MECCA_20-02_web New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 1 10.14311/mecdc.2020.02.01 New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ NEW ADVANCED METHODS IN SIDE CRASH TESTING JAKUB JELÍNEK TÜV SÜD Czech, Novodvorská 994/138, +420606613227, jakub.jelinek@tuvsud.com !eské vysoké u"ení technické v Praze, Fakulta strojní, Technická 4, +420606613227, jakub.jelinek@fs.cvut.cz MILAN R!"I#KA !eské vysoké u"ení technické v Praze, Fakulta strojní, Technická 4, milan.ruzicka@fs.cvut.cz AL"B$TA KAFKOVÁ TÜV SÜD Czech, Novodvorská 994/138, +420725894624, alzbeta.kafkova@tuvsud.com !eské vysoké u"ení technické v Praze, Fakulta strojní, Technická 4, alzbeta.kafkova@fs.cvut.cz ABSTRACT This work follows up the previous work [1] regarding the used methodology in the field of passive safety, ie. crash testing. The work is based on experience gained in the Active Lateral Impact Simulator (ALIS) project and describes complete process. The main focus has been given to the fine-tuning of the boundary conditions and loading of the system in order to ensure correct biomechanical loads. KEY WORDS: CRASH TEST, FINITE ELEMENT METHOD, DESIGN OF EXPERIMENT, BIOMECHANICAL LOADS, DYCOT, ALIS SHRNUTÍ Tato práce navazuje na p#ede$lé p#ísp%vky [1] t&kající se metodiky v oblasti pasivní bezpe"nosti, a zejména crash testování. Tento "lánek vychází ze zku$enosti získané v rámci projektu bo"ních náraz' a za pou(ití systému Active Lateral Impact Simulator (ALIS) a popisuje cel& postup. Hlavní d'raz je kladen na jemné lad%ní po"áte"ních podmínek a náhradního zatí(ení p'sobícího na cel& systém a k dosa(ení po(adovan&ch biomechanick&ch kritérií. KLÍ#OVÁ SLOVA: NÁRAZOVÁ ZKOU%KA, METODA KONE#N&CH PRVK!, NÁVRH EXPERIMENTU, BIOMECHANICKÉ ZATÍ"ENÍ, DYCOT, ALIS 1. INTRODUCTION This work proposes a new advanced approach of combined virtual and physical testing. The main idea is to reduce development time and associated costs by using sled testing which used to be used mainly for physical simulation of frontal crashes. Simulation of side crash in sled environment is not a brand-new topic, but certainly very complex one. This method is not really used on regular basis especially due to predictability issues and low accuracy. This work presents new approach of combination both virtual and physical testing. The whole process starts with full crash simulation, goes through conversion of virtual model to reduced sled model, sled testing and finally is wrapped up with full vehicle crash. 2. MAIN SECTION 2.1 DYCOT TÜV SÜD Czech has recently invested a large sum to test lab equipped with sled system (catapult) – DYnamic COmponent Testing (DYCOT) [2]. Sled test system consists of sled with grid holes and pusher sled, where all electronics and measurement equipment is mounted as also shown on Figure 1. The pusher sled is being pushed by CSA catapult, equipped with hydraulic piston that can accelerate the sled by up to 90G to total velocity of 100kph with payload of 1000kg. When fully loaded (payload of 5000kg), the piston is capable of accelerating the sled up to 35G. Maximum force is equal to 2.5MN. Maximum acceleration gradient is 14G/ms. New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 2 FIGURE 1: DYCOT system during the acceleration of the test sample OBRÁZEK 1: Systém DYCOT p#i urychlení zku$ebního vzorku It is usually used for frontal crash test where the occupant safety is being tested. It can also be used for testing of crash-landing of any small airplane that would fit in the lab. Latest addition to the service portfolio is battery pack testing for any battery packs up to 1000kg. 2.2 ALIS The capabilities of DYCOT sled system have been significantly increased by adding ALIS into serie, right next to the sled platform see Figure 2. It uses up to 6 hydraulic cylinders in order to correctly simulate the door intrusion kinematics during the side crash. It enables one to use only small part of the car together with dummies and restraint systems and carry out simulation of the side crash with focus on restraint system and biomechanical loads. The system may seem as a "train of trolleys". The driven sled trolley is mounted to the main hydraulic system that generates the main acceleration pulse. ALIS is mounted on the separate trolley, attached to the sled. The whole structure is shown on Figure 3, where main components are identified. The lateral system consists of additional pneumatic system directly attached to several pneumatic cylinders, ALIS primary structure and control system, linear guiding system and "impact break-in structure". The main reason for testing is to fine-tune the restraint system in order to get the best biomechanical loading in cheaper and quicker way – on sled. The fact that sled tests with only several trim parts and seats are used instead of fully equipped crash vehicles makes this approach very effective. We are definitely talking about tens of percents. Door structure deforms and biomechanical loads are reached Se at , d um m y, c am er as ALIS sample palette ALIS actuators G lo ba l A CC p ul se Catapult ALIS on board system and structure Im pa ct m ec ha ni sm Li ne ar g ui di ng s tr uc tu re FIGURE 3: DYCOT + ALIS concept OBRÁZEK 3: Koncept DYCOT + ALIS FIGURE 2: Active Lateral Intrusion Simulator (ALIS) OBRÁZEK 2: Active Lateral Intrusion Simulator (ALIS) New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 3 2.3 METHODOLOGY The whole process starts with FE simulation of full vehicle crash and is shown in Appendix A. It is also very important to mention that usually testing consists of two sets of tests. The first one inputs are based on virtual model and results only and gets the initial recommendations for the first crash test. The second loop inputs are already based on this crash test and requires further development and tuning of ALIS. 2.4 DESIGN OF EXPERIMENT (DOE) [3] The main objective is to develop a virtual method that would allow reducing full crash into sled crash via ALIS, defining complete ALIS setup and give highly accurate results, while reducing costs. The DoE method is advanced mathematical method that uses n-dimensional mathematical surface for response values prediction based on combination of input parameters. The aim is to get ideally perfect match between full crash model as given at the beginning of the project and ALIS reduced model. Amount of input parameters is very often high. One of the ways how to put up with them might be Design of Experiment (DoE) with response surface creation or "step-by-step" iteration with subsequent physical validation as shown in Figure 4. Such method would reduce number of runs and predicts multiple results based on input parameter combinations. Such pulses have to fulfill feasibility criteria of the cylinders and catapult. 2.4.1 PULSE TUNING PROCEDURE There are several pulses that come into the whole simulation and subsequent physical test. In order to identify and tune pulses two main steps have been chosen. Firstly, contribution of every pulse needs to be determined and secondly chosen pulses have to be fine-tuned in a special manner that will ensure both physical feasibility and biomechanical responses. 2.4.2 PULSE IDENTIFICATION Currently there are three hydraulic cylinders available at the ALIS system. One is 120kN and other two are 60kN and therefore three pulses are available. Additional pulse comes from the catapult that represents overall pulses during the side crash. That makes it four pulses available for the first stage of DoE testing. Each pulse has got several parameters such as scale factor for both abscissa and ordinate and also offset values for both abscissa and ordinate. All four pulses have following set of parameters as shown in Figure 5. FIGURE 5: List of design variables OBRÁZEK 5: Seznam vstupních prom%nn&ch FIGURE 4: DoE response surface (top), step-by-step process (bottom) OBRÁZEK 4: DoE povrch (naho#e), postupn& proces ALIS #e$ení (dole) New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 4 Following variable abbreviations are used: • ASD_SY – scale factor of sled • ASD_OA – pulse offset of sled • DBB_SF – scale factor of actuator at B-pillar bottom • DBB_OA – pulse offset of actuator at B-pillar bottom • DBU_SF – scale factor of actuator at B-pillar upper • DBB_OA – pulse offset of actuator at B-pillar upper • DDD_SF – scale factor of actuator at door structure • DDD_OA – pulse offset of actuator at door structure Since there are 8 variables, the resultant design space will be 8D. Since there is no simple way of illustrating the 8D interactions, we have to go down to 3D visualisation. When always 3 variables are selected and can be switched for any other variable. All 200 experiments (simulations) have to be run It has to be pointed out that as there are 8 variables, then 8-dimensional surface will be created based on the responses and hence the complete surface is so complex that cannot be displayed. TABLE 1: List of responses TABULKA 1: Seznam vyhodnocovan&ch odezev ID Type Name Component Units 90079631 BAR First thorax rib Compression mm 90079632 Second thorax rib Compression mm 90079633 Third thorax rib Compression mm 90079634 First abdomen rib Compression mm 90079635 Second abdomen rib Compression mm 90000002 NODE Head acc Acceleration, velocity mm ms-2 / mm ms-1 90015619 T1 Lower neck acc Acceleration, velocity mm ms-2 / mm ms-1 90021212 T4 first thorax acc Acceleration, velocity mm ms-2 / mm ms-1 90023825 T12 second abdomen acc Acceleration, velocity mm ms-2 / mm ms-1 90029764 Pelvis acc Acceleration, velocity mm ms-2 / mm ms-1 FIGURE 6: Comparison of initial ALIS vs full crash results (ribs) OBRÁZEK 6: Porovnání úvodních v&sledk' ALIS s fyzickou zkou$kou ((ebra) New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 5 2.4.3 RESPONSES For response surface determination it is necessary to get responses respective to our objectives. Responses are resultants of any measurements such as force, displacement, acceleration, angle, etc. Response list is given by the scope of the sensitivity study. In all crash simulations, the most important are biomechanical loads that describes the behaviour of a human body during the crash event. The requirements differ very much from case to case so it is always unique set of criteria that are ideally to be matched. In our pole strike, it is ribs compression. Nowadays, most of the dummies and solvers are able to calculate and/or evaluate these criteria directly via sensors/points of interests. In our case several node and bars have been selected. Nodes are FIGURE 7: The response trends based on initial variable combination (top) and response trends based on update variable combination (bottom) OBRÁZEK 7: Trendy odezev v úvodním nastavení (naho#e) a trendy zalo(ené na upraven&ch parametrech (dole) New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 6 used for tuning of controlled trim deformation and its velocity. Simply the velocity and deformation of the trim ensures the same initial conditions as per full crash. Bar then are used for force (shoulder) and displacement (rib compression) evaluation. This metric is the most important for most of the safety crash engineers. Responses are used for response surface modelling and results evaluation. In our case there are several responses taken into account. They have been chosen according to the requirements of the customer and also EuroNCAP. Responses that have been used are shown in Table 1. 3. RESULTS OF THE VIRTUAL EXPERIMENTS So far we have been preparing ourselves for the main task. To choose suitable variables from all available sources to achieve the intended responses. Now, when the response surface has been created and validated, the selection of variable that would fit the intended values follows. The main reason of the virtual experiments is to perform sensitivity analyses that would later give a good knowledge of the system behaviour. This is particularly useful during the physical testing, when quick response to the current behaviour and recommendation of the next steps is highly expected and TABLE 2: Final variable values TABULKA 2: Seznam finálních hodnot prom%nn&ch Label Name Value Initial values ASD_SY scale factor of sled 1.02 No ASD_OA pulse offset of sled 0 Yes DBB_SF scale factor of actuator at B-pillar bottom 1.11 No DBB_OA pulse offset of actuator at B-pillar bottom 0 Yes DBU_SF scale factor of actuator at B-pillar upper 1.03 No DBU_OA pulse offset of actuator at B-pillar upper 0 Yes DDD_SF scale factor of actuator at door structure 0.98 No DDD_OA pulse offset of actuator at door structure 1 No FIGURE 8: Comparison of initial and final ALIS pulses OBRÁZEK 8: Porovnání úvodních a finálních puls' ALIS New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 7 there is no time for further simulations. In order to get ideal pulse configurations for respective biomechanical responses, it is necessary to set the target. EuroNCAP assessment is based on scoring system of the maximal biomechanical loads. For illustration there is a comparison of initial ALIS run, with all variables equal to 1, and full crash model shown on Figure 6. The match is not ideal one at the moment and our goal is to get better match. Hence there has to be an update done of some or all available pulses (scale factor or offset). The suitable variable combinations can be found by user to achieve his requirements. LS-OPT can easily predict response values based when one changes the input variables as indicated on Figure 7. This is exactly the way how to better understand mutual interaction between input variables and responses. In our case, when the five ribs are of interest, we get desired response with following variables written in Table 2. As these values are predicted, another testing run has to be to verify the suitability. Updated three pulses for ALIS and one for sled are shown on Figure 8. Updated ALIS results of dummy biomechanical criteria compared to full crash data are displayed on Figure 9. The comparison shows rather good match of both simulation approaches. Reduced model is and always will be only approximation and can only get close to the full crash simulation model. Four pulses with reasonable match, which is usually considered within deviation of 10%, to the full crash model have been found and hence the first objective is complete. Secondary objective was to get a good knowledge of the system behaviour and it has also been done. It will become very useful in upcoming testing. 4. CONCLUSION This paper has shown how to handle ALIS project within the virtual part. The main objective (pulses identification) has been achieved. Controlled pulses have become input parameters into the physical sled test. It is very important to get a good knowledge of the whole system behavior and how biomechanical responses are affected by variation of input as this helps the tuning procedure during early physical testing. Without it, one would not be able to recommend further steps to improve the results accuracy. Future work is to cover the last remaining part and it is the physical testing and results validation. LIST OF NOTATIONS AND ABBREVIATIONS ALIS – Active Lateral Impact Simulator ASD_SY – scale factor of sled ASD_OA – abscissa offset DBB_SF – B-pillar bottom scale factor DBB_OA – B-pillar bottom abscissa offset FIGURE 8: Comparison of initial and final ALIS pulses OBRÁZEK 8: Porovnání úvodních a finálních puls' ALIS New Advanced Methods in Side Crash Testing JAKUB JELÍNEK, MILAN R!"I#KA, AL"B$TA KAFKOVÁ MECCA 02 2020 PAGE 8 DBU_SF – B-pillar upper scale factor DBU_OA – B-pillar upper abscissa offset DDD_SF – door scale factor DDD_OA – door abscissa offset DoE – design of experiment DYCOT – Dynamic Component Testing ENCAP – European New Car Assessment Programme REFERENCES [1] Jelinek J., R'(i"ka M., Kalinsk& M. Advanced Methods in Crash Safety Testing, 56th Conference on Experimental Stress Analysis Proceedings 2018, ISBN 978-80-270-4062-9 [2] )otola M., 2016. DYCOT presentation, TÜV SÜD Czech, pages 3-7 [3] Jelinek J., R'(i"ka M. Advanced methods in crash safety testing, 24th Workshop of Applied Mechanics 2018, ISBN 978-80-01-06453-5 APPENDIX A – METHODOLOGY Output is to be biomechanical loads, intrusion and kinematics of important structural parts such as doors, A- and B-pillars. Size reduction of FE model comes next. The most important outcome of this phase is determination of the ALIS settings. This includes number of cylinders used, their timing and also design of the impact structure. Amount of input parameters is countless. Other two phases are related to the physical testing. FIGURE 10: Real crash to ALIS reduction procedure [3] (Courtesy of )koda Auto) OBRÁZEK 10: Proces redukce z reálného crash testu po ALIS [3] (S laskav&m dovolením )koda Auto) Full Car FE Simulation Output: Biomechanical Loads lntrusion Kinematics Size-Reduced FE Simulation Output: Number of cylinders Cylinder positions Force distribution (shape and magnitude) Impact structure design Reduced Physical Crash Test Output: Biomechanical Loads Physical Crash Test Output: Biomechanical Loads Model Reduction ALIS Parameters Application Correlation