The THUMSTM Human Models

- Overview -

Dirk Fressmann
DYNAmore GmbH

Infotag Human Modeling

Stuttgart, Juni 2016

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 1
Human Models …

• developed as direct model of the human body

• represent additional tools to evaluate injury risks and
develop/improve passive and active safety systems
• „vehicle optimisation w.r.t. to humans, rather than dummies“
• reproduce anatomical geometry and biomechanical properties
of the human body
• e.g. geometry, skeletal structure, joints, stiffness- and mass distribution, etc.
• AM50, AM95, AF05, 6YO, (individual)
• are used in crash, ergonomics, seating
comfort, sport sciences, etc.
• simulation of the kinematics
of the human body
• stress- and strain evaluations
in bones and joints
• recent, more detailed models
may also allow deeper analysis
of organ injuries or more general
injury mechanisms
Comparison WorldSID, Hybrid III,
THUMS V3, THUMS V4

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 2
From Dummy to Human Model
• From human to dummy and to virtual dummy model
• From human to virtual (numerical) human model

Experiment Numerical Simulation

numerical
human Model human
model
Model

permanent development

numerical
Crash Test Model dummy-
Dummy
model

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 3
 Available Human Models
• THUMSTM – Total HUman Model for Safety
• developed by Toyota Motor Corporation and Toyota Central R&D Labs. Inc. since 2000
• additional research institutes involved (e.g. WSU, Detroit/Michigan)
• 2 versions with 2 levels of detailing
• GHBMC-Models – Global Human Body Model Consortium
• Members: Chrysler, GM, Honda, Hyundai, Nissan, Peugeot, Renault, Takata
• development at various US universities (Wake Forest, Uni of Virginia, Uni Waterloo,
IFSTTAR)

simplified and detailed

GHBMC occupant models THUMS V4
[Courtesy by www.ghbmc.com] pedestrian and occupant models

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 4
Model Variants and Versions

• Versions not available anymore

• 1.0 – first version (since ~2000)
THUMS V5,
• 1.4/1.6 – first usable generation (since 2004) relaxed and
braced state
• 3.0 – third generation (beginning of 2008)
• 4.0x – fourth (current) generation (end 2010)
current version: Version 4.02
• Version 5.0 – based on Version 3, including muscle modeling
• academic and commercial versions available
• only civil usage permitted

THUMS THUMS THUMS THUMS

V1.x V3 V4 V4.02

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 5
Model Variants and Versions
THUMS Occupant Model
• occupant simulations, belt
development, airbags, etc.
• higher biofidelity
→ front/side/rear crash situations
• driver & co-driver postures
• interest in “THUMS Family”
AM95, AM50, AF05, etc.

THUMS Pedestrian Model

• pedestrian safety simulations (head impact time
and location, qualitative injury evaluation)
• variation of posture, stance or model size
• additional interest in „THUMS Family“
(different model sizes – AM95, AM50, AF05, 6YO, ...)

• basically same modelling techniques for occupant and

pedestrian with slight modifications
(V3: internal organs, shoulder, material
properties + failure behaviour)
Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 6
Model Variants and Versions
THUMS Model Versions 1.x and 3.0
• mostly based on literature data (geometry and material properties)
• simple materials (mostly elastic, elastic-plastic, viscoelastic)
• AM50 model size, comparable to size of corresponding dummy models
• exclusively used for kinematical evaluations

Versions 1.4/1.6 (ca. 2004-06) Version 3.0 (beginning of 2008)

• kinematical model (skeletal structure, joints, • refined head model (based on CT-scans)
flesh, simplified organs, simple head model) • also: material adaptations, slight
geometrical changes
simple head model • theoretically head injury simulations
THUMS V1.x
possible

enhanced head model

THUMS V3

THUMS 2nd Generation

THUMS 1st Generation Version 3.0
Version 1.x

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 7
Model Variants and Versions
Current Version: THUMS Model Version 4.x
• no model update → new model rebuilt from scratch
• geometry obtained from medical CT scans
• basis: 39 year-old male (173cm, 77.3kg, BMI 25.8)
• scaled to AM50 model (178.6cm, 74.3kg) → realistic geometry
• very high detailing of joints, internal organs, head, …
• model parameters
• element size 3-5mm, 1.8Mio elements, 630,000 nodes
• mainly solid elements (hexa/tetra mesh) and some shell meshing

THUMS 4 occupant and

occupant upper body pedestrian thorax pedestrian models

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 8
Model Variants and Versions

THUMS THUMS THUMS V3.0 THUMS V4.0 THUMS

occupant V1.61 (not available) V4.02
models (not available) (current)
parts 1,350 1,576 1,273 1,293
nodes 66,729 104,489 628,358 762,997
elements 91,204 143,044 1,755,284 1,921,772
- deformable 70,019 118,484 1,749,575 1,916,310
- rigid 21,185 24,560 5,709 5,462
contacts 176 220 19 9
- tied 21 30 9 0
- sliding 155 190 10 9
time step size 8.55e-4 ms 3.88e-4 ms 1.45e-4 ms 4.97e-5 ms

THUMS V1.x THUMS V3 THUMS V4 THUMS V4.02

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 9
THUMS Validation Basis

• Geometry (V1/3) and materials mainly from literature

• Yamada H., Strength of Biological Materials, Williams & Wilkins Company, 1970
• Clemente, C.D., Gray‘s Anatomy, 30th American Edition of the Anatomy of the Human body
by Henry Gray, Lea & Febiger, PA, 1985
• Schneider, L.W. et al., Development of anthropometrically-based design specifications for an
advanced adult anthropomorphic dummy family, Volume 1, UMTRI-83-53-1, NHTSA, 1983
• and others.
• Experiments on human material ethically highly problematical
• Standard-Pendulum tests validated by Cadaver Tests (Ethics?)
• Thorax – lateral, frontal; Pelvis – lateral
• Leg – lateral knee impact
• Head/neck – lateral and frontal impact
• Evaluation of test corridors, thus upper and lower bounds of experimental data, mainly in
the form of force-intrusion curves

• Problem: Validation sources partly old and reliability/validity often unknown

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 10
THUMS Modeling Details – Skeletal Structure
bone structure
- trabecular bones (solids)
- cortical bones (shells)

THUMS V4 + V3
THUMS V3 + V4 occupant model pedestrian models

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 11
THUMS Modeling Details – Head and Cranium

THUMS
V3/4 eye

THUMS V1.x
head

THUMS V3/4
THUMS V3/4 head
cranium
THUMS V3/4 brain

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 12
 THUMS Modeling Details – Spine and Thorax
Source:
Sobotta

THUMS V3/4
vertebrae

THUMS V3 + V4 pedestrian
THUMS V3/V4
spine + thorax
pedestrian thorax

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 13
THUMS Modeling Details – Lower Extremities
Source:
Sobotta

femural THUMS V3/V4

bone pelvis area

knee with
patella, THUMS V3/V4
meniscus foot detail
and
ligaments

tibia and
fibula
bones

THUMS V3/v4
pedestrian knee detail

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 14
THUMS Modeling Details – Internal Organs

THUMS
Version 4.0

THUMS Version 1.x/3.0

• coarse organ modelling in THUMS v1.x-3.0

• due to coarse meshing and required model stability
• (fine) organ modelling in THUMS version 4.0

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 15
Application Example – Pedestrian Model

THUMS V3 from left and right side THUMS V4 left impact and zoom on
different kinematical behaviour stress distribution in lower extremities

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 16
Application Example – Occupant Model
Occupant Barrier Impact – THUMS 3 vs THUMS 4

THUMS V3 impact from left total model and THUMS V4 impact from left total model and
zoom on shoulder belt zoom on shoulder belt

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 17
 Special Topic:
Model Shape Modification and Positioning

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 18
 Model Positioning - Introduction / Motivation
AM50 THUMS v4
2 areas of interest in standard postures
• model positioning
• match different (non-standard) postures or seat geometries
• necessary for virtually all load cases
• model scaling/morphing
• human models like THUMS available only in standard body size, shape
and posture (AM50, AM95, AF05, 6YO)
• however: influence of individual body shapes is hardly accounted
for (skinny/obese body shapes, changes due to ageing)
• standard body sizes may not be representative any more
• necessary only sometimes, combines with positioning
 Q: how to quickly modify available human models to
create different body shapes or postures?

example for
B. Allen, B. Curless, Y. Popovic: The space of human body shapes: Reconstruction and posture change
parameterization from range scans, University of Washington, 2003

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 19
Model Positioning - Introduction / Motivation
Geometric Model adaptation rather than simulations
- straight forward approach: use FE simulations
- apply appropriate boundary conditions (impactors,
string pulling technique, etc.) to adapt the posture
- perform simulations and grab desired position from
the result files
- merge new nodal coordinates into original model file
- however:
- sometimes difficult to estimate required
boundary conditions -> iterative approach
- can be time consuming and numerically expensive
- mesh quality deteriorates after positioning simulation
-> can lead to problems in actual crash simulation
 use geometric smoothing procedures rather
than simulations
- based on control-point based non-linear interpolation approach
- apply constraints – e.g. translate/rotate body limbs to final position
- use smoothing process on interfacial parts (joints, covering flesh/skin)
- pure relocation of nodes, no change of the mesh connectivity
 required: smoothing procedure to adapt deformable interfacial parts

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 20
Interpolation Method – Problem Description
Mathematical Problem
- required: control-point based interpolation of
multidimensional data with exact fit of data points
- given: set of N data points xj (j=1,…,N) and corresponding
data values f(xj)
- N is number of control points / landmarks
- data points – nodal coordinates, data values – 3D displacements
λj – interpolation weights
ψj – interpolation function

Example: ψj – linear: - NI iso-parametric shape functions used in FE analyses

- interpolation via morphing (morphing boxes)
- only limited approximation possibilities using linear interpolation functions
- no large local deformations can be realized
 refine mesh or use higher order shape functions

- choice of interpolation function (nonlinear)

- radial basis functions - a real-valued radially symmetric function which
value only depends on the distance r to a given point
- kriging approach - geostatistical technique, based on minimization of
a Lagrangian to compute weights λj

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 21
Interpolation Method – Mathematical Background
Interpolation based on Interpolation based on the
Radial Basis Functions Kriging Approach
- choose conditions
1. minimize the scattering of the estimation error

- augmented RBF approach

2. match of expected value

 minimize Lagrangian functional

with polynomial extension pk(x)
- leads to linear equation system

 leads to linear equation system for λ and μ

- possible radial basis functions

- linear/cubic
 rearrangement leads to a dual formulation:
- “thin plate spline”
 matrix contains initial coordinates of control points
- multiquadratic
 rhs contains new coordinates of control points
 simple theory/implementation
 good results depending
on polynomial  complex theory and implementation
extension  very good results and stable
 stable system

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 22
Interpolation Method - Example
Test Example – interpolation of a test box motion
- fixation at 20 nodes at the end of the box fixation at given motion
10 nodes at 2 nodes
- displacement of two nodes in vertical-direction
- test of different interpolation procedures

 strong distortion for RBF interpolation with

constant extension
 good results for tri-linear extension
fixation at
 best results: kriging + cubic RBFs 10 nodes

interpolation with
tri-linear extension

interpolation with
constant extension

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 23
Interpolation Approach - Full Interpolation
Holistic interpolation of the Thorax
- control-point based interpolation of the whole thorax
- required
- good distribution of control points, avoid extrapolations
- exact match of control points necessary, otherwise
local distortions -> very difficult
 very fast and simple method to adapt the thorax
control point distribution
 however good shaped elements difficult to obtain
and ensure
unrealistic
sternum
shape
rib cross
section not retained

deformation of
the cervical spine

 holistic interpolation hardly suitable for FE analyses

 unrealistic deformations / bad element shapes
initial and interpolated THUMS
thorax geometry
 highly depends on distribution of control points

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 24
Interpolation Approach - Multi-Step Interpolation
 apply multistep approach to adapt the geometry in different steps
Development of an Interpolation Tool Box
- different geometric modification methods (Python)
- create batch-based geometry adaptation process

*DEFINE_SEGMENT
define (rigid) segment to be moved or aligned

*MOVE *SCALE
move nodes/parts with scale nodes/parts with
optional smoothing optional smoothing

*ALIGN *ALIGN_CSYS
align parts according to align parts according to
the motion of two the motion of a ref
reference points coordinate system

*SMOOTHING_PARTS *PROJECT_NODES
smoothing of parts with project nodes to given
given boundary base part and transform
conditions nodal positions

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 25
Multi-Step Thorax Interpolation

Multistep Interpolation of the Thorax

- here: simplified parameterization of the thorax
- statistic evaluation of a CT database
- given: fix of sternum position and shape
- given: thorax width in each rib plane
- assumption: no spine deformation
 automatic geometry adaptation using tool box

- adaptation in 7 steps
1. sternum position and shape (given)
2. adaptation of rib base
3. fix thorax width (given)
4. reconstruct ribs (keep thickness and shape)
5. thoracic skin, flesh and organs
6. transitions to head, pelvis/abdomen
7. model fixes: remove contact penetrations,
element distortions etc.

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 26
Multi-Step Thorax Interpolation
Step 1 – Adaptation of Sternum
- given: points of sternum (s1-s7)
- describe sternum position and shape
- motion of sternum points and
smoothing process

initial and adapted geometry comparison

RBF (cubic) and kriging

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 27
Multi-Step Thorax Interpolation
Step 2 – Adaptation of Rib Basis initial and adapted
geometry
- given: sternum motion
- interpolation of rib basis and
inner costal pleura
- sternum and vertebrae as control
parts (boundary conditions)

comparison
RBF(cubic) and kriging

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 28
Multi-Step Thorax Interpolation
Step 3 – Fix of thorax width initial and adapted
geometry
- given: new rib basis + sternum
- adaptation of thorax in each
rib plane (nodal displacements)
- interpolation of rib basis and inner
costal pleura, sternum and vertebrae
as control points

comparison
RBF (cubic) and kriging

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 29
Multi-Step Thorax Interpolation
Step 4 – Reconstruction of Rib Geometry initial and adapted
geometry
- given: new rib base and sternum
- reconstruction of rib
- projection of “old” rib onto rib base
and reconstruction on “new” rib base
- minimize of rib deformation
- retain rib cross section

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 30
Multi-Step Thorax Interpolation
Step 5 – Adaptation of the Thorax (flesh, skin, organs, shoulder belt)
- given: new rib, sternum and vertebrae
- adaptation of skin, flesh, organs and
shoulder belt using kriging
- rib base and costal pleura as
control parts

initial and adapted

geometry

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 31
Multi-Step Thorax Interpolation
Step 6 – Adaptation of the transitions (neck/abdomen)
- given: costal pleura and thorax flesh
- kriging of neck and abdomen/pelvis parts

Summary
 steps 1-6 can be performed automatically
 variants possible by changing of
parameters in input file
 model quality is very good

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 32
Multi-Step Thorax Interpolation
Step 7 – Model Fixes
- merge of new nodal coordinates into original model
- fix of extreme element distortions (only few in abdomen)
- fix of initial contact penetrations (only few)

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 33
Remarks & Outlook
Some Remarks
• dramatically risen interest in human body modelling in automotive industry
• currently frequent use of old THUMS V3.0 model
• primary concern: model kinematics in various crash situations → THUMS4 too detailed (expensive)
• THUMS 3 model is easier to handle (numerically and biomechanically, validation issue)
• THUMS V1-4 only passive models, THUMS V5 first active model → to be evaluated …
• no injury criteria yet available for THUMS model(s)
• direct simulation of injuries desirable, but difficult to realize (injury mechanisms, model validation)
• validation only w.r.t. crash situations, rather than biomechanical injury mechanisms
 we are still at the beginning of human body modelling in automotive applications !!!

Outlook
• increase validation database for all body regions
• increase biomechanical (user) knowledge required for result extraction
• first step: establishment of a THUMS Users Community (TUC)
• join forces in THUMS development, gather biomechanical knowledge and develop/establish
useable injury criteria
• virtually all German automotive companies involved
 first project finished, follow-up project in preparation

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 34
The End …
Thank you for your Attention

Aktuelles von den THUMS Menschmodellen, Dirk Fressmann, 2016, DYNAmore GmbH 35