TNO report. Automotive Schoemakerstraat 97 P.O. Box JA Delft. T F

Transcription

1 Evaluation and comparison of the biomechanical responses of MADYMO Finite Element 5th percentile Human body model and Hybrid III 5th percentile dummy model in static OOP simulations Report No.: MT06.13

2 TNO report Evaluation and comparison of the biomechanical responses of MADYMO Finite Element 5th percentile Human body model and Hybrid III 5th percentile dummy model in static OOP simulations Automotive Schoemakerstraat 97 P.O. Box JA Delft T F info-ient@tno.nl Date February 7, 2006 Author(s) Reyes Orozco G.A. Assignor Project number Classification report Title Abstract Report text Appendices Number of pages Number of appendices 67 (incl. appendices) All rights reserved. No part of this report may be reproduced and/or published in any form by print, photoprint, microfilm or any other means without the previous written permission from TNO. All information which is classified according to Dutch regulations shall be treated by the recipient in the same way as classified information of corresponding value in his own country. No part of this information will be disclosed to any third party. In case this report was drafted on instructions, the rights and obligations of contracting parties are subject to either the Standard Conditions for Research Instructions given to TNO, or the relevant agreement concluded between the contracting parties. Submitting the report for inspection to parties who have a direct interest is permitted TNO

3 TTNO report 2 / 64 Summary The present work presents a comparison between the MADYMO Hybrid III 5 th %tile dummy model and the HUMOS2 5 th %tile finite element human body model in Out Of Position (OOP) driving situations. The comparison is based on the evaluation of the biomechanical responses of both models; furthermore, biomechanical responses are evaluated by means of injury criteria. Prior to the evaluation of the responses, a literature review was done, with which the student could familiarize with the adopted numerical models, the OOP standarized tests and the biomechanical injury criteria used to evaluate the models responses. Two standarized tests are used on this research, namely the chin on module and chin on rim OOP tests. These tests are part of the Federal Motor Vehicle Safety Standard No. 208 (FMVSS 208) also called occupant crash protection, which was published by the National Highway Traffic Safety Administration in At the present time this is the only available international regulation recognized to evaluate this type of crash situations. Both models (Hybrid III and the HUMOS2) are placed in both OOP positions according to the regulation giving place to four different results. The results of the simulations with the Hybrid III 5 th percentile model and the HUMOS2 5 th percentile model are analyzed, compared to each other and finally evaluated according to injury criteria. To finalize this report conclusions as well as some recommendations are given.

4 TTNO report 3 / 64 Contents Summary Introduction MADYMO Introduction Pre-processing Multibody module Finite element module Finite element airbag model Force interaction models Contact definition Out of Position Situations Definition Background OOP Regulations Injury criteria Literature review on the development and validation of the FE human body model Introduction FE human models HUMOS2 model Validation of HUMOS2 model th percentile HUMOS2 model Application of Hybrid III dummy model in the standarized OOP test environment Dummy model positioning Simulation results th 6 Application of 5 percentile FE Human body model in the standarized OOP test environment Human model positioning OOP simulation set-up Simulation results TH 7 Comparison of FMVSS208 simulations with HUMOS2 and HYBRIDIII 5 percentile dummy FMVSS 208 chin on module position FMVSS 208 chin on rim position Discussion HUMOS2 OOP simulations Comparison with dummy model Conclusions... 61

5 TTNO report 4 / Recommendations References... 63

6 TTNO report 5 / 64 1 Introduction 1.1 Background As an increasing concern among governments, car manufacturers and their suppliers, car occupant safety has become in the past few years a big area of research. Car occupant safety can be divided in two well known parts in the automotive world: Active and Passive Safety. Active Safety can be defined as all the measures taken to prevent a crash from happening. Passive Safety comes in to action when an accident is inevitable; in such a situation Passive Safety features such as seat belts, airbags, pretentioners, etc. are designed to prevent or minimize injuries to the vehicle s occupants. Nowadays passive safety requirements are determining factors in the car designing process. Until now most of these safety requirements are evaluated by means of mechanical human substitutes (also called crash test dummies or anthropomorphic test devices). Biofidelity is the ability of any human surrogate to mimic human physical characteristics such as size, shape mass, stiffness, and energy absorption and dissipation [7]. It has been observed that anthropomorphic test devices (ATDs) lack of biofidelity [15] and of ability to provide organ specific injury response information [10]. For the lack of biofidelity and the fact that most of these tests are rather expensive, there is an increasing need for cheaper tools and ways to evaluate the safety of a car. Here, the computer is widely used but again the numerical models developed from the dummies are as biofidelic as a dummy can be, thus not biofidelic enough since they are based on the physical crash test dummies and not on humans. The HUMOS II (Human Models for Safety II) project, in which the present work was developed, is a European project whose main goal is to deliver Human body numerical models that can be used in the evaluation of safety requirements for cars and restraint systems, making the evaluation process less time consuming as well as cheaper. Another goal of the HUMOS II project is to evaluate the developed models injury prediction capacity in several impact conditions. The results of the present work are part of the investigation of the models injury prediction ability in Out Of Position (OOP) tests. Out-of-position is defined by Rekveldt (2001) as: An out-of-position occupant is an occupant who is not situated in his normally assumed seating position but is located in the deployment space of the airbag [22]. OOP tests here evaluated are part of the standard FMVSS 208 also called occupant crash protection. From this standard, the positions known as chin on module and chin on rim are investigated. Main aim of this work is to obtain a comparison between the numerical 5 th %tile HUMOS2 model

7 TTNO report 6 / 64 developed during the HUMOS II project and the existing 5 th %tile Hybrid III dummy model from an injury prediction point of view. 1.2 Report s structure This report is divided in 8 different chapters in which the activities realized during the internship are summarized in accordance with the plan agreed at the beginning of the internship. Chapter 1 gives a general introduction about the background and the goals of this research work followed by a brief description of the used software (MADYMO) given in Chapter 2, in which the most important features implemented for the realization of the project are commented. In Chapter 3, background and definition of the Out-Of-Position (OOP) problem are given. Additionally, the regulations and injury criteria used to asses it are described. In Chapter 4, a summary is given of the relevant literature on the development and validation of the FEM human body models. In Chapter 5 a general overview of the positioning procedure (in both standarized test positions) of the 5 th %tile Hybrid III dummy model is described. Chapter 6 deals with the description of the procedure followed to position the HUMOS2 model in the standarized OOP test environment. Chapter 7 presents the comparison of the obtained results where the Hybrid III simulations results are compared with the HUMOS2 simulations. Chapter 8 deals with the analysis of the results with focus on the injury criteria as well as some recommendations.

8 TTNO report 7 / 64 2 MADYMO 2.1 Introduction Mathematical models of dummies and humans in combination with a mathematical description of the vehicle appear to offer a very efficient method for the numerical analysis of crash responses. In the present chapter some of the most important features of the adopted software (MADYMO) for the simulations realized in this investigation will be highlighted. For an extended reference the reader is referred to the MADYMO theory manual [5]. The version of MADYMO used in this investigation was MADYMO (Mathematical Dynamic Model) is a simulation package developed in Europe by TNO Automotive in Delft the Netherlands. It simulates dynamic behaviour of physical systems concentrating in vehicle collisions and injury prediction. MADYMO combines the MB (multibody) as well as FE (finite element) techniques. Multibody is used when the user is mainly interested in the kinematics of the system; it provides a faster analysis but without detailed information about the structural behaviour of the system. FE is used when de user wants to make a more in depth analysis of the structural behaviour of the system. 2.2 Pre-processing Many pre-processing packages are available to set up the MADYMO simulations. Among others XMADGIC and XML Spy are some of the most used packages in the Safety department within TNO Automotive. The MADYMO input files have XML extension (Extensible Mark-up Language). The xml file contains all the needed information relative to the simulated environment. A simulated environment is formed by several components that are described in the following paragraphs. 2.3 Multibody module The multibody technique is the simulation approach adopted in one of the modules within MADYMO. This module analyses the motion of systems of rigid bodies. A rigid body is defined by its inertial properties (mass, moments and products of inertia) and the location of its centre of gravity. Rigid bodies can be interconnected by kinematic joints such as spherical joints, revolute joints, universal joints, and translational joints, which restrict the relative motion of the connected bodies. In addition, simple geometrical shapes (planes, ellipsoids, cylinders) can be attached to a body and used to represent the surface of a specific body. The surfaces are used for visualization purposes and allow the definition of contact interactions.

9 TTNO report 8 / Reference space and Multi-body systems The motion of any rigid body in MADYMO is calculated with respect to an inertial system. This system is called the reference space. The orientation and origin of the reference space can be chosen arbitrarily. All multi-body systems are defined relative to this system. A multibody system is a system of rigid bodies, in which any pair of bodies is interconnected by one kinematic joint. In MADYMO the multibody structures are called tree structures. A system of bodies is defined by: - The bodies: the mass, the inertia matrix and the location of the centre of gravity. - The kinematic joints: the bodies they connect, the type, and the location and orientation - The Initial conditions. Furthermore the shape of bodies may be needed for contact calculations or post-processing (graphic) purposes Kinematic joints A kinematic joint is a link between any two elements of the tree structure and limit the motion of the bodies defining their degrees of freedom. Within MADYMO a library of kinematic joints is available: Figure 2.1 summarizes several of the most frequently used joint types within MADYMO. A local reference system is attached to each of the two bodies, in this way, for every joint, two different reference systems are defined, one rigidly connected to the first body (called parent body ) and one rigidly connected to the second body (called child body ). The relative motion/rotation between the two systems represents the joint degrees of freedom. Figure 2.1 Examples of kinematic joints in the MADYMO software [5]

10 TTNO report 9 / Facet surfaces A facet model is a multi-body model where non-deformable meshes made of membrane-type massless contacts elements are used to define the model s outer surface. These surfaces are fully connected to rigid and/or flexible bodies allowing for a more accurate geometry representation. Facet surfaces are MB features. Deformation is represented by a force or stress-based contact, where contact forces are exchanged for each facet which comes into contact and not in one single point as it happens in the case of planes or ellipsoids. However, the more detailed contact definition causes an increment on the computation time. A facet surface simulation time is larger than a simple multi-body (planes, ellipsoids) simulation but is smaller than a complete FE simulation time. 2.4 Finite element module In the finite element module MADYMO features full FE capabilities for structural impact analysis. Within this module truss, beam, membrane, shell and brick elements are implemented. Several material models such as elastic, elasto-plastic, Mooney-Rivlin, etc. can be used. Several finite element models can be used within one simulation. A MADYMO model can consist of only multi-body systems, only finite element models or both Time integration A reciprocal interaction between the finite element module and the multibody module allows the use of different time integration methods for the equations of motion for the finite element and the multi-body modules. For short duration crash analysis, explicit integration methods are preferred. For a MADYMO analysis with a finite element model, the 4 th order Runge- Kutta or Euler method must be used for the time integration of the equations of motion of the multibody module. The central difference method is used for the time integration of the equations of motion of the finite element models. Actual positions and velocities at each time step of the central difference method determine the support and contact forces. The forces acting on the multi-body system are accounted for in each main time point of the 4 th order Runge-Kutta and each time step of the Euler method. Explicit methods are used within MADYMO for solving the equations of motion of the finite element models. In an explicit method the displacements and velocities are calculated from quantities at previous points in time only. Explicit methods are conditionally stable, this means that the time step must be small enough to avoid that the solution grows without bound. The time step for undamped linear systems is limited by the Courant stability condition [16]: Δt 2/ω (2.1) Where ω is the maximum eigenfrequency appearing in the mesh. Due to the fine spatial discretization often required, a much smaller time step is needed for finite element models than for multi-body models. The finite element

11 TTNO report 10 / 64 analysis is sub-cycled with respect to the multi-body analysis using a different constant time step to increase the efficiency of the entire analysis. For contact between different finite element models the time step is identical Material models All structural materials are elastic to a certain extent, if the applied loads do not exceed a certain limit; the deformation disappears with the removal of the loads. When loaded beyond the elastic limit, plastic deformations remain after removal of the loads. Often the material is assumed to be: - homogeneous = the smallest part cut from the body possesses the same specific mechanical properties as the body - isotropic = the properties are the same in all directions. Usually, structural materials do not satisfy all the above mentioned assumptions. Even metals consist of crystals that vary in size and orientation. However, as long as the dimensions of a body are large in comparison to the dimensions of a single crystal the assumption of homogeneity can be used. If the crystals are randomly oriented the material can be treated as isotropic. Within MADYMO the following material models are available: elastic, elasto-plastic, woven fabric, interface material, hyper-elastic, visco-elastic, sandwich, solid foam, spot-weld. Damage can also be implemented as well as honeycomb material behaviour. [5] Element types With the use of the finite element method, a continuous system is reduced to a discrete numerical model, where the actual continuum is separated into a collection of finite elements. The elements are assumed to be interconnected at a discrete number of points, the nodes of the elements. Within MADYMO there are many types of elements available for example: trusses, beams, membranes, shells and solids. The element formulation is based on linear displacement interpolation and integration at a single point at the centroid of the element. For some elements this leads to zero-energy or hourglass modes which are instabilities in the elements. The term zero-energy mode refers to a nodal displacement vector that is not a rigid-body motion but nevertheless produces zero strain energy. These modes appear because of shortcoming in the element formulation, such as use of a low-order Gaussian integration [17]. Hourglass can occur in a single element or a mesh of elements. MADYMO implements an hourglass control algorithm to suppress these modes. 2.5 Finite element airbag model The MADYMO finite element module includes gas thermodynamics to make it suitable for airbag applications. The discretization of the airbag fabric in finite elements allows for a more accurate description of the motion of the fabric during airbag deployment and the contacts with objects. Effects of inertia, bag slap, and pressure forces on objects that come into contact are accounted for during airbag deployment. The gas in an airbag chamber is treated as a mixture of ideal gases and the state variables, pressure and

12 TTNO report 11 / 64 temperature, are assumed to be uniform throughout the chamber [5]. As an option, the pressure distribution in the airbag can be converted to a gas flow model by implementing a gas jet in the model where the inflator location and gas flow direction are defined [18]. This second option was used in this investigation Airbag finite element mesh The fabric skin of an airbag must be modelled with membrane elements, since membrane elements show a more stable behaviour under large distortions which is the case when an airbag is folded or deployed. The elements used to model an airbag chamber must form a closed surface so that the volume inside the camber can be defined. The volume of an airbag chamber is the sum of all the volume contributions of the membrane elements that make up the skin of the chamber. [5] Initial metric method Two different mesh configurations of the airbag fabric must be defined in MADYMO. The first configuration must represent the airbag in the initially folded configuration where each node s initial coordinates must be specified (initial configuration, folded airbag). The second mesh also called reference or design configuration represents the airbag fabric in the undeformed configuration, which is when the airbag is unfolded and no gas pressure is present in the airbag. By means of software (i.e. PAM-CRASH/Safe editor, MADYMO/Folder) the design configuration can be turned in to the folded configuration. In some cases parts or the whole airbag are simply scaled from the design configuration to obtain the airbag in its initial configuration. Since both of these processes will always introduce distortion to the elements a method is needed to account for these distortions. This method is called the Initial Metric Method (IMM) and is used when the initial and the undeformed configuration of the airbag are specified. The method establishes a mapping between the two configurations calculating strains and stresses with respect to the reference configuration Airbag inflation process The inflation process in the airbag results from the inflators-supplied gas mass. As gas mass is injected into the airbag, the airbags volume is increased and the airbag fabric skin is loaded with internal pressure. Due to exhaust orifices and fabric own permeability gas can flow in and out of the airbag. The mass balance in an airbag chamber is calculated as: Where: m m m m s i ex. m = inflator-supplied gas mass = the inflowing gas mass = exhausted gas mass = gas mass in the chamber... = m x + mi m ex (2.2)

13 TTNO report 12 / 64. Where the m, ṁ s, ṁi and ṁ ex are the time derivatives of m, m respectively. [5] 2.6 Force interaction models m, and m s, i ex The motion of a system of joint-connected bodies is caused by applied forces. MADYMO offers a set of standard force-interaction models. The various categories of force-interaction models are summarised below: - Acceleration field model - Spring-damper elements - Muscle models - Contact models - Belt model - Dynamic joint models - Additionally the user can define and link his own routines to the MADYMO multibody module. 2.7 Contact definition Within MADYMO, three different ways are used to define contact interaction between surfaces: - Multi-body Multi-body (MB_MB) - Multi-body Finite Element (MB_FE) - Finite Element Finite Element (FE_FE) The basic principle of the contact model in MADYMO is that a contact force is generated between two colliding surfaces. The two contacting surfaces are called master and slave surface. Depending on the type of contact, different contact models are available to calculate de contact force. For MB_MB, the contact force is a function of the penetration of the two surfaces (plane, cylinder or ellipsoid) as well as of the relative velocity in the contact area in this definition the Elastic contact model is used. In MB_FE, the contact force can be calculated using the kinematic contact model where no penetration is allowed or the elastic contact model where a FE node is allowed to penetrate the master contact surface. For FE_FE contact force calculation, two contact models are available: Penalty based contact and Elastic characteristic contact. In order to calculate the reaction forces due to contact, for all surfaces involved in contacts the mechanical characteristics should be defined. The reaction force is the sum of the elastic forces (function of the penetration), the viscous damping forces (defined by a damping factor and dependent of penetration velocity) and the friction forces (which are assigned to every pair of surfaces). In order to specify contacts groups must be defined in a list called GROUP.MB or GROUP.FE depending whether the group defined is formed by multibody or finite element surfaces.

14 TTNO report 13 / CONTACT.MB_MB In this type of contact all contacts between ellipsoids, cylinders and planes are considered. In these contacts the slave surface is defined as ellipsoids and the master surface is defined as planes, cylinders and ellipsoids. Only the elastic contact model can be applied [5]. A positive value of the resulting contact force is applied to the contacting bodies in a single point. The contact force s point of application depends on the geometry of the contacting surfaces and on the way the contact characteristic has been defined. A surface is treated as rigid or deformable depending on weather its contact characteristic is used; if the contact characteristic of a surface is used this surface is considered as deformable and any other surface involved in the contact is considered as rigid CONTACT.MB_FE This type of contact is defined between finite element surfaces and multibody surfaces. In these contacts, the slave surface is defined by the nodes of the finite element surface and the master surface is defined as planes, cylinders and ellipsoids. Contacts can be calculated using either kinematic or elastic contact model. Specifically, the kinematic contact model does not allow penetration between the surfaces in contact and is only available for the contact between nodes of a finite element model and ellipsoids, cylinders and planes (MB_FE contacts). The contact force is based on an inelastic impact of the node with the contact surface. When contact occurs, a normal impulse is applied to the node and to the rigid body to which the MB contact surface is attached so that the relative velocity component perpendicular to the contact surface becomes zero CONTACT.FE_FE Contact FE_FE is used when defining contacts between two finite element surfaces. The finite element contact algorithm searches for contact between a master surface and a slave surface. The master surface is defined as a group of contact segments that are formed by one or more finite element groups. The slave surface is defined as a group of contact nodes that are formed by one or more finite element groups. Contacts between elements are found with the help of two algorithms in MADYMO: - The first method finds the contacts by intersections of slave surface nodes through the master surface contact segments. This algorithm is mainly used for contact between rigid surfaces where the surface stiffness is defined in the contact force characteristic. [5] - The second algorithm finds the contacts based on penetration of slave nodes in the contact thickness (gap) of the master surface contact segments. This algorithm is mainly used for deformable FE structures where penetration has to be kept as low as possible. In the contact algorithm, three different phases can be distinguished:

15 TTNO report 14 / 64 - the search phase - the detailed search phase - the force calculation phase In the search phase the algorithm checks each contact node on the slave surface for intersection of the master surface contact segment during the current time step, to make this calculations positions of both, the contact node and contact segment of the current and previous time step are used. In the detailed search phase all contacts that happened before the present time step and new contacts (contacts found in the search phase) are checked to see if the contact node is still penetrating the contact segment. In the force calculation phase the penetrations of the contacts nodes on the contact segment are calculated. The calculation is made as the minimum distance between the node and the segment. [5] When contact thickness (gap) is used, the master surface elements are surrounded on both sides by limit lines placed at a distance defined by the user as shown in figure 2.2. The contact force is generated for the nodes of the slave surface penetrating the gap which, is handled by the contact algorithm as a thickness contact calculation is calculated in the same way as for any other penetration; the only difference is that now contact forces are generated earlier. Two contact force models are available: penalty based contact and elastic characteristic based contact. [5] Figure 2.2 Contact with gap Penalty based contact In the penalty based model, the contact force is calculated by using the bulk modulus of the contact segment (master surface) as follows [5]: F = ψλ (2.3) ( i K / V ) A 2 0 i i

16 TTNO report 15 / 64 where K is the bulk modulus of the contact segments penetrated V is the initial volume of the contact segment 0 A is the area of the contact segment i ψ is the penalty factor λ i is the penetration of the contact node Elastic characteristic contact In this model one surface is considered rigid and the other as deformed ; the hysteresis in the contact forces are stored for the deformed surface. The deformed surface can be either the master or the slave depending on the contact type. These two cases are illustrated in figure 2.3. Figure 2.3 Contact type: (a) slave surface deformed, (b) master surface deformed [5] Choosing master and slave surface When defining a FE_FE contact the choice of the master and the slave surface should be made with care since wrong selection of either one can lead to incorrect results in the calculation. When calculating FE_FE contacts the contact algorithm checks for contacts between a master and a slave surface, the segment of the master surface is taken for which the projection of the contact node lies within that segment. It is highly recommended to choose the contact surface with the most curvature as the slave surface. Otherwise incorrect results are obtained because in a FE_FE contact the penetrations are calculated using the projection of the contact nodes on the master surface. Thus when a curved contact surface is used as master surface the penetration measured with respect to this master surface is incorrect, as it can be seen in figure 2.4. Figure 2.4 Choice of master and slave surface [5]

17 TTNO report 16 / 64 3 Out of Position Situations 3.1 Definition Out-of-position (OOP) is a fairly new area of investigation within the automotive industry. OOP refers to a car driver or occupant that is not in his normal driving or seating position but is in the deployment space of the airbag [22]. When an occupant is in OOP position the airbag can no longer act as a protective system, on the contrary it poses a threat due to the high forces involved in its deployment. In particular belted children tend to be moving for example to play with the radio, placing their selves in a potential OOP situation. 3.2 Background Out-of-position research was originated after the introduction of the airbag in the automobile. Airbags can be very beneficial for car occupants that are involved in a car accident, nevertheless the energy required to inflate an airbag can injure and even cause death to a person situated in OOP. In particular, fatalities and severe injuries were reported especially in low speeds crashes, which would normally not have been fatal in absence of airbags [1]. Figure 3.1 shows results of statistical studies realized by D.S. Zuby, et.al [24], this study was made taking in account 116 fatalities that occurred in the years recorded in the National Automotive Sampling System Crashworthiness Data System (NASS/CDS). Unknown 19% Airbag 16% Other 3% Nondeployment 1% Ejection 13% Interior surface 6% Intrusion 42% Figure 3.1 Causes of death to drivers of airbag equipped vehicles in frontal crashes. [24] The graph shows the distribution of causes of fatality to drivers of vehicles equipped with airbags in frontal crashes. This study indicates that airbag related deaths account for 16% of all the fatalities.

18 TTNO report 17 / OOP Regulations In the last decades, different research centers, car manufacturers and international organizations have carried out studies about airbag aggressiveness. As a result different documents have been published where OOP risk assessment is treated. The most important are summarized in the following paragraphs SAE J1980 The Society of Automotive Engineers (SAE) published in 1990 the SAE J1980 document where guidelines for OOP dynamic and static crash test procedures with different dummy sizes where described [25]. This document was the first written about OOP injury assessment and it was made without having the intention of being a recommendation or a law ISO/TR In 1998 the International Organization for Standardization published the ISO/TR [26] taking in consideration the research previously done by SAE. This was not intended as a standard but as a technical report due to the inexperience on airbag testing and the lack of real world accident data correlation. After analysis ISO indicated the following scenarios in which OOP takes place: - Unbelted driver, passenger and children - Children standing near windshield during travel - Pre impact events (Panic braking, sickness, drowsiness) - Normal driving position close to the steering wheel for small size people. ISO describes two static pre-positioned OOP driver tests and three acceleration-induced OOP dynamic tests. Two dummies are proposed for adult driver simulations the 50 th percentile Hybrid III male and the 5 th percentile Hybrid III small female FMVSS 208 In 2000 the US National Highway Traffic Safety Administration (NHTSA) published a new version of the Federal Motor Safety Standard (FMVSS) Nr. 208 where OOP tests were included. This standard is the only world wide available regulation on OOP assessment and the only one prescribing exact procedures for new vehicle design and testing. All car manufactures that want to sell their products in the USA must comply with this regulation. The FMVSS208 prescribes real crashes with Hybrid III 5 th percentile (small female) and various child dummies. This was decided because it was seen that OOP is a problem that involves mainly small occupants like children or small females. People with short height and/or low weight are more likely to get in OOP because of their small inertia. For example when braking a child is more easily moved towards the airbag module than an adult. Another

19 TTNO report 18 / 64 reason is that a small female is generally more likely to sit closer to the steering wheel than a larger male, thus potentially at higher risk of OOP. Within the FMVSS208 two different positions for static OOP tests with the 5 th %ile Hybrid III dummy are described in part S [27]; one called chin on module and the second one called chin on rim. In the chin on module position the dummy is placed in such a way that its chin is the body part closest to the airbag container. In this test most of the deployment energy from the airbag is directed towards the head of the dummy. In the chin on rim test, the chin of the dummy is placed on the upper edge of the steering wheel, and the chest of the dummy is placed against the airbag module. In this positioning, most of the airbag deployment energy is received by the thorax. Since this is the only recognized regulation by car manufacturers as well as research centers, the procedures contained in FMVSS208 for static OOP tests with the 5 th %ile Hybrid III dummy were adopted for the tests realized in this investigation with the Hybrid III and the HUMOS2 model. 3.4 Injury criteria In OOP situations the airbag deployment energy is received mainly by the upper part of the occupant body, namely: head, neck and thorax. For each of these body parts specified biomechanical responses are used to asses whether injury will occur and its severity. In this way for the head the HIC (Head Injury Criteria) is used, for the neck the Nij (Biomechanical Neck Injury predictor) is used. Finally, for the thorax, accelerations and deflections are measured to assess if injury will occur or not. All the criteria mentioned above have certain thresholds, which serve as a reference to find out if an injury will happen. Each of these criteria will be described in short in the following paragraphs: HIC, Head Injury Criteria The head injury criterion was developed based on the linear acceleration of the skull, impacting a rigid surface. The automotive industry has used HIC with a certain degree of success despite the fact that most impacts involve an angular acceleration component. The HIC is defined by the U.S. government as follows: 2.5 t max 2 1 HIC = T 0 t1 t2 TE R( t) dt ( t2 t1) t2 t (3.1) 1 t1 Where T0 is the starting time of the simulation, TE is the end time of the simulation, R(t) is the resultant head acceleration in g s (measured at the head s centre of gravity) over the time interval T 0 t1 t2 TE, t1 and t 2 are the initial and final times (in s) of the interval during which the HIC attains a maximum value. [5]

20 TTNO report 19 / Nij, Biomechanical Neck Injury predictor This injury predictor is a measure of the injury due to load transfer through the occipital condyles. In this parameter the neck axial force Fz and the flexion/extension moment My measured in the occipital condyles are combined. It is assumed that the condyles bracket joint is oriented in agreement with SAE J221/1. The axial force Fz is obtained from the constraint force in the joint z-direction. As shear force Fx, the component of the constraint force in the joint x direction is used. And as bending moment the constraint moment My about the joint η -axis is used. My is calculated from the following equation: My = My - e * Fx (3.2) Nij is a collective name describing four injury predictors corresponding to four axial forces and bending moments: - tension extension (N TE ) - tension flexion (N TF) - compression extension (N CE ) - compression flexion (N CF ) The Nij is calculated from: Nij = (F z /F zc ) + (M y /M ) (3.3) yc Where F zc and M yc are constants and depend on the dummy and the loading conditions. The sum of all predictors may not exceed a value of one Thorax injury criteria For the thorax different injury criteria are available. Among others, the 3ms (three milliseconds) injury criterion, the Thoracic Trauma Index (TTI), the Viscous Injury Response (VC) and the Combined Thoracic Index (CTI) are used. For this investigation the criterion used was the one stated in the FMVSS 208. The Federal Motor Vehicle Safety Standard 208 states that the compression deflection of the sternum relative to the spine shall not exceed 52 [mm] (2. [in]). And that the resultant acceleration calculated from the thoracic instrumentation shall not exceed 60 g s. These last two parameters where used to assess the models tested in this investigation.

21 TTNO report 20 / 64 4 Literature review on the development and validation of the FE human body model 4.1 Introduction Crash test dummies are used in all standard regulations as the main evaluation tool for restraint systems and injury assessment in car crash safety. The expensive full-scale testing needed and limited biofidelity in different loading conditions are disadvantages of these tools [15]. Virtual testing, that is mathematical computations of crash tests, offers greater flexibility on loading conditions applied, and is nowadays playing an increasing role in vehicle design. Numerical models representing the physical dummies have been developed, validated and used providing significant advances in the automotive safety area, not mentioning the lower costs that the use of numerical models provides with respect to physical experimental testing. Numerical models of the human body offer a great potential in vehicle design, especially for loading conditions for which biofidelic dummies are not available (for example oblique impact). Furthermore, the use of mathematical human body models could provide insight into the injury mechanisms (also on a material level) that cannot be detected with the use of dummies. Mathematical modelling of the real human body potentially offers improved biofidelity compared to crash dummy models and allows the study of aspects like body size, body posture, muscular activity and post fracture response. In addition detailed human body models potentially allow analysis of injury mechanisms on material level. A large number of mathematical models describing specific parts of the human body have been published but only a few models describing the response of the entire human body in impact conditions [7]. In the present investigation the human model developed within the HUMOS2 project was used. Nevertheless, in order to get better insight on the status of the developments of human body modelling other models developed in other research programs will be mentioned in this chapter. The following sections will give a summary of the developments on FE human body modelling as well as information on the validation of these models. 4.2 FE human models As the importance of numerical modelling became clear in crash testing, industrial firms began developing their own models. Toyota developed the THUMS-AM50 (Total HUman model for Safety) [8]. This model represents the 50%ile American adult male in a seating posture as shown in figure 4.1. In figure 4.1 the whole structure of the THUMS model is depicted with some

22 TTNO report 21 / 64 soft tissues removed to expose the skeletal structure. The response of the model has been validated for the following simulations - thoracic frontal impact - thoracic side impact - pelvic side impact and - abdominal frontal impact All these simulations have been validated using cadaver impact tests data. In the case of the thoracic frontal simulation data published by Kroell et al. (1971 and 1974) was used, for thoracic side impact experimental the test conducted by Bouquet et al (1994) was used. For pelvic side impact the experiment reported by Viano (1989) was simulated, and for the abdominal frontal impact the cadaver impact tests conducted by Nusholtz et al (1988) were simulated. Figure 4.1 THUMS model with some soft tissues removed [8]. In later work a model of a small female was also developed and validated for frontal impact using data published by Kroell et al. (1971) called THUMS- AF05. The head, neck, spine and extremities were obtained by scaling down the THUMS-AM50 The Ford Motor company developed a detailed full human body model for the prediction of human thoracic impact responses and injuries [10]. The thoracic model was based on the model originally developed by Wang (1995) with significant improvements to geometry, articulation and internal organs. Individual abdominal organs were modelled representing the liver, spleen, kidneys, abdominal aorta and inferior vena cava, while the remaining abdominal organs were modelled using a single compressible solid. The pelvis was modelled with accurate geometry and material properties with connection made to a detailed leg model as previously developed by Schuster et al. (2000).

23 TTNO report 22 / HUMOS2 model Within the European project HUMOS a FE Human Model was developed. In the follow up project HUMOS2 finite element meshes representative of the th th th 5 female and 50, 95 male percentiles in driving and standing positions were developed with the help of scaling and positioning tools. The basic assumption of the HUMOS programme is that a biofidelic model shall be structurally very close to the real human body [19]. With this assumption a male body (1.73m, 80kg) representative of the mean European man (50 th percentile) was used to develop the numerical model. The model includes bones, ligaments, muscles, thoracic and abdominal organs, skin, etc. The small female (5 th percentile) was used in this investigation. This model results from scaling the geometry and inertial properties of the 50%ile Human model using the scaling tool developed in HUMOS2. Figure 4.2 th th th shows the HUMOS2 5, 50 and 95 percentile human model in seating position. (a) (b) (c) Figure 4.2 HUMOS2 5 th, 50 th th and 95 %ile human model in seating position Mesh update and refinement Within HUMOS2 some meshes where refined in order to obtain successful scaled models. The defects on the meshes were identified during crash simulations done by end-users ending with the following improvements: Head and brain improvements Neck geometry and mesh modification Elbow mesh refinement Elbow ligaments implementation Ribs mesh improvement Aorta insertion Thoracic and abdominal organ mesh improvements Lower limbs bones improvement Lower limb flesh modifications Below figure 4.3 shows some examples of refined, modified and/or improved meshes. Figure 4.3 (a) shows the old neck (left) and the modified neck (right). Figure 4.3 (b) shows the old thorax (left) and the modified thorax (right). In both models is possible to observe a more refined mesh.

24 TTNO report 23 / 64 (a) (b) Figure 4.3 Old and modified meshes, neck (a) and thorax (b) [20] Internal contact definition As mentioned in the previous paragraph, in the HUMOS2 model all the internal organs are modelled and suitable for contact definition. The detailed internal organ modelling, allows the HUMOS2 model to provide information on the magnitude of the forces that the organs are subjected to in a crash event. Contacts between internal organs are defined in MADYMO by the use of force penalty contacts, which minimizes the penetration between the surfaces in contact while accounting for a stable solution. Since all the organs are made of FE meshes, a surface to surface contact method is used. Figure 4.4 shows an upper body sagittal section of the HUMOS2 body model where the internal organs can be seen (brain, lungs, etc). An overview of the HUMOS2 internal contacts is given in Appendix A Figure A.4 Figure 4.4 Sagittal section of the mesh of the upper body [19].

25 TTNO report 24 / Validation of HUMOS2 model The HUMOS2 50 th percentile model (occupant and pedestrian) has been validated on a segment and full body level using volunteer and PMHS test data, to assess its capacity in predicting injuries. Dynamic as well as static tests have been used Rib validation The HUMOS2 rib models behaviour was first checked against static and dynamic tests carried out on isolated human ribs. The corridors published by Lizée et.al of the human thorax [11, 12] were used for the validation. The HUMOS2 ribs are modelled using shell elements for the cortical bone and solid elements for the trabecular bone. Node sharing was assumed for the description of the connection between both tissues. Figure 4.5 below, shows validations results for the HUMOS ribs. a) b) th Figure 4.5 HUMOS2 50 percentile rib subject to: quasi-static loading a), dynamic loading b). [19] Thorax, abdomen and pelvis validation Figure 4.6 shows the results obtained for the thorax as compared to the experimental results published by Kroell et Al. [13, 14]. In these experiments, the thorax was hit by a 23.4 kg cylindrical impactor at 9.9 m/s. Again the red lines represent the corridor and the blue line the results obtained by the HUMOS thorax. Specific validation for rib fractures was performed. The predicted stresses in the ribcage where in accordance with the fractures registered in autopsies. The viscous criterion indicated similar fractures as predicted by the stresses. From these good correlations the model was found god for prediction of severity of ribcage injuries.

26 TTNO report 25 / 64 a) b) th Figure 4.6 HUMOS2 50 percentile thorax results: Deflection vs. time (a), impactor force vs. time (b). [19] Abdomen validation has been carried out using the results published by Cavanaugh et al. (1986) and Viano (1989). Cavanaugh realized 2 frontal tests using PMHS, one at 6.9 [m/s] and another at 9.4 [m/s]; the impactors used weighed 31.4 [kg] and 63.6 [kg] respectively. Viano made 3 oblique impacts using PMHS varying the speed from [m/s], only one impactor was used weighing 23.4 [kg]. For the validation of the pelvis, results published by Bouquet et al. (1994) and Viano (1989), both used PMHS to realize the tests Sled tests The HUMOS2 model was also validated using available sled tests that were carried out at the Institute of Legal Medicine in Heidelberg [23]. The following sled tests were used. - 30kph, medium sled deceleration 20G - 40kph, medium sled deceleration 20G - 50kph, medium sled deceleration 20G - 60kph, medium sled deceleration 15G Figure 4.7 shows the validation setup. Figure 4.7 Frontal sled test set-up with HUMOS2 50 th percentile. The HUMOS2 model is already included in the MADYMO human models manual (version 6.2.2) in its 50 th percentile FE occupant model [21]. The model has been validated and is now available for commercial use.

27 TTNO report 26 / th percentile HUMOS2 model The HUMOS2 5 th percentile model is a scaled version of the 50 th percentile model, from which all the mass and inertial properties have been scaled. For this investigation it is assumed that the 5 th percentile model behaves in a realistic way from a biomechanical point of view since it is derived from the validated 50 th percentile model.

28 TTNO report 27 / 64 5 Application of Hybrid III dummy model in the standarized OOP test environment 5.1 Dummy model positioning Positioning of any model whether human or dummy is very important in OOP situations since alignment and offset of the occupant with respect to the airbag model can have a big influence in the magnitude of the injuries that the dummy or human model undergoes due to the airbag deployment [1]. The present chapter describes the steps and the methology used to position the 5 th percentile Hybrid III dummy model in both standarized OOP tests from the FMVSS 208 [4] namely 1nhtsa ( chin on module ) and 2nthsa ( chin on rim ). Within TNO a preliminary investigation about OOP applications has already been done [1] for which MADYMO version was used. Facet models were used for the human and dummy model numerical simulations. In this first investigation, the MADYMO models were validated with respect to experimental tests. This previous investigation was used as basis for the present work, which is developed using MADYMO version The Hybrid III 5 th percentile model used for this investigation is the same as the one used in the previous research, nevertheless it has been translated to be used in MADYMO Though no real positioning of the Hybrid III model was made, other changes were implemented in the previously used simulation model in order to obtain a working application. In particular, due to numerical instabilities originated by the change of MADYMO version, the original airbag used in the previous study was substituted with the generic airbag model available in the MADYMO database. The generic MADYMO model had to be modified in order to comply with the specifications of the original airbag. The comparison between the results obtained in the present work and the results of the previous TNO investigation show a very good correlation, giving confidence that the substitution of the original airbag model with the MADYMO generic database airbag model did not compromise the level of validation of the OOP simulations Environment configuration Thought the dummy positioning was already realized for the facet model in the previous TNO investigation [1], in the present study the environment was changed because it was considered that the previous environment configuration included some parts that were not necessary for an OOP simulation; since no interaction takes place between the occupant and these parts. In the new configuration the following SYSTEM.MODELs are present:

29 TTNO report 28 / Driver 2.- Seat 3.- Dashboard 4.- Windshield 5.- Foot plate 6.- Steering Wheel Figures 5.1 and 5.2 show the configuration adopted in the previous investigation and the configuration adopted in the present work. Figure 5.1 previous environment configurations 1nhtsa, 2nhtsa respectively [1] Figure 5.2 Current environment configurations 1nhtsa, 2nhtsa respectively Airbag implementation The OOP application for both investigated positions used in the previous investigation (figures 5.1 and 5.2) was translated to the newest version of MADYMO (version 6.2.2); unfortunately this resulted in numerical instabilities of the airbag. Many attempts were made to make it stable again without good results. Finally, a decision was taken to replace the airbag with a numerically stable airbag model. The airbag used was the generic airbag model provided in the MADYMO database and was different from the original one in the following aspects:

30 TTNO report 29 / Dimensions 2.- Volume 3.- Mesh refinement 4.- Container 5.- Positioning 6.- Folding method Dimensions and volume The volume of the generic airbag available in the MADYMO database was smaller than the one from the original airbag by almost 50%, (original airbag 62 [lt], generic airbag 25 [lt]). Therefore the mesh of the generic airbag was scaled using the option SCALING option within FE_MODEL. In order to obtain a starting suitable value for the scaling factor the airbag volume was approximated with a sphere. The first factor obtained for scaling was 1.4; this factor was used in X, Y and Z direction. In addition, the numerical parameters that control the inflation process had to be updated. These parameters are shown in table 5.1. Parameter Value before Value after added changed Block_flow x Amb. Temp x Jet shape circular x CDEX x Prop. MEM3 1 3 x Polytropic constant 0 1 x Table 5.1 Parameters changed and added within the airbag model. Where: Block flow; is a factor controlling the gas outflow through those elements of the airbag that are in contact, range: [0, 1]. Amb. Temp; represents the ambient temperature [K] Jet shape circular; one of the two shapes with which the shape of a jet can be defined, range [0, ]. The second possible shape is a rectangular shape. CDEX; Discharge coefficient for the exhaust openings or area scale factor. Prop. MEM3; Linear three node triangular membrane element. Polytropic constant; is the ratio between the constant pressure heat capacity C p and the constant volume heat capacity C v of the inflator gas mixture. By changing parameters like Block Flow and CDEX (both reduced) less gas was lost through the airbag vent holes and pores of the fabric, therefore increasing the inner volume of the airbag. The polytropic constant was added and set to a value of 1, with this the gas coming from the inflator has a

31 TTNO report 30 / 64 constant temperature; thus having a more uniform gas flow into the airbag chamber. By increasing the jet shape the output velocity of the gas was decreased, and at the same time a more similar volume increment was achieved in relation with the original airbag used see figure 5.3. Final values for the scaling parameter were: X= 1.4, Y= 1.65 and Z= The figure below (figure 5.3), show plots of the resulting volume from the generic and original airbag respectively. Figure 5.3 Airbag volume comparison Contact definition For the OOP simulations performed with the Hybrid III model, contacts were defined between the airbag and the head, neck, thorax, arms, upper legs and lower legs. Furthermore contacts between the seat and the Hybrid III were also defined. In addition to these contacts, contacts between the airbag and the airbag container were defined as well as airbag inner contacts. The airbag inner contacts or self contacts are contacts occurring between various parts of the airbag during deployment. Having this contact defined helps to simulate a more realistic deployment of the airbag avoiding compenetrations in the airbag fabric. The actual contact definition used in the Hybrid III simulations for both positions is shown in Figure A3 in appendix A. Two different contact methods where used: NODE_TO_SURFACE and SURFACE_TO_SURFACE. Most contacts were defined as NODE_TO_SURFACE since most contacts were defined between the airbag (which is a FE feature) and the Hybrid III model (modelled as facet surface); additionally the characteristic of the Hybrid III was used to

32 TTNO report 31 / 64 calculate the contact forces between the airbag and the Hybrid III. For the contacts between the Hybrid III and the seat the characteristic from the seat was used. Furthermore, the RELEDG option was also used, so that edge contacts between elements were ignored. The SYMMETRIC option was also used in contacts between the airbag and the Hybrid III to change the contact method to SURFACE_TO_SURFACE. The SURFACE_TO_SURFACE method as such, was used just once to define the inner contacts in the airbag; additionally a gap function is defined and contact force penalty is used to calculate the contact forces Folding method The generic MADYMO database airbag adopts the IMM.METHOD1. With this method the airbag is folded. This method is normally used in OOP simulation since accurate folding of the airbag is very important for an accurate simulation of the deployment. This is not possible in the case when IMM.METHOD2 is used because this method adopts a scaled initial configuration Airbag container implementation The airbag container, besides containing the airbag also guides its deployment. The container was modelled as a rigid FE model. Also the mesh of the container had to be scaled. The scaling parameters for the container are the following: X= 1.4, Y=1.6 and Z=1.6. Contacts between the airbag and the container were defined using surface to surface contact method where force penalty is used. 5.2 Simulation results Among all the different contacts happening in an OOP crash situation, two major loading phases can be recognized. The first loading phase occurs when the airbag contacts the occupant, the second loading phase occurs when the occupant pushed backwards by the airbag gets into contact with the seatback (also called rebound phase ). The two FMVSS208 tests were simulated for 60 [ms] because the present study was focussed on the first loading phase (occupant interaction with the airbag) which is the cause of the most severe injuries and is the phase that current regulation focuses on. In the following paragraphs kinematics of the model as well as the measured biomechanical responses will be described nhtsa Simulation ( chin on module position) Figures 7.2, 7.4 and 7.5 show the output signals for these simulations, they are included in chapter 7 where they are compared also against the results obtained from the human model simulations.

33 TTNO report 32 / Kinematics Figure 5.4 shows the kinematics of the chin on module simulation performed with Hybrid III 5 th percentile at various times. The interaction of the unfolding airbag and the occupant is initially concentrated in the upper thorax, neck and chin area (10 [ms]). Extensive contact with the dummy shoulders occurs after 20 [ms]. Contact with the arms occurs after 30 [ms]. At 40 [ms] the dummy starts to move towards the seatback. Figure 5.4 Hybrid III 5 th percentile dummy in FMVSS208 chin on module position (1nhtsa) Peak values Peak values obtained for the chin on module simulation are shown in table 5.2. Table 5.3 is shown only for comparison with the results obtained in previous investigations realized at TNO. All of the values here described are related to the airbag contacts. For the head the maximum resultant acceleration represents the contact between the airbag and the dummy s head. For the neck tension and compression force, the force was mostly applied in tension. The neck maximum shear force is registered when the airbag is reaching its maximum volume. For the neck moments two peak values are registered one with positive value corresponding to the neck s flexion and another with negative value corresponding to the neck s extension.

34 TTNO report 33 / 64 RESPONSE VALUE TIME [ms] UNITS Head Acceleration 32,6 37 [G] Thorax Resultant Acceleration 15,1 38 [G] Chest deflection [mm] Neck tension-compression force [N] Neck shear force [N] Bending moment (My-extension) -33,6 48 [Nm] Bending moment (My-flexion) 14,8 26 [Nm] Table 5.2 Peak values chin on module (1nhtsa). RESPONSE VALUE TIME [ms] UNITS Head Acceleration [G] Thorax Resultant Acceleration [G] Chest deflection [mm] Neck tension-compression force [N] Neck shear force [N] Bending moment (My-extension) [Nm] Bending moment (My-flexion) [Nm] Table 5.3 Peak values chin on module (1nhtsa) values as obtained in simulations made by R. Guidetti, Delft 2003 [1] nhtsa Simulation ( chin on rim position) Figures 7.7, 7.8 and 7.9 show the output signals for these simulations, they are included in chapter 7 where they are compared against the results obtained from the human model simulations Kinematics Figure 5.5 shows the kinematic results of the FMVSS208 chin on rim test with the Hybrid II 5 th percentile dummy. The interaction between the airbag and the occupant begins just before 10 [ms] with a concentrated contact of the airbag with the thorax and the right upper arm. At 20 [ms] the neck and upper legs make contact with the airbag while dummy begins to be pushed backwards.

35 TTNO report 34 / 64 Figure 5.5 Hybrid III 5 th percentile dummy in FMVSS 208 chin on rim position (2nhtsa) Peak values Table 5.4 contains the peak values obtained for the chin on rim simulation results. Table 5.5 is shown only for comparison with the results obtained in previous investigations realized at TNO. The head acceleration peak value occurs when the airbag is surrounding the neck completely. From the neck bending moments is possible to see that the neck undergoes flexion and extension. RESPONSE VALUE TIME [ms] UNITS Head Acceleration 30 44,2 [G] Thorax Resultant Acceleration 19,5 37 [G] Chest deflection [mm] Neck tension-compression force [N] Neck shear force [N] Bending moment (My-extension) [Nm] Bending moment (My-flexion) 15,7 60 [Nm] Table 5.4 Peak values chin on rim (2nhtsa). RESPONSE VALUE TIME [ms] UNITS Head Acceleration [G] Thorax Resultant Acceleration [G] Chest deflection [mm] Neck tension-compression force [N] Neck shear force [N] Bending moment (My-extension) [Nm] Bending moment (My-flexion) N.A. N.A. [Nm] Table 5.5 Peak values chin on rim (2nhtsa) values as obtained in simulations made by R. Guidetti, Delft 2003 [1].

36 TTNO report 35 / Conclusions From the comparison of the results obtained in this investigation and the one in the previous TNO investigation [1], it can be seen that the results obtained in this work correlate well with the obtained previously by R. Guidetti, Delft, 2003 [1]. This gives confidence that the substitution of the original airbag model with the MADYMO generic database airbag did not compromise the level of validation of the OOP simulations. As previously mentioned the airbag used in this investigation includes a detailed modelling of the airbag unfolding pattern which was not adopted in the previous investigation. This explains most differences in the results shown in tables

37 TTNO report 36 / 64 th 6 Application of 5 percentile FE Human body model in the standarized OOP test environment. 6.1 Human model positioning As already mentioned in paragraph 5.1 of the previous chapter, positioning of the occupant within the environment is one of the most important steps when trying to recreate a real accident or experimental test with the use of computer simulation tools. The present chapter explains all the steps and the methodology used to position the FE Human body model in both standarized FMVSS 208 tests Positioning tool The positioning of the FE Human model in the chin on module (1nhtsa) and in the chin on rim (2nhtsa) position was achieved by using a new software tool developed within the HUMOS 2 project for positioning of finite element models. The positioning of a dummy model doesn t pose much difficulty since all the body parts of the dummy model behave as rigid bodies. When a rotation or displacement is prescribed in any body part, the body part moves as a whole. When positioning the HUMOS2 model, this approach is not possible because its body parts are made of deformable finite elements. Therefore when a rotation or displacement is prescribed to a joint to re-position any body part, the node and the elements of the FE mesh belonging to the body part must be repositioned thus originating stretched or compressed elements with respect to the original configuration. The positioning tool was used for the firs time in the present work and makes use of MADYMO EXCHANGE environment. This tool provides the user a graphical user interface (GUI) where the model that is to be positioned can be visualized along with any environment. The environment (in this case the car interior) is present only for visualization purposes since no contacts or any other type of interaction between the human and the environment are defined in the positioning tool. The positioning process starts from the default position in which the HUMOS2 model is in seated driving position see figure 6.1 (up). The HUMOS2 model consists of both a FE part and a multi-body part. The FE part is mainly used to represent all the soft tissues like flesh muscles and internal organs. The multi-body part is constituted by rigid bodies (defined with very low masses and inertial properties, so they hardly affect the kinematics of the FE structure) to which the FE skeletal structure is supported. Kinematic joints are defined between these bodies that model the articulation between the skeletal structures [21]. Thanks to the above described MB-FE duality of the HUMOS2 model, is possible to edit the joint s degrees of freedom (rotations, displacements) of the body parts and of the complete human body figure 6.1 (up, down) within

38 TTNO report 37 / 64 the GUI. Once the desired position is specified, a pre-simulation is run where the real position of every node in the whole human body mesh is calculated. After the pre-simulation a model of the human body is obtained, containing the new position for the nodes and elements, thus the new positioned model. Figure 6.1 Positioning tool before positioning (up) after positioning (down). By using the positioning tool an updated view of the MB-structure (skeletalstructure) is shown in the graphical user interface (GUI) with every subsequent change to the joint s degrees of freedom. Since the positioning tool was used for the first time in the present study, a few problems were encountered related to the visualization of the car environment which was not displayed properly (some parts where scattered in the space around the human model). The problems were due to the structure of the input file and were solved by adopting a sequential order of the parts in the MADYMO input file. Another problem encountered while using the positioning tool was that after the pre-simulation, in some parts of the body the skin did not properly follow

39 TTNO report 38 / 64 the positioning of the bones, resulting in initial penetrations between the skin (pink) and the bones (violet), as it can be seen in the tibia area in figure 6.2. This originated initial contact forces between the skin and the bones which caused numerical instabilities in the positioned model. This problem was solved by using the MADYMO INITIAL_PEN_TRACK option, which ignores all forces generated by initial penetrations of the contacts. Figure 6.2 bone uncovered on HUMOS2 leg after positioning Positioning procedure The following steps were realized to achieve the desired position for the chin on module and chin on rim position: As a first step, the initial orientation of the whole human body with respect to the reference space was given; that was done by editing the original position of the human joint. This corresponds to a rigid body motion (translation and rotation) of the whole human body with respect to the reference space. The human joint (also known as H- point) is a joint in which the position and orientation of the complete human model can be changed with respect to the reference space. After that, the positions of the various joints from the different body parts were specified (shoulders, hips, elbows, etc). All joints (excepting the human joint) are positioned by means of a positioning method called Rotation angle method. In the rotation angle method the final orientation of the local joint coordinate system is the results of up to three successive rotations, in tables 6.1 and 6.2 most joints have an R beside it, and this is to indicate rotation. A more detailed description of this method can be found in [5]. Finally, a pre-simulation is made in MADYMO Exchange from which the final position of the human model is obtained. As a result from the simulation an xml file is obtained that contains the new positioned mesh. This file is now ready to be used in the simulations together with the car environment.

40 TTNO report 39 / 64 The final values for the joints positioning in the chin on module positions are shown in table 6.1. Figure 6.3 shows the final OOP set up with the airbag included for the chin on module (1nhtsa) position. Joint Value Joint Value Joint Value Human X 1483,566 Right Hip R2-63,3 Left Elbow R3-78 Human Y -409,256 Right Hip R3-165 Right Elbow R3 78 Human Z 275,2735 Left Knee R2 53 sacrum L5 R2 0 Human R1 0 Right Knee R2 56,6 L5-L4 R2 0 Human R2-18 Left Shoulder R1-80 L4-L3 R2 0 Human R3 181,1 Left Shoulder R2-7 L3-L2 R2 0 Left Hip R1 160 Left Shoulder R3-27,4 L2-L1 R2 0 Left Hip R2-65,3 Right Shoulder R1 78 Left Ankle R2-90 Left Hip R3 165 Right Shoulder R2-7 Right Ankle R2-90 Right Hip R1-160 Right Shoulder R3 27,4 Note: R stands for rotation. X.Y,Z, are displacements. Values are in degrees and milimiters respectively Table 6.1. Final joint positioning for position chin on module (1nhtsa). Figure 6.3. Positioned human model with car interior, position chin on module (1nhtsa). Table 6.2 shows the different values obtained for the positioning of the human model in the chin on rim position. Figure 6.4 shows the visualization of the final position of the human body with its environment included in the chin on rim position.

41 TTNO report 40 / 64 Joint Value Joint Value Joint Value Human X 1455,534 Right Hip R2-75,3 Left Elbow R3-81,9 Human Y -401,709 Right Hip R3-148 Right Elbow R3 81,9 Human Z 370,4191 Left Knee R2 43,9 sacrum L5 R2 0 Human R1 0 Right Knee R2 43,9 L5-L4 R2 0 Human R2-20 Left Shoulder R1-80 L4-L3 R2 0 Human R3 181,5 Left Shoulder R2-7 L3-L2 R2 0 Left Hip R1 147 Left Shoulder R3-18,4 L2-L1 R2 0 Left Hip R2-75,3 Right Shoulder R1 80 Left Ankle R2-90 Left Hip R3 148 Right Shoulder R2-7 Right Ankle R2-90 Right Hip R1-147 Right Shoulder R3 18,4 Note: R stands for rotation. X.Y,Z, are displacements. Values are in degrees and milimiters respectively Table 6.2. Final joint positioning for position chin on rim (2nhtsa). Figure 6.4. Positioned human model with car interior, position chin on rim (2nhtsa). 6.2 OOP simulation set-up Once the model is positioned in both OOP positions, the boundary conditions of the test can have to be defined. For both positions a similar procedure was followed, because the human model positioning is in principle the only difference, both input files are quite similar Contact Definition Once the groups are defined, it is possible to define and edit the contacts. For this application, in both OOP positions FE_FE contacts were defined as well as MB_ FE contacts. In figure A1 and A2 in Appendix A, a table with the

42 TTNO report 41 / 64 actual contact definitions are given. Contacts between the airbag and the human model s most important body parts for this research were defined, e.g. Airbag and head, neck, arm (L, R), thorax, pelvis, etc. The SURFACE_TO_SURFACE contact method was used. This method was also used in the human model FE_FE contact definition as well as in the main input file. Additionally a gap function (see 2.7.3) was used along with the definition of the CONTACT_FORCE_PENALTY element. 6.3 Simulation results In this paragraph the results of the FMVSS 208 simulations with HUMOS2 5 th percentile are presented in terms of kinematics and biomechanical responses. The most significant biomechanical responses for the specific loading condition (static frontal OOP impact) are chosen and analysed per body part FMVSS 208 chin on module position Kinematic response Figure 6.5 shows the kinematics of the chin on module simulation at different timing. After 10 ms the airbag is unfolding and in contact with the model s thorax, neck and chin. At 20 ms the airbag is in contact with the shoulders, neck and thorax while the human begins to be pushed backwards. At 60 ms the airbag is completely inflated and in contact with the lower arms while the human is moving towards the seatback. Figure 6.5 Kinematics of the HUMOS2 5 th percentile simulation in FMVSS208 chin on module position Head response Figure 6.6 shows the head c.o.g. resultant acceleration of the HUMOS2 5 th percentile model in FMVSS208 chin on module position. The airbag is

43 TTNO report 42 / 64 triggered at 5 [ms]. The effect of the airbag contact on the human head is seen at 7 [ms]. The resultant acceleration peak value of 57.5 [G] is reached at 35 [ms]. The unloading phase is followed after 52 [ms] by an increment in the acceleration due to the airbag s inflation which ends at 53 [ms]. Figure 6.6 Head C.o.g. resultant acceleration HUMOS2 5 th percentile FMVSS208 chin on module position Neck response Figures 6.7 to 6.9 show the human neck responses as measured in the upper neck load cell. Figure 6.7 shows the Y-moment on the upper neck load cell. The contact between the airbag and the human neck causes flexion on the upper neck until 20 [ms]. After 20 [ms] extension is present due to direct action of the inflation on the chin of the human. The peak value of 12.5 [Nm] is reached at 31 [ms]. In the last 10 [ms] the neck is again subjected to flexion (20 [Nm] at 60 [ms]): the airbag contact is mainly acting at the thorax level while the human chin is unloaded. The shear force loading on the neck is shown in figure 6.8. The maximum value of 900 [N] is reached at 33 [ms]. The neck is subjected to negative shear through the first 45 [ms]. The neck is subject to tension force with peak value of 2380 N at 33 [ms] as shown in figure 6.9.

44 TTNO report 43 / 64 Figure 6.7: Upper neck flexion-extension torque (MY) of the HUMOS2 5th percentile in FMVSS208 chin on module position Figure 6.8: Upper neck shear force (FX) of the HUMOS2 5th percentile in FMVSS208 chin on module position Figure 6.9: Upper neck tension force (FZ) of the HUMOS2 5th percentile in FMVSS208 chin on module position.

45 TTNO report 44 / Thorax response Figure 6.10 shows the thorax deflection for the chin on module position. The chest deflection is measured as the relative displacement in the global x- direction between a node on the skin in the thorax and a node on the skin in the back of the human located at the 4 th rib level and at the 8 th rib level. Due to the method used to measure the thorax deflection [1], a wave effect caused by the airbag deployment is observed in the skin where the deflection is measured. Due to this behaviour a negative deflection is present from 5 to 15 [ms] in the upper and lower part of the thorax 4 th rib level and 8 th rib level respectively. At 8 [ms] for the 4 th rib and 14 [ms] rib for the 8 th rib, direct contact with the airbag occurs, showing a compression that reaches a peak of 0.031[m] and 0.036[m] at 32 [ms] and 36[ms] for the 4 th and 8 th rib respectively. After 43 [ms] a negative deflection is present in rib 4 caused by the rigid body rotation that the thorax experiences. Figure 6.10 Thorax deflection for the HUMOS2 in FMVSS 208 1nhtsa ( chin on module position). In figure 6.11 the Von Mises stress distribution for the chin on module position is shown. A peak stress value of 8.32 e07 [Pa] is found at 32 [ms] which is caused by the folding pattern of the airbag. Furthermore the nodes where the maximum and the minimum Von Mises stress values take place are indicated. Stress concentration can be seen the on the sternum area and on the back side of the ribs.

46 TTNO report 45 / 64 Figure 6.11: Thorax stress distribution for the HUMOS2 5 th percentile in FMVSS208 chin on module position FMVSS 208 chin on rim position Kinematic response Figure 6.12 shows the kinematics of the HUMOS2 5 th percentile simulation of FMVSS208 chin on rim position at different timing. The airbag trigger time is 5 [ms]. The kinematics at 10 [ms] show a concentrated airbag loading on the thorax. At 20 [ms] the airbag contacts the neck, left and right upper arms and the human starts to be pushed backwards. Contact with the upper and lower arms starts at 30 [ms]. At 55 ms the airbag is fully inflated and the human is being pushed towards the seatback. Figure 6.12 Kinematics of the HUMOS2 5 th percentile simulation in FMVSS208 chin on rim position.

47 TTNO report 46 / Head response th Figure 6.13 shows the HUMOS2 5 percentile head resultant acceleration in the FMVSS208 chin on rim position. The airbag is triggered at 5 [ms]. The effect of the airbag contact on the human head is seen after 10 [ms]. The resultant acceleration peak value of 42 [G] is reached at 37 [ms]. Figure 6.13: Head C.o.g. resultant acceleration HUMOS2 5 th percentile FMVSS208 "chin on rim" position Neck response In figures 6.14 to 6.16 show the human neck responses as measured in the upper neck load cell. Figure 6.14 shows the Y-moment on the upper neck load cell. From 15 to 20 [ms] neck flexion is present due to the airbag force on the thorax while the human head is not yet significantly interacting with the airbag. This also corresponds to negative shear in the neck from 15 to 20 [ms]. From 20 [ms] until the end of the simulation extension due to the interaction of the airbag with the human chin is present reaching a maximum of 22.5 [Nm] at 40 [ms]. The shear force loading on the neck is shown in figure The maximum value of 1112 [N] is reached at 36 [ms]. The neck is subject to tension force with peak value of 1490 N at 34 [ms] as shown in figure 6.16.

48 TTNO report 47 / 64 Figure 6.14: Upper neck flexion-extension torque (MY) of the HUMOS2 5 th percentile in FMVSS208 chin on rim position Figure 6.15: Upper neck shear force (FX) of the HUMOS2 5 th percentile in FMVSS208 chin on rim position Figure 6.16: Upper neck tension force (FZ) of the HUMOS2 5 th percentile in FMVSS208 chin on rim position.

49 TTNO report 48 / Thorax response Figure 6.17 shows the thorax deflection at the 4 th and the 8 th rib level. The method used to measure the deflection is the same as in point At 7 [ms] rib 8 makes contact with the airbag, for rib 4 this happens at 9 [ms]. The chest deflection is measured as the relative displacement in the global x- direction between a node on the skin in the thorax and a node on the skin in the back of the human located at the 4 th rib level and at the 8 th rib level. As expected for this position, rib 8 has a higher peak deflection by almost 100%: this is caused by the fact that in this position the airbag deployment is directed to the lower thorax. At 47 [ms] rib 4 th undergoes negative deflection in the rebound phase caused by the rigid body rotation that the thorax experiences under the action of the airbag deployment. Figure 6.17 Thorax deflection for the HUMOS2 in FMVSS 208 2nhtsa ( chin on rim position). In figure 6.18 the thorax Von Mises stress distribution for the chin on rim position is shown. A higher stress value is found in comparison with the chin on module position, since in this position the deployment of the airbag mainly interests the thoracic area. Thus more deployment energy is transferred from the airbag to the thorax. The peak value of 1.88 e08 [Pa] is reached at 43 [ms] which is caused by the folding pattern of the airbag. Nodes where the maximum and the minimum stress take place are also indicated.

50 TTNO report 49 / 64 Figure 6.18: Thorax stress distribution for the HUMOS2 5 th percentile in FMVSS208 chin on rim position.

51 TTNO report 50 / 64 7 Comparison of FMVSS208 simulations with HUMOS2 and HYBRIDIII 5 TH percentile dummy. The comparison of the results obtained with the HUMOS2 5 th percentile model and the Hybrid III 5 th percentile model in the FMVSS208 chin on module and chin on rim positions was performed in terms of kinematics and biomechanical responses per body part. The following responses have been evaluated: 1.- Head CoG resultant acceleration 2.- Upper neck-tension-compression force (z-axis) 3.- Upper neck-shear force (x-axis) 4.- Upper neck flexion-extension bending moment (around y-axis) 5.- Chest deflection (x-axis) 7.1 FMVSS 208 chin on module position Kinematic results In figure 7.1 the kinematics of the HUMOS2 model and the Hybrid III model are shown for comparison at different timing. Though in the initial position both models overlay each other perfectly (see also figure 7.3), significant differences are found on the motions of the HUMOS2 model in comparison with the Hybrid III dummy model. At 20 [ms] the thorax, neck and upper arm of the Hybrid III and HUMOS2 model enter in to contact with the airbag. At 40 [ms] the HUMOS2 shows initial movement and flexion in the arm whereas the motion on the Hybrid III is minimal. At 60 [ms] it is evident that the kinematic of the HUMOS2 upper body has higher range of motion then the one form the Hybrid III.

52 TTNO report 51 / 64 Figure 7.1 Kinematics of HUMOS2 5 th percentile and Hybrid III in FMVSS 208 chin on module at different timing (0-60 [ms]).

53 TTNO report 52 / Head results Figure 7.2 shows the comparison of the head resultant acceleration of the Hybrid III 5 th percentile dummy and the HUMOS2 5 th percentile model. Figure 7.2 Head resultant acceleration Hybrid III-HUMOS2 comparison in FMVSS 208 chin on module position. Similar behaviour of the head resultant acceleration is observed. The contact with the airbag occurs at the same time for both the Hybrid III and the human. The maximum human head acceleration is 70% higher than the dummy head acceleration. Figure 7.3 shows the initial position of the HUMOS2 model and Hybrid III model (superposed) in the OOP environment. The two models have the same head position with respect to the steering wheel. Figure 7.3 Hybrid III-HUMOS2 head initial position comparison in FMVSS 208 chin on module.

54 TTNO report 53 / Neck results In figure 7.4 the measured responses in the upper neck load cell in both Hybrid III and HUMOS2 model are shown. All three plots show similar behaviour with different magnitudes. All the dummy neck responses are delayed with respect to the human response, which indicates a difference in stiffness of the two models. With respect to the neck tension force, the human model shows a peak value 65% higher than the dummy, furthermore the human response shows the rebound phase which does not occur yet for the dummy. For the shear force similar behaviour is observed with delayed peak values of the dummy with respect to the human. The human response reaches the highest peak values, being 30% and 50% higher than the dummy in the negative and positive shear respectively. In the Y bending moments the dummy has the highest peak values for the positive (flexion) and negative part (extension) of the plot, having almost 2 times higher extension and more than 5 times higher flexion.

55 TTNO report 54 / 64 Figure 7.4 Neck responses Hybrid III-HUMOS2 comparison in FMVSS 208 chin on module position.

56 TTNO report 55 / Thorax results Figure 7.5 shows the deflection results for the HUMOS2 and Hybrid III model in the "chin on module position for comparison. Overall dummy and human model responses are similar; nevertheless time delay and different peak values for dummy and human model are present in the response. The HUMOS2 peak deflections are higher by 50% in comparison with the Hybrid III. Due to differences in the way the deflection is measured in both models negative deflection observed in the HUMOS2 from 5 to 15 [ms] is not present in present in the Hybrid III. Figure 7.5 Thorax responses Hybrid III-HUMOS2 comparison in FMVSS 208 1nhtsa ( chin on module position). 7.2 FMVSS 208 chin on rim position Kinematic results Figure 7.6 shows the kinematics of the HUMOS2 and Hybrid III model for comparison at different timing. Significant differences are observed in the motion of the HUMOS2 model in comparison with the Hybrid III dummy model. At 20 [ms] the partially unfolded airbag contacts the neck, thorax and upper arm of the dummy and the HUMOS2 model. At 40 [ms] the HUMOS2 arm and upper body s motion is visible, furthermore the neck is in tension due to relative motion of the upper body with respect to the neck, while the Hybrid III motion is relatively small with respect to the human model. At 60 [ms] airbag is in contact with upper leg of dummy and the human. The dummy neck undergoes visible flexion. Both models are in motion towards the backseat.

57 TTNO report 56 / 64 Figure 7.6 Kinematics of HUMOS2 5 th percentile and Hybrid III in FMVSS 208 chin on module at different timing (0-60 [ms]).

58 TTNO report 57 / Head results Head acceleration results for HUMOS2 model and Hybrid III model are shown below in figure 7.7. Both curves follow similar paths though with different timing and magnitudes. The airbag contact starts first on the Hybrid III due to differences in the initial position (see figure 7.6). The human s peak acceleration is 40 % higher than the dummy head acceleration registered at around 36 [ms] and 44 [ms] respectively. Figure 7.7 Head resultant acceleration Hybrid III-HUMOS2 comparison in FMVSS 208 chin on rim position Neck results The neck responses as measured by the upper neck loads cells are shown in figure 7.8. Similar behaviour is observed in all responses with a delay of the dummy response with respect to the human model. The neck tension forces have comparable peak values. The human has the highest tension peak value with 1490 N at 34 [ms]. For the shear force the human model has a negative peak force 30% higher than the dummy. The Y-bending moment responses show similar magnitudes and timing.

59 TTNO report 58 / 64 Figure 7.8 Neck responses Hybrid III-HUMOS2 comparison in FMVSS 208 chin on rim position Thorax results Figure 7.9 shows the deflection results for the HUMOS2 and Hybrid III model in the chin on rim position for comparison. A similar behaviour with delay and magnitude differences for dummy and human model is shown. In the HUMOS2 8 th rib a 174% higher deflection is found. Peak

60 TTNO report 59 / 64 deflection values occur at 48 and 31 [ms] for the Hybrid III and HUMOS2 model respectively. Figure 7.9 Thorax deflexion Hybrid III-HUMOS2 comparison in FMVSS 208 2nhtsa ( chin on rim position) Hybrid III HUMOS2 injury criteria results for chin on module and chin on rim OOP positions. Table 7.1 shows the different values obtained for the injury criteria in the simulations. Results for the human model as well as the ones for the Hybrid III are presented for both positions mentioned in the FMVSS208. Only the HIC15 is presented since is the one required by the regulation, the limit values are also presented for comparison. For the HIC and tensioncompression forces, higher values are obtained for the human than for the dummy. Since no compression forces are present in the dummy during the simulation, no compression-tension/flexion values are registered, whereas for the human values of 0.1 are found. For most of the injury criteria here presented the maximum values are not exceeded. Only the HUMOS2 model exceeds once the maximum deflection in the 2nhtsa position by 50%, and once the maximum peak tension in 1nhtsa by 15%. INJURY CRITERIA DRIVER chin on module (1nhtsa) chin on module (2nhtsa) Hybrid III Humos2 Hybrid III Humos2 Max values Head injury criteria HIC Nij Neck injury criteria Nte Ntf Nce Ncf Peak tension [N] Peak compression [N] << << 2520 Chest deflection 4th rib (Humos2) [m] th rib (Humos2) [m] x axis [m] Table 7.1 Injury criteria results for HUMOS2 and Hybrid III model in 1nhtsa and 2nhtsa OOP positions.

61 TTNO report 60 / 64 8 Discussion 8.1 HUMOS2 OOP simulations The HUMOS2 model was tested in the two positions of the FMVSS 208 standard: 1nhtsa ( chin on module ) and 2nhtsa ( chin on rim ). A significant difference in the response of the model was observed, showing the influence of the initial position on the model response. For the head acceleration higher values were obtained in the chin on module position where the head is closer to the airbag at deployment with respect to the chin on rim position. Higher neck shear forces and neck extension bending moments are found in the chin on rim position with respect to the chin on module, due to the fact that in the chin on rim the airbag deployment is concentrated in the thorax area, thus more energy is transferred from the airbag to the thorax and to the neck while the head is not yet interacting with the airbag. This phenomenon was also observed in the chest deflections, where the chin on rim position showed higher values in comparison with the chin on module and in the Von Mises stress distribution, which showed a higher and more localized value of the stress in the chin on rim position than in the chin on module. 8.2 Comparison with dummy model The Hybrid III and HUMOS2 models were tested and compared in chin on module and chin on rim positions of the FMVSS 208 regulation. The comparison of the kinematic results shows a larger motion of the HUMOS2 model with respect to the Hybrid III model. However, a similar trend is observed in all the responses for the HUMOS2 and Hybrid III models. Nevertheless, significant difference in magnitude of the responses and in the timing was seen between the two models. In general, the HUMOS2 model undergoes higher levels of accelerations with respect to the Hybrid III model. The upper neck Y-bending moments in the HUMOS2 model are lower with respect to the Hybrid III model, whereas the upper neck tension and shear forces are higher in the HUMOS2 model than in the Hybrid III model. With respect to the thorax deflections, the results show bigger deflection values for the HUMOS2 model than for the Hybrid III model in both investigated positions. The observed different behaviour is mainly related to the differences between the two models in terms of geometry, inertial properties and material characteristics (stiffness and damping). With reference to the geometrical properties, as shown in figure 8.1, below, the HUMOS2 5 th percentile model has wider shoulder and upper body than the Hybrid III, thus offering a bigger contact surface. Specifically, since the

62 TTNO report 61 / 64 Hybrid III upper body is smaller than the HUMOS2, the airbag deployment energy is concentrated in a smaller area, causing the airbag to expand laterally, which does not happen in the HUMOS2 simulation; consequently less energy from the deployment of the airbag is transferred to the Hybrid III upper body. This explains the fact that the responses of the Hybrid III occur with a delay with respect to the HUMOS2 and have lower peak values, since a higher deployment energy absorbed is translated in higher kinetic energy and higher loads for the HUMOS2, and less deployment energy absorbed is translated in lower kinetic energy and lower loads for the Hybrid III. Figure 8.1 Superimposed Hybrid III (blue) and HUMOS2 (red) 5th percentile models. With respect to the differences in inertial properties, it is observed that they have a significant influence on the head and neck behaviour. The mass of the human head is 4.7 [kg] and the Iyy= [kg*m^2] whereas the mass of the dummy head is 3.5 [kg] and the Iyy=0.015 [Kg*m^2], therefore the human neck undergoes higher inertial loading than the dummy neck. The different responses observed for the Hybrid III and the HUMOS2 model are of course also caused by the different material characteristics of the two models, which represent two different physical objects: a crash test dummy and a human being. 8.3 Conclusions In this study the response of the HUMOS2 small female model was investigated in FMVSS 208 OOP positions. In addition, the response was compared with the responses of Hybrid III 5 th percentile model in the same test conditions. The most important conclusions of this study are: - It was possible to use the HUMOS2 5 th percentile model in Out-Of- Position assessment. The HUMOS2 model showed the same

63 TTNO report 62 / 64 differences between the two different OOP conditions as the Hybrid III model. However, significant differences in the responses were observed between the Hybrid III model and the HUMOS2 model among other due to differences in geometry and inertial properties. - The HUMOS2 model was able to show stress distributions in the ribcage which were consistent with the load distribution on the thorax in the OOP applications. As a consequence the HUMOS2 model could be a useful tool in the design phase of restraint systems. 8.4 Recommendations From the results obtained in the simulations the following recommendations can be made for the HUMOS2 model: Further insight and validation of the HUMOS2 model can be achieved by comparing the simulation results to experimental human data. The material properties of the soft tissues should be investigated. The biofidelity of the soft tissue material models and properties should be improved. With the choice of the material models also the robustness should be taken into account. Geometrical and inertial properties of the HUMOS2 head should be checked with literature. Also the geometry of the HUMOS2 5 th percentile model should be made more female like. Additional contacts between internal organs should be defined where missing.

64 TTNO report 63 / 64 9 References [ 1 ] R. Guidetti, Hybrid III Model to Human Model Comparison in Static Out-of-Position (OOP) Simulations. TNO Report 03.RASID3705.1/RME, Delft, [ 2 ] Agnieszka Zmijewska, Validation of a Finite Element Human Model for Prediction of Rib Fractures TNO Report, Delft, [ 3 ] Response corridors of human surrogates in lateral impacts M.R. Maltese et al Stapp Car Crash Journal, Vol 46 (November 2002), pp [ 4 ] National Highway Traffic Safety Administration, Federal Motor Vehicle Safety Standard No. 208: Occupant crash protection. NTHSA US Regulations 49 CFR part 572, [ 5 ] TNO automotive, MADYMO Theory manual, version MADYMO manual release 6.2, Delft [ 6 ] J.S.H.M. Wismans, E.G. Janssen, M. Beusenberg, W.P. Koppens, R. Happee, P.H.M. Bovendeerd, Injury biomechanics (4J610). College dictaat Technische Universiteit Eindhoven. [ 7 ] Paul Du Bois, Clifford C. Chou, Bahig B. Fileta, Tawfik B. Khalil, Albert I. King, Hikmat F. Mahmood, Harold J. Mertz, Jac Wismans, Vehicle crashworthiness and occupant protection Automotive applications committee, AISI, [ 8 ] Oshita, F., Omori, K., Nakahira, Y., Miki, K. (2002) Development of a finite element model of the human body Proc.7 th International LS DYNA Users Conference, Detroit. [ 9 ] Jay (Zhijijan) Zhao, Gopal Narwani, Development of a human body finite element model for restraint system R&D applications, Takata Automotive Systems Laboratory, Inc. Paper No [ 10 ] Patrick A. Forbes, Development of a Human Body Model for the analysis of Side Impact Automotive Thoracic Trauma, University of Waterloo, Ontario, Canada, [ 11 ] Lizée, E., Robin, S., Besnault, S. et Al., Development of a 3D finite element model of the human body, proceedings of the 42nd Stapp Car Crash Conference, 1998, SAE Paper N [ 12 ] Lizée E., Song, E., et al., Finite element model of the human thorax validated in frontal, oblique and lateral impacts: a tool to evaluate new restraint systems, proceedings of the International IRCOBI conference, [ 13] Kroell C. K., Schneider D. C. and Nahum M., Impact Tolerance and response of the human thorax, SAE Paper N , Proceedings of the 15th Stapp Car Crash Conference, 1971 [ 14] Kroell C. K., Schneider D. C. and Nahum M., Impact Tolerance and response of the human thorax II, SAE Paper N , Proceedings of the 18th Stapp Car Crash Conference, 1974 [ 15 ] PSN Workshop on Human body modelling, March 29 and 30, 2001, Eindhoven University of Technology, Eindhoven, The Netherlands.

65 TTNO report 64 / 64 [ 16 ] Courant, R., Friedrichs, K. and Lewy, H., On the Partial Difference Equations of Matematical Physics., Math. Ann, 100, pp ,1928 [ 17 ] Liang-Wu Cai, Finite Elements, Department of Mechanical Engineering, Kansas State University, Kansas, USA, Spring [ 18 ] Honglu Zhang, Srini Raman, Madana Gopal and Taeyoung Han, Evaluation and comparison of CFD integrated airbag models in LS- DYNA, MADYMO, and PAM-CRASH, Delphy Corporation, 2004 SAE World congress. [ 19 ] HUMOS, Human Model for Safety, Public report, 7PSA/011231/P1/DA, [ 20 ] Jan Pierre Verriest, Philippe Vezin, Development of a set of numerical human models for safety, HUMOS2 consortium. [ 21 ] TNO automotive, MADYMO Human Models Manual, version MADYMO manual release 6.2.2, Delft [ 22 ] M.G.C. Rekveldt, Out-of-position: introductory study. TNO report 01.OR.BV.020.1/MRE, Delft, [ 23 ] Kallieris, D., Riedl, H., Human mechanical properties. Global validation data. Final report. University of Heidelberg, [ 24 ] D.S. Zuby, S.A. Ferguson, M.X. Cammisa, Analysis of driver fatalities in frontal crashes of airbag-equiped vehicles NASS/CDS. Proceedings of airbag Technology conference 2001 (SP-1615), SAE paper No ,2001. [ 25 ] SAE International, Guidelines for evaluating out-of-position vehicle occupant interactions with deploying airbags. SAE Surface vehicle Information Report J1980, [ 26 ] International Organization for Standardisation, Road vehicles test procedures for evaluating out-of-position vehicle occupant interactions with deploying air bags. ISO Technical Report 10982, [ 27 ] National Highway Traffic Safety Administration, Federal Motor Vehicle Safety Standard No. 208: Occupant crash protection. NHTSA US Regulations 49 CFR Part 572, [ 28 ] TNO automotive, MADYMO Model Manual, version MADYMO manual release , Delft [ 29 ] [ 30 ]

66 TTNO report 65 / 64 APPENDIX A CONTACT DEFINITIONS Figure A.1: HUMOS2 Input file contact definition (1nhtsa) Figure A.2: HUMOS2 Input file contact definition (2nhtsa)

67 TTNO report 66 / 64 Figure A.3: HYBRID III Input file contact definition

68 TTNO report 67 / 64 Figure A.4: Contact definition human model (Contact definitions are equal for both positions)