You are on page 1of 9
OCCUPANT SIMULATION MODELS:EXPERIMENT AND PRACTICE PRIYA PRASAD Biomechanics and Advanced Safety CAE Ford Motor Company Dearborn, Michigan, USA Abstract ‘This paper is an overview of the evolution of math modeling as it pertains to cccupantrestraint interactions. Specifically, math models involving lumped mass, finite element, and hybrid methods are examined. ‘The process by which math models of both the dummy and restraints are developed, validated, and integrated into the vehicle design process is also presented. Finally, the progression of modeling impact responses of instrumented dummies to that of modeling actual human responses is motivated by way of example. 1, Introduction With the advent of computers, mathematical modeling of crashes involving vehicles, ‘occupants and pedestrians has been pursued over the last thirty-three years. Whereas early models were one- and two-dimensional in nature, the emergence of multi-body dynamics and non-linear finite element techniques, coupled with the development of supercomputers and workstations, has led to the current availability of three- dimensional rigid body and finite element models of vehicle structures, occupants, and pedestrians. ‘These models have moved on from research tools in the hands of a few ‘analysts to vehicle design tools utilized by many engineers involved in the development ‘of new vehicles. In most automotive companies and their suppliers, these models are being used as necessary tools for design decisions leading to optimization of designs and shortening of product design time. ‘Whereas the ultimate goal of mathematical modeling is to eliminate the testing of prototype hardwares, testing has been a necessary part in the development of ‘mathematical models. Most models require a substantial number of experiments to establish their validity. Comparisons between experimental and model results lead to further improvements in modeling techniques and test methods. Therefore, experimentation and modeling are in an iterative loop which facilitates further development of both technologies. However, with continual model validation exercises, enough confidence in the predictive capability of models will be gained, thereby leading to substantial reduction in hardware testing, The objective of this paper is to briefly review the following: Current ‘occupant/pedestrian simulation models, ther utilization in vehicle design, methods for ‘experimental verification, and their future involvement in the biomechanical field. Bey J.A.C.Ambrsio et al ed) CCrashorthness of Transportation Systems: Srutra Impact and Occupant Protection, 209-219 (©1997 Kluwer Academic Publishers, 210 2. Brief Overview of Occupant Simulation Models Almost thirty-three years have elapsed since McHenry (1) proposed one of the first ‘mathematical simulation models to describe the dynamic response ofa vehicle occupant involved in a collision. Since then, many more sophisticated models have been developed for simulating occupant kinematics in crashes. During the past decades, a reat deal of emphasis has been placed on the use of mathematical models in research and development in the field of automotive safety. A review of several "Gross-motion Simulators" was made by King and Chou (2) in 1975. One of the simulation codes reviewed, namely CAL3D, has gone through extensive use and development. A second code called MADYMO2D/3D has been developed by TNO in Holland in 1979. The basic features of MADYMO2D were described and reviewed by Prasad (3) in a benchmark study (4). Further developments in these gross-motion simulators and others such as SOMLA (5) as well as some one- dimensional and special purpose restraint system models were reported by Prasad and Chou (6, 7). All the above reviews concentrated on mathematical models derived on the basis of rigid body dynamics. Although the SOMLA program had integrated an occupant ‘model with a finite element seat model for light aircraft, use of finite element analysis ‘in occupant simulation was limited in the past due to lack of both software and computational speed. Recent developments in finite element analysis and coupling of finite element analysis codes with rigid body dynamics codes have opened a new era in ‘occupant dynamics simulation. Finite element codes are now available in which the entire occupant or parts of the ‘occupant can be modeled using finite elements, making it possible to model the ‘occupant body deformations due to contact loadings more accurately than with rigid body dynamics codes. Although the majority of studies so far have concentrated on ‘modeling dummies, the use of mathematical models to explore injury mechanisms and injury criteria is also being undertaken. These new developments of finite element dummy models and finite element human injury models have been reviewed extensively by Prasad and Chou (8). The above detailed reviews of simple to very detailed occupant/pedestrian models will not be repeated in this paper. However, some ‘of the more useful developments are worth reporting ‘Model occupant responses are sensitive to restraint system modeling, An early limitation in belt system modeling was that the belts were constrained to pass over several fixed points on the occupant torso. As a result, even in conditions where a belt system would slip off an occupant segment, the models could not predict such belt motions. Additionally, situations where occupant kinematics would dictate complete disengagement of the belt system, e.g. the upper torso belt in a rear impact or far side impact, could not be accurately predicted. ‘These phenomenon can now be modeled by the new finite element belt models and an advanced belt routine incorporated in the ATB (9), as well as in the MADYMO codes. Another useful development in restraint system modeling has been the introduction of finite element airbag models, Whereas earlier airbag models utilized the final shape of the airbag for volume and pressure predictions, and membrane effects were simulated with highly simplified assumptions, the current finite element airbag ‘models can simulate the shape of the bag as it inflates and membrane effects are included in the formulation. Additional advances in bag folding and unfolding during inflation, have led to more realistic simulations. The Euler-Lagrange coupling in MSC- au Dytran (10) has led to much more realistic airbag deployment predictions in terms of fabric leading edge velocity and airbag/occupant contact forces. Application of this technology has been reported by Prasad and Laitur (11). ‘The development of hybrid finite element and rigid body modeting is leading towards better discretization of flexible and deformable elements in the human body. as well as vehicle structures (12). Coupling of finite element and rigid body dynamics has been utilized by Ruan and Prasad (13) to study the effect of various restraint systems ‘on skull stress and brain strains (14, 15) and will be described later in this paper. ‘In summary, the field of occupant simulation is evolving rapidly as a direct result of new simulation tools and faster computers. Implementation of this growth into the design process will now be discussed. 3. Utilization of Models During Vehicle Design Process Occupant responses in crashes are determined by vehicle structural responses, interior ‘geometry, stiffnesses of occupant contact surfaces, and characteristics of the restraint system. In frontal impact, the structural responses of interest are the acceleration time history of the passenger compartment and the deformation of the passenger compartment. In side impacts, the side structure acceleration and velocit important parameters. The interior geometry of the vehicle dictates the initial conditions of the occupant and distances that the occupant can travel within the vehicle without contacting vehicle components. Although the belt system and airbags are the ‘commonly recognized restraint systems, the seat and all surfaces contacted by the ‘occupant complement the restraint system, Therefore, the shape and stiffness of these ‘contact surfaces are also important, Restraint system parameters that affect occupant responses are generally belt stiffness and geometry, airbag shape, vent sizes, and inflator characteristics like gas mass inflow rates and temperature. "All of the above restraint and vehicle interior parameters combine as a system to affect occupant responses ‘The challenge to the vehicle designer is to integrate all design parameters to ‘control occupant responses in various modes of impact. Some of the impact modes are in safety regulations worldwide and are shown in Figure 1. Some regulations and requirements have strong opposing interactions. For example, the roof crush requirement demands stiff upper structures; the head impact requirement demands softer structures. The low speed damageability requirements demand stiffer structures; pedestrian requirements demand softer structures. The United States’ FMVSS 208 passive restraint frontal impact standard require softer front ends; the proposed European offset crash standards require stiffer front ends. In general, safety standards hhave a tendency to increase vehicle mass, but lower weight designs are required by the Corporate Average Fuel Economy Standard in the US. To achieve the right balance of conflicting requirements while meeting occupant protection goals, mathematical models of the occupant and structure are essential tools during the development phases of the vehicle. Early in the design process lumped mass, multi-body dynamic models reported by Chou, et a. (16), Prasad, et al. (1) Low, et al (18), and Midoun et al., (12) are used to evaluate and guide the development of the basic architecture of the vehicle. As the ‘design progresses, sizing of the vehicle components involved in crash are determined by non-linear beam analysis methods proposed by Mahmood and Paluzny (19, 20). At this 212 stage, detailed sub-system finite element models are utilized to further guide the designs. As hardwares of sub-systems become available, dynamic tests are used 10 evaluate the adequacy of the sub-system designs. The structure and occupant models, are updated to reflect the latest level of hardware designs, and further iterations of vehicle designs are performed with the models, until the next level of hardware is available, The transition from lumped mass models to detailed finite element models of the total vehicle is a gradual process involving several levels of modeling and design verification, with the final full-vehicle FE model representing the finished product. Crash / CAE Integration Frontal us NCAP Angular Offset Seasor Bumper Pedestrian [Roof Crush] Side Headform] 4 Occupant US Static Impact | [Responses] »s Diane Pole Rear Fall barrier Carto-car Bumper Figure 1. CAE integration in vehicle crash design 213 4, Experimental Validation Of Occupant Models 4.1 GENERAL PRINCIPLES To validate models experimentally, some general principles followed by the author are ‘worth discussing. A single test validation of any model, structural or occupant, may be misleading, since itis possible that the model parameters may be valid for only that test. ‘As a result, validation should be conducted over various test severities that cover the range of expected severities in which the model may be utilized. This method of validation recognizes that model or test results may be non-linear in nature. In general, tests themselves may not be repeatable. As a result, several tests should be conducted at the same severity. In general sled tests are more repeatable than vehicle crash tests. ‘Therefore, wherever possible, occupant models should be validated against sled tests The reproducibility of test results are also important since the results may vary due to test operator, test site, and dummy/vehicle differences. ‘Therefore, validation of models ‘may require many tests to determine the stability of test results. Assuming that repeatability and reproducibility of model results are not in question, the best that any ‘model can predict is within the known variability of test results ‘The "goodness" of a model in predicting test results can be entirely in the mind of the modeler. An experienced modeler can utilize a model for design purposes if the ‘modeler can account for systematic differences between model and test results Objective criteria for determining the "goodness" of a model do not currently exis, although an attempt to develop such objective criteria was made by a sub-committee of the SAE, This sub-committee was composed of occupant modelers from the automotive industry, code developers, and the academia. A proposed Validation Index ‘was agreed upon in principle, but various issues dealing with the methods for comparing two time varying responses, the total number of required tests, and the region of validation were left unresolved. As a result the Validation Index was not finalized. The proposed Validation Index consisted of various levels of validation of a model, with higher levels implying higher quality of model predictions of a reference test event(s). This proposed Validation Index is shown in Table 1, hoping that the ‘modeling community can further refine this Index. TABLE 1. Summary of proposed validation index levels Chass Characins Tevel 0 No agreement baween predictions of model and "reference event™ Level | Qualitative agreement: 8) Trends of prodiced parameters same ) Kinematics correspond qualitatively ) Contacts between the occupant and vehicle interior are the same in general Level 2 HICand sinarinfcators pected by simon ae within 2% of tose obtained inference event Level Peak values of important occupant responses limited toa relative exor of 20% (20% on vector magnitude, 11.31 deg on vector direction) Level4 Same as level 3 except 5% Level $ Timing of peaks of important vector responses limited to 5% relative difference Level All peaks and valleys in the duration of time-dependent predictions must match the reference ‘event within 10% Level 7 Same as level 6 except 5% Level §__ 1% relative error, poat-by-point, over the durations ofthe reference and predicted events 24 4.2 DUMMY DATA GENERATION As occupant models have developed over the last thirty-five years, some level of ‘experimental validation studies have also followed. Depending on’the geometrical discretization and linkage system describing the occupant, different types of data are required. The most extensive validation studies have been conducted for the test devices currently or previously utilized in crash testing of vehicles, e.g., the Hybrid II and the Hybrid IIT dummies for frontal impacts, and the SID and Eurosid dummies for side impacts, In general, the dummy segments lend themselves to be described as a collection of rigid links.’ The size, shape, and the inertial properties of the dummy linkages are required to describe the dummy. The joints and their elastic, viscous, and friction properties are further required to describe their interactions during relative movements. Commonly accepted test methods to obtain the above data have been developed. The dummy surfaces and their stiffness characteristics are also required to predict the interaction of the dummy with vehicle components and the restraint systems. ‘Standardized test procedures for the determination of the dummy surface stiffness characteristics are not currently available since they may be dependent on the type/mode of crash being simulated. However, with the greater use of finite element techniques, itis expected that advanced foam models of the dummy surfaces will lead to generalized description of the dummy surfaces. An extensive study to develop a dataset for modeling to Hybrid II dummy, previously used in frontal crash testing worldwide, has been reported by Fleck, et al. (21). This report contains the test methods for determining the geometrical, inertial, and stiffness properties of the dummy. Whereas the Hybrid IT dummy has now been superceeded by the Hybrid IIL dummy for frontal crash testing worldwide, the test methods to determine dummy ‘modeling parameters have been used by the Armstrong Air Force Medical Research Laboratory to determine the modeling parameters for the Hybrid III dummy (22). Modeling parameters determined by the AAMRL are now in datasets for the ATB (22). ‘A parallel study has been conducted by TNO jin the Netherlands and has resulted in datasets for the Sth, 50th and the 95th percentile dummies for the MADYMO occupant simulation models (23), Datasets describing the side impact dummies are also available in MADYMO. 43 EXPERIMENTAL VALIDATION ‘The datasets currently available for describing the Hybrid IH dummy have been utilized by various organizations to determine their predictive capability in frontal sled tests. A benchmark test series using a rigid seat and rigid floor pan was conducted by Prasad (24) in 1985 to determine the dynamic responses of the Hybrid IT and the Hybrid ITT dummy in frontal sted tests. Three tests each were conducted at three severity levels Three-point belt restraint systems of known stiffness characteristics were used. Occupant-to-vehicle contacts were limited to seat/pelvis and foot/oe board only. This arrangement dictated that the dummy responses were not influenced by other possible vehicle frontal component contacts. ‘The exact geometry of the belt system and their ‘initial contacts with the dummies were controlled in each test. ‘The results of these tests ‘were released to a sub-committee of the Society of Automotive Engineers for validation exercises of the ATB and the MADYMO occupant simulation models. ‘These exercises have been reported by Obergefell, et al. (25), Wismans and Hermans (26) and Khatua, et al. (27) at the 1988 SAE Congress. ‘The author's own simulations of these tests 2s utilizing the MADYMO software and dummy joint characteristics measured at Ford are seen in Figure 2 and Table 2, and were discussed at the SAE Congress. The simulation results were well within test-to-test variability of important responses like the Head Injury Criteria (HIC), chest acceleration, chest deflection, and belt loads, The overall kinematics of the dummy were reproduced well, as measured by head displacements, ‘The time varying responses were also within test corridors. Further improvements in the dummy datasets have followed as a result of further usage in different impact conditions. Details of the development and experimental validation of the side impact dummy, SID, have been reported by Low and Prasad (29). In this series of validation exercises, repeat tests of the SID dummy against rigid load cells were conducted at three impact velocities - 10, 15, and 20 mph. The efficacy of the dummy model was further investigated in crash tests at three different velocities (17, 18). For model validation purposes, test methods to determine the stiffnesses ofthe side structure and ccupant/side interactions were developed and are reported by Prasad, et al. (17) and Low, et al. (18). Other validation studies have been cited in Reference 8. We ar chest accEL, 6 (CHEST DEFL, mm Seer —= ——_ PEAK SLED ACCEL (avg), @ Figure 2, Test and mode! esl of sled svies TABLE 2. Test and model results (Head) Head Digphcement WRT, Seat Back (a) conga ‘Vertal Peak Sted Accel 5 Tea Woe Tea Model 25237 046-087 00s Dano nama oa 025.028 031 321-326 049 03 034 216 5. Biomehanical Modeling With the available modeling technology, the logical progression is to advance from dummy modeling to modeling of the human itself. Modeling of various parts of the hhuman body has been an essential activity in the field of biomechanics to gain better understanding of human response to impact and associated injury criteria. Rigid body dynamic models, as well as finite element models, have been developed for many years. ‘A review of such biomechanical modeling activities has been presented by King and Chou (2) in 1975. Since 1975, substantial improvements in rigid body dynamic ‘modeling and finite element analysis techniques have been made. These improvements have led to the development of detailed anatomical finite element models of the human body. There is considerable activity worldwide in biomechanical modeling. Human skull and brain models are being developed and may be considered in future rulemaking. The goal is to measure the linear and angular accelerations of the dummy head in crash tests, and to utilize these measurements to excite a human skull/orain model. The model responses such as skull stresses and/or brain stresses and strains, predicted by the model, may be subjected to control. This method may lead to the replacement of commonly-utilized Head Injury Criteria (HIC) or peak head acceleration for injury assessment. Some candidate models are those developed by DiMasi et al., (29) and Ruan, etal. (30). ‘The development of a human chest model is being pursued by Plank and Eppinger 31). The head and chest models can be integrated with the model of the spine (32) and the pelvis (33). There is considerable activity in modeling the long bones of the lower and upper extremities. ‘Therefore, itis envisioned that all the human body anatomical models can be integrated into one model of the total human body. Once developed, scaling techniques can be utilized to generate models of various size occupants. Since “tuning” of these models to known biofielity response corridors are somewhat easier than that of hardware models, itis quite possible that test dummies will be replaced by ‘mathematical human models. These whole body human models can be integrated with vehicle crash models and ‘may lead to “virtual” testing, instead of hardware testing, inthe long term. To get to the stage of "virtual" testing, considerable development in computer technology is required towards faster model building and analysis times. Further theoretical development in ‘biomechanical material models and potentially more advanced element formulations are needed to describe the impact behavior of biological tissues. The integration of the currently-available human anatomical models is in its infancy. These models are being utilized for hypothesis testing to gain better understanding of human responses to impact and injury mechanisms, One such example is discussed below. ‘Scientific interest in head injuries continues due to the high incidence of fatality rates from head impact and the related high treatment costs of these injuries. Head injury mitigation has been a high priority in the designs of current vehicles. Most regulations worldwide utilize the Head Injury Criterion (HIC) as the indicator of potential injuries to the skull and brain. The Head Injury Criterion is based on the ‘Wayne State University Tolerance Curve first proposed by Patrick, etal. (34) in 1965, and later verified by Ono, et al. (35) in 1980. A shortcoming of the HIC identified by ‘many researchers is that it is based on the measurement of only the resultant linear acceleration at the center of gravity of the head. It is well known that angular a7 accelerations can also cause head injuries. As a result, the control of HIC may not be a ‘necessary and sufficient condition to minimize head injuries. To examine the above ‘hypothesis, it was decided to utilize a detailed finite element model of the skull and brain developed by Ruan, et al. (30). As.a first step, the skull/orain model was coupled toa detailed rigid body model of the Hybrid IIT dummy through a mult-segment neck ‘model. Short duration acceleration pulses ranging between 1 to 8 m secs were used to excite the skull. A constant skull tensile stress curve was generated by varying the average acceleration and time duration. It was found that a constant 100 MPa tensile stress curve corresponded to the skull fracture curve suggested by Ono, et al. (35). A ‘constant brain shear strain curve representing 5% shear strain was also generated by model exercises. This curve also was in good agreement with proposed brain concussion tolerance curves of Patrick, etal. 34) and Ono, et al.(35). These modeling studies have been reported by Ruan and Prasad (13,15). ‘The coupled skull/orain and the Hybrid III dummy model was further used to simulate frontal impacts, with no head contact, airbag/head contact and head contacts with a soft, medium stiffness, and semi-rigid steering hub. It was found that in the cases of no head contact, airbag and soft hub contacts, the HIC’s were nearly the same (553-665), angular accelerations were low (2700-3700 rad/S* ) and the maximum shear siress in the brain was below 12 kPa. With the medium stiffness head/hub contact, the ‘maximum shear stress was 15 kPa and was associated with HIC of 1543 and angular acceleration of 19500 rad/S*. The semi-rigid hub contact yielded a maximum shear stress of 22 kPa associated with a HIC of 2359, and an angular acceleration of 15,500 rad/S*, It could be seen that whereas the HIC rated the increasing severity of head contact as evidenced by the shear stress in the brain, the angular acceleration did not. Based on the above simulations of vehicle frontal crashes, it appears that HIC rates the severity of head impact properly and reduction of HIC in these cases should be associated withthe reduction of brain shear stresses, 6. Conclusions ‘Occupant simulation models have been discussed from several standpoints: 1) Their evolution from two to three dimensions, 2) their refinement as a consequence of the availability of finite element methods, computational advances, and extensive validation efforts, 3) their introduction into the automobile design process, and 4) their evolution into the field of biomechanics. 7. References 1. Meblenry, R. R. (1963) Analysis of the Dynamics of Automobile Passenger restnnt Systems, Pro, 7th Stapp Car Crash Conference, SAB, Warrendale, Pennsylvania. 2. ‘King’ A. I. and C. C.Chou (1975) Mathematical Modeling, Simulation and Experimental Testing ‘of Biomechanical Sytem Crash Response, AIAA 11th Annual Meeting and Technical Display, also in: 1976 Journal of Biomechanics, p- 301-317 3. Prasad, P. (1984) Overview of Major Occupant Simulation Models, SAE Paper No. 4055 in SAE -146, Mathematical Simulation of Occupant and Vehicle Kinematics

You might also like