Dummy model validation and its assessment Sebastian Stahlschmidt, Alexander Gromer, Yupeng Huang DYNAmore GmbH, Germany 25 th January 2012 1
Introduction The design of occupant safety systems by using numerical simulations became an essential part of the vehicle development processes. Especially the optimization of safety systems as well as robustness studies of these systems benefit from the progress of the simulation. Hence, the requirements to computational dummy models increased over the past years significantly. Thus the validation process and the validation level is focused very strong. Also a reliable quality rating could ease the assessment of these models. This Presentation should give a overview about the current validation process of the FAT / PDB Dummy models and gives some ideas how different dummy models could be compared in the future. 2
FAT / PDB Dummy Models FAT (German Association for Research on Automobile Technology) Participants of the first dummy modeling project PDB (Partnership for Dummy Technology and Biomechanics) Theses models are developed by DYNAmore. 3
Dummy meshing with all important details. Mass validation is not very complex if the geometry is captured accurate. Only the right density of the single materials has to be used. The right mass distribution and the inertias era then captured well. 4
Material Tests All relevant materials were testes Specimen were taken from new original parts Some vinyl skins specimen were taken from repair kits. 5
Material Tests Rubbers, foams, Vinyl, Nitinol (WorldSID), rib damping material (WorldSID), Neoprene (ES2), Test types: Static tension and compression Dynamic tension and compression (0.001 1/s, 20 1/s,100 1/s, 400 1/s) For rubber-like materials additional compression tests with constrained lateral expansion 6
Component Tests (The tests were performed with varying speeds, masses and angles): Pendulum tests on rib assembly Pendulum tests for neck and lumbar spine Dynamic shear for the lumbar spine Pendulum tests for the abdominal insert Head drop tests Partial thorax impact tests Impact tests for the pelvis Sled tests for the lumbar spine Impact tests for abdomen component 7
Outline of ES2 rib assembly Component Tests: Different masses, such that rib intrusions are: 10, 20, 30, 40, 50 mm Different impactor speeds Different impact locations Different angles With/without damper unit In total 25 different test 8
Outline of ES2 rib assembly Component Tests: 20 mm rib intrusion 30 mm rib intrusion..test..test model..test..test model 40 mm rib intrusion 50 mm rib intrusion..test..test model..test..test model 9
Outline of WorldSID 50% rib assembly Component Tests: Similar tests as for ES-2 More oblique loads Shoulder rib inner band only Spine box 10
Outline of WorldSID 50% first thorax rib assembly Component Tests: 20 mm rib intrusion all tests model 60 mm rib intrusion 11
Outline of BioRID spine assembly Component Tests: Test with different pulses Test with different support of spine Test with and without springs and dampers In total 10 different tests Each tests is performed with 3 different dummies 12
Pendulum Tests on fully assembled Dummy: Test easy to perform Different pendulum masses and velocities Loading of specific areas possible Loading may differ from crash load 13
Pendulum Tests on fully assembled Dummy: Validation of rib extension by using simple pendulum tests 14
Pendulum Tests on fully assembled Dummy: Rib intrusions with low velocity of pendulum..test..test model Rib intrusions with higher velocity of pendulum..test..test model 15
Sled tests on fully assembled Dummy: Loading similar to a crash load Test setup is still simple Investigate interaction of parts 16
Sled tests on fully assembled Dummy: all tests model 17
Sled tests on fully assembled Dummy: all tests model 18
Sled tests on fully assembled Dummy: all tests model 19
Sled tests on fully assembled Dummy: all tests model 20
Methodology of Development: Why component tests? Investigate the behavior of substructures Validate parts with simplified geometry Simple tests Difficult get relevant loading of part Why so much sled tests? Investigate the interaction of parts Possible to get relevant loading comparable to a crash Barriers designed in respect to - different car sizes - different regulations - different vehicle design 21
Methodology of Development: Example for difficulty to find appropriate test Lumbar spine calibration test - Pre-load in tension - Pure bending test - Significant bending angles Lumbar spine during side impact simulation - Pre-load in compression - Combined torsion and bending - Small deformations 22
Methodology of Development: Example for difficulty to find appropriate test 23
Methodology of Development: Example for difficulty to find appropriate test new component test: 24
Methodology of Development: Example for difficulty to find appropriate test new component test: Test ES2 v4.5 ES2 v5.0 25
Methodology of Development: Example for difficulty to find appropriate test new component test: Test ES2 v4.5 ES2 v5.0 26
Methodology of Development: Example for difficulty to find appropriate test new component test: Test ES2 v4.5 ES2 v5.0 27
Methodology of Development: accurate mesh and detailed geometry material tests generate a first model predict the loads by sled test simulation enhance the FE model define appropriate tests 28
Assessment of the validation states The major work of validation is to evaluate hundreds of signals. It must also assess which variant of the model is the best one in many load cases. Number of tests for validation of WSID 50%: - 246 individual test - for one test multiple signals have to be evaluated Component Nec k Arm Ribs Lumba r Spine Iliac wings Configuration Notch 1 Notch 2 Inner Complete half full band assembly Type Sled Pendulu m Pendul um Pendul um Pendulu m Sled Pendul um Pendul um Loads 2 2x1 2x1 3x2 3x2 2 2x2 2x2 (velo x mass) Load 2 1 1 3 5 2 1 1 directions Impact points 1 3 2 1 1 1 2 1 Specimen 2 1 1 3+2 4 2 1 1 Indiv. Tests 8 6 4 90 120 6 8 4 Reruns 2 2 2 2 2 2 2 2 Total tests 16 12 8 180 240 12 16 8 This number of signals is a very high effort to evaluate rating tools can help to handle this number of tests. 29
Method The application of an objective rating tool was the base of this study. There are a few rating tools on the market and even more published in the literature. Each of the existing tools and algorithms has pros and cons. This study used the CORA ( CORrelation & Analysis ) approach. [Gehre, C. et al; Objective rating of signals using test and simulation responses ; 21st International Technical Conference on the Enhanced Safety of Vehicles Conference (ESV); Stuttgart; Germany; 2009; Paper 09-0407]. The findings of this study should be valid if another rating tool is used. 30
Method CORA (CORrelation & Analysis) uses two different methods to assess the correlation of signals. Corridor method calculates the deviation between curves by using corridors. Cross correlation method analyses specific curve characteristics like phase shift or shape of the signals. The rating results ranges from 0 (no correlation) to 1 (perfect match). 31
Dummy Models The FAT ES-2 Dummy Models are used to demonstrate the rating. The models are developed by the German car makers and suppliers. They are used in production and they are accepted all over the world. The following releases are used: Release v2.0 Release v4.5 Release v5.0 32
Dummy Models Releases v2.0 Released in spring 2004. Model was derived from the EuroSID model. For validation material tests of some foam parts are used and mainly full assembled tests like pendulum and sled tests. The focus was on a good overall performance. 33
Dummy Models Releases v4.5 Released in summer 2009. Optimisation by using the validation tests of release 2.0. Feedback of customers and some small additional component tests are used to improve the model. The robustness of this release is increased significantly. 34
Dummy Models Releases v5.0 Released in spring 2011. Initiated by the PDB (Partnership for Dummy Technology and Biomechanics ) in 2009. The model is improved in nearly all body regions like: Shoulder Abdomen Lumbar spine New Material, component and sled tests are used. Geometry of internal parts is captured accurately. 35
Load Cases Certification Tests: 36
Load Cases Component Tests: Clavicle Test two different impact velocities for each test Abdomen Test different speeds on different impact locations and angles for each test 37
Load Cases Component Tests: Lumbar Spine Test bending shear torsion two different impact velocities for each test 38
Load Cases Sled Tests: PDB Sled Tests FAT Sled Tests 39
Results Certification Tests R2.0 R4.5 R5.0 Shoulder 0.562 0.645 0.825 Thorax 0.841 0.919 0.911 Abdomen 0.532 0.576 0.774 Lumbar spine 0.394 0.397 0.568 Pelvis 0.748 0.625 0.785 Component Tests R2.0 R4.5 R5.0 Clavicle 0.551 0.594 0.776 Abdomen 0.690 0.714 0.750 Lumbar spine 0.675 0.562 0.731 40
Results Sled Tests R2.0 R4.5 R5.0 PDB tests 0.390 0.579 0.666 FAT tests 0.552 0.592 0.668 PDB Sled Tests R2.0 R4.5 R5.0 D1 P barrier 0.536 0.509 0.617 D3 P barrier 0.634 0.616 0.724 D4 P barrier 0.000 0.612 0.657 R2.0 R4.5 R5.0 FAT Sled Tests D1 F barrier, v1 0.587 0.652 0.775 D1 F barrier, v2 0.537 0.590 0.587 D3 F barrier, v1 0.574 0.560 0.695 D3 F barrier, v2 0.634 0.600 0.697 D4 F barrier 0.479 0.602 0.667 D5 F barrier 0.545 0.561 0.657 D6 F barrier, v1 0.516 0.525 0.640 D6 F barrier, v2 0.634 0.723 0.794 D7 F barrier 0.464 0.514 0.497 41
Discussion Certification Tests The progress of dummy releases 4.5 compared to v2.0 is noticeable. Version v5.0 is a big step forward. The certification tests are clearly better. The rating process based on certification tests is not reliable. It is possible to tune a model to correlate well to the certification tests by neglecting the overall performance. Component Tests Dynamic tests of sub parts are an important part of any assessment. But they can not replace full dummy tests. (ES-2 v4.5 lumbar performance decreased but allover constant/slightly better). For a rating process based on component tests, tests of all relevant dummy parts must be available. 42
Discussion Sled Tests Both sled test configurations show the same evolution of the different ES-2 models. They are a solid base for a dummy rating procedure. Sled Tests might not recognize improvements of single parts of a model. Minor updates may not be relevant in sled tests but might help in vehicle tests. This indicates that component test ratings complete the allover rating. Vehicle Test Seems to be the best choice for evaluation of dummy model in theory. The influence of vehicle and restraint-system validation is too large to distinguish between dummy effects and effects caused by surroundings. 43
Conclusion Study gives first impression of an objective dummy rating procedure. The results mostly correlate to the user s experiences. A rating procedure must combine different kinds of tests. Certification tests give a limited impression on the overall quality of a dummy model. Component tests can only be used to assess the performance of single components. Sled tests seem to be the right choice to assess the performance of dummy models. But they might miss improvements of single parts. Vehicle tests are probably too complex to integrate them into a rating scheme. Combination of certification, component and sled tests seems to be right mix. Geometry, mass, inertias as well as robustness and modeling techniques cannot be assessed by an rating tool. 44
Thank you! 45