanism is caused by tensile failure. The procedure to calibrate the
maximum principal strain is presented in [40,107]. In LS-DYNA®, the number 123rd material model –
*MAT_123, which is similar to *MAT24 except for the additional maximum principal strain criteria.
For two grade steels in this study, the calibrated damage criteria are as follows:
 Equivalent plastic strain 20.1% (S235) 19.6% (S355)
 Maximum principal strain 17.7% (S235) 17.4% (S355)
In the FEA software, the damage criteria usually collaborate with the element erosion (also known as
deletion) function. For example, the shell element has five integration points through-thickness in this
study, and the element will be deleted once all the integration points have failed.
5/1.6.4 Fracture and mesh dependence
The estimation of damage also shows strong mesh-size-dependency, similar to that of the strain
rate. In [189,190], the authors carried out a series of numerical studies and discussed the fracture
and mesh dependency. For example, Figure 5/1-9, reproduced from [189,190], shows a deformed
plate section at the maximum indentation level in both testing and simulation. The figure shows the
finer mesh model can accurately capture both the deformed shape and fracture path, but the coarser
mesh model can only capture the fracture initiation point.

5-10
Figure 5/1-9 (a) Testing Results of Failure Mode; (b) Numerical Results of Failure Modes
and von Mises Stress Distributions for Four Different Mesh Sizes (i.e., Length/Thickness
= 1, 2, 5, and 10)

5/2 FRACTURE AND FATIGUE

5/2.1 Introduction
Several years ago, oil shipping companies started using double-hull tankers to replace single-hull
tankers based on Annex I of MARPOL (Maritime Pollution) convention – “phase out single-hull
tankers around the world by 2026” [191]. A client built a double bottom hull tank in the year 2000,
and cracking zones have been observed throughout the double bottom after five-year services. This
study investigates the cause of cracking in the double bottom using a fatigue analysis.
Figure 5/2-1 shows a crack is located at the toe of a welded connections of a double bottom floor
stiffener. It should be noted that the cracking plane is atypical for longitudinal loading. In Figure
5/2-2, the cracking incidents observed in the double bottom are highlighted in the green shading. It is
seen that the cracking spreads throughout the double bottom and is away from the bulkhead.

5-11
 Figure 5/2-1 Cracking Occurring on a Horizontal Plane at Floor Stiffener to Longitudinal
Connection

Figure 5/2-2 Locations of Cracking Incidents in the Double Bottom

The objective of this study is to understand the root cause of the cracking by carrying out numerical
investigations on structural behaviors under two different loading scenarios. The investigations
include two steps of analyses: static linear elastic FEA analysis and fatigue analysis, where the
fatigue cyclic loadings are interpreted based upon the stress results from the Step-1.
5/2.2 Engineering model
5/2.2.1 FEA software
The general FEA software – ANSYS® Mechanical version 7.0 with double precision was employed
for calculating the stress distributions. FlawCheck® [192] developed by BMT-Canada is used to

5-12
carry out the fatigue analysis.
FlawCheck® is a comprehensive integrity assessment tool for metal structures designed to support
the development and implementation of integrity assurance and maintenance programs. The current
study uses FEA-based structural loading data to estimate the fatigue crack growth and to determine
the durability life. In this study, the fatigue crack growth and fatigue life are evaluated by following
the procedures described in BS 7910-2013 [12]. Figure 5/2-3 shows the FlawCheck’s GUI (graphic
user interface) and fatigue analysis module.

Figure 5/2-3 FlawCheck GUI® and Fatigue Analysis Module

5/2.2.2 System units

In this study, the metric unit system is used:
1. Length millimeter [mm]
2. Mass metric tons [t]
3. Time second [s]
4. Force Newton [N]
5. Pressure megapascal [MPa] or [N/mm2]
6. Stress megapascal [MPa] or [N/mm2]
7. Strain no unit [%]
8. Velocity millimeter per second [mm/s]
9. Energy millijoule [mmJ]
10. Fatigue life year [y]
5/2.2.3 Geometry assumptions
The geometric dimensions of this double bottom oil tanker are as follows:
1. Length B.P.: 245 m

5-13
 2. Depth MLD: 20.4 m
3. Breadth MLD: 44.8 m
A typical hold of the vessel is modeled (see Figure 5/2-4), which includes rough scantlings, a
longitudinal plane of symmetry, and a forward and aft bulkhead.

Figure 5/2-4 FE Model of a Typical Hold of a Vessel

Another level of the FE model with finer meshing (local submodel) is prepared to capture the
variations of stress gradients in the hot spot. The loading and boundary conditions for the local
model are interpreted from the results of the above-mentioned coarse meshing model. Figure 5/2-5
shows the finer meshing model and locations of hot spots.
Note that in the global analysis, only one hold model is used (see Figure 5/2-4), and Figure 5/2-5
shows the hierarchy structure the different levels of FE meshing.

Figure 5/2-5 Local FE Model with Finer Meshing for Capturing Variations
of Stress Gradients Close to Hot Spots

5/2.2.4 Material properties

For the linear elastic FE analysis, the material properties of ship grade steel are listed as follows:

5-14
 1. Young’s modulus E [MPa] 210,000
2. Poisson’s ratio v [-] 0.3
A linear SN curve for a corrosive environment is used for the fatigue analysis, shown in Figure 5/2-6.
Note that the failure locations are assumed in ballast tanks. The slope and intercept points of the SN
curve are 3 and 11.705, respectively. The yield strength of the base metal is 360 MPa, and the limit
of thickness effect is 25 mm.

Figure 5/2-6 SN Curve of Ballast Steel for Corrosive Environment

with Slope = 3 and Intercept Point = 11.705

5/2.2.5 Loading and boundary conditions

In the FEA analysis, three different load cases are considered in this study:
 Hull girder bending 2.1×107 N·mm
 Internal pressure – ballast see Figure 5/2-7 Illustrations of the pressure loads of
(a) Ballast and (b) Cargo loaded cases

 External pressure – loaded see Figure 5/2-7 Illustrations of the pressure loads of
(a) Ballast and (b) Cargo loaded cases


5-15
 Figure 5/2-7 Illustrations of the pressure loads of (a) Ballast and (b) Cargo loaded cases

The loading and fixations of three load cases are schematically shown in Figure 5/2-8 Loading and
Fixation Conditions of Three Load Cases. In Figure 5/2-7 (a), the yellow highlighted plane defines
the longitudinal symmetry, and in Figure 5/2-7 (b) and (c), the green highlighted regions refer to
fixation.

Figure 5/2-8 Loading and Fixation Conditions of Three Load Cases

In the analysis of the local submodel with finer meshing, the loading and fixation conditions for all the
load cases are displacement-control, i.e., relative displacement, which are interpreted from results
from the coarser meshing model. The concept is similar to the submodeling technique described in
PART 3.
5/2.3 FE models and simulation setups
5/2.3.1 Element
The first-order shell elements (i.e., 3- and 4-nodes for tris and quads, respectively) are employed in
both coarse and fine meshing models. The average element size for the coarse meshing model is
about 200 mm, and that for the fine meshing model is about 12 mm that is consistent with the steel
wall thickness.

5-16
5/2.3.2 Mesh design and quality control
The mesh was designed and tuned up by following DNVGL-CG-0127 [9], the mesh of global and
local FE models is revised based on the criteria shown in Table 5/2-1 and Table 5/2-2, respectively.
Note the criteria are developed based on the description and discussion in PART 3. The mesh
quality of the revised model meets all the criteria listed in Table 5/2-1 and Table 5/2-2.
Table 5/2-1 Fatigue Problem with 200mm Element Size

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 200 154 106.8 60 48.4
2 max length 200 260 316 400 514.4
3 aspect ratio 1 2 4 5 10
4 warpage 0 3.75 11.25 15 30
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 1.5 2.25 3 4.5
13 taper 0 0.2 0.5 0.6 0.9

Table 5/2-2 Fatigue Problem with 25mm Element Size

# Criterion Ideal Good Warn Fail Worst

0 penalty value 0 0 1 2 10
1 min length 25 19.25 13.35 7.5 6.05
2 max length 25 32.5 39.5 50 64.3
3 aspect ratio 1 2 4 5 10
4 warpage 0 3.75 11.25 15 30
5 max angle quad 90 110 140 150 165
6 min angle quad 90 70 40 30 15
7 max angle tria 60 80 120 130 155
8 min angle tria 60 50 30 20 10
9 skew 0 10 50 60 75
10 jacobian 1 0.9 0.6 0.5 0.35
11 chordal dev 0 0.3 0.8 1 2
12 % of trias 0 1.5 2.25 3 4.5
13 taper 0 0.2 0.5 0.6 0.9
The FE models with coarse and fine meshing consist of about 55,000 and 85,000 nodes,
respectively. The number of triangular elements is less than 3% of the total element numbers. Note
that the mesh quality criteria for the fatigue analysis are different from those for the impact analysis

5-17
presented in Example-1 (Table 5/1-1).
5/2.4 Analysis control and solution options
A special post-processing technique is adopted to estimate the stress levels in the point of interest to
reduce the impact of the stress singularity at the notch tip. As above-mentioned, the shell element is
used to construct the structure, and therefore the weld toe (see blue shading in Figure 5/2-9) is not
practical to consider in the model. However, because of the presence of the geometric
discontinuities – unreal notch, the estimated stress at the notch (notch) can be significantly higher
than the actual value (g). Figure 5/2-9 schematically shows the technique to estimate the value of
g from stress values at t/2 and 3t/2, where t refers to the plate thickness.

Figure 5/2-9 Hot Spot Stress Extrapolation in FE Models

when Weld Toe (Blue-Color Shading) is not Modeled, Since the Use of Shell Elements

In the fatigue analysis, the stress range is calculated from results at hot spots. The fully reversed
loading is assumed for the bending load case, and the stress range (𝜎𝑟𝑎𝑛𝑔𝑒 ) is two times of the first
principal stress (𝜎1 ) of hot spots:
𝜎𝑟𝑎𝑛𝑔𝑒 = 2|𝜎1 |
For the pressure loading case, the stress ranges of upper and lower connections are the difference
between the first principal stress (𝜎1 ) of hot spots from fully loaded and ballast states:
𝐹𝑢𝑙𝑙𝑦−𝑙𝑜𝑎𝑑𝑒𝑑
𝜎𝑟𝑎𝑛𝑔𝑒 = 𝜎1𝐵𝑎𝑙𝑙𝑎𝑠𝑡 − 𝜎1

𝐹𝑢𝑙𝑙𝑦−𝑙𝑜𝑎𝑑𝑒𝑑
where 𝜎1𝐵𝑎𝑙𝑙𝑎𝑠𝑡 > 0 > 𝜎1 .
In a unit structure (i.e., a local submodel), there are two hot spots (see Figure 5/2-10), and there are
four possible predominant load paths (A, B, C, and D). Note that the load path depends on the
loading and boundary conditions. The values of the principal stress are extracted along the four load
paths to calculate the stress (g) at hot spots. The hot spots at both upper and lower connections are
examined. Note that all the results are based on the nodal values.

5-18
 Figure 5/2-10 Possible Four Load Paths to Two Hot Spots at the Lower Connection

5/2.5 Results
5/2.5.1 Displacement results
The deformed shape and displacement results of the global model under longitudinal bending are
shown in Figure 5/2-11. Note that the deformed shape in Figure 5/2-11 is exaggerated, aiming to
qualitatively confirm the sense of loading and boundary conditions. The displacement results from
the global model are used to model the structural response in the local sub-model. The same
procedures are followed for the other two load cases.

Figure 5/2-11 Displacement Results of Global Model Due to Longitudinal Bending (Load
Case – 1)

5/2.5.2 Stress Results

In the analysis of the local sub-model, the distribution of the maximum principal stress due to the
longitudinal bending is shown in Figure 5/2-12. It is clearly seen that the stress levels at two hot
spots are significantly higher than the rest of the structure. The distributions of maximum principal
stress due to ballast load and cargo load are shown in Figure 5/2-13 and Figure 5/2-14, respectively.
In Figure 5/2-12, Figure 5/2-13, and Figure 5/2-14, the stress levels at point A are relatively higher
than that at point B. The maximum principal stress in all three load cases is aligned with the stiffener

5-19
axis. That indicates the fatigue failure mechanism is in Mode-1 (see Figure 3/2-17 of PART 3) and
caused by cyclic tensile loading.

Figure 5/2-12 Maximum Principal Stress Due to Longitudinal Bending (Load Case – 1)

Figure 5/2-13 Maximum Principal Stress of Local Sub Model Due to Ballast Load
Conditions (Load Case – 2)

Figure 5/2-14 Maximum Principal Stress of Local Sub Model Due to Cargo Load
Conditions (Load Case – 3)

5-20
The stress values of hot spots along four load paths are extracted based on the method above
mentioned (see 5/2.4 and Figure 5/2-9). The results for three load cases are summarized in Figure
5/2-15.
At this point, all the results are preliminary for the subsequent fatigue analysis. The stress ranges
(see green shading in Figure 5/2-15) are calculated from those hot spot stresses. Note that for the
bending load case, it is a cyclic moment loading and the stress transfer functions are used instead of
the stress range for the fatigue life calculation in FlawCheck®. The stress transfer function is defined
as follows:
𝑇𝑟𝑎𝑛𝑠𝑓𝑒𝑟 𝐹𝑢𝑛𝑐𝑡𝑖𝑜𝑛 = 𝑆𝑡𝑟𝑒𝑠𝑠 𝑅𝑎𝑛𝑔𝑒⁄𝐴𝑝𝑝𝑙𝑖𝑒𝑑 𝑀𝑜𝑚𝑒𝑛𝑡
On the other hand, the stress transfer function for the cyclic membrane tension (i.e., load cases – 2
and – 3) equals the stress range, where the applied moment = 1.
For the load case – 1, the worst case is through load path A at lower connection, i.e., the maximum
stress range = 2.95 × 10-3 MPa, and the stress transfer function = 1.4 × 10-10 MPa/(N·mm), see
Figure 5/2-15. For the load cases – 2 and – 3, the worst case is through load path C at the upper
connection, i.e., the maximum stress range = 635.6 MPa.
From the real-time measurement, the sample loading spectrum for the load case - 1 in a year is
shown in Table 5/2-3. The screenshot of the loading spectrum setup in FlawCheck is shown in
Figure 5/2-17. The loading spectrum for load cases – 2 and 3 (combined for fatigue analysis) is unit,
which means the stress range occurs once per transit.

5-21
Figure 5/2-15 The stress values of hot spots for four load paths from FEA and corresponding stress ranges at upper and
lower connections

5-22
 Figure 5/2-16 Defining SN Transfer Functions for Load Case – 1 (bending)
in FlawCheck®

Table 5/2-3 Loading Spectrum Under Longitudinal Bending Moments in a Year

Bin No. Bin Size – M [MN·m] Number of occurrences

1 150 2.0 × 106
2 200 1.0 × 106
3 250 5,000
4 300 500
5 350 5

5-23
 Figure 5/2-17 Setup of Loading Spectrum in FlawCheck®

5/2.5.3 Fatigue Results

The estimated fatigue life for the moment loading is approximately 11.8 years, which is equivalent to
3.6 × 10-7 cycles. It should be noted that this result can vary with a different applied load spectrum.
The SN accumulated damage is shown in Figure 5/2-18 in terms of the number of years.

Figure 5/2-18 Fatigue Life Under the Moment Loading, where the Accumulated Damage
Plots against Time

For the ballast and cargo pressure loading, the estimated fatigue life is about 1,975 cycles, which is
equivalent to 5.41 years of service. Note that the result may vary with the refinement of local
geometry, e.g., including weld toes.
The fatigue results from FEA and fatigue life analysis are similar to those from real-life observed

5-24
observation. The ballast pressure and full cargo pressure are the major cause of short service life. It
is recommended to revise the design close to the hot spots to reduce the stress level or stress
concentration.
5/2.6 Discussions and Comments
The fatigue analysis method presented in the above section is based on the method of SN damage
analysis. As discussed in PART 3, the fracture-mechanics-based fatigue analysis is another method
that is usually adopted in an integrity analysis (including a fatigue life analysis). In this section, the
fatigue life is evaluated by using the fracture-mechanics-based method. The procedures also follow
the rules of BS 7910 [12].
It is assumed that there is a pre-existing flaw with 30 microns (i.e., 0.03 mm) and will propagate
along the most critical load path C at the upper connection, as shown in Figure . The duration of this
flaw to grow through the plate (either through-thickness or through-width direction) is estimated. The
width and length of the interested plate (i.e., purple in Figure ) are 200 mm and 12 mm, respectively.

Figure 5/2-19 Direction of Crack Growth along the Load Path C at the Upper Connection

The material properties of ballast steel are as follows:

 Yield strength [MPa] 360
 Ultimate tensile strength [MPa] 400
 Paris’ law constant – C [-] 2.3 × 10-12
 Paris’ law constant – m [-] 3
Note that the crack growth constants of Paris’ law are recommended by BS 7910.
Using the same stress range and loading spectrum from the ballast and cargo pressure loading, the
estimated fatigue life is about 2,436 cycles. That means it takes about 6.67 years to grow through
the plate thickness. Figure 5/2-20 shows that the crack growth depth is about 90.09 mm along the
plate width direction. This result is similar to that from the SN based fatigue life estimation. Note that
this result will vary with a refinement of the initial flaw in terms of size and orientation.
Note that the effects of residual stress and mean stress need to be addressed in fatigue crack
growth analysis. A crack may initiate in an area of high residual tensile stress but grow into an area
of lesser stress, where the rate of crack growth will be slowed. On the other hand, as a crack grows
across a ship, the stress levels increase as less structure is effective, and the rate of crack growth

5-25
increases. If the finite element code does not adjust for the changes in the stress fields with crack
growth, the analyst must do so by reinitiating the analysis.

Figure 5/2-20 Fatigue Life Estimation Using the Fracture-Mechanics-Based Fatigue

Analysis Method, Where the Flaw Length Plots Against Time

5/3 WHOLE SHIP ANALYSIS

5/3.1 Sample Application Description
An example is given to show how to perform a whole ship hull girder strength analysis. The sample
application notional frigate has a length between perpendiculars L=135m, beam B=17m, depth
D=9.4m, and displacement of 4,000 tons. This analysis uses MAESTRO 11.7.0, a finite element
analysis-based toolset tailored to the design of ships and other floating platform structures by naval
architects.
5/3.2 Finite Element Model Development
The notional frigate model was created using the NAPA-Steel initial design structural modeler, as
shown in Figure 5/3-1 (a). The whole ship finite element model has over 61,000 nodes and 125,000
elements, as shown in Figure 5/3-1 (b). This model is an example of a finer mesh model, where the
stiffeners are modeled explicitly as offset beams at their exact locations, the web plate of the frames
and girders are modeled as shell elements, and the flange plate of the frames and girders is
modeled as flat bar elements. One shell element is modeled between stiffeners, and three elements
are modeled between frames. Figure 5/3-2 illustrates the finite element mesh density of the frigate
model.

5-26
 (a) (b)
Figure 5/3-1 A Notional Frigate

Figure 5/3-2 A Notional Frigate Meshing Approach

The full hull and superstructure are modeled. The model would typically be developed based on the
geometry of a 3D hull model, general arrangements drawings, and scantling drawings at typical
frames. This full-ship FE model and its mesh density are typical for use in designing hull, deck, and
bulkhead scantlings for the final design. A simpler coarse mesh model could have been used in
early-stage design to establish initial structural topology and scantlings.
5/3.3 FE Model Mass Properties and Hydrostatic Loading
The tank boundary elements and the hull wettable elements can be defined interactively using
element grouping features. The frigate tank boundary definitions and wettable surface elements are
shown in Figure 5/3-3 and Figure 5/3-4, respectively.

5-27
 Figure 5/3-3 Liquid Tank Definition of the Notional Frigate

Figure 5/3-4 Frigate “Wettable” Elements

In addition, the weight distribution, tank loads, and floating conditions defined in the initial design
modeler software hydrostatic module are also translated into the sample application FEA. Figure
5/3-5 and Figure 5/3-6 show how tank loads and lightship weight was entered in the sample
application FEA software.

5-28
Figure 5/3-5 Frigate Full Load Tank Definition

5-29
 Figure 5/3-6 Frigate Lightship Weight Distribution

5/3.4 Boundary Conditions

Three traditional constraints are placed in the model as described in 3/2.7.5.1 and shown in Figure
5/3-7. For the still water condition, the total reaction force in the vertical direction is 1.6 tons, which is
less than 1% of the ship's gross weight. Therefore, the constraints are considered satisfactory.

5-30
 Figure 5/3-7 Typical Full-ship Boundary Constraints

5/3.5 Hull Girder Design Waves

Hull girder sagging and hogging design loads are generated using four different methods,
1. Quasi-static trochoidal waves
2. SPECTRA universal RAO
3. Frequency domain extreme loads
4. Time-domain extreme loads
Using the sample application FEA software internal hydrostatic balance feature, the prescribed
trochoidal wave profiles are automatically adjusted to have proper sinkage, trim, and heel of the
model into a quasi-static equilibrium between the defined weight and the buoyancy provided by the
wettable elements. Figure 5/3-8 shows the wave pressure distribution of the frigate subjected to
quasi-static sagging and hogging trochoidal waves.

Figure 5/3-8 Quasi-static Pressure Under Sagging and Hogging Waves

The second method is to use SPECTRA to generate a hull girder lifetime extreme bending moment
distribution envelope. Alternatively, one can create a design wave bending moment using equations
provided by class society rule books. To simulate the bending moment distribution on a 3D finite
element model, the sample application FEA software automatically places a large set of vertical
nodal forces on the model such that the resulting hull girder longitudinal bending moment distribution
matches the prescribed bending moment distribution. The third method uses the sample application
FEA software hydrodynamic kernel to generate extreme design waves. This approach is based on
ABS’s DLA procedure. The frigate model has 19,266 wetted finite elements. The wetted elements

5-31
can be further reduced to 8,202 evaluation patches, as shown in Figure 5/3-9. Those patches are
then used for hydrodynamic loads calculations.

Figure 5/3-9 Frigate Model Evaluation Panels

For the frequency domain analysis, it is assumed the ship has a forward speed of 0, 5, 15, and 25
knots, 7 different headings, and 30 wave frequencies. A unit wave response database is generated.
The database includes wave-induced accelerations, panel pressures, nodal forces due to hydrostatic
restoring correction, and the hull girder loads. A design wave is generated based on a desired
Dominant Load Parameter (DLP), a sea state scatter diagram, an operating profile, and the return
period. The dynamic design wave, which has a perfect equilibrium (because the equations of
motions are formulated on the FEA mesh rather than a hydrodynamic mesh), combined with the
static loads, becomes a DLP static analysis load case. Figure 5/3-10shows an example of
generating the extreme vertical bending moment design waves using the general Atlantic scatter
diagram, combatant operating profile, Pierson Moskowitz wave spectrum, and 20 years return
period. Figure 5/3-11 shows the hydrodynamic pressure and vertical bending moment distribution of
the corresponding linear design waves.

5-32
Figure 5/3-10 Example of Generating an Extreme Vertical Bending Moment Design Wave

Hogging Pressure distribution Sagging Pressure distribution

Hogging bending moment distribution Sagging bending moment distribution

Figure 5/3-11 Hydrodynamic Pressure and Vertical Bending Moment of the Linear Design
Waves

The fourth method is to use the sample application FEA software weakly nonlinear time-domain
simulation to generate a hull girder extreme bending moment. From the frequency domain load
analysis results, the expected most probable extreme hull girder vertical bending moment occurs
when the ship sails head sea at 15 knots in sea state 8. A 20-minutes weakly nonlinear time-domain
simulation at this condition was conducted. The maximum hull girder sagging and hogging moments
were extracted from the time history. A load scale factor was determined using the ratio of the peak
time-domain hull girder loads and the frequency domain hull girder loads. The panel pressures, point
forces, and accelerations at the peak response time-step of each time-domain simulation were then
scaled such that the magnitude of the DLP was equal to that predicted from the frequency domain
analysis. If the peak of the time domain run exceeded the frequency domain extreme value, the peak
of the time domain run was to be used for that particular DLP. Figure 5/3-12 shows an example of
setting up a weakly-nonlinear time-domain simulation in a sea state. Figure 5/3-13 shows an

5-33
example of vertical bending moment time history at mid-ship. Figure 5/3-14 shows the ship speed,
heading, sea condition, and DLP VBM values where the time-domain simulation peaks occurred,
and Figure 5.3-15 shows a snapshot of the maximum bending moment in the nonlinear simulation
time history and the corresponding hydrodynamic pressure.

Figure 5/3-12 Example of Setting Up a Weakly-Nonlinear Time Domain Simulation

Figure 5/3-13 Example of Vertical Bending Moment Time History at Mid-ship

5-34
 Figure 5/3-14 Maximum Vertical Bending Moments in this Time Domain Simulation

Figure 5/3-15 Pressure Distribution when the Maximum Vertical Bending Moments Occur

The extreme sagging and hogging moments computed by the four methods are listed in Table 5/3-1.
For SPECTRA, a 20-year design life, the combatant operating profile, Pierson-Moskowitz wave
spectrum and General Atlantic sea scatter diagram were used to generate the dynamic hull girder
loads
Table 5/3-1 Design Bending Moments

Sagging(ton*m) Hogging(ton*m)
Quasi-Static Trochoidal Design Wave 22,614 28,856
SPECTRA 37,455 50,850
Frequency Domain 22,618 41,455
Time Domain 43,138 40,957

5-35
5/3.6 Limit State or Failure Criteria
To assess the strength of whole ship primary structures, the ABS High-Speed Naval Craft (HSNC)
2017 [86] limit states have been integrated into the sample application FEA software limit state
evaluation and optimization frameworks, and are used in this example. The acceptance criteria
include not only stress limits but also buckling rules and other limit states. Figure 5/3-16 shows the
limit state evaluation panels automatically defined in the sample application FEA software for the
entire ship’s structure. Each colored group of elements represents a single evaluation patch, which
is evaluated against each of the selected limit states for all load cases defined.

Figure 5/3-16 Frigate Evaluation Panels

Figure 5/3-17 presents the buckling, and ultimate strength limit states evaluated in the sample
application FEA software during the optimization process or limit state evaluations outside of
optimization. The HSNC 2017 buckling and ultimate strength limit states reference the criteria in 5-1-
5/5.3.1 and 5-1-5/5.3.2 of the guide ABS Rules for Building and Classing Steel Vessels [65] .

5-36
 Figure 5/3-17 ABS HSNC Buckling and Ultimate Strength Limit States

5/3.7 Finite Element Analysis and Results

A finite element and limit state analysis of the frigate’s structural design was conducted using the
eight load cases created using the methods discussed in Section 5/3.5. The results for the extreme
vertical sagging bending moment (time-domain weakly nonlinear simulation under full load, sea state
8, 15 knots, head sea, Sagging VBM = 43,138 ton*m) are shown here. The colored elements shown
in Figure 5/3-18 indicate those elements that do not meet the maximum allowable stress limits as
specified in the HSNC Rules. It should be noted that, for the most part, except hatch corners, the
stresses are within the ABS HSNC limits. That is not true of the structure’s evaluation against the
HSNC buckling limits. Figure 5/3-19 shows those elements that fail to meet, in this load case alone,
the HSNC’s various buckling and ultimate limit state criteria as outlined in Section 5/3.6 (Figure
5/3-17).

5-37
Figure 5/3-18 Allowable Stress Under Time Domain Sagging Condition

Figure 5/3-19 All Limit States Under Time Domain Sagging Condition

5-38
5/3.8 Sample Showing the Significance of Equilibrium for Whole Ship Analysis
For a floating structure, it is important to obtain equilibrium before performing a finite element
analysis. Imbalance in a model causes errors in the results. To illustrate this, a design exercise is
given in this section.
A frigate model is given in section 5/3.1. The hydrostatic model was balanced using the hydrostatic
software in the design stage prior to a finite element analysis. The floating condition of a draft, trim,
and heel was given from the hydrostatic software. The balanced floating condition had a draft of 3.96
m, with no heel and -0.382 degree trim. In regular finite element analysis, this floating condition is
used to load the finite element model. Figure 5/3-20 and Figure 5/3-21 show the weight and
buoyancy distribution of NAPA (the hydrostatic software) and MAESTRO (the sample FEA
software), respectively. While the weight distribution of these software has a good agreement, the
buoyancy distribution of the sample FEA software and the hydrostatic software does not have good
agreement. Such a discrepancy may be caused by several possible reasons: A) Rudders and
propellers are usually not modeled in the finite element model for hull girder strength analysis. B)
The integration schemes are different. The buoyancy calculation in the hydrostatic software is
volume-based, using continuous curves and/or surfaces. For finite element analysis packages, the
buoyancy calculation is pressure based, as a result of integrating the hydrostatic pressure over the
faceted shell elements in the finite element model.
Weight Comparison
50
Weight Density(Kg/m)x1000

NAPA
40
MAESTRO
30

20

10

0
x(m)

Figure 5/3-20 Weight Distribution Comparison

Hydrostatic Buoyancy Comparison

6
Buoyancy(kg/m) x 10000

NAPA
4 MAESTRO

0
-10 40 x(m) 90 140

Figure 5/3-21 Buoyancy Distribution Comparison

To solve the above finite element model, three nodal constraints were placed near the longitudinal
neutral axis of the model to prevent the rigid body motion, with two located at the stern and one at
the bow. If the model is properly balanced, the restraining forces will be negligible. In order to check
the balance, the sample application FEA software computes and plots hull girder responses before
the restraints are applied. Figure 5/3-22 shows the vertical bending moment, and vertical shear force
before the restraints are applied. Since the curves are not closed, they reveal that the model is not
well balanced. Figure 5/3-23 shows the resulting bending moment distribution, which includes the
reaction forces due to the constraints. Figure 5/3-24 shows the deflection and stress distribution. The

5-39
lack of buoyancy at the stern and the resulting non-negligible restraint forces cause an incorrect
change of sign in the bending moment and in the curvature of the deflected hull.

(a) Bending moment

(b)Shear force

Figure 5/3-22 Bending Moment and Shear Force Distribution

before Applying Restraining Forces

Figure 5/3-23 Bending Moment Distribution

after Applying Restraining Forces

Figure 5/3-24 Deflection and Stress Distribution

If the analyst is using a general-purpose FEA program, then they must work to resolve as much of
the force and moment imbalance as possible prior to evaluating any FEA results. Either the vessel
draft and trim may need to be modified in the FEA model, or the applied loads may need to be
reviewed for accuracy. The results shown in Figure 5/3-24 would be questionable.
Finite element tools specifically developed for floating structures usually provide a “hydrostatic
balancing kernel” by which an imbalanced finite element model can be automatically rebalanced by
adjusting draft, heel, and trim. After this hydrostatic balancing, the finite element model has a draft of
4.07m and a trim of 0.437 degrees, which are small changes comparing to the original hydrostatic
software’s floating condition (0.397m draft, and -0.382 degree trim). Once the model is balanced, the
buoyancy discrepancy between the hydrostatic software and the FEA software is corrected, as
shown in Figure 5/3-25. Likewise, the distributions of vertical bending moment and shear force are
correctly closed at both ends, as shown in Figure 5/3-26, and the corresponding deflection and
stress distributions are as expected, as shown in Figure 5/3-27.

5-40
 Hydrostatic Buoyancy Comparison
NAPA
MAESTRO,Draft=3.96,Trim=-0.382
6

Buoyancy(kg/m)x 10000
MAESTRO,Draft=4.07,Trim=0.437
5
4
3
2
1
0
-10 40 x(m) 90 140

Figure 5/3-25 Buoyancy Distribution Comparison with MAESTRO Auto-Balancing

(a) Bending moment

(b) Shear force

Figure 5/3-26 Bending Moment and Shear Force Distribution after Hydrostatic Balance

Figure 5/3-27 Stress Distribution after Hydrostatic Balance

Hydrostatic imbalance can also result when the loads are derived from a linear seakeeping analysis,
where the mean water surface plane is prescribed. There are two main reasons for this as follows:
(1) hydrostatic balancing can only achieve equilibrium in heave, heel and trim, but not in surge, sway
and yaw, and(2) a hydrostatic rebalance would cause a change of the mean water surface plane,
which would require re-running the linear seakeeping analysis.
The “inertia relief” method is commonly used to resolve imbalance forces and moments. Figure
5/3-28 shows the distributions of vertical bending moment and shear force after using “inertia relief,”
and Figure 5/3-29 shows the corresponding deflection and stress distribution.

5-41
 (a) Bending moment

(b) Shear force

Figure 5/3-28 Bending Moment and Shear Force Distribution after Inertia Relief

Figure 5/3-29 Stress Distribution after Inertia Relief

The hull girder responses may be different depending on whether the finite element model is loaded
using inertia relief, hydrostatic balancing within the finite element model, or hydrostatic balance
outside the finite element model. Figure 5/3-30 shows the longitudinal bending moment distribution
of (a) the model is balanced in the original hydrostatic model (b) the model is balanced using inertia
relief, and (c) the model is hydrostatically balanced using the finite element geometry and weight
distribution.

Bending Moment Distribution

Bending Moment(Nm)x1E6

12 NAPA
10 Inertia Relief
8
Hydrostatic Balance
6
4
2
0
0 50 x(m) 100 150

Figure 5/3-30 Comparison of Bending Moment Distribution

5-42
5/4 FREQUENCY RESPONSE VIBRATION ANALYSIS
5/4.1 Sample Application Description
The following sample application is an example of how to perform a frequency response vibration
analysis (i.e., a forced vibration analysis). The sample application is a modified version of a report
issued to a commercial customer. Some vessel identifying information has been removed.
5/4.2 Introduction
This report validates the main engine and gear box foundation design vibration characteristic for a
vessel with a conventional low-speed diesel engine propulsion system. This report is an example of
how to perform a forced vibration analysis. Note that this report is a modified version of a report
issued to a commercial customer; some vessel identifying information has been removed.
5/4.3 Method
5/4.3.1 General
This report evaluates the subject foundation based on [14] ABS Guidance Notes on Ship on Ship
Vibration criteria for local structure. This analysis uses ANSYS 19.0, an industry-leading finite
element analysis software.
The analysis model was created based on the project foundation drawing and the available project
ship structure drawings. The analysis model extends from main engine room bulkhead to main
engine room bulkhead and includes the tank top, floors, girders, and shell plate. It also includes the
main engine room stanchions. The edges of the model are simply supported. Images of this model
are shown in Figure 5/4-1 to Figure 5/4-4.
The model is created using shell elements except for the main engines and gearboxes. The main
engine and gearbox are modeled using brick elements. The main engine and gearbox models
approximate the correct equipment masses, centers of gravity, and connections with the foundation.
The model boundary is the engine room bulkheads and the tops of the stanchions. These
boundaries have a zero-displacement boundary condition.
The foundation’s lowest natural frequencies are first determined using a modal analysis. A frequency
response (forced vibration) analysis is then performed to determine the subject foundation
responses to the excitation forces.
The modal superposition method is used for the frequency response analysis. The frequency
response analysis calculates the main engine induced maximum vibrations and propeller-induced
maximum vibrations separately. The separate vibrations values are added together where
applicable. This is conservative where added together because the maximum motions likely do not
occur at the same location in the same direction. The total vibration results are then compared
against the criteria in [14].
The analysis is done for one case with just the port side power train operating, and for three cases
with both power trains operating. The three cases with both power trains operating have the
starboard main engine 0 degrees, 90 degrees, and 180 degrees out of phase of the port engine.
This was done to test the impact of out of phase main engine excitation. The propeller excitation is
kept in sync for all these cases.
The Main Engine Technical Information and Propeller Blade Design Report describe the main
engine and propeller excitation forces, respectively. The excitation frequencies range from 0 to 25.9
Hz. There are three cyclical loads of concern, and they are as follows:
 The main engine is rated for up to 15 Hz (900 rpm).
 The gearbox ratio is 4.64:1, and the propeller has four blades. The blade rate, therefore, is
12.9 Hz (776 rpm).

5-43
  The prop has a v-strut in front of it, so a frequency of twice the blade rate or 25.9 Hz may
also be an important forcing frequency. Note also that this cyclical load may be occurring
simultaneously with the other two.
Reference [88] recommends that a range of frequencies +/- 5 to 10% be evaluated surrounding each
excitation frequency and that the worst-case frequency be used. Therefore for simplicity, the
frequencies were evaluated at 5 Hz increments surrounding the target excitation frequencies. In
practice, this works out to taking the highest response from three frequencies for each of the cyclical
loads. As an example, in the case of the main engine cyclical load, the three frequencies that are
relevant are the 10 Hz, 15 Hz, and 20 Hz results.
Further, the vibration is assumed to be linear, and so, therefore, the maximum individual response
values for each cyclical load can be summed together to get a conservative maximum total
response. The maximum total response in velocity, therefore, is taken as the summation of three
maximum individual responses in velocity: the main engine response velocity, the propeller blade
rate analysis response velocity, and the propeller twice the blade rate response velocity. The result
is compared against the ABS criteria maximum allowable velocity.
The maximum total velocity is determined without consideration for where on the model, the
maximum individual velocities are occurring. This is only because the results were much less than
the allowable criteria. Had the results been closer to the limiting criteria, then the locations of these
maximums would have been taken into consideration.
5/4.3.2 Model Size Variation
This report also includes additional results using a smaller model to show the impact of the model
size. The smaller model only includes the equipment and foundation structure above the tank top on
the port side. Images of this model are shown in Figure 5/4-5 to Figure 5/4-8.
5/4.4 Given and Assumed Parameters
5/4.4.1 Equipment
Equipment List:
Main Engine: Make/Model - Mass
Gear Box: Make/Model – Mass
5/4.4.2 Vibration Criteria
ABS Guidance Notes on Ship on Ship Vibration Section 7 Acceptance Criteria, Part 5 Vibration
Limits for Local Structures [14] says the following:
 For each peak response component (in either vertical, transverse, or longitudinal direction),
from 1 Hz to 5 Hz, the displacement is recommended below 1.0 mm, and the damage is
probable above 2.0 mm
 For each peak response component (in either vertical, transverse, or longitudinal direction),
from 5 Hz and above, the velocity is recommended below 30 mm/sec, and the damage is
probable above 60 mm/sec.
The peak excitation forces occur above 5 Hz. Therefore it is required that the maximum structural
velocity is less than 30 mm/sec.
ABS Guidance Notes on Ship on Ship Vibration Section 7 Acceptance Criteria, Part 7 Vibration
Limits for Machinery [14] provides the Table 5/4-1 machinery vibration limits.

5-44
 Table 5/4-1 ABS Machinery Vibration Limits [14]

Therefore it is required that the maximum diesel engine velocity is less than 13 mm/sec, and the
maximum thrust bearing and bull gear hub (gearbox) is less than 5 mm/sec.
The peak displacements are also reported for reference.
5/4.4.3 Excitation Forces
5/4.4.3.1 Main Engine Excitation Forces
The Main Engine Technical Information reports describe the main engine excitation forces as
follows:

Therefore a rotating force of 500 lbs is applied about the longitudinal axis of the engine. This force is
applied to the forward and aft face of the main engine.
This force will peak at 15 Hz, as stated.
5/4.4.3.2 Propeller Excitation Forces
Propeller Blade Design Report determines the propeller excitation force amplitude is 1.8 kN or 405
lbs. Therefore, a dynamic force of 405 lbs is applied in the direction of the shaft centerline. This force
is applied to the aft face of the gearbox in this analysis.
The blade rate, therefore, is 12.9 Hz.
The prop has a v-strut in front of it, so a frequency of twice the blade rate may also be an important
forcing frequency; 25.9 Hz.
5/4.5 Results
5/4.5.1 Bulkhead to Bulkhead Model
The modal analysis determined the 12 lowest modal frequencies as listed in Table 5/4-2 below for
the bulkhead to bulkhead model. The lowest natural frequencies are at the high end of the highest
excitation frequencies. Therefore it is expected, and the frequency response results show that the
most significant vibration motions occur as a result of the high-end excitation frequencies.
Table 5/4-2 Bulkhead to Bulkhead Modal Frequencies

Mode No. Nat. Freq [Hz] Description

1 24.9 Aft centerline stanchion surging near boundary condition.
2 25.2 Middle centerline stanchion surging near boundary condition.
3 27.9 Small outboard port side floor unconnected to bottom shell
(modeling error).
4 28.0 Small outboard starboard side floor unconnected to bottom shell
(modeling error).

5-45
 Mode No. Nat. Freq [Hz] Description
5 28.7 Starboard outboard aft stanchion surging between its end
connections.
6 28.7 Port outboard aft stanchion surging between its end connections.
7 32.5 Shell and tank top heaving below main engines.
8 33.2 Main engines swaying side to side in tandem.
9 34.2 Middle centerline stanchion surging between boundary conditions.
10 34.6 Aft centerline stanchion surging between boundary conditions.
11 35.0 Port outboard aft stanchion swaying between its end connections.
12 35.1 Starboard outboard aft stanchion swaying between its end
connections.
Table 5/4-3 and Table 5/4-4 show the bulkhead to bulkhead model maximum vibration-induced
displacements & velocities based on the frequency response analysis. The data is provided for a
range of forcing frequencies near the predominant 15 Hz and 25 Hz forcing frequencies.
Table 5/4-3 shows the bulkhead to bulkhead model results for the cases with both power trains
operating. Table 5/4-4 shows the results with only the port power train operating. Table 5/4-3 cases
with both power trains operating results in more severe motions. The most severe motions within the
cases with both power trains operating depend on the forcing frequency.
The main engine maximum individual velocity response is 0.408 mm/s. The blade rate maximum
individual response is 0.099 mm/s. The twice blade rate maximum individual response is 0.562
mm/s. The maximum total velocity response is, therefore, 1.069 mm/s. This is significantly less than
the least allowable maximum velocity of 5 mm/sec in [14]. The foundation, therefore, will not have
vibration issues.
Table 5/4-3 Bulkhead to Bulkhead Model Vibration Results: Port and Starboard Power
Trains in Sync

5-46
 Table 5/4-4 Bulkhead to Bulkhead Model Vibration Results: Port Power Train Only

5/4.5.2 Tank Top Model

The tank top port side model main engine maximum individual velocity response is 0.206 mm/s. This
is only 53% of the same result for the bulkhead to bulkhead model results with just the port side
forces. The blade rate maximum individual response is 0.093 mm/s. This is only 88% of the same
result for the bulkhead to bulkhead model results with just the port side forces. The twice blade rate
maximum individual response is 0.144 mm/s. This is only 49% of the same result for the bulkhead to
bulkhead model results with just the port side forces. In summary, the maximum individual
responses are all significantly less for the tank top model. The extra effort to develop the larger
bulkhead to bulkhead model was, therefore, worthwhile in order to avoid reporting a non-
conservative result.
Table 5/4-5 Tank Top Model Vibration Results

5/4.6 Conclusions
The calculated vibrations are acceptable. They do not exceed the allowable limits defined in [14].

5-47
5/4.7 Appendix A: GEOMETRY PLOTS

5-48
Figure 5/4-1 Bulkhead to Bulkhead Model Image 1

Figure 5/4-2 Bulkhead to Bulkhead Model Image 2

5-49
Figure 5/4-3 Bulkhead to Bulkhead Model Image 3

Figure 5/4-4 Bulkhead to Bulkhead Model Image 4

5-50
Figure 5/4-5 Tank Top Model Image 1

Figure 5/4-6 Tank Top Model Image 2

5-51
Figure 5/4-7 Tank Top Model Image 3

Figure 5/4-8 Tank Top Model Image 4

5-52
5/4.8 Appendix B: LOADING AND BOUNDARY CONDITION PLOTS

5-53
Figure 5/4-9 Main Engine Forcing Loading Condition Image 1

Figure 5/4-10 Main Engine Forcing Loading Condition Image 2

5-54
Figure 5/4-11 Gear Box Forcing Loading Condition Image

Figure 5/4-12 Boundary Conditions Image

5-55
5/4.9 Appendix C: DISPLACEMENT PLOTS

5-56
Figure 5/4-13 Main Engine Maximum Displacement at 15 Hz

Figure 5/4-14 Main Engine Maximum Displacement at 20 Hz

5-57
Figure 5/4-15 Gear Box Maximum Displacement at 15 Hz

Figure 5/4-16 Gear Box Maximum Displacement at 20 Hz

5-58
Figure 5/4-17 Gear Box Maximum Displacement at 25 Hz

5-59
5/4.10 Appendix D: VELOCITY PLOTS

5-60
Figure 5/4-18 Main Engine Maximum Velocity at 15 Hz

Figure 5/4-19 Main Engine Maximum Velocity at 20 Hz

5-61
Figure 5/4-20 Gear Box Maximum Velocity at 15 Hz

Figure 5/4-21 Gear Box Maximum Velocity at 20 Hz

5-62
Figure 5/4-22 Gear Box Maximum Velocity at 25 Hz

5-63
 CONCLUSIONS AND RECOMMENDATIONS
There has been much development in the theory and application of FEA to ship structures since the
1996 SSC-387 was published. The relative quantity of references in this updated version versus the
original version attests to the level of development. The work presented in this report provides an
updated guide to those that are faced with the problem of evaluating the FEA work performed by
other parties. As an aid to the evaluation process, a comprehensive, systematic, updated
assessment methodology is presented in this report. It is designed to be flexible in terms of the level
of skill expected of the evaluator, and in terms of the size and complexity of the FEA that the
methodology can be applied to.
The methodology is structured on three levels. The first level is essentially an overview checklist of
features of an FEA that needs to be evaluated. A more detailed checklist, based on the first level, is
presented in the second level of the methodology. The third level provides guidance in narrative and
illustrative form and is structured to match the first and second level checklists. Further guidance is
provided through a series of illustrative examples which show the influence of varying finite element
modeling practice on FEA results. These are intended to help the evaluator in assessing the levels
of accuracy that might be attained in the FEA that is being evaluated.
The benchmarking information from the SSC-387 guide has been retained here, but it is typically no
longer necessary for well-established FEA software that has been used repeatedly already on ship
structure type problems. However, new FEA software or FEA software that has not been used on
ship structure type problems still requires benchmarking. The benchmark problems of the type
presented in this report can be regarded as a further level of qualification for such software.
These benchmark problems are intended to test the ability of software to provide accurate solutions
for structural assemblies typical of ship structures. Unlike the typical verification problem used by
software houses, benchmark problems consider non-ideal configurations, multiple element types,
several load cases, etc.
FEA codes are large and complex and hence can never be guaranteed to be free of errors.
However, it is suggested that FEA software that has been thoroughly tested by the vendor at the
verification example level, will, by successfully yielding solutions for the benchmark problems,
provide another level of assurance that the software is fit for performing ship structure FEA.
Several recommendations are presented below for consideration:
1. The scope could be broadened to include composite materials and the application of the
many varied types of contacts available in modern FEA software.
2. The benchmarks presented in this report might be considered as a starting point for building
a library of benchmark problems that include advanced analysis problem benchmarks. These
problems could also include high-quality and well-documented experiments on ship structure
assemblies.
3. There may be a benefit in a new study comparing the tradeoffs of using solid elements
versus using shell elements. There has been a significant advancement in computing power,
and FEA software capabilities since the marine industry generally adopted the use of shell
elements for ship structure problems. The automotive and aerospace industry is tending to
use more solid element models because less analyst intervention is required. Such a new
study would determine using current technology which types of ship structure FEA problems
are more efficiently done using solid elements versus shell elements.

6-1
 Appendix A
Evaluation Forms for Assessment of Models and
Results

A-2
Figure A/0-1 Overall Evaluation Methodology Checklist

A-3
FINITE ELEMENT ANALYSIS ASSESSMENT I PRELIMINARY CHECKS
Project No. I Project Title :

Company Name: IDate:

Analyst: I Checker:

A/1.1 Documentation Requirements

In order to perform a comprehensive assessment of an FEA, certain essential information
must be provided in the documentation submitted.
Refer to
Finite Element Analysis Assessment Check Guideline Result Comments
Section
1.1.1 Has the following information been provided in 3/1.1
the FEA documentation?
a. Scope and objective of analysis.
b. Timeline for delivery of analysis
c. Reference codes, manuals, and/or standards required.
d. Physical problem documentation references
e. FEA software used.
f. Contractor qualifications.
g. Analysis type(s).
h. Description of physical problem.
i. Description of engineering model.
j. Description of the estimated range of stress and displacements
expected from the FEA analysis.
k. Description of the FEA model.
l. Plots of full FEA model and local details.
m. System of units.
n. Coordinate axis systems.
o. Material properties.
p. Stiffness and mass properties.
q. Loads and boundary conditions.
r. Element type(s).
s. Meshing idealizations/ assumptions/ representations/
simplifications.
t. Meshing criteria for 2D and 3D elements.
u. FE loads and boundary conditions.
v. Solution options and procedures.
w. Results.

A-4
x. Comparison of results with acceptance criteria.
y. Accuracy assessment.
z. Conclusions and recommendations for amendments.

Based on the above checks, answer Question 1.1, and enter the result in Figure A/0-1. Result
1.1 Is the level of documentation sufficient to perform an assessment of the FEA?
Comments

A-5
A/1.2 Job Specification Requirements
Perform these checks to ensure that the analysis addresses the objectives, scopes, requirements,
and intent of the job specification (e.g., the contract document, work specification, statement of work,
etc.).
Refer To
Finite Element Assessment Check Guideline Result Comments
Section
1.2.1 Is the job specification identified and referenced in 3/1.2
the analysis documentation?

1.2.2 Are the objectives and scopes of the analysis 3/1.2

clearly stated, and are they consistent with those of the
job specification?
1.2.3 Are the analysis requirements clearly stated, and 3/1.2
are they consistent with those of the job specification?

1.2.4 If certain requirements of the job specification have 3/1.2

not been addressed (such as certain load cases), has
adequate justification been given?

1.2.5 Are the design / acceptance criteria clearly stated, 3/1.2

and are they consistent with those of the job
specification?
1.2.6 Is there reasonable justification for using FEA for 3/1.2
this problem?

1.2.7 Has the advantage been taken of any previous 3/1.2

experimental, analytical, or numerical works that are
relevant to this problem? If so, are these previous works
cited in the subject analysis report?

Based on the above checks, answer Question 1.2, and enter the result in Figure A/0-1. Result
1.2 Does the analysis address the job specification requirements?

Comments

A-6
A/1.3 Finite Element Analysis Software Requirements
The FEA software should meet certain minimum standards to be considered acceptable for ship
structural analysis applications.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
1.3.1 Is the FEA software appropriate for the application 3/1.3
or on a list of approved programs for ship structural
analysis applications in the organization?

If the answer to Check 1.3.1 is "Y," you may skip Checks 1.3.2 and 1.3.3.

1.3.2 Are the capabilities and limitations of the FEA 3/1.3

software used to perform the required analysis stated in
the analysis documentation?
1.3.3 ls evidence of this capability documented and 3/1.3
available for review (e.g., verification manual, results of
ship structure FEA benchmark tests, previously
approved FEA of similar problems)?
1.3.4 Does the vendor of the FEA software have a 3/1.3
quality system to ensure that appropriate standards are
maintained in software development and maintenance.
1.3.5 Other concerns: 3/1.3
 Does the FEA solver have both single and
double precisions?
 Does the FEA solver include implicit and/or
explicit solvers?
 Does the FEA solver have a HPC (high
performance computing) version? Parallel
processing: SMP (Shared memory parallel)
version? DMP (distributed memory parallel)
version?
 Does the GPU (graphics processing units)
hardware precision match the solver’s precision?
Note that the double-precision GPU is
recommended for those CPU and GPU based
simulations
 Does RAM have an error checking and
correcting (ECC) function?

Based on the above checks, answer Question 1.3, and enter the result in Figure A/0-1. Result
1.3 Is the FEA software qualified to perform the required analysis?
Comments

NOTE: Part 4 of this report presents benchmark problems for the purpose of assessing the quality
and suitability of FEA software for performing ship structural analysis. On its own, the successful
performance of the candidate FEA software in exercising the benchmark problems is not sufficient

A-7
evidence of the quality and suitability of the software. The assessor should, in addition, be able to
answer the other questions in the table above affirmatively.

A-8
A/1.4 Personnel Qualification Requirements
The personnel should possess certain minimum qualifications for performing ship structure FEA. In
addition, the personnel should adhere to a Quality Assurance (QA) system to ensure that proper
management, administrative, and checking procedures have been applied in the analysis.
Refer To
Finite Element Assessment Check Guideline Result Comments
Section
1.4.1 Do the personnel have adequate academic training 3/1.4
and experience qualifications to perform finite element
analysis?

1.4.2 Do the personnel have adequate engineering 3/1.4

experience qualifications for performing ship structural
design or analysis?

1.4.3 Do the personnel have adequate professional 3/1.4

certification qualifications?

1.4.4 Does the personnel have a working system of 3/1.4

Quality Assurance (QA) procedures and checks that are
satisfactory for the requirement?

1.4.5 Do the personnel have adequate experience with 3/1.4

the FEA software used for the analysis?

Based on the above checks, answer Question 1.4, and enter the result in Figure A/0-1. Result
1.4 Is the personnel adequately qualified for performing ship structure FEA?

Comments

A-9
FINITE ELEMENT ANALYSIS ASSESSMENT ENGINEERING MODEL CHECKS

Project No. I Project Title :

Company Name: IDate:

Analyst: I Checker:

A/2 Engineering Model Checks

A/2.1 Analysis Type
Perform these checks to ensure that the assumptions used in developing the engineering model or
idealization of the physical problem are adequate
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.1.1 Does the engineering model employ enough 3/2.1
dimensions and freedoms to describe the structural
behavior (e.g., 1-D, 2-D, 3-D, or a combination of these)?
2.1.2 Does the engineering model address the 3/2.1
appropriate scale of response for the problem (e.g.,
global, intermediate, or local response)?
2.1.3 Is the type of analysis appropriate for the type of 3/2.1
response and loading of interest (e.g., linear, static,
modal vibration, frequency response vibration, buckling,
nonlinear, quasi-static, dynamic, crash, impact, fracture,
fatigue, crack, creep, damage, residual fields, pre-
loaded, DoE (design of experiments), optimization –
topology / tomography / etc.)?
2.1.4 Does the engineering model address all the 3/2.1
required results parameters (e.g., critical stress / strain /
displacement / force values, frequency, permanent
deformation, penetration depth, buckling load, contact
force, strain energy, deformation energy, crack driving
force(s) – J / CTOD / CTOA, stress intensity factor – KI /
KII / KIII, weight, stiffness, strength, velocity, acceleration,
etc.)?
2.1.5 Are all assumptions affecting the choice of 3/2.1
engineering model and analysis type justified (watch for
unusual assumptions)?
2.1.6 Is the level of detail, accuracy, or conservatism of 3/2.1
the engineering model appropriate for the criticality of the
analysis and type of problem?
2.1.7 Does the analysis employ a consistent set of units? 3/2.1
2.1.8 Does the analysis employ a consistent global 3/2.1
coordinate axis system, and if applicable, a consistent
local coordinate axis system?

Based on the above checks, answer Question 2.1, and enter the result in Figure A/0-1. Result

A-10
2.1 Are the assumptions of the type of analysis and engineering model acceptable?
Comments

A-11
A/2.2 Analysis Geometry
A/2.2.1 Analysis Geometry General
Perform the following checks to ensure that correct procedures have been followed for defining the
geometric properties of the structure.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.2.1.1 Does the extent of the model geometry capture 3/2.2
the main structural actions, load paths, vibrations (if
applicable), failure mode(s), critical value(s), and
response parameters of interest?
2.2.1.2 If the analysis is vibration analysis, is the model 3/2.2
boundary conditions far enough away from the vibration
response zone(s), e.g. potential maximum
displacement/force location, that the model is not overly
stiff?
2.2.1.3 Are correct assumptions used to reduce the 3/2.2
extent of model geometry (e.g., symmetry, boundary
conditions at changes in stiffness)?

2.2.1.4 Will the unmodelled structure (i.e., outside the 3/2.2

boundaries of the engineering model) have an
acceptably small influence on the results?

2.2.1.5 Are there the impacts of geometric simplifications 3/2.2

(such as omitting local details, cut-outs, holes, stiffeners,
openings, etc.) on the accuracy of the analysis
acceptable?
2.2.1.6 For local detail models, have the aims of Saint- 3/2.2
Venant's principle been satisfied?

2.2.1.7 Do the dimensions defining the engineering 3/2.2

model geometry correspond to the dimensions of the
structure adequately?

2.2.1.8 For buckling analysis, does the geometry 3/2.2

adequately account for discontinuities and imperfections
affecting buckling capacity?

2.2.1.9 Has corrosion been considered over the 3/2.2

applicable design life when selecting the material
thickness? Is this decision in accordance with any
selected evaluation criteria guidance on corrosion?
2.2.1.10 Is the criteria for the use of material thickness in 3/2.2
the model (i.e. net scantlings or gross scantlings)
provided?

Based on the above checks, answer Question 2.2.1 and enter the result in Figure A/0-1. Result

A-12
2.2.1 Are the geometry assumptions in the engineering model acceptable?

Comments

A-13
A/2.2.2 Mass and Added Mass
Perform the following checks to ensure that correct procedures have been followed for defining
the mass and added mass properties of the structure.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.2.2.1 Are all components that have a significant effect 3/2.2.2
on the mass of the structure accounted for in the
engineering model?
2.2.2.2 Has the material density properties been 3/2.2.2
appropriately defined (see also Check 2.3.3)?

2.2.2.3 Has the added mass of entrained water been 3/2.2.2

adequately accounted for with structure partially or
totally submerged?
2.2.2.4 Are lumped mass representations of structural 3/2.2.2
mass and / or equipment correctly consolidated and
located?
2.2.2.5 If rotational inertia is expected to be important, 3/2.2.2
are mass moments of inertia properties correctly defined
for masses?
2.2.2.6 If element erosion is to be considered, are the 3/2.2.2
criteria for mass reduction provided?

2.2.2.7 If the nodal mass element approach is adopted, 3/2.2.2

does its location match the physical observation without
creating additional loads?
2.2.2.8 Are the values of the mass properties considered 3/2.2.2
supported by acceptable calculations and / or
references?
2.2.2.9 If relevant, has fluid-structure interaction been 3/2.2.2
accounted for? Has the added mass been included in
the model?
2.2.2.10 Are the units for the stiffness and mass 3/2.2.2
properties consistent with the system of units for other
parts of the analysis?

Based on the above checks answer, Question 2.2.2, and enter the result in Figure A/0-1. Result
2.2.2 Are the assumptions and data defining the mass and added mass properties
acceptable?
Comments

A-14
A/2.2.3 Shock Analysis Mass Modeling Reduction
In shock analyses, it is sometimes desirable or necessary to reduce the size of the problem by
reducing the number of dynamic degrees of freedom (DOF). Perform these checks to ensure that
the correct procedures have been followed for selecting dynamic degrees of freedom.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.2.3.1 Are dynamic DOF defined along enough 3/2.2.3
directions to model the anticipated dynamic response
behavior of the structure?

2.2.3.2 Are the number of dynamic DOF at least three 3/2.2.3

times the highest mode required (e.g., if 30 modes are
required, at least 90 DOF are needed)?

2.2.3.3 Are the dynamic DOF located where the highest 3/2.2.3
modal displacements are anticipated?

2.2.3.4 Are the dynamic DOF located where the highest 3/2.2.3
mass-to-stiffness ratios occur for the structure?

2.2.3.5 Are dynamic DOF located at points where forces 3/2.2.3

are to be applied for dynamic response analyses?

2.2.3.6 Are the number of dynamic DOF such that at 3/2.2.3

least 90% of the structural mass is accounted for in the
reduced model in each direction?

Based on the above checks, answer Question 2.2.3 and enter the result in Figure A/0-1. Result
2.2.3 Are the assumptions and data defining shock analysis mass modeling acceptable?

Comments

A-15
A/2.3 Material Properties
Perform the following checks to ensure that correct procedures have been followed for defining the
material properties of the structure.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.3.1 Are all materials of structure crucial to the problem 3/2.3
included in the engineering model? (Note: it is preferable
to number each material separately as well as associated
components in the assembly of FE models)
2.3.2 Are the assumed behaviors valid for each material 3/2.3
(eg. linear elastic, nonlinear elastic, elasto-plastic,
isotropic, anisotropic, orthotropic, hardening, fatigue,
fracture, creep, failure, damage, pre-loaded, relaxation,
rate- or temperature dependency, composites)?
2.3.3 Are the required material parameters defined for 3/2.3
the type of analysis? For example:
 Density
 Thermal conductivity (if applicable)
 Electric resist (if applicable)
 Young’s modulus (moduli, if temperature
dependent)
 Poisson’s ratio
 Stress-strain curves or models (if temperature
and/or strain rate dependent)
 Failure criteria (e.g., element erosion,
accumulated strain)
 Damage criteria (e.g., Weibull stress, Lematre
model, and Gurson model)
2.3.4 For cellular type materials (i.e., metallic and non- 3/2.3
metallics), are the required material parameters defined
for the type of analysis? For example:
 Matrix materials
All listed material parameters listed in 2.3.3
 Composite materials
All listed material parameters listed in 2.3.3
 Cellular material (e.g., honeycomb, open- or
close-foams, wood, sandwich structure, glass- or
carbon-fiber-reinforced composite), as well as
associated parameters, defined the material
(e.g., density ratio, strength ratio, representative
cells, interface strength)
2.3.5 Are orthotropic and / or layered properties defined 3/2.3
correctly for non-isotropic materials such as metallic
foam?

2.3.6 Are orthotropic properties defined correctly where 3/2.3

material orthotropy is used to simulate structural
performance?

A-16
2.3.7 Are the correct material models are adopted for 3/2.3
non-isotropic materials?

2.3.8 If strain rate effects are expected to be significant 3/2.3

for this problem, are they accounted for in the material
property data?

2.3.9 If temperature effects are expected to be significant 3/2.3

for this problem, are they accounted for in the material
property data?

2.3.10 If electrical and/or magnetic effects are expected 3/2.3

to be significant for this problem, are they accounted for
in the material property data?

2.3.11 Are the values of the material property data 3/2.3

traceable to an acceptable source or reference (e.g.,
handbook, mill certificate, coupon tests)?

2.3.12 Are the units for the material property data 3/2.3
consistent with the system of units adopted for other
parts of the analysis?

2.3.13 For the analysis requiring plasticity correction 3/2.3

(e.g., low cycle and high cycle fatigue, or yielding state
analysis), is the Neuber correction adopted and
appropriately defined?
2.3.14 If the user-defined material model is adopted in 3/2.3
FE simulation, has the material model been verified
against testing data as well as validated in terms of
reliabilities and accuracies? If possible, the user-defined
material should be enclosed in the submitted report (e.g.,
source code or encrypted format).

Based on the above checks, answer Question 2.3, and enter the result in Figure A/0-1. Result
2.3 Are the assumptions and data defining the material properties acceptable?

Comments

A-17
A/2.4 Loads and Boundary Conditions
Perform the following checks to ensure that correct procedures have been followed for defining the
loads and boundary conditions of the problem.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.4.1 Are all required loadings/load cases accounted for 3/2.3.5
and has sufficient justification been provided for omitting
certain load effects?

2.4.2 Are the loading assumptions stated clearly, and are 3/2.3.5
they justified?

2.4.3 Has an assessment been made of the accuracy 3/2.3.5

and / or conservatism of the loads?

2.4.4 Are the procedures for combining loads/load cases 3/2.3.5

(e.g., superposition) described adequately, and are they
justified?

2.4.5 Have the boundary conditions assumptions been 3/2.3.5

stated clearly, and are they justified?

2.4.6 Do the boundary conditions adequately reflect the 3/2.3.5

anticipated structural behavior?

2.4.7 Has an assessment been made of the accuracy of 3/2.3.5

the boundary conditions, and if they provide a lower or
upper bound solution?

2.4.8 Is the information on loads and boundary conditions 3/2.3.5

provided? For example:
 Loading history and loading sequence (force-
based or displacement-based or other-types)
 Whether inertial relief is needed
2.4.9 If nonlinear loads or boundary conditions (e.g., 3/2.3.5
follower forces or contacts) are considered, is the loading
information converted adequately to represent the
physical problem.
2.4.10 If the analysis includes pre-tension (e.g., bolt 3/2.3.5
preload tension) and/or pre-loaded (e.g., residual stress
and strain), are the input values as well as the
corresponding directions adequate and sufficiently
representing the physical observations.
2.4.11 Are the units for the data of loads and boundary 3/2.3.5
conditions (e.g., forced displacement) consistent with the
system of units adopted for other parts of the analysis?

A-18
2.4.12 If the engineering problem involves contact(s), 3/2.3.5
which type(s) of contact(s) should be adopted? For
example:
 self-contact
 point-to-point contact
 point-to-surface contact
 surface-to-surface contact
2.4.13 If required, are contact model(s) defined and 3/2.3.5
applied appropriately? Are the related parameters
provided and documented? Note that for the frictional
contact, the dynamic and static frictional coefficients
should be distinguished.
2.4.14 If required, have the load and boundary conditions 3/2.3.5
as well as the mesh designing of the contact point(s)
and/or surface(s) been revised to be compatible with the
adopted contact type(s) and model(s)?
2.4.15 If required, have the normal direction(s) and slave 3/2.3.5
and/or master been reviewed for the contact pair(s)?

2.4.16 If a moving rigid barrier(s) is/are used, is the 3/2.3.5

coordinate defining loads or boundary conditions
consistent with the input data, e.g., local or global
coordinate?

Based on the above checks, answer Question 2.4, and enter the result in Figure A/0-1. Result
2.4 Are the assumptions and data defining loads and boundary conditions reasonable?

Comments

A-19
A/2.5 Impact and Plasticity
Perform the following checks to ensure that correct procedures have been followed for an impact
and plasticity problem.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.5.1 Is the study an impact analysis? 3/2.5

2.5.2 Is the study a static or quasi-static analysis 3/2.3

involving plastic deformation?

2.5.3 For the impact analysis, do the material properties 3/2.3.3

depend on the strain rate?

2.5.4 Is it clear which level of impact analysis the study 3/2.5

focuses on?
 Local cross-section
 Component or sub-structure
 Total system
2.5.5 Does the force- or displacement-controlled loading 3/3.5
input provided? For example,
 Quasi-static analysis
o Maximum displacement (displacement
control)
o Maximum force (force control)
 Impact analysis
o Velocity and time curve
o Displacement and time curve
2.5.6 Is the geometric imperfection considered in the 3/3.5
model?

2.5.7 For a perfect geometric shape, is the deformation 3/3.5

trigger considered in the model?

Based on the above checks, answer Question 2.5, and enter the result in Figure A/0-1. Result
2.5 Are the assumptions and data for this impact and plasticity analysis acceptable?

Comments

A-20
A/2.6 Fatigue and Fracture Analysis
Perform the following checks to ensure that correct procedures have been followed for fatigue and
fracture analysis problems.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.6.1 Is the study a fatigue analysis? For example: 3/2.6.1
 Stress life fatigue analysis
 Strain life fatigue analysis
 Fracture mechanics-based fatigue analysis
 Vibration fatigue analysis
2.6.2 Is the study a fracture analysis? For example: 3/2.6.3
 Linear elastic fracture mechanics
 Elastic-plastic fracture mechanics
o Small-scale yielding
o Large-scale yielding
 Elastic-plastic fracture mechanics with multiple
constraint parameters
 Damage included (crack propagation)
2.6.3 Are the material properties provided, including 3/2.3
conditions such as the ambient temperature or corrosive
environment?
 General material properties
 Fatigue analysis
 Stress-cycle (SN) or strain-cycle (EN) curve
 Fracture analysis
2.6.4 Are the loading histories provided? 3/2.6.2.1

2.6.5 Are the geometric detail profiles provided for the 3/3.3.5.3
fatigue analysis?

2.6.6 Are the geometric detail profiles provided for the 3/3.3.5.2
fracture analysis?

2.6.7 Is the crack explicitly included in the model? 3/2.6.2.1

2.6.8 Is the crack extension considered in the fatigue life 3/2.6.2.1

estimation?

2.6.9 Is the initial crack or flaw size correct or calibrated 3/2.6.2.1

using NDT (nondestructive testing) tools?

A-21
2.6.10 If the submodel technique is used for the fatigue 3/2.6.2.1
analysis, does the constraint equation and mesh
refinement defined appropriately?

2.6.11 For the fracture mechanics-based fatigue 3/2.6.2.1

analysis, is the accumulative damage model used, i.e.,
Paris’s law?

2.6.12 For low cycle fatigue analysis, is the Neuber 3/2.3.5

correction method used?

2.6.13 For a fracture analysis, is it clear which 3/2.6.4.2

parameter(s) will be considered? For example:
 Stress intensity factor – K
 J-integral – J
 Crack-tip opening displacement – CTOD

Based on the above checks, answer Question 2.6, and enter the result in Figure A/0-1. Result
2.6 Are the assumptions and data for this fatigue and fracture analysis acceptable?

Comments

A-22
A/2.7 Whole Ship
Perform these checks to ensure that the correct procedures have been followed for a whole ship
analysis.
If the analysis is not a whole ship analysis, you may proceed directly to Part A/2.8.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.7.1 Is the scope of the whole-ship analysis is clear? 3/2.7
Specifically, is it clear which criterion the model is
evaluated against, and how many sub-sections of the
model will be used for the local stress analysis?

2.7.2 If buckling is of interest, is an appropriate buckling 3/2.7

criterion used for the given vessel type and risk profile?

2.7.3 Are the hydrostatic loads applied to all surfaces 3/2.7

correctly?
2.7.4 Is a hull girder design wave applied if applicable? 3/2.7

2.7.5 Does the model mass match the loading condition 3/2.7
being evaluated?

2.7.6 Is the model is hydrostatically balanced, or have 3/2.7

the residual forces been minimized and an appropriate
rigid body displacement restraints been applied?

2.7.7 Does the mesh sizing match the requirements of 3/3.3.1

the applied evaluation criteria?

2.7.8 Are the beam and shell element properties 3/3.3.2

appropriate for the given mesh design?

2.7.9 Is the deflected hull girder shape qualitatively 3/2.7

correct?

Based on the above checks, answer Question 2.7, and enter the result in Figure A/0-1. Result
2.7 Are the assumptions and data defining the whole ship analysis acceptable?

Comments

A-23
A/2.8 Frequency Response Vibration Analysis
Perform these checks to ensure that the correct procedures have been followed for a frequency
response vibration analysis (i.e., a forced vibration analysis).
If the analysis is not a frequency response vibration analysis, you may proceed directly to Part A/3.1.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.8.1 Are all relevant cyclical excitation forces included in 3/2.7
the analysis? If the analysis is a main machinery
analysis, have the (1) the main engine excitation, (2)
propeller alternative thrust excitation, and (3) propeller
hull pressure excitation been considered?
2.8.2 Are the excitation force magnitudes based on 3/2.7
acceptable references?
2.8.3 Have the directions of the excitation forces 3/2.7
provided by the equipment manufacturer been
considered?
2.8.4 Do propeller hull pressure excitation forces have 3/2.7
due consideration for cavitation?
2.8.5 Is the model extent selected with due consideration 3/2.7
for all the aforementioned natural and excitation
frequencies, so that all relevant resonant conditions can
be captured in the model?
2.8.6 Have the hull-girder natural frequencies, main 3/2.7
machinery/shafting system longitudinal vibration natural
frequencies, superstructure fore-and-aft vibration natural
frequencies all been considered?
2.8.7 Does the model include all relevant masses? This 3/2.7
includes the mass of all lightship weight, all loading
condition weight such as cargo and tank liquids, and
added mass where appropriate.
2.8.8 Has the inclusion of damping been considered 3/2.7
appropriately?
2.8.9 Has a free vibration (modal) analysis been 3/2.7
performed to identify both the hull-girder natural
frequencies and the local natural frequencies that may
impact the results? Are these natural frequencies
compared against the excitation frequencies?
2.8.10 If there is more than one cyclic load and these 3/2.7
loads occur at different frequencies simultaneously, are
two analyses run and the results combined during post-
processing?
2.8.11 If there is more than one cyclical load and these 3/2.7
loads occur at the same frequencies simultaneously, are
any phase angle shifts applied correctly? Alternatively,
are the worst-case phase angles used?
2.8.12 Are ranges of frequencies surrounding the 3/2.7
excitation frequencies evaluated as required by Section
3/2.8.3?

A-24
Based on the above checks, answer Question 2.8, and enter the result in Figure A/0-1. Result
2.8 Are the assumptions and data defining vibration analysis acceptable?

Comments

A-25
FINITE ELEMENT ANALYSIS ASSESSMENT FINITE ELEMENT MODEL CHECKS

Project No. I Project Title :

Company Name: IDate:

Analyst: I Checker:

A/3 Finite Element Model Checks

A/3.1 CAD Importing
Perform the following checks to ensure that correct procedures have been followed for importing
CAD. Relevant, material, loading, and boundary condition checks need to be completed in addition
to general geometry considerations listed here.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.1.1 Is the CAD model compatible with the FEA 3/3.1
software?

3.1.2 Does the CAD model include all the geometric 3/3.1
information for FE modeling?

3.1.3 Does the CAD model need to clean up for 3/3.1

meshing?

3.1.4 Does the CAD model need to perform geometric 3/3.1

defeaturing?

3.1.5 Does the CAD model interfere or overlap with other 3/3.1
components in the assembly? If not, do the gap distance
between surfaces recorded?

Based on the above checks, answer Question 3.1, and enter the result in Figure A/0-1. Result
3.1 Are the assumptions and data for CAD importing acceptable?

Comments

A-26
A/3.2 Element Types
Perform these checks to ensure that the correct types of elements have been used to model the
problem.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.2.1 Are all the different types of elements used in the 3/3.2
FEA model identified and referenced in the analysis
documentation?
3.2.2 Are the element types available in the FEA 3/3.2
software used appropriately to ship structural analysis?

3.2.3 Do the element types support the kind of analysis, 3/3.2

geometry, materials, and loads that are of importance for
this problem?
3.2.4 If required, do the selected beam element types 3/3.2
include capabilities to model transverse shear and / or
torsional flexibility behavior?
3.2.5 If required, do the selected beam element types 3/3.2
include capabilities to model tapered, off-set, or
unsymmetric section properties?
3.2.6 If required, do the selected beam element types 3/3.2
include capabilities for nodal DOF end releases (e.g., to
model partially pinned joints)?
3.2.7 If required, do the selected plate element types 3/3.2
include capabilities to model out-of-plane loads and
bending behavior?
3.2.8 If required, do the selected plate element types 3/3.2
include capabilities to model transverse shear behavior
(i.e., thick plate behavior)?
3.2.9 If the model is 2-D, are the selected element types 3/3.2
(or options) correct for plane stress or plane strain
(whichever case applies)?
3.2.10 If required, can the selected element types model 3/3.2
curved surfaces or boundaries to an acceptable level of
accuracy?
3.2.11 If the model includes multiple element types, do 3/3.2
the connected or transient elements have formulations
that allow them to be connected and produced correct
results? Check the software literature if there are
questions about this.
3.2.12 In a linear fracture mechanics analysis, does the 3/3.2
set of 2-D or 3-D elements in the vicinity of the crack tip
have 25% offset towards the tip and coincided/collapsed
nodes, which can represent the stress singularity (i.e.,
1/sqrt(r))?

A-27
 Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.2.13 In a nonlinear fracture mechanics analysis, is the 3/3.2
high-order element type used to capture the crack-tip
blunting?
3.2.14 If required for automatic mesh refinement, does 3/3.2
the element type is compatible with user-defined mesh
splitting or mesh refinement algorithm?
3.2.15 In an analysis with element erosion or structural 3/3.2
damage (e.g., flaw coalescence and crack propagation),
is the element type compatible with material models?
Have the element parameters been appropriately
defined?
3.2.16 In certain special analyses, hybrid or user-defined 3/3.2
types of elements (note: user compiling needed) need to
be adopted. Does the definition of elements meet the
analysis requirements and can the analysis be performed
accurately?
3.2.17 Have the limits of each element type in the model 3/3.2
been documented?

3.2.18 For shell elements, are the number of integration 3/3.2

points in-plane and through-thickness sufficient to
capture tension/compression and bending behaviors?
Have the locations of these integration points been
recorded?
3.2.19 For solid elements, are the number of integration 3/3.2
points sufficient to capture tension/compression and
bending behaviors? Have the locations of these
integration points been recorded?
3.2.20 For lower-order elements, are the numbers of the 3/3.2 ,
DOF of each node (especially the rotational DOF)
appropriately defined to keep the calculation accuracy?
3.2.21 For 2-D shell elements, if the thickness offset 3/3.2
feature is activated, has the offset value been verified
and documented?
3.2.22 If the automatic mesh refinement feature is 3/3.2
activated, is the algorithm compatible with the current
element type without inducing element inconsistency?
3.2.23 If the element erosion feature (associated with 3/3.2
material model) is activated, is the calculation of critical
values verified without inducing inaccuracy?
3.2.24 If a special element, e.g., a cohesive element, is 3/3.2
adopted, are the related variables appropriately defined
based on the FEA codes? Note: there are differences
among FEA codes when dealing with special elements.
3.2.25 If a user-defined element is adopted, is the 3/3.2
element verified and documented? It is recommended to
enclose the source code or encrypted file in the report.

A-28
Based on the above checks, answer Question 3. 2 and enter the result in Figure A/0-1. Result
3.2 Are the types of elements used in the FEA model acceptable?
Comments

A-29
A/3.2.1 Stiffness and Thickness Properties
Perform the following checks to ensure that correct procedures have been followed for defining the
stiffness and thickness properties of the structure.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.2.1.1 Are all components that have a significant effect 3/3.2.1
on the stiffness of the structure accounted for in the
engineering model (under-considered load cases)?
3.2.1.2 Are the assumed stiffness behaviors valid for 3/3.2.1
each structural component (e.g., linear, membrane,
bending, shear, and torsion)?
3.2.1.3 Are the required stiffness parameters defined for 3/3.2.1
each component and each contact pair, e.g.:
 Truss members – A
 Beams, bars – A, Iyy, Izz, others
 Plates, shells – t (uniform or varying, or offset)
 Springs – K (axial or rotational)
 Contact springs – K, friction coefficient
3.2.1.4 Do the section properties of stiffeners (where 3/3.2.1.1
modeled with beams) include correct allowances for the
effective plate widths?
3.2.1.5 If torsion flexibility is expected to be important, 3/3.2.1.1
are torsion flexibility parameters correctly defined for
beam sections?
3.2.1.6 If shear flexibility is expected to be important, are 3/3.2.1.1
shear flexibility parameters correctly defined for beam
and/or plate elements?
3.2.1.7 Are the values of the stiffness and thickness 3/3.2.1.1
properties data supported by acceptable calculations
and / or references?
3.2.1.8 Are the units for the stiffness and thickness 3/3.2.1.1
properties data consistent with the system of units for
other parts of the analysis?
3.2.1.9 Are the orientations of the beam element axes 3/3.2.1.1 3.2
correct for the defined section properties?

3.2.1.10 Are the cross-section of the beam elements 3/3.2.1.1 3.2

correct for the defined section properties?

Based on the above checks, answer Question 3.2.1, and enter the result in Figure A/0-1. Result
3.2.1 Are the assumptions and data defining stiffness and thickness properties acceptable?

A-30
Comments

A-31
A/3.3 Mesh Design
As a finite element method is essentially a piece-wise approximation technique, the accuracy is
dependent mainly on the mesh design and corresponding quality. Perform the following checks to
ensure that the finite element mesh is acceptable.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.3.1 Does the mesh design adequately reflect the 3/3.3
geometry of the problem (e.g., overall geometry, stiffener
locations, and details)?
3.3.2 Does the mesh design adequately reflect the 3/3.3
anticipated structural response (e.g., stress/strain
gradients, stress/strain concentrations, deflections, mode
shapes, failure modes, load paths)?
3.3.3 Is the mesh free of unintentional gaps or cracks, 3/3.3
overlapping, or missing elements?

3.3.4 Is the mesh free of duplicate elements? 3/3.3

3.3.5 Is proper node continuity maintained between 3/3.3

adjacent elements (also continuity between beam and
plate elements in stiffened panels)?
3.3.6 Does the analysis documentation state or show that 3/3.3
there are no "illegal" elements in the model (i.e., no
element errors or warnings)?
3.3.7 Are element aspect ratios acceptable (e.g., < 5 for 3/3.3
shell), particularly near and in the areas of interest?

3.3.8 Are element taper angles acceptable (e.g., < 0.6 for 3/3.3
shell), particularly near and in the areas of interest?

3.3.9 Are element skew angles acceptable (e.g., < 60 for 3/3.3
shell), particularly near and in the areas of interest?

3.3.10 If flat shell elements are used to model curved 3/3.3

surfaces, are the curve angles < 10° for stresses, or < 15°
for displacement results?
3.3.11 If flat shell elements are used for double or tapered 3/3.3
curve surfaces, is warping avoided or appropriately
controlled (e.g., small curve angles, use of triangles, or <
20 for quad shell)?
3.3.12 Does the analysis documentation state or show that 3/3.3
there is an upper limit of the percentage of triangle shell
and/or tetra solid elements in the model?

A-32
 Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.3.13 Are differences in rotational DOF I moment 3/3.3
continuity for different element types accounted for (e.g.,
beam joining solid)?
3.3.14 Are the element shapes in the areas of interest 3/3.3
acceptable for the types element used and degree of
accuracy required?
3.3.15 Are mesh transitions from coarse regions to areas 3/3.3
of refinement acceptably gradual?

3.3.15 Are mesh transitions from coarse regions to areas 3/3.3

of refinement having no disconnected nodes/elements?

3.3.16 Are nodes and elements correctly located for 3/3.3

applying loads, support and boundary constraints,
connections to other parts, and pairs of contacts?
3.3.17 Are the outward normal directions for plate/shell 3/3.3
elements of a surface in the same direction?

3.3.18 If the analysis is a vibration analysis, is the mesh 3/3.3.5.4

size sufficient to model the relevant vibration modes.

Based on the above checks, answer Question 3.3, and enter the result in Figure A/0-1. Result
3.3 Is the design of the finite element mesh acceptable?

Comments

A-33
A/3.4 Substructures and Submodeling
Substructuring or submodeling techniques may be employed to reduce the size of the problem for
computing and / or to take advantage of repetitive geometry in the structure. Perform the following
checks to ensure that the acceptable procedures have been followed.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.4.1 Is the overall substructure or submodeling scheme 3/3.3.8
or procedure adequately described in the analysis
documentation?

3.4.2 Are all individual substructure models, global 3/3.3.8

models, and refined submodels identified and described in
the analysis documentation?
3.4.3 Are the master nodes located correctly, and are the 3/3.3.8
freedoms compatible for linking the substructures?

3.4.4 Are the master nodes located correctly for the 3/3.3.8
application of loads and boundary conditions upon
assembly of the overall model?
3.4.5 Are loads and boundary conditions applied at the 3/3.3.8
substructure level consistent with those of the overall
model?
3.4.6 Does the boundary of the refined submodel match 3/3.3.8
the boundary of coarse elements/nodes in the global
model in the region of interest?
3.4.7 If the technique of super-element is used, does the 3/3.3.8
boundary as well as the node numbers of submodel
match the super-elements at the region of interest,
especially in the use of including FEA files (e.g., Nastran
*.pch files)?
3.4.8 Is the boundary for the submodel at a region of 3/3.3.8
relatively low-stress gradient or sufficiently far away from
the area of primary interest?
3.4.9 Does the refined submodel correctly employ forces 3/3.3.8
and/or displacements from the coarse model as boundary
conditions?
3.4.10 Does the submodel include all other loads applied 3/3.3.8
to the global model (e.g., surface pressure and
acceleration loads)?
3.4.11 Have stiffness differences between the course 3/3.3.8
global mesh and refined submodel mesh been adequately
accounted for?
3.4.12 Are the super-element(s) created correctly and 3/3.3.8
verified by trial runs? (Note: it is recommended to carry
out a preliminary modal analysis, and it is commonly seen
a certain degree of difference in terms of eigenvalues.)

A-34
 Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.4.13 Is the gravity center of the submodel or 3/3.3.8
substructure reviewed?

3.4.14 Does the model size (in terms of the total DOF of 3/3.3.8
global models, including submodel or substructure or
super element) meet the hardware capacity?

Based on the above checks, answer Question 3.4, and enter the result in Figure A/0-1. Result
3.4 Are the substructuring or submodeling procedures acceptable?

Comments

A-35
A/3.5 Loads and Boundary Conditions
Perform the following checks to ensure that correct procedures have been followed for defining the
loads and boundary conditions of the finite element model.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.5.1 Are point load forces applied at the correct node 3/3.5
locations on the structure, and are they defined with the
correct units, magnitude, and direction?

3.5.2 Are distributed loads applied at the correct 3/3.5

locations on the structure, and are they defined with the
correct units, magnitude, and direction?

3.5.3 Are surface pressure loads applied at the correct 3/3.5

locations on the structure, and are they defined with the
correct units, magnitude, and direction?

3.5.4 Are translational accelerations defined with the 3/3.5

correct units, and do they have the correct magnitude
and direction?

3.5.5 Are rotational accelerations defined with the correct 3/3.5

units, magnitude, and direction and about the correct
center of rotation?

3.5.6 Are prescribed displacements applied at the correct 3/3.5

locations on the structure and are they defined with the
correct units, magnitude, and direction.

3.5.7 Are the displacement boundary conditions applied 3/3.5

at the correct node locations?

Based on the above checks, answer Question 3.5, and enter the result in Figure A/0-1. Result
3.5 Are the loads and boundary conditions applied correctly?

Comments

A-36
A/3.6 Analysis Controls and Solution Options
Perform the following checks to ensure that correct solution options, techniques, or procedures have
been used for the finite element model.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
3.6.1 Have any special solution options and procedures 3/3.6
been used, and, if so, have they been documented?

3.6.2 If non-standard options been invoked, have they 3/3.6

been documented, and the reasons for their use been
explained?

3.6.3 If the problem is a dynamic analysis, is the method 3/3.6

for eigenvalue and mode extraction appropriate?

Based on the above checks, answer Question 3.6, and enter the result in Figure A/0-1. Result
3.6 Are the solution options and procedures followed for the FEA acceptable?

Comments

A-37
FINITE ELEMENT ANALYSIS ASSESSMENT FINITE ELEMENT RESULTS CHECKS

Project No. I Project Title :

Company Name: IDate:

Analyst: I Checker:

A/4 Finite Element Results Checks

A/4.1 General Solution Checks
Perform these checks to expose any gross errors. Most program output values of gross parameters
associated with the solution process. These parameters typically include summed applied loads and
reactions, total mass, the position of the center of gravity, time steps, and other related parameters.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.1.1 Are all error and warning messages issued by the 3/4.1
software reviewed and understood?

4.1.2 Is the magnitude of the mass of the finite element 3/4.1

model approximately as expected?

4.1.3 Is the location of the center of gravity of the model, 3/4.1

as calculated by the program, reasonable?

4.1.4 Are the applied forces in equilibrium with the 3/4.1

applied reactions?

4.1.5 Are the reaction loads and moments at any 3/4.1

supports added to prevent rigid body motion minimized?

4.1.6 Does the solution converged within a number of 3/4.1

time steps?

Based on the above checks, answer Question 4.1, and enter the result in Figure A/0-1. Result

4.1 Are the general solution parameters acceptable?

Comments

A-38
A/4.2 Post Processing Methods
Perform these checks to ensure that the methods and their limitations, used by the program to
post-process the results are understood.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.2.1 Are the methods for reducing the analysis results 3/4.2
described (e.g., calculation of safety factors and other
parameters calculated by manipulating raw output)?

4.2.2 Are the methods for "correcting" FE results 3/4.2

described (e.g., correction factors and smoothing
factors)?

4.2.3 Are any path plotting of data and similar data 3/4.2
collection tools appropriately applied to collect data on
the correct surface or mid thickness location of the
element?

Based on the above checks, answer Question 4.2, and enter the result in Figure A/0-1. Result
4.2 Is the methodology used for post-processing the results satisfactory?

Comments

A-39
A/4.3 Displacement Results
Perform these checks to ensure that the displacement results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.3.1 Are the displacement results described and 3/4.3
discussed?

4.3.2 Are plots of the deformed structure (or mode 3/4.3

shape) presented?

4.3.3 Are the directions of displacements consistent with 3/4.3

the geometry, loading, and boundary conditions?

4.3.4 Do the magnitudes and sense/direction of 3/4.3

displacements make sense?

4.3.5 Is the deformed shape (or mode shape) smooth 3/4.3

and continuous in the area of interest?

4.3.6 Are unintentional slits or cuts (indicating elements 3/4.3

not connected where they should be) absent?

Based on the above checks, answer Question 4.3, and enter the result in Figure A/0-1. Result

4.3 Are displacement results consistent with expectations?

Comments

A-40
A/4.4 Force Results
Perform these checks to ensure that the force results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.4.1 Are the force results available for review? For 3/4.4
example, the results are exported into text files.

4.4.2 Do the force magnitudes and directions make 3/4.4

sense? Are they explainable?.

4.4.3 Do the forces and moments balance? For example: 3/4.4

 Global-level
 Local-level
 Contact forces
 Following forces
4.4.4 For strain-life fatigue analysis, are the force levels 3/4.4
suitable for using Neuber correction method?

4.4.5 For the fracture analysis, are the force results using 3/4.4
for validating fracture toughness J-integral?

Based on the above checks, answer Question 4.4, and enter the result in Figure A/0-1. Result
4.4 Are force results consistent with expectations?
Comments

A-41
A/4.5 Stress Results
Perform these checks to ensure that the stress results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.5.1 Are the stress results described and discussed? Is 3/4.5
any stress index used (i.e. equivalent von Mises)?

4.5.2 Are stress contour plots presented? In the stress 3/4.5

plots, are the stress parameters or components defined
using appropriate units and scale/fringe levels?
4.5.3 Is the method of smoothing stress results, or 3/4.5
averaging stress results described (e.g., element
stresses vs. nodal average stresses)?
4.5.4 Are the units of stress parameters consistent? 3/4.5

4.5.5 Are the magnitudes of stresses consistent with 3/4.5

intuition?

4.5.6 In cases where there are adjacent plate elements 3/4.5

with different thicknesses, does the method for averaging
stresses account for the differences?

4.5.7 Are the stress contours smooth and continuous, 3/4.5

particularly in the region of primary interest?

4.5.8 Are the stress contours at boundaries consistent 3/4.5

with the boundary conditions applied (e.g., stress
contours perpendicular to the boundary if there is a
symmetry boundary)?
4.5.9 Are stresses local to the applied loads reasonable? 3/4.5

4.5.10 Do stress levels at the boundaries (applied loads 3/4.5

and fixtures) make an engineering sense by comparing it
with the area of interest? Note that the area of interest
should be sufficiently remote from boundaries (applied
loads and fixtures).
4.5.11 Are there areas in which stresses are above yield 3/4.5
(which would invalidate linear elastic analysis)? Are
these areas small enough to be ignored?

Based on the above checks, answer Question 4.5, and enter the result in Figure A/0-1. Result
4.5 Are stress results consistent with expectations?

A-42
Comments

A/4.6 Strain Results

Perform these checks to ensure that the strain results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.6.1 Are the strain results described and discussed? 3/4.6

4.6.2 Are strain contour plots presented? In the strain 3/4.6

plots, are the strain parameters or components defined
using appropriate units and scales/fringe levels?
4.6.3 Is the method of smoothing strain results, or 3/4.6
averaging strain results described (e.g., element strains
vs. average nodal strain)?
4.6.4 Are the units of strain parameters consistent? 3/4.6

4.6.5 Are the magnitudes and directions of strains 3/4.6

consistent with intuition?

4.6.6 In cases where there are adjacent plate elements 3/4.6

with different thicknesses, does the method for averaging
strain account for the differences?

4.6.7 Are the strain contours smooth and continuous, 3/4.6

particularly in the region of primary interest?

4.6.8 Are the strain contours at boundaries consistent 3/4.6

with the boundary conditions applied (e.g., strain
contours perpendicular to the boundary if there is a
symmetry boundary)?
4.6.9 Are strains local to the applied loads reasonable? 3/4.6

4.6.10 Are there areas in which strains are over yield 3/4.6
(which would invalidate linear elastic analysis)?

Based on the above checks, answer Question 4.6, and enter the result in Figure A/0-1. Result
4.6 Are strain results consistent with expectations?

A-43
Comments

A-44
A/4.7 Energy Results
Perform these checks to ensure that the energy results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.7.1 Are the energy-related parameters exported into 3/4.7
text files? For example:
 Strain energy
 Internal energy
 Kinetic energy
 Contact energy
 Hourglass energy
 Damping energy
4.7.2 Are the internal and external energy balanced? 3/4.7

4.7.3 Are the amount of hourglass and/or damping 3/4.7

energy within the limit?

4.7.4 For the fracture analysis, are the energy-related 3/4.7

fracture toughness parameter employed?

Based on the above checks, answer Question 4.7, and enter the result in Figure A/0-1. Result
4.7 Are energy results consistent with expectations?
Comments

A-45
A/4.8 Fracture Results
Perform these checks to ensure that the fracture results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.8.1 Are the fracture mechanics parameters exported? 3/4.8
For example:
 J (J-integral)
 K (stress intensity factor)
 CTOD (crack-tip opening displacement)
 CTOA (crack-tip opening angle)
 C* (creep fracture mechanics parameter)
4.8.2 Are the results used for a fatigue analysis? Are the 3/4.8
results adequate for evaluating the fatigue crack growth
rate?
4.8.3 Are the results used for an integrity analysis? If yes, 3/4.8
which design code does the analysis follows? Is the
analysis procedure documented?
4.8.4 Are the results adequate for evaluating parameters 3/4.8
in integrity analysis, e.g., the FAD method?

Based on the above checks, answer Question 4.8, and enter the result in Figure A/0-1. Result
4.8 Are fracture results consistent with expectations?
Comments

A-46
A/4.9 Fatigue Results
Perform these checks to ensure that the fatigue results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.9.1 Are the fatigue life calculation performed manually, 3/4.9
in software or explicitly in the FEA?

4.9.2 Which index is used for fatigue life estimation? For 3/4.9
example:
 Absolute or signed maximum shear stress
 Absolute or signed von Mises stress
 Absolute or signed maximum principal stress
4.9.3 What is the safety factor (SF)? For example: 3/4.9
 SF = 2

4.9.4 Does the estimated fatigue life agree with that from 3/4.9
testing?
4.9.5 Does the explicit crack growth direction or stress 3/4.9
direction make sense?
4.9.6 Is the crack tip plasticity reasonably localized? 3/4.9

4.9.7 Are the node release criteria respected? 3/4.9

4.9.8 Are the load shedding and boundary conditions of 3/4.9

sub-models behaving appropriately?

Based on the above checks, answer Question 4.9, and enter the result in Figure A/0-1. Result
4.9 Are fatigue results consistent with expectations?
Comments

A-47
A/4.10 Vibration Results
Perform these checks to ensure that the vibration analysis results are consistent with expectations.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
4.10.1 Are the frequencies expressed in the correct 3/4.10
units?

4.10.2 Are the magnitudes of natural frequencies 3/4.10

consistent with the type of structure and the modal
number?

4.10.3 Are the mode shapes smooth? 3/4.10

4.10.4 Are results being reported for a range of 3/2.8.3

frequencies near the excitation frequency? 3/4.10.2

4.10.5 Do the results and criteria being used consider 3/4.10.2

that the local response magnitudes may be off by a
factor of 2 to 3?

Based on the above checks, answer Question 4.10, and enter the result in Figure A/0-1. Result
4.10 Are vibration results consistent with expectations?

Comments

A-48
FINITE ELEMENT ANALYSIS ASSESSMENT CONCLUSIONS CHECKS

Project No. I Project Title :

Company Name: IDate:

Analyst: I Checker:

A/5 Conclusion Checks

A/5.1 FEA Results and Acceptance Criteria
Perform these checks to ensure that the results are in a form suitable for comparison with specified
acceptance criteria.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
5.1.1 Are the results presented in units and summarized 3/5.1
in a manner that allows comparisons with acceptance
criteria, or alternative solutions or data?

5.1.2 Are satisfactory explanations provided where the 3/5.1

results do not meet acceptance criteria, or where they
differ significantly from other comparable solutions or
data?

Based on the above checks, answer Question 5.1, and enter the result in Figure A/0-1. Result
5.1 Are the results presented in sufficient detail to allow comparison with acceptance
criteria?

Comments

A-49
A/5.2 Load Assessment
Perform these checks and evaluations to ensure that the loads applied in the FEA, and their
accuracy, are understood.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
5.2.1 Has an assessment been made of the accuracy or 3/5.2
degree of conservatism of the loads used in the FE
model with respect to the following aspects:

a. types of loads/load cases that were included and excluded

b. basis or theory used to derive loads (e.g., linear strip theory for
sea motion loads, base acceleration vs. DRS for shock, drag
coefficients for wind loads, etc.)
c. magnitudes of loads

d. loading directions included / excluded

e. load combinations

f. load factors

g. boundary conditions

Based on the above checks, answer Question 5.2, and enter the result in Figure A/0-1. Result
5.2 Are the accuracy and conservatism, or otherwise, of the applied loading modeled
understood?

Comments

A-50
A/5.3 Strength / Resistance Assessment
Perform these checks and evaluations to ensure that an adequate assessment of the capability of
the structure has been made.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
5.3.1 Has an assessment been made of the accuracy or 3/5.3
degree of conservatism of the strength or resistance of
the modeled structure with respect to the following
aspects:
a. failure theory, failure criteria, allowable stresses, safety factors,
etc.
b. section properties

c. material properties

d. allowances for imperfection, misalignment, manufacturing

tolerances
e. allowances for corrosion

Based on the above checks, answer Question 5.3, and enter the result in Figure A/0-1. Result
5.3 Has an adequate assessment been made of the capability of the structure?

Comments

A-51
A/5.4 Accuracy Assessment
The checks listed below are intended to ensure that an attempt has been made to assess the
accuracy of the FEA.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
5.4.1 Has an assessment been made of the scale of the 3/5.4
FE model and its level of detail and complexity?

5.4.2 Have the types of behavior modeled and not 3/5.4

modeled (e.g., membrane only instead of membrane plus
bending) been assessed?

5.4.3 Has the influence of mesh refinement on accuracy 3/5.4

been considered?

5.4.4 Has a comparison with other results (e.g., other 3/5.4

solutions, and experiments) been made?

5.4.5 Based on the above, has an overall assessment of 3/5.4

the accuracy of the relevant results been made?

Based on the above checks, answer Question 5.4, and enter the result in Figure A/0-1. Result
5.4 Has an adequate assessment of the accuracy of the analysis been made?

Comments

A-52
A/5.5 Overall Assessment
The checks listed below are to ensure that the overall conclusions and recommendations resulting
from the FEA have been presented and are generally satisfactory.
Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
5.5.1 Are conclusions from the FEA provided, and are 3/5.5
they consistent with the material presented?

5.5.2 If appropriate, has a way ahead or potential 3/5.5

solutions been presented?

5.5.3 Based on consideration of all previous checks, is 3/5.5

the overall assessment that the FEA is acceptable?

Based on the above checks, answer Question 5.5, and enter the result in Figure A/0-1. Result
5.5 Is the finite element analysis generally assessed satisfactorily?

Comments

A-53
 Appendix B
Example Application of Assessment Methodology

B-1
B1.0 INTRODUCTION
The purpose of this Appendix is to illustrate the application of the FEA
assessment methodology, and the guidelines presented in Parts 2 and 3 of
this document.
An example finite element analysis (FEA) of a web frame from an Arctic-going
tanker design subject to ice loads is used for this purpose. The approach used
to illustrate the assessment methodology and guidelines includes :
• a sample report of the Arctic tanker web frame FEA annotated with
references to relevant sections of the FEA assessment methodology and
guidelines; and
• completed checklists as required by the assessment methodology.
The annotated report and the completed checklists are presented in
Annexes 8-1 and 8-4, respectively.
B2.0 EXAMPLE FINITE ELEMENT ANALYSIS

The example FEA is adapted from an analysis for an actual design1 of an

icebreaking tanker. The tanker is double-hulled. Transverse strength is provided
by a series of closely spaced web frames, and the longitudinal load transfer is
achieved through several longitudinal stringers. The design requirements are
based on current Canadian rules.
The primary interest tor this analysis is the behavior of a typical web frame in
response to ice loads. Other loads are ignored as negligible compared with the
ice loads. The analysis was performed to ensure that the side structure that
directly resists the ice loads responds in the manner expected by the designers
and that the structure is as optimized as possible.
This example illustrates several aspects of finite element modeling common in
ship structures, including:
• behavior of stiffened plate structures
• openings in structures
• discontinuities often found in ship structures
• integrated nature of typical ship structures
• use of most types of elements commonly used in the FEA of ship structures.
For reasons explained in Annex B-1, it was necessary to make modifications
to the original analysis, particularly in regard to the level of ice load, to make it
suitable for the purposes of the present work.

1
The design was undertaken by Canarctic Shipping Co. Ltd., Ottawa, Ontario, Canada under

B-2
 contract to the Transportation Development Centre, Montreal, Quebec, Canada
B3.0 ANNOTATED REPORT

Annex B-1 presents a sample report of the Arctic tanker web frame FEA that has
been prepared by a contractor ("BB Engineering") and has been subjected to the
assessment methodology. For illustrative purposes, the report has been
annotated with short descriptions identifying the relevant part of the assessment
methodology presented in Parts 2 and 3 of this document. Except for the
annotations, the report is meant to be typical of the documentation that an
evaluator of FEA might receive.

B4.0 CHECKLISTS

A sample of completed FEA evaluation checklists for the report in Annex B-1 is
presented in Annex B-4.

Acknowledgment

The finite element analysis described in the following pages is adapted from an
analysis performed by MIL Systems Engineering, Ottawa, Ontario, for Canarctic
Shipping Ltd., Ottawa, Ontario, under a contract awarded by the Transportation
Development Centre, Montreal, Quebec.

Warning

This example is presented solely for the purpose of illustrating the

assessment methodology described in Part 2. As such, it is not necessarily
complete in all details, particularly in regard to parameters such as number
of loading types, design criteria, and number of structural responses
considered. Furthermore, this example should not be construed as
representative of the requirements for finite element analysis of other marine
structures.

B-3
 Annex B-1
Finite Element Analysis of Arctic Tanker Web
Frame

BB Engineering Ltd.
13-1300 Finite Drive
Ottawa, Ontario
XXX XXX

B-4
 May 1995

B-5
1.0 INTRODUCTION 8-6
2.0 PRELIMINARY INFORMATION 8-6
2.1 JOB SPECIFICATION 8-6
2.2 RATIONALE FOR USING FINITE ELEMENT METHOD 8-7
2.3 FEA SOFTWARE 8-7
2.4 CONTRACTOR AND ANALYST QUALIFICATIONS 8-7
3.0 ENGINEERING MODEL 8-7
3.1 ANALYSIS TYPE AND ASSUMPTIONS 8-7
3.2 GLOBAL GEOMETRY OF 50000 DWT TANKER 8-8
3.3 FRAME SELECTED 8-8
3.4 EXTENT OF MODEL 8-9
3.5 MATERIAL PROPERTIES 8-9
3.6 INTERACTION WITH ADJACENT STRUCTURE .............................. B-10 3.7 LOADS
8-11
3.8 BOUNDARY CONDITIONS 8-12
4.0 FINITE ELEMENT MODEL 8-12
4.1 GENERAL INFORMATION 8-12
4.2 ELEMENT SELECTION 8-12
4.3 MESH DESIGN 8-13
4.4 FINITE ELEMENT ATTRIBUTES AND SPRING CONSTANTS . . . . . . . . . . . . . . . . . . . . . . B-
14
4.5 FE MODEL LOADS AND BOUNDARY CONDITIONS 8-16
4.6 FE MODEL CHECKS 8-17
4.7 FE SOLUTION OPTION AND PROCEDURES 8-18
5.0 ANALYSIS RESULTS 8-18
5.1 GENERAL SOLUTION CHECKS 8-18
5.2 POST PROCESSING METHODS ..................................... B-18
5.3 STRUCTURAL RESPONSE 8-19
6.0 CONCLUSIONS 8-20
7.0 REFERENCES . . . . . . . . . . . . . . .. . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . .
. . . B-20
Annex 8-2 Company and Personnel Qualifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-
35
8-2.1 Contractor Qualifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-35
B-2.2 Personnel Qualifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B-35
Annex 8-3 FEA Results Verification ............................................................................ 8-36
Annex 8-4 Sample Completed Assessment Methodology Forms 8-37

..... .... -
- _

B-6
FINITE ELEMENT ANALYSIS OF 50000 DWT TANKER
SINGLE MIDBODY WEB FRAME

1.0 INTRODUCTION

AA Shipping Company Limited has developed a design for

a 50000 DWT Arctic tanker. The focus of the work has been
to design cost-optimized midbody and bow structures.
The BB Engineering Co Ltd. (BBE) has been tasked to
undertake a finite element analysis (FEA) of a typical midbody
web (diaphragm) frame. The purpose of the FEA reported in
this report is to assess the response of the midship structure to
ice loads.
Section 2 of this report provides a summary of the
requirements for the analysis, and data on the software and
the resources applied to the problem. The engineering model
is described in Section 3. This section includes a discussion of
the subject structure and the assumptions made in developing
the engineering model. Section 4 describes the finite element
model, and Section 5 presents the results of the analysis.

2.0 PRELIMINARY INFORMATION

2. 1 Job Specification
The job specification calls for a static, linear elastic, FEA of a web frame from the
midbody section of the 50000 DWT tanker at a design ice load of 4435 kN.
The finite element model is based on the drawings provided in Arctic Tanker
Structural Evaluation - Midship Sections, Bow Sections, and Repair Drawings (Ref.
2).
The acceptance criteria for the analysis are as follows:
1. maximum stress not to exceed the material yield stress except as noted in item
2.
2. very localized stresses in excess of yield stress are considered acceptable

B-7
2.2 Rationale for using Finite Element Method
The structure under investigation is too complex to be
analyzed by hand calculation particularly in regions of high-
stress concentrations.
FEA Software
2.3 FEA Software Para. 1.3.1
ANSYS finite element software (Version 5.1), developed and
supported by ANSYS Inc. of Houston, PA, was used for the
finite element work performed and presented here. ANSYS
is a well-established FEA package that has a proven track
record in analyzing structures of the type under
consideration. BBE currently has maintenance and technical
support contract with ANSYS, Inc.
The software updates and error reports received from ANSYS
are reviewed by all BBE staff involved in FEA and filed along
with other ANSYS documents. ANSYS's shell and beam
elements have been validated by BBE for use in ship structural
analysis. ANSYS has been evaluated against benchmarks
designed to test the capability of the software to perform ship
structural FEA.
2.4 Contractor and Analyst Qualifications
Information on qualifications of the contractor, the analysts,
and the supervisor, to perform the required FEA is provided Contractor I Personnel
in Annex B-2 of this document. Qualification
Para. 1.4
3.0 ENGINEERING MODEL
3.1 Analysis Type and Assumptions
Since the stresses are limited to the yield stress, the material
behavior is assumed to be linear. Similarly, because large
deflections are not expected geometric behavior is assumed to
be linear as well.
Analysis Type &
Justification for using FEA
Assumptions
Para. 1.2.6 Para. 2.

The load is assumed to be static, and interest is centered on the

strength of the frame. Hence, the dynamic behavior of the frame is
not within the scope of this analysis. Instability behavior is also not
considered in this analysis. However, it should be considered as part
of the design process.

B-8
B-9
The overall strength of the frame is the primary focus of this
analysis, and therefore the analysis is not optimized to
examine stress concentrations at structural discontinuities
such as those that will exist around openings for example.
Again, these should be addressed as part of the normal
design process.
3.2 Global Geometry of 50000 DWT Tanker
The 50000 DWT tanker has a waterline length of 242 meters, Geometry
a breadth of 34.6 meters and a depth of 18.1 meters. The Assumption
vessel has seven cargo tanks. In the cargo tank region of the
s Para. 2.2
vessel the distance between transverse bulkheads is 19.2
meters. Each cargo tank has approximate dimensions of 18 m
x 30.6 m x 14.6m.
The vessel is double-hulled. The distance between the inner
and outer hulls is 2000 mm. The bottom structure wraps
around the turn of the bilge and connects to the side shell
structure at a point
4.0 meters above baseline. The side shell structure connects
with the deck structure at a point of 15.0 meters above the
baseline. Therefore, the side shell structure vertically spans a
distance of
11.0 meters. The structure is transversely framed by web
frames (diaphragms) spaced at 1000 mm intervals.
Longitudinal framing is provided by several stringers spanning
between bulkheads.
The midship section is shown in Figure1 3.1
3.3 Frame Selected
The ice load for the 50000 DWT tanker is approximately 22 Extent of Model
meters in length and, therefore, if centrally positioned, spans Para. 2.2.1
across a pair of bulkheads. The ice load applied to side
structure is resisted by the transverse frames (each acting
essentially as a ring), the deck structure, the bottom structure
and the transverse bulkheads.
Any transverse loads applied to the side structure are
distributed vertically to the bottom and deck structures by
transverse frames,
· and longitudinally to bulkheads through stringers.

B-10
 The most severe loading case for a web frame is from ice load
1 Figures are presented at the end of this document

B-11
applied to the frame midway between bulkheads and
centrally disposed with respect to the frame. The
characteristics of the load are discussed in Section 3. 7.

3.4 Extent of Model

The structure of the vessel, between transverse bulkheads, Extent of Model

is a series of ring frames comprising inner and outer hull Para. 2.2.1
plating with a stiffened plate diaphragm connecting them.
These frames are connected by all longitudinally oriented
An alternative method
structure (framing members and plating). to account for the
It is sufficient to model a single transverse ring frame if the influence of the
correct boundary conditions are applied as discussed in surrounding structure
would be to model
Section 3.6. Due to the symmetry (structure and load) that
adjacent web frames
exists along the vessel centerline it is also sufficient to
and stringers
model one half of the ring frame.
approximately.
This ring frame extends from the bottom of the ship at
centerline around to the vessel centerline at the deck. The
width of the model needs to be the frame spacing (1000
mm) and will include the inner and outer shell plating and
the stiffened plate diaphragm.

Figure 3.1 illustrates the midbody frame that was analyzed.

Figure
3.2 shows the outer dimensions for the frame.

3.5 Material Properties

Figure 3.2 indicates that the vessel material in the outer

shell plating is Grade EH50 and that the inner shell and
framing components are Grades DH36 and EH36. Table 3.1 Material Properties
lists the relevant material properties as taken from Para. 2.3
Reference 3 for these steel grades.
The Young's Modulus was taken as 208,700 MPa for all
steel types. Parameters such as initial imperfections and
residual strains were not included in the analysis, and no
allowance is made for corrosion. These assumptions are
consistent with the design criteria.

B-12
 TABLE 3.1: Steel Mechanical Properties

Property Steel Grade

EH50 DH36,EH36

Yield Stress (min.) (MPa) 500 355

Tensile Stress (MPa) 610-770 490-620

Elongation % 16 21

Young's Modulus (MPa) 208700 208700

Poisson's Ratio 0.3 0.3

3.6 Interaction with Adjacent Structure

The midbody web frame is part of an integrated structural Influence of

system comprising the inner and outer shells, the transverse unmade/led
frames and longitudinal girders. However, for the reasons structure Para. 2.2.3
discussed above, it is reasonable to isolate a single web
frame for analysis provided that the interaction with adjacent
structure is accounted for.
The primary interaction with adjacent structure (for the load
pattern of interest to this analysis) is through load transfer via
longitudinal structure. A reasonable approximation for this
configuration is to account for the support provided by the
longitudinal structure by using springs representing the
stiffness of this structure.
With reference to Figure 3.1, springs are required at the following
locations:
1. Centerline of Main Deck to account for the
deck centerline longitudinal girder
(vertically);
2. On Main Deck to account for the inboard side
girder (vertically);
3. On Main Deck to account for the outboard side
girder (vertical and horizontal components);
4. On side shell to account for the upper
stringer (horizontal);

B-13
 5. On side shell to account for the lower stringer at
the top of the turn of the bilge (horizontal);
6. Bottom structure to account for the girders (3 locations
- vertically);
7. Centerline of bottom structure to account for
the centerline girder (vertically); and
8. Bottom structure to account for the bottom
shell longitudinals (vertically).

Spring constants for the above items have been calculated as

the inverse of the deflection at the midspan of the longitudinal
member being evaluated (list above) due to a unit point load
placed at each of the points of intersection with a midbody web
frame along its length. The ends of the longitudinal member(s)
have been conservatively assumed as pinned. If a fixed end
condition had been assumed, the stiffness of the longitudinal
structure would have been overestimated resulting in a greater
load transfer from the midbody web frame than would be the
case in reality.

Spring constants calculated and used in the FE model are listed in

Section 84.4 Beam Section Properties.

3.7 Loads
The ice load2 is a function of vessel displacement, power of the Loads
vessel, the region of the ship, and the Arctic Class. Taking Para. 2.6
account of the various factors associated with ship parameters Para. 3.4
the total load applied to the web frame is 4435 kN. This is
applied as a uniform pressure of 1 meter width (which equals
the web frame spacing) and 2.85 meter height. This translates Influence of
to a pressure of Extent of
Model Para.
1.556 MPa. As required by the standard the pressure patch is
2.2.1
positioned such that 10% of its height is above the waterline.

The load applied is illustrated in Figure 3.3.

2
The ice loads are adapted from Ref. 1. The structural design philosophy of this standard is based on plastic
design. Hence design loads calculated from this standard will, for a well-designed structure, result in extensive
yielding. For the purposes of this example FEA, which assumes linear elastic behavior, the load applied has been
arbitrarily halved to ensure the structure remains elastic.

B-14
B-15
3.8 Boundary Conditions

Symmetry is assumed about a vertical plane through the Boundary

longitudinal axis of the ship. Therefore, symmetry boundary Conditions Para. 2.6
conditions are applied to all nodes along the outer (longitudinal)
Para. 3.4
edges of the plates. This provides translational restraint along
the longitudinal axis of the vessel, and rotational restraint about
the other two axes.
Symmetrical boundary conditions are applied to the bottom
structure and the deck structure intersecting the vertical plane
through the longitudinal axis of the ship. In addition, the bottom
shell plating along the centerline is fixed in the vertical
translation to avoid rigid body motion
4.0 FINITE ELEMENT MODEL
4.1 General Information
SI units were used throughout the finite element model.
Therefore, the units of length, area, moment of inertia, Units
Young's Modulus, and pressure were mm, mm2 , mm4 , MPa, Para. 2.1.7
and MPa, respectively. Global axes
The global coordinate system for the problem is as follows: system Para.
2.1.8

Global X-axis : athwartshi

Global Y-axis : p vertical
Global 2 axis :
parallel to ship CL

4.2 Element Selection

The elastic shell element (SHELL63) of ANSYS was selected Element Types
and used for modeling the web frame, and stiffeners from the Para. 3.1
bottom stringer of the side shell structure at the top of the turn
of the bilge to the start of the sloped section on the outboard
edge of the main deck. The stiffeners in other areas were
modeled using 3-D elastic beam elements (BEAM44) of
ANSYS. The stiffness of longitudinal girders was modeled using
linear spring elements (COMBIN14).

The SHELL63 element is well suited for modeling linear behavior

of flat or warped, thin to moderately thick, shell structures.

B-16
The element has six degrees of freedom at each node:
translations in the nodal x, y, and z directions and rotations about
the element x, y, and z axes. The deformation shape is linear in
the two in-plane directions. The out-of-plane motion is predicted
using a mixed interpolation of tensorial components. The element
is defined by four corner nodes, four thicknesses, and the
orthotropic material properties (if required). A triangular-shaped
element may be formed by defining the same node numbers for
the third and fourth nodes. Pressure load may be applied as
surface loads on the element.
The stiffeners in the deck and bottom structure of the mid-body
section have been modeled using 3-D elastic offset beam
elements (BEAM44). BEAM44 is a uniaxial element with tension,
compression, torsion, and bending capabilities. This element also
has six degrees of freedom per node. The stiffeners in the side
structure diaphragms were modeled using shell elements
(SHELL63).
To simulate the overall stiffness of the rest of the structure, as
discussed in Section 2.4, the connection points of the frame to
other structures were modeled with linear springs (COMBIN14)
elements. COMBIN14 elements are uniaxial tension-compression
elements with up to three degrees of freedom at each node:
translations in the nodal x, y, and z directions. Two sets of
elements, one for springs in the horizontal direction and the other
for springs in the vertical direction, were defined.
4.3 Mesh Design

The response of the side shell structure is of primary interest, Mesh Design
particularly in the vicinity of the loading. Therefore, the frame
Para. 3.2
structure has been modeled with a fine mesh of shell elements in
the following areas:

1. side shell structure between the turn of the bilge and the
side shell upper stringer; and
2. outer edge of the deck structure between the side shell
upper stringer and the deck angled outboard girder.

The remainder of the frame has been modeled using a coarse mesh
of shell and beam elements. This ensures that the stiffness of this
part of the structure is reasonably modeled in an

B-17
economical manner.

The mesh, consisting of beam and shell elements, used for the
frame analysis is shown in Figure 4. 1. The mesh design is
consistent with the results expected from the finite element model,
that is, a fine mesh is provided in the regions where a high-stress
gradient is expected with a coarse mesh provided elsewhere. The
mesh is the densest around openings, which are sources of
stress concentrations. Since the primary interest is in establishing
overall adequacy of the structure, the mesh density adopted is
designed to yield stresses that are accurate for this purpose.
Based on preliminary analyses the mesh around these openings
should allow the prediction of peak stresses with an accuracy of
roughly ± 5%.

The finite element model contains 3758 elements, 3578 nodes,

and 18131 total active degrees of freedom.

4.4 Finite Element Attributes and Spring Constants

The attributes of the elements used in the model are listed in Table Stiffness and Mass
4.1 . The spring constants calculated based on the stiffness Properties
properties of the adjacent structure are listed in Table 4.2. Para. 2.4

To avoid ill-conditioning in the stiffness matrix, ANSYS prints a

warning if the ratio of largest to smallest stiffness value is greater
than 1.0e08. The largest stiffness in the stiffness matrix being 4.
179e + 11, the smallest stiffness allowed is 4179 N/mm.
Therefore, springs with stiffness less than 4179 N/mm were not
used. Because of their relatively low stiffness values, these springs
will have a negligible effect on the overall behavior of the web
frame.

B-18
B-19
 TABLE 4. 1: Finite Element Attributes

Element Thicknes lzz lyy

Item Description Type Met. Real s or x10" x103 TKZT1 TKYT1
No. & Type Cons. No mm" mm" mm mm
No. Area
&No. mm/mmi
1 Diaphragms / Web Plating She1143 EH36 101 16

2 Floors - Web Plating " " 102 26

3 Deck Transverses - Web 1500x12 " " 103 12

4 Deck Plating Shell43 EH36 104 14

5 Outer Shell Plating " EH50 105 36

6 Bottom Shell Plating " AH36 106 29

7 Deck Transverses - Flange Shell43 EH36 107 19

8 Inner Deck Plating " .. 108 14

9 Inner Shell Plating " " 109 16.5

10 Inner Shell Plating - Bilge " " 110 17

11 Tank Top Plating " " 111 13

12 Transverse Stiffeners - Diaphragms Shell43 EH36 112 16

13 Stringers " " 113 16

14 Transverse Stiffeners - Tank Top Beam44 AH36 114 5700 38.58 190.0 10 142.5

15 Girders - Tank Top Shell43 AH36 115 15

16 Deck Transverse Stiffeners Beam44 EH36 116 1575 2.95 14.47 5.25 75

17 Side Girders Shell43 EH36 117 14

18 Deck Plating (with openings) Shell43 EH36 118 9.34

19 Beam Elements for stiffeners at Beam44 EH36 119 6576 92.56 140.3 8 205.5

20 Beam Elements for the bilge and Beam44 EH36 120 6576 92.e6 140.e4 8 205.5

21 Vertical Springs - to account for Combin14 - see Table 4.2 for spring stiffness

22 Horizontal Springs - to account for Combin14 - see Table 4.2 for spring stiffness

B-20
 TABLE 4.2 Spring Stiffness Calculated Based on Stiffness of Adjacent Structure

Real Spring
Spring Element
Description Constant Stiffness
Direction Type
No. N/mm

Deck Centerline Girder Vertical 5 121 231

Inboard Side Girder Vertical 5 122 3785
Outboard Side Girder Vertical 5 123 3012
Outboard Side Girder Horizontal 6 124 56
Upper & Centre Stringer Horizontal 6 125 7151
Lower Stringers Horizontal 6 126 7151
Bottom Girder -
Vertical 5 127 6508
Outboard

Bottom Girders Vertical 5 128 5913

Bottom Centre Line
Vertical 5 129 3631
Girder
4.5 FE Model Loads and Boundary Conditions

General information on the applied load is provided in Section 3. 7. Loads and Boundary
The design ice load was applied as a pressure of 1.556 MPa. Conditions

Para. 2.6
The finite element model boundary conditions are as explained in
Section 3.8. Referring to the global coordinate system described in Para. 3.4
Section 4.1, all nodes with Z - co-ordinate of +500or -500mm have
symmetry boundary conditions along the Z-axis. This provides
translation restraints in the Z-axis and rotational restraints in the X
and Y axes. All nodes along the bottom center
line have symmetry boundary conditions along the X - axes, i.e.,
translations restrained in the X and rotations restrained in the Y &
Z axes. The nodes along the bottom centerline for the bottom
shell plating were also restrained in the Y direction. For the top
centerline, all nodes have symmetry boundary conditions along
the X-axis.

B-21
4.6 FE Model Checks

Before the finite element model was run, the following prerun
Finite Element Model
checks were performed on the FE model :
Checks
- consistent units Para. 3.0
coordinate system
- element attributes and real
constants boundary conditions and
loads

The following prerun checks were conducted using the graphical

user interface provided by ANSYS. ANSYS provides a listing of
requested information for specifically selected entities. Also,
symbols can be turned on/off to view various aspects, such as
boundary conditions, loads, element connectivity, etc., of the
model.

- nodal coordinates of extremities of model

- free edge plots to check for structural
discontinuities element shape; aspect ratio,
taper, skew, orientation shrink plots and element
edge plots to check element connectivity
- checks for property assignment to elements - using
color coding based on element type, material type,
physical property type, etc.
- element plot showing element coordinate system to
check for element orientation
- true scale 3D plot of beam elements to ensure
correct beam size, orientation, and offsets
- boundary conditions - using model plots with
boundary condition symbols
- pressure load magnitude and direction (using arrows)
The following prerun checks are built into ANSYS and are
performed during the data checking process. Warning or error
messages are issued when the model fails to pass the check. The
output from such a data check run was reviewed for warning
and/or error messages.

- nodes not connected to structure

- elements not connected to structure
- missing material properties
- missing physical properties

B-22
 - element aspect
- ratio element
- warping element
- skewness

4. 7 FE Solution Option and Procedures

The following solution options and procedures used were: Solution Options and
- New Analysis Procedures
- Static Analysis Para. 3.5
- No Stress Stiffening
- Small Deflections
- Store all results for all load steps
- Print all output to a listing file
5.0 ANALYSIS RESULTS
5.1 General Solution Checks
The following post-run checks were performed:
comparison with simple hand calculations to ensure that the
results are reasonable (these calculations are included as
Annex 8-3)

equilibrium between the applied load and the reactions

inspection of the displaced shape of the structure to ensure
that there were no discontinuities in the model

inspection of stress contours to ensure the adequacy of the

mesh used
General Solution
All error and warning messages output by the program were Checks
investigated and resolved.
Para. 4.1
The total applied load in the X direction is 4434.9 kN. No forces
are applied in the Y and Z directions. The summed reactions in
the X, Y, and Z directions are 4434.9 kN, 0 kN, and 0 kN,
respectively. Post-processing
5.2 Post Processing Methods Methods
The ANSYS graphical post-processor was extensively used to
review stress and displacement results. Listings were reviewed to

B-23
obtain specific magnitudes for various quantities. In all of the Para. 4.2
stress contour plots, nodal averaging was used. For the shell Para. 4.3
element used in the model, the nodal values are calculated by
Para. 4.4
extrapolating from the values at the integration points.
5.3 Structural Response
The deflected shape of the structure is shown in Figure 5.1, where
the displacements are scaled up by a factor of 20. The maximum
vertical displacement at the top centerline of the vessel is 1 24
mm. The maximum horizontal displacement is 51.08 mm and
occurred on the inner shell in the vicinity of the load application.
The out of plane displacement, which was relatively small at 1.96
FEA Results and
mm, occurred in the diaphragm between the side shell and the
Acceptance Criteria
opening, also in the area of load application.· This displacement
Para. 5.1
occurred between two stiffeners indicating a possible location for
shear buckling. This possibility should be checked using classical
methods.
The von Mises stress plot for the area of interest is shown in
Figure 5.2. The contours are arranged, such that color orange
indicated stresses past yield (355 MPa) in all areas except the
outer shell. Dark red shading is used to indicate stresses past
yield (500 MPa) in the outer shell. It is clear from the figure that at
the applied load the overall structure remains elastic, except for a
small area around the openings where the stresses are past-
yield. The maximum stress recorded here is 573 MPa.
Figure 5.3 shows contours of bending stress, Sy. The outer shell is in
compression with a maximum compressive stress of 307 MPa. The
inner shell has maximum tensile stress of 330 MPa. High bending
stresses, past yield stress, were again observed around openings.
Clearly the bending stresses in the outer and the inner shells are
below the yield stress.
A contour plot of shear stresses in the diaphragm is shown in
Figure 5.4. The maximum and minimum stresses recorded were
188 and 164 MPa, respectively. The yield stress in shear being
205 MPa, the structure remains elastic at the applied load. Figure
5.5 contains an enlarged view of shear stresses around the
opening. which is directly under the load. The stress
concentrations around the opening are clearly visible in the figure.

B-24
The smoothness of the contours suggests that the mesh density is
probably adequate for the purposes of this study.

6.0 CONCLUSIONS

The midbody framing section of the 50000 DWT tanker as Overall Assessment
designed and analyzed meets the acceptance criteria. At the Para. 5.5
applied load, the structure remains predominantly elastic except in
a very localized region around openings. The tendency towards
an out-of-plane displacement in the diaphragm, between two
stiffeners in the area of an opening, could result in instability at
higher loads. This needs further investigation.

7 .0 REFERENCES

1. PROPOSED EQUIVALENT STANDARDS FOR THE

CONSTRUCTION OF ARCTIC CLASS SHIPS; Arctic Ship
Safety, (AMNB) Canadian Coast Guard - Northern; Dated
March 1993.

2. Arctic Tanker Structural Requirement Evaluation MIDSHIP

SECTIONS, BOW SECTIONS AND REPAIR DRAWINGS; AA
Shipping Company Limited; Dated June 1994.

3. LLOYD'S REGISTER - RULES FOR THE MANUFACTURE,

TESTING, AND CERTIFICATION OF MATERIALS; Dated
January 1993

B-25
.

B-26
 0

3364

B-27
B-28
FIGURE 4.1 Finite Element Model of Web Frame

B-29
FIGURE 5.1 Deflected Shape of Web Frame

B-30
B-31
B-32
B-33
B-34
B-35
B-36
 \...._ , ,,·

B-37
B-38
B-39
B-40
 Annex B-2
Company and Personnel Qualifications

B-2 COMPANY AND PERSONNEL QUALIFICATIONS

B-2.1 Contractor Qualifications
BB Engineering (BBE) is an ISO 9001 compliant company with a firm commitment to quality. It is also
certified by the Association of Professional Engineers of Ontario. BBE's primary business is Ship
Design and Analysis. It has several qualified professional structural engineers and naval architects on
its staff.
BBE performs all its finite element analysis on either a Dec Station 5000, running on Ultrix operating
system, or on a 60 MHz, 486 PC. For the current analysis the Dec Station 5000 was used. The finite
element software used is called "ANSYS". ANSYS is a well-established finite element software with a
large user base. It has been successfully used by BBE in several of its ship structure finite element
analyses. ANSYS provides all the required features for the current task and hence deemed adequate.

B-2.2 Personnel Qualifications

Analyst
Mr. J. S. is the finite element analyst assigned to this task. He has a Ph.D. in Structural Engineering,
and is registered as a Professional Engineer in the province of Ontario. He has taken two courses in
finite element analysis at the graduate level and has eight years experience in using finite element
method as an analysis tool. JS has a total of five years experience in using ANSYS, out of which three
years are ship structure specific. Information on specific finite element analysis problems that JS has
worked on in the past is available on request.
Checker
Ms. J. B. is the project engineer for this project and holds the responsibility of checking the finite
element analysis. JB has a Masters' Degree in Structural Engineering and is registered as a
Professional Engineer in the province of Ontario. She has taken one graduate level course in finite
element analysis and has six years experience in finite element analysis. JB has gained ten years
experience in the design and analysis of ship structures and has supervised several finite element
analysis projects. JB has three years experience in using ANSYS. Information on projects that JB
has worked in the past is available on request.

B-41
 Annex B-3
FEA Results Verification
B-3 FEA RESULTS VERIFICATION

The FEA results were compared with hand calculations. Two Accuracy Assessment
analyses have been performed as follows: Para. 5.4

1. An elastic beam analysis of the frame with a span of 11000

millimeters, ends fixed, openings ignored, subjected to a
uniformly distributed load of length 2850 millimeters equal
to 3.112 MN/m (9.373*0.8*0.5*0.83), for a total load of 8869
kN.

The structure has a bending stress of 550 MPa at the top

support in the inner hull plating. Shear stresses in the
portion of structure above the load are 195 MPa.

This structure reached first yield (in bending) at a load of

approximately 5700 kN.

2. An elastic frame analysis of the structure was FE modeled,

except that the inner shell and bottom structure was
analyzed with a flange width equal to 40 times the plate
thickness and the frame was assumed to be fixed on
centerline at the deck and at the bottom. In this analysis
side sway of the frame was ignored. The bending moments
calculated were within a few percents of those found in the
first analysis.

By comparison, the FEA predicts first yield of the inner hull

plating at the top of the 11000 mm portion of the side shell
framing at a load of approximately 4835 kN. This
comparison suggests that the FEA results are broadly
consistent with the results from the approximate simplified
analyses.

B-42
 Annex B-4

Sample Completed
Assessment Methodology Forms

EVALUATION OF FINITE ELEMENT MODELS AND RESULTS

Project#: xxxx

Project Title: Finite Element Analysis of Arctic Tanker Web Frame

Project
Description: Linear. static analysis of web frame tp ensure the adequacy of frame

ice load

Contractor:
BB Engineering ltd.

Result of
Evaluation: Generally satisfactory. Final approval subject to the supply of data

on some details of the model

Evaluator: John Doe

B-43
Date: May 1995

B-44
B-45
 FINITE ELEMENT ANALYSIS ASSESSMENT PRELIMINARY CHECKS

Project No. xxxx I Project Title : FEA of Arctic Tanker Web Frame
Company Name: BB Engineering Ltd I Date: May 1995

Analyst: JS IChecker: JS

Documentation Requirements
Refer to
Finite Element Analysis Assessment Check Guideline Result Comments
Section
1.1.1 Has the following information been 3.1.1
provided in the FEA documentation?

a) Objectives and scope of the analysis. 

bl Analysis requirements and acceptance criteria. 
cl FEA software used. 
d) Description of physical problem. 
e) Description of engineering model. 
f) Type of analysis. 
g) System of units. 
h) Coordinate axis systems. 
i) Description of FEA model. 
j) Plots of full FEA model and local details.  Some detail missing*
k) Element types and degrees of freedom per node. 
I) Material properties. 
m) Element properties (stiffness & mass properties). 
n) FE loads and boundary conditions. 
o) Description and presentation of the FEA results. 
pl Assessment of accuracy of the FEA results. 
q) Conclusions of the analysis. 
r) List of references. 

Based on the above checks answer Question 1. 1 and enter result in Figure 1.O. Result
1.1 Is the level of documentation sufficient to perform an assessment of the
FEA?
Comments

*Request additional detail on stiffener/web connection

B-46
Job Specification Requirements

Refer To
Finite Element Assessment Check Guideline Result Comments
Section

1.2.1 Is the job specification identified and 3-1.2

referenced in the analysis documentation? 

1.2.2 Are the objectives and scope of the analysis 3-1.2

clearly stated and are they consistent with 
those of the job specification?
1.2.3 Are the analysis requirements clearly stated 3-1.2
and are they consistent with those of the 
job specification?
1.2.4 If certain requirements of the job specification 3-1.2
have not been addressed (such as certain
N/A
load cases), has adequate
justification .been given?
1.2.5 Are the design / acceptance criteria clearly 3-1.2
stated and are they consistent with those -of 
the job specification?
1.2.6 Is there reasonable justification for using 3-1.2
FEA for this problem? 

1.2.7 Has advantage been taken of any previous 3-1.2

experimental, analytical, or numerical works N/A
that are relevant to this problem?

Based on the above checks answer Question 1.2 and enter result in Figure 1.0. Result

1.2 Does the analysis address the job specification requirements? 

Comments

B-47
Finite Element Analysis Software Requirements

Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section

1.3.1 Is the FEA software on the list of approved 3-1.3
programs for ship structural analysis
applications?

If the answer to Check 1.3.1 is "Y", you may skip Checks 1.3.2 and 1.3.3.

1.3.2 Are the capabilities and limitations of the FEA 3-1.4
software used to perform the required analysis
stated in the analysis documentation?

1.3.3 Is evidence of this capability documented and 3-1.3
available for review (eg. verification manual,
results of ship structure FEA benchmark tests,
previous approved FEA of similar problems)?

1.3.4 Does the vendor of the FEA software have a

quality system to ensure that appropriate

standards are maintained in software
development and maintenance.

Based on the above checks answer Question 1.3 and enter result in Figure 1.0. Result

1.3 Is the FEA software qualified to perform the required analysis? 

Comments

B-48
1.4 Contractor I Personnel Qualification Requirements

Refer To
Finite Element Assessment Check Guideline Result Comments
Section

1.4.1 Do the contractor personnel have adequate 3-1.5

academic training and experience qualifications 
to perform finite element analysis?

1.4.2 Do the contractor personnel have adequate 3-1.5

engineering experience qualifications for 
performing ship structural design or analysis?

1.4.3 Do the contractor and contractor personnel 3-1.5

have adequate professional certification 
qualifications?

1.4.4 Does the contractor have a working system of 3-1.5 Not documented but
Quality Assurance (QA) procedures and checks X using well established
that are satisfactory for the requirement? software

1.4.5 Do the contractor personnel have adequate 3-1.5

experience with the FEA software used for the 
analysis?

Based on the above checks answer Question 1.4 and enter result in Figure 1.0. Result

1.4 Is the contractor adequately qualified for performing ship structure FEA? 

Comments

B-49
 FINITE ELEMENT ANALYSIS ASSESSMENT I ENGINEERING MODEL
CHECKS

Project No. xxxx I Project Title : FEA of Arctic Tanker Web Frame
Company Name:
BB Engineering Ltd I Date:
May 1995

Analyst: JS I Checker: JB

2.1 Analysis Type and Assumptions

Refer To
Finite Element Analysis Assessment Check Guideline Result Comments
Section
2.1.1 Does the engineering model employ enough 3-2.1
dimensions and freedoms to describe the 
structural behavior (eg. 1-0, 2-D, or 3-0)?

2.1.2 Does the engineering model address the 3-2.1

appropriate scale of response for the problem (eg. 
global, intermediate, or local response)?
2.1.3 Is the type of analysis appropriate for the type of 3-2.1
response and loading of interest (eg. linear, 
static, dynamic, buckling analysis)?
2.1.4 Does the engineering model address all the 3-2.1
required results parameters (eg. stress, 
displacement, frequency, buckling load)?
2.1.5 Are all assumptions affecting the choice of 3-2.1
engineering model and analysis type justified 
(watch for non-standard assumptions)?
2.1.6 Is the level of detail, accuracy or conservatism of 3-2.1 Appears marginal - may
the engineering model appropriate for the criticality require more data on
of the analysis and type of problem? results to complete
evaluation
2.1.7 Does the analysis employ a consistent set of 3-2.1
units? 

2.1.8 Does the analysis employ a consistent global 3-2.1

coordinate axis system? 

Based on the above checks answer Question 2. 1 and enter result in Figure 1.0. Result
2.1 Are the assumptions of the type of analysis and engineering model acceptable? 
Comments

See above

B-50
 Appendix C
Examples of Variations in FEA Modeling Practices and
Results

C-1
INTRODUCTION

The purpose of this Appendix is to illustrate the effect of varying certain FEA modeling
parameters on the results using typical ship structure example problems.

Three typical ship structure examples are used. The first example, presented in Section C1,
concerns the modeling of stiffened panels. Four different approaches for modeling stiffened
panels are considered and the results presented. In the second example, presented in Section
C2, the modeling of stress concentrations arising from openings in a deck structure is
considered. In the third example, presented in Section C3, variations in the approach to
modeling a truss type mast structure are illustrated. A brief introduction is provided for each
problem, followed by a pictorial overview of the FEA model and results. A brief discussion of the
results is provided at the end of each example.

It is not the intention of this Appendix to endorse any particular modeling method. Rather, it
represents an effort to illustrate various modeling practices and present the variations in results.
This should provide some insight into the consequences of adopting a particular modeling
approach. The choice of the appropriate method, for a given problem, depends on the
purpose and objectives of the FEA.

In all cases the ANSYS program was used. The following element types were used:

• four-node membrane shell elements

• four-node shell elements with bending capabilities
• eight-node shell elements with bending capabilities
• two-node 3-D beam elements
• two-node 3-D truss elements
• mass elements

In certain cases, converged solutions are referred to. These solutions result from very fine mesh
models which are known to have converged (by comparison with less fine mesh models).

C-2
C1.0 Stiffened Panel

The majority of the structural weight in conventional ship structures is stiffened panels that
comprise the shell, decks, bulkheads, and superstructure. The panels are stiffened with
structural sections that are usually spaced in a regular fashion. The appropriate modeling
approach for stiffened panels depends on both the scale of the response (i.e., local or global
response) and the main structural actions of interest. Two main structural actions typically
modeled are 1) bending action due to loading normal to the panel surface, and 2) membrane
action due to loading in the plane of the panel. The first part of this section deals with bending
action and hence focusses on stiffened plate subjected to transverse loading. Membrane
action in a stiffened plate as a result of in-plane loads is briefly examined in the second part.

C-3
C-4
C-5
C-6
C-7
C-8
C-9
C-10
C-11
C-12
C-13
C-14
C-15
C-16
C-17
C-18
C2.O Multiple Deck Openings
A deck with multiple openings is used as an example to illustrate the influence of mesh density
and the element type on deflection and stress results. The mesh density is gradually increased
from coarse to fine, Two types of elements, 4-node membrane elements and 8-node shell
elements were used. The example also illustrates the effect of varying element aspect ratio.
The results obtained from the various trials are tabulated and compared with the converged
solution.

C-19
C-20
C-21
C-22
C-23
C-24
C-25
C3.0 MAST
A major factor in the modeling of lattice masts is the modeling of the connection details,
Depending on the type of connection, the joints can be modeled with full rigidity at the joint,
or some or all members can be modeled as pinned (hinged) joints. A simple truss-type mast
structure is used to illustrate both these options. In the case of a rigid jointed structure, the
mesh density (i. e., the number of elements per member of the mast) was varied to investigate
the influence on the results. Both static and dynamic analyses were performed on all these
models.

C-26
C-27
C-28
C-29
C-30
C-31
C-32
C-33
C-34
C-35
C-36
C-37
 Appendix D
Ship Structure Benchmark Problems for Assessing
FEA Software

D-1
D-2
D-3
D-4
D-5
D-6
D-7
D-8
D-9
D-10
D-11
D-12
D-13
D-14
D-15
D-16
D-17
D-18
D-19
D-20
D-21
D-22
D-23
D-24
D-25
D-26
D-27
D-28
D-29
D-30
D-31
D-32
D-33
D-34
REFERENCES
[1] NAFEMS. NAFEMS - Quality System Supplement to BS EN ISO 9001 Relating to
Engineering Analysis in the Design and Integrity Demonstration of Engineered Products.
1999.
[2] Noor AK. Bibliography of Books and Monographs on Finite Element Technology. Appl
Mech Rev 1991;44:307–17. doi:10.1115/1.3119505.
[3] Morris AJ. A practical guide to reliable finite element modelling. John Wiley & Sons; 2008.
[4] Mac Donald BJ. Practical stress analysis with finite elements. Glasnevin Publishing;
2011.
[5] Brauer JR. What every engineer should know about finite element analysis. M. Dekker;
1993.
[6] Meyer C, American Society of Civil Engineers. Structural Division. Task Committee on
Finite Element Idealization. Finite element idealization for linear elastic, static, and
dynamic analysis of structures in engineering practice. The Society; 1987.
[7] NAFEMS. NAFEMS Guidelines to finite element practice 1984.
[8] Steele JM. Applied finite element modeling : practical problem solving for engineers. M.
Dekker; 1989.
[9] DNV GL. DNVGL-CG-0127 Finite element analysis. 2015.
[10] Altair. Altair OptiStruct 2019 Verification Problems. 2019.
[11] Altair. Altair HyperWorks 2019 Hardware Recommendations and Certifications, Version
1.5. 2019.
[12] BSI. BS 7910: Guide to methods for assessing the acceptability of flaws in metallic
structures. BSI; 2015.
[13] American Petroleum Institute. API 579-1 / ASME FFS-1. 3rd ed. 2016.
[14] ABS. Guidance Notes on Ship Vibration #147. 2018.
[15] Altair University. Practical Aspects of Finite Element Simulation - A Study Guide. 2019.
[16] Siemens. Superelement User’s Guide. 2019.
[17] Sönnerlind H. Applying and Interpreting Saint-Venant’s Principle | COMSOL Blog 2018.
https://www.comsol.com/blogs/applying-and-interpreting-saint-venants-principle/
(accessed June 4, 2019).
[18] Fung YC, Tong P. Classical and computational solid mechanics. World Scientific; 2001.
[19] Park G-J. Equivalent Static Loads Method for Non Linear Static Response Structural
Optimization. 2010.
[20] Haufe A, Feucht M, Neukamm F, Dubois P. Recent Enhancements to the GISSMO
Failure Model in LS-DYNA. 2011.

References-1
[21] Basaran M, Weichert D. An Extension of the GISSMO Damage Model Based on Lode
Angle Dependence. 2010.
[22] Callister WD, Rethwisch DG. Materials Science and Engineering: An Introduction. 9th ed.
2013. doi:10.1007/3-540-29970-X_12.
[23] ASTM. E8/E8M − 16a Standard Test Methods for Tension Testing of Metallic Materials 1
2018. doi:10.1520/E0008_E0008M-16A.
[24] Dieter GE, Bacon DJ. Mechanical metallurgy. McGraw-Hill; 1988.
[25] Bonet J, Wood RD. Nonlinear continuum mechanics for finite element analysis.
Cambridge: Cambridge University Press; 2008. doi:10.1017/CBO9780511755446.
[26] Ramberg W, Osgood WR. Technical Note No. 902: Description of stress–strain curves by
three parameters. Washington DC.: 1943.
[27] Lemaître J, Chaboche J-L. Mechanics of solid materials. Cambridge University Press;
1990.
[28] Hallquist JO. LS-DYNA ® Theory Manual. 2006.
[29] Lemaitre J. A Continuous Damage Mechanics Model for Ductile Fracture. J Eng Mater
Technol 1985;107:83. doi:10.1115/1.3225775.
[30] Lemaitre J. How to use damage mechanics. Nucl Eng Des 1984;80:233–45.
[31] Gurson AL. Continuum Theory of Ductile Rupture by Void Nucleation and Growth: Part
I—Yield Criteria and Flow Rules for Porous Ductile Media. J Eng Mater Technol
1977;99:2. doi:10.1115/1.3443401.
[32] Opera Magistris. Opera Magistris (Elements of Applied Mathematics). 2016.
[33] Conle FA. Correcting Elastic FEA Results for Local Plasticity 2019.
http://fde.uwaterloo.ca/Fde/Notches.new/neuber.html (accessed June 11, 2019).
[34] Neuber H. Theory of Stress Concentration for Shear-Strained Prismatical Bodies With
Arbitrary Nonlinear Stress-Strain Law. J Appl Mech 1961;28:544. doi:10.1115/1.3641780.
[35] Topper TH, Wetzel RM, Morrow J. Neuber’s Rule Applied to Fatigue of Notched
Specimens. 1967.
[36] Tipton SM. A Review of the Development and Use of Neuber’s Rule for Fatigue Analysis.
SAE Trans 1991;100:143–8. doi:10.2307/44581113.
[37] Chu CC. A three-dimensional model of anisotropic hardening in metals and its application
to the analysis of sheet metal formability. J Mech Phys Solids 1984;32:197–212.
doi:10.1016/0022-5096(84)90008-5.
[38] Conle A, Oxland T, Topper TH. Computer-Based Prediction of Cyclic Deformation and
Fatigue Behavior. Low Cycle Fatigue, 100 Barr Harbor Drive, PO Box C700, West
Conshohocken, PA 19428-2959: ASTM International; 1988, p. 1218-1218–9.
doi:10.1520/STP24548S.
[39] Forrest PG. Fatigue of metals. 1962.
[40] DNV GL AS. FE Model Library for Collision Analysis: Description and Results (2015-

References-2
 0984.Rev1). 2016.
[41] Zerbst U, Ainsworth RA, Beier HT, Pisarski H, Zhang ZL, Nikbin K, et al. Review on
fracture and crack propagation in weldments - A fracture mechanics perspective. Eng
Fract Mech 2014. doi:10.1016/j.engfracmech.2014.05.012.
[42] NAFEMS. Fatigue and Fracture Mechanics in FEA - Part-1. 2001.
[43] NEFEMS. Fatigue and Fracture Mechanics in FEA - Part-4 2001.
[44] Turnquist M, Smith I, Eng P. A Life-Cycle Approach to the Assessment of Pipeline Dents.
2016.
[45] Lee Y-L, Barkey ME, Kang H-T. Metal fatigue analysis handbook : practical problem-
solving techniques for computer-aided engineering. Butterworth-Heinemann; 2011.
[46] Hui CY, Ruina A. Why K? High order singularities and small scale yielding. Int J Fract
1995;72:97–120. doi:10.1007/BF00042823.
[47] Anderson TL. Fracture mechanics : fundamentals and applications. 4th ed. 2018.
[48] Irwin GR. Onset of fast crack propagation in high strength steel and aluminum alloys.
1956.
[49] Tada H, Paris PC, Irwin GR. The Stress Analysis of Cracks Handbook, Third Edition.
Three Park Avenue New York, NY 10016-5990: ASME; 2000. doi:10.1115/1.801535.
[50] ASME. ASME Boiler and Pressure Vessel Code Complete Set. 2019.
[51] Hutchinson JW. Singular behaviour at the end of a tensile crack in a hardening material. J
Mech Phys Solids 1968;16:13–31. doi:10.1016/0022-5096(68)90014-8.
[52] Rice JR, Rosengren GF. Plane strain deformation near a crack tip in a power-law
hardening material. J Mech Phys Solids 1968;16:1–12. doi:10.1016/0022-
5096(68)90013-6.
[53] ASTM. ASTM E1820 - 18ae1 Standard Test Method for Measurement of Fracture
Toughness. 2018.
[54] ASTM. ASTM E399 - 17 Standard Test Method for Linear-Elastic Plane-Strain Fracture
Toughness KIc of Metallic Materials. 2017.
[55] ASTM. ASTM E1290 - 08e1 Standard Test Method for Crack-Tip Opening Displacement
(CTOD) Fracture Toughness Measurement (Withdrawn 2013). 2013.
[56] BSI. BS 7448: Fracture Mechanics Toughness Tests Part 1, 2, 3 and 4. 2005.
[57] ISO. ISO 12737: Metallic materials: determination of plane-strain fracture toughness.
British Standards Institution; 2011.
[58] Williams ML. On the Stress Distribution at the Base of a Stationary Crack. J Appl Mech
1935;24:109–14.
[59] O’Dowd NP, Shih CF. Family of crack-tip fields characterized by a triaxiality parameter—
I. Structure of fields. J Mech Phys Solids 1991;39:989–1015. doi:10.1016/0022-
5096(91)90049-T.
[60] O’Dowd NP, Shih CF. Family of crack-tip fields characterized by a triaxiality parameter—

References-3
 II. Fracture applications. J Mech Phys Solids 1992;40:939–63. doi:10.1016/0022-
5096(92)90057-9.
[61] Yang S, Chao YJ, Sutton MA. Higher order asymptotic crack tip fields in a power-law
hardening material. Eng Fract Mech 1993;45:1–20. doi:10.1016/0013-7944(93)90002-A.
[62] Chao YJ. J-A 2 Characterization of Crack-Tip Fields: Extent of J-A 2 Dominance and Size
Requirements. Int J Fract 1998;89:285–307.
[63] EDF Energy. R6-Revision 4, Assessment of the integrity of the structures containing
defects, Amendment 11. Gloucester: 2015.
[64] ABS. Rules for Building and Classing Marine Vessels. Houston,TX: 2019.
[65] ABS. Rules for Building and Classing Steel Vessels. Houston, TX: 2019.
[66] Hughes OF, Paik JK, Béghin D. Ship structural analysis and design. Society of Naval
Architects and Marine Engineers; 2010.
[67] Cook RD, Malkus DS, Plesha ME. Concepts and applications of finite element analysis.
1989.
[68] Jayadevan KR, Thaulow C, Østby E, Berg E, Skavllerud B, Holthe K, et al. Structural
integrity of pipelines: T‐ stress by line‐ spring. Fatigue Fract Eng Mater Struct
2005;28:467–88. doi:10.1111/j.1460-2695.2005.00889.x.
[69] MAESTRO. MAESTRO Version 11.7. Manual. 2018.
[70] NAPA. Structural Design 2019. https://www.napa.fi/software-and-services/ship-
design/structural-design/ (accessed September 11, 2019).
[71] Lee Y-L, Barkey ME, Kang H-T. Metal Fatigue Analysis Handbook: Practical Problem-
solving Techniques for Computer-aided Engineering. vol. 0. 2011.
[72] NAVSEA. Handbook for The Design and Assessment of Naval Surface Ship Structure,
MIL-HDBK-519. 2009.
[73] Michaelson RW. User’s Guide for SPECTRA: Version 8.3 (NSWCCD-65-TR-2000/07)
2000.
[74] Sikora JP, Michaelson RW, Ayyub BM. Assessment of Cumulative Lifetime Seaway
Loads for Ships. Nav Eng J 2002;114:167–80. doi:10.1111/j.1559-3584.2002.tb00129.x.
[75] Sikora JP. Cumulative lifetime loadings for naval ships 1998;56:299–305.
[76] Ma M, Zhao C, Hughes O. A practical method to apply hull girder sectional loads to full-
ship 3D finite-element models using quadratic programming. Ships Offshore Struct
2014;9:257–65. doi:10.1080/17445302.2013.770725.
[77] SINTEF. MARINTEK, VERES Version 3.01.0096 2012.
https://www.sintef.no/en/software/shipx/ (accessed September 11, 2019).
[78] WAMIT. Wamit, Inc. - The State of the Art in Wave Interaction Analysis. 2019.
[79] Bureau Veritas. HydroStar 2019. https://marine-offshore.bureauveritas.com/software
(accessed September 11, 2019).
[80] Daalen E van, Sireta F. PRECAL_R User Manual 2014.

References-4
[81] ABS. ‘Dynamic Loading Approach" for Floating Production, Storage and Offloading
(FPSO) Installations. 2017.
[82] Malenica S, Stumpf E, Sireta F-X, Chen X-B. Consistent Hydro-Structure Interface for
Evaluation of Global Structural Responses in Linear Seakeeping. Vol. 6 Nick Newman
Symp. Mar. Hydrodyn. Yoshida Maeda Spec. Symp. Ocean Sp. Util. Spec. Symp.
Offshore Renew. Energy, ASME; 2008, p. 61–8. doi:10.1115/OMAE2008-57077.
[83] Ma M, Zhao C, Danese N. A Method of Applying Linear Seakeeping Panel Pressure to
Full Ship Structural Models, Conference on Computer Applications and Information
Technology in the Maritime Industries, Leige, Belgium, 2012.
[84] Ochi MK, Cambridge University Press. Ocean Waves : the Stochastic Approach.
Cambridge University Press; 1998.
[85] Paik JK, Thayamballi AK. Ultimate limit state design of steel plated structures. J. Wiley;
2003.
[86] ABS. High-Speed Naval Craft 2017.
[87] Gerrit Visser. Frequency Response Analysis - What is It? 2018.
[88] Joseph Servidio RA, Quinn Director JP, Gascon J, Baetsen K, Lantz J, Curry A, et al.
SSC-470 Finite Element Modeling Methods: Vibration Analysis for Ships. 2015.
[89] Cook RD (Robert D. Finite element modeling for stress analysis. Wiley; 1995.
[90] Thompson MK, Thompson JM. ANSYS Mechanical APDL for Finite Element Analysis. 1st
ed. Butterworth-Heinemann; 2017.
[91] ADINA R&D. ADINA Handbook. 2017.
[92] Kuna M (Meinhard). Finite elements in fracture mechanics : Theory--Numerics--
Applications. Springer; 2013.
[93] Kardestuncer H, Norrie DH, Brezzi F (Franco). Finite element handbook. McGraw-Hill;
1987.
[94] ADINA R&D. Theory and Modeling Guide Volume II: Thermal. vol. II. 2010.
[95] Siemens. FEA - Powering today’s most advanced engineering analysis environment.
2019.
[96] Dassault Systems. ABAQUS Unified FEA Simulation Realistic Performance with
Advanced Multiphysics Solutins. 2019.
[97] BETA Simulation Solution. ANSA - The Advanced CAE Pr-Processing Software for
Complete Model Build Up. 2019.
[98] Altair University. Introduction to Nonlinear Finite Element Analysis using Optistruct. 2018.
[99] Kuna M. Finite Elements in Fracture Mechanics. vol. 201. 2013. doi:10.1007/978-94-007-
6680-8.
[100] Petti JP, Dodds RH. Ductile tearing and discrete void effects on cleavage fracture under
small-scale yielding conditions. Int J Solids Struct 2005.
doi:10.1016/j.ijsolstr.2004.11.015.

References-5
[101] Healy B, Gullerud A, Koppenhoefer K, Roy A, Roychowdhury S, Walters M, et al.
WARP3D-Release 17.9.3 3-D Dynamic Nonlinear Fracture Analyses of Solids Using
Parallel Computers. 2018.
[102] Koppenhoefer KC, Dodds RH. Ductile crack growth in pre-cracked CVN specimens:
Numerical studies. Nucl Eng Des 1998. doi:10.1016/S0029-5493(97)00218-5.
[103] Neukamm F, Feucht M, Haufe A. Considering damage history in crashworthiness
simulations. 2009.
[104] Connor J a. WG. Computer-Aided Teaching of Finite Element Displacement Method (MIT
Report 69). 1969.
[105] Bala S, Day J. General Guidelines for Crash Analysis in LS-DYNA ® Modeling Guidelines
for Crash Analysis. 2003.
[106] Det Norske Veritas. DNV-RP-C208 Determination of Structural Capacity by Non-linear FE
analysis Methods. 2013.
[107] As DG. DNVGL-RP-C204 Design against accidental loads. 2017.
[108] ADINA R&D. Theory and Modeling Guide Volume IV: EM. vol. IV. 2005.
[109] Wang E, De Waele W, Hertelé S. A complementary η_pl approach in J and CTOD
estimations for clamped SENT specimens. Eng Fract Mech 2015.
[110] BSI. BS 7608: Guide to fatigue design and assessment of steel products. BSI; 2015.
[111] Hadley I, Horn A. Treatment of constraint in BS 7910:2013, ISO 27306 and DNVGL-RP-
F108. Int J Press Vessel Pip 2019;169:77–93. doi:10.1016/J.IJPVP.2018.11.015.
[112] ASME. ASME B31G - 2012 Manual for Determining the Remaining Strength of Corroded
Pipelines. ASME International; 2017.
[113] American Petroleum Institute. API RP 1176: Recommended Practice for Assessment and
Management of Cracking in Pipelines. 1st ed. API; 2016.
[114] American Petroleum Institute. API 579-2 / ASME FFS-2. 1st ed. API; 2009.
[115] Zerbst U, Ainsworth RA, Madia M. Reference load versus limit load in engineering flaw
assessment: A proposal for a hybrid analysis option. Eng Fract Mech 2012;91:62–72.
doi:10.1016/J.ENGFRACMECH.2011.10.018.
[116] Pyttel B, Varfolomeyev I, Luke M, Berger C, Siegele D. FKM guideline “fracture
mechanics proof of strength for engineering components” - overview and extension
topics. Weld World 2007. doi:10.1007/BF03266576.
[117] Wang E, De Waele W, Hertelé S. A complementary ηpl approach in J and CTOD
estimations for clamped SENT specimens. Eng Fract Mech 2015;147.
doi:10.1016/j.engfracmech.2015.07.043.
[118] Brocks W, Cornec A, Scheider I. Computational Aspects of Nonlinear Fracture
Mechanics. 2002.
[119] Brocks W. Elastic-plastic fracture mechanics. Solid Mech Its Appl 2018:49–84.
doi:10.1007/978-3-319-62752-6_5.

References-6
[120] Siegele D, Schmitt W. Determination and simulation of stable crack growth in ADINA.
Comput Struct 1983;17:697–703. doi:10.1016/0045-7949(83)90083-4.
[121] Brocks W, Yuan H. Numerical investigations on the significance of J for large stable crack
growth. Eng Fract Mech 1989;32:459–68. doi:10.1016/0013-7944(89)90317-2.
[122] Brocks W, Eberle A, Fricke S, Veith H. Large stable crack growth in fracture mechanics
specimens. Nucl Eng Des 1994;151:387–400. doi:10.1016/0029-5493(94)90183-X.
[123] Jayadevan KR, Berg E, Thaulow C, Østby E, Skallerud B. Numerical investigation of
ductile tearing in surface cracked pipes using line-springs. Int J Solids Struct
2006;43:2378–97. doi:10.1016/J.IJSOLSTR.2005.06.098.
[124] Berg E, Skallerud B, Thaulow C. Two-parameter fracture mechanics and circumferential
crack growth in surface cracked pipelines using line-spring elements. Eng Fract Mech
2008;75:17–30. doi:10.1016/J.ENGFRACMECH.2007.03.023.
[125] Needleman A. An analysis of decohesion along an imperfect interface. Int J Fract
1990;42:21–40. doi:10.1007/BF00018611.
[126] Needleman A. Some Issues in Cohesive Surface Modeling. Procedia IUTAM
2014;10:221–46. doi:10.1016/J.PIUTAM.2014.01.020.
[127] Yuan H, Lin G, Cornec A. Verification of a Cohesive Zone Model for Ductile Fracture. J
Eng Mater Technol 1996;118:192. doi:10.1115/1.2804886.
[128] ADINA R&D. Theory and Modeling Guide Volume I : ADINA Solids & Structures. vol. I.
2017.
[129] Hao S, Brocks W. The Gurson-Tvergaard-Needleman-model for rate and temperature-
dependent materials with isotropic and kinematic hardening. Comput Mech 1997;20:34–
40. doi:10.1007/s004660050213.
[130] Sun X, Wimmer SA, Karrt DG. Shear Band Initiation of Brittle Damage Materials. Int J
Damage Mech 1996;5:403–21. doi:10.1177/105678959600500404.
[131] Tvergaard V. On localization in ductile materials containing spherical voids. Int J Fract
1982;18:237–52. doi:10.1007/bf00015686.
[132] As DG. CLASS GUIDELINE Fatigue assessment of ship structures. 2018.
[133] DNV GL. Recommended Practice DNV GL AS - Fatigue design of offshore steel
structures. 2016.
[134] DNV GL. Plus-extended fatigue analysis of ship details. 2016.
[135] Basu RI, Kirkhope KJ, Srinivasan J. SSC-387 Guideline for Evaluation of Finite Element
and Results. 1996.
[136] United States Naval Academy. EN358 :: Naval Architecture &amp; Ocean Engineering ::
USNA. Annapolis: 2009.
[137] NSWCCD. Modeling Guidelines Full Ship FEM Development (Revision G). 2013.
[138] Avi E, Lillemäe I, Romanoff J, Niemelä A. Equivalent shell element for ship structural
design. Ships Offshore Struct 2015;10:239–55. doi:10.1080/17445302.2013.819689.

References-7
[139] SSC. FINITE ELEMENT MODELING METHODS: VIBRATION ANALYSIS FOR SHIPS.
2015.
[140] Altair University. Learn Dynamic Analysis with Altair Optistruct. 2018.
[141] Irons B, Ahmad S. Techniques of finite elements. E. Horwood; 1980.
[142] Steele JM. Applied finite element modeling : practical problem solving for engineers. M.
Dekker; 1989.
[143] Crisfield MA. Non-linear finite element analysis of solids and structures. Vol. 1,
Essentials. Wiley; 2000.
[144] Belytschko T, Liu WK, Moran B. Nonlinear finite elements for continua and structures.
Wiley; 2000.
[145] Bathe K-J. Finite element procedures in engineering analysis. Prentice-Hall; 1982.
[146] Bathe K-J, Bathe K-J. Finite element procedures. Prentice Hall; 1996.
[147] Bucalém ML, Bathe K-J. The mechanics of solids and structures : hierarchical modeling
and the finite element solution. Springer; 2011.
[148] Chapelle D (Dominique), Bathe K-J. The finite element analysis of shells. Fundamentals.
Springer; 2011.
[149] Bathe K-J. Finite element procedures. 1996.
[150] Holzapfel G. Nonlinear solid mechanics: a continuum approach for engineering science.
Meccanica 2002. doi:10.1023/A:1020843529530.
[151] De Souza Neto EA, Perić D, Owen DRJ. Computational Methods for Plasticity: Theory
and Applications. Comput Methods Plast Theory Appl 2008.
doi:10.1002/9780470694626.
[152] Implicit: Checklist — Welcome to the LS-DYNA support site 2019.
https://www.dynasupport.com/howtos/implicit/implicit-checklist (accessed June 11, 2019).
[153] Hilber HM, Hughes TJR, Taylor RL. Improved numerical dissipation for time integration
algorithms in structural dynamics. Earthq Eng Struct Dyn 1977;5:283–92.
doi:10.1002/eqe.4290050306.
[154] Paik JK. Practical techniques for finite element modeling to simulate structural
crashworthiness in ship collisions and grounding (Part I: Theory). Ships Offshore Struct
2007;2:69–80. doi:10.1533/saos.2006.0148.
[155] Paik JK. Practical techniques for finite element modelling to simulate structural
crashworthiness in ship collisions and grounding (Part II: Verification). Ships Offshore
Struct 2007;2:81–5. doi:10.1533/saos.2006.0149.
[156] Abambres M, Arruda MR. Finite element analysis of steel structures – a review of useful
guidelines. Int J Struct Integr 2016;7:490–515. doi:10.1108/IJSI-07-2015-0020.
[157] Amdahl J, Ehlers S, Leira BJ. Collision and grounding of ships and offshore structures :
proceedings of the 6th International Conference on Collision and Grounding of Ships and
Offshore Structures, ICCGS, Trondheim, Norway, 17-19 June 2013. CRC
Press/Balkema; 2013.

References-8
[158] Storheim M. Structural Response in Ship-Platform and Ship-Ice Collisions. NTNU; 2016.
[159] Benson DJ. The History of LS-DYNA®. 2007.
[160] ADINA R&D. ADINA Primer. 2017.
[161] Altair University. Introduction to Explicit Analysis unsing RADIOSS - A Study Guide.
2018.
[162] Hourglass — Welcome to LS-DYNA Examples 2019.
https://www.dynaexamples.com/introduction/process_simulation/hourglass (accessed
July 22, 2019).
[163] LSTC. Mass scaling. LSDYNA Support 2019.
https://www.dynasupport.com/howtos/general/mass-scaling (accessed July 23, 2019).
[164] LSTC. What is mass scaling? DYNA Support 2019.
https://www.dynasupport.com/faq/general/what-is-mass-scaling (accessed July 23,
2019).
[165] Bala S. Overview of Mass-Scaling in LS-DYNA. D3view 2006.
https://www.d3view.com/2006/10/overview-of-mass-scaling/ (accessed July 23, 2019).
[166] Altair University. Learn Fatigue Analysis with Altair OptiStruct. 2018.
[167] Goodman J. Mechanics Applied to Engineering. Longmans, Green & Company; 1899.
[168] Hertzberg RW, Vinci RP, Hertzberg JL. Deformation and Fracture Mechanics of
Engineering Materials. John Wiley & Sons, Inc; 2013. doi:10.1093/nq/s6-XII.297.185-c.
[169] Wang E. A study on CMOD-based compliance calibration functions for clamped single-
edge tension specimens. J Test Eval 2015;43. doi:10.1520/JTE20130333.
[170] Wang E, Zhou W, Shen G, Duan D. An experimental study on J(CTOD)-R curves of
single edge tension specimens for X80 steel. Proc. Bienn. Int. Pipeline Conf. IPC, vol. 4,
2012. doi:10.1115/IPC2012-90323.
[171] Xue Thinkviewer L, Xue L, Wierzbicki T. Verification of a New Fracture Criterion Using
LS-DYNA Sandia fracture challenge View project Ductile fracture modeling and
formulation of damage plasticity theory View project Verification of a New Fracture
Criterion Using LS-DYNA. 2006.
[172] SSC. SSC-472-Survivability of Hull Girder in Damaged Condition. 2015.
[173] SSC. SSC-459-Reliability-based Performance Assessment of Damaged Ships. 2011.
[174] SSC. SSC-409-Guide to damage toelrance analysis of marine structures. 2000.
[175] SSC. SSC-386-Ship Maintenance Project - Vol. 2 Corrosion Damage Evaluation. 1995.
[176] SSC. SSC-386-Ship Maintenace Project - Vol.1 Fatigue Damage Evaluation. 1995.
[177] Poudret J, Huther M, Jean P, Transport G, Vauhan H. SSC-81symp10-Grounding of a
Membrane Tanker; Correlation Between Damage Predictions and Observations. 1981.
[178] Ainsworth RA, Hooton DG. R6 and R5 procedures: The way forward. Int J Press Vessel
Pip 2008;85:175–82. doi:10.1016/J.IJPVP.2007.10.003.

References-9
[179] Kanninen MF, Broek D, Marschall CW, Rybicki E. Mechanical fracture predictions for
sensitized stainless steel piping with circumferential cracks [NP-192]. 1976.
[180] CSA. CSA Z662-15. 6th ed. 2015.
[181] CEPA. CEPA Recommended Practices for Managing Near-neutral pH Stress Corrosion
Cracking 3 rd edition. 2015.
[182] Hogström P, Ringsberg JW, Johnson E. Survivability analysis of a struck ship with
damage opening – influence from model and material properties uncertainties. Ships
Offshore Struct 2011;6:339–54. doi:10.1080/17445302.2010.500052.
[183] Hogström P, Ringsberg JW. An extensive study of a ship’s survivability after collision – A
parameter study of material characteristics, non-linear FEA and damage stability
analyses. Mar Struct 2012;27:1–28. doi:10.1016/J.MARSTRUC.2012.03.001.
[184] Storheim M, Notaro G, Johansen A, Amdahl J. Comparison of ABAQUS and LS-DYNA in
simulations of ship collisions. Iccgs-2016 2016:239–46.
[185] Storheim M, Amdahl J. Design of offshore structures against accidental ship collisions.
Mar Struct 2014;37:135–72. doi:10.1016/J.MARSTRUC.2014.03.002.
[186] Storheim M, Amdahl J. On the sensitivity to work hardening and strain-rate effects in
nonlinear FEM analysis of ship collisions. Ships Offshore Struct 2017.
doi:10.1080/17445302.2015.1115181.
[187] Cowper GR, Symonds PS. Strain-Hardening and Strain-Rate Effects in the Impact
Loading of Cantilever Beams. 1957. doi:10.21236/AD0144762.
[188] Johnson GR, Cook WH. Fracture characteristics of three metals subjected to various
strains, strain rates, temperatures and pressures. Eng Fract Mech 1985;21:31–48.
doi:10.1016/0013-7944(85)90052-9.
[189] Storheim M, Alsos HS, Hopperstad OS, Amdahl J. A damage-based failure model for
coarsely meshed shell structures. Int J Impact Eng 2015;83:59–75.
doi:10.1016/J.IJIMPENG.2015.04.009.
[190] Storheim M, Amdahl J. Non-Conservative Consequences of “Conservative” Assumptions
in Ship-Platform Collision Analysis. Vol. 4B Struct. Saf. Reliab., ASME; 2014, p.
V04BT02A034. doi:10.1115/OMAE2014-24457.
[191] MARPOL73-78: Brief history - list of amendments to date and where to find them.
MARPOL73-78 Br Hist - List Amend to Date Where to Find Them 2012.
[192] BMT. FlawCheck Structural Integrity Assessment Tool. Ottawa, Canada: 2019.

References-10