Skip to main content
European Commission logo print header

Development of a Finite Element Model of the Human Thorax and Upper Extremities

Final Report Summary - THOMO (Development of a finite element model of the human Thorax and upper extremities)

Executive summary:

In 2009, in the European Union, there were 34.817 road fatalities and 1.190.448 accidents involving injuries. The annual socio-economic cost of road crashes within EU 27 is 130 M EUROS (ONISR 2010/Database CARE). Considering the 2009 figures in comparison with the 2008 statement, it appeared a decrease of -10.6% in the number of killed people and - 4% in the number of the injured people on road accidents. Nevertheless, the amount of casualties is still so important that it shall be reduced by all the available ways.

Presently, the vehicle safety devices are mainly developed with anthropomorphic dummies which, even if they have given concrete answers, show limitations with recent restraint systems and in various cases which are not taken into account by the regulation. Their "biofidelity" has still to be improved to allow the emergence of more protective systems for all the involved people, the youngest and the elderly of any corpulence. Because it seems illusory to develop physical dummies representing each person with this own characteristics, and because the regulations cannot represent all the possible crash car situations, numerical human body models could be used to assess injury risks for a wider population in wider situations, and therefore can improve the existing regulations without even so complicating them. The THOMO project aimed to develop a finite-element model of the human thorax, as a part of the Global Human Body Model (GHBM), initiated by a consortium of world-wide car manufacturers and suppliers. To this end, most of the results have been disseminated through the University of Virginia (UVa) who is the centre of expertise (COE) for the thorax.

WP1: after having obtained a license to the GHBM model and a simplified version of the thorax part through the cooperation with UVa, the thorax model has been cleaned up and its numerical stability improved in realistic test conditions. Additional information regarding the shoulder complex and the use of a simplified 2D model has contributed to a better understanding of the thorax mechanics in realistic loading conditions.

WP2: dedicated to the acquisition of new human data, very significant results were obtained. Indeed, a set of eighteen post-mortem human subjects were tested in dynamic on the thorax in different test conditions (loading type, force direction). A specific effort was done on the acquisition of rib strain profiles, in order to define new target corridors and to define typical rib fracture scenarios. Medium male and small female specimens were considered. The acquisition of new geometrical data regarding the thorax through CT-scan imagery, and µCT-scan technique for the ribs, was performed successfully. The data, complemented with specific mechanical tests on ribs were widely used inside the THOMO project.

WP3: a complete comparison of the model result with the target corridors, defined thanks to an in-depth literature review, was performed and has given good results as a whole. The most important conclusion is that the rib fracture mechanisms are well simulated. Using the data acquired in THOMO, a set of six personalized models (four males, two females) have been built and have shown promising results regarding the influence of inter-individual differences on the rib fractures.

WP4: the analysis of geometrical data acquired in THOMO and from the literature has allowed proposing scaling ratios for the anthropometrical ribcage data for building several scaled models family. The results of these models were compared with the target corridors, scaled with the usual methods. The result analysis has shown inconsistencies in several cases. Further research is needed to deepen the knowledge on this topic. The robustness of the thorax model is also to be improved in order to become a key tool for research on road user safety.

Project Context and Objectives:

Project context
Deaths and injury as a result of road accidents are now acknowledged to be a global phenomenon. On the 7th April of 2004, during the World Health Day, the World Health Organization recognized that traffic-related deaths and injuries, as well as non-intentional injury in general, are important contributors to the global burden of disease and must figure highly in public health efforts. In the 2000 Global Burden of Disease Study, road traffic injuries are considered the ninth leading cause of all deaths world-wide. By the year 2020 forecasts suggest that they will move up to the third place in terms of loss of healthy life years. According to the European Union road Federation, in the twenty-five Member States of the European Union, in 2004, there were 42193 road fatalities and 1213300 road accidents involving injuries. Road fatalities by participation in traffic in the EU 15 were distributed in 2004 as follows: 54% involved passenger cars, 20% powered two-wheelers, 14% pedestrians, 5% cyclists and 7% other vehicles. The European Road Safety Observatory believe that the socio-economic cost of these crashes to the EU 15 is about 2% of EU countries' gross domestic product - around Euro 180 billion and twice the EU's annual budget. Body regions the most severely and frequently injured in frontal accidents are the head and the thorax. In car accidents with airbag deployment, the injury severity of head injuries decreases considerably, while the percentage of severe thorax injuries remains constant: it follows from the accident data that severe thoracic injuries are mainly organic injuries and fracture of the ribs. This statement has also been verified for powered two-wheelers. For pedestrians shorter than 120 cm, the body regions most frequently injured by the bonnet leading edge of the vehicle were the arms (35%) and the thorax (20%). Thoracic injuries are observed in 40 to 50 % of injured non-belted drivers and in 50 % of public road fatalities: they are the first leading cause of death among children and young adults.

Based on the above accident data, it is obvious that in spite of significant improvements in recent years in vehicle safety, the current number of deaths and casualties added to the social and economic costs is still unacceptable. Fatalities and injuries shall be reduced by all the available ways: public regulation, prevention/education of road users, road infrastructure, compatibility between vehicles, active, passive and tertiary safety devices. With the rapid advances in finite element modelling and computational technology, human body models could become the mainstream tool for assessing vehicle safety designs. Their use could significantly minimize the need for hardware prototyping and testing, which reduces the vehicle's development time and cost, de facto enlarges the market and increases the competitiveness of vehicle industry. Furthermore, numerical models directly impact European consumers: these innovation tools generate progress in the field of passive safety through the development of new devices directly implemented in the vehicles. At last, the modelling of an occupant representative of the human body will provide better predictive capabilities of the injury risks in different crash situations.

The present project proposes to create and maintain biofidelic finite element models of the human thorax including upper extremities which will be usable at the segment, organ and tissue level. It will focus on the research, development, and validation of the human thorax region models, which include the body components of the thorax, the shoulder complex and the upper extremities for the 5th percentile female, 50th and 95th percentile male. These tools will represent every road user anthropometry, be used for every test configuration (either lateral or frontal) with every loading device type and assess injuries at the organ and tissue levels.

In 2004, a worldwide project named, Global Human Body Model has been initiated by American, Asian and European automotive companies and automotive suppliers. To run this project, a limited liability company, called Global Human Body Models Consortium, LLC ("GHBMC") was established on the 7th of April 2006 under American Law. On behalf of this company, a request for proposal was sent on the 1st of March 2007 to institutes or companies, which could be considered as potential centres of expertise (COE) in biomechanics and modelling. The objective of this company is to consolidate all the current and future research and development activities in the area of human body modelling into a single global effort. The ultimate goal is to create and maintain the world's most biofidelic human body models. To do so, a three phase research and development program is planned.

During Phase One, a set of six adult models will be developed and validated. The models will represent an average 5th percentile female, 50th and 95th percentile male. The models will be developed with existing and new biomechanics data and modelling technologies to be created. Since it is very difficult to acquire post-mortem human subjects (PMHS) with the exact sizes as desired, the model development work will have to rely on biomechanical specimens and data from PHMS with different genders, ages, and body sizes and shapes. This suggests that some data interpolation and scaling will be necessary. Two different postures will be developed: the "occupant" is in a typical automotive seated posture and the "pedestrian" is in a typical walking posture. The models should be able to simulate both "tensed" and "relaxed" occupants and pedestrians. It may be necessary to conduct tests in a variety of scenarios to obtain data on muscular involvement and posture of volunteers in anticipation of impact. In the second phase of the research and development ("Phase Two"), the Models will be further refined and a method for scaling the refined models to adult models with any age, body shape and size will be generalized and validated. With such a scaling capability, one could create adult human body models representing the population of any world market segment as desired. In the third phase of the research and development ("Phase Three"), child models will be developed.

Considering the magnitude of the research and development program outlined above, financial support of governmental authorities is expected by the GHBMc, which seeks to establish worldwide multiple biomechanics research institutes, named "Centre of Expertise" ("COE"), to share the workload and to ensure success. Under the direction of the Technical Committee of the GHBMC, these COEs will be responsible for executing research and development plans in their assigned work area and working closely with the other COEs. Two different kinds of GHBMC centres of expertise, namely Centre of Expertise in Full Body Models ("Full Body Model COE") and Centre of Expertise in Body Region Models ("Body Region Model COE"), will be established by the GHBMC.

Answers to this request for proposal shall be sent no later than the 1st of June 2007.

At the same time, CEESAR, a non-profit making organization established under French Law, is willing to answer the GHBMC request for proposal through the University of Virginia and the FP7 call for proposal with a project on the same topic. Since CEESAR does not gather all the required skills to hold the project by itself, a collaboration with European biomechanical laboratories is necessary to achieve the project objectives. As SMEs are not fully financed by the EC and as the EC encourages partners of European projects to exploit the results obtained in these EC financed projects, CEESAR is thinking on exploiting these results through the Centre of Excellence of the Thorax and Upper Extremities of the GHBMC. Combining both a FP7 European project in biomechanics and an answer to the GHBMC request for proposal with the University of Virginia seems to be appropriate and, under the general conditions for FP projects and the rules established by the GHBMC, feasible. Considering the above-mentioned challenge and in view of the "Global human body model" proposal, the potential project coordinator decided to focus on only one human body region model: the thorax.

Since the GHBMC needs public financial support, similar steps have been taken by other organisations identified as potential "Centre of Expertise": for instance, the NHTSA, the Michigan Economic Development Corporation (MEDC) or the Bill Gates foundation as well as Japanese governmental institutions have been approached in the United-States or in Japan.

Main objectives
Considering the "FP7 Cooperation Work Programme: Theme 7-Transport call fiche" and the topic "SST.2007.4.1.2 Human physical and behavioural components" the scientific and technological objectives are as follows:
-Identification of all necessary anatomical details in the models to be developed in order to reproduce body kinematics, mechanical responses, and all target crash induced injuries during dynamic loading of crash events in computer.
-Definition of the kinematics of the models which is critical to accurately simulate thorax and upper extremity injuries.
-Examination of the mechanical responses in order to accurately reproduce those of PMHS testing.
-Development of the tissue-level injury criteria that can be used to assess the risk of the crash induced injuries identified for the thorax, shoulder complex and upper extremities.
-Determination of the failure criteria, i.e. fracture/rupture force/moment, stress/strain and/or other parameters in the model so that failures are simulated at the same timing and condition as in the PMHS tests.
-Identification at just necessary detailed anatomical and mechanical level of the injuries that shall be predicted to determine the appropriate countermeasures.

Project Results:

Work package 1: Development of a 50th percentile thorax model including upper extremities.

The main activities which have been performed in the WP1 are summarized in the following chart.

After having obtained an access to a license of the GHBM models for all the THOMO partners and a simplified version of the thorax through the cooperation with the University of Virginia, the main tasks of the WP 1 have been successfully accomplished. Several significant results were obtained even then the access to the license arrived later than expected.

A clean and numerically stable thorax model in realistic test conditions.

The simplified thorax model has been developed by the University of Virginia. It is composed of 300000 shell and solid elements. All the bones are represented, the main inner organs and vessels as well. It constitutes the most advanced finite-element thorax model in the field of impact biomechanics to date.

In order to contribute to the improvement of the numerical stability of the thorax, THOMO has introduced several changes which regarded the mesh and the numerical parameters.

The clean-up process has contributed to detect and solve the inevitable deficiencies due to the heaviness of the model.
During the first review of the thorax mesh model, the following deficiencies were identified:
-asymmetry of provided thorax mesh which potentially may influence (and false) model's outcome analyses under dynamic loading.
-inconsistent modelling approach – for some parts of bone models sole SOLID type elements were used and for some mixed SOLID-SHELL type.
-model not split into parts indicating different physical type of entities (e.g. whole rib-cage treated as one part)
-empty spaces and/or overlapping between elements or parts of elements – potential problems in contact definitions.
-minor mesh quality problems like distorted elements, duplicate elements, missing elements and or large differences in mesh density depending on area.

The work performed in THOMO has been disseminated to the centre of Expertise in order to improve the mesh and correct the mistakes.

For the numerical stability, which is usually an important task for finite-element modelling, the thorax model has been tested in realistic frontal conditions. A set of twenty runs were necessary to propose a stabilized model, in LS-DYNA code. The main changes regarded the shell formulation and the numerical parameters for the contacts between the ribs and lungs, and between the muscles and skin.

During the first numerical simulations, high distortion of the elements occurred in the model and the analysis terminated after 27.5ms due to negative volumes in solid elements. At the end of the stabilization process, the model run with normal termination at 70 ms and with a satisfactory energy balance, the Hourglass energy representing less than 10% of the deformation energy. At a higher speed (6.7 m/s) which represents a very severe test condition, the runs terminated with an error at around 60 ms. The results have shown that further work should be needed to reinforce the numerical stability for which the solutions strongly depend on the LS-DYNA code.

A visual description of the shoulder complex through a filmed autopsy and an analysis of existing finite-element shoulder models of the literature.

The modelling of the shoulder complex remains an issue for human body models used in a car crash context. Whatever the impact direction is, the involvement of the shoulder cannot be ignored. In side impact, the shoulder transmits loads from the seat or the door panel to the thoracic ribcage. In front impact, the belt rests on the clavicle which transmits loads onto the ribcage. The main difficulties rely on several facts:
-complex anatomy
-large movements
-complex interactions between the ribcage and external loadings from restraint systems (belt, door panel, airbag...).

The main goals of the task were:
-to show the involvement of the different shoulder complex muscles in case of simplified large movements.
-to help the modeller in order to only take into account the useful parts (=greater than potential simplifications regarding a shoulder complex model).

For that purpose, a visual description of the shoulder functional and behavioural anatomy has been performed through a filmed biomechanical autopsy. A document explains the methodology used to reproduce several typical shoulder movements and to show the action of different muscles. Some anatomical reminders are also given for a better understanding of the video document and a description of several existing human shoulder complex models are given as well. The results have shown that simplifications can be made regarding the involvement of the muscles, which depend on the impact direction.

A pre-study on a 2D rib model for a better understanding of rib fracture mechanisms.
In order to improve the knowledge on the rib fracture mechanisms, a study on a 2D rib model has been developed. The costal arch is represented by a perfect circular ring, for which the geometrical and mechanical properties are constant along the rib. Several distributions of reaction forces were studied, and the force direction as well. The results have shown that maximum (or minimum strains) appear at specific locations, depending on the force direction. Moreover, bending mechanism is preponderant, compared to compression-tension or shearing mechanisms.

Work package 2: Human mechanical properties

The WP2 was dedicated to the analysis of existing data and acquisition of new human data, in order to define validation targets regarding the geometrical and mechanical rib properties, regarding the global mechanical behaviour in various test conditions, and at last regarding the rib fracture mechanisms. The most significant results are the following.

Relevant literature reviews.

Validation database for the thorax model
Two significant literature reviews were performed in the WP2. The first one presents a synthesis of the biomechanical tests performed on the thorax or sub-system of the thorax (no sled test configuration considered). The literature review included the Stapp Car Crash Journal, the International Research Council on Biomechanics of Injury (IRCOBI) Proceedings, the Enhanced Safety Vehicles (ESV) proceedings, the Journal of Biomechanics, laboratory report, European project reports, dissertations, and Society of Automotive Engineers (SAE) publications. It represents an amount of 57 publications.

They regard whole body thorax tests in frontal and side directions with different loading conditions (airbag, rigid surface and belt), rib cage tests, and isolated rib tests. For each configuration, the boundary conditions have been described as precisely as possible. The number of post-mortem human subjects (PMHS) per configuration, with their characteristics (age, weight...) is given, as well as the injuries. The mechanical responses, in the form of corridors (if they exist) are also supplied. The normalization process (if applicable) is at last mentioned. It has represented a very good starting point for the validation database for the thorax model. On the basis of biomechanical expertise, taking into account the objectives of the model validation (numerical stability, injury mechanism, gross motion analysis...) the configurations have been prioritised to speed up the validation process. Some of them have been removed; the reasons have been given and explained in a second report. In order to reinforce complementarities, coordination with the COE Thorax of the GHBM was necessary and effective.

Mechanical rib properties
The report is an overview of the main previous studies which aimed at evaluating mechanical properties of rib cortical bone. The review is divided in two main parts with regard to the type of test: (1) 'material' tests and (2) 'structure' tests. Moreover, it contains a synthesis which has directed the choice of the mechanical tests to perform for characterizing cortical bone material of the post-mortem human subjects used in THOMO. The main results were that the tensile coupon test is the most appropriate for defining mechanical rib cortical bone properties, and that a refined description of the geometrical properties of the coupons using µ-CT scan imagery is necessary because of the very limited accuracy of other techniques.

New human data regarding the geometrical properties on complete body
One of the issues of cadaver tests is the scattering of the geometrical and mechanical data even then a selection has been performed in order to define groups of 50th percentile and 5th percentile specimens. It must be emphasized that the choice of the specimens depends on several anthropometrical parameters, weight and stature being the main ones, but which cannot be respected at once in most cases. By definition, a perfect 50th percentile male does not exist. In practice, the choice is still more limited, because of the other parameters as the age of the specimen, positive serologic tests, and the cause of death.

Rather than to be too restrictive, it has been consider that the geometrical parameters should be characterized as systematically as possible. Theses ones can be very useful to explain dispread results regarding physical parameters as forces, displacements and injuries. For these reasons, each post-mortem human subject to be tested has been characterized using computerized tomography scanning technique. This work was possible thanks to collaboration with the Cochin-Saint Vincent de Paul hospital, the high-school "Arts et Métiers ParisTech" in Paris and LAB PSA-Peugeot-Citroën-Renault SA.

CT scan slices of 18 post-mortem human subjects (PMHS) were collected in erect position in order to obtain a positioning of the inner organs adapted to biomechanical studies. Because the maximum displacement of the tray could not cover the complete specimen in one shot, several sets of acquisition were performed:
-head + neck
-thorax + abdomen + pelvis (including the upper limbs)
-pelvis + lower legs

Two resolution levels were used. The three sets of slices were acquired with a resolution of 1 mm/pixel. For the head, a higher resolution of 0.5 mm was used also. For all the datasets, the thickness of each slice is 0.5 mm. The THOMO database is composed of eleven medium male and seven small female specimens. For nine out of 18 PMHS, standards were used for further analysis regarding the bone mineral density.

New dynamic PMHS tests in frontal, side and oblique impact.

Human data regarding the rib strain profiles in side and oblique impact

A first set of 12 dynamic tests has been successfully performed, in various conditions representing automotive loadings, id est rigid and soft (by airbag) impacts in different directions, and also by means of a diagonal belt or a 4-point harness in frontal impact. Four of them have been disseminated to the THORAX project, under the COVER umbrella project.

A specific attention was given to the instrumentation of the thorax in order to exhibit the rib fractures mechanisms. Then, for all the tests (n=12), the ribcage was instrumented with more than 100 strain gauges per PMHS glued on the ribs, cartilage and sternum. Accelerometers were also fixed onto the vertebrae (T1, T4, T12) and sternum, and at last, pressure transducers (aorta, pleurae, inferior veina cava, trachea, stomach) were put inside the ribcage in order to analyse the pressure wave transmission through the different organs. Finally, the forces applied on the thorax have been measured to quantify the impact violence. Use of 3D arm positioning system, in complement to X-ray imagery has allowed to well define the impact location and high-speed cameras have been used to detect specific events potentially useful for the analysis. After each test, an in-depth autopsy was done on the thorax in order to describe and locate accurately the fractures of the ribcage and the soft tissue injuries thanks to an Indian ink marking by injection during the test. Several additional experiments were performed on rib samples in order to characterize the rib geometry, by measuring the mineral content.

A second set of 6 dynamic tests has been performed successfully on small female specimens in side and forward oblique impact, for comparison with the medium male, with exactly the same instrumentation used for the medium males. The impactor surface (130 mm) was adapted in order to involve the same number of ribs than for the medium male.

Two types of analysis were performed. One regarded the mechanical analysis of gross motion (forces, deflection); the other was focused on the analysis of rib strain profiles.

In complement with existing data, corridors of strain profiles have been built, for each rib in side and forward oblique impactor test condition, For the fifth rib in airbag test condition. The second major result is that typical rib fracture scenarios (chronology and location of rib fractures) have be defined. The third major finding is that the strain profiles and in good accordance with those obtained with a simplified 2D model. The results confirm the strain as a good failure criterion for the rib cortical bone material.

Results on the small female specimens
The comparison between small female and medium has shown 66 % lower forces for the small females, and 14 % higher deflections. For hard tissues, the injuries are comparable. The soft tissue injuries are more numerous for the small females. The strain profiles are very similar.

New refined geometrical rib properties
Refined geometrical data on the ribs have been acquired using µCT-scan imagery. If standard CT-scanning imagery gives acceptable results regarding the global parameters for the thorax, the resolution is not enough high to well define the rib cortical bone thickness, which can be very low (0.1 mm), specifically in the anterior part of the rib.

A specific procedure has been developed for 3D reconstruction of human ribs in order to define the internal and external rib 3D surfaces. The internal surface is provided by the 3D reconstruction of each rib segment using the micro-CT imager, the surfaces between segments being interpolated; then, the external surface of each rib obtained by laser-scanning is used in order to re-position each rib segment in a local costal frame; this re-positioning is performed by minimizing distances (least squares method) between micro-CT external surface and laser-scanning external surface. Then, the external surface of each rib obtained by CT-scan is used in order to re-position each rib (both external and internal surfaces) in the global thoracic frame; this re-positioning is performed by minimizing distances (least squares method) between CT-scan external surface and laser-scanning external surface. Finally, the geometric model of the rib consists in an external surface (the one provided by the CT-scan) and an internal surface (addition of internal surfaces provided by the micro-CT and the interpolated surfaces).

Using this methodology and data acquisition performed on the hemi-thorax of two post-mortem human subjects, sixteen ribs have been fully reconstructed.

The analysis of the data has allowed for comparing the shell thicknesses assigned in the baseline model to the data coming from THOMO. The conclusion is that the rib geometry of the thorax model (local and global parameters) is consistent, when compared to the data acquired in the THOMO project.

New mechanical rib properties
In order to characterize the cortical bone material of each specimen, a protocol has been developed and applied on machined rib samples.

Tensile tests were carried out on the different coupons which were instrumented with a strain gage. The geometry of each coupon was acquired with a µ-CT scan device in order to define the real cross-sections. The data process has given the Young's modulus, ultimate stresses and strains. Different ribs of two specimens and coupons from the fifth rib, for the others were considered. The results have shown inter and intra-individual dispersions.

The most important findings are:
-the refined cross-section should be considered to compute the stress.
-the strains computed from the strain gauge or from the machine displacement are not consistent, due to the sliding of the coupon in the grip.
-the young's modulus is higher on the internal part of the rib than on the external part.
-the rib fracture location is in good agreement with the location of the minimum cross-section.

Work package 3: Model validation

Thanks to the numerically stable model developed in the WP1, the objectives of the WP3 have been achieved on the whole. Figure 15 shows the main tasks which have been completed. The most significant results regard the comparison of the model results with the validation targets for the gross motion and the personalized models for the validation of rib fracture mechanisms.

Comparison of the model results the target corridors.

A list of chosen configurations for thorax validation (gross motion)

For that purpose, the simplified thorax model improved in the THOMO project (WP1) has been used. Thanks to the literature review, several configurations were chosen in close cooperation with the centre of expertise.

These tests were chosen according to the availability of all the data needed for validation setup and resulting validation corridors. The sub-systems tests used a rigid impactor (23.4 kg) in frontal, oblique or side impacts at a speed of 4.3 m/s. The outputs were the force applied to the thorax and the thoracic deflection.

A thorax model with the arms up for a direct thorax impact

It must be emphasized that the simplified thorax model was in driver position. Because most of the chosen configurations regarded a direct thorax impact with no arm involvement, it was necessary to modify the mesh of the template model. In the same time, numerous improvements were brought on the other body regions which were suspected to influence the mechanical behaviour of the thorax (ex: the lumbar spine). The model was also prepared for an easy post-processing of the strain fields.

Results regarding the gross motion (force and deflection)
The results have shown mixed results, depending on the impact direction. In frontal or forward oblique impact, the model is stiffer than expected. In side impact, the results are satisfactory, whatever the impact speed is, 2.5 or 4.3 m/s. The following figures illustrate the main conclusions of the model-corridor comparison.

In order to suggest avenues of improvement, a sensitivity analysis has been performed, including variation of the intercostal muscles stiffness (from 0.21 MPa, to 21 MPa), and variation of material law for the heart and lungs. In frontal impact, improvements have been found, but which were not satisfactory regarding the numerical stability. The main conclusion of the study is that the stiffness of the thorax is mainly driven by the soft tissues, which were too stiff, but also too elastic.

Validation of rib fracture mechanisms
Nevertheless, the analysis of rib fracture mechanisms has been performed successfully. The strain profiles have been compared with the corridors built in the THOMO project. Here after an example of comparison on the fifth rib in side impact. Globally, the results have shown the ability of the model to reproduce the rib strain profiles. Because, maximum strain has been assigned in the model as a valuable failure criterion, the rib fracture location corresponds to the location of the minimum or maximum strain. One can also see on the figure that on the opposite side, the strain is too high, which reinforces the conclusion of a too elastic thorax. Nevertheless, the model has been used as a baseline model for the personalization of six specimens tested in THOMO, in order to see the influence of the overall thorax geometry and of the geometrical rib properties on the applied force, deflection and number of rib fractures.

A set of six personalized models.
Using the "dual kriging algorithm", a specific methodology has been developed in THOMO for personalizing the thorax model.

The process was done based on "dual kriging algorithm. It can be divided into two separate phases, which includes additional steps:
1. Preparation phase
2. Models generation phase

Before generation of personalized geometry, preparation of the data and reference model was needed. It is the most important step as a correct determination of new positions of control points is a key issue in personalization process. To ensure appropriate deformation of the ribcage a set of carefully selected, corresponding to data taken on CT geometry of the subjects, control points were defined.

Each of ribcage parts got its own control points described below in more detail:
1. Ribs - 20 control points for each of them - 4 for each cross section at 0%, 25%, 50%, 75% and 100% of rib length, in total 480
2. Thoracic spine - 11 control points for each vertebra - 8 on vertebral body, 1 for each transverse process and 1 on spinal process - in total 132.
3. Sternum - 8 control points, one at upper, lower frontal and lateral part, and 2 at each side.
4. Costal cartilage - 2 for each side in the part below sternum.

The geometry preparation phase consisted of pointing out appropriate points, in total 624, on the ribcage, which corresponds to control points defined in the model. This step was performed for all chosen subjects. 3D geometries, derived from computer tomography scans acquired in THOMO (see WP2), were used. After selecting appropriate points, the specific numbering of them was applied. It was necessary for further step to correctly calculate control points displacements.

The last step was to personalize the rib cortical bone thickness. It has been shown, in a first step, that the cortical bone thickness distribution on the baseline model was relevant. For that purpose, the data from THOMO have been used (see WP2). Therefore, it has been chosen to apply a scaling factor to the shell elements defined with the cortical bone thickness representative of each specimen. This scaling factor varied from 0.77 to 1.19.

The following figure presents the personalized models of four specimens in front and side views. The external soft tissues are removed for a better view of the ribcages, underlining the different shapes of the ribcages.

Several parameters were chosen (ex: rib length, ribcage volume, antero-posterior distance...) in order to check the results. No major difference was found. The main conclusion is that the chosen control points were relevant to well describe the complete ribcage.

Mixed results on the personalized models
The comparison between the personalized models and the PMHS tests has given the following conclusions:
-for the forces applied onto the thorax, the differences observed initially on the baseline model were also seen on the personalized models, even then, for several cases, the differences are softened.
-the strain profiles were in good accordance with those from the PMHS tests.
-the number of rib fractures on the personalized model is more consistent with the PMHS tests. On the baseline model, no fracture was observed.
-the rib fracture locations are consistent with those observed in the experiments.

Perspectives for further research
Following these mixed results, a document has been written, suggesting several improvements which are the following:
-to continue the clean-up process regarding the meshing, in order to increase the numerical robustness. Another option is to mesh the complete thorax with simplifications, in order to obtain a continuous mesh between the hard and soft tissues. It is clearly a challenge if no simplification of the thorax is done.
-to define robust interfaces between the independent meshes.
-to modify the material properties specifically on the energy dissipation, as a last step.
-to continue the analysis on the personalized models. It is highly probable that the benefit of this technique, for a better understanding of the inter-individual differences, is hidden by the global too stiff response.
-to develop positioning tools, in order to focus the effort on the validation process.

Work package 4: Models family

The figure 32 presents the work performed in the workpackage 3. The most significant results regard the building of a complete family of models by the scaling of the baseline 50th percentile male models, the calculation of scaling ratios for the cortical bone thicknesses based on the new data acquired in THOMO, and the photography of the comparison between the models results and usual scaled targets.

Definition of the geometrical skeleton properties for the models family
Global parameters
The objective was the definition of geometrical skeleton properties for the 5th percentile and 95th percentile models. Thanks to the all measured parameters on the THOMO specimens, tables have been generated to classify the different THOMO subjects and compared them with the anthropometric parameters established by Bertrand. The different subject classes are referred to the 5th, 50th and 95th percentile male and female. By this way, scaling factors have been defined for all the parameters (n=40) needed for scaling the thorax models with a good accuracy regarding the ribcage. For the large male, because no data were acquired in the THOMO project, the scaling factors were applied by extrapolation, using the results from Bertrand.

Proposal for the cortical bone thickness scaling ratios (small female and large male)
The analysis of the data regarding the geometrical properties of the THOMO specimens has shown that scaling factors can be defined between the medium males and the small females, with the rib 5 as a reference. It means that the geometrical rib properties can be estimated with a satisfactory accuracy, knowing the local geometrical properties of the 5th rib.

Due to the lack of data concerning the 95th Male, it was not able to find a relationship with the 50th Male cortical bone thickness. To solve this issue, photos of rib cross-sections from the Highway Safety Research Institute (University of Michigan) were analyzed and a scaling factor has been estimated.

As a conclusion, a scaling scheme is proposed in order to obtain the average cortical bone thickness in the ribs, for all the FE models, given this local data for the 50th percentile model.

Several models family (5th and 95th percentile)
Development of a methodology
The scaling procedure was based on modification of the selected external anthropometric measures of the human body. The internal segments, i.e. bones, muscles and internal organs were scaled proportionally to the deformation of the external contour. The pre-processing (trial series of scaling) was performed basing on data obtained from GEBOD software (Generator of Body Data), followed later on literature based measures of the human based either on REBIFFE data or newly developed by BERTRAND in his thesis set of measurements.

The work was done in a couple of consecutive steps. At first, both THOMO full body 50% models (driver position and arms up) were prepared for scaling. It included appropriate control points placement, definition of dependencies between them and definition of reference points and coordinate systems, in which scaling factors should be applied. When the model was ready, the scaling factors calculated based on chosen anthropomorphic measurements were applied to it to generate scaled 5th female and 95th male geometries.

Each of the groups consist of control points selected on outer shape of the body and also on the internal body structure (bones) which corresponds to body landmarks.

In case of the circumferential dimensions of limbs it has been estimated that the cross section has the elliptical shape, so it would be enough to scale only the semi-axes lengths to obtain the desired scaled shape. Due to this assumption, four control points lying with the good approximation on the ellipsoids' semi-axes have been chosen.

For the full body model, a set of coordinate systems describing the scaling directions has been also defined. There were:
- four systems for the upper extremity - hand, wrist, forearm and upper arm,
- four systems for lower extremity - foot and ankle, lower leg. knee and upper leg,
- one for the neck and head,
- one general for the whole thorax and pelvis, and
one for pelvis (buttock) depth.

The scaling process consisted of a series of changes applied to the location of certain control points. Such an approach required a definition of relationship between chosen markers in such a way to force a desired displacement of the control points when the other ones were modified and to keep the distances between them reasonable during scaling. The dependencies were given for all body regions separately.

In the applied methodology, a HyperMorph module was used to scale the model to the dimensions of 5th percentile female and 95th percentile male. The scaling parameters were defined as the ratio between the target anthropometric measurements and the data taken on reference 50th percentile model or in some cases data related to 50th percentile male, when some dimensions of the reference model did not correspond to the database.

Although the 5th percentile models were based on the female body dimensions, it still corresponded to the male body constitution, with an athletic chest in the neck and clavicle area and male shape of the ribcage. However, without additional information or some post-processing it was not possible to model the breasts for the female geometry correctly with the method developed in THOMO. For the completion of the objectives defined in the WP4, it has not been an issue.

Several models' families (three variants)

The set of scaled models' family of 5th and 95th geometries in 3 variants (based on different anthropometrical ref. measurements) and for two frontal and lateral applications, have been successfully generated on the basis of THOMO reference models.

The assessment and approval of the developed geometries has been conducted on the basis of simple comparison between measured and received values of chosen quantities.

Geometries of scaled models are quite different at certain areas from each other depending on scaling data and methodology applied.

Assessment of the models robustness.
THOMO project goal was development and validation of a family of human body models based on 50th percentile GHBMC model focusing on upper torso. One of the tasks, described within this report, referred to the sensitivity and stability survey of the model in a variety of test configurations and models parameters. The study covered calculations performed for reference THOMO 50th male model, which is a simplification of the GHBMC one, as well as scaled 95th large male and 5th small female models, and in the most personalized models. There were no specific sensitivity tests prepared apart from minor exceptions, but large set of different calculations was used and evaluated with respect to widely understood robustness.

The stability check was performed from different perspectives. It was examined by simple computation termination state check, from the model's energy balance and in terms of finite element mesh quality after scaling and personalization process.

The mesh quality check was done with the use of standard set of criteria proposed in HyperMesh software at the level of full model, upper body part and only ribcage component. The study shows some drawbacks in the mesh of reference model, but it can be caused by application of mismatched criteria. Scaling and personalization methods developed in the project do not introduce any significant decrease of the model's quality level.

It can also be noted that model stability is a problem. It seems the model is very close to the limit of stability, so it crashes depending on platform configuration on which it is computed. The change of different parameters, e.g. solver precision set and number of cores used, but also distortion of impact point, impact speed modification, material model parameters and mesh properties can lead to error termination of the calculations.

The sensitivity part was performed with respect to different modifications, applied inside the model as well as in testing conditions. It is an expansion of typical approach to robustness issue, as checking the susceptibility of the model to internal parameters change was directed mainly to people dealing in the future with further development of the model. Both time history response of the model and influence of applied changes on number and location of rib fractures were tested.

Analysis of the model was also done for changed impact point position, which can be called a real sensitivity study. Only two special tests were prepared, one with small and the other with substantial impactor shift. Such number of tests should be considered as a disadvantage of the analysis and should be supplemented with more variation of changes. Nevertheless, it was shown that for selected tests the model is not susceptible to small inaccuracy in defining test conditions, e.g. impactor positioning. However, for the large deviations of these parameters differences can be significant, as to be expected.

A comparison of the scaled models with scaled corridors.
The scaled models were tested in six different impactor test conditions, including frontal, side and forward oblique force direction:
-frontal speed at 4.3 and 6.5 m/s
-OSU lateral and oblique tests at 2.5 m/s
-THOMO lateral and forward oblique tests at 4.3 m/s

The results were compared to the scaled corridors, according to the usual Mertz method, which takes into account ratios of impactor and total body masses, sitting height, and bending bone modulus.

The following figures show two examples of results obtained with the scaled regarding thorax force.

The results show that the THOMO model can be a good basis for a range of variable geometries. However, the performance of the model is dependent on the impact scenario chosen. The contact impactor forces and thorax deflections show that the models are too stiff, especially in lateral and oblique setup. In case of frontal impact, significant differences have been observed among the three types of scaled modes, especially in thorax deflection. For the lateral and oblique impact, the differences were not that evident. The force and deflection scaling factors obtained with the scaled models are not consistent with the scaling factors based on Irwin et al. The factors depend on the impact scenario and the scaled model. It shows further researches for a better understanding of scaling effect on the biomechanical responses.

The limitations of the study can be seen in the fact that the scaled models carry the challenges related to the reference model. Therefore high care should be taken in the preparation of the reference 50th percentile model.

Even though the scaled models have a good performance in the side tests, they do not seem to give a reliable response in case of the frontal impact. Excluding the frontal impact validation scenario might lead to a serious misinterpretation of the results obtained with the model.

Application of the human body model to the virtual testing procedures and research and development process requires a careful choice of the validation experiments. The model has to be validated in conditions that represent settings required during the final (target) test. Thinking about the model for side impact applications, it is not possible to rely on the frontal impact validation procedures only. Same way, developing a model for the frontal impact, it is not recommended to base the model's biofidelity on the side impact response. In the real world situations, the load exerted on the human body is usually multi-directional, therefore the human body model should be reliable in different (or even combined) impact scenarios.

Potential Impact:

Contribution towards the expected impacts listed in the work programme

The expected impacts listed in the work programme are:
1. Halving the number of road fatalities by 2010 (respect to 2001 levels) and reducing number and severity of injuries caused by road accidents,
2. Halving the number of fatalities in rail transport by 2020 (respect to 2000 levels),
3. Contribute towards further reducing the risk to human life and environment associated to maritime transport,
4. Ensuring that the level of safety and security of the transport system will respond to the increasing mobility demand and crime emergence,
5. Decrease the level of human error.

The European Union is the largest car producing area in the world and the largest car market. The growing demand for greater mobility in Europe has made transportation an essential feature of modern living. The problems arising from the growth of road traffic are impacting European citizen daily life by prejudicing its quality: road users are claiming for cleaner, safer and high-performance vehicles but at competitive prices. This is a real technological challenge for the European industry especially as they try to maintain or even increase its competitiveness without putting aside transport safety: research and technological development might be the key solution to render the impact that motor vehicles have on our society more acceptable.

With this traffic increase, the risk of becoming involved in a serious accident has also risen. In fact, road-related deaths and injuries are still a major socio-economic issue in the European Union, even if road fatalities have declined by more than 17 % and road injuries 14.5% in the EU 25 since 2001. According to the European Union Road Federation, road remains, with 42'193 road fatalities and 1'213'300 road accidents involving injuries in 2004 , the least safe mode of transport in the twenty-five Member States of the European Union. These road crashes costs the EU 15 about 180 billion EUROS a year, twice the EU's annual budget. Beyond their economic cost, these accidents can hardly be accepted at the societal level because of the young age of the casualties: if the data of the International Road Traffic and Accident Database are considered, the driving people under 25 years old represented in 2004 in the OECD countries 10% of the overall population but 27% of the driver fatalities. For the 15-24 age group, traffic crashes are in 70% of the cases the cause of the death. In the view of the ERF 2006 road statistics, road fatalities are twice more numerous in the 18-20 and 21-24 than in the 15-17 and 35-44 age groups. Road accidents affect people who are in their most productive years. The loss of breadwinners and the long-term care of people disabled in road accidents drive many families into poverty, particularly in the developing countries. From this perspective, the potential consequences for the sustainable development are clear: loss of human capital and productive capacity, rehabilitation and family loss and property damage.

The proposed development of a set of biofidelic numerical models of the thorax including upper extremities is a response to the expected impact n°1: "halving the number of road fatalities by 2010 (respect to 2001 levels) and reducing number and severity of injuries caused by road accidents". These numerical models of the human thorax will contribute to road user safety because these very fine tools allow generating new concepts with a whole bunch of tests: the return on efficiency is very short in comparison with physical tests. The overall objective of the project is to realise new tools to be used in transport safety assessment and simulations and in newly designed car approval in accordance with regulations: the project focuses on the numerical simulation of the human thorax and upper extremity response to multidirectional impacts, with a tool fine enough to reproduce biofidelically the response of the bones, the soft tissues and their interactions during an impact, to potentially re-orientate transport safety assessment and vehicle design and to hypothetically influence current regulations on vehicle approval. This tool concerns any transport mode users independently of gender, size and age.

Specific objectives are:
-to contribute to the improvement of transport safety by providing a numerical tool more biofidelic than existing dummies in terms of injury risk assessment and road user representativity, and,
-to reduce the costs related to safety regulation requirements thanks to vehicle and restraint system evaluation tools more affordable, more easily produced and more accessible to all.

The necessary steps to bring about these impacts will be the:
1.Development and maintenance of a biomechanical database of post-mortem human subject (PMHS) tests at the segment (thorax) and organ (heart, lungs, aorta) levels with the necessity to
-define the mechanical validation criteria of the model using PMHS tests to be performed with global and local internal measures (strain fields of the rib cage) and for loads representative of vehicle crashes,
-improve the knowledge of the mechanical behaviour of the organs and of the mechanical and geometrical properties of the rib cage.

2-Development of numerical models of the thorax and of the upper extremities with the necessity to:
-define a model architecture allowing its validation at the mechanical and injury levels, with a previous study of the interaction model to be chosen (interaction, on the one hand, between the skeleton and the soft tissues and, on the other hand, between the organs), and to quantify the numerical and mechanical consequences, particularly at the rib fracture level,
-personalize the models from the global and local geometries obtained by scan.
-mechanical and injury validation of the thorax and of the upper extremities, i.e.
-validation from tests coming from the literature or performed during the project,
-validation at the entity level (organ, rib),
-validation at the segment level (thorax).

European and international dimension and factors influencing the attainment of the impacts

To realise the expected impacts for road safety in the future, the potential participants of GHBMC firmly believe that it is necessary to integrate at worldwide level the research capacities currently existing or emerging at national and worldwide levels. Safer road transport can only be accomplished through a joint effort in the areas of research and development, testing, harmonisation of regulations, dissemination and transfer of knowledge. The expected results will only become available if efforts are gathered and coordinated at a larger scale than Europe.
As previously explained, the GHBMC was established on the 7th of April 2006 under American Law to run the GHBM worldwide project whose main objective is to consolidate all the current and future research and development activities in the area of human body modelling into a single global effort. The ultimate goal is to create and maintain the world's most biofidelic human body models. To do so, all the centres of expertise in biomechanics have been contacted worldwide.

CEESAR, which has been recognized potential centre of expertise in biomechanics, took into account current European research such as past FP projects as well as national and international programmes in order to plan ambitious but realistic research objectives and to avoid any doubling of work and any resource and time waste through the integration of the project into the GHBM programme. Since CEESAR does not gather all the required skills to hold the project by itself, this project is based on the collaboration of some of the main and the best biomechanical laboratories in Europe to achieve the project objectives, especially as recognized biomechanical centres are rare at national and European levels: a high level of expertise, which has been brought together in this project from universities and passive safety research institutes, would not be available in a single country, like the development costs of the project which certainly exceed the financial capabilities of national groups. Moreover, CEESAR, through its proposed project THOMO, is willing to make European biomechanical centres work together to represent the European work force in the GHBM project.

At last, the University of Virginia, which is, with CEESAR, the most prolific organisation regarding scientific publications about the thorax in the fields of biomechanics and passive safety, will cooperate to the project and therefore increase its international dimension: in the past ten years, the University of Virginia and CEESAR are authors in most of the SAE, STAPP and ESV publications. Collaboration between the two most active organisations in this topic is of particular interest since their competencies will be pooled to achieve the project objectives.

List of Websites: