
Prerequisites for a Generic Rear Seat Test Rig
Master’s thesis in Applied Mechanics

KARL-JOHAN LARSSON
IDA FRANSÉN

Department of Applied Mechanics
CHALMERS UNIVERSITY OF TECHNOLOGY
Göteborg, Sweden 2015


MASTER’S THESIS IN APPLIED MECHANICS

Prerequisites for a Generic Rear Seat Test Rig

KARL-JOHAN LARSSON
IDA FRANSÉN

Department of Applied Mechanics
Division of Material and Computational Mechanics
CHALMERS UNIVERSITY OF TECHNOLOGY

Göteborg, Sweden 2015


Prerequisites for a Generic Rear Seat Test Rig
KARL-JOHAN LARSSON
IDA FRANSÉN

c© KARL-JOHAN LARSSON, IDA FRANSÉN, 2015

Master’s thesis 2015:37
ISSN 1652-8557
Department of Applied Mechanics
Division of Material and Computational Mechanics
Chalmers University of Technology
SE-412 96 Göteborg
Sweden
Telephone: +46 (0)31-772 1000

Cover:
An overview of the designed generic rear seat test rig concept

Chalmers Reproservice
Göteborg, Sweden 2015


Prerequisites for a Generic Rear Seat Test Rig
Master’s thesis in Applied Mechanics
KARL-JOHAN LARSSON
IDA FRANSÉN
Department of Applied Mechanics
Division of Material and Computational Mechanics
Chalmers University of Technology

Abstract

In the early stages of a car development project, limited design information is available regarding the safety
components and their set-up in the rear seat. If a generic rear seat test rig is developed, which allows for
physical testing of different solutions in the early stages of a project, information can be gathered as to how the
rear seat environment should be designed to assure a safe environment in the event of a crash. Furthermore,
if a virtual analysis model of the same test rig is developed, this virtual model can be used to find preferred
solutions before physical testing is performed. In this thesis the frontal crash behaviour of a currently used
car platform is studied in two virtual crash tests, in order to determine which components that needs to be
represented in a generic rig. Based on the discovered behaviour in these models and investigation of other
related parameters, together with given requirements from engineers at the Volvo Cars Safety Center, a concept
design for the generic rear seat test rig is produced, along with a simplified virtual crash test model containing
the functional parts of the designed concept. In the concept design, the compliance as well as the positioning of
different parts are adjustable, and deformations occurring in a real car crash is captured by small replaceable
deforming plates. The positioning and size of the deforming elements in the virtual model of the concept
rig are tuned using system identification to reproduce the behaviour of the two studied virtual model crash
tests. Results show that the virtual model of the concept rig has capability to reproduce the results with
good accuracy, especially the crash test with smaller forces present during the crash. In the other case, where
larger forces are present, several more components in the real car deform in an interdependent manner, and
the results from the simplified rig model deviates slightly. However, the results produced by the rig model in
this case follows the existing trends in the reference model and the rig model should be sufficient to evaluate
the effects of design changes, in a trend based manner. Furthermore, an existing preprocessing tool for the
virtual crash test model seatbelts is implemented, which significantly reduces the preprocessing time required
to test different seatbelt configurations. Using this tool, a study on the influence of the seatbelt anchoring
points positioning on crash test dummy data is performed and presented.

Keywords: Crash testing, sled testing, test rig, virtual crash testing

i


Sammanfattning

I ett tidigt skede av ett bilutvecklingsprojekt är mängden tillgänglig information begränsad gällande desig-
nen av säkerhetskomponenter i baksätesmiljön. Om en generisk testrigg för baksätesmiljön utvecklas, som
möjliggör fysiska tester av olika lösningar i ett tidigt skede av projektet, kan information erh̊allas gällande hur
baksätesmiljön bör utformas för att säkerställa en säker miljö i händelse av en krock. Dessutom, om en virtuell
analysmodell av samma testrigg utvecklas, kan denna modell användas för att finna fördelaktiga lösningar innan
fysiska tester utförs. I det här arbetet studeras frontalkrockbeteende i tv̊a olika modeller av en, för närvarande,
använd bilplattform, för att avgöra vilka komponenter som behöver vara representerade i en generisk testrigg.
Baserat p̊a studierna av hur dessa modeller beter sig samt utvärderingar av andra relaterade parametrar,
tillsammans med givna krav fr̊an ingenjörer p̊a Volvo Cars Safety Center, producerades ett koncept för den
generiska testriggen tillsammans med en förenklad virtuell krocktestmodell best̊aende av de funktionella delarna
av framtagna konceptet. I den konceptuella riggen finns det möjlighet att justera styvhet samt positionering
av olika komponenter, och deformationer som förekommer i baksätesmiljön vid en verklig krock f̊angas av
sm̊a utbytbara samt deformerbara plattor. Positioneringen och storleken p̊a de deformerbara elementen i den
virtuella modellen av den konceptuella riggen ställs in med hjälp av systemidentifiering för att återskapa
beteendet för de tv̊a virtuella krocktestmodellerna som studerats. Resultaten visar att den virtuella modellen
av riggkonceptet har kapacitet att återskapa resultaten med god noggrannhet, särskilt krocktestet med lägre
krafter under krocken. I det andra fallet, där högre krafter förekommer, deformeras fler komponenter i den
verkliga bilen, beroende av varandra, och resultaten fr̊an riggmodellen avviker n̊agot. Däremot följer resultaten
fr̊an riggmodellen i det här fallet de befintliga trenderna i referensmodellen och riggmodellen bör vara tillräcklig
för att utvärdera effekter av konstruktionsändringar p̊a ett trendbaserat sätt. Vidare implementerades ett
befintligt preprocessorverktyg för säkerhetsbälten i den virtuella krocktestmodellen, vilket reducerar förarbetet
avsevärt vid test av olika bälteskonfigurationer. Med hjälp av det här verktyget utfördes en studie av hur olika
positionering av bältesinfästningspunkterna p̊averkar krocktestdockan.

ii


Preface

The thesis work has been performed at the Volvo Cars Safety Center (VCSC), rear interior safety group, at
Volvo Car Corporation, Torslanda Göteborg. We, the authors, have developed a concept design for a generic
rear seat test rig based on results obtained when studying and modifying virtual crash test models developed
by the staff at VCSC, as well as given requirements by the staff. Furthermore a virtual model of the concept
rig has been implemented, as well as studies of parameters that has been pointed out as interesting by our
supervisors. Since the state of the art crash analysis virtual models used at VCSC are immensely complicated,
this work could not have been completed without the help of the staff, who readily answered any question we
had.

Special thanks goes out to supervisors Mathias Retzlaff and Magnus Björklund, as well as examiner Martin
Fagerström, who have shown great support and interest in our work.

iii


iv


Contents

Abstract i

Sammanfattning ii

Preface iii

Contents v

1 Introduction 1

1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.3 Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Frame of reference 3

2.1 Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.2 Crash test dummies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.2.1 Dummy instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.3 Seatbelts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3.1 Three point seatbelts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3.2 Risks associated with seatbelts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3.3 Common safety features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.4 Safety rating organisations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.4.1 Crash tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.5 Sled testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2.6 CAE crash testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.6.1 Features of a car crash simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.6.2 LS-DYNA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.6.3 Ansa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 The generic rig concept 15

3.1 Given requirements and design guide-lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2 Finding requirements through model studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.2.1 Simplified models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3.2.2 Rear seat backrest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.2.3 Seat cushion attachment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.3 How to capture necessary behaviour with the rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.3.1 Foot stop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.3.2 Tilting seat bottom plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.3.3 Summary of found requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.4 The generic rig concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.4.1 Generic seat bottom structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.4.2 Seatbelt fixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.4.3 Assumptions made for the rig concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.5 Results from the rig FE model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.5.1 Set-up of rig FE model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3.5.2 P4 rig results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

3.5.3 P5 rig results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

3.5.4 Rig results for offset deformable barrier crash test . . . . . . . . . . . . . . . . . . . . . . . . . . 42

3.6 Fulfilment of the requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

v


4 Implementation of the modelling concepts 45
4.1 Ansa seatbelt module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.1 LS-DYNA seatbelt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.2 Setting up a dynamic Ansa seatbelt entity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.1.3 Compatibility issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1.4 Example model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2 Automatic parameter configuration using LS-OPT coupled with Ansa and LS-DYNA . . . . . . . 48
4.2.1 Ansa and optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.2 LS-DYNA input files and optimisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.3 Setting up an LS-OPT optimisation task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2.4 Example optimisation task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5 Parameter study 53
5.1 Seatbelt mounting positions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Dummy friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3 Stiffness of seat bottom plate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6 Concluding discussion 67
6.1 Fulfilment of goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.1.1 Describe components and phenomena affecting the rear seat safety . . . . . . . . . . . . . . . . . 67
6.1.2 Develop a conceptual design for a generic rear seat test rig . . . . . . . . . . . . . . . . . . . . . 67
6.1.3 Develop a CAE model of the rig that can be used to evaluate different safety solutions . . . . . . 68
6.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.2.1 Describe components and phenomena affecting the rear seat safety . . . . . . . . . . . . . . . . . 68
6.2.2 Develop a conceptual design for a generic rear seat test rig . . . . . . . . . . . . . . . . . . . . . 68
6.2.3 Develop a CAE model of the rig that can be used to evaluate different safety solutions . . . . . . 69

References 69

Appendix A NCAP crash tests 70
A.1 Euro NCAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
A.1.1 Euro NCAP full width frontal impact test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
A.1.2 Injury criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
A.2 China NCAP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
A.2.1 China NCAP tests and scoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

Appendix References 73

Appendix B Simplified model 74

Appendix C Ansa scripts 76
C.1 Reverse order of slipring sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
C.2 Move edgeset script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
C.3 Update sets of DBCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Appendix D Auto-recreating seatbelt model 79
D.1 Tutorial: Seatbelt example model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
D.2 Set up of example model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

Appendix E Start LS-DYNA from LS-OPT 85
E.1 LS-DYNA start script . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
E.2 Setting up LS-OPT solver stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

vi


1 Introduction

This thesis concerns car crash safety evaluation and is performed at the Volvo Cars Safety Centre (VCSC) at
Volvo Car Corporation (VCC). At VCSC, crash safety evaluation is performed both with computer simulated
crashes using CAE (Computer Aided Engineering) models, as well as physical crash tests. In the physical tests
both full scale crash tests, where a complete car is crashed in a test scenario, and sled tests are used. In the
sled tests, a rig representing part of the car that is being investigated is built according to the test case onto a
sled that is accelerated in a crash simulator. In the crash simulators, the rig is stationary at the starting point,
and accelerated such that the test corresponds to a reversed crash.

The main intention behind the work presented in this thesis is to develop a generic test rig for the rear
seat passenger safety evaluation. In this work, three different parts can be identified; the development of a
conceptual generic test rig implemented as a simplified CAE model, the implementation of existing software
tools that can be used for future work with the generic rig model and, in the third part, a number of parameter
studies relevant to rear seat safety evaluation is presented.

1.1 Background

The current designs of the rear seat test rigs used at VCSC are based on prototypes or the final version of
the car that is being evaluated, and does not easily allow for changes in the test set-up, such as changing
components and their positioning. In early concept phases there is a limited availability of CAD (Computer
Aided Design) data to build up representative CAE models since the final interior is still unknown.

This makes it hard to evaluate the influence of a certain component and its position. A generic crash test
rig would make it possible to perform crash tests in early design phases, and thus evaluate different design
options, and even contribute to design guidelines and requirements for the future car models.

For VCSC, this rig is also interesting in other aspects, especially from a correlation point of view. The
current CAE models used consist of several different components, obtained from both internal and external
sources such as different design departments, seat and seatbelt manufacturers. In full scale crash test models
there are many factors interacting, which make it very difficult to validate that a certain component behaves
accurately.

If a CAE generic test rig is developed, that is well correlated to the physical generic crash test rig, it would
be possible to design test set-ups to isolate component behaviour and investigate the validity of these models.
In that way, the test rig can be used both as a tool for CAE model validation and for decision making in the
early design phase.

1.2 Purpose

This thesis work is the first step towards the generic rear seat test rig. Based on phenomena existing in a crash,
a virtual generic test rig shall be developed, which shall be realisable as a physical test rig.

1.3 Goals

In order to fulfil the purpose of the thesis work a number of goals have been formulated.

Describe components and phenomena affecting the rear seat safety
A car consists of many components and it is of interest for the development of a generic test rig to
investigate which components of these that have a significant impact on the rear seat safety, and also
what that impact is, in order to determine which simplifications can be made when constructing a test rig.

Develop a conceptual design for a generic rear seat test rig
Based on components and phenomena affecting the rear seat safety environment, a concept for a generic
test rig shall be developed. The design of the concept shall be such that the crash behaviour of the rig is
possible to change. It shall be possible to test different components and their relative positioning.

1


Develop a CAE model of the rig that can be used to evaluate different safety solutions
Along with the designed concept a CAE model that allows for easy evaluation of different test set-ups
before performing physical tests shall be produced.

1.4 Limitations

This thesis concerns adult occupant rear seat crash safety and the rig will only be designed with regard to
frontal impact crash testing, a limitation set by the thesis supervisors. The crash behaviour of the test rig will
be based on the behaviour observed from CAE models of a current Volvo platform carrying adult occupants in
frontal impact crash simulations, although care will be taken to ensure that the rig is adjustable to represent
other designs. These CAE models will be regarded as accurate and the accuracy of the rig CAE model will be
evaluated against these. Furthermore, this thesis focuses on delivering a conceptual design proposal, with focus
on the components affecting the rear seat safety, which means that detailed dimensioning and drawings of the
surrounding structures is out of scope. Only software available at VCSC will be used, and CAE models will be
constructed using existing Finite Element (FE) modelling guidelines regarding element types, mesh densities,
material models and contact definitions.

2


Figure 2.1: The definition of the global coordinate system in the car

2 Frame of reference

The crash safety of Volvo cars is evaluated by both VCC internally and by different safety rating organisations
around the world. Since safety is a core value at VCC when designing cars, their internal safety standards and
goals are very high. Since a car can crash in many ways and can carry differently sized occupants, there is no
general best set-up for the different safety components in a car, such as airbags and seatbelts, and every new
car design needs to be carefully investigated. Several physical crash tests, as well as thousands of computer
simulated crashes are performed to ensure that every car has a high level of safety in many types of crashes.

Furthermore, the safety rating achieved by different safety rating organisations are very important for the
Volvo brand and thus the crash testing cases evaluated here must also be carefully evaluated for each car by
VCSC during the design process. In Section 2.4 the different rating organisations and test cases pertinent to
this thesis is presented.

Section 2.1 briefly introduces nomenclature used, and in Section 2.2 the crash test dummies used when
evaluating occupant safety are described, followed by Section 2.3, where seatbelts and their terminology, safety
features as well as associated risks are described.

To evaluate crash safety during the design process at VCC, CAE testing with FE models are utilised,
but also physical testing. The physical testing is performed both as full scale testing, where complete cars
or prototypes are crashed in different cases and with sled tests, where part of the car environment under
investigation is built into a rig that is placed on a sled, which then is accelerated with a pulse corresponding to
a reversed crash. The sled testing procedure of interest in this thesis is treated in Section 2.5 and the general
features of the CAE virtual crash test models are introduced in Section 2.6.

2.1 Conventions

In order to uniformly refer to positions and directions in the car a global coordinate system is defined, see
Figure 2.1, where X-direction is defined from the front of the car to the rear, the Y-direction is from driver left
hand side to the right hand side and is zero in the centreline of the car and Z-direction points from the road
level and upwards. The directions of the axes is the same between different car models, but the position of the
Z,X origin may differ.

Each seat position in the car is associated with a unique number and the numbering convention can be seen
in Figure 2.2. This convention is used when referring to which seat position a particular crash test dummy is
seated at in a test. In this thesis, position numbers four and five are mainly referred to, using the convention
4th and 5th position respectively.

2.2 Crash test dummies

Crash test dummies are important components in crash safety evaluations. A crash test dummy is an
anthropomorphic standardised measuring instrument designed to evaluate risk of human injury in an accident.
Since different types of collisions cause risks for different types of injuries, there are dummies specially designed
and instrumented for different cases, such as frontal impact, side impact and rear impact. There are dummies
representing children of different ages as well as adults. For frontal impact the Hybrid III (HIII) type dummies
are used to evaluate risk of injury. Since adult occupants are differently sized the HIII type adult dummy is
available in three different sizes, corresponding to the 95th percentile of the adult male population, the 50th
percentile of the male adult population and the 5th percentile of the female adult population, see Figure 2.3,

3


Figure 2.2: The numbering convention for seat positions

Figure 2.3: Computer models of Hybrid III crash test dummies. From left to right: 95th percentile male, 50th
percentile male and 5th percentile female

where computer models of the three dummies are shown. The design and instrumentation of Hybrid III 50th
percentile male and 5th percentile female dummies is regulated by U.S Department of Transportation Code of
Federal Regulations (CFR) Part 572 Subpart E and O respectively. The 95th percentile male dummy represents
the largest segment of the adult population and is not yet incorporated in the Federal Code but is still used in
testing and in particular for seatbelt integrity testing [1].

2.2.1 Dummy instrumentation

To measure the risk of injury the HIII dummies are instrumented at several key locations. The measurements
recorded from these instruments during a crash are then used to estimate risk of injury according to different
injury criteria. Instrumentation include accelerometers in the head, thorax and pelvis, loadcells that measure
normal forces and moments in the neck, spine, femur, knee, iliac crests and tibia and also a potentiometer that
measures chest compression. Each of these measurement devices are mounted rigidly inside the dummy and
will measure normal forces and moments with regard to a coordinate system that moves with the body part it
is mounted in, see Figure 2.4, where some local body part coordinate systems are defined. It can be seen that
the coordinate systems move with the limb they belong to as displayed with the leg coordinate system as the
dummy is seated.

In Figures 2.5 and 2.6, positive force directions are defined for the force transducers. These directions are
coincident with the coordinate system directions in the respective dummy body part. Positive accelerometer
output corresponds to an acceleration in the respective positive coordinate system axis direction, e.g an impact
that would cause the dummy pelvis to accelerate in the pelvis negative z-direction will result in a negative
pelvis z-acceleration measurement. When referring to dummy measurement data in this thesis, the direction
given will always refer to the local body part coordinate system, i.e chest z-acceleration is the acceleration
measured in the z-direction of the chest local coordinate system that moves with the chest, not to be confused
with Z, that refers to the global car coordinate system Z-axis.

4


Figure 2.4: Examples of coordinate systems in a dummy that are fixed with regard to a limb, picture from
CFR Part 572

Figure 2.5: Positive forces definition in the dummy lumbar spine. The figure shows the dummy pelvis and the
position of the force transducer inside the pelvis, picture from CFR Part 572

5


Figure 2.6: Positive forces definition in the dummy lower neck, picture from CFR Part 572

6


Figure 2.7: The basic parts of a three point seatbelt

2.3 Seatbelts

A component that has a large influence on crash safety is the seatbelt. The main function of the seatbelt is to
restrain the occupant to the seat during an accident, and thus reducing the risk of injuries for the occupant by
colliding with interior parts or even getting thrown out.

2.3.1 Three point seatbelts

The seatbelts in this thesis are three point seatbelts and are called so because it is attached to the car at three
different points: the buckle, anchor and retractor positions, see Figure 2.7. It is also possible to locate the
retractor in other positions and re-route the belt using linking points such that the shoulder belt still runs
over the shoulder of the occupant. It is not unusual that this linking point is adjustable in height in order
to provide comfort and good fit to differently sized occupants. The retractor is responsible for retracting the
webbing when the seatbelt is not in use, and also keep a slight tension in the webbing to reduce slack when the
occupant is wearing it. The retractor can also lock the feed out of additional webbing when certain criteria are
met, such as high accelerations, if the car is tilted to a large degree, or signals from other systems in the car
depending on retractor model and car.

2.3.2 Risks associated with seatbelts

Certain safety risks need special attention when designing a seatbelt system for a car, such as submarining,
slip-out, maximum forward head displacement and chest compression.

Submarining is the term that describes the phenomena where the occupant slides under the lap belt, causing
the lap belt to run over the abdominal area of the occupant. This poses a great risk for severe internal injuries
for the occupant in the event of a crash [2].

Slip-out occurs when the occupants shoulder slips out of the shoulder belt and as the torso is no longer
constrained by the shoulder belt, this increases the risk of the occupant of hitting the head in the car interior,
another occupant or itself, typically in the chest or knee.

The maximum forward head displacement of the occupant is also important to control during a crash when
the occupant is belted. It is desirable to use the seatbelt as an energy absorber to minimise the acceleration of
the occupant and thus forces exerted by the belt during a crash. If the seatbelt is too compliant the occupant
may risk impact of the head with the interior of the car, such as the steering wheel or, in the case of rear seat
occupants, the front seat.

7


The compression of the chest, as well as the velocity of the chest compression, can cause injuries to the ribs,
heart, lungs and major vessels in the chest [2, 3].

2.3.3 Common safety features

Common additional safety features in seatbelts are load limiting and pre-tensioning.
Load limiting means that the locking force of the retractor is limited to a certain threshold value. If this

value is exceeded the retractor will feed out additional webbing, and thus limit the contact forces exerted on
the passenger in the event of a crash. A major advantage of load limiting is that it allows for the seatbelt to
decelerate the passenger over some distance and is an effective way to obtain energy absorption in the seatbelt.

Pre-tensioning is a rapid tensioning of the webbing in the event of a crash. This reduces the slack of the
seatbelt, often caused by loosely fitted clothing worn by the passenger. This means that the passenger will be
firmly restrained early in the crash, and thus the energy absorption provided by the load limiting can work over
a larger distance. Pre-tensioning may also help to position the passenger more correctly and can reduce the
risks of submarining and slip-out [2]. The pre-tensioning can be incorporated at the buckle, anchor or retractor
position, or at several locations simultaneously.

2.4 Safety rating organisations

Around the world there are several organisations that evaluates the safety of cars. The New Car Assessment
Programme (NCAP) aims to provide customers with an informed choice regarding the safety of popular cars
on the market, as well as stimulate car manufacturers to produce safer cars. A series of standardised tests are
used that determines the total safety rating for a car. A safety rating is provided on a five star scale, where five
stars is the highest safety rating.

To assure a high rating by these organisations, it is desirable to evaluate the tests performed in a sled test
during the design process. The crash test cases considered in this thesis are used by the European NCAP
(Euro-NCAP) and China NCAP (C-NCAP) organisations as part of the total safety rating evaluation. In
these tests, crashes are performed with dummies in the car, and scores towards the final rating of the car is
calculated from measurements obtained from the dummy instruments.

2.4.1 Crash tests

The tests evaluating rear seat occupant safety pertinent for this thesis are the Euro-NCAP full width frontal
impact test, the C-NCAP full width frontal impact test and the C-NCAP frontal impact test against deformable
barrier with 40% overlap. In both of the full width frontal impact tests, the car is frontally crashed against

Figure 2.8: Offset deformable barrier crash test (picture from Euro-NCAP press room)

a rigid barrier with an initial velocity of 50 km/h. However the dummies used and the scoring procedures
differ between the two organisations. For a detailed description of the test setup and scoring procedure, see
Appendix A. In the C-NCAP frontal impact test against deformable barrier with 40% overlap the car is crashed

8


against a deformable barrier that overlaps 40% of the width of the car with an initial velocity of 64 km/h, see
Figure 2.8. This test is also used in the Euro-NCAP rating, but only with child dummies as rear seat occupants.
A more detailed description of this test is also presented in Appendix A. In the crash test, scores towards the
final rating are determined from measurements from the following dummy instruments:

• Head accelerometer, resultant acceleration

• Upper neck load cell, Fx, Fz and bending moment My.

• Chest potentiometer, total chest compression and chest compression velocity

• Femur load cell, Fz

Additional factors that affects the final score are applied if submarining occurs, or if any part of the seatbelt
fails.

2.5 Sled testing

Figure 2.9: A rig representing the front interior of a car (grey) being prepared for a sled test. The sled (black)
can be seen sitting on the rails under the rig

Since full-scale car crash tests are destructive and very expensive, sled testing with crash simulators is often
utilised to test subsystems of the car. In a sled test, a rig is built representing the environment investigated,
which is then fixed onto a rail-borne sled, see Figure 2.9. A propulsion system then accelerates the sled, and
thus the car environment backwards, corresponding to a car crash played in reverse. The propulsion system
can deliver a controlled acceleration pulse to the sled derived from either measurements in full-scale physical
tests or computer simulated crashes.

VCSC have two sled crash test simulators, where the generic rig may be used for sled testing. KS1, with a
power rating of 1.1 million horsepower is the most powerful. It can accelerate a rig from 0 to 90 km/h in 0.1
seconds depending on weight, deliver controlled pulses with a maximum acceleration of 55 g and can also pitch
(rotation about the car Y-axis) the rig during the crash simulation using vertical cylinders under the sled rails.
The older crash simulator, KS2, has a power rating of 100 000 horsepower, and cannot pitch the rig during the
crash simulation.

During e.g, the frontal impact 40% overlap crash tests, the car will often rotate about the Z-axis, due to the
non-uniform impact forces over the car-front. The crash simulators cannot simulate this movement, but the rig
can however be mounted onto the sled with a fixed rotation about the car Z-axis, see Figure 2.10, and be tested
in this configuration, which can approximate the sideways (Y) forces present in such a crash. In Figure 2.11 a
test rig previously used for evaluating rear seat safety can be seen. The photo shows that part of the actual
rear car body have been built into a stiffening cage structure. The cage structure strengthens the car body
such that the rig may be used for several crash simulations.

9


Figure 2.10: A rig in the KS1 sled crash simulator. The rig is mounted with a fixed rotation about the Z-axis
onto the sled.

Figure 2.11: A partly dismounted previously used rear seat test rig.

10


Figure 2.12: Frontal crash rear seat safety evaluation model.

2.6 CAE crash testing

At VCSC computer simulated crashes are an important tool when evaluating the safety of a car model under
development. In this section the general features of a car crash simulation is introduced, as well as the
preprocessor Ansa, [4], and the FE-solver LS-DYNA, [5], which have been used in this thesis.

2.6.1 Features of a car crash simulation

In a car crash, large forces occur under a relatively short period of time. The large forces are sufficient to cause
several non-linear phenomena to occur, such as material yielding and breakage, large deformations causing
geometrical non-linearities, contact between different parts or even self-contact of parts, which give rise to
sliding friction or penetrations between parts.

At VCSC the aim is to model the car as accurate as possible, while still keeping the computational costs
reasonable. Internal guidelines for mesh density, element formulations, material models, connections modelling
and contact definitions have been developed to ensure consistent high quality models, and are followed in this
thesis.

Simulation model

A modern car is a complex product consisting many different materials, parts and components. Depending on
what is investigated, different parts are modelled with varying detail in different sub-models. A typical rear
seat frontal crash safety evaluation FE model can be seen in Figure 2.12. Included in this model is part of the
car body, the rear and front seats, crash test dummy and seatbelt model. The front most part of the car is
evaluated in other models, and the resulting accelerations obtained in these simulations is recorded at reference
points in the car body, which can be reapplied as boundary conditions in later models.

To simulate the crash, a rigid plate under the car body gets a prescribed acceleration, see Figure 2.13, and
nodes in the front and rear of the car body is constrained to move with this plate. This boundary condition
effectively means that the car body will not be compressed in the X-direction during the simulation. Modern
Volvo cars have a strong safety cage structure surrounding the passenger compartment. This structure, when
crashed frontally together with the doors have been proven to obtain very little compression in the X-direction
in a real crash, which justifies this type of boundary condition. This also allows for the doors to be omitted,
and thus faster solution times can be obtained.

Even with these sub modelling simplifications, more than 1 million shell elements make up the rear car
body in this particular model.

Dummy model

For each HIII dummy type an advanced and verified LS-DYNA FE-model provided by the dummy manufacturer
is available. In the dummy model, the dummy parts, materials, joints and instruments are modelled, in order to
obtain accurate predictions of the dummy measurements during a physical crash, see Figure 2.14. Each dummy

11


Figure 2.13: The rigid plate that is accelerated, seen under the car body. Nodes constrained to follow the
motion of this plate marked in pink at the front and rear of the car body

Figure 2.14: (Left) FE-model of HIII 95th percentile male dummy. (Right) Same model with some external
parts removed

12


Figure 2.15: The seatbelt webbing mesh seen placed around the dummy

model has a kinematic definition, consisting of all the joints in the dummy. For each new crash simulation case
being set-up, the dummy needs to be seated correctly in the car during the preprocessing stage. The kinematic
definition in the dummy model can be recognised by different preprocessors, and manipulation of different
dummy limbs can be obtained by changing parameters for the defined joints.

As a seated dummy will, in a physical test, depress the seat, this depression has to be obtained in the
simulation model once the dummy has been seated, and is obtained by a small simulation just focusing on
dummy seat depression. The simulation will result in the seat being deformed, and the stresses in the seat
due to this deformation can be saved and prescribed as initial stresses in a simulation, however they are often
neglected. Once this operation is performed, the seated dummy and the depressed seat configuration can be
saved, to be used in later crash simulations. This means that the dummy positioning and the seat depression
only needs to be performed once for every dummy-seat-position combination.

Seatbelt model

The seatbelt model consists of retractor, webbing, buckle and anchor. The functionalities of the retractor, such
as locking, load limiting, and webbing feed-out is obtained by a retractor model, often provided by the retractor
manufacturer. That is, the components of the retractor is not modelled in full detail, but the functionalities are
represented by special model entities designed to provide the same behaviour as the physical retractor. Since
this is the case, it is desirable to be able to verify the retractor model with physical testing.

The seatbelt webbing model consists of shell elements, with the webbing material properties. In a prepro-
cessing stage, the webbing elements must be carefully placed around the dummy, see Figure 2.15. The webbing
mesh cannot be placed too loosely around the dummy since this would introduce extra slack in the seatbelt,
and care must also be taken that the webbing mesh does not intersect other meshes, such as the dummy or
seat mesh, since elements may become locked together due to the contact modelling between different parts.
Since the webbing mesh is so tightly positioned around components, this means that if any component that the
webbing wraps is moved, the webbing needs to be re-positioned in the preprocessing stage.

2.6.2 LS-DYNA

LS-DYNA is an FE solver with capabilities to solve highly non-linear transient dynamic FE problems, which
makes it suitable for crash analysis. Special entities are available to model common car safety features, such as

13


airbags and their inflation, seatbelts with retractors and pre-tensioners [5]. LS-DYNA has no graphical user
interface and is entirely command line driven, and all input to the solver must be performed with ASCII text
files. These text files are most often generated by a preprocessor capable of producing LS-DYNA formatted
files. Version 971 release 6.85274 of LS-DYNA is used in this thesis.

2.6.3 Ansa

Ansa, version 15.2.3, is the preprocessor used to write the LS-DYNA input files used in this thesis. In Ansa, all
LS-DYNA required input can be defined. Ansa also has several tools implemented to aid in the preprocessing
of a car crash analysis, such as tools to define airbags and seatbelts, and an integrated rigid body kinematics
solver, which can be used to position CAE models of dummies or other entities,[4]. Furthermore, Ansa has a
Morphing tool that can be used to change the geometry of an FE-mesh, which can be used to quickly generate
a number of alternative designs. It is possible to couple Ansa with optimisation software, such that Ansa can
perform design changes to the structure being evaluated to find an optimum configuration. Ansa also allows
the user to define scripts that can be used to perform operations that is not implemented.

14


3 The generic rig concept

This section contains a description of the final rig concept and the development process towards this. It
includes the requirements on the rig, both those given by VCSC at the starting point of this work and also
those found during the process of studying and simplifying existing CAE models. During the process, different
sub-concepts were evaluated and these will also be presented in the following section. Some of these resulted in
new requirements while others were further developed.

3.1 Given requirements and design guide-lines

The following requirements and requests are based on the initial problem description together with consultations,
with experienced engineers and technicians at VCSC, about different phenomena and concerns that is of interest
in a rig environment. Some of these are quantifiable and can be measured, others are not and the fulfilments of
these will be discussed in Section 3.6.

Since the field of car safety is under constant development there shall be possibilities to test new components
in the rig. This implies that the rig shall be easily modified to test different components and set-ups, including
different dummies. The rig shall be generic in terms of e.g. car model, seat geometry and seatbelt positioning.
This means that seats from different car models can be used in the same rig. Also, the seatbelt attachments
shall be adjustable, preferably continuously. To reduce the costs as many parts as possible shall be reusable.
The components that needs to be changed or moved between tests shall be easy to remove and attach. They
shall also be easily produced at the workshop at VCSC at a reasonable cost.

The rig shall also be possible to use for correlation with CAE models of the rig and the components included.
Phenomena which are typically hard to measure in real life, and thus hard to model accurately in a CAE
environment, such as contacts and friction between sliding parts, need to be kept to a minimum.

Since the test rig will be a simplification of a full scale car, it is not expected that all data gathered from a rig
test will be fully comparable to a real car crash test. However, some crash test dummy data is considered more
important to capture correctly to estimate occupant safety and the following list is given by the supervisors:

• Pelvis resultant acceleration

• Chest resultant acceleration

• Upper neck Fx and Fz

• Chest deflection

These should not differ significantly from the behaviour in a corresponding full-scale crash test.

Another stated requirement is that the compliance of the seatbelt fixtures shall be possible to adjust, since
different car models have dissimilar rigidity of the parts responsible of carrying this load, and there is a need to
evaluate the effect of this from a safety point of view.

3.2 Finding requirements through model studies

In order to investigate further requirements for the rig, several questions needed to be answered. How much of
the actual crash physics must be captured by the rig in order to gather sufficient data to make design decisions?
Which parts deform and how much do they deform? Does this deformation have a significant impact on the
passenger safety? To answer these questions it was necessary to gather information about important dynamics
in the crash regarding the rear seat safety and a first step was therefore to study already performed full scale
rear seat model tests, from now on referred to as full models. From this point, mainly two reference models
were studied. One with a 5th percentile female dummy at the 4th position in 50 km/h and one with a 95th
percentile male dummy at the 5th position in 64 km/h, both subjected to crashes with a full frontal rigid
barrier. The first case will be referred to as the P4 model and and the later to the P5 model.

15


Figure 3.1: Set-up for the simplified P4 model, a model with a HIII 5th female dummy at the 4th position

Figure 3.2: Set-up for the simplified P5 model, a model with a HIII 95th male dummy at the 5th position

3.2.1 Simplified models

Starting from the dummy, some components are obvious to have significant impact on the safety, e.g. the
seatbelt and the rear seat. These are mounted on other parts of the car which in turn are mounted on other
parts and so on, each one with certain behaviour and purpose which may differ between different existing and
future car models. Thus, there was a need to decide which components should be represented in the rig, and
how.

Seat bottom plate

The previous full scale simulations suggested that there should be possible to neglect components at a certain
distance from the passenger, thanks to the stiff cage environment surrounding the passenger compartment.
Simplified models were set up where only the components immediately surrounding the respective crash test
dummy were extracted from the reference models and fixed relative to the rigid sled plate, see Figure 3.1 and
Figure 3.2. The following components were included:

• A rigid sled plate used to accelerate the rig

16


• The crash test dummy

• The seatbelt system including webbing, retractor model, buckle and anchor. The seatbelt components are
attached by the same fixture elements as in the full models, and theses fixture elements are constrained
to follow the rigid sled plate, resulting in rigid belt fixture points

• The rear seat with the seat bottom plate that the bottom rests on and necessary fixtures for holding the
backrest in position. The seat bottom plate is modelled as rigid and the fixtures are constrained to follow
the sled

• The front seat at the 1st position for the P4 model and the front seats at the 1st and the 3rd position for
the P5 model

• A rigid plate replacing the original floor

The results from an analysis with these simplified models showed that modelling the seat bottom plate as
rigid had significant impact on the data obtained from the dummy. When hitting the bottom plate the
dummy was subjected to a much higher impact force as measured by the pelvis and chest accelerometers
(Force = mass× acceleration). This is due to the fact that a rigid bottom plate offers no energy absorption
by deflection. The result from the P4 model can be seen in Figure 3.3 where the curves labelled ”Simplified –

Figure 3.3: Chest and pelvis acceleration in the z-direction for the dummy in the P4 model. The graphs show
the full model and two simplified models, one with rigid and the other with non-rigid seat bottom plate

rigid bottom plate” shows the chest and pelvis acceleration in the z-direction, for this simulation. The chest
and pelvis accelerations in the z-direction for the P5 model are shown in Figure 3.4. The peak value for the
accelerations at 50 ms represents the bouncing into the seat bottom plate and it affects all subsequent data
which makes it hard to ignore this effect. The same behaviour can be observed for other dummy data as well
and the reason is assumed to be the shock from the impact with the rigid bottom plate that goes through the
whole body. For more dummy data from these runs see Appendix B.

Figure 3.4: Chest and pelvis acceleration in the z-direction for the dummy in the P5 model. The graphs show
the full model and two simplified models, one with rigid and the other with non-rigid seat bottom plate

17


Figure 3.5: Deformed and undeformed seat bottom plate, showing the difference between a rigid seat bottom
plate and a non-rigid for the P4 model. The z-deflection at an arbitrary point is highlighted in the figure

Figure 3.6: The components contributing to the total response from the seat bottom plate at the 5th position,
and a cross sectional cut through the 5th position showing the deformation profile in the P5 reference model

Figure 3.5 shows the significant difference between the deformed non-rigid seat bottom plate compared
to the rigid undeformed plate. According to experienced analysts at VCSC this specific seat bottom plate is
unusually compliant, and in order to investigate how the compliance of the seat bottom plate affects the results
obtained from the dummy, the stiffness of the bottom was varied in a parameter study presented in Section 5.3.

New runs were performed with the same set-ups except for the bottom plate which now was assigned with
the original material properties and fixed rigidly to the sled by the existing weld points. The results for the P4
model confirmed the assumption made for the seat bottom plate. The peak values observed in the previous run
were now gone and the results can be seen in Figure 3.3, curves labelled “Simplified – non-rigid bottom plate”.
For more results see Appendix B.

The conclusion drawn from this was that the structure underneath the rear seat need to be somehow
deformable, otherwise certain data will be significantly affected. When studying the P5 model it can be seen
that the new simplified model with the non-rigid seat bottom plate did not give as good result as the P4 model,
see Figure 3.4, curves labelled ”Simplifed - non-rigid bottom plate”.

The impact forces from the larger dummy cause deflection to occur in several more components underneath
the seat bottom plate. Thus, more components under the seat bottom would need to be included in order to
get as close as the P4 model. In Figure 3.6 the components involved in the total response at the seat bottom
plate in the reference P5 model are displayed, together with the deformation profile of a cross section in the
seat bottom.

Seatbelt mounting

In the two simplified models, the seatbelt mounting points are attached rigidly to the sled, hence they allow for
no deformation. This deformation might be significant for the results and could partly explain why the results
for the simplified P5 model were more off than for the P4 model. The 95th dummy caused more deformation
in the buckle and anchor attachments than the 5th dummy, due to its weight. Therefore, the P5 model was
run again in full scale but this time the seatbelt was mounted in the same way as in the simplified model. The

18


Figure 3.7: Chest and pelvis acceleration in the z-direction for three variants of the P5 model.

result can be seen in Figure 3.7.

The curves in Figure 3.7 indicates that the deformation in the seatbelt attachment points are significant for
the 95th dummy at the 5th position. For the reference model in this case the deformation in the Z-direction
for the buckle attachment point is more than twice as large as for the anchor attachment point. Thus, the
compliance of the buckle fixture point is deemed as significant.

When it comes to the retractor, the difference in attachment is that for the simplified model the retractor is
allowed to bend down when the belt is loaded. In the original model it is prevented from doing this because of
a supporting basis, compared to the simplified model where the rear part of the retractor is fixed to the sled by
a small beam element. To see how much this rotation affects the results another simulation was performed
where the whole retractor was rigidly constrained to the sled. The results from this shows that this difference
does not affect the dummy data significantly, thanks to the compensation from the load limiting in the seatbelt.
Some results can be seen in Figure 3.8 which show the upper neck forces, Fx and Fz. To summarise, for this
car platform, the buckle attachment point should be deformable but the anchor as well as the retractor may be
rigidly attached to the sled. This confirms the importance of the given requirement that the compliance of the
seatbelt attachments shall be adjustable, at least for the buckle.

Figure 3.8: Upper neck forces, Fx and Fz, for the simplified P5 model when the retractor is rigidly constrained
compared to when it is allowed to tilt

3.2.2 Rear seat backrest

In the initial position, the dummy rests against the backrest. Simulations from front crashes show that
immediately after the crash the dummy leaves the backrest, see Figure 3.9. Thus, in a frontal crash testing
point of view, the main function of the backrest is to give the initial position to the dummy. Depending on
how the seatbelt is positioned it may affect the belt force and the dummy due to friction and the fact that
the shoulder belt passes the backrest and may come into contact. Therefore it may be possible to replace the
original backrest with a simplified component in a frontal crash test rig.

19


Figure 3.9: Interaction between dummy and rear seat backrest for the P4 model. 1. Initial position 2. After 40
ms 3. After 80 ms

3.2.3 Seat cushion attachment

In the original seat bottom model the seat cushion is only fastened to the seat bottom plate in three points in
the front end of the seat. This allows a large area of the seat cushion to slide against the seat bottom plate
during the crash. To find out if this affects the results significantly a simulation was performed where the
cushions were modelled as glued to the bottom plate, by not allowing any relative movement between the lower
seat cushion surface nodes and the seat bottom plate. The simplified models with non-rigid seat bottom plates
were used. Figure 3.10 show the chest deflection and the pelvis acceleration in the z-direction for the P4 model.
It can be seen that the two set-ups give similar results. This could also be seen for the rest of the dummy
data. The pelvis acceleration for the P5 model can be seen in Figure 3.11 and this shows the same result. The
conclusion drawn from this is that the mounting of the seat cushion to the seat bottom has less significance.

Figure 3.10: Chest deflection and pelvis acceleration in the z-direction for the simplified P4 model with non-rigid
seat bottom plate, using the original seat cushion mounting compared to gluing the seat cushion to the bottom
plate

3.3 How to capture necessary behaviour with the rig

In the previous section the focus was to find what characteristics are most important for the rig to have. This
section will instead focus on how to capture these characteristics and still fulfil the given requirements described
in Section 3.1. The approach was to replace components with solutions having a similar behaviour and which

20


Figure 3.11: Pelvis acceleration in the z-direction for the simplified P5 model with non-rigid seat bottom plate,
using the original seat cushion mounting compared to gluing the seat cushion to the bottom plate

can be part wise correlated. The reason was to avoid special-case solutions, i.e. combinations of many solutions
happens to give a good result for one case but while, in general, it is hard to adjust for an arbitrary case.

3.3.1 Foot stop

In certain rig tests the front seats are not included, and in such cases something is needed to restrain the
dummy’s legs realistically, otherwise they will not behave as in a crash and will thus affect the obtained results,
see Figure 3.12. Some different kinds of foot stops were tested at different distances and the best result was
obtained using a rigid bar, see Figure 3.13. Using this kind of foot stop bar had large impact on the legs and
thus it can not be used if leg related data are of interest.

Figure 3.12: Simulation without front seat or anything else to stop the dummy’s legs

Mainly two phenomena was observed when studying the results using the foot stop. The first has to do
with the initial impact from the bar and it appears as distinct spikes in e.g. the pelvis acceleration. Figure 3.14
shows the pelvis acceleration in the x-direction, where the spikes mentioned can be seen between 40 and 50 ms.
The second phenomenon observed was that the feet may get stuck in an unnatural way, especially if the bar is
placed at an inappropriate position, see Figure 3.15. This foot stop concept was not further evaluated but
future results will be studied with these results in mind.

21


Figure 3.13: This figure shows the bar used as foot stop at the 4th position when the front seat was not included

Figure 3.14: Pelvis acceleration in the x-direction for the P4 model when using front seats compared to a foot
stop bar

3.3.2 Tilting seat bottom plate

Simulations of the deformable seat bottom plate were studied in order to see if it would be possible to use a
rigid seat bottom plate and still catch the right behaviour by letting it tilt or move in a predefined path. In
Figure 3.16 two lines are drawn, for the P4 model, to clarify the main deformation shape of the seat bottom
plate and how it is related to its initial shape. The arrow points out where the dummy pelvis acts upon this
rear part. These lines gave the idea of a tilting rigid bottom plate employing some kind of deformable element
which can be exchanged for each run.

22


Figure 3.15: Foot stop bar is placed too close to the dummy, causing the legs to get stuck, hindering the pelvis
and femur to move forwardly as they would have otherwise

Springs

To investigate this idea of a tilting seat bottom plate, a first step was to replace the welds, attaching the seat
bottom plate, with a hinge in the rear part and two linear elastic springs in the front part, see Figure 3.17. As
reference model in this case, the simplified model, with non-rigid seat bottom plate, was used. The reason
was that this model has the same set-up except for the seat bottom plate and thereby this difference could be
evaluated separately.

Different spring stiffness were tested through rough estimations and some of the results can be seen
in Figure 3.18. The chest acceleration shows a similar negative peak value, but the elastic springs causes
significantly more throwback and overshoots the reference model results, which has plastic deformation in the
seat bottom. Between 80 and 100 ms higher values can be observed for both the acceleration and the moment.
This has to do with the elasticity of the springs and the shape of the seat bottom plate. After maximum
displacement of the seat, the springs pushed the dummy back which was not the case for the plastically deformed
original seat bottom plate. Figure 3.19 shows the difference in motion, of the dummy’s hip and thighs, between
the reference model and two different spring models. The two images are taken just before and after the peak
values shown in the graphs just described. For both spring models the pelvis location was rather close to the
reference but the thighs had tilted down too much at maximum displacement. However, after 30 ms the thighs
had been pushed up to a similar position as for the reference model. In this case the original shape was used
for the rigid bottom plate in order to keep the model clean and free from special-case solutions. The shape
makes the plate more stiff in the front part, hence a pure tilting solution could not capture the right behaviour.
The conclusion from this investigation is that some kind of permanent deformation in the seat bottom plate is
desirable and the the seat bottom shall appear somehow more stiff in the front.

3.3.3 Summary of found requirements

Except for the given requirements described in Section 3.1, the simplified models and the sub-concepts described
in previous and current sections contributed to additional requirements. A summary of these are listed below.

• The seat bottom need to be deformable with the possibility of permanent deformation

• The buckle mounting shall be deformable to some extent

23


Figure 3.16: Main deformation shape of the seat bottom plate for the P4 model. The arrow points out where
dummy pelvis acts upon this rear part

Figure 3.17: Set-up for the simplified model with rigid seat bottom plate mounted with two springs in the front
and a hinge joint at the rear part of the plate

Figure 3.18: Chest acceleration in the z-direction and the upper neck bending moment, My, for the P4 model
when using linear elastic springs compared to the simplified model with non-rigid seat bottom plate

24


Figure 3.19: Difference in motion of the dummy’s hip and thighs at two different moments. The two gray
shapes are from spring models and the black shape is the reference

• The front seats may be replaced with some kind of foot stop, depending on what is studied. In that case,
care must be taken when placing it, since the dummy’s legs might get stuck, which can affect other data

• Rear seat cushion mounting has minor importance for the results

3.4 The generic rig concept

The rig concept displayed in Figure 3.20 is designed to be generic and allow for easy changes in the test set-up,
while capturing the effects present in a real car environment during a crash. The rig concept is based on an
available sled test plate, with screw holes spaced 100 mm apart in X and Y-direction, see Figure 3.21. As every
component is screwed to this plate, changing a component is easily obtained, which allows for easy updating
of the test rig. The rig concept has two main functions, emulate the car structure under the rear seat (seat
bottom plate), and provide adjustable fixing points for the seatbelts.

3.4.1 Generic seat bottom structure

The rig seat bottom consists of three main parts:

A rigid bottom plate
Forms the base of the whole seat bottom structure and serves to obtain the main deformation path
identified in Section 3.3.2. The plate is divided into two parts; a front part rigidly attached to the hole
plate and a rear part mounted on the hole plate with hinge joints, see Figure 3.22.

A seat bottom foam
Expanded PolyPropylene (EPP) foam is attached to the bottom plate. The upper side of the foam is
shaped like the seat bottom plate in the full scale reference models, to keep the initial condition for the
dummy, see Figure 3.23. Different foam density can be used in the front and rear part respectively to
adjust the stiffness. In the current concept, a lower density softer foam is used in the rear part and a
higher density stiffer foam is used in the front part. The EPP foam material allows for easy creation
of different seat bottom geometries, since it can be easily cut into different shapes that can be glued
together.

Exchangeable plates
Small plates, referred to as bendplates, supporting the rear part of the seat bottom plate. Spikes attached
to the rigid bottom plate rests on the plates, see Figure 3.24. When the bottom plate is loaded, the
bendplates deform resulting in a tilting motion of the rigid rear seat bottom plate. The geometry, in
terms of length, width and thickness of the bendplates can be customised to capture the right behaviour.
The plates have supporting rollers which can be moved along the plates to change the point of bending
relative to the spikes. The fixture point of the plates can also be adjusted relative to the spikes.

25


Figure 3.20: The designed generic rig concept

Figure 3.21: A plate with screw holes used today for sled testing

26


Figure 3.22: Rigid bottom plate

Figure 3.23: Seat bottom foam made out of EPP foam

The seat bottom is shaped to allow for a variation in seatbelt positioning, see Figure 3.25. If there is a need to
perform a test with a rigid seat bottom, the spikes and the bendplates can be replaced with a rigid structure
under the seat bottom, thus locking the tilting motion.

3.4.2 Seatbelt fixtures

In order to adjust the seatbelt attachment points three different concepts will be presented; for the retractor,
the buckle and the anchor respectively. All of them allow for movements in the X, Y and Z-direction.

Retractor

Figure 3.26 shows the overall concept for the retractor attachment. Two towers with an interconnecting beam
form the backbone. On top of each tower there is a hat resting on a screw in the tower. It can be extended 50
mm in the Z-direction by adjusting the height of the screw. The beam is attached to the hat via a slider device,
which allows for the bar to be moved ±50 mm in the X-direction, see Figure 3.26. The slider can be fixed to
the hat at two different levels in the Z-direction 50 mm apart, giving a total flexibility of 100 mm. The slider
can be seen in detail in Figure 3.27 and Figure 3.28. Since the main load will act in the sliding direction there
is also a concept of stop screws to avoid unwanted sliding in the X-direction, see Figure 3.29. The retractor is
then mounted on the beam according to Figure 3.26. It can therefore be placed at an arbitrary position with
respect to the Y-direction.

Anchor

The anchor holder is based on two sliders similar to the one described for the retractor. They allow moving in
the X and Y-direction, ±50 mm in each direction. For the Z-direction the position can be varied ±50 mm
using shims to obtain the desired height. An overview of the anchor holder can be seen in Figure 3.30.

27


Figure 3.24: Deformable plates under a tilting seat bottom plate

Figure 3.25: Seat bottom from above

28


Figure 3.26: Towers for retractor attachment

Figure 3.27: Detail of a slider attachment

29


Figure 3.28: Exploded view of the slider mounting of the beam between the two towers

Figure 3.29: Concept of a slider attachment with extra screws to prevent it from slide when tightened

30


Figure 3.30: Anchor holder

Figure 3.31: Buckle holder

Buckle

The concept for the buckle holder is similar to that described for the anchor. There are two sliders for positioning
in the X and the Y-direction respectively and shims are used to adjust the height. The main difference is that
the deformation in the buckle attachment point was found in Section 3.2.1 to be significant. To capture this
behaviour, a concept with a deformable plate, similar as for the seat bottom plate, is applied here. The buckle
is attached to a bendplate which in turn is mounted on the top slider. When the buckle is loaded, the plate
will deform and the buckle is pulled up, see Figure 3.31. This deformation can be adjusted by changing the
position of the supporting rollers relative to the buckle fixing point as well as the thickness and the width of
the deformation plates.

An overview of the anchor, buckle, plate and hinge holders can be seen in Figure 3.32 and Figure 3.33 shows
how the different components can be attached to the hole plate. Please note that it is possible to add front
seats to the rig if there is a need for that and there is also room for measurement equipment behind the seat.

3.4.3 Assumptions made for the rig concept

There are some uncertainties regarding the rig concept and a few assumptions were made:

31


Figure 3.32: Buckle, anchor, plate and hinge holders

Figure 3.33: An overview of the hole plate showing how the holes have been utilised for all components to fit
and be fixed to the hole plate. 1: Retractor towers, 2: Retractor tower supporting legs, 3: Anchor holder, 4:
Rear seat bottom hinges, 5: Buckle holder, 6: Front seat bottom, 7: Seat bottom bendplate fixtures

32


• The weight of the rig, when fully dimensioned, is assumed to be at an acceptable level with respect to the
crash simulator which will generate the acceleration pulse

• The concepts described are possible to produce and assemble at the workshop at VCSC

• The components described as rigid can be dimensioned to be reusable and rigid in the sense that during
the tests there will they will be subjected to small elastic deformations

3.5 Results from the rig FE model

In this section the results from the rig FE models will be presented. The positioning and sizing of the deforming
elements in the rig models were tuned using target data from the reference models using system identification
in LS-OPT as described in Section 4.2.

3.5.1 Set-up of rig FE model

The rig FE model consist of the functional parts in the designed concept together with the rear seat, dummy
and seatbelt model pertinent to each case. All components and how they are constrained are listed below:

• The hole plate modelled as rigid. The hole plate is accelerated to obtain the crash boundary condition.

• The rigid seat bottom plate, divided into two parts. The frontmost part is fixed relative to the hole plate,
and the tilting rear part is mounted with joints to the hole plate to allow the tilting motion.

• The spikes applying force to the bend plates, fixed to the tilting seat bottom plate.

• The EPP foams, fixed to their respective seat bottom plate.

• The bendplates underneath the seat bottom plate and at the buckle position. The bendplates are clamped
in one end that will follow the motion of the hole plate.

• The rollers providing a second bend point for the bendplates are modelled as rigid and are fixed relative
to the hole plate.

• The rear seat FE model, including backrest and seat cushion. Depending on which dummy is present,
one of two different models of the same rear seat is used since the seat cushions are depressed according
to which dummy is seated in them. The rear seat is fixed to the hole plate using the same fixture points
as in the reference models.

• Seatbelt FE model including webbing, anchor, buckle and retractor model. One of two different seatbelt
models is used, depending on which dummy is used. The buckle is attached to the buckle bendplate. The
retractor and anchor is attached by existing screw elements (modelled by beam elements), whose end
nodes are constrained to move with the hole plate.

• Crash test dummy FE model. A HIII 95th percentile male seated at the 5th position when testing the P5
case, and a HIII 5th percentile female seated at the 4th position for the P4 case.

In order to obtain a fair evaluation of the designed rig concept, the same floor and front seat set-up existing
in the respective reference model was also included in the rig FE model, see Figure 3.34. Since the dummies will
come into contact with these parts during the crash, they will have an effect of the obtained result. Exchanging
these parts with a foot stop device is possible, but that will, as described in Section 3.3.1, change the results.

Different configurations for the deforming plates under the rig seat bottom are obtained when trying to
replicate the result from the P4 and P5 reference models with system identification, even though they have
the same seat bottom plate. This is due to the fact that the shape of the seat bottom plate as well as the
underlying components differ between the 4th and 5th seat position.

The configuration for the P4 seat bottom bendplates can be seen in Figure 3.35. The bendplates are clamped
72 mm from the contact point of the spikes, and the center of the lower rollers is located 44 mm from the
spikes. The bendplates have a width of 16 mm and are 2.5 mm thick.

For the P5 model, the lower roller is located under the spike, which means that the bendplates will not
deform. This was intentionally allowed in the system identification procedure, to allow for a configuration

33


Figure 3.34: The front seats and floor parts interacting with the dummy in the P5 reference model, now
included in the rig FE model. The deforming bendplates under the seat bottom of the rig can be seen in the
lower right corner.

Figure 3.35: The configuration for the bendplates under the rig seat bottom in the P4 model

Figure 3.36: The locked bend plates under the P5 rig tilting seat bottom. The rear seat cushion is hidden from
this view to show the deforming EPP foams. The deformation of the buckle fixture bendplate can be seen to
the right

34


where the tilting motion of the rig seat bottom is locked. This means that the EPP foams in the rig seat
bottom alone will represent rig seat bottom response, see Figure 3.36.

The bendplate and roller for the buckle were configured so that the fixture point of the buckle should obtain
the same Z-displacement as in the P5 reference model, where the large 95th percentile male dummy and the 64
km/h initial velocity cause significant displacements to occur at the buckle fixture point during the crash. The
obtained Z-displacement of the buckle fixture point in the rig can be seen in Figure 3.37.

Figure 3.37: Z-deformation of the buckle fixture point for the P5 reference model and the buckle used in the
P5 Rig.

In Figure 3.37 it can be seen that the same maximum Z-displacement is reached at the same time, however
the Z-deformation of the rig buckle lags somewhat in time compared to the reference and the rig buckle fixture
exhibits less springback. This has to do with the design of the deforming buckle fixture. With this design, it is
simply not possible to achieve both the same displacement and shape. In the reference models the buckle is
fixed to large supporting brackets under the seat bottom, and when loaded the reaction forces keeping the
buckle in place are distributed over a larger area, causing a more elastic response. In the deforming buckle
fixture concept for the rig, this load is only carried by the bendplate. Since both the compliance and the
position of the buckle fixture should be possible to adjust in the generic rig, the size of the deforming fixture
for the buckle is limited, in order to fit in the rig for every configuration.

In Figure 3.38, the configuration for the buckle fixture used in both rig models can be seen. The buckle is
fixed at a distance of 40 mm from the clamped end of the 6 mm thick and 60 mm wide bendplate. The center
of the bendplate roller is located 55 mm from the buckle fixture point. The reason that the same configuration
is used in both models is that both the P4 and P5 reference models use the same fixture point for the buckle
and thus, there should be no difference in compliance for this fixture point in the rig either.

Figure 3.38: The configuration of the buckle fixture bendplate used in both rig models

35


3.5.2 P4 rig results

In Figures 3.39 and 3.40 the results from several dummy instruments can be seen together with the reference
model results. The results obtained from the rig model in this case can be seen to closely approximate the
reference model, regarding both the overall shape and amplitude of the time signals.

36


Figure 3.39: Chest deflection and the resultant acceleration of pelvis and chest for the P4 reference model and
the P4 rig model

37


Figure 3.40: Upper neck forces for the P4 reference model and the P4 rig model

38


3.5.3 P5 rig results

In Figures 3.41 and 3.42, the results from several dummy instruments can be seen together with the reference
model results. In Figure 3.41 it can be seen that the overall shape and amplitude of the dummy instrument
time signals in the rig model matches the corresponding time signals in the reference model, although the pelvis
and chest curves seem to have their peaks somewhat shifted in time relative to the reference. This phenomenon
can also be seen for the chest deflection result, although less pronounced. This could be an effect of the fact
that the buckle fixture is deforming with a slight delay in the rig model, causing the restraint forces on the
dummy from the seatbelt to also be shifted in time. Since both the rig seat bottom and the compliance of the
buckle fixture influence the results from the dummy in the rig model, it is hard to distinguish which part of the
rig is responsible for which effect in the obtained result.

In Figure 3.42, the result from the upper neck My bending moment sensor in the rig model has a good fit to
the reference. The upper neck Fx and Fz force sensors however have lower peak magnitudes, and also reach
their peak magnitudes earlier than the reference model. It is believed that this effect is a result of the rig seat
bottom and the deformation of the buckle fixture in the rig, and not related to how the retractor is mounted in
the rig. In Section 3.2.1 it is shown that how the retractor is mounted does not influence the results in such a
way (see Figure 3.8).

39


Figure 3.41: Chest deflection and the resultant acceleration of pelvis and chest for the P5 reference model and
the P5 rig model

40


Figure 3.42: Upper neck forces for the P5 reference model and the P5 rig model

41


3.5.4 Rig results for offset deformable barrier crash test

Since it is desirable to use the rig for evaluating the Offset Deformable Barrier (ODB) crash test case, both the
reference models and the rig models were analysed with an acceleration pulse corresponding to an ODB crash,
with initial velocity of 64 km/h. The results obtained from the P4 reference model and the P4 rig can be seen
in Figure 3.43.

Figure 3.43: Results from the P4 reference model (P4 Ref) when crashed in an ODB case, and the corresponding
results obtained from the P4 rig model for the same case.

It is clear that the P4 rig behaviour is less similar to the P4 reference in an ODB crash test case than in
the full frontal rigid barrier crash test, although the rig model data follow the general trends present in the
corresponding reference model data. There is also a hump in the data from the rig model that appears at 20 ms
on all dummy sensors presented in the figure. The reason for this is unknown. In the P4 ODB case, the resultant
acceleration is a misguiding measurement, since deviations obtained in different accelerometer directions at
different times evens out the resultant. In Figure 3.44, the dummy pelvis acceleration measurements are plotted
component wise to demonstrate this effect. The P4 rig dummy pelvis is the body part where this effect is the
most prominent.

The deviating behaviour is believed to be caused by the rig seat bottom. As the rear rig seat bottom plate
is rigid and constrained to move in a tilting motion, the resistance to deformation in the Z-direction is uniform
along the Y-direction. The seat bottom plate in the reference model is, due to its shape, stiffer towards the
outer ends in the Y-direction. This difference can explain the deviation in the dummy pelvis local y-direction
accelerometer data, as seen in Figure 3.44. The dummy pelvis y-coordinate axis is parallel to the car Y-axis in
the initial seated position, and little rotation occur during the crash.

In Figure 3.45, the ODB crash test results can be seen from the P5 reference and rig model respectively.
Also for the P5 rig model the results deviate slightly, but they follow the general trends in the reference model.

42


Figure 3.44: Results from the dummy pelvis accelerometer, from the P4 reference model (P4 Ref) when crashed
in an ODB case, and the corresponding results obtained from the P4 rig model for the same case. Time is
measured in milliseconds

Figure 3.45: Results from the P5 reference model (P5 Ref) when crashed in an ODB case, and the corresponding
results obtained from the P5 rig model for the same case.

43


3.6 Fulfilment of the requirements

The generic rig concept is designed with two purposes in mind; to allow for easy set-up of different cases and to
capture the behaviour found in the reference models. The behaviour of the seat bottom plate in the reference
models was shown to have a significant impact on the obtained results. The seat bottom plate is a part with
complicated geometry fabricated with metal stamping that requires pressing tools to be constructed for every
geometry. It is not easy to produce alternatively designed seat bottoms with this method since the tools are
expensive and take time to develop. Furthermore, the seat bottom plate undergoes large plastic deformations
in a crash, and would have to be replaced between every test. As such, a real seat bottom plate is not suitable
to use in the rig, since it cannot easily be modified to investigate effects of design changes. The rig seat bottom
plate utilises EPP foam material to obtain the correct shape and also part of the deformation, and the large
plastic deformations are captured by the tilting seat bottom and the deforming bendplates. The EPP foam
material can be cut to shape in different sections, which are glued together to obtain the final shape. With this
method differently shaped seat bottoms can be built for the rig, and thus different seat cushions can be tested
to evaluate different seat designs. The bendplates under the seat bottom can easily be exchanged, and their
bend points moved relative to the spikes to obtain different deformation behaviour.

All three seatbelt fixture points are movable, and will thus allow for testing different seatbelt configurations.
For every specific type of retractor, an adapting fixture will need to be built according to the design of the
retractor before mounting it to the rig. The buckle fixture point has adjustable compliance, since deformations
significant for the results was found to occur at this fixture point. The same concept can also be applied for
the anchor fixture point and, with some small modifications, at the retractor position as well. The rig buckle
fixture however cannot obtain the same deformation behaviour as observed in the reference model, and might
need some further evaluation or a complete redesign. If the buckle fixture does not need to be as movable as in
the designed concept, more freedom is available when designing another concept. The current design is the
result of a trade-off between movability and adjustable compliance.

As the rig concept is based on the existing hole plate, all parts can be independently replaced if other designs
are needed in the future. There is also sufficient room left on the hole plate to mount frontseats. All parts are
also designed with simple geometries, and will not require special tools in order to be built; an angle grinder, a
milling machine and a welder is sufficient to construct the designed parts. However, the non-deforming parts in
the rig have not been fully dimensioned, only an approximate initial design has been developed. The resulting
total weight of the rig is therefore unknown.

The rig should also, due to it’s mechanically simple design be easy to correlate with CAE models of the
rig. The tilting seat bottom joint will have a certain friction, and sliding causing friction will also occur at
the bendplates, when sliding relative to the rollers and the spikes. From a CAE correlation point of view, the
frictional coefficients needed can easily be introduced. The bendplate contact with the circular rollers and
circular point of the spikes can be modelled using analytical geometries for the circular shapes, resulting in a
very accurately predicted contact. If the values of the frictional coefficients are uncertain, data can be gathered
from testing and e.g. system identification can be used to estimate their values to obtain a good prediction.
It is also assumed that all screw connections can be tightened sufficiently, such that no sliding at the screw
fastening positions will occur. For the adjustable seatbelt fastening points, the principle of stop screws can be
applied at all positions if needed, except for the retractor Y-direction on the beam. Here additional clamps can
be introduced next to the retractor mounting if needed to stop unwanted sliding.

The data calculated in the rig FE-model closely follows the existing trends in the reference models, especially
the P4 full frontal crash simulation. For the P5 full frontal crash test, the rig FE model also closely follow the
reference model results, although with a small time shift at the peak amplitudes, believed to be caused by the
delayed buckle fixture deformation in the rig model. Also for the ODB crash test, the rig models follow the
trends in the reference models well, but it is believed that the results can be improved for the P4 case. If the
rig seat bottom is updated to obtain stiffer behaviour towards the outer ends in the Y-direction, perhaps by
using a higher density, stiffer, EPP foam or excluding the outer ends from the tilting motion by redesigning the
rig seat bottom rigid plates, the rig FE model might produce more accurate results.

44


4 Implementation of the modelling concepts

In this chapter, CAE tools implemented during the thesis work is presented. The successful implementation
of these tools means that the CAE generic rig model can be used more efficiently, but these tools can be
implemented in other CAE models as well.

4.1 Ansa seatbelt module

The seatbelt webbing can be very time consuming in the pre-processing stage, since it has to be placed just
as a physical webbing around the dummy. Since the seatbelt webbing is just a shell element mesh placed in
space, these elements might need to be repositioned if the dummy or other parts that the webbing wraps is
moved. This is to secure correct webbing placement. In Ansa there is a seatbelt module designed to aid in the
pre-processing work of setting up the seatbelt model. By using this module, Ansa can set up a seatbelt entity
which allows both Ansa and the user to treat the seatbelt elements dynamically. With a correctly defined
seatbelt entity, Ansa can automatically update the seatbelt webbing if parts move, and the user can manually
manipulate the belt which will behave much like a string.
However, usage of this seatbelt module in Ansa has not been fully implemented at the VCSC rear interior group,
partly because lack of time and partly because there are some compatibility issues between the FE-model
generated by this module and LS-DYNA. This section will cover the steps needed to successfully set-up the
seatbelt module to automatically update the seatbelt entity, as well as the steps needed to obtain a valid
LS-DYNA seatbelt FE-model.

4.1.1 LS-DYNA seatbelt

LS-DYNA has several entities defined in order to aid in the simulation of seatbelts. The pertinent entities for
this section and their required configuration are briefly presented, and a full description is available in [5], from
where the information in this section has been gathered. The most important entities are:

*ELEMENT SEATBELT

In LS-DYNA two types of elements are available for seatbelt modelling, either a single degree of freedom
element connecting two nodes, or a four node shell element. For a single degree of freedom seatbelt
element, a tension force will be applied between the two nodes if the current distance between the nodes
is greater than the initial distance. Seatbelts modelled with single degree of freedom elements are termed
1d seatbelts, and if seatbelt shell elements are used they are termed 2d seatbelts. A 2d seatbelt must
have a logically regular quadrilateral mesh (nodal numbering must follow fixed intervals in transverse and
longitudinal seatbelt direction), and this mesh cannot be disjoint.

*ELEMENT SEATBELT RETRACTOR

Models the behaviour of a physical retractor, such as load-limiting, webbing feed-out and pull-in, as well
as keeping a small tension in the seatbelt. This entity is not mandatory in a seatbelt definition, i.e the
user may implement the retractor behaviour using custom elements and entities.

*ELEMENT SEATBELT SLIPRING

Used at locations where the seatbelt webbing is wrapped around a part with a sharp angle, typically
at the buckle. This entity is used to pass the seatbelt around a linking point with friction using a force
balance between the entering and exiting elements.

*DATABASE CROSS SECTION SET, (DBCS)
As the tensional forces in different parts of the seatbelt webbing is interesting when evaluating different
safety aspects and seatbelt behaviour, the user can specify the cross-sectional force of the seatbelt to be
output during the solution. This is obtained by defining a set of transverse nodes over the seatbelt, which
will define the cross section, and one set of elements whose internal forces will be used to calculate the
total force. A DBCS using these sets may then be defined.

LS-DYNA requires the user to specify ordered nodal sets at each slipring location and at the edge of the
non-disjoint seatbelt mesh (edgeset). For each slipring, a set of nodes defining the slipring location must be

45


Figure 4.1: Illustration of index ordering in required sets, as well as longitudinal and transverse direction in a
seatbelt webbing mesh

defined, as well as one set for the entering seatbelt elements and one set for the exiting elements, see Figure 4.1.
The transverse direction of index ordering must be the same for these sets, and this order must be consistent
for all sliprings in the seatbelt. This ordering must also be followed for the seatbelt edgeset and retractor
set. If a retractor element definition is used, the location of the edgeset nodes should be coincident with the
retractor element nodes, but if no retractor elements are used, the edgeset can be placed in any end of the
seatbelt webbing mesh.

4.1.2 Setting up a dynamic Ansa seatbelt entity

The Ansa seatbelt module comes with a graphical user interface, in which the seatbelt entity is defined in a
step-by-step process, which here is covered briefly. Each Ansa seatbelt consists of parts, typically a starting
part at the retractor, a shoulder belt part and a lap belt part, although an arbitrary number of parts may be
defined if needed. Each part has a starting nodal point and an ending nodal point. The starting and ending
nodal point must be existing FE nodes. At these nodal points, the user may specify if a slipring or retractor
should be placed. Additional nodal points between the starting and ending point may be defined, and during
the creation of the seatbelt entity, the algorithm creating the seatbelt webbing will use these points as targets
for the seatbelt to pass over, although it may not always succeed to do so. In addition, for 2d seatbelts, an
entry and exit vector must be defined to orient the webbing mesh. It is important that consecutive parts of the
seatbelt ends and starts in the same nodal point, otherwise the seatbelt mesh created will become disjoint.

When all parts of the seatbelt have been defined, the next step is to define the parts that the seatbelt
should wrap, typically some part of the seat, the dummy torso and pelvis. These parts are stored in a list
called Parts to wrap. The webbing fitting algorithm uses the shapes of these parts to determine how the
seatbelt webbing should be placed. The final step is to create the seatbelt, which will result in that Ansa
creates each seatbelt part consecutively, pausing after each part and allowing the user to manually manipulate
the positioning of the webbing. In this process, all necessary LS-DYNA entities as described in Section 4.1.1 is
created, except for the DBCS’s which the user must manually define on the created webbing mesh. In order to
achieve that the webbing placement updates automatically if parts are moved, the user has to turn on the
so-called Auto-recreate feature for the created seatbelt entity.

46


Figure 4.2: The arrows represent index ordering of required sets transverse the 2d seatbelt. Top: LS-DYNA
expected uniform ordering. Bottom: Ansa generated ordering

Auto-recreate

The Auto-recreate feature for the seatbelt entity will position the seatbelt parts correctly by simply recreating
all the seatbelt parts onto the moved configuration, if any of the parts defined in the Parts to wrap list are
moved. When the seatbelt mesh is recreated, the seatbelt elements and nodes will be created anew, with new
numbering, and thus any defined DBCS’s will no longer be valid, since the sets used refer to the numbers in
the old seatbelt mesh. In order for the Auto-recreate feature to react, a part in the Parts to wrap list must be
moved using Ansa kinematic definitions; it is not enough to simply change the nodal coordinates of the part.
However, this is not a significant issue that requires a lot of extra work, since most dummy models already
have a kinematic definition that is recognised by Ansa. However, the seatbelt buckle, anchor and retractor or
the rear seat does not have this predefined. If the user wishes to update the positioning of these parts and
thereby also wants the seatbelt positioning to automatically update, kinematic joints and actuators that can
obtain the desired movement of the parts must be defined in Ansa.

4.1.3 Compatibility issues

In version 15.2.3 of Ansa, used in this thesis, the seatbelt FE model that is created by the Ansa seatbelt module
will not be recognised as correct by LS-DYNA, and therefore will not run. The issue arises with the ordered
nodal sets at the sliprings and the edgeset of the seatbelt mesh, see Figure 4.2. The edgeset might not be
correctly placed, and the order of indexing of sets belonging to a slipring sets might become reversed relative to
the edgeset indexing. However, these issues can be circumvented, either by manually redefining the sets or by
utilizing Ansa Generic Entity Builders (GEB) together with Ansa’s scripting capabilities to further speed up
the creation of seatbelt models, and minimise the need for manual input.

Ansa GEB

The Ansa GEB is an aid for the user to define rules based on spatial coordinates. The GEB will search for
specified Ansa entities in a predefined volume around the GEB in the model space. Once the requested entity
has been found, a number of predefined rules can be applied, or the GEB can provide a user defined script

47


with information about the found entity. In order to automatically place the edgeset of the seatbelt at one of
the ends, a GEB can be placed near the endpoint of this seatbelt part and search for seatbelt elements. Once
the element closest to the endpoint has been located, a script can be provided with sufficient information to
re-locate the edgeset.

Ansa Scripts

Ansa scripts are written in the Python coding language. Scripts that rely on Ansa GEBs to relocate the edgeset
of the seatbelt, as well as setup the DBCS’s on the newly created seatbelt mesh have been created. A script
that reverses the indexing of the slipring sets has also been created. The scripts are presented in Appendix C.

4.1.4 Example model

An Ansa model that contains an automatically updating seatbelt entity has been created, and was used to
create the seatbelt configurations used in Section 5.1. A description of how the model was set-up, as well as a
tutorial on how to use it is available in Appendix D.

4.2 Automatic parameter configuration using LS-OPT coupled with
Ansa and LS-DYNA

Designing for crashworthiness can be very difficult. Large, non-linear, time dependent forces cause materials to
yield, break and come into contact, which makes the analysis non-linear and effects of design changes hard
to predict. Furthermore, design objectives, such as maximum structural strength and lowest possible mass,
is often in conflict and no clear choice of the best design choice exists. However, if mathematical objective
functions that depend on design choices made can be constructed, these objective functions can be subjected
to optimisation.

LS-OPT is an optimisation software with built in capabilities to handle single and multiple objective
optimisation. It can be coupled with several different preprocessors, which can be used to obtain design changes
in an FE-model, and solvers to obtain the resulting objective function value. In this section, the steps needed
to couple LS-OPT with Ansa and LS-DYNA are covered, and also an example model where LS-OPT has been
used to configure parameters for the deforming elements in the generic rig FE-model is presented.

4.2.1 Ansa and optimisation

Ansa has built in capabilities to be coupled with optimisation software, and a predefined optimisation scheme,
defined as an Ansa optimisation task, see Figure 4.3. In evaluating this task Ansa will perform three default steps:

Read design parameters
An ASCII file containing values for parameters will be read. It is by writing values to this file optimisation
software can command Ansa to make requested design changes.

Make design changes to FE-model The parameters read will be applied in the Ansa model. If wanted,
the parameters can first be treated in mathematical expressions. Several different Ansa entities can be
parametrised, such as material properties, nodal coordinates as well as Ansas morphing tools.

Output FE-file
The changed model is written to a requested FE-solver format.

Anywhere in this task, the user may specify other Ansa commands to be carried out, such as re-meshing parts
whose elements may have become distorted, reconfigure constraints or invoke user defined scripts. Morphing in
Ansa is a tool to change the geometrical shape of CAD geometries and FE-model meshes. Several morphing
methods exist, as covered in [4], and morphing is an efficient too