STATUS OF NHTSA S EJECTION MITIGATION RESEARCH. Aloke Prasad Allison Louden National Highway Traffic Safety Administration

Transcription

1 STATUS OF NHTSA S EJECTION MITIGATION RESEARCH Aloke Prasad Allison Louden National Highway Traffic Safety Administration United States of America Stephen Duffy Transportation Research Center United States of America Paper Number ABSTRACT Federal Motor Vehicle Safety Standard (FMVSS) No. 226, Ejection mitigation, set requirements for ejection mitigation systems to reduce the likelihood of complete and partial ejections of vehicle occupants through side windows during rollovers or side impact events. At the time of the final rule, the agency was not in a position to extend coverage to roof glazing because of the need to research a viable performance test procedure. This paper presents the results of initial assessments of the test as a viable performance test procedure and of the performance of current production roof glazings in preventing occupant ejections. The assessment of ejection protection offered by laminated glazing roof panels in production vehicles was made primarily using a guided impactor (18 kg) directed toward pre-broken roof glazing from inside the vehicle. Test procedures followed those developed in the FMVSS No. 226 regulation. Test speeds were 16 and 20 km/h (10 and 12.5 mph). Three vehicles with laminated glass sunroofs were selected: a 2008 Ford Flex and a 2013 Ford CMAX, both equipped with a panoramic laminated glass roof that is fixed to the vehicle structure, and a 2013 Subaru Forester with a moveable laminated glass inbuilt sunroof. When tested at the 16 km/h impact speed, the displacements for both vehicles with fixed glass roof structures were within the 100 mm criterion specified for side windows in FMVSS No. 226, although the results from the Ford Flex were at or just slightly below the criterion. As expected, higher values were seen at the 20 km/h speed. The Ford CMAX displacements were slightly below the criterion, while the Ford Flex exceeded the criterion for all tests conducted at the higher speed. There was no incidence of bonding material failure at the glass/roof structure interface, and no damage was seen to the roof sheet metal in either vehicle. All tests on the Subaru Forester were conducted with the sunroof in the fully closed position, and all displacement values exceeded 100 mm at both test speeds. The number of vehicle designs tested was limited by the availability of laminated glazing used in production sunroof designs. Extensive vehicle preparation was required to ready them for testing with the impactor used for side window ejection evaluations. Prasad 1

2 INTRODUCTION The purpose of Federal Motor Vehicle Safety Standard (FMVSS) No. 226 Ejection Mitigation was to establish requirements for ejection mitigation systems to reduce the likelihood of complete and partial ejections of vehicle occupants through side windows during rollovers or side impacts. In the January 2011 final rule (76 FR 3212, January 19, 2011), the National Highway Traffic Safety Administration (NHTSA) said it was interested in learning more about roof ejections and would like to explore this area further. NHTSA also stated in the final rule that mitigating roof ejection was determined to be potentially cost effective, but the agency was not in a position to extend coverage to roof glazing in the final rule due to the lack of a proven performance test procedure for roof glazing. Vehicle and Buck Description Figure Ford Flex Three vehicles were selected that contained roofs with laminated glass composition. The 2009 Ford Flex (shown in Figure 1) has a panoramic laminated glass roof that is fixed to the vehicle structure. The movable sunroof above the front row seats is made from tempered glass and was not tested. The headliner divides the laminated glass into three distinct daylight openings (as defined in FMVSS No. 226): two symmetrical openings above the 2 nd row seating position and a larger opening above the 3 rd row seating position, as seen in Figure 2. The 2013 Ford CMAX shown in Figure 3 also has a fixed panoramic glass roof. The exterior dimensions of the glass are 1.5 m in length by 1.5 m in width. The headliner reduces the daylight opening resulting in an area measuring 1 m by 1 m. Figure 2 - Ford Flex Interior Showing Three Distinct Daylight Openings Prasad 2

3 Figure Ford CMAX The 2014 Subaru Forester has a moveable laminated glass inbuilt sunroof (shown in Figure 4). It is an inbuilt sunroof since the operable glass panel slides between the vehicle roof and headliner. A small motor mechanically opens and closes the power sunroof. Attached to the sunroof are small rods called cable guides, which in turn are attached to the sunroof motor at the opposite end. When the sunroof motor is activated, the motor spins which in turn pushes or retracts the rods connected to the sunroof. The kinematics of the inbuilt design also allow the sunroof to be rotated upward at the rear edge for venting purposes. Figure 5 illustrates the different modes of operation for this sunroof type. The daylight opening measures 1.5 m by 1.5 m. Figure 5 Operation of Subaru Forester s Inbuilt Sunroof The ejection impactor used in this project meets FMVSS No. 226 specifications and was originally designed to extend across a vehicle to impact the side window and cannot be articulated inside a vehicle. This required the vehicles to be prepared so that the impactor could be aimed at the roof structure. The vehicles were turned 90 degrees and secured to a rigid steel base (see Figure 6). All components not integral in providing rigidity to the roof were removed. This included all sunshades. A portion of the floor was removed to allow the ejection impactor to be inserted into the vehicle. Also, the vehicle was secured to the impactor frame using steel tubing to limit vehicle motion during impact. Figure Subaru Forester Figure 6 - Vehicle Prepared to Accommodate the Ejection Impactor TEST PROCEDURE AND EQUIPMENT Prasad 3

4 Ejection Impactor Description The component test involved use of a guided linear impactor specified for use in FMVSS No. 226 (shown in Figure 7). The device used met the friction and static deflection characteristics therein. It was designed to replicate the loading of a 50th percentile male occupant s head and upper torso during ejection situations. The ejection mitigation test device consisted of an impactor and propulsion mechanism. The ejection impactor consisted of a headform attached to a shaft. The featureless headform was originally developed to be a free-motion headform for use in interior impact testing. The width and height dimensions as well as the contour of the headform face were chosen based on biomechanical data from mid-sized adult males. The impacting face of the headform had dimensions which are the average of the front and side of a human head. The ejection impactor has a mass of 18 kg ± 0.05 kg. In addition to low friction characteristics, the impactor was capable of obtaining the desired velocity in a highly repeatable manner and maintaining the desired velocity over the travel length. Impact velocity was measured by an optical sensor that recorded the time a beam of light was interrupted when a flag, attached to the impactor rod, passed through it. A linear variable differential transformer (LVDT) recorded the displacement of the impactor mass and calculated the velocity to provide a redundant impact speed. The impactor had a maximum stroke length of 700 mm. Test Description A series of tests using the ejection impactor was conducted on the vehicles roof glazing structures to determine their retention characteristics. The impactor was positioned perpendicular to the roof, with the direction of travel being from inside the vehicle towards the outside, and aligned with the predetermined target locations. Target locations were selected to challenge different aspects of sunroof design. There were two targeted impact regions: the centermost point on the glazing area and a point in the upper rear area of the glazing. A single test was also performed on the upper forward corner of the moveable sunroof. Impacts to the centermost points were intended to primarily test the strength of the polyvinyl butyral (PVB) interlayer of the laminated glazing, while impacts in the corners were intended to primarily test the mounting between the laminated glazing (movable or fixed) and the vehicle. The selected impact locations for the Ford Flex, Ford CMAX, and Subaru Forester are shown in Figures 8, 9, and 10, respectively. The headform was aligned such that its longitudinal axis was perpendicular to the vehicle s longitudinal axis. Figure 11 shows a typical setup for the ejection mitigation component test. Impacts were conducted at 16 and 20 km/h. Data from the displacement transducer was captured with a data acquisition system sampling at 20,000 Hz. The linear potentiometer recorded the impactor face displacement measured from first contact of the impactor headform with the interior glazing surface through maximum dynamic displacement. Primary and redundant accelerometers recorded the impact pulse for force computation. Figure 7 - Ejection Impactor Figure 8 - Impact Locations for Ford Flex Prasad 4

5 Figure 9 - Impact Locations for Ford CMAX Figure 11 - Typical Setup for Ejection Impacts Figure 10 - Impact Locations for Subaru Forester After establishing the daylight opening, an offset line was marked 25 mm inside the daylight opening. The offset from the window daylight opening provides buffer to assure that the impactor does not strike any vehicle structure surrounding the glass. Prior to testing, the glazing was broken using the prescribed method outlined in FMVSS No. 226 to reproduce the state of glazing in an actual rollover crash. The method uses a 75 mm offset pattern, with a 75 mm by 75 mm pattern on the outside surface of the glazing and the same pattern, offset by 37.5 mm horizontally, on the inside surface (see Figure 12). A spring loaded center punch was used to break the glass. The fixed glass roof panels on the Ford CMAX and Flex were replaced between tests by a professional glass installer using typical aftermarket glass replacement technique. In-house personnel at the Vehicle Research and Test Center (VRTC) replaced the glass panels and associated hardware on the Subaru Forester s moveable sunroof. Figure 12 Pre-Broken Glazing Photographs were taken to document the test set-up and post-test observations. High-speed video was used to capture the impact during each test. The roof structure profile at the point where the glazing is bonded to the roof structure was measured pre- and post-test with a 3-D coordinate measuring system to determine if damage to the roof occurred. EJECTION TEST RESULTS One goal of this test series was to assess the performance of a small sample of current production vehicles with laminated glass roof structures to determine their retention characteristics under loading with the ejection impactor. The results of the tests are tabulated in Table 1. Prasad 5

6 Table 1 Results for Ejection Testing Test Number FF01 FF02 FF03 FF04 FF05 FF06 Impact Position Center of Daylight Opening Center - Upper glazing area over 2nd row seat Top Rear Corner - Glazing area over 3rd row seat Center of Daylight Opening - Glazing area over 3rd row seat Center of Daylight Opening Center - Upper glazing area over 2nd row seat Top Rear Corner - Glazing area over 3rd row seat Center of Daylight Opening - Glazing area over 3rd row seat Impact Speed (km/h) Displacement Beyond Glass Plane (mm) 2009 Ford Flex Peak Dynamic Force (N) , , , , , ,280 Comments Some tearing of PVB interlayer ~ 4 mm; no glass/roof bond separation Some tearing of PVB interlayer ~ 4 mm; no glass/roof bond separation No tearing of PVB interlayer; no glass/roof bond separation Some slight tearing of PVB interlayer < 4 mm; no glass/roof bond separation Some tearing of PVB interlayer at side of impact area; no glass/roof bond separation Some tearing of PVB interlayer ~ 5 mm; no glass/roof bond separation 2013 Ford CMAX FC01 Center of Daylight Opening ,625 FC02 Center of Daylight Opening ,637 FC03 Top Rear Corner , Subaru Forester SF01 Top Rear Corner ,865 SF02 Top Rear Corner ,516 SF03 Center of Daylight Opening ,851 SF04 Center of Daylight Opening ,440 SF05 Upper Forward Corner ,836 No tearing of PVB interlayer; no glass/roof bond separation No tearing of PVB interlayer; no glass/roof bond separation No tearing of PVB interlayer; no glass/roof bond separation Glazing material separated from frame above impact point; Sunroof guide rails pulled out from channel creating large gap (< 100 mm) at rearward edge; no tearing of plastic interlayer Glazing material separated from frame above impact point; no tearing of plastic interlayer Pressure variance in propulsion unit produced higher impact speed; Sunroof guide rails pulled out from channel creating large gap (> 100 mm) at forward edge; no tearing of plastic interlayer Sunroof guide rails pulled out from channel creating large gap (> 100 mm) at forward edge; no tearing of plastic interlayer Sunroof guide rails pulled out from channel creating large gap (> 100 mm) at forward edge; no tearing of plastic interlayer When tested at the 16 km/h impact speed, the displacements for both vehicles with fixed glass roof structures were within the 100 mm criterion specified for side windows in FMVSS No. 226, although the results from the Ford Flex were at or just slightly below the criterion. As expected, higher values were seen at the 20 km/h speed. The Ford CMAX displacements were slightly below the criterion, while the Ford Flex exceeded the criterion for all tests conducted at the higher speed. The table notes if tearing to the PVB interlayer occurred during the test and to what extent. However, the impactor was fully contained (plastic interlayer showed minor tears but not holed ) in all tests despite the presence of tearing. Also, there was no incidence of bonding material failure at the glass/roof structure interface, and no damage was seen to the roof sheet metal in either vehicle. This was verified by the 3-D coordinate measuring system and the professional glass installation procedures. Prasad 6

7 All tests on the Subaru Forester were conducted with the sunroof in the fully closed position, and all displacement values exceeded 100 mm. The displacement measurement was a combination of both the glazing material and moveable system parts. Impacting the upper corner of the forward edge at 16 km/h (SF05) produced the highest displacement value (shown in Figure 13). The failure mode was in the system designed to move the sunroof, as shown in Figure 14. In this system, the forward edge of the glass panel is attached to the aluminum frame through the cable guide. The cable guide travels in a U-channel on the aluminum frame. Finally, there was no discernable difference in the peak impact loads between the center of glazing and corners. Figure 13 - Maximum Dynamic Excursion on Subaru Forester Movable Panel SUMMARY AND OBSERVATIONS NHTSA evaluated the ejection impactor specified in FMVSS No. 226 for use in testing roof openings. Testing by rotating the vehicle and using the impactor through the floor appears to be feasible. Three vehicles with production roof laminated glass panels were tested. Test were conducted to selected targets at 16 and 20 km/h. o The ejection impactor was fully contained by the glazing in all fixed sunroof panel tests and four of five movable sunroof panel tests. The PVB inner layer showed minor tears in some tests but was not holed. o There was no damage to the roof sheet metal in any test. o For the fixed panoramic designs: When tested at 16 km/h, all displacements were 100 mm or less. When tested at 20 km/h, the displacements ranged from 92 to 130 mm. There was no failure of the glazing to roof bonding. o For the movable sunroof design, when tested at the center of the daylight opening and top rear corner: When tested at 16 km/h, the displacements were 103 and 105 mm. When tested at 20 km/h, the displacements were 150 and 167 mm. o For the movable sunroof, there was damage to the system designed to move the sunroof, resulting in large gaps at the periphery in four of five tests. Some modification to the system would be needed to achieve displacements below 100 mm. Figure 14 - Typical Failure Mode at the Rail for the Movable Glass Panel Prasad 7

8 INVESTIGATING POTENTIAL CHANGES TO THE IIHS SIDE IMPACT CRASHWORTHINESS EVALUATION PROGRAM Matthew L. Brumbelow Becky Mueller Raul A. Arbelaez Insurance Institute for Highway Safety USA Matthias Kuehn GDV German Insurers Accident Research Germany Paper Number ABSTRACT Ninety-seven percent of 2016 model year vehicles evaluated in the Insurance Institute for Highway Safety (IIHS) side impact test received good ratings. Good-rated vehicles have lower side impact fatality rates than other vehicles, but additional crashworthiness improvements may be possible. In a previous analysis of real-world cases, most serious injuries in good-rated vehicles resulted from crashes with impacts centered farther forward than the IIHS configuration and/or crashes that produced greater intrusion at the occupant location. The current study examines whether the occurrence of real-world injury in a different crash configuration can be identified in the laboratory, how injury risk in such a configuration compares to the current IIHS test, and whether current vehicle designs already offer improvements over the vehicles in the real-world cases (median model year was 2007). A NASS-CDS crash of a 2007 Honda Fit struck by a 1999 Toyota Camry was chosen for laboratory replication. The nearside impact location was centered forward of the front axle and the 75-year-old driver occupant sustained fatal thoracic injuries. A WorldSID-50th percentile male ATD with a RibEye deflection measurement system was used to record injury measures, and these were compared to measures from four additional tests. In the first, the case vehicle was struck by the IIHS MDB at the standard test location and speed (50 km/h). In the second, the reconstruction test was repeated using a 2015 Honda Fit as the struck vehicle. The third and fourth tests involved the IIHS MDB impacting the 2015 Fit at the standard location at 50 km/h and 60 km/h, respectively. The reconstruction test of the 2007 Fit produced structural damage comparable to the real-world case. Compared to the standard IIHS test, the torso airbag deployment time was similar, the ATD loading was later due to the longer crash pulse, and there was less intrusion at the occupant position. Despite these differences, the injury measures recorded by the ATD were broadly similar and indicated elevated injury risks consistent with the observed realworld injuries. Compared to the 2007 model, the 2015 Fit produced much lower intrusion and injury measures in the reconstruction and standard IIHS tests. The greatest injury risks in all five tests were recorded when the 2015 Fit was impacted by the IIHS MDB at 60 km/h. The loading and intrusion patterns in the real-world reconstruction differed from the standard IIHS test, but did not translate to large differences in predicted injury risks. Furthermore, tests of the newest generation Fit suggest some of the risk factors observed in the real-world crash have been mitigated by more recent crashworthiness improvements. However, the benefit of these improvements was more than offset by the increased severity of a 60 km/h test. Simply increasing the severity of the current IIHS test may be more effective at producing additional real-world improvements than a test configuration that has a different impact location but does not result in increased intrusion. However, more research would be needed to ensure that a higher severity test does not promote countermeasures with reduced protection in less severe crashes. Brumbelow 1

9 INTRODUCTION The Insurance Institute for Highway Safety (IIHS) began its side impact crashworthiness evaluation (SICE) program in In the SICE test, the stationary tested vehicle is struck on the left side by a 1,500 kg moving deformable barrier (MDB) at 50 km/h. One of four ratings is assigned based on a combination of structural performance, injury measures recorded on dummies in the driver and left rear passenger seat, and observations of the restraint system and kinematics of the anthropometric test device (ATD). Of the models tested in the program, 27 percent received the highest rating of good, while 41 percent received the lowest rating of poor. For models, these proportions had changed to 93 and 1 percent, respectively (Figure 1). Based on analysis of real-world side impacts, Teoh and Lund [1] found that when a left-side crash occurred, drivers of good-rated vehicles were 70 percent less likely to die than drivers of poor-rated vehicles. When combined with other changes in the fleet, driver behavior, and environmental factors, improved crashworthiness has helped contribute to a decline in side-impact driver fatality rates in 1-3 year old vehicles from 22 per million in 2005 to 5 per million in 2015 [2]. Good Acceptable Marginal Poor 100% 80% 60% 40% 20% 0% Figure 1. IIHS side impact ratings by vehicle model year Despite these improvements, side impact crashes accounted for 5,593 passenger vehicle occupant fatalities in These fatalities occurred in vehicles with a median model year of 2003, meaning that most were not rated in the IIHS program. This suggests that side impact fatality rates will continue to fall as the fleet continues to turn over, given the relationship between good test performance and real-world experience. At the same time, however, 49 percent of the rated vehicles with 2015 side impact fatalities were rated good. It is possible that the existing IIHS test configuration could be modified or supplemented in order to encourage additional countermeasures that improve the real-world crashworthiness of the passenger vehicle fleet. In order to identify changes to the IIHS test that have the potential to provide additional benefit, a previous study focused on crashes that produced serious or fatal injuries to occupants in vehicles with good ratings [3]. Queries of the National Automotive Sampling System Crashworthiness Data System (NASS-CDS) and Crash Injury Research and Engineering Network (CIREN) identified 109 occupants in crashes from Differences between the real-world crashes and the IIHS test were categorized through in-depth analysis of each case. Table 1 shows the potential for various changes to the IIHS test configuration to affect the injury outcome for the study population. No single change to the current test configuration would have been relevant to more than around one-quarter of the occupants. When considering combinations of two changes, a more severe test with an impact centered farther forward on the vehicle had the greatest potential relevance. This assumes such a test would encourage countermeasures that benefit occupants in crashes that differ from the current configuration in either or both of these ways. While the NASS-CDS/CIREN study was restricted to good-rated vehicles, it still is possible that the sample does not represent the current fleet. The median model year for vehicles in the sample was 2007, and 91 percent of the occupants were in vehicles built before Countermeasures introduced since then may have reduced the risk of injury in some of the specific crash scenarios identified in the study. Brumbelow 2

10 Change or combination of Case occupants changes affected Forward impact location 28% Increase severity 17% Adjust injury criteria 9% Include far-side occupant 9% Increase severity and forward 62% impact location Increase severity and include 37% far-side occupant Table 1. Potential relevance of test changes to NASS-CDS and CIREN occupants with serious injury in goodrated vehicles Even if the relevance of potential test changes shown in Table 1 holds for the current fleet, it does not necessarily follow that a modified side-impact test could predict the real-world injuries that were observed. This is a particular concern for oblique impacts or for perpendicular impacts that are offcentered from the occupant compartment and produce oblique ATD loading or kinematics. Existing side impact dummies have been designed for and validated against perpendicular lateral impacts. Some work has been done to document the response of specific body regions under oblique loading (e.g. [4]) but injury reference values have not been established, nor has the kinematic response of the dummies been validated in oblique conditions. In addition to possible limitations of the ATDs, there may be additional challenges to replicating the vehicle loading conditions observed in real-world cases in a laboratory setting Average crush (cm) For example, even among good-rated vehicles, manufacturers have continued to make structural improvements. Figure 2 shows the average B-pillar crush measurements in the IIHS test by model year. Injury risks also may have been reduced due to the oblique pole tests introduced by the National Highway Traffic Safety Administration (NHTSA) in Federal Motor Vehicle Safety Standard 214 and the New Car Assessment Program. Even improvements in other modes such as the IIHS small overlap test or the roof strength test may carry over to provide benefit in side impacts Figure 2. Average B-pillar crush in SICE tests of goodrated vehicles by model year. Crush is measured relative to the precrash centerline of the driver s seat, with negative values indicating crush does not reach the centerline. The current study was conducted to explore the potential for modified crash tests to predict injury outcomes observed in the real-world that may be different from the risks identified in the existing SICE test. In addition, tests of a current vehicle design were used to investigate whether some of the risks associated with a modified configuration have been mitigated by more recent vehicle redesigns. METHODS The NASS-CDS and CIREN cases previously analyzed [3] were filtered to select a case for laboratory replication. The inclusion criteria were a near-side vehicle-to-vehicle crash centered farther forward than the existing SICE test. In addition, a case occupant sustaining thoracic injuries with a level of 3 or higher on the Abbreviated Injury Scale (AIS) was required due to the prevalence of injuries to that body region in the overall analysis. Finally, photographic documentation and measures of structural deformation for the striking vehicle were necessary in order to facilitate and assess the agreement between the test configuration and the real-world case. Based on the inclusion criteria, NASS-CDS case was selected for replication. Details of this case are shown in Table 2. The initial impact was the primary event, with the front of the 1999 Toyota Brumbelow 3

11 Camry striking the left side of the 2007 Honda Fit close to the front axle. The coded direction of force for the Fit was 290 (20 oblique towards the rear). The coded case information was used to reconstruct the crash using the PC-Crash software [5]. This resulted in calculated impact speeds of 88 km/h and 33 km/h for the Camry and Fit, respectively Honda Fit 1999 Toyota Camry 75-year-old male, 185 cm, 104 kg, belted, fatally injured AIS 2 injuries AIS 5 Bilateral flail chest AIS 4 Trachea perforation AIS 3 Pulmonary artery laceration AIS 3 Left lung contusion, laceration, hemothorax AIS 2 Spleen laceration Table 2. Details of NASS-CDS case the three-dimensional displacements for each of three LEDs installed on each rib. Figure 4 shows the installation of the LEDs on a rib. Struck vehicle Striking vehicle Case occupant The striking and struck vehicles in the replication test were the same generation as those in the NASS-CDS case. Due to the technical challenges of conducting an oblique test with both vehicles moving, two alternative tests were conducted. In the first test, the Fit was stationary but rotated 20 to represent the assumed direction of force in the real-world crash. In the second test, both vehicles were moving but aligned perpendicularly at impact. Based on the damage patterns to both vehicles, the second configuration was selected as the best match to the real-world crash. Another limitation of the IIHS crash propulsion system required the Camry s speed to be reduced from the 88 km/h estimated in the NASS-CDS case to a test speed of 80 km/h. The test speed for the Fit was 32 km/h. Figure 3 shows the orientation of both vehicles at impact. The horizontal centerline of the Camry was aligned 19 cm forward of the Fit s left front axle. A WorldSID 50th percentile male ATD was used to assess the injury risks for the driver occupant. The ATD was positioned according to the IIHS SICE protocol while following the seat positioning procedure for a 50th percentile male [6],[7]. The ATD was equipped with a RibEye Multi-Point Deflection Measurement System [8]. The RibEye system reports Figure 3. Impact orientation in replication test Figure 4. RibEye LEDs installed on WorldSID rib [8] The three-dimensional displacement measurements were converted to a resultant deflection for each of the three LED locations on each rib. The resultant deflection was defined relative to the centerline of each rib horizontally and vertically and to the centerline of the dummy laterally. In other words, the calculated deflection would match the reading from a potentiometer or IR-TRACC that was attached between the location of the LED and the center of the ATD at the x-coordinate of the rib centerline. In addition to the resultant deflection measurements, the peak lateral displacement was calculated. Figure 5 illustrates the deflection and displacement measurements. Brumbelow 4

12 the NASS-CDS/CIREN analysis, the replication and SICE tests of the Fit were repeated using the Fit design. Finally, the new Fit was evaluated in a SICE test with the impact speed increased to 60 km/h. This allowed a comparison of injury risks between two crash modes that differed from the SICE test in the ways most commonly identified in the analysis of real-world crashes. The complete test matrix is shown in Table 3. ID Struck vehicle A 2007 Fit, 33 km/h B 2007 Fit, stationary C 2015 Fit, 33 km/h Figure 5. Definition of the deflection and displacement measurements for one of the RibEye LEDs. Deflection is defined as the difference between L 0 and L1. Readings from the ATD were compared to the injury risk curves published by the International Organization for Standardization [9]. For the thoracic and abdominal risks, the greatest deflection was used from all measurement locations, even though the standard IR-TRACC would not record deflection at the anterior or posterior locations. Adjusted risks were calculated for a 75-year-old, since this was the age of the occupant in the NASS-CDS case being replicated, as well as for a 45-year-old. Because no injury risk curve has been published for the head, risks were assessed using the HIC-15 curve published for the Hybrid III 50th percentile male ATD in frontal crashes [10]. Injury risks from the replication test cannot directly be compared to the original SICE test with a 5th percentile female SID-IIs ATD. In order to isolate differences introduced by modifying the test configuration, a second 2007 Fit was tested according to the SICE procedure (50 km/h MDB test) but with the WorldSID ATD in the driver position. To explore the effect of the latest crashworthiness improvements that may not have been captured in D 2015 Fit, stationary E 2015 Fit, stationary Impact configuration 1999 Camry centered 24 cm forward of front axle, 88 km/h MDB centered 145 cm rearward of front axle, 50 km/h 1999 Camry centered 24 cm forward of front axle, 88 km/h MDB centered 145 cm rearward of front axle, 50 km/h MDB centered 145 cm rearward of front axle, 60 km/h Table 3. Test matrix RESULTS Figure 6 shows a comparison of crush measurements from the real-world NASS-CDS case and from the reconstruction test. The bumper bar of the striking Toyota Camry had more deformation in the realworld crash than in the test. The lateral crush measurements on the struck Honda Fit were similar. Figure 7 shows a comparison of lateral crush measurements for all 5 crash tests. Almost all the intrusion in Tests A and C occurred forward of the pretest ATD H-point position, while the tests in the SICE configuration had intrusion profiles centered between the H-point and the B-pillar. The tests of the Fit had less crush than the paired tests with the earlier design. In fact, for the tests in the SICE configuration, the B-pillar intrusion for the current design in the 60 km/h test (Test E) was less than the intrusion for the old design in the 50 km/h test. Several of the RibEye readings had data drop-outs, potentially caused when the line of sight between an LED and a sensor was obstructed. Usually these dropouts occurred after peak loading, or were of short Brumbelow 5

13 enough duration that linear interpolation still allowed the data to be used. However, at times the drop outs were longer and none of the output from a given sensor was usable. Table 4 lists these sensors. Figure 7. Vehicle lateral crush profiles taken at the mid-door height (cm). The origin is the intersection of the front axle and vehicle centerline. The crush profile is not shown for Test E because the driver door opened and affected the measurement. The Bpillar deformation is reported at the mid-door height. There was no B-pillar deformation in Tests A or C. Test ID B C D E E E Figure 6. Vehicle crush measurements (cm) from the reconstruction test (Test A) and from the NASS-CDS case, shown at the test impact point. Measurements of the striking vehicle were taken on the front bumper bar and the origin is the front center of the bumper. Measurements of the struck vehicle were taken near the frame rail height and the origin is the intersection of the front axle and vehicle centerline. Rib Sensor position(s) Shoulder Anterior Shoulder Anterior Shoulder All three Shoulder Anterior st 1 thoracic Anterior 2nd thoracic Anterior Table 4. RibEye sensor locations where data loss prevented valid measurements Peak injury measures for all 5 tests are shown in Table 5. Figure 8 shows the injury risks for a 45-year-old and for a 75-year-old calculated using published risk curves. With the exception of the abdominal body region, the highest injury values were recorded in Test E. Among the other four tests, tests A and B tended to have higher injury risk than the paired tests with the newer Fit. One exception was the higher shoulder force recorded in the SICE test in the newer vehicle. Brumbelow 6

14 A B C D E HIC Shoulder force (kn) m 41-p 33-m * 51-m 51-m 36-m Shoulder deflection (mm) Shoulder lateral displacement (mm) Max thoracic deflection (mm) Max thoracic lateral displacement (mm) Max abdominal deflection (mm) Max abdominal lateral displacement (mm) Pubic force (kn) Airbag deployment (ms) Max thoracic deflection time (ms) 100% 90% A 59-p 80% B * 65*-m 70% 38-m 43-p 25-p 33-p 57-p 60% 42-a 44-m 29-a 32-m 59-m 50% 36-p 46-p 26-p 31-p 44-p 40% 34-p 45-m 26-m 31-a 42-p 30% % % Figures 9-11 show the two-dimensional X-Y displacement of the RibEye LEDs at all three measurement locations on the rib. Only the thoracic and abdominal ribs with the highest deflection are shown. While some of the ribs in Tests A and C showed anterior-to-posterior oblique loading initially, the overall peak displacements were oblique from the posterior-to-anterior direction. Among the three measurement locations on each rib, peak threedimensional deflections were always recorded at the posterior or middle locations (Table 5). But peak lateral displacements were recorded at each of the three locations, and often at a different location than the peak three-dimensional deflection on the same rib. D E 20% 1.3 Table 5. Summary injury measures and timing by test ID. The RibEye sensor locations recording the peak rib deflections and displacements are indicated by: a (anterior), m (middle) or p (posterior). The * indicates either a complete loss of data or a partial loss where the peak value may have been higher. C Thorax Abdomen Shoulder AIS3+ AIS2+ AIS2+ Pelvis AIS2+ Head AIS4+ Figure 8. Injury risks predicted by the WorldSID ATD in each test. The background bars show the risk for a 75-year-old. The thicker foreground bars show the risk for a 45-year-old. (Head injury risks are based on the Hybrid III injury curve and are not adjusted for age.) DISCUSSION Reconstruction vs. Current SICE Configuration The reconstruction of the NASS-CDS case produced generally similar damage patterns to the struck Fit. The measured crush for the striking Camry was less than that measured in the real-world crash, likely due to the required constraint on the test speed. In the real-world crash, the driver sustained fatal thoracic injuries, and in the reconstruction test the thoracic deflections measured with the ATD correlated to a 62 percent risk of AIS 3 injury. In addition, an elevated shoulder force suggests the possibility of other load paths that may have contributed to the injuries observed in the crash. Despite the general agreement between the outcomes in the NASS-CDS case and the test, there were no unique crashworthiness deficiencies identified in the reconstruction test. With the exception of the shoulder, injury metrics to all body Brumbelow 7

15 Figures Displacement of RibEye LEDs in the rib X-Y plane for the shoulder, thoracic, and abdominal ribs, respectively. Only the thoracic and abdominal ribs with the greatest calculated deflection in each test are displayed. Figure 5 illustrates the coordinate system used. regions were lower in this configuration than in the standard SICE configuration for the older Fit design (Tests A and B). The main difference between the two crash modes was the longer crash pulse in the reconstruction test and the later peak loading times. Because the airbag deployed at a similar time, it may have had reduced capacity for energy absorption by the time of peak loading. Furthermore, the forward impact location did not produce a reversal in the predominant direction of the obliquity of rib loading. While there was some movement in the anterior-to-posterior direction early in the crash, the direction had reversed by the time of intrusion and peak deflection. This may at least partially be due to the design of the WorldSID ATD Brumbelow 8

16 ribs. Yoganandan et al. observed displacements in the posterior-to-anterior direction during pure lateral load wall tests of the ATD, but they did see movement in the opposite direction during anterior oblique wall tests [4]. In the current study, video from Test A also suggested that the ATD rotated around the pretensioned seat belt and it is possible that this could produce twisting of the ATD spine about its vertical axis. If the ribs were partially constrained by loading from the airbag and door, such spinal rotation would be equivalent to moving the RibEye LEDs anteriorly relative to the RibEye sensors. Regardless of the explanation, the ATD was not able to identify a potential injury mechanism unique to this alternative crash configuration. A comparison of the reconstruction and SICE tests for the new Fit design (Tests C and D) yields similar conclusions. With the exception of the pubic force, injury measures in the SICE configuration were greater than those in the more forward impact. In the forward impact, the thoracic and abdominal rib deflections lacked even the initial indication of anterior oblique loading that was visible in the test with the old Fit. New Fit vs. Old Fit While the 2007 Fit in the NASS-CDS case was a goodrated design, the paired tests of this design and the design illustrate how crashworthiness improvements have continued beyond the level required to obtain a good rating. This suggests that if there were sufficient cases to replicate the NASSCDS/CIREN study [3] with only the newest vehicle designs, the relevance of specific changes to the SICE test would differ. Specifically, the injury risks that may have been relevant to the occupant in the replicated NASS-CDS case were much lower in the new Fit design, with the risk for a 75-year-old falling from 62 to 5 percent for an AIS3+ thoracic injury and from 68 to 14 percent for an AIS2+ shoulder injury. While limited to a single vehicle design, if this trend held for the rest of the fleet, it is likely that occupants continuing to sustain serious injuries in newer vehicles would be involved in proportionally fewer crashes with forward impact locations and more higher severity compartment. impacts to the occupant Potential SICE Changes As stated above, the test results for these two vehicle designs do not indicate potential value for a crashworthiness evaluation in the more forward impact at 80 km/h. In fact, justifying such an evaluation would have required injury risks that were substantially greater than those observed in the current SICE configuration. This is because there is no indication in the field data that side impacts are more frequently centered forward of the occupant compartment than near the B-pillar. Therefore, a test with an increased speed is most likely to drive meaningful improvements at whatever location currently produces the highest injury risk. Without exception, the 60 km/h impact of the new Fit at the current SICE configuration produced greater injury measures than the 80 km/h more forward impact (Tests E and C). The 60 km/h impact speed in Test E represents a 44 percent increase in impact energy over the SICE test. The published risk curves for a 45-year-old indicated that the increased speed results in a 90 percent greater risk of AIS2+ shoulder injury and a 55 percent greater risk of AIS3+ thoracic injury. Injury risk to the head, abdomen, and pelvis increased by 6 percent or less. Maximum intrusion at the B-pillar increased from 16.9 cm to 23.1 cm. However, this was still less than the intrusion in the 50 km/h SICE test of the older Fit model, and when compared to the precrash centerline of the seat only would have been 1 cm away from a good structural rating. This suggests that a 60 km/h SICE test would encourage more changes to vehicle restraint systems than to structure. While restraint changes may benefit occupants in higher severity crashes, they have a greater potential to induce injuries in lower severity side impacts. Any potential tradeoff would need to be evaluated prior to introducing a higher severity test. The most suitable impact speed for a higher severity test also would require further study. In the NASSCDS/CIREN study, the maximum crush of the occupant compartment in each real-world case was Brumbelow 9

17 compared to the maximum produced in the SICE test of the same vehicle. The cases with greater crush were categorized as being more severe than the test. For cases in this category, the median crush was 56 cm compared with a median of 31 cm in SICE tests of good-rated vehicles [3]. On its own, this would suggest that the 60 km/h test speed used in the current study is still too low to match the majority of real-world crashes producing serious injury. However, the median crush values are another metric that likely would change if the real-world study could be replicated with only the newest generation of vehicles. A different severity metric, such as door intrusion velocity, may be a better predictor of injury, but establishing a real-world baseline would require a large number of case reconstructions through simulated or physical testing. CONCLUSIONS Side impact crashworthiness, as measured in the IIHS SICE test, continues to improve beyond the level required for a good rating. While real-world crashes of different configurations can produce serious injury in good-rated vehicles, the tests conducted for the current study have not demonstrated that a test with a more forward impact configuration would identify unique injury risks. Increasing the impact speed of the current test is more likely to drive continued crashworthiness improvements that are relevant in real-world crashes. However, potential tradeoffs of more aggressive or complex restraint systems would need to be evaluated to minimize any disbenefit in low and moderate severity side impacts. REFERENCES [1] Teoh, E.R. and Lund, A.K. IIHS Side Crash Test Ratings and Occupant Death Risk in Real-World Crashes. Traffic Inj Prev. 2011;12: [2] Insurance Institute for Highway Safety. Fatality Facts: Passenger Vehicle Occupants. Arlington, VA: Author; fatalityfacts. Accessed Feb 22, [3] Brumbelow, M.L.; Mueller, B.C.; and Arbelaez, R.A. Occurrence of Serious Injury in Real-World Side Impacts of Vehicles with Good Side-Impact Protection Ratings. Traffic Inj Prev. 2015;16:S [4] Yoganandan, N.; Humm, J.R.; Pintar, F.A.; and Brasel, K. Region-Specific Deflection Responses of WorldSID and ES2-re Devices in Pure Lateral and Oblique Side Impacts. Stapp Car Crash Journal. 2011;55: [5] PC-Crash. Computer software. Version 9.0. Dr. Steffan Datentechnik. Linz, Austria. [6] Insurance Institute for Highway Safety. Side Impact Crashworthiness Evaluation Crash Test Protocol (Version IX). Ruckersville, VA: Author; [7] Insurance Institute for Highway Safety. Guidelines for Using the UMTRI ATD Positioning Procedure for ATD and Seat Positioning (Version V). Ruckersville, VA: Author; [8] Boxboro Systems. Hardware User s Manual: RibEye Multi-Point Deflection Measurement System 3-Axis Version for the WorldSID 50th ATD. Boxboro, MA: Author; [9] International Organization for Standardization. Road vehicles Injury Risk Curves for the Evaluation of Occupant Protection in Side Impact Tests. Geneva, Switzerland: Author; ISO/TR [10] Mertz, H.J.; Prasad, P.; and Irwin, A.J. Injury Risk Curves for Children and Adults in Frontal and Rear Collisions. Warrendale, PA: Society of Automotive Engineers; SAE Technical Paper Series Brumbelow 10

18 PEER REVIEW PAPER This paper has been peer reviewed and published in a special edition of Traffic Injury Prevention 18(S1), by Taylor & Francis Group. The complete paper will be available on the Traffic Injury Prevention and websites on June 5, To access all Peer-reviewed papers please copy and paste the link below into your browser to access the papers.

19 Comparison of Thorax Responses between WorldSID-5th and SID-IIs in Lateral and Oblique Impacts Miwako Ikeda Hiroyuki Mae Honda R&D Co., Ltd. Automobile R&D Center Japan Paper Number ABSTRACT Recently, enhancing the biofidelity of the WorldSID-5 th percentile adult female dummy (WorldSID-5 th ), which is an acceptable worldwide fifth percentile adult female side impact dummy, has been investigated and incorporating WorldSID-5 th in the GTR no.14 pole side impact as a substitute for SID-IIs is considered. Since the torso design and instrumentation for measuring thorax deflection are different between these two dummies, it is expected that WorldSID-5 th can indicate the improved performance of evaluating thorax injuries. The aim of this study was to clarify a difference of performance in evaluating severity of thorax injuries between WorldSID-5 th and SID-IIs by comparing thorax responses in lateral and oblique impacts. In order to understand deformations of ribs, thorax impact simulations were conducted by using WorldSID-5th small female dummy FE model v2.0.3 and SID-IIs dummy FE model SBLD v3.3.2, which are developed by Humanetics Innovation Solutions Inc. A kilogram pendulum with mm face was impacted into two dummies at the speed of 4.3 and 2.0 m/s, similar to the biofidelity test for thorax without arm shown in 49 CFR Part 572, Subpart V. The centerline of the pendulum was aligned at the level of the centerline of the middle thorax rib in the most lateral side of each dummy. The directions of impacts were set to 0 (pure lateral), ±5, ±10 and ±15. Results from SID-IIs simulations in both high and low speed impacts showed that a thorax deflection measured by potentiometers in pure lateral loading is larger than that in oblique loadings. In contrast, thorax deflections measured by 2D IR-Tracc from WorldSID-5 th simulations in high speed impacts were generally constant with loading directions, those in low speed impacts in pure lateral loading are smaller than that in oblique loadings. According to published papers, it is known that human thorax response shows larger deflections in the anterolateral oblique loadings than that in the pure lateral loadings. Therefore, WorldSID-5 th is supposed to be able to represent characteristics of human thorax more adequately compared to SID-IIs. Since human thorax response in postero-lateral oblique impacts has not been thoroughly investigated, further validation of WorldSID-5 th will be needed. It was clarified that WorldSID-5 th can represent human characteristics of thorax response more appropriately than SID-IIs. Furthermore, it was shown that SID-IIs has a possibility of underestimating thorax deflection in oblique impacts. Therefore, it can be expected that the vehicle performance of occupant protection will be enhanced by introducing WorldSID-5 th into side impact test protocols sometime in future. Ikeda 1

20 INTRODUCTION In the United States, according to Fatality Analysis Reporting System (FARS) provided by National Highway Traffic Safety Administration (NHTSA), the number of passenger vehicle occupant fatalities in 2015 was decreased by 26.9% compared with that in Although the number of passenger fatalities was dropped by 31.2%, that of driver fatalities was only reduced by 23.8% in these ten years [1]. The analysis of fatality and serious injury rate of driver by using National Automotive Sampling System General Estimates System (NASS-GES) provided by NHTSA [2] shows that a decrease of the fatal or serious injured driver rate in side crash accidents seems to be small compared to that in frontal crash accidents (Figure 1). In addition, the number of fatal or serious injured drivers in Vehicle-to-Pole/Tree type side crash accidents is only 2,479, while that in Vehicle-to-Vehicle type side crash accidents is 17,414 in However, the fatal or serious injured driver rate in Vehicleto-Pole/Tree accidents is 6.9% while that of Vehicle-to-Vehicle accidents is 1.0%. This suggests that mitigating the number of fatal or serious injured drivers in Vehicle-to-Pole/Tree accidents must be focused on, as well as that in Vehicle-to-Vehicle accidents. Rate (%) Frontal FRONTAL crash Side SIDE crash Calendar year Figure 1. Fatal or serious injured driver rate in U.S. It is known that the distribution of direction of force in Vehicle-to-Pole/Tree accidents in which occupants sustaining AIS3+ injuries shows that the pure lateral accounts for 50.8% and the anterolateral oblique accounts for 40.0%, respectively [3]. Additionally, thorax is the most frequent severe injured body region in Vehicle-to-Pole/Tree accidents [4, 5]. For this reason, not only human thorax responses against pure lateral impacts but also those against antero-lateral oblique impacts have been investigated. Shaw et al. [6] conducted thorax impact tests by using seven Post Mortem Human Subjects (PMHSs) in which a kilogram pendulum impacted to the level of the forth interspace of the sternum at the speed of 2.5 m/s. Based on results from seven pure lateral impact tests and seven antero-lateral oblique impact tests, corridors of thorax force-deflection responses for each two impact configurations were developed. The comparison of the averaged maximum forces and the averaged maximum deflections between those two corridors shows that the averaged maximum force in the pure lateral impact is larger than that in the antero-lateral oblique impact; in contrast, the averaged maximum deflection in antero-lateral oblique impact is larger than that in pure lateral impact. Baudrit et al. [7] conducted twelve thorax impact tests in which a 23.4-kilogram pendulum impacted to the middle of the sixth rib of PMHSs at the speed of 4.2 to 4.4 m/s in pure lateral directions and antero-lateral oblique directions. Based on these results, four thorax force-deflection corridors by combination of two physical sizes and two impact directions were developed; the 50 percentile adult male and the 5 percentile adult female; pure lateral and antero-lateral oblique. Similar to results from Shaw et al., it was shown that the averaged maximum force in a pure lateral impact is larger than that in antero-lateral oblique impacts, and the averaged maximum deflection is larger than that in pure lateral impacts. In the aim of mitigating occupant injuries in real world side crash accidents, side impact test protocols have been introduced. There are two principally different test configurations for side impact tests. One is called the Moving Deformable Barrier test (MDB test) simulating a crash accident where the vehicle is collided by the other vehicle in its side. The other is called Pole test simulating a crash accident where a vehicle collides into a utility pole or tree. In United States, those tests are introduced by legal requirements FMVSS214 and the consumer information tests U.S. new car assessment program (U.S. NCAP) and Insurance Institute for Highway Safety. In order to assess severities of occupant injuries, Anthropometric Test Devices (ATDs) have been developed. ES2-re and SID-IIs, which were introduced by FMVSS214 NPRM released on May 2004, are used in side crash tests introduced presently in United States. As for the replacement of those ATDs, WorldSID-50 th adult male dummy (WorldSID-50 th ), developed by ISO task group in Ikeda 2

21 1997, is planned to be introduced in the future U.S. NCAP protocol [8]. Moreover, introducing of WorldSID-5 th adult female dummy (WorldSID-5 th ) which has been developed by WorldSID 5 th TEG in the GTR pole test is considered [9]. Each rib of ES2-re and SID-IIs which are adopted in current side crash test protocols is designed to represent a pair of human s left and right rib by using one rib. Thorax deflection selected as an index for evaluating thorax injuries is measured as a unidirectional deflection between the left and right sides of rib for ES2-re, and a unidirectional deflection between the most lateral side of rib and the spine for SID-IIs. By contrast, WorldSID-50 th and WorldSID-5 th have been designed as a more human-like thoracic structure, ribs are separately into left and right ribs whose anterior end is connected to the sternum and posterior end is connected to the spine, respectively. Thorax injury measure of WorldSID-50 th which is specified in Euro NCAP s protocol is a lateral deflection calculated by using outputs measured by 2D Infra- Red Telescoping Rod for the Assessment of Chest Compression (IR-Tracc). 2D IR-Tracc is capable of measuring a change of a distance between the most lateral point of the rib and the spine, and a change of an angle at the most lateral point of the rib relative to the spine. Then, the lateral deflection is defined as a pure lateral compression of the rib calculated in terms of these two measurements. Hence, it can be said that a performance of evaluating severities of thorax injuries is different between current ATDs and modern ATDs; ES2-re and SID-IIs; WorldSID-50 th and WorldSID-5 th. Yoganandan et al. [10] compared thorax responses of ES2-re and WorldSID-50 th in pure lateral and oblique side impact loadings by conducting fullscale sled tests. The result shows that WorldSID- 50 th better sensed the oblique loading than ES2-re. However, thorax responses from 5 percentile female dummies; WorldSID-5 th and SID-IIs have not been compared. The objective of this study was to clarify a difference of performances of thorax injury evaluation between WorldSID-5 th and SID-IIs by comparing patterns of rib deformation and thorax injury values. THORAX RESPONSES IN DIFFERENT ANGLE IMPACTS Thorax Impact Simulation Since rib components of full-scale physical dummies are covered with jackets, it is physically impossible to obtain patterns of whole rib s deformation. Therefore, LS-Dyna R6.1.2 finite element (FE) simulations by using WorldSID5th Small Female Dummy v2.0.3 [11] and SID-IIs dummy SBL D v3.2.2 [12] developed by Humanetics Innovative Solutions, Inc. were conducted in order to capture patterns of rib deformation located inside ATDs. Because it is known that a difference of arm positions affects values of thorax deflection [13], thorax without arm impact test s configuration similar to that shown in 49 CFR Part 572 Subpart V [14] was selected in this study. The seatback of a certification bench was cut off at the height of 300 mm in order not to interfere with a pendulum s movement and modeled as a rigid surface. A WorldSID-5 th, while raising the arm to a vertical orientation, was seated on the bench in order that the top of the lower neck bracket was horizontal, and its pelvic tilt sensor showed 19.5 degrees. SID-IIs removed its arm was seated on the bench in order that the thoracic fore/aft plane measured 24.6 degrees and the back of the thorax touched the seatback. It was estimated that no friction force is generated in physical tests because the seat back and base is covered with PolyTetraFlourEthylene sheets. Therefore, a coefficient of friction force of contact characteristic between the bench and the dummy was set to zero in order that the dummy model can glide over the bench model smoothly. A circular cylindrical pendulum was modeled as a rigid surface with a mm face diameter and a 12.7 mm edge. A kilogram mass was applied at the center of the shape. The pendulum was made to collide with the dummy at 4.3 m/s similar to the speed specified in 49 CFR Part 572 Subpart V, or 2.0 m/s which is an estimated impact speed that induces negligible thorax deflection. As for the relative location between the dummy and the pendulum, the height of the center of the pendulum s face was aligned to the height of the centerline of the middle thoracic rib at the most lateral side of the dummy. In the pure lateral impact simulation, the pendulum was positioned so that its centerline was centered vertically on the centerline of the middle thoracic rib. Setups of thorax impact simulation for WorldSID-5 th and SID-IIs in pure lateral impact are shown in Figure 2. As shown in Figure 3, the probe was rotated by ±5, ±10 and ±15 relative to the center of the Ikeda 3

22 spine box in each dummy in an antero-lateral or a postero-lateral oblique impact. Thorax impact simulations were carried out by impacting WorldSID-5 th or SID-IIs FE model with a pendulum model. Seven impact directions, two impact speeds and two dummy models were combined to create twenty eight impact simulations. WorldSID-5 th SID-IIs CFC600, then lateral deflection was calculated in accordance with WorldSID-5 th physical dummy manual [15] by using equations 1 to 3. Symbols used in above equations are shown in Figure 4 and Table 1, in the way of Y direction representing ATD s lateral direction and X direction representing ATD s fore/after direction. Figure 5 show comparisons of outputs between results from simulation and physical tests of WorldSID-5 th. In accordance with SID-IIs physical dummy manual [16], an impact force calculated as in the case with WorldSID-5 th, lateral accelerations at T1 and T12 filtered at CFC180, output of potentiometer for each rib filtered at CFC600, were compared. Figure 6 shows the comparison of outputs between results from simulation and physical tests of SID-IIs. Figure 2. Setups of thorax impact simulation for WorldSID-5t h and SID-IIs. WorldSID-5 th (Top view) SID-IIs (Top view) 0 0 Figure 3. Impact directions for pure lateral and oblique impacts. Comparison between Physical Dummies and FE Dummy Models in Lateral Impacts In order to confirm accuracies of thorax responses from the results of FE simulations, thorax impact tests using physical dummies were conducted and results from FE simulations were compared to those from physical tests. Thorax without arm impact test in pure lateral direction at the speed of 4.3 m/s was selected as an impact configuration for this comparison because this is the configuration specified in 49 CFR Part 572 Subpart V [14]. Two physical tests for each dummy were conducted. An Impact force, lateral accelerations at T4 and T12, and lateral deflections of thorax were compared for WorldSID-5 th. An Impact force was calculated by multiplying a longitudinal acceleration of the pendulum filtered at CFC180 by its weight. Time histories of lateral accelerations at T4 and T12 were filtered at CFC180. As for the lateral deflection of thorax, time histories of compression and rotation from each 2D IR-Tracc s output were filtered at Figure 4. Symbols used in equations for calculating lateral deflection [15]. Table 1. Calculation- parameters, symbols, and description [15] Parameter Description t 0 [s] Time zero L 0 [mm] Reference length at t 0 Dyi [mm] IR-Tracc compression at ti ϕxyi [degrees] IR-Tracc angle at time i (positive angle indicated) X [mm] Calculated x displacement w.r.t x 0 (time zero x) Y [mm] Calculated y displacement w.r.t y 0 (time zero y) R [mm] Calculated resultant displacement w.r.t R 0 (time zero R) x y L0 sin xyi L L cos (Equation 1) i d yi (Equation 2) i 0 0 d yi i 2 i 2 i R x y (Equation 3) xyi Ikeda 4

23 Simulation Physical test #1 Physical test #2 Physical #1 CFC180 Impact F Physical #1 DRIV. T-4 Y Physical #1 DRIV. T-12 Y Physical #2 DRIV. T-12 Y Physical Impact #2 CFC180 ForceImpact F Physical T4 #2 Lateral DRIV. T-4 Acceleration Y T12 Lateral Acceleration Force (N) Acc. (G) Acc. (G) Time (msec) Time (msec) Time (msec) Physical #1 CFC600 Rib1 2D Physical #1 CFC600 Rib2 2D Physical #1 CFC600 Rib3 2D DefY Physical #2 CFC600 Rib1 2D Upper Thoracic Rib Deflection Middle Thoracic Physical #2 CFC600 Rib3 2D DefY DefY Rib Deflection Lower Thoracic Deflection CAE_WS05 # m/s Calc Def Rib3 mm Def. (mm) Def. (mm) Def. (mm) Time (msec) Time (msec) Time (msec) Figure 5. Comparison of results between simulation and physical tests of WorldSID-5 th. Force (N) #6507_ Filtered T1 spine accel. Impact Force Acc. (G) #6507_ Filtered T1 spine accel. Acc. (G) 20 T1 Lateral Acceleration Simulation Physical test #1 Physical test #2 10 #6507_ Filtered T12 spine accel. T12 Lateral Acceleration Time (msec) Time (msec) Figure 6. Comparison of results between simulation and physical tests of SID-IIs Time (msec) #6507_ Filtered #6507_ Filtered #6507_ Filtered Upper Thoracic Upper rib Rib disp. Deflection Middle Thoracic Middle rib Rib disp. Deflection Lower Thoracic Lower rib Rib disp. Deflection Def. (mm) Def. (mm) Def. (mm) Time (msec) Time (msec) Time (msec) Although time histories of results from SID-IIs simulation matched well with those from physical tests, only maximum levels of each output from WorldSID-5 th simulation matched to those from physical tests. In addition, comparing time histories from physical tests between WorldSID-5 th and SID-IIs (Figures 5 and 6), it seems that WorldSID-5 th has a possibility to have a poor repeatability. For this reason, parametric study by conducting simulation was selected in this study. It is known that the thoracic component of SID-IIs FE model is validated in terms of oblique impacts [17]. However, those validations for WorldSID-5 th FE model have not been reported yet. Therefore, results from WorldSID-5 th simulation were compared to those of physical tests from the published study in which thorax impact tests similar to the simulation in this study were shown. Been et al. [18] conducted thorax impact tests with WorldSID-5 th revision 1 dummy where the head, Ikeda 5

24 Lateral side (mm) arm and jacket were removed. The dummy was seated on the platform, and impacted by a kilogram pendulum at the speed of 2.5 m/s in the antero-lateral oblique impact (+15 degree) and the postero-lateral oblique impact (-15 degree). The wooden block was fitted to the front of the pendulum so that the first contact point was the most lateral aspect of the upper thoracic rib. Since the heights of impact level were different between the tests and this study, lateral deflections of the upper thoracic rib in Been et al. and the middle thoracic rib in this study were selected as outputs used in a comparison for thorax deflection. Table 2 shows the comparison of the maximum impact forces, and Table 3 shows the comparison of the maximum thoracic lateral deflections. Table 2. Comparison of impact force between results from physical tests (Been et al. [18]) and CAE simulations (this study) Impact Direction Antero-lateral (15 degree) Pure lateral (0 degree) Postero-lateral (-15 degree) Been et al. This study 909 N 1599 N 904 N 2125 N 835 N 1511 N Table 3. Comparison of lateral deflections between results from physical tests (Been et al. [18]) and simulations (this study) Impact Direction Antero-lateral (15 degree) Pure lateral (0 degree) Postero-lateral (-15 degree) Been et al. This study 27.8 mm 8.5 mm 29.5 mm 9.8 mm 18.1 mm 7.2 mm The weights of pendulums in both studies were similar. However, the impact speeds were faster in the simulations than in the physical tests, and the impact forces applied to the dummy were more concentrated in the physical tests than in the simulations. For this reason, the levels of impact forces were thought to be higher in the simulation, and the levels of deflections were thought to be higher in the physical tests. Nevertheless, both results of physical tests and simulations show higher impact forces in oblique impacts and higher deflections in pure lateral impacts. Therefore, it is qualitatively confirmed that WorldSID-5 th FE model used in this study can estimate a response of physical WorldSID-5 th ATD. Thorax Responses from WorldSID-5 th Simulation Trajectories of the most lateral points of each inner rib relative to the spine box for 4.3 m/s impact simulations were shown in Figure 7, in which red lines show trajectories at upper thoracic ribs, yellow lines show those at mid thoracic ribs and green lines show those at lower thoracic ribs, respectively. Figure 8 shows the deformations of the middle rib in 15 antero-lateral impact, pure lateral impact and -15 postero-lateral oblique impact at 4.3 m/s impact simulations of WorldSID- 5 th. Anterior 60 Upper rib Middle rib Lower rib Anterior side (mm) Lateral Left (mm) Figure 7. Trajectories of most lateral points of thoracic ribs of WorldSID-5 th in 4.3m/s impacts. Additionally, Figures 9 to 11 show time histories of compressions and rotations from 2D IR-Tracc, and Figures 12 to 14 show time histories of lateral deflections and impact forces, in the cases of 4.3 m/s impacts in 15 antero-lateral impact, pure lateral impact and -15 postero-lateral oblique impact, respectively. 0 Ikeda 6

25 0 ms 4 ms 8 ms 12 ms 16 ms 20ms Antero-lateral impact (+15 ) Pure lateral impact (0 ) Postero-lateral impact (-15 ) Figure 8. Middle thoracic rib deformations of WorldSID-5 th in 4.3m/s impact (top view). Upper rib, compression Middle rib, compression Lower rib, compression Upper rib, angle Middle rib, angle Lower rib, angle Upper rib, compression Middle rib, compression Lower rib, compression Upper rib, angle Middle rib, angle Lower rib, angle Comp. (mm) Time (msec) Angle (deg.) Comp. (mm) Angle (deg.) Time (msec) Figure 9. Time histories of IR-Tracc outputs in 4.3m/s, antero-lateral oblique impacts (15 ). Upper rib, compression Middle rib, compression Lower rib, compression Comp. (mm) Upper rib, angle Middle rib, angle Lower rib, angle Time (msec) Angle (deg.) Figure 10. Time histories of IR-Tracc outputs in 4.3m/s, pure lateral impacts (0 ) Figure 11. Time histories of IR-Tracc outputs in 4.3m/s, postero-lateral oblique impacts (-15 ) Def. (mm) Upper rib, Lateral deflection Middle rib, Lateral deflection Lower rib, Lateral deflection Impact force Time (msec) Force (N) Figure 12. Time histories of lateral deflections and force in 4.3m/s, antero-lateral oblique impacts (15 ). 0 Ikeda 7

26 Lateral side (mm) Def. (mm) Upper rib, Lateral deflection Middle rib, Lateral deflection Lower rib, Lateral deflection Impact force Time (msec) Force (N) Figure 13. Time histories of lateral deflections and force in 4.3m/s, pure lateral impacts (0 ) Def. (mm) Upper rib, Lateral deflection Middle rib, Lateral deflection Lower rib, Lateral deflection Impact force Time (msec) Figure 14. Time histories of lateral deflections and force in 4.3m/s, postero-lateral oblique impacts (-15 ) In the antero-lateral oblique impact, compression and angle output of each 2D IR-Tracc reach its maximum values almost simultaneously (Figure 9). For this reason, the most lateral points of ribs draw sharp edges when those outputs reach to their maximum values (Figure 8). Because the ATD s sternum displaces on the anterior side after the rib was compressed in the postero-lateral oblique impact (Figure 7), the time when the angle output reaches its maximum value occurs later than the time when the compression output reaches its maximum value (Figure 11). Therefore, the most lateral points of ribs move in the large range (Figure 7). In addition, trajectories of the upper, middle, lower ribs in same loading condition are quite different. This suggests that each rib moves individually. 0 Force (N) Thorax Responses from SID-IIs Simulation Trajectories of the end points of each potentiometer relative to the spine for 4.3 m/s impact simulations were shown in Figure 15, in which red lines show trajectories at upper thoracic ribs, yellow lines show those at mid thoracic ribs and green lines show those at lower thoracic ribs, respectively. Figure 16 shows the deformations of the middle rib in 15 antero-lateral impact, pure lateral impact and -15 postero-lateral oblique impact at 4.3 m/s impact simulations for SID-IIs. Anterior 40 Upper rib Middle rib Lower rib Anterior side (mm) Lateral Left (mm) Figure 15. Trajectories of most lateral points of thoracic ribs of SID-IIs in 4.3m/s impacts. Figures 17 to 19 show time histories of thoracic deflections which are resultant deflections measured by potentiometers and are specified as a thorax injury measure for SID-IIs, and calculated impact forces, in cases of 4.3 m/s impacts in 15 antero-lateral impact, pure lateral impact and -15 postero-lateral oblique impact, respectively. Trajectories of the end points of each potentiometer in same load direction show similar shape (Figure 16). In addition, time histories of thoracic deflections in the upper, middle and lower ribs change their values uniformly (Figure 17 to 19). For this reason, it seems that three thoracic ribs deform with conjunction with each other. Ikeda 8

27 0 ms 5 ms 10 ms 15 ms 20 ms 25ms Antero-lateral impact (+15 ) Pure lateral impact (0 ) Postero-lateral impact (-15 ) Figure 16. Middle Thoracic rib deformation of SID-IIs in 4.3m/s impact (top view). Upper rib, Deflection Middle rib, Deflection Lower rib, Deflection Def. (mm) Impact force Force (N) Upper rib, Deflection Middle rib, Deflection Lower rib, Deflection Def. (mm) Impact force Force (N) Time (msec) Figure 17. Time histories of SID-IIs outputs in 4.3m/s, antero-lateral impacts (15 ) Upper rib, Deflection Middle rib, Deflection Lower rib, Deflection Def. (mm) Impact force Time (msec) Force (N) Figure 18. Time histories of SID-IIs outputs in 4.3m/s, pure lateral impacts (0 ) Time (msec) Figure 19. Time histories of SID-IIs outputs in 4.3m/s, postero-lateral impacts (-15 ) DISCUSSION Comparison of Normalized Deflections between WorldSID-5 th and SID-IIs Since levels of thoracic deflections are different between WorldSID-5 th and SID-IIs, even under same impact speed and same impact direction, all of the output values are normalized by using its values at pure lateral impact in each combination of impact speed and dummy. Figures 20 to 25 show comparisons of normalized values for thoracic deflections of the upper, middle and lower thoracic ribs, the averaged thoracic deflections between three ribs, the maximum thoracic deflections between three ribs and the maximum impact forces. Ikeda 9

28 SID-IIs, 4.3 m/s SID-IIs, 2.0 m/s Impact direction (deg.) WorldSID-5 th, 4.3 m/s WorldSID-5 th, 2.0 m/s Def. ( - ) Def. ( - ) Figure 20. Comparison of normalized deflection of upper thoracic rib. SID-IIs, 4.3 m/s SID-IIs, 2.0 m/s Impact direction (deg.) WorldSID-5 th, 4.3 m/s WorldSID-5 th, 2.0 m/s Def. ( - ) Def. ( - ) Figure 21. Comparison of normalized deflection of middle thoracic rib. SID-IIs, 4.3 m/s SID-IIs, 2.0 m/s Impact direction (deg.) WorldSID-5 th, 4.3 m/s WorldSID-5 th, 2.0 m/s Def. ( - ) Def. ( - ) Figure 22. Comparison of normalized deflection of lower thoracic rib. SID-IIs, 4.3 m/s SID-IIs, 2.0 m/s Impact direction (deg.) WorldSID-5 th, 4.3 m/s WorldSID-5 th, 2.0 m/s Def. ( - ) Def. ( - ) Figure 23. Comparison of normalized average deflection between three thoracic ribs. Max(RIB1-3) SID-IIs, 4.3 m/s SID-IIs, 2.0 m/s Impact direction (deg.) MAX(RIB1-3) WorldSID-5 th, 4.3 m/s WorldSID-5 th, 2.0 m/s Def. ( - ) Def. ( - ) Figure 24. Comparison of normalized maximum deflection between three thoracic ribs. SID-IIs, 4.3 m/s SID-IIs, 2.0 m/s Norm WorldSID-5 th, 4.3 m/s Impact direction (deg.) WorldSID-5 th, 2.0 m/s Force ( - ) Force ( - ) Figure 25. Comparison of normalized impact force. Ikeda 10

29 The upper, middle and lower thoracic ribs of SID- IIs are connected to the same part called the upper bib-ribs, and the thorax pad covers the upper bibribs with cable tie wraps (Figure 26). For this reason, all three thoracic ribs deform in conjunction with each other, and this results in that comparison of three ribs shows same tendency like that thoracic deflections in pure lateral loadings show generally larger than that in oblique impacts (Figures 20 to 22). 22), on the other hand, the deflection of the upper thoracic rib in the postero-lateral oblique impact is 1.5 times larger than that in the pure lateral loading (Figure 20). This suggests that a rib component of WorldSID-5 th is easy to deform in postero-lateral impacts. Remove jacket Remove jacket and thorax pad Remove jacket Remove jacket and thorax pad Velcro Thorax pad Cable tie wrap Thorax pad Figure 27. WorldSID-5 th Thorax Component. Upper bib-ribs Figure 26. SID-IIs Thorax Component. Although, a comparison of the middle thoracic rib deflection for WorldSID-5 th (Figure 21) shows same tendency as that for SID-IIs, deflection of the upper thoracic rib decreases as the dummy is impacted in more anterior direction (Figure 22). In addition, that of the lower thoracic rib decreases as the dummy is impacted in more posterior direction (Figure 23). As for the design of assembling thorax component of WorldSID-5 th, lateral sides of three thoracic ribs and two abdominal ribs are only connected to the thorax pad by using Velcro (Figure 27). Since, lateral sides of ribs are not connected firmly, thoracic ribs of WorldSID-5 th seem to be able to deform independently. Although the anterior and the posterior ends of inner ribs and the posterior ends of outer ribs are rigidly connected to the same spine box, the anterior ends of outer ribs for left and right thorax are only linked to the sternum, which is divided by each rib location. For this reason, the anterior part of the outer rib of the right thorax moves forward along with the anterior part of the outer rib of left thorax, especially in the case that left thorax is applied in the postero-lateral oblique loading. The pendulum initially engaged with the upper thoracic rib in postero-lateral oblique impacts (Figure 11). By contrast, it is initially engaged with the lower thoracic rib in antero-lateral oblique impacts (Figure 9). The deflection of the lower thoracic rib in the antero-lateral oblique impact is 1.2 times larger than that in the pure lateral loading (Figure In comparisons of SID-IIs, deflections in pure lateral loadings show the largest deflection both in the comparison of the averaged and the maximum rib deflection (Figures 24 and 25). In those of WorldSID-5 th, deflections in pure lateral impacts show the largest deflection in the comparison of the averaged rib deflections (Figure 24), the deflections in oblique impacts are as large as or equal to that in pure lateral impacts in the comparison of the maximum rib deflection (Figure 25). A thoracic deflection is not included as an injury measure in the current protocol of neither FMVSS214 nor U.S. NCAP, however, the maximum thoracic rib deflection is introduced as an injury measure in the future U.S. NCAP protocol [8]. If WorldSID-5 th is introduced as a dummy instead of SID-IIs and the maximum thoracic deflection is selected as an injury measure in the future, thorax injuries seem to be evaluated more severely compared to the present. Lateral Component of SID-IIs Thoracic Deflection A thoracic rib deflection of SID-IIs is specified as a unidirectional deflection between the most lateral point of the rib and the spine box. In contrast, that of WorldSID-5 th is specified as a lateral component of deflection between them. Since it is possible that the difference of measurements causes the difference of characteristics of thorax responses between SID-IIs and WorldSID-5 th, lateral components of thoracic deflection are additionally measured from results of SID-IIs simulations. Those outputs can be measured in physical dummy tests by using an optical system Ikeda 11

30 named RibEYE which can measure three dimensional movements of ribs relative to the spine [19]. Figure 28 shows the comparison of normalized averaged deflection between original thoracic deflection and lateral component of thoracic deflection. Deflections used as denominators in normalization were, 16.6 mm for the original deflection and 16.6 mm for the lateral component of thoracic deflection in 2.0 m/s, 42.8 mm for the original deflection and 43.2 mm for the lateral component of thoracic deflection in 4.3 m/s, respectively. All of the deflections used as denominators are output of the middle thoracic ribs. Original (unidirectional deflection) Lateral component of thoracic deflection Def. ( - ) Def. ( - ) Figure 28. Comparison of Normalized Maximum deflection between three thoracic ribs of SID-IIs. Although each lateral component of thoracic deflection shows larger deflection compared to the original deflection in each impact configuration, lateral deflections in oblique impacts show smaller than or equal to that in the lateral impact in both impact speeds. This suggests that a difference of thorax responses between SID-IIs and WorldSID- 5 th is not because of the difference of measured physical quantities but the difference of thoracic design. Biofidelity Evaluation 2.0 m/s 4.3 m/s Impact direction (deg.) In order to clarify whether WorldSID-5 th or SID- IIs can represent more human-like thoracic response, results from this study were compared against the published data. Shaw et al. [6] conducted thorax impact tests in which a 23.8-kilogram pendulum impacted to the level of the forth interspace of the sternum at the speed of 2.5 m/s. Based on the results from seven pure lateral impact tests and seven antero-lateral oblique by 30-degree impact tests, corridors of force-deflection responses for two impact directions were developed. Average values of the maximum thoracic deflection and an impact force scaled into the midsized adult male show that a thoracic deflection in the antero-lateral oblique impact is 1.27 times as large as that in the pure lateral impact, and an impact force in the anterolateral impact is 0.72 times as large as that in the pure lateral impact. As for the thorax response in high-speed impacts, Baudrit et al. [7] conducted twelve thorax impact tests in which a 23.4-kilogram pendulum impacted to the level of the middle of sixth rib at the speed of 4.2 to 4.4 m/s. Then, four corridors of thorax responses by combinations of two physical sizes and two impact directions were developed; 50 percentile adult male and 5 percentile adult female; pure lateral loadings and antero-lateral loadings by 30-degree. Based on the averaged responses for 5 percentile adult female, the maximum thoracic deflection in the antero-lateral oblique impact is 1.25 times as large as that in the pure lateral impact, the maximum impact force in the anterolateral oblique impact is 0.8 times as large as that in the pure lateral impact. Proportions of the maximum thoracic deflection or impact force in antero-lateral oblique impacts to those in pure lateral impacts shown in Shaw et al., Baudrit et al. and results from simulation in this study are compared in Table 4. Table 4. Proportion of maximum thorax deflection or maximum impact force in antero-lateral oblique impacts to that in pure lateral impacts Source PMHS (Shaw et al.) PMHS (Baudrit et al.) WorldSID-5 th (this study) SID-IIs (this study) Impact velocity Proportion deflection force 2.5 m/s m/s m/s m/s m/s m/s In both of the impact speeds, simulation results for SID-IIs show a smaller deflection and a larger impact force in antero-lateral oblique impacts than Ikeda 12

31 in pure lateral impacts. By contrast, those for WorldSID-5 th show a larger deflection and a smaller impact force in antero-lateral oblique impacts than in pure lateral impacts. Since the impact angle used in both of the PMHSs studies was 30 degrees and it is larger than the impact angle used in the simulation of this study, proportions should be compared qualitatively. However, it can be said that proportions for WorldSID-5 th are more similar to those from PMHSs studies than SID-IIs. Consequently, it can be said that WorldSID-5 th can represent more human-like thoracic responses than SID-IIs. LIMITATION At present, WorldSID 5 th TEG has a plan to enhance the biofidelity of WorldSID-5 th female dummy, and the modification of thoracic design has been discussed. However, the basis of its design is not supposed to be a major modification. Therefore, it can be asserted that WorldSID-5 th can represent more human-like thoracic response compared to SID-IIs in future. There is a limitation of published data showing human thoracic responses against various impact directions, the biofidelity evaluation in this study is limited to responses in pure lateral and anterolateral oblique impacts. Accordingly, a biofidelity of WorldSID-5 th in postero-lateral impacts must be evaluated in the future. CONCLUSIONS In this study, thorax impact simulations were conducted by varying impact speeds and directions. As a result, the following conclusions were reached; Three thoracic ribs in SID-IIs tends to deform in conjunction with eatch other, by contrast, those in WorldSID-5 th deform independently. SID-IIs shows similar values in the maximum thoracic deflection and the averaged thoracic deflection. However, the maximum thoracic deflection in WorldSID-5 th shows larger values compared to the averaged thoracic deflection. SID-IIs has a possibility to underestimate the severities of thorax injuries in oblique impacts regardless of a method of mesurement compared to WorldSID-5 th. Based on a proportion of a thoracic deflection and an impact force in the antero-lateral oblique impact to that in the pure lateral impact, it can be said that WorldSID-5 th represent human chraracteristics of thorax reseponse more adequately than SID-IIs. REFERENCES [1] National Highway Traffic Safety Administration, , Fatality Analysis Reporting System (FARS) [2] National Highway Traffic Safety Administration, , National Automotive Sampling System General Estimates System (NASS-GES) [3] Zaouk, A. K., Eigen, A. M., Digges, K. H., 2001, Occupant Injury Patterns in Side Crashes, SAE Technical Paper [4] Pinter, F. A., Maiman, D. J., Yoganandan, N., 2007, Injury Patterns in Side Pole Crashes, 51th Annual Proceedings Association for the Advancement of Automotive Medicine [5] Samaha, T. T., Elliott, D. S., 2003, NHTSA Side Impact Research: Motivation for Upgraded Test Procedures, Paper No. 492, Proceedings of the 18 th Conference on the Enhanced Safety of Vehicles (ESV) [6] Shaw, J. M., Herriott, R. G., McFadden, J. D., Donelly, B. R., Bolte, J. H., 2006, Oblique and Lateral Impact Response of the PMHS Thorax, 50 th Stapp Car Crash Conference, Pp [7] Baudrit, P., Trosseille, X., 2015, Proposed Method for Development of Small Female ad Midsize Male Thorax Dynamic Response Corridors in Side and Forward Oblique Impact Tests, 59 th Stapp Car Crash Conference, Pp [8] National Highway Traffic Safety Administration, Docket No. NHTSA [9] UN GTR No Pole side impact (ECE/TRANS/180/Add.14) [10] Yoganandan, N., Humm, J. R., Pinter, F., A., Brasel, K., 2011, Region-Specific Deflection Responses of WorldSID and ES2-re Devices in Pure Lateral and Oblique Side Impacts, 55 th Stapp Car Crash Conference, Pp [11] Humanetics Innovative Solutions, Inc., 2012, SID-IIs Dummy LS-DYNA Model Version User s Manual Ikeda 13

32 [12] Humanetics Innovative Solutions, Inc., 2016, WorldSID Small Female Dummy LS- DYNA Model Version [13] Kemper, A. R.,McNally, C., Kennedy, E. A., Manoogian, S. J., Duma, S. M., 2008, The Influence of Arm Position on Thoracic Response in Side Impacts, 52 nd Stapp Car Crash Conference, Pp [14] National Highway Traffic Safety Administration, 2006, 49 CFR Part 572, Subpart V, Section SID-IIs D Side Impact Crash Test Dummy, Small Adult Female [15] Humanetics Innovative Solutions, Inc., 2015, User Manual WorldSID Small Female [16] Humanetics Innovative Solutions, Inc., 2011, User Manual SID-IIs Small Side Impact Dummy (SBL D) [17] Humanetics Innovative Solutions, Inc., 2015, SID2s SBLD Dummy Model LS-DYNA Release Version 4.0 [18] Been, B., Waagmeester, K., Trosseille, X., Carroll, J., Hynd, D., 2009, WorldSID Small Female Two-Dimensional Chest Deflection Sensors and Sensitivity to Oblique Impact, Paper No, , Proceedings of the 21 st Conference on the Enhanced Safety of Vehicles (ESV) [19] Jensen, J., Berliner, J., Bunn, B., Pietsch, H., Handman, D., Salloum, M., Charlebois, D., Tylko, S., 2009, Evaluation of an Alternative Thorax Deflection Device in the SID-IIs ATD, Proceedings of the 21 st Conference on the Enhanced Safety of Vehicles (ESV) Ikeda 14

33 An Evaluation Method of Cervical Spinal Injury for Car Passenger in Dynamic Rollover Using a Human FE model Yasuhiro Dokko Shigeo Tobaru Honda R&D Co., Ltd. Japan Kazuki Ohashi Honda Technofort Co., Ltd. Japan ABSTRACT Although the electronic stability control devices have reduced the number of dynamic rollover accidents, it still occupies non-negligible portion of the traffic accidents with fatality and severe injuries. The principal body region of fatal or severe injury in dynamic rollover is cervical spine, while there have been no recognized injury criteria of cervical spinal injuries using existing ATDs for such a loading condition. In this study, the authors tried to establish the method to evaluate cervical spinal injury of the car passengers in dynamic rollover using a human FE model. The human FE model that the authors had developed to be capable of predicting whole body kinematics and the injuries on thorax, lumbar spine, and lower extremities of car occupant in frontal and side impact was adopted as a baseline model. Since the cervical spine part of the model had been constructed by jointed rigid bodies, it could not be used to predict injury level under loading. Therefore, the model was modified to be capable of injury prediction. First, each vertebral body of the cervical spine was modified to deformable, and the deformable intervertebral disk (IVD) was inserted between each pair of vertebral bodies. Second, each isolated vertebra or IVD models were exposed to static compression in the same conditions as the experiments from the literature to find the critical stress corresponding to the fracture or rupture in the experiments. Next, the kinematics and these critical stress values were validated against the whole body inverted drop tests from the literature. Finally, the critical stress values were examined to be available in several different angles of impact in two series of head-neck drop tests from the literature. In the whole body inverted drop, the kinematics of the cervical spine was well replicated by the model and the critical stress values could well divided the impact velocities with or without injuries. In the head-neck drop with different angles of impact, the model could well predict injurious or non-injurious conditions of the tests. In addition, existing anthropomorphic test devices (ATDs) were examined if their neck structures and corresponding injury criteria could be used for evaluating cervical spinal injuries in rollover compared with the human model. It was found that there were large differences between the predicted injury by the modified human model and those by ATDs output based on the injury assessment reference value (IARV). As a result, the human FE occupant model modified to have deformable vertebral bodies and IVDs instead of jointed rigid bodies appeared to be capable of predicting cervical spinal injury in dynamic rollover. On the other hand, it could be mentioned that further investigation on ATD neck structure and/or injury criterion is necessary to establish a physical evaluation method for occupant protection in dynamic rollover. Dokko

34 INTRODUCTION As a countermeasure for occupant protection of motor vehicle rollover accident, many kinds of policies, such as improvement in wear rate of seat belt, enhancement of roof strength, and adoption of inflatable restraints for ejection mitigation by FMVSS226, have been introduced until now. Furthermore, fatality rate in rollover accident of SUV has decreased by widespread equipment of the Electrical Stability Control System in recent years. However, the rollover accidents forms about 30 percent of all the fatal accidents of passenger vehicle s in the U.S. [1], that means the reduction of fatalities in rollover accidents is still one of the big issues. The main causes of death in rollover accidents are resulted from ejection. By applying the above FMVSS 226, it is expected to have an effect on ejection mitigation at the time of a rollover accident. On the other hand, when the head is impacted by the inside of the roof of a vehicle during a rollover accident, injuries tend to occur in the head and/or neck. In such a case, the rate of occurrence of failure in the head and neck is high. Even those occupants wearing seatbelts that did not eject outside the vehicle have been injured. Although the countermeasures against these injuries are expected, criteria and dummies that evaluate the measures do not exist. Existing ATDs are the tools to evaluate injuries caused by frontal collisions or side impacts. However, injury severity levels of cervical spine caused by impact input from multiple directions like rollover accidents cannot be evaluated because there is no criterion. In this study, the evaluation method of the injury criteria of the cervical spine was examined using the Human FE model. METHODOLOGY The human FE model that the authors had developed to be capable of predicting whole body kinematics and the skeletal injuries on thorax, lumbar spine, and lower extremities of car occupant in frontal and side impact was adopted as a baseline model.[3]-[7] Figure 1 shows the overview of the baseline model. Since the cervical spine part of that model had been constructed by rigid bodies connected by spring elements representing intervertebral disks (IVDs), it could not be used to predict injury level by stress or strain on each element under loading. Therefore, the part of the model was modified to be capable to to that. Figure 1 Overview of the Baseline Model Modification of Cervical Spine Model First, each vertebral body model of the cervical spine was modified from rigid body to deformable model with shell and solid elements for cortical and trabecular bones respectively. And the deformable solid elements for IVD was inserted between each pair of vertebral bodies. Fig.2 shows the cervical spine models of the baseline model (left) and the modified model (right). Rigid Bodies: C1-C7 Figure 1. the cervical spine models of the baseline model (left) and the modified model (right) From one of the anatomies[8], it was found that the cylindrical bodies have extensive cancellous interior with a thin shell of compact bone, while pedicles, articular and transverse processes are mainly compact bone. Based on this knowledge, the thickness of the shell elements of anterior and posterior surface and upper and lower endplates was set thin (0.5mm) and that of other parts including pedicles and processes was set thick (2.0mm) as shown in Fugure 2. Thin Shell (0.5mm) Deformable C1-C7 and IVDs IVDs Represented by Spring Elements Thick Shell (2.0mm) Figure 2. Assignment of Thickness for Vertebral body (C6 as an example) Dokko

35 All the ligaments surrounding the cervical spine were modeled by elastic membrane elements like Sato et al. [9]. Since, in this study, the focus was whether any injury occurs or not, it was thought enough to see the stress level over vertebral bodies and IVDs if more or less than those of critical levels. Therefore, it was decided that either vertebral body or IVD could be treated as of simple elastic materials. The elastic moduli of the lumbar spine model from Dokko et al.[4] were adopted for them. The explicit FE solver PAM-CRASH TM [10] was used. Determination of Critical Stress Second, each isolated vertebra or IVD model was exposed to (quasi) static compression in the same conditions as the experiments from Sonoda[11] as shown in Figure 3. Sonoda[11] showed maximum forces of C3 through C7 for twenty-two Japanese PMHSs from twenties to seventies, from which averaged maximum forces of twenties through fifties were calculated. Sonoda [11] also showed an averaged maximum force of IVDs for forties through fifties. As Dokko et al. [3] did, these forces were scaled to be those for AM50%ile body size by the scale factor of 1.1, resulting in 1.21 for cross section, considering the body size of old Japanese and AM50%ile. Table 1 shows these derived numbers. Vertebral Body Static Force Fixed Cemented IVD Cemented Static Force Fixed Figure 3. Static Compression of Lumbar Vertebral body (left) and IVD (right) Table 1. Averaged Maximum Forces of C3 through C7 and IVD from Sonoda [11] and those after Scaling Averaged Maximum Force (N) Scaled x1.21 (N) C3 3,484 4,216 C4 3,680 4,453 C5 3,646 4,411 C6 3,827 4,631 C7 3,856 4,666 IVD 3,136 3,795 Vertebral Body Under the same loading condition, the maximum stress over the shell elements corresponding to each force level was determined as the critical stress. Figure 4 shows an example for C4. Because of lack of the data for C1 and C2, average of C3 through C7 was put for them. The derived critical Stresses are shown in Table [MPa] 4,453 [N] Figure 4 Determination of the Critical Stress Corresponding to the Maximum Force (Example of C4) Table 2. Determined Critical Stress for Vertebral Bodies # Critical Stress [MPa] C1 218 C2 218 C3 248 C4 204 C5 218 C6 210 C7 208 Table 3. Determined Critical Stress for IVDs # Critical Stress [MPa] C2-C3 15 C3-C4 15 C4-C5 13 C5-C6 12 C6-C7 13 C7-T1 12 Validation of the Modified Model and Critical Stresses The kinematics and the critical stress values derived above were validated against the whole body inverted drop tests from the litereature. Roberts et al.[12] performed the series of full body inverted drop tests as shown in Figure 5 to see the kinematics and injuries around the cervical spine in low (2.0 m/s) and high impact velocity (4.4 m/s) using five male PMHSs of 47 through 71 y.o. with neary AM50%ile body sizes. In addition, the four other results from Kerrigan et al.[13] of the similar condition but medium velocities were supplementaly adopted. Table 4 shows the PMHS Dokko

36 physical information and test matrix of those two series of the tests. surface were picked. Table 5 shows the test matrix picked in this study. Impact velocity was around 3.2m/s. Releaser Harness +θ PMHS Padded Impact Plate Load cell Figure 5. Full Body PMHS Inverted Drop Test Table 4. Physical Information and Test Matrix for PMHS from Roberts et al. [12] and Kerrigan et al. [13] Subject # Age (y.o.) Height (cm) Mass (kg) V1 (m/s) V2 (m/s) V3 (m/s) Note: V1-V3 means the impact velocity in first to third tests Upper five subjects are from Roberts et al. [12] and other four from Kerrigan et al. [13] Finally, the critical stress values were examined to be available in several different angles of impact in two series of head-neck drop tests from Nightingale et al [14] and Toomey et al. [15]. In both series, soft tissue around cervical spine was removed, T1 was fixed into the rigid-like pot, and the mass of carriage including a load cell and a pot was set 16kg representing the effective body mass under T1. Nightingale et al. [14] performed twenty-two head/neck drop tests as shown in Figure 6 varying the conditions, i.e., rigid or padded impact surface with four different anterior/posterior angles. Because of lack of detailed information on the pad to specify to the model, only ten cases with rigid impact Figure 6. Test Set-up of Head-neck Drop from Nightingale et al. [14] Table 5. Test Matrix Picked from Nightingale et al. [14] Impacting onto Rigid Surface Test # PMHS Age (y.o.)/sex Θ (deg) Impact Velocity (m/s) N05-R+30 36/M N18-R+15 -/M D41-R+15 69/M I32-R+15 78/M N26-R+0 65/M N24-R+0 62/M N22-R+0 71/M N11-R-15 55/M N13-R-15 35/F UK3-R-15 62/M On the otherhand, Toomey et al. [15] performed the series of five tests in similar condition to Nightingale et al. [14] but with either lateraly angled impact surface or laterally tilted neck as shown in Figure 7. Table 7 shows the test matrix of them. 15deg 15deg Figure 7. Test Set-up of Head-neck Drop from Toomey et al. [15] with Two Conditions; Laterally Angled Impact Surface (upper right) and Angled Neck (lower right) Dokko

37 Test # Table 6 Test Matrix from Toomey et al. [15] PMHS Age [y.o.] PMHS Height [cm] PMHS Mass [kg] Impact Velocity [m/s] Condition Laterally angled Surface Surface Surface Neck Neck Note: Surface means impact onto the angled surface. Neck means tilted neck. tests. In the tests, assuming the restraint by three point seat belt, the secondary strap was provided to prevent the lower body bearing on the neck, while the simulation did not consider that. It might cause such difference in the second peak of impact forces. Considering those limitation, the principal responses from the human model simulation look satisfactorily representative of those from PMHS tests. RESULTS Either full body inverted or head/neck drop tests were replicated by the human FE model, results of which were compared with the tests. Full Body Inverted Drop Figure 8 shows the human FE model replicating the full body inverted drop tests. As same as the PMHSs used in the tests, upper extremities including clavicles and scapulae were removed. Initial velocity of each case was given to the whole body. From the results, first, overall kinematics around cervical spine was checked. Figure 9 shows the comparison between the high-speed Xray image of the subject 582 from Roberts et al. [12] and the corresponding status from the human model simulation. The simulation result shows the characteristic motion of the cervical spine, i.e., extension in upper and flexion in lower portion, similar to that seen in the test. Next, to confirm the kinematics and responses from the human model representing those from the tests, head and T1 vertical accelerations and head impact force onto the impact plate were compared with those of the test results. Figure 10 through Figure 12 show them. For the tests in 2.0m/s or 4.4m/s, it was possible to develop the corridors of 1SD by three or five data sets for each condition, while, for 3.0m/s or 3.5m/s, only two data sets for each were not enough to do that. Therefore, comparisons were to those two data sets as they were for 3.0m/s and 3.5m/s. Looking at these comparisons, it was found that, in higher velocity, i.e., 3.5m/s and 4.4m/s, the peaks of T1 vertical accelerations from the simulation were higher than those of the tests. It may be caused by elastic modeling of cervical spine without fracture that makes the responsive forces increase linearly while fracture makes it drop in PMHSs, resulting in such a diffirence in T1 acceleration. In higher velocity, it was also found that the second peaks from the simulation were greater than those of the Initial Velocity Figure 8. The Human FE Model of Full Body Inverted Drop Timing 0 Timing 1 Padded impact surface Timing 2 Figure 9. Comparison of Overall Kinematics around Cervical Spine between the test (left) and the Model (right) Dokko

38 Figure 10. Comparisons of Head Vertical Acceleration between Human Model and PMHSs in Inverted Drop Test Figure 11. Comparisons of T1 Vertical Acceleration between Human Model and PMHSs in Inverted Drop Test Dokko

39 For injury prediction, Roberts et al. [12] and Kerrigan et al. [13] described the injuries diagnosed after the tests, which were depicted in Figure 13. There was no injury in any tests in 2.0m/s. Maximum von-mises stress from the human model simulation was checked if greater than the critical stress determined in the previous section for each condition as shown in Figure 14. The elements in C1 and C7 in the velocity higher than or equal to 3.0m/s showed higher von-mises stresses than the critical stresses. Injuries in the tests and predicted from the model for each case are listed in Table 7. From them, it would be mentioned that the human model used with the critical stresses could predict the occurrence of fracture in full body inverted drop condition. (3.0m/s and 3.5m/s) (4.4m/s) Figure 12. Comparisons of Impact Force between Human Model and PMHSs in Inverted Drop Test Fracture Ligament Rupture Dislocation Fugure 13. Diagnosed Injuries in the Full Body Inverted Drop Tests Dokko

40 (2.0m/s) (3.5m/s) (Lower Part) (3.0m/s) (4.4m/s) laterally asymmetric loading conditions. As same as full body inverted drop cases, predicted injuries from the simulation of each condition were compared with those described in Nightingale et al.[14] and Toomey et al.[15]. Figure 15 depicts the von-mises stress contour at the timing of its maximum for each condition highlighting the elements with higher von- Mises stresses than the critical stresses. Injuries in the tests and predicted from the model for each case are listed in Table 8. The condition with fracture was well predicted and the tendency of more fractures at left side was represented by the model. (2.0m/s) (Upper Part) (3.0m/s) (Flat) (+15deg) (3.5m/s) (4.4m/s) Circles show the elements of von-mises stress greater than the critical stress. Figure 14. Predicted Fractures from the Human Model Simulation of Full Body Inverted Drop Table 7. List of Injuries Diagnosed in the Full Body Inverted Drop Tests and those Predicted by the Human Model Simulation Vel. (m/s) PMHS ID Fracture Ligament Dislocation Predicted Fracture C C7 C C1 C1, C7 534 C3, C7, T1 C7-T1-582 C C6, C7 C6-7, C7-T1 - C1, C7 610 C7 C3-4, C7-T1 C7-T C7-T1 C6-7 (+30deg) (Tilted Surface) (-15deg) (Tilted Neck) Head/neck Drop Further validation of the model with the critical stresses was tried if they were applicable for other conditions as shown in Table 5 and Table 6 including Circles show the elements of von-mises stress greater than the critical stress. Figure 15. Predicted Fractures from the Human Model Simulation of Head/neck Drop Dokko

41 Table 8. List of Injuries in the Head/neck Drop Tests and those predicted by the Human Model Simulation Seriese ID Angle (deg) Fracture IVD Ligament Dislocation Predicted Fracture N05-R C3 C3-4 C3-4ALL, C4-5ALL - C1,C2 N18-R C1, C2 C2-3 C2-3ALL C6-7 D41-R C1,C2,C3,C7 C5-6, I32-R C5-6 - Nightingale Cap&ALL et al. N26-R N24-R+0 0 C1, C C1,C3,C6,C7 N22-R+0 0 C N11-R N13-R UK3-R (Surface) (Surface) T1 - - T1-T2 Toomey et al. 3 15(Surface) L-side: C1, C4, C5, C R-side C5 4 15(Neck) C4 - C3-C4-5 15(Neck) L-side: C5, C6 R-side: C C1, L- side:c2,c6,c 7 R-side:C2,C3 C1, L- side:c2,c3,c 4,C6,C7 4,000N for AM50 ATD, was based on the reconstruction of the injurious accident in tackling drill for football by Mertz et al.[19]. It is no doubt a football player is in his maximum muscle tense when he charges the target, that makes his neck stiffer than loose state, resulting in higher loading on cervical spine. At present, within the author s knowledge, it is not clear whether tense or loose state of the neck is likely to suffer cervical spinal injury in the same level of loading. Further investigation is necessary in both points above to establish a physical evaluation method for cervical spinal injury for dynamic rollover, that is, structure of ATD and/or injury criterion. DISCUSSION The possibility of the prediction of cervical spine fracture by the partly modified human FE model with the determined critical stresses was aforementioned. On the other hand, in the development of motor vehicles, it is still necessary to evaluate the performance for occupant protection by physical, not on computer, test using an ATD. In this study, such two existing ATDs as Hybrid III and THOR were examined via their FE models if they were possible to be used for injury evaluation in rollover. The FE models used in this section were as follows. - Humanetics H3-50 th v8.0.1 [16] - Humanetics THOR-50 th Metric v1.3 [17] The explicit FE solver was LS-DYNA [18]. Both models were applied with the same loading conditions as the full body inverted drop test in previous sections as shown in Figure 16. From the results, upper neck force time histories of 2.0m/s are shown in Figure 17. Even in such a low velocity that no injury occurred in any PMHS tests as aforementioned, an upper neck force of either Hybrid III or THOR FE model indicated higher value than the IARV for neck compression force. This inconsistency should be discussed considering two points. First is the structural difference of the cervical spine between the human and ATDs. The stiff and straight-shaped cervical spine of ATDs produces higher axial force in axial loading, while less stiff and curved multi-segmented human cervical spine should ease the force by its flexibility in deformation. Second is no muscle tense in PMHS. As is well known, the IARV for neck compression, (Hybrid III) (THOR) Figure 16. Hybrid III (upper) and THOR (lower) Models in Inverted Drop Condition Dokko

42 Figure 17. Upper Neck Vertical Forces from ATD Models in Inverted Drop (2.0m/s) CONCLUSIONS The human FE occupant model was modified to have deformable vertebral bodies and IVDs instead of jointed rigid bodies for cervical spine, which resulted in capable of predicting cervical spinal injury in dynamic rollover condition by comparing stress level among cervical spine to the determined critical stress. On the other hand, large differences were found between the injury prediction by the modified human model and ATD models in the same loading condition. It has become clear that further investigation on ATD neck structure and/or injury criterion is necessary to establish a physical evaluation method for occupant protection in dynamic rollover. ACKNOLEDGEMENT The authors would thank to ESI Japan for providing the material parameters of the pad tuned for the drop test results REFERENCES IARV for Neck Compression [1] Traffic Safety Fact 2014,NHTSA [2] Ridella S. A., Eigen A, Kerrigan J., Crandall J.R An Analysis of Injury Typr and Distribution of Belted, Non-ejected Occupants Involved in Rollover Crashes, AAAM Vol. 53 [3] Dokko Y., Ito O., Ohashi K. 2009, Development of Human Lower Limb and Pelvis FE Models for Adult and the Elderly, SAE [4] Dokko Y., Kanayama Y., Ito O., Ohashi K., 2009, Development of Human Lumbar Spine FE Models for Adult and the Elderly, SAE [5] Ito O., Dokko Y., Ohashi K., 2009, Development of Adult and Elderly FE Thorax Skeletal Models, SAE [6] Ito Y., Motozawa Y., Dokko Y., Ohashi K., Mori F., 2012, Kinematics Validation of Age-Specific Restrained 50 th Percentile Occupant FE Models in Frontal Impact, SAE [7] Dokko Y., Yanaoka T., Ohashi K., 2013, Validation of Age-specific Human FE Models for Lateral Impact, SAE [8] Williams P. L., 1995, Gray s Anatomy, Thirty- Eighth Edition, Churchill Livingstone [9] Sato F., Antona J., Ejima S., Ono K., 2010, Influence on Cervical Vertebral Motion of the Interaction between Occupant and Head Restraint/Seat, based on the Reconstruction of Rear- End Collision Using Finite Element Human Model, 2010 IRCOBI Conference Proceedings [10] Explicit Solver Reference Manual, PAM- CRASH, 2009, ESI Group [11] Sonoda T., 1962, Studies on the Strength for Compression, Tension and Torsion of the Human Vertebral Column, Journal of Kyoto Prefectural University of Medicine, Vol. 71, pp (1953) (in Japanese) [12] Roberts C., Kerrigan J., 2015, Injuries and Kinematics: Response of the Cervical Spine in Inverted Impacts, ESV [13] Kerrigan J.R., Foster J.B., Sochor M., Forman J.L., Toczyski J., Roberts C.W., and Crandall J.R., 2014, Axial Compression Injury Tolerance of the Cervical Spine: Initial Results, Short Communications from AAAM s 58 th Annual Scientific Conference, Traffic Injury Prevention, 15:sup1, S238- S269, DOI: / [14] Nightingale R.W., McElhaney J.H., Camacho D.L., 1997, The Dynamic Response of the Cervical Spine: Buckling, End Conditions, and Tolerance in Compressive Impacts, 41 st STAPP Car Crash Conference Proceedings, SAE [15] Toomey D.E., Mason M.J., Hardy W.N., Yang K.H., Kopacz J.M., 2009, Exploring the Role of Lateral Bending Postures and Asymmetric Loading on Cervical Spine Compression Responses, Proceedings of the ASME 2009 International Mechanical Engineering Congress & Exposition, IMECE [16] User s Manual, Hybrid III 50th Percentile Male Dummy, LS-DYNA Model Version 8.0.1, 2013, Humanetics Innovative Solutions, Inc. [17] THOR-50TH Dummy Model LS-DYNA, Release Version 1.3, 2016, Humanetics Innovative Solutions, Inc. [18] LS-DYNA KEYWORD USER'S MANUAL, Version 971 R6.0.0, 2012, JSOL Corporation Dokko

43 [19] Mertz H.J., Hodgson V.R., Thomas L.M., Nyquist G.W., An Assessment of Compressive Neck Loads under Injury-producing Conditions, 1978, The Physician and Sports Medicine 6 Dokko

44 Rollover Simulation Using an Active Human Model Allen Chhor Damian McGuckin Pacific ESI Australia Hyung Yun Choi ManYong Han Hongik University South Korea Inhyeok Lee Hankook ESI South Korea Paper Number ABSTRACT Today s anthropomorphic test devices (ATD) derive their behavior from cadaver test data. This same statement also applies to numerical models of these physical ATDs, and equally to the more sophisticated numerical human body models. Across the wide spectrum of automotive and aerospace crash scenarios, the prediction of occupant responses relies mainly on joint properties that are inherent in such behavioral representations. These do not account for the muscle reflexes of tensing and bracing. Most ATDs, and especially the Hybrid IIIs, are relatively rigid and their response will effectively represent occupants subjected to high speed impacts. A series of numerical active human (AH) body models have been developed for the 5th, 50th and 95th percentile of human subjects using multi-body modelling that incorporates joints with active torque behavior. In addition to the standard joint torque resistance, active joint behavior is implemented numerically in these AH-models using proportional-integral-derivative (PID) control methods to deliver torque resistance representative of active muscle responses. Active torque behavior for selected human body joints is achieved by optimizing PID gain parameters to correlate with test responses of human volunteer test data. The result of this work was applied to this first generation of active joint human models. The potential of human body models with active joints is demonstrated in a vehicle rollover situation. The specific case of vehicle rollover provides a crash scenario where the occupant s accident awareness response is likely to influence tensing and active joint behavior at various stages during the accident. These simulations highlight the influence of muscle tensing and joint bracing on potential injury risk. This method of modelling the active joint torque seeks to mimic the complex behavior of muscles. It provides an efficient modelling technique that can be used to simulate long duration events (such as vehicle rollover) that in the past may have been considered less than optimal for the more complex human models. The ability to activate or deactivate the joint behavior to account for conscious muscle tensing will allow the analysis of various occupant awareness states during a rollover accident. It is anticipated that the addition of active joint behavior will provide a more accurate numerical representation of human body kinematics and hence improve the quality of the prediction of the risk of injury that can be deduced from simulations. The ability to activate joint behavior to account for conscious muscle reflexes will also extend the range of crash scenarios which can be modelled effectively. Chhor

45 INTRODUCTION Occupant safety regulations for the crashworthiness testing of vehicles rely on the physical representation of a human occupant by an anthropomorphic test device or ATD. These ATDs, more generally referred to as crash test dummies, exist in various customized forms that are designed to perform biomechanically for specific types of crash configurations. ATDs derived their behavior from cadaver test data [1,2] which can be considered similar to a flaccid human. However, if one manipulates the various joints or bend the neck, it becomes very clear that the dummy is very stiff. This is because ATDS are tuned to represent humans in high speed (about 60km/h), potentially injurycausing crashes. In a low velocity impact test, an ATD s response is unrepresentative because of their overly stiff nature. The Hybrid III dummy is the current international standard for frontal crashes, while various other ATDs have been specifically designed for alternative test configurations such as side impact. To adequately represent the range of human sizes and proportions, these ATDs also exist as a series of anthropometric scales. Hence, such ATDs are implicitly limited by their characterization of actual complex humans in terms of their own geometric representation and mechanical response. ATDs are mechanical surrogates designed to represent a particular demographic according to gender, size, and age. In addition, they are designed to exhibit a biofidelic response for specific loading conditions (e.g. principal direction of force and severity). The responses of these devices are not validated for alternate loading conditions and thus may not produce biofidelic responses beyond their intended design specifications. [3] Simple but computationally efficient multi-body models as well as very detailed finite element (FE) models of the various ATDs have been developed over the years to simulate real world ATDs. Having numerical equivalents of the various ATDs has complemented real world tests by allowing more cost effective, efficient, and detailed analyses of ATD behavior in a wider ranges of test scenarios. The limitations of these models however, is that they can only be as accurate as the ATD s representation of real humans. For these reason, there have been extensive parallel development of numerical human body models that attempt to simulate real humans rather than ATDs. Detailed numerical computational models of the human body have been developed by various research groups to allow more detailed study into biomechanical issues. Unlike ATDs, humans have functioning circulatory and respiratory systems, resting muscle tone and active bracing capabilities, continuous neural responses, and the ability to perform cognitive functions [3]. Numerical human models offer not only modelling flexibility but more exact characterizations of both varying human anthropometry and their biomechanical responses in a wide range of loading conditions. Some of these human models that have been developed include the H-MODEL (Hongik University s Human Body Model) [4], THUMS (Toyota s Total Human model for Safety) [5], and GHBMC s human body models (Global Human Body Models Consortium) [13]. In more recent years, attention has also been directed towards modelling active muscle behavior to account for an occupant s bracing reflex in the event of awareness of the approaching accident. It has been shown that tensed muscles can change the initial posture, kinematics, and subsequently the kinetics during an automotive collision and as a consequence, the resulting injury patterns may be altered based on muscle activation [8]. This study illustrated that muscle activation has a significant influence on the biomechanical response of human occupants in low-speed frontal sled tests. MODEL DEVELOPMENT Active human body models (ah-model) are currently being developed with body size and weight representative of accepted automotive industry standards for occupant safety testing. A series of three ah-models currently under development include the 5 th percentile female (ah-f05), the 50 th percentile male (ah-m50) and the 95 th percentile male (ah-m95). The motivation for developing these ah-models is to study the contribution of active reflexive human responses to accident events and its potential to affect injury risk. The computational implementation takes advantage of the simpler but extremely efficient modelling techniques of multiple rigid body segment connected by joint elements with defined moment resistance. This traditional technique of modelling a joint is complemented by incorporating active joint behavior via a torque Chhor

46 actuator at the joint. The torque actuator is implemented using a numerical closed loop proportional-integral-derivative (PID) controller. The efficiency of the ah-models with its moderate CPU demand will allow the simulation of more complex and long duration crash and test scenarios where active responses are more likely to play a role in the kinematics of the occupant. Human Geometry For each of the human models, the outer surface geometry of their body segments were meshed from 3D surface data. The 3D human surfaces were produced from geometric scans of selected human subjects postured in a seated position. The three scanned human subjects were selected by size and weight, to be representative of the previously mentioned size categories. The selection criteria were based on the Size USA 2002 datasets. Parts Segmentation and Joint Locations The human geometric scans were discretized into a finite element (FE) mesh, and then subsequently segmented into the fifteen commonly-accepted anatomical body parts. For each mesh of the human model category, a consistent segmentation method was applied across the three ah-models to create these rigid body segments. Figure 1 shows the three human models and the segmentation scheme used to create the various anatomical parts. A skeletal mesh within the surface mesh, as illustrated in figure 2, provides a reference to aid in the body segmentation process. In particular, while the head/neck/trunk is represented by five body segments that are separated by defined spinal positions, they can only be physically located using the spine as a reference. The skeletal articulation also assists with the location of the joint position of the shoulder, hip, and joints of the limbs. Table 1 summarizes the modelled joints and their anatomical positions. Figure 2. structure Human models including skeletal Table 1. Joints and body segments. # Joint DOF Anatomical position 1 Head-neck 3 OC joint 2 Neck-Upper trunk 3 C7/T1 3 Upper-Center trunk 3 T12/L1 4 Center-Lower trunk 3 L5/S1 5 Upper trunk-arm, R 3 Right Shoulder 6 Upper-Lower arm, R 1 Right Elbow 7 Upper trunk-arm, L 3 Left Shoulder 8 Upper-Lower arm, L 1 Left Elbow 9 Lower trunk-leg, R 3 Right hip joint 10 Upper-Lower leg, R 1 Right Knee 11 Lower leg-foot, R 3 Right Ankle 12 Lower trunk-leg, L 3 Left hip joint 13 Upper-Lower leg, L 1 Left Knee Figure 1. Human models meshed and segmented from the Size USA 2002 data 14 Lower leg-foot, L 3 Left Ankle Chhor

47 Active Joint Modelling Each joint is modelled using a kinematic FE joint element consisting of an angular stiffness function, damping, and a PID torque actuator. All of these joint constraints act in parallel to represent various muscle conditions. Using Choi s hypotheses [6,7] of active joint responses, the rotational stiffness of a joint is complemented by different joint damping profiles depending on whether muscles are tensed or relaxed, and a PID torque control is applied to the joint to model whether the subject is aware or unaware during loading. The active torque of the joint represents the resultant actions of all muscles, ligaments, tendons, and other human tissues which affect that joint behavior. states of muscle tensing can be reasonably modelled by a damping moment. In addition to the muscle being relaxed or tensed, there are also the potential joint responses to the cognitive states of being aware or unaware of both an impending load or the actual loading per se. Awareness generally leads to muscle tensing but during long duration loading events (such as a rollover accident), human instinct and reflexes mean that the subject cognitively tries to correct their posture and limb configurations to counter the forces acting on them. This reflex reaction is modelled by applying a torque actuator to the joint via a PID closed loop control method. At every cycle of the computation, the PID function makes an assessment of the proportional, integral, and derivative behavior of the joint relative to the target position. A very general explanation of PID control would say that the proportional is the current behavior, the integral is the historical behavior, and the derivative is the projected future. Based on this, the PID controller attempts to correct the system based on the error calculated at each cycle multiplied by the gain constants (k p,k i,k d ). The PID function is summarized by the function below. Muscle model with Closed-Loop Control (PID) Tensed (co-contraction) Relaxed (single contraction) Figure 3. Elbow jerk loading test by Choi[8] To show this, volunteer tests were conducted [6] to measure the response of the elbow joint to the application of an initial static load followed by a jerk load while the muscles are tensed or relaxed. The experiment produced different joint damping resistances as shown by the graphs in Figure 3. The tensed state of muscle co-contraction produces both a flexion resistance (-1.5kNms/rad) to the static load and an equivalent extension resistance (1.5kNms/rad) when the jerk loading is applied. When the subject is relaxed with the muscles only in the single contraction state to resist the static load, the same flexion resistance is observed (-1.5kNms/rad), although the extension resistance to the jerk loading is greatly reduced (0.5kNms/rad). Therefore, the various y(t): current state r(t): reference u(t): control signal k p : proportional gain k i : integral gain k d : derivative gain The gain parameters k p, k i, and k d are obtained through an optimization processes by correlating the model s response to the active response exhibited by an aware subject in a volunteer test. The joint stiffness properties of the ah-models exploit the knowledge gained from past human body model studies [6,7,8]. Each ah-model consists currently of 15 rigid body segments connected by 14 articulated joints. All joints are modelled with three rotational degrees of freedom with the exception of the elbow joints and the knee joints which only have one rotational degree of freedom. Active torque capability is modelled for Chhor

48 all joints in all their degrees of freedom, for the ah-models similar to the elbow model described. Joint Stiffness Scaling The 50th percentile male is the most common human size used for compliance standards. This has led to it being the most widely documented and tested. For this reason, the ah-m50 model is the first to be calibrated and correlated with test data. The ah-m50 serves as the reference for the stiffness of the various body joints for this first generation of ah-models. To estimate the stiffness of joints for the ah-f05 and ah-m95, some general scaling can be made from the ah-m50. The stiffnesses of human joint articulations are dependent on the gender, body size, muscular structure, anatomical proportions and hence, some scaling assumptions can be used. The same can be said for the joint damping which models the muscle tensing strength. Therefore, the joint stiffness of the various joints for the ah-f05 and the ah-m95 can within reason be initially scaled and interpolated from the ah-m50 until additional data for calibration becomes available. The ah- M50 joint properties will be re-tuned regularly as more up-to-date data becomes available with the same being done on the ah-f05 and ah-m95. Joint reflex modelled by PID controls are less scalable however, as instincts are not dependant on body size or muscle strength. Nonetheless, similar scaling of PID gain parameters can be used as a good first estimate. thoracic/lumbar spinal system, and the limbs. This was done by applying a relative low acceleration/deceleration pulse in selected loading directions, to the various anatomical systems. When the joints are defined only by angular stiffness functions, and using the head/neck system as an example, the segments would oscillate indefinitely in the direction of loading. When some damping was applied to the joints, the oscillations would come to rest after 1 to 2 cycles. A further application of torque actuators to the joints, through the PID closed loop control is applied to the system to represent natural human reflexes of an aware subject to stabilize oneself and resist the loading. The PID gain parameters were optimized with the objective of damping out the oscillations within approximately one cycle. This was considered a good first approximation of the human reflex contributions to resist such loading. Figure 3 shows some time history frame grabs of the head/neck system for non-active and active joints when the local joint systems are loaded in lateral bending. When the joints are active, the maximum lateral bending of the head/neck is reduced and is stabilized quicker, as well as returning to the neutral target position defined by the PID function. CALIBRATION AND OPTIMISATION start Accel decel Non-active head/neck joints stop Calibration and verification of the ah-m50 is currently being undertaken by correlating the models response to available test data and published experimental data. Published data of post mortem human subject (PMHS) tests [10] and volunteer tests [3,11] provide some excellent references for correlating and assessing the performance of the ah-m50 model. Some basic calibrations were initially performed on selected joint groups of local anatomy to optimize PID gain values that represent those of an aware subject. The joints were optimized locally for the head/neck system, the start Accel decel stop Active head/neck joints Figure 3. PID calibration of active joints in the head/neck system Figure 4 shows a similar calibration where the lateral pulse loading is applied to the pelvis. Only the thoracic and lumbar joints are free to deform. As with the head/neck system calibration, PID activation of the thoracic/lumbar joints produced lower maximum deformations, earlier Chhor

49 stabilization, and the model is returned to the target neutral posture. Similar active torque calibrations were carried out for the hip/lower-limb and the shoulder/upperlimb systems. As data becomes available, particularly those regarding non-injurious loading of volunteers, such active joint torque calibrations can be refined to better represent bracing and reflex responses. more than 1.5 seconds, where active joints are expected to play a role in the kinematics of the ah-model. Figure 5 shows the rollover test rig developed by the University of Virginia s Center for Applied Mechanics with occupants in the leading and trailing seat positions. start Accel decel stop Non-active thoracic/lumbar Joint start Accel decel stop Active thoracic/lumbar Joint Figure 4. PID calibration of active joints in the Thoracic/lumbar system Figure 5. DRoTS system from the University of Virginia Center for Applied Biomechanics Figure 6 shows some frame grabs at 0 o, 90 o, 180 o, and 270 o of the simulation for pure dynamic roll, with both the leading and trailing occupants. As in the PMHS tests, the ah-m50 models are each restrained with a three-point lap sash belt. In addition, the left and right hand are strapped to their respective left and right upper leg with the lower legs and feet restrained to the test rig. HUMAN MODEL PEFORMANCE IN DYNAMIC ROLLOVER TEST SYSTEM (DRoTS) DRoTS tests of PMHS [10] were used as an initial reference test to assess the performance of the full ah-m50 model. Several defined tests can be performed on this rollover test system. They include, a quasi-static test with 180 o rotation, an upside down drop and catch with 0 o rotation, a pure dynamic roll with 360 o rotation, a leadingside drop with 360 o rotation, and a trailing-side drop with 360 o rotation. The case of pure dynamic roll, over 360 o was simulated with the ah-m50 in the leading and trailing seat positions. The loading conditions of this test allow the simulation of a rollover of a full rotation, which occurs over an extensive period of Figure 6. DRoTS simulation with two ah-m50 models Of interest in the DRoTS test of pure roll is the kinematics of the spine during the entire rolling event. The overall performance of the ah-m50 can be assessed relatively simply by comparing the lateral bending of the head/neck system during a Chhor

50 PMHS physical test and its simulation using the ah-m50 model. Images at 45 o intervals of angular rotation during the rollover, taken from high speed camera footage, were used to make these qualitative comparisons with the simulation. Cadaver non-active Active Test (cadaver) (Human) Pure Roll Test (360 o ) Leading-Side Position Figure 7 shows the comparison between the PMHS leading-side position test (column1) and its model equivalent of the ah-m50 with non-active joints (column2). Simulation results of an ah-m50 with active joints representing an aware human that is tensed with reactive reflexes (column 3) is also presented to study variations in kinematic response compared to an ah-m50 with non-active joints. The amount of lateral bending of the head/neck cannot be clearly observed from the photo frames at 225 o and 270 o for the leading-side position due to what appears to be visual obstruction of the camera view. A comparison of the head/neck lateral bending response between the leading-side PMHS test and the ah-m50 model with non-active joints show good general qualitative agreement in terms of the amount of head/neck angular rotation in lateral bending as well as being in phase. When the joints are activated in the ah-m50 model to simulate tensing and reflexive behavior, the amount of lateral bending deformation is reduced in all the frames shown. This reduction in lateral bending deformation of the head/neck system is most obvious at the frame rotations at 90 o, 180 o, and 360 o. The probable reason for this is because the largest variation between an active and non-active model is observed at the higher loads levels, when the joint PID torque actuator is most affective. At lower load levels where the head/neck lateral bending displacements are lower, less variation is also observed between the ah-m50 with and without active joints. Pure Roll Test (360 o ) Trailing-Side Position Figure 8 shows the same comparison of the DRoTS pure roll test for the trailing-side position. The PMHS test results are shown in column 1 with the non-active ah-m50 in column2 and the active ah- M50 human model in column 3. For this trailingside position test, the view of the PMHS head/neck kinematics is obscured in the 270 o and 315 o photo frames. Figure 7. Pure dynamic roll of PMHS and ah-m50 in the leading-side position Chhor

51 Cadaver non-active Active Test (cadaver) (Human) Qualitative comparison between the test results and the ah-m50 with non-active joints show good agreement of the lateral bending of head/neck in terms of general magnitude and timing. Unlike the leading-side position analyses, when the ah-m50 joints are activated for the trailing-side position analysis, the reduction in head/neck lateral bending is less pronounced. The only frames that showed an obvious difference were the 45 o and 360 o frames. What this highlights is that the loads experienced in the leading and trailing positions are quite different. These two controlled pure rollover cases illustrate is that the ah-m50 model with active joints generally shows lower maximum bending of corresponding joints. Not obvious in these frame grabs is the earlier recovery of the joint angular displacements displayed by the active human model to try to return to its defined neutral (or target) position defined by the PID controller. Earlier recovery of joint angular displacements leads to changes in human body kinematics during the loading event. Taking the case of a vehicle rollover as an example, these kinematic changes can result in different levels of maximum human body joint deformations, timing of head impact or human body contact to the vehicle interior, and can change the location of the impact which can ultimately change the type of injury potential. VEHICLE ROLLOVER Vehicle rollovers by their nature are complex crash events that can be triggered by various combinations of driver behavior, road surface type and its interaction with the vehicle, and the size/height/weight of the vehicle. Rollover crashes occur over a long duration measured in seconds as opposed to general car crash durations of only a few hundred milliseconds. For these reasons, the numerical analysis of occupant behavior in vehicle rollovers have been less viable due to the high computing demands required to simulate such long duration events. Figure 8. Pure dynamic roll of PMHS and ah-m50 in the trailing-side position The numerical efficiency of the ah-m50 model has meant that simulating a long duration rollover analysis is attainable. Trial analyses were performed to assess the viability of simulating a full vehicle rollover with the ah-m50 human occupant model. The rollover arrangement Chhor

52 simulated, involves the vehicle flipping over at 48km/h inducing approximately three rolls of the vehicle before coming to rest over a duration of five seconds. Figure 9 shows the various vehicle rollover states during one of the rollover simulations. Two simulations were carried out using this loading configuration, one with the ah- M50 with active joints and the other with nonactive joints. The ah-m50 human models are seated in the driver side position for these simulations. 48km/h Figure 9. Whole vehicle rollover simulation with the ah-m50 model. Two main factors directly affect the risk of injury to an occupant in a rollover. The first is the ability of the vehicle to maintain its structural integrity during the crash thereby preserving the occupant space. The second is the performance of the vehicle s internal safety and restraint systems such as seatbelts that restrain occupants and keep them away from hard surface impacts. With occupant ejection from the vehicle being a concern in rollover accidents, seatbelts also play a role in preventing this type of phenomena. More recently, the prevention of occupant ejection has been a secondary consideration in the design of curtain airbags [12]. For these trial simulations, a simple analysis of the results was undertaken to compare the effect of using a human body model with fully active joints against one that is non-active. The active joints were maintained throughout the rollover simulation in the first rollover case. This represents one extreme of the occupant being fully aware for the whole rollover duration. The second rollover case with non-active joints simulate the other extreme of the occupant being unaware throughout the accident duration. The graph in Figure 10 shows the contact force magnitude of the head to the interior surfaces of the vehicle for the first 2.5 seconds of the rollover. It can be seen that when the joints are modelled as active, the force of the first major contact (~600ms) made between the head and the interior of the vehicle is about 30% lower in comparison to the non-active model. This can be attributed to the resistive reflex nature of the active joints at the neck to correct its posture and reduce the amount of head/neck motion, all of which leads to a lower contact force. It is also noteworthy that the timing of the initial head contact force between the 2 cases is already slightly out of phase due to variations in the active joint response of the head/neck system. Beyond this first impact event, the head contact forces are significantly higher for the active joint case. The reasoning for this is that after the first main head impact, the kinematics of the two cases have varied enough that subsequent head contacts (between the two cases) are no longer in phase and are contacting the vehicle interior at different locations as well as from different velocity vectors from the head. A general hypothesis that can be made is that relatively reliable quantitative comparisons between models can be made up to the first major contact event. Beyond this, only qualitative comparisons are reasonable due to growing variations of the occupant kinematics with time after the crash event. This is not to say that these later occupant kinematics, are any less important as they determine the interaction of the occupant with the interior environment, interactions with other occupants, or even occupant retention. Figure 10. Contact force between the ah-m50 head with the vehicle interior The example presented is an ideal extreme of active and non-active joints on the ah-m50 human Chhor

53 model. In a real-world rollover, it is more likely that an occupant will experience different states of awareness throughout the rollover. One possible scenario could involve the driver bracing (active joints) at the beginning of the accident, but losing consciousness (non-active joints) during the rollover due to a head impact to the vehicle interior. Alternatively, a passenger may not be aware of an impending rollover and will brace much later during the rollover. These are just two examples of many possible scenarios in this type of long duration accident. added to the model. Recalling the motive for developing these active human models, the model s efficiency should be maintained when possible to set it apart from existing complex and very detailed human models. Calibration and correlation is ongoing and will be updated as new data becomes available. It is anticipated that a database of tensing parameters and optimized PID gain values for reflexive strength will eventually be developed for specific loading conditions and severity. FUTURE MODELLING This first series of ah-models has taken advantage of past human body models and their joint stiffnesses. In addition, the models have been correlated to available muscle tensing studies, and optimized for projected active reflexive behavior. Although still relatively early in its development, their performance in these initial correlation studies show promise, especially in regard to the implementation of active joint torque behavior. Incorporating translational degrees of freedom (with stiffness properties) to selected spinal joints is the next logical step to refining the biomechanical response of these ah-models. To better understand the reflexive behavior of aware subjects, it would be essential to calibrate such responses to human volunteer data. Volunteer experiments to measure joint reflexes would need to be performed at loads that are non-injurious. Although not ideal, such data will still be invaluable for projecting the anticipated reflex behavior under genuinely injurious loads using mathematical interpolation techniques. The current models have been developed in a manner that increased sophistication can be retrofitted to the model. Possible retrofit options may include the addition of spinal complexity through additional joint articulations, or whole cervical or thoracic spine replacements. Some deformability of the pelvis or torso may also be developed as a retrofit option to better model seat and seatbelt interactions. However, the model s efficiency, which has been achieved by using simple modelling techniques would gradually be sacrificed as more complexity is CONCLUSIONS Three active human models with active joint behavior have been developed. The series of active human models consist of the 5 th percentile female, the 50 th percentile male, and the 95 th percentile male. The performance of the 50 th percentile model has been correlated to a limited selection of published PMHS and human volunteer test data with good general agreement. The models will need to undergo further correlation and calibration to extend their validity over a wider range of loading severities and loading types. This will involve further optimization of the active joint behavior with existing and future PMHS and human volunteer test data. The ah-m50 model has been used to simulate a vehicle rollover undergoing three full rotations over a relatively long duration of five seconds. The implementation of active joints in the active human models allows the simulation of human joint tensing and reflex behavior resulting from different states of human awareness. Active joints coupled with the overall model efficiency will allow the analysis of longer duration accident scenarios that account for complex human awareness reactions and ultimately will broaden potential fields of application. ACKNOWLEDGEMENTS The authors gratefully acknowledge the efforts of Dr Tom Gibson and Dr Lex Mulcahy, in reviewing this paper and contributing to its content with their insightful comments. Chhor

54 REFERENCES [1] Mertz H, Patrick L, Strength and Response of the Human Neck. SAE Technical Paper No , 15 th STAPP Crash Safety Conference (1971), SAE 1971 Transaction V80- E. [2] Horsch J, Schneider D, Bio-fidelity of the Hybrid II Thorax in High -Velocity Frontal Impact. SAE Technical Paper No , SAE International Congress and Exhibition (1988). [3] Stephanie M. Beeman, Andrew R. Kemper, Michael L. Madigan, Christopher T. Franck, Stephen C. Loftus, Occupant kinematics in low-speed frontal sled tests: Human volunteers, HybridIII ATD, and PMHS. Elsevier Accident Analysis and Prevention 47 ( ) [4] Hyung-Yun Choi, Hong-Won Eom, Soon-Tak Kho, In-Hyeok Lee, Finite Element Human Model for Crashworthiness Simulation, Digital Human Modeling for Design and Engineering International Conference and Exposition, The Hague, The Netherlands, May 18-20, [5] Eberhard Haug, Muriel Beaugonin, Nicole Montmayeur, Christian Marca, Hyung-Yun Choi, Towards Legal Virtual Crash Tests For Vehicle Occupant Safety Design Using Human Models. Invited presentation: ICD 2003, Dec 2-4, Lille France [6] M. Han, H.Y. Choi, Elbow Joint Model with Active Muscle Force. Journal of Mechanical Science and Technology 30/ [7] Hyung Yun Choi, Manyong Han,Inhyeok Lee, Jungtae Yang, Wiro Lee, Active Human Body Model. IRCOBI Asia. [8] H. Y. Choi, S. J. Sah, B. Lee, H. S. Cho, S. J. Kang, M. S. Mun, I. Lee, J. Lee, Experimental and Numerical Studies of Muscular activation of Bracing occupant. Proc. Of Enhanced Safety of Vehicles, Washington D.C., USA [9] Eberhard Haug, Hyung-Yun Choi, Stéphane Robin, Muriel Beaugonin,2004. Human Models for Crash and Impact Simulation. Computational Models for the Human Body Copyright 2004 Elsevier B.V. HANDBOOK OF NUMERICAL ANALYSIS, VOL. XII [10] David J. Lessley, Patrick Riley, Qi Zhang, Patrick Foltz, Brian Overby, Sara Heltzel, Mark Sochor, Jeff Crandall, Jason R. Kerrigan, Occupant Kinematics in Laboratory Rollover Tests: PMHS Response. Stapp Crash Journal, Vol 58 (November) [11] Stephanie M. Beeman, Andrew R. Kemper, Michael L. Madigan, And Stefan M. Duma, Effects of Bracing on Human Kinematics in Low-Speed Frontal Sled Tests. Annals of Biomedical Engineering, Vol 39, No. 12, December, pp [12] Eung-Seo Kim, Dae-Young Kwak, Hyeong- Ho Choi, Han-Il Bae, Seung-Hui Yang, Seung- Man Kim, Dong-Jun Lee, Kwang-Soo Cho, 2011 A Study Of Curtain Airbag Design Factors For Enhancement Of Ejection Mitigation Performance. Paper Number , 22nd International Technical Conference on the Enhanced Safety of Vehicles (ESV) [13] Maika Katagiri, Jay Zhao, Jason Kerrigan, Richard Kent, Jason Forman, 2016 Comparison of Whole-Body Kinematic Behaviour of the GHBMC Occupant Model to PMHS in Far-Side Sled Tests. IRC-16-88, IRCOBI Conference 2016 Chhor

55 PROTECTION OF CHILDREN IN CHILD RESTRAINT SYSTEMS IN OBLIQUE IMPACTS: RELATIVE MOTION OF THE CHILD AND CHILD RESTRAINT Hans W. Hauschild John R. Humm Medical College of Wisconsin United States Kristy B. Arbogast Matthew R. Maltese The Children s Hospital of Philadelphia University of Pennsylvania Perelman School of Medicine United States Frank A. Pintar Narayan Yoganandan VA Medical Center Medical College of Wisconsin United States Bruce Kaufman Children s Hospital of Wisconsin Medical College of Wisconsin United States Paper No ABSTRACT Objective. The objective was to determine the relative contribution of occupant versus child restraint system (CRS) kinematics to overall lateral head excursion for children in forward facing CRS (FFCRS) during oblique side impacts. As a secondary objective, the effect of the tether was investigated. Methods and Data Sources. Sled tests were conducted with a FFCRS and Q3s Anthropomorphic Test Device (ATD) secured to a vehicle seat via LATCH, utilizing the center seat position. The vehicle seat and a simulated intruded door were secured to the sled at two angles (60 and 80 degrees from full frontal). Tests were conducted at 35 km/h delta-v, with and without a tether. Three-dimensional motion capture cameras captured kinematics of the ATD, FFCRS and vehicle seat. Head accelerations, neck forces and moments, and LATCH belt forces were obtained. The analysis focused on the relative contribution of the FFCRS motion versus the ATD motion with respect to the FFCRS on global lateral head excursion. Results. The overall median lateral head excursion of the Q3s relative to the sled was 430 mm; approximately half of the excursion was the displacement of the head relative to the FFCRS (median 223 mm). Head angular motion relative to the FFCRS (median roll, pitch and yaw were -79, -55, and 34 degrees respectively) was greater than the overall angular motion of the CRS (median roll, pitch and yaw relative to the vehicle seat were -18, 5, and -17 degrees). Tether use influenced the FFCRS motion, but not the head motion within the FFCRS. Observations were similar across both test angles. Discussion and Limitations. In order to gain a better understanding of side impact occupant protection for those restrained in FFCRS, this research examined both overall FFCRS motion as well as occupant motion within the FFCRS. Previous kinematic analyses typically examined only occupant motion relative to the vehicle frame of reference. A large proportion of the occupant s lateral head excursion was due to the head movement relative to the

56 FFCRS suggesting interventions that address both aspects of lateral kinematics movement of the FFCRS as well as lateral bending/forward flexion of the occupant s torso/neck relative to the FFCRS might result in overall injury mitigation. It was important to note that while tether use reduced FFCRS motion, it did not significantly increase the motion of the head relative to the FFCRS due to increased restraint of the FFCRS. Limitations include testing one FFCRS, one delta-v, and FFCRS attachment with a flexible LATCH system. Conclusion. Occupant lateral head excursion and angular kinematics in oblique side impact crashes are related both to movement of the FFCRS as well as significant motion of the occupant relative to the FFCRS. This finding suggests two pathways for design intervention to mitigate overall occupant lateral excursion and potential impact with intruding structures, a common injury causation scenario for children in these crashes. INTRODUCTION Recent development of regulatory test procedures worldwide have been focused on evaluation of child restraint systems (CRS) in side impacts, but that work has mostly been focused on near side child occupants (Brown et al. 1997; NHTSA 2014b; Sullivan et al. 2011; Sullivan and Louden 2009). Research data suggests non-near side child occupants are being injured in side impact crashes as well (Huntley 2002; Arbogast et al. 2010; Brown et al. 2002; McCray et al. 2007; Orzechowski et al. 2003; Sherwood et al. 2003; Sullivan and Louden 2009). Injuries occurred when the child occupants contacted the vehicle interior, other CRS, their own CRS, and other occupants (Arbogast et al. 2010; Sullivan and Louden 2009; McCray et al. 2007; Charlton et al. 2007; Sherwood et al. 2003; Brown et al. 2002). Previous work studied the protection of non-near side children in FFCRS during side impacts and examined the role of large side structures or side wings designed to provide a means by which to limit lateral head excursion in oblique side impact loading. (Hauschild et al. 2015). Results demonstrated the side wings did not provide adequate head restraint as the ATD head rolled out of the FFCRS and displaced far enough to place the occupant at risk of impacting intruding side vehicle components. Other research examined the role of the tether in controlling lateral head excursion in similar lateral oblique loading scenarios (Hauschild et al. 2016). To better assess the potential for head impact, intrusion was simulated by including a door structure on the test buck. All tests without a tether resulted in head contact with the simulated door, and two tests at near pure lateral (80 degree) impact direction with a tether also resulted in head contact. No head to door contact was observed in two tests at 60 degrees from full frontal utilizing a tether. High speed video showed the FFCRS rotated and tipped (yawed and rolled) which caused the head to roll out of the FFCRS head side wings and make contact with the simulated intruded door. The research described above as well as others related to child occupants in CRS has focused on the occupant motion relative to the vehicle seat fixture or vehicle used for testing (Ghati et al. 2009; Brown et al. 1997; Klinich et al. 2005; Hu et al. 2014; Hauschild et al. 2016; Sullivan and Louden 2009). In order to target strategies for improved design, it is important to understand the contributions of FFCRS motion to the overall motion of the head. Thus, the objective of this research was to quantify the relative occupant motion within the FFCRS during lateral oblique impacts compared to the FFCRS motion relative to the vehicle seat fixture for a center or farside positioned occupant. Additionally, the role of the FFCRS tether attachment on head excursion and FFCRS roll and yaw was analyzed. METHODS The research methods for the current study were presented in previous research (Hauschild et al. 2016). A summary follows below. A forward facing child restraint system (FFCRS) utilizing a Q3s ATD was secured to a reinforced production vehicle bench seat with a center LATCH for a series of nine sled tests. FFCRS were installed per the CRS manufacturer instructions and according to FMVSS 213 procedures (NHTSA, 2014a) when applicable. The FFCRS was secured to the bench seat in the center seating position using the available LATCH belt (Figure 1).

57 Figure 1. Vehicle seat fixture, FFCRS, and simulated door set up. A simulated intruded door was secured on the left side of the bench seat. The static intrusion level was based on side impact New Car Assessment Program tests of small sport utility vehicles at the mid rear door crush level (163 to 296 mm, average 220mm, SD 40 mm). Door and armrest padding on the simulated door was similar to that utilized in other testing, Dow Ethafoam (2.2 lb/cu ft density) and Armacell Oletx (4.0 lb/cu ft density) respectively (Sullivan et al. 2011, Hauschild et al. 2013). The simulated door panel surface was located 508 mm from the centerline of the center LATCH anchors. Tests were conducted at oblique side impact angles. The vehicle bench seat was set at 80 degrees and 60 degrees from pure frontal. Four tests were conducted at 80 degrees and five tests were conducted at 60 degrees. Test angles selected for this series were based on previous research (Arbogast et al. 2005; Hauschild et al. 2015; Maltese et al. 2007; McCray et al. 2007; Sullivan et al. 2011). The test matrix is presented in table 1. Sled Test No. Angle from Front Table 1. Test matrix Int. Door Tether Delta v (km/h) Peak G Avg. G Yes Yes Yes Yes Yes No Yes No Yes No Yes Yes Yes Yes Yes No Yes No The FFCRS 5 point harness was utilized to secure the Q3s ATD. CRS manufacturer instructions and FMVSS 213 procedures (NHTSA 2014a) were followed as applicable. ATD head accelerations, and upper neck loads and moments were collected according to SAE J211 recommended practices (Society of Automotive Engineers, 2014) and head injury values (HIC15) were calculated. Each test was conducted utilizing the proposed FMVSS 213 side impact pulse (NHTSA 2014a) scaled to a target 35 km/h delta-v. Pulse width remained at a maximum of 60 milliseconds (Figure 2).

58 Figure 2. Sample pulse including upper and lower boundaries for proposed FMVSS 213 side impact pulse. Three-dimensional motion capture cameras (TS40, Vicon, Denver, CO) recorded kinematics of the ATD, FFCRS and seat. The ATD, FFCRS and vehicle seat fixture had retroreflective markers secured on each in a noncollinear pattern. Ten markers were secured on each FFCRS. The ATD, FFCRS, and vehicle seat position were measured using a coordinate measuring machine (CMM) (FARO Technologies, Lake Mary, FL). Data from the CMM and 3-D motion cameras were processed to create local coordinate systems on each item of interest. The ATD head center of gravity was calculated from the markers on the ATD head. Markers on the FFCRS were processed and utilized to create a FFCRS coordinate system to determine ATD motion with respect to the FFCRS. Head angular motion was calculated from the collected displacement data. Positive directions follow SAE conventions; positive X, Y, and Z directions are forward, right and down. To calculate FFCRS motion, the child restraint was treated as a rigid body, and the roll, pitch and yaw with respect to the vehicle seat were calculated from the retroreflective markers secured on the FFCRS. Locations of the ATD head and FFCRS were offset to their starting position for each test. A Wilcoxon rank sum test was performed to evaluate the effect of tether use on the observed kinematics. The analysis was conducted on STATA/IC 13.1 for Mac revision 19 Dec 2014 (StataCorp, College Station, TX). RESULTS Overall kinematic results and discussion of the potential for head impact from this test series has been previously described (Hauschild et al, 2016). Results presented here will be limited to data examining the relative motion of the Q3s ATD within the FFCRS and the FFCRS motion relative to the vehicle seat fixture. Both the ATD and FFCRS moved toward the input pulse on the left of the Q3s. In all tests the FFCRS motion was limited by the simulated intruded door, and in 7 of 9 tests, including all tests without a tether, the ATD head contacted the simulated intruded door. Head The head lateral excursion relative to the FFCRS and relative to the vehicle seat fixture is presented in table 2. Head center of gravity forward, lateral and vertical excursion with respect to the FFCRS was not significantly different (p=0.07, 0.81, and 0.46) for seats restrained by the tether compared to those which were not. This finding was similar for forward and vertical head displacements with respect to the vehicle seat fixture (p =0.62 and 0.33 respectively); however the lateral head excursion with respect to the vehicle seat fixture did significantly differ with tether use (p=0.05). The lateral displacement of the FFCRS accounts for 31% of the lateral ATD head excursion for tethered FFCRS and 51% for the untethered FFCRS. Lateral excursion results are presented in table 2 below.

59 Table 2. Head CG and FFCRS lateral displacements sorted by tether use. Test Angle Tether Head Impact Max Head Max Head Max Top CRS Target Ratio of Lateral Disp. Lateral Disp. Lateral Disp. Lateral Disp. wrt FFCRS wrt SLED wrt SLED wrt SLED mm mm mm CRS/ Head Yes No % Yes No % Yes Yes % Yes Yes % No Yes % No Yes % No Yes % No Yes % No Yes % The head roll angle relative to the FFCRS and relative to the vehicle seat fixture is presented in table 3. Head center of gravity roll, pitch and twist angles with respect to the FFCRS were not significantly different (p= 0.14, 0.90 and 0.46 respectively) for seats restrained by the tether compared to those which were not. When the tether was utilized the FFCRS roll angle decreased and the yaw angle increased. FFCRS angular motion was significantly different depending on tether use; roll, pitch and yaw (p = 0.01, 0.02 and 0.01 respectively). The FFCRS maximum roll, pitch and yaw angles and associated timing are presented in Table 4. Table 3. Head CG and FFCRS roll angles sorted by tether use. Max Head CG Roll wrt FFCRS Max Head CG Roll wrt SLED Max FFCRS Roll wrt SLED Ratio of Roll Angle wrt SLED Test Angle Tether Head Impact deg deg deg CRS/ Head Yes No % Yes No % Yes Yes % Yes Yes % No Yes % No Yes % No Yes % No Yes % No Yes %

60 Table 4. FFCRS roll, pitch and yaw maximum angles sorted by tether use. FFCRS Roll wrt SLED FFCRS Pitch wrt SLED FFCRS Yaw wrt SLED FFCRS Angle Resultant wrt SLED Time Time Time Time Head Test Angle Tether Impact Deg. ms Deg ms Deg ms deg ms Yes No Yes No Yes Yes Yes Yes No Yes No Yes No Yes No Yes No Yes Upper Neck Upper neck tension and flexion/extension moment (Y) were not significantly different based on tether use (p= 0.81 and 0.27 respectively). Upper neck lateral bending moment (X) was significantly different based on tether use (p=0.01). Detailed upper neck results were presented in Hauschild et al. (2016). DISCUSSION In order to improve performance of FFCRS in lateral oblique crashes, this research examined the occupant motion with respect to the FFCRS as well as the overall FFCRS motion. Previous research studies examining this crash direction have only examined the child ATD motion with respect to the seat fixture (Ghati et al. 2009; Brown et al. 1997; Klinich et al. 2005; Hu et al. 2014; Hauschild et al. 2016; Sullivan and Louden 2009). This study is an extension of previous research examining the influence of the tether and the possibility for head impact during oblique side impacts (Hauschild et al. 2016). Previous studies examining head excursion in oblique side impact testing found excursion levels which could potentially expose the center positioned child occupant to intruding vehicle components or a farside child occupant to impacts with adjacent occupants (Arbogast et al. 2010, Ghati et al. 2009; Hauschild et al. 2015; Sherwood et al. 2003). This study highlighted that the displacement of the FFCRS is a factor in the peak lateral head excursion values. One-third to one-half of the overall lateral head excursion is derived from the motion of the FFCRS itself demonstrating the importance of controlling the FFCRS motion for limited occupant excursion. When considering angular motion, the contribution of the FFCRS motion was less, representing only % of the overall head roll angle. This study also examined the influence of the FFCRS tether on these relationships. The head CG lateral excursions with respect to the vehicle seat fixture were significantly lower when the tether was utilized (median 400 mm with tether and 442 mm without tether). The tether is primarily designed to control forward excursion of the FFCRS in frontal crashes; these data demonstrate that it is also effective in controlling lateral motion in side impacts. In contrast the head CG lateral excursions with respect to the FFCRS trended slightly higher when the tether was utilized (median 234 mm with tether and 216 without the tether). This is likely due to the lateral restraining force the tether provides to the FFCRS such that the head s response to the crash energy requires it to move farther relative to the child restraint. The FFCRS displacement contributed approximately 50% of lateral excursion when the tether was not utilized; whereas when the tether was used that ratio was smaller (approximately 30%). When the tether was utilized, the FFCRS had increased yaw and decreased roll with respect to the

61 vehicle seat fixture, which corresponded to less lateral displacement at the top of the FFCRS and less overall lateral head displacement. The roll angles and lateral displacements were limited by the interaction with the simulated intruded door and may have been higher if the tests had been conducted without the simulated door or if the FFCRS was placed in a farside seat position where interaction with the door is less likely. Tests with higher FFCRS roll angles and lateral head excursions trended to higher HIC values due to the intruded door contact (Hauschild et al. 2016). The kinematic observations are summarized in Figure 3 below displaying the FFCRS motion. The figure demonstrates the additional lateral movement of the upper portion of the untethered FFCRS (dotted lines) which leads to greater ATD lateral head excursions. The figure also shows the untethered (dotted lines) FFCRS has more lateral travel and rebound past its starting position. This may lead to injury for a far side occupant who impacts the adjacent vehicle side structure on rebound. The study by Brown et al. (1997) also indicated that the far side occupant could potentially have injuries on rebound. The FFCRS motion also influences upper neck moments. The lower roll angles and higher yaw angles of the tethered FFCRS had higher upper neck lateral bending moments which directly correlate to lateral head displacements within the FFCRS (r2 = 0.89). In contrast, there was not a relationship between neck moments and lateral head excursion with respect to the vehicle seat fixture. As the FFCRS is held in place by the tether the occupant tends to roll out more relative to the FFCRS thereby increasing the lateral bending neck moment. Although there currently is no criteria for neck lateral bending moments (X) specifically for the Q3s ATD, values are approaching or in some case exceeding the recommended IARV for a 3 year old in lateral bending (32 Nm) (Mertz et al. 2003). Neck flexion/extension moments (Y) were slightly higher with tether use but did not exceed the recommended maximum IARV (21 Nm) (Mertz et al. 2003). The results suggest other pathways related to vehicle and FFCRS interaction to control the head and neck motion may be required to reduce neck tension and moments. Figure 3. Exemplar 2-dimensional displacement of the targets on the FFCRS from the 80 degree tests, demonstrating FFCRS motion in lateral and vertical planes (solid lines tether use; dashed line non-tether use). Select marker locations are circled in image on right.

62 Limitations Limitations of this study include the testing of one FFCRS, one input pulse, one delta-v, and a single vehicle seat fixture. The FFCRS was a common child restraint and had no distinguishing design features which suggest its response would be different from others. The FFCRS was attached using the available single LATCH belt and tether webbing system. It is likely that the results would be different for other lower attachment methods; this should be the focus of future exploration. A stock production bench seat from a small SUV type vehicle was used for testing. Other seat types may have an effect on occupant response. CONCLUSION AND RECOMMENDATIONS This research found the occupant lateral displacement and angular kinematics are related to the motion of the FFCRS in oblique side impact crashes. A substantial proportion of the occupant s lateral head excursion was due to the FFCRS movement relative to the vehicle seat fixture. This result varied by tether use such that 50% of the overall lateral head excursion was due to FFCRS motion in untethered FFCRS while only 30% of the overall lateral head excursion was due to the FFCRS motion when tethered. It was noted that while tether use reduced FFCRS motion, it did not significantly increase the motion of the head relative to the FFCRS due to increased restraint of the FFCRS. Interventions that address both aspects of lateral kinematics movement of the FFCRS as well as lateral bending/forward flexion of the occupant s torso/neck relative to the FFCRS might mitigate overall lateral excursion and potential impact with intruding structures, a common injury causation scenario for children in these crashes. ACKNOWLEDGEMENTS The authors would like to acknowledge the National Science Foundation (NSF) Center for Child Injury Prevention Studies at The Children s Hospital of Philadelphia (CHOP) and the Ohio State University (OSU) for sponsoring this study and its Industry Advisory Board (IAB) for their support, valuable input and advice. The views presented are those of the authors and not necessarily the views of the NSF or the IAB members. The authors would like to acknowledge the assistance of Paul Gromowski with sled test setup, data acquisition, and high speed video acquisition during the testing series. REFERENCES Arbogast KB, Locey CM, Zonfrillo MR, & Maltese MR. Protection of Children Restrained in Child Safety Seats in Side Impact Crashes. J. Trauma, Injury, Infection and Critical Care Oct; 69(4): doi: /TA.0b013e3181e883f9. Arbogast, KB, Ghati Y, Menon RA, Tylko S, Tamborra N., & Morgan RM. Field investigation of child restraints in side impact crashes. Traffic Injury Prevention. 2005; 6(4), Brown J, Griffiths M, Paine M. Effectiveness of child restraints: the Australian experience. Australian New Car Assessment (ANCAP) Brown J, Kelly P, Griffiths M. A comparison of alternative anchorage systems for child restraints in side impacts. (No ). SAE Technical Paper Charlton JL, Fildes B, Taranto D, Laemmle R, Smith S, Clark A, Holden GM. Performance of booster seats in side impacts: effect of adjacent passengers and ISOFIX attachment. Annu Proc Assoc Adv Automot Med. 2007; 51: Charlton JL, Fildes B, Laemmle R, Smith S, Douglas F. A preliminary evaluation of child restraints and anchorage systems for an Australian car. Annu Proc Assoc Adv Automot Med. 2004; 48:73-86 European New Car Assessment Programme (Euro NCAP) Assessment protocol child occupant protection. Version December Ghati Y, Menon RA, Milone M, Lankarani H, Oliveres G. Performance evaluation of child safety seats in far-side lateral sled tests at varying speeds. Annu Proc Assoc Adv Automot Med. 2009; 53:

63 Hauschild HW, Humm, JR, Pintar FA, Yoganandan N, Kaufman B, Maltese MR, & Arbogast KB. Protection of children in forward-facing child restraint systems during oblique side impact sled tests: Intrusion and tether effects. Traffic injury prevention, 2016; 17(sup1), Hauschild HW, Humm, JR, Pintar FA, Yoganandan N, Kaufman B, Maltese MR, & Arbogast KB. The influence of enhanced side impact protection on kinematics and injury measures of far-or centerseated children in forward-facing child restraints. Traffic injury prevention, 2015; 16(sup2), S9-S15. Hu J, Klinich KD, Reed MP, Ebert-Hamilton SM, & Rupp JD. Characterizing Child Head Motions Relative to Vehicle Rear Seat Compartment in Motor Vehicle Crashes (No. DOT HS ). Huntley M. Federal Motor Vehicle Safety Standard no. 213 Child Restraint Systems. NHTSA. May 15, Klinich KD, Ritchie NL, Manary MA, Reed MP, Tamborra N, Schneider LW. Kinematics of the Q3S ATD in a Child Restraint Under Far-Side Impact Loading. In Proceedings: International Technical Conference on the Enhanced Safety of Vehicles 2005 (Vol. 2005). National Highway Traffic Safety Administration. Paper Maltese MR, Locey CM, Jermakian JS, Nance ML, Arbogast KB. Injury causation scenarios in beltrestrained nearside child occupants. Stapp Car Crash J Oct; 51: SAE Technical Paper No McCray L, Scarboro M, Brewer J. Injuries to children one to three years old in side impact crashes. In 20th International technical conference on the enhanced safety of vehicles conference (ESV) Paper Number Lyon, France. Mertz HJ, Irwin AL, & Prasad P. Biomechanical and scaling bases for frontal and side impact injury assessment reference values. 2003; Stapp car crash journal, 47, 155. NHTSA (a), Laboratory Test Procedure for Federal Motor Vehicle Safety Standards 213 Child Restraint Systems. February 16, NHTSA (b), National Highway Traffic Safety Administration. Notice of Proposed Rule Making. 49CFR Part 571 Docket no. NHTSA Federal Motor Vehicle Safety Standards; Child Restraint Systems Side Impact Protection. Washington, DC Orzechowski KM, Edgerton EA, Bulas DI, McLaughlin PM, Eichelberger MR. Patterns of injury to restrained children in side impact motor vehicle crashes: the side impact syndrome. J Trauma Jun;54(6): Sherwood CP, Ferguson SA, Crandall JR. Factors leading to crash fatalities to children in child restraints Annu Proc Assoc Adv Automot Med. 2003; 47: Society of Automotive Engineers. SAE J211-1 Instrumentation for impact test, Surface vehicle recommended practice. SAE International Sullivan, L. K., & Louden, A. E. NHTSA's Initial Evaluation of Child Side Impact Test Procedures. In Proceedings: International Technical Conference on the Enhanced Safety of Vehicles (Vol. 2009). National Highway Traffic Safety Administration. Paper Sullivan, L. K., Louden, A. E., & Echemendia, C. G. (2011). NHTSA S Evaluation of a Potential Child Side Impact Test Procedure. National Highway Traffic Safety Administration. United States. Paper Number

64 LIGHT METAL-PLASTIC BODY-IN-WHITE SOLUTIONS FOR AUTOMOTIVE Dinesh, Munjurulimana Anil Tiwari SABIC Research & Technology Pvt. Ltd. India Tom Koning Joel Thambi SABIC, Bergen Op Zoom Netherlands Matthew M Delaney Dhanendra Nagwanshi SABIC, Wixom United States Paper Number ABSTRACT Engineered systems in today s automobiles are often designed and built to meet conflicting and complex requirements. While mobility is a car s primary function, accomplishing that in an energy-efficient manner and ensuring the safety of the occupants are critical requirements. Automotive OEMs, therefore, are aggressively working on making vehicles lighter without compromising its safety. Meeting such complex requirements often requires solutions encompassing innovative designs, manufacturing processes and multimaterial systems. This paper focuses on the development of lightweight metal-plastic body-in-white (BIW) solutions. A generic vehicle validated for high-speed crash scenarios such as full frontal impact, side deformable barrier impact, side pole impact and rollover (roof crush resistance) is chosen for the feasibility study of developing hybrid lightweight solutions using metals and thermoplastics. Various weight reduction opportunities by either replacing the existing metal reinforcements in the BIW or by replacing a complete sub-system such as B-pillar were explored using metal-plastic hybrid combinations. Developed reinforcements include those in the floor rocker, rails, floor etc. A combination of high heat unfilled thermoplastic resins (tough and ductile) or fiber reinforced thermoplastic resin (high stiffness and strength) and metal are chosen appropriately depending on the requirements. For instance, an unfilled thermoplastic resin over-molded with multiple metallic inserts was chosen to replace the incumbent energy-absorbing members in the floor rocker for side impact, and fiber compounded thermoplastic resin over molded with a metallic insert is chosen to replace the existing B- pillar with comparable crash performance. The developed lightweight hybrid B-pillar replaces a multi-piece B-pillar made of high-strength steel. The metal inserts in the hybrid systems are exploited for assembly ease in the BIW structure. Such a solution not only offers part integration possibilities with equivalent crash performance as that of the baseline system, but also opens the door for replacing the high-strength steel used in the BIW with a medium-strength steel. A significant weight reduction potential (approximately 30%) is observed as the baseline BIW structures were down-gauged with overmolded thermoplastics. Thermoplastic material overmolded on steel plays a crucial role in avoiding localized buckling of the BIW structures and in absorbing impact energy as and when required. The developed solutions validated using CAE studies are further correlated using component level studies with a generic 800 mm long metal-plastic system weight 1.6 kilograms. This system is subjected to 3-point Munjurulimana

65 bending and force vs. deflection characteristics and the deformation kinetics in the above loading scenario is correlated using sub-system level CAE studies. Munjurulimana

66 INTRODUCTION Automotive safety regulations, in general, can be categorized as those that ensure occupant safety and those that regulate pedestrian safety. While the latter is achieved by designing an optimum bumper and a bonnet, the former warrants a combination of appropriate design of vehicle body-in-white (BIW) and incorporation of additional safety features such as airbags, seat belts, etc. inside the vehicle. As the BIW accounts for majority of the mass of a vehicle, it also plays an important role in defining the energy/fuel needs of a vehicle. While there are significant developments in solar energy, fuel cells and other such renewable sources of energy, fossil fuels still remain the most common and preferred source of energy for automobiles. This continues and the ever-increasing use of fossil fuels has serious undesirable impact to our environment resulting in global warming and more importantly on the sustainability of humankind. Thus, to make sure that the current usage of fossil fuels does not jeopardize the potential for people in the future to meet their energy needs, the U.S. government (later supported by other regulatory bodies in different parts of the world) introduced the concept of Corporate Average Fuel Economy (CAFE) standards in 1975 [1]. Its primary objective is to reduce the energy consumption by increasing the fuel economy of light trucks and cars, which also indirectly results in reducing greenhouse gas emissions. An in-depth study of worldwide statistical data indicates that automobile manufacturers need to come up with bold solutions in the next few years, as CO2 emission reduction targets for the next 10 years are nearly double of what has been achieved in the last 10 years [2]. Studies and surveys performed by several institutes [3] show that light weighting is so far the most promising option for automobile manufacturers to address 2025 CAFE industry standards (refer Figure 1). It is worthwhile to note that some of the survey respondents focus on multiple technologies and hence a cumulative score of more than 100% as one could observe in the figure. Lightweighing, though seemingly relatively simple, is not the most convenient option to implement in a vehicle due to several factors mentioned below. vibration and harshness (NVH) performance of the vehicle. 2. Strength and stiffness of application/part being replaced with a lighter solution should not be compromised as it can negatively impact the long-term performance and more importantly the crash performance of the vehicle. Others Adopting Bio-fuel programs Downsizing vehicles Electrifying the vehicle Lightweighting 3% 15% 7% 10% 11% 13% 26% 39% 49% Figure 1.Technologies being focused by industries to help to meet 2025 industry standards. Adapted and recreated from [3]. This paper investigates one of the short-term lightweighting approaches for automobiles. The approaches explained in the paper are focused on replacing hang-on parts with a lightweight and optimally designed system while making sure that this replacement does not result in reduced crash performance of the car. The remaining part of this paper is divided into five sections as follows. The first section explains why the body-in-white (BIW) reinforcements are targeted for the lightweighting of automobiles. The next section of the paper deals with identifying a realistic weight reduction potential in an automobile using BIW reinforcement concepts. This is performed by developing solutions for one of the vehicle platforms for which a validated computer aided engineering (CAE) model was developed by the National Crash Analysis Center at George Washington University. The third section includes the preliminary crash performance evaluation of the conceived lightweight vehicle, and the comparison of the performance with the baseline solution. The next section explains how the performance of such lightweight BIW reinforcement solutions can be validated using component level tests. The last section contains an overall summary, thoughts on future work required and some concluding remarks. 1. It can have an adverse effect on other factors such as the dynamic stability and noise, Munjurulimana

67 AUTOMOTIVE LIGHTWEIGHTING It is estimated that a vehicle s typical subsystem mass distribution is led by the body [4]. On average, it amounts to 37% of the total mass of a vehicle. This is followed by the chassis (30%), powertrain (14%), interior (12%), electrical (4%), and the remaining 3% contributed by Heating, Ventilating and Air Conditioning (HVAC) and powertrain cooling systems. A similar distribution are reported by other research papers too [5-6]. Though the numbers and ranking reported by other studies can vary, most of those studies unanimously show that body and chassis contribute to roughly 65% of the total mass of the car. It is, therefore, important that one view the BIW as a major lightweighting region of the car. Numerous options including alternate materials, optimum geometrical configurations and diverse manufacturing methods are being investigated in the literature to take out the mass from the BIW without compromising the car s performance [7-10]. Automobile manufacturers also need to make sure that the resulting increase in cost is maintained within acceptable levels. Each application in an automobile is unique in its own way. Interior trim applications, which are typically made of plastics, need not offer high stiffness and strength, but should provide the aesthetics and premium looks for the occupant sitting inside the car and should also have provisions for sufficient storage holders. Similarly, polyurethane foam used in seats should offer the passenger sufficient comfort and cushioning effect. Likewise, a car s BIW has to provide sufficient support and mountings to other parts in a car including the engine and powertrain, suspension, body panels, glazing and so on. The BIW is also the major energyabsorbing member in an automobile in the event of a high-speed crash. Figure 3 shows typical materials used and a few major relevant applications using the same in an automobile. As you would notice, each material has its own pros and cons, making it more appropriate or not appropriate for certain applications. For example, it would be highly challenging to achieve the required cushioning and comfort of a seating system using high strength steel (HSS). Similarly, it would be tough to imagine polyurethane foam replacing the BIW, which is typically made using steel, HSS or aluminum. Certain applications can be designed and made of multiple materials. The BIW of a car is a one such application. A typical low-cost and heavy vehicle uses conventional stamped steel parts to constitute its BIW. More expensive and probably lighter cars use HSS or aluminum for manufacturing its BIW. Even more expensive cars such as sports cars, which demand the lightest possible vehicle with superior dynamic stability, use composites predominantly to make most of its parts. Figure 4 shows cost implications and lightweighting potential in a car using different materials. It is worth noting that the conventional medium strength steel is used as the baseline for this comparison. Figure 2. Mass distribution in a typical automobile. Adapted and recreated from [4]. Figure 3. Various materials, applications and why those materials are used for those applications in automobiles. Adapted and recreated from [10]. Why BIW Reinforcements for lightweighting? As mentioned in the earlier section, a vehicle s BIW is typically made using stamped steel parts. Several stamped steel parts are welded together to form the complete BIW. In general, it is difficult to achieve local stiffening or softening effect in a stamped steel parts. This is primarily because the raw material used for the stamping or metal forming operation is a blank with uniform thickness. The only way to vary the stiffness along the length of a stamped steel part is by smart geometrical variations, which beyond a limit is infeasible as it is limited by the draw ratio. This is true with other materials, too, such as Munjurulimana

68 aluminum wherein the parts are typically made using an extrusion process. Needless to mention, achieving local stiffness variation in an extruded aluminum part is even more challenging. Automobile designers, therefore, generally make use of local reinforcements in a car s BIW to improve the stiffness and strength at certain selective locations. The center portion of the B-pillar, roof, A-pillar center and rocker as shown in Figure 5 are a few, examples of such reinforcements [11]. Similar reinforcements exist in other parts of the BIW such as rails, floor and C-pillar. These reinforcements are typically made of HSS and are separately welded onto the part. 3. One does not need to be concerned about joining techniques as the same welding process or adhesives can be used to join the new solution to the BIW. Figure 6 shows a schematic representation of few potential BIW reinforcement applications using a thermoplastic, metal-plastic or composite-plastic solution. Details of the development of such solutions for a realistic vehicle platform and the potential weight reduction possibilities is demonstrated in the next section. Figure 4. Impact of lightweight materials on the part cost for a typical automotive application. Adapted and recreated from [10]. Each reinforcement in the BIW has different functions. For example, in the case of the B-pillar, the reinforcement is provided in the center to prevent the undesired local buckling of the B-pillar during a side impact and a roof crush/roll over scenario. The rocker reinforcement absorbs the greatest share of energy during a pole impact event. A reinforcement in a rail can absorb energy during a high-speed frontal crash. It may also provide an additional local stiffening effect in the vertical direction at engine mount locations in the rails. These reinforcements in the rails, therefore, can also reduce the transfer of engine vibration to the BIW of the vehicle to a greater extent. Considering all these factors, one can say that BIW reinforcements can be appropriate applications to target for lightweighting in an automobile as: 1. Replacement of BIW reinforcements with lighter and hybrid reinforcements does not require any major changes in the existing assembly line. 2. Potential weight reduction possibilities are significant as multiple reinforcements are present in a vehicle. Figure 5. Few BIW reinforcements in typical automobile. Adapted from [11]. Figure 6. A realistic representation of BIW reinforcements in a vehicle using plastic, metalplastic and other hybrid concepts. DEVELOPMENT OF BIW REINFORCEMENTS This section aims to demonstrate the weightreduction potential in a realistic vehicle platform by replacing a few of its BIW reinforcements by lighter plastic or hybrid solutions. A finite element model of one of the car models developed by FHWA/NHTSA National Crash Analysis Center at George Washington University [12] is used as a baseline vehicle for this study. The identified vehicle is a Munjurulimana

69 sedan weighing approximately 1070 kg. This vehicle was chosen for the study because: 1. This vehicle was one among the well correlated vehicle finite element (FE) models from the set of several available. 2. This vehicle model has reinforcements in the A- pillar vertical member, front rails, B-pillar, and rocker. Thus, the weight reduction potential by replacing all four reinforcements can be studied. 1. Metal sheets are typically downgraded to at least 1 mm or 0.8 mm depending on the grade of the steel used. This provides a significant weight saving as the baseline solutions are typically 1.5 or 2 mm thick. 2. Plastics honey combs are over molded on this down gaugeed steel stamp parts to avoid the local buckling of these structures Figure 7. Identified vehicle models for the lightweight BIW reinforcement development study. Adapted from [12]. As mentioned earlier, four reinforcements were selected to study the lightweighting potential in this vehicle platform. These are floor reinforcements, reinforcements in the vertical A-pillar, front rail reinforcements, an integrated and lightweight rocker solution and a metal-plastic B-pillar system replacing four out of the existing 5-piece B-pillar in the vehicle. While lighter solutions are achieved in the first two applications purely by replacing the existing steel inserts by injection molded thermoplastic systems, the last three applications realize the weight reduction by combining a multiple steel stamped solution to a single-piece metal-plastic over molded solution. The metal in the metal-plastic solutions are down-gauged significantly compared to the existing solutions, and thermoplastics are molded onto it to compromise the reduced stiffness as a result of the down-gauging of the steel part. These solutions, therefore, not only offer significant lightweighting opportunities, but also offer part integration possibilities in many cases. Figure 8 to Figure 11 show the details of the conceived lighter solutions. Appropriate meshing, morphing and preprocessing software [13] was used to conceive these solutions so that they fit within the packaging space available in the vehicle. The engineering techniques/approach used to reduce the mass of reinforcements are as follows. Figure 8. Existing reinforcement in the A-pillar and proposed thermoplastic reinforcement. Total weight reduction of 1.6 kg/car. Reinforcement dimensions 450 mm * 110 mm * 60 mm. Figure 9. Existing 3-piece steel front rails and proposed metal-plastic lighter front rails. Total weight reduction of 2.0 kg/car. Reinforcement dimensions 550 mm * 120 mm * 70 mm. Munjurulimana

70 Based on these developed solutions, it is observed that approximately 15.6 kg of a car weighing close to 1070 kg can be reduced by this technology. Similar technologies, when evaluated for another vehicle can yield different numbers, but the message remains the same. Figure 10. Existing 4-piece steel B-pillar and proposed metal-plastic single-piece B-pillar. Total weight reduction of 6.6 Kg/car. Reinforcement dimensions 1000 mm * 140 mm * 115 mm. It is worth noting that the assembly sequence is only minimally altered by introducing these lighter solutions. For example, in the case of the B-pillar, the top and bottom portions still have isolated metal sections, which can be welded onto the existing BIW. In many cases, the assembly is actually made simpler as several parts are integrated in the proposed solutions. Furthermore, based on the requirements, materials can be selected capable of passing the e- coat bath for anti-corrosion. Figure 11. Existing rocker outer and reinforcement solution and proposed metal-plastic rocker outer & reinforcement. Total weight reduction of 5.4 Kg/car. Reinforcement dimensions 1650 mm * 150 mm * 120 mm. Figure 12 shows a summary of weight reduction possibilities achieved in the considered vehicle using the aforementioned four lightweight solutions. Figure 12. Summary of weight reduction potential in a car using BIW reinforcement concepts. NUMERICAL PERFORMANCE EVALUATION As mentioned in the earlier sections, BIW caters to multiple functionalities in an automobile. This section, however, only focuses on evaluating a few major high-speed crash scenarios during which the BIW plays a crucial role in absorbing a significant portion of the impact energy, and thus mitigating the injury of the occupant to a greater extent. A performance evaluation for the secondary functionalities of the BIW is beyond the scope of this paper. The vehicle with the conventional BIW inserts and the same vehicle with the four newly proposed, alternate lightweight solutions are subjected to 56 km/h full frontal impact [14] 50 km/h side IIHS deformable barrier impact [15], 30 km/h pole impact [16] and roof crush impact scenario [17]. LSDYNA a commonly available explicit solver is used for these simulations [18]. The metal and plastic parts of the vehicle were modelled using MAT 24, a commonly used piecewise linear plastic material model in LSDYNA. A strain-based failure model was used to model the failure of these parts. To avoid the complexity, the delamination of plastics in the metal-plastic hybrid inserts was not modelled. Figure 13 shows the expected impacts of replacing the BIW parts with the conceived lighter solutions on the four major crash scenarios explained above. A tick mark in any column or row indicates that the corresponding solution (shown in the respective row) can have a significant impact on the respective Munjurulimana

71 (shown in the column) crash impact performance of the vehicle. Figure 13. Replacement of BIW parts with lighter solution and their impacts on the crash performance of a car during various crash scenarios. High speed full frontal Impact The acceleration experienced by the occupant or measured by the accelerometer positioned close to the left/right rear seat floor is an important criteria evaluated in a full frontal impact at 56 km/h. Typically, for a safer car, the deceleration levels have to be maintained below 40 g. The deformed/crushed vehicle configurations at the maximum intrusion points are shown in Figure 14. Figure 15 shows the deformed configuration of the relevant part of rail which is replaced with a lighter solution. Both figures indicate that the vehicle behavior and its performance is not drastically affected. The acceleration measured near the left rear seat as shown in Figure 16 also substantiate this fact. The reduced acceleration in the case of the proposed solution is mainly because the rails absorb more energy than just buckling about a point in the rear as in the case of the baseline solution. Side impact to a vehicle using Insurance Institute for Highway Safety (IIHS) deformable barrier impact emulates a crash scenario between two vehicles in orthogonal directions. The function of BIW components including the B-pillar, A-pillar, door, rocker, etc. in this case is to limit the side intrusion in a vehicle. This is extremely important as there is hardly any space available between the occupant and the side structural parts of the car, and any direct contact of the structural member to the occupant s body can cause severe injuries to the occupant. It is also important that these side members are not over-designed so that the occupant will experience high side accelerations in these cases. In order to make sure that the proposed solutions do not adversely affect the side impact performance of the vehicle, IIHS deformable barrier is impacted to both vehicle configurations at 50 km/h. Figure 17 shows the deformed configurations (at maximum intrusion point) of the baseline solution and the proposed solution respectively. As one can make out from the figure, the performance of the vehicle with lightweight systems is very much comparable to that of the original configurations. Section views (refer Figure 18) along the width of the car at the B-pillar location also demonstrate that the lightweight B-pillar, rocker and floor reinforcements behave very similar to the behavior of those respective parts in the baseline vehicle configuration. This is further supported by the force vs. intrusion curves during the side impact scenario as shown in Figure 19. The proposed solution generates higher peak force and relatively higher force levels towards the end of the impact mainly because of the additional stiffness from the plastic over molds in the B-pillar inserts. Figure 14. Deformed configuration of the baseline vehicle and a vehicle with lighter front rail insert solutions in a high-speed frontal impact Side 50 km/h using IIHS barrier Munjurulimana

72 Figure 15. Deformed configurations of the left rails in case of baseline and the proposed solution in a high-speed frontal impact Figure 16. Normalized acceleration measured near the left rear seat in a high-speed frontal impact. This gives an indication of acceleration experienced by the occupant. The proposed solution shows better crushing resulting in reduced occupant acceleration. Side Rigid Pole Impact at 29 km/h Pole impacts are performed in a vehicle mainly to safeguard the occupant during the event of its impact with a rigid tree or any other relatively slender, vertical and rigid structures on either sides of the road. The major challenge in this case is to make sure that the required amount of energy is absorbed by the vehicle s structural members before the pole comes in direct contact with the occupant s body. The rocker, one of the very first portions that comes in contact with the rigid pole, plays a crucial role in limiting the intrusion in a vehicle during such an event. The two vehicle configurations are therefore compared for its rigid side pole impact performance. Figures 20, 21 and 22 show the performance of the vehicle in this impact situation. Observations and conclusions from these results are no different from what was observed in the earlier impacts. It is worth noting that the force levels are again higher in the case of proposed solution. These force levels can be reduced, if required, to reduce the acceleration experienced by the occupant. Softer plastic honeycombs will help to achieve this. Figure 17. Deformed configuration (at maximum intrusion time) of the baseline vehicle and a vehicle with lighter solution subjected to IIHS deformable barrier side impact. Figure 18. Sections views along the width of the car at the B-pillar section. Both baseline and the proposed configurations seem to behave in a similar way. The proposed solution also shows promises of reducing the side impact intrusions. Figure 19. Normalized Force versus Intrusion curves during the IIHS side deformable barrier impact. Munjurulimana

73 . Figure 20. Deformed configuration (at maximum intrusion time) of the baseline vehicle and a vehicle with lighter solution subjected to rigid pole impact. The objective of roof crush performance evaluation is to make sure that the occupant has sufficient headroom before the vehicle s structural parts (mainly the roof, A-pillar and B-pillar) deform to absorb the energy during a rollover situation. Different regulations evaluate the roof crush in different ways. In general, the vehicle is supposed to be performing well for the roof crush requirements if it can generate a peak force of at least 2.5 times (> 2.5 times the weight of the vehicle marginal performance and > 4 times the vehicle s weight good performance) the weight of the car before the roof intrudes by 5 inches. Rigid plate impacts to the roof of the baseline and lightweight vehicle are performed as per the regulatory protocols and the performance curves and vehicle behaviors are shown in Figure 23 to Figure 25. Results again demonstrate that a lightweight BIW reinforcement solution does not necessarily reduce the roof crush performance of the vehicle. The maxim value of the strength to weight ratio in the case of the proposed solution is higher within 5 inches (about 127 mm) of intrusion. This is mainly because of the additional stiffness from the plastic honeycomb parts. Figure 21. Sections views along the width of the car at the B-pillar section during a rigid pole impact. Both the baseline and proposed configurations seem to behave in similar ways. The proposed solution also shows promise of reducing side impact intrusions. Figure 23. Deformed configuration (at maximum intrusion time) of the baseline vehicle and a vehicle ith lighter solution subjected to roof crush impact. Figure 22. Normalized Force versus Intrusion curves during the rigid pole impact. Figure 24. Sections views along the width of the car at the B-pillar section during a roof crush impact. Roof Crush Resistance Munjurulimana

74 Figure 25. Strength to weight ratio versus Intrusion curves during the roof crush EXPERIMENTAL STUDIES Component-level validation Full vehicle level tests demand higher investment in hardware and take longer time. Therefore, in this section, component level tests on a representative all-plastic and metal-plastic hybrid beam are performed for both static and dynamic scenarios, and are correlated with CAE results. The chosen test specimen is a C- section filled with plastic ribs. For an all-plastic beam, both the channel section and ribs are made of plastic whereas for metal-plastic hybrid beam channel section is made of metal which is then over molded with plastic to form inner ribs. These test samples are represented in Figure 26. Two load cases (static & dynamic) as shown in Figure 27 are considered. 1. Three point bending load case with indenter moving at 10mm/min. As the speeds are relatively low, this scenario may be considered as static loading. 2. An impact with indenter weighting 23 kg with a speed of 13.5 kmph. This represents a dynamic impact scenario. Figure 26. All-plastic and metal-plastic samples used for experimental validation. Experimental Setup: Static Loading Two custom-made fixtures support BIW inserts and an indenter loads the insert as shown in the Figure 28. This 3-point bending scenario represents most of the loading that is being applied in BIW insert in the event of a crash. In the component level, these loads are applied using a 100 kn capacity hydraulic press with integrated measurement system from INSTRON. An indenter as shown in the Figure 28 is used to transfer load from hydraulic press to the specimen. A thick layer of foam covering the indenter ensures uniform transfer of load from indenter to the specimen. Force cells with displacement sensors are mounted on the indenter to capture the force and displacement. The beams are loaded with the bottom support fixed and the indenter is allowed to move in the downward (bending) direction at a speed of 10m mm/min. To capture and understand how the specimen fails during these loading scenarios, the bottom face of the beam is focused within the scope of a high-speed camera. Experimental Setup: Impact Loading A medium energy uniaxial impactor is used to apply impact load onto the BIW insert. Impact energy supplied to BIW insert can be adjusted to desired level by adjusting mass and/or speed of impactor. The hardware can be adjusted for different impact Munjurulimana

75 objects and clamping possibility. High-speed camera is again used to capture deformation and failure of BIW insert subjected to impact loads. The uniaxial impactor can slide along a uniaxial guidance system. The guidance system is mounted on an impactor frame. A Hydraulic cylinder launches the impactor frame with a specified speed. Launch of impactor frame triggers data acquisition system to measure force, intrusion and acceleration. BIW inserted is supported in an orientation such that the impact happens at the center of the beam. A schematic representation of such a system is shown in the Figure o C to 200 o C for 20 minutes to 30 minutes. It is also worth noting that BIW can be exposed to different environmental conditions (humidity, excess temperature) in the use phase of a car. Hence, it is important to ensure that metal-plastic BIW reinforcements are immune to such conditions. To study the combined effect of e-coat curing and moisture absorption of the metal-plastic hybrid BIW insert, the BIW insert is exposed to the following conditioning cycle: 1. Oven is preheated to 200 o C. 2. BIW insert is kept in the preheated oven for 30 minutes. 3. Insert is cooled to room temperature. 4. Insert is exposed to 95% relative humidity for 40 hours 5. Insert is kept at 50% relative humidity until equilibrium or for 40 hours Conditioned specimens are also tested for both static and impact loading and their performances are compared against non-conditioned samples. Figure 27. Three point bend and impact load applied to BIW insert. Effect of e-coat on metal plastic hybrid: As most of the vehicles have BIW made of steel, they are often subjected to electrophoretic painting process (e-coat) to mitigate the potential long-term corrosion issues and to improve the adhesion of paints on its surface. In this process, BIW is typically immersed in an aqueous solution containing paint emulsion. Paint emulsion gets condensed onto the part by applying electrical voltage. All the surfaces where the solution can reach get painted. Coating thickness is controlled by the magnitude of applied voltage. One of the most important steps of e-coat process is curing. Depending upon the type of paint used, curing temperature can be anywhere between Figure 28. Experimental setup of the 3-point bend loading for BIW inserts. Munjurulimana

76 Figure 29. Schematic representation medium energy uniaxial impactor. Static Loading Samples are tested with and without e-coat conditioning to study the effect of e-coat cycle on metal plastic hybrid sample. Three samples are tested with e-coat conditioning and three samples are tested without e-coat conditioning at 21 o C. An average representative of the three test iteration is been considered for reporting. The test result in the form of a normalized force vs. a normalized intrusion curve is shown in Figure 30. It is worth noting that the e-coat conditioning cycle slightly improves the load-bearing capacity of the metal-plastic hybrid insert at room temperature (21 o C). One possible explanation of such behavior is that the e-coat conditioning cycle causes annealing to the molded region of the hybrid insert. This results in relaxation of the process-induced residual stress, which ultimately results in improving the strength of molded part and thus improving the load-bearing capacity. Research [19] confirms improvement in mechanical properties of the molded part due to different level annealing temperature. Figure 30. Normalized Force vs. Normalized Intrusion curve for static test considering e-coat sample and non e-coat sample at 21 deg C. Another set of testing is performed at -20 o C with and without e-coat of metal plastic hybrid sample. Two samples are tested with e-coat and two samples are tested without e-coat. An average representative of the three test iteration is been considered. Test result in the form of normalized force vs. normalized intrusion curve is shown in Figure 31. It shows that the e-coat cycle has very little effect on performance at minus 20 o C. Figure 31. Normalized Force vs. Normalized Intrusion curve for static test considering e-coat sample and non e-coat sample at -20oC. To confirm that e-coat has little or no influence on the performance of metal-plastic reinforcements, one more set of experiments are performed at 80 o C with and without e-coat conditioning of metal plastic hybrid sample. Two samples are tested with e-coat and two without. The test result (refer Figure 32) shows that the e-coat cycle has very little effect on performance at 80 o C. Munjurulimana

77 Figure 32. Normalized Force vs. Normalized Intrusion curve for static test considering E-coat sample and non E-coat sample at 80 deg C. Figure 33 shows comparison of the static performance of the metal-plastic BIW insert at -20 o C, 21 o C and 80 o C. As one would expect, energy absorption of the sample is best at 21 o C, whereas load-bearing capacity is highest at -20 o C. Figure 33. Normalized Force vs. Normalized Intrusion curve for static test at -20 deg C, 21 deg C and 80 deg C. Impact Loading The effect of the e-coat cycle on impact performance of the metal plastic hybrid insert is also studied. Impact testing as described in the previous section, is performed at 21 o C for samples with and without e- coat. Test results in the form of normalized force vs. normalized intrusion curve is shown in Figure 34. These results reinforces that the e-coated metalplastic hybrid insert shows similar performance as that of non e-coated sample at room temperature. Figure 34. Normalized Force vs. normalized Intrusion curve for impact test considering e-coat sample and non e-coat sample at 21 deg C. Correlation Studies Experimental validation of predictive methodology is required to build confidence in vehicle level simulation. Finite element (FE) simulations are performed for static three-point loading scenario for all-plastic beam and metal-plastic beam and the results are compared with the experimental results explained in the previous section. CAE versus test results for plastic beam is plotted in Figure 35. The effect of foam is ignored in normalized force vs. normalized intrusion for both prediction and testing. Correlation of normalized force vs. normalized intrusion for the all-plastic beam is found to acceptable. There is an excellent match for stiffness (initial slop of the curve) of the insert. Predicted strength (maximum force) is within 2% of tested strength, and predicted intrusion at failure is within 4% tested intrusion at failure. FE model could also predict the location and nature of failure with reasonable level of accuracy. Figure 36 shows predicted and tested failure location for all-plastic insert. CAE versus test results for metal-plastic beam is plotted in Figure 37. Effect of foam is ignored in normalized force vs. normalized intrusion for both prediction and testing. Correlation of normalized force vs. normalized intrusion for metal-plastic insert is reasonably good. The FE model could predict the stiffness of the metal-plastic insert (initial slop of the curve) with very good accuracy. Predicted strength (maximum force) is within 6% of tested strength. Predicted intrusion is about 28% off as compared to the test. The discrepancy in prediction of failure mode is primarily due to the unknown adhesion property at the metal-plastic interface. The interface Munjurulimana

78 was modeled using cohesive surface behavior in ABAQUS with stiffness and strength of interface as 20% of stiffness and strength of plastic. A detailed investigation of adhesion between metal-plastic hybrids is beyond the scope of this paper, and hence excluded from subsequent sections. Figure 35. Predicted versus experiment correlation for an all-plastic insert subjected to three-point bending. A very good correlation is observed for all-plastic insert. Figure 37. Predicted vs. experiment correlation for metal-plastic insert subjected to three-point bending. Failure prediction can be improved by improving the adhesion between metals and plastics in PMH. Figure 36. Failure modes and zones for all plastic insert. It is worth noting that the failure happens almost at the same region, and is always in tension. Predicted location of failure is correlating reasonably well with actual failure location observed in the test for metal-plastic insert. Figure 38 shows predicted and tested failure location for the metal-plastic insert in three-point bending scenario. Figure 38. Failure zones for metal-plastic insert showed reasonable correlation between prediction and experiements. CONCLUSIONS This paper focused on the development of BIW reinforcement solutions using multi-material systems including engineering thermoplastic materials and metals. Various design and material configurations including plastic and metal-plastic structural members mounted on the BIW are evaluated through CAE studies for various crash scenarios such as high-speed frontal crashes, side Munjurulimana

79 impact, pole impact and rollover. These CAE studies performed on a generic vehicle shows that up to 15 kg weight can be taken out by replacing four reinforcements from a midsize sedan weighing nearly 1070 kg. Approaches to correlate the CAE studies using component level testing and validation of generic reinforcements are also investigated. Data from all of this work indicate that the use of lighter metal-plastic BIW reinforcements can achieve significant weight reduction (up to 30%) in a vehicle, while also ensuring no compromise in crash performance. Further work can include detailed validation of component level high-speed tests, investigation of the assembly of proposed BIW reinforcements and their performance evaluation for secondary requirements such as NVH, creep, longterm durability and so on. REFERENCES [1] nability/corporate-average-fuel-economy-cafestandards,accessed April, [2] accessed April, [3] April, [4] Buchholz, Kami., Materials lead the way to vehicle mass reduction, SAE article [5] Sah, S., Bawase, M., and Saraf, M., "Lightweight Materials and their Automotive Applications," SAE Technical Paper , 2014, doi: / [6] Makino, K., "Advanced Requirements for Fuel Efficient Cars by Creating Efficient Body," SAE Technical Paper , 2011, doi: / [7] Lei, F., Chen, X., Xie, X., and Zhu, J., "Research on Three Main Lightweight Approaches for Automotive Body Engineering Considering Materials, Structural Performances and Costs," SAE Technical Paper , 2015, doi: / [8] Zhou, G., Li, G., Cheng, A., Wang, G. et al., "The Lightweight of Auto Body Based on Topology Optimization and Sensitivity Analysis," SAE Technical Paper , 2015, doi: / [9] Lewis, A., Keoleian, G., and Kelly, J., "The Potential of Lightweight Materials and Advanced Combustion Engines to Reduce Life Cycle Energy and Greenhouse Gas Emissions," SAE Technical Paper , 2014, doi: / [10] Lightweight, heavy impact McKinsey & Company, Advanced Industries, February [11] Nagwanshi D.K., Parameshwara, A., Bobba, S., Marur, S.R, Marks, M.D, Reinforced body in white and method of making and using the same U.S. Patent 2014/ A1, Aug. 24, [12] National Crash Analysis Center at George Washington University, Finite Element Model Archive, accessed Oct [13] HyperWorks User Manual, Version 12.0, 2012, Altair Incorporation, [14] Frontal full wall NHTSA 5677, hicle/yaris-v1m.pdf, accessed Oct [15] Side impact crash worthiness evaluation, Crash test protocols test protocols (version VII) Insurance Institute for Highway Safety, May [16] AP/Side-Pole_TP_NCAP.pdf, accessed April, [17] Laboratory test procedures for FMVSS 216 Roof crush resistance -TP , November 16, [18] LS-Dyna User Manual, Version 971, 2006, Livermore Software Technology Corporation. [19] Wang. L, Zhang. Q, Wang. J, Yang. B, Yang. M and Feng. J, Effects of annealing on the hierarchical crystalline structures and mechanical properties of injection-molded bars of high-density polyethylene, Research Article, Wiley Online Library, May 20, 2013 ACKNOWLEDGMENTS The authors would like to acknowledge National Crash Analysis Center at George Washington for providing FE model of the car model used in the paper. The car model, which is validated for most of the high-speed crash scenarios, is extensively used in this paper to demonstrate the concepts presented in the paper. Authors also would like thank various internal reviewers for providing valuable and constructive feedbacks about the manuscript for further improvement. SABIC and brands marked with are trademarks of SABIC or its subsidiaries or affiliates. Munjurulimana

80 Any brands, products or services of other companies referenced in this document are the trademarks, service marks and/or trade names of their respective holders. Munjurulimana

81 THERMOPLASTIC CARBON FIBER REINFORCED BODY-IN-WHITE STRUCTURES FOR VEHICLE CRASH APPLICATION Lennart Keuthage BMW AG Germany Dirk Heider John W. Gillespie, Jr. Bazle Z. Gama Haque John J. Tierney Shridhar Yarlagadda Center for Composite Materials, University of Delaware USA Adam Campbell BMW AG Germany Derek Rinehardt BMW North America USA Paper Number ABSTRACT Carbon Fiber Reinforced Plastic (CFRP) composites are becoming one of the possible solutions for vehicles to achieve overall weight reduction in order to meet fuel economy and emission standards while maintaining safety requirements. Carbon fiber thermoplastic composites offer several additional advantages over their thermoset equivalents: higher levels of ductility and specific energy absorption, rapid processing and recyclability. The Department of Transportation s National Highway Traffic Safety Administration (NHTSA) awarded the National Center for Manufacturing Sciences (NCMS) a contract to research potential materials and to evaluate their impact on vehicle crash safety and weight savings. In a joint research Project, the University of Delaware Center for Composite Materials (UD-CCM) and Bayerische Motoren Werke (BMW) investigated available computational tools for the design, optimization and manufacture of carbon fiber thermoplastic body-in-white structures for vehicle crash applications. A vehicle B-pillar was developed to meet the FMVSS No. 214 standard, a US vehicle safety requirement for side impact. In addition, BMW internal structural integrity requirements as well as geometrical requirements were met. The design process demonstrated the capabilities of a computational tool chain, including geometrical design, carbon fiber layout, draping, material property management and dynamic impact simulation. Following this approach, a weight reduction of 60% compared to a metal baseline could be achieved. A thermoplastic B-pillar was manufactured at UD-CCM using infusion as well as thermoforming processes for differing parts of the assembly, which could be scaled to meet industry requirements. In the final drop tower test series, the B-pillar was proven to meet all considered safety requirements. In addition the computational predictive engineering approach could be validated using the test results. Keuthage 1

82 INTRODUCTION Previous studies have shown that composite structures deform in a different manner than similar structural components made of conventional materials like steel and aluminum. The micromechanical failure modes, such as matrix cracking, de-lamination, fiber breakage etc. constitute the main failure modes of composite structures. These complex fracture mechanisms make it difficult to analytically and numerically model the collapse behavior of fiber reinforced composite structures. However, they provide potential for weight saving while maintaining a high level of safety in vehicle crash applications. Crashworthiness of carbon composites Several studies [1] [2] have shown that carbon composites are superior to conventional metal structures with respect to energy absorption per unit weight in a dynamic impact event. Further investigations led to the conclusion that crash performance and energy absorption of composite structures are influenced by a wide range of design parameters as well as material properties and loading conditions [3]. Therefore the design and dimensioning of composite structures for vehicle crash application requires sophisticated computational models as well as a flexible, tailored design process. Structural integrity is a main driver of dimensioning of body-in-white structures for crashworthiness. BMW conducted a series of investigations, showing that composite structures offer at least an equal level of safety compared to conventional materials, with regard to structural integrity subsequent to a crash event. [4] Thermoplastic Carbon Composites offer specific advantages over their thermoset counterparts. Material properties of many available thermoplastic matrix materials offer greater ductility and thus may provide advantages concerning energy absorption in an impact event. In addition to the mechanical properties some thermoplastic composite materials show great potential for recyclability compared to most thermoset composites. REQUIREMENTS energy consumed by the B-pillar as well as the deformation behavior. It was assumed that as long as the composite B-pillar shows an equal or smaller intrusion during the course of the impact, equivalent or greater occupant safety can be achieved by applicable restraint systems. In addition to the intrusion requirement, the need to ensure structural integrity after the crash event permits a valid B-pillar design from separating completely or de-bonding completely from neighboring parts during the crash. DESIGN PROCESS In this joint research UD-CCM und BMW aim to showcase a design process (Figure 1) suitable to support the above mentioned advantages thermoplastic carbon fiber materials offer, while addressing the challenge of computational modelling. This includes evaluation of commercially available software. Figure 1. Design Process for Carbon Composite Components in Vehicle Crash Application (see also appendix 1). Geometrical Design A parametric CAD model of a simple B-pillar was developed utilizing a generic design derived from a BMW vehicle model. This model helped to establish the design space or envelope available for composite design and optimization. A wide variety of shapes and associated composite designs were evaluated. This led to the development of a twopart closed Hat section. This design considered two composite parts with a smooth Spine laminate bonded to a Hat laminate as shown in Figure 2. To obtain suitable performance goals for the composite B-pillar component, full vehicle crash simulations featuring a conventional steel B-pillar have been conducted to measure the amount of Keuthage 2

83 Figure 4, features cutouts in the Hat section. These cutouts significantly reduced strain concentrations in the Spine and also resulted in a composite design that was lighter and more manufacturable. Figure 2. Spine-Hat Adhesively Bonded Composite B-pillar. The Spine was designed to survive the impact event without experiencing major failure, while the high-elongation nylon-based Hat structure absorbed the majority of the impact energy by means of deformation and crushing. To achieve progressive crushing behavior, off-axis dominant laminates were prescribed in the Hat s sidewalls. To reduce overall weight, the laminate thickness drops in the vertical axial direction as less material was needed in the less-loaded Hat upper section. Transition regions were automatically built between these regions using rules and ply transitions defined with the Grid Method in CATIA. The different composite regions and transitions zones of the Hat laminate are shown in Figure 3. Figure 3. Hat Composite Design with Discrete Functionality. Geometrical design of the composite B-pillar evolved during iterative design loops through feedback from finite element (FE) simulation of the crash event. The final TAB Design, shown in Figure 4. Finalized TAB Design. CATIA Composites Engineering Design (CPE) and Composite Design for Manufacturing (CPM) provided process-oriented tools dedicated to the design of composites parts from preliminary to engineering detailed design to direct generation of manufacturing data. CPE offers three methods for composite definition that vary in function and complexity and robustness. For this effort, all composite components were defined using the Grid Based Method to capture and transition the discrete functionality within various regions within the structure. Materials Carbon fiber reinforced thermoplastics were chosen as candidate materials for the design, analysis, prototyping, test and evaluation of a B- pillar. An initial assessment was performed to evaluate material forms and thermoplastic resin combinations with potential for scalable manufacturing processes in the automotive industry. A materials requirements document was created to source carbon fiber thermoplastic materials from suppliers in the composites industry, describing the fiber, resin and material form criteria for this effort. For down-selection of promising candidate systems for detailed assessment, a preliminary materials screening strategy was adopted for all materials sourced in this effort. The strategy centered on the measurement of three key mechanical properties: Keuthage 3

84 0-degree tension for translation of fiber properties 90-degree tension for processed laminate quality and sizing or fiber-matrix adhesion ±45-degree tension for in-plane shear and ductility assessment In addition to these tests, ultrasonic scans for panel quality, fiber volume fraction and density measurements were performed for all material systems. Based on this screening procedure two material systems were down selected for the two regions of the B-pillar (Figure 5). Tencate PA6 / Hexcel AS4 12K was selected for the Hat section due to its superior shear elongation. Arkema Elium / T700 was selected for the Spine part because of its good tensile properties and advantages of good manufacturability and relatively low material cost. Figure 5. Relative Key Properties of Selected Material Systems. Throughout the design of the composite B-pillar, these initial properties measured from quasi static coupon tests were complemented by more in depth analyses of material properties. As the B-pillar design evolved, additional coupon compression tests as well as a sub component test program, featuring quasi static crush tests of head section profiles, were carried out to measure the post damage behavior of the selected materials. For bonding of the two B-pillar sections, Plexus MA530 adhesive was selected. SMARTree software was used for material property management. The software assisted calculating different sets of nonlinear material data from test results as well as distributing them to the design team members. Finite Element Modelling and Simulation During the iterative design process, a large quantity of design variants of the composite B- pillar were evaluated for crash performance using dynamic, explicit FE simulation. To allow for quick evaluation of design changes and thus an efficient design process, the generation of the FE model was automated. Altair Hypermesh was used to create an FE-Mesh based on the geometric shape derived from CATIA. The composite layup information is mapped onto this mesh and a subsequent draping analysis is performed within Hypermesh to obtain an accurate estimate of the ply angle for each ply and element. To covert the generated FE Model into the LS-Dyna format, an automated interface was implemented. The FE simulations were carried out in LS-DYNA due to the variety of material models available for composites in this explicit FE code. In order to model the multi-layer composite laminates, layered shell elements were used. MAT54 was downselected as the material model of choice for the composite B-pillar application offering a linear elastic behavior up to failure and a drop off to a limit value after failure. Although other material models in LS-Dyna offer nonlinear material behavior, the control over nonlinearity in these models is limited. Selection of MAT54 was considered a conservative approach. Throughout all project stages the available material data was used to calibrate the material models, thus leading to increasing prediction capabilities as the B-pillar design evolved. Starting with basic properties from material screening up to post failure damage behavior captured from sub component tests. A non-congruent solid adhesive layer was modeled at all adhesive locations using *MAT_COHESIVE_MIXED_MODE element formulation in conjunction with tie constraints. The adhesive model was fitted to test data obtained from tension and lap-shear tests. MANUFACTURE AND TESTING To prove the validity of the documented design process and to demonstrate the feasibility of the developed B-pillar component five B-pillars were manufactured and tested by UD-CCM. Manufacture Due to the different material systems chosen for Spine and Hat section, production utilized different Keuthage 4

85 processing approaches. Both the liquid molding process for Spine production and the forming process for Hat manufacture have potential to be scalable for mass production. The mass production equivalent for the Spine vacuum infusion being a resin transfer molding process, which is well established for thermoset CFRP. During liquid composite molding (LCM) processing of the Spine a dry fiber textile is impregnated with a liquid low-viscosity resin. The applied pressure gradient between the injection and vent gate allows resin infiltration of the fiber reinforcements. Flat pattern of the preform design was generated and cut from uni-directional carbon fabric (Figure 6). gelation in the resin bucket was observed. This minimized vapor generation and thus dry-spot development. Figure 7. Dry-Spots after Infusion. Over the project period 10 Spines were produced, with 5 being used for impact testing. Figure 6. Preform Kitting. An adhesive layer was added through a heat treatment to minimize any loss of individual tow bundles as the material was cut and handled. The final assembly was heat treated under vacuum at 120 C for an hour and resulted in a good dimensional stable preform which was further processed for final infusion. Initial infusions showed significant void space close to the injection locations. The pressure in the injection gates at the end of infusion will drop to almost atmospheric pressure. It was speculated that this allows generation of vapors in the infusion areas resulting in an increase in vapor pressure pushing the resin out of the preform area locally. The effect led to the observed dry-spot development (Figure 7). In all further experiments, the injection ports were inverted to a vent as Hat Section Manufacturing: Forming of continuous uni-directional carbon fiber thermoplastic parts is still in its infancy with limited thermoplastic uni-directional prepreg material availability coupled with no established simulation tool to predict the forming process. The major processing challenge is the forming of the heated but still viscous blank material over the tool. The three major processing steps include consolidation of an engineered blank and heating of the blank followed by forming of the blank in a die. The NHTSA program established a three-stage thermoforming system at UD-CCM which integrated a 54 kw infrared (IR) heater station, blank preparation station with a shuttle in a 150- ton press system (Figure 8). The system allowed placement of the blank into the shuttle, rapid heating of the blank under the IR heater followed by forming in the press section. The system was used to produce flat components for mechanical tests and small-scale Hat sections for sub-element testing. During full-scale Hat production, the engineered blank was heated in the press using convection and then pressed in shape. Keuthage 5

86 B-pillar Assembly was achieved by adhesively bonding Spine and Hat. Therefore the tool used to manufacture the Hat in the thermoforming process also served as a jig for assembly and bonding of the Hat to the Spine. After both Hat and Spine were fabricated by their respective processes and trimmed to final shape adhesive dispensing was performed with the UD-CCM robot (Figure 10). Figure 8. Schematic of Forming Station. For production of the full-scale Hats, engineered blanks based on the final B-pillar design were assembled manually (Figure 9). Twenty-six prepreg layers were arranged to form the blank with the majority of the off-axis plies being located on the vertical walls of the Hat. Plies were manually cut using a pattern master and pieces were attached to the main body using a point welding process. The final blank was consolidated under vacuum. Figure 9. Assembly of Prepreg Pieces to Form Hat Blank. The IR heating cell, shown in Figure 8, was extended to accommodate a larger blank but uniform heating was difficult to accomplish using the heater bank system. Heat gradient between individual heater units as well as temperature losses along the cell edges resulted in unacceptable temperature gradients of more than 10 C. Thus, the press and molds were insulated and the cavity was heated using the integrated press and external convection heaters. A blank was placed on the mold surface and the system was heated. The stamping process was initiated at 230 C. After forming, the part was actively cooled to roomtemperature. A total of six parts were manufactured successfully using this approach. Figure 10. B-pillar Assembly. Testing As final assessment of the composite B-pillar, a drop tower test was performed at UD-CCM. The B-pillar test setup included a generic steel rocker to provide realistic boundary conditions for the B-pillar compared to the FMVSS No. 214 side impact crash test. Figure 11 shows the clamping fixture for the steel rocker as well as the complete assembly including the composite B-pillar, the generic steel rocker and the fixtures on either side of the rocker and on the top of the B-pillar. The top of the B-pillar was clamped directly in this drop tower setup. The steel rocker was a single use component and partially impacted and deformed during the drop tower tests. Figure 11. Drop Tower Setup. The B-pillar was impacted by a rigid steel impactor which allowed for a small angular rotation around the vehicle x-axis. In previous FE analyses this configuration had been found to represent the FMVSS No. 214 loading conditions closely considering the limitations of a drop tower facility. Keuthage 6

87 The tests were conducted at an impacting mass of kg and an impact velocity of 7.26 m/s which yielded an impact energy of kj. During the tests impactor displacement as well as forces at the impactor were measured. Additionally the B-pillar deflection and strain field was measured on the far side of the B-pillar with a digital image correlation system. Additional force measurements were conducted at the B-pillar and rocker fixtures. Figure 12. Time History of Impactor Force. Experimental Results were used to judge the B- pillar performance as well as to validate the virtual prediction and the design process. Figure 12 shows time history of impact contact forces for all five experiments. Investigating the high-speed photography of the first test (TOP-LVI on B-pillar B1) revealed that the roof clamp fixture, although firmly bolted to the floor, slid inward towards the rocker during impact. Additional measures were taken such that sliding of the roof clamp fixture was prevented. The inadequate boundary conditions in the first test resulted in additional compliance, longer duration impulse loading and a reduction in the peak load. The validity and functionality of all other instrumentation was proven in this initial test. Subsequent tests were compared to the FE simulation model prediction as shown in Figure 13. While force over time as well as deformation shape show very similar peak results and a good over all correlation between test results and model prediction, the impactor displacement is significantly higher in the test data. Since the B- pillar far side deflection obtained from the test is equal to the prediction, the difference in impactor displacement indicates a higher amount of B-pillar crushing taking place in the drop tower tests. Figure 13. Comparison of FE Model (red / dotted) and Test Results (blue / solid). To judge the performance of the B-pillar with respect to structural integrity after the impact, the specimens were thoroughly inspected visually. As predicted, the Hat section showed significant amounts of crushing and fiber damage as shown in Figure 14. Keuthage 7

88 Figure 14. Crushing of Hat Section. Figure 15 shows the adhesive bond between Hat and Spine which failed locally in the areas predicted by the simulation model. Figure 15. Debonding of Hat and Spine. The Spine did not show any fiber damage, though delamination in the Spine was visible at the rocker bonding location (Figure 16). However, a significant fraction of the Spine laminate was still adhesively connected to the rocker, which led to the B-pillar being able to arrest impactor movement after rebound and supporting the static load. Figure 16. Spine Delamination. CONCLUSIONS UD-CMM and BMW investigated thermoplastic carbon fiber reinforced materials for vehicle side frame structures. The proposed B pillar was designed to meet structural and crash safety requirements (e.g., FMVSS No. 214 barrier) using thermoplastic composites which offers significant advantages (e.g., recycling, joining) compared to thermoset with the potential for improved crash performance. Novel side-impact crash concepts maximizing crash performance have been developed and commercial available thermoplastic materials were characterized to define appropriate material models and to evaluate energy absorption mechanisms. Predictive engineering at all levels, from coupon to sub-element to full-scale, guided the material down-selection. The same CAE tools simulate full vehicle to component & test setup behavior and were used to optimize manufacturability and structural / crash performance. Sub-components and B-pillars were fabricated using stamp forming and infusion processes, allowing scalability with the potential to meet automotive production rates in the future. The UD-CCM high energy drop tower was used to validate the predictive engineering tools and crash performance of the proposed B-pillars under realistic side-impact crash conditions. The B-pillar design was spatially optimized for energy absorption (ductility), stiffness, and strength while maintaining part producibility and vehicle integration. BMW established B-pillar performance metrics derived from full-vehicle crash simulations and other design and integration requirements. UD-CCM provided a full range of capabilities in materials selection and evaluation, composite design, analysis and crash simulations, process development and manufacturing (tooling, part production, trimming), full-scale pillar Keuthage 8

89 assembly and high-energy impact testing. This project has demonstrated design, materials, manufacturing and joining methods with continuous carbon fiber thermoplastics, at technology readiness level (TRL) 4-7 to meet automotive industry and government safety specifications. Key achievements from this project are summarized as follows: Successful fabrication and manufacture of an all thermoplastic composite B pillar that is 60% lighter than the existing metallic design while meeting project requirements for NHTSA FMVSS No. 214 side-impact crash. State-of-the-art CAE tools were evaluated (with internally developed data translation) simulating full vehicle to component impact (Dassault Systemes CATIA, Altair HyperWorks & LSTC LS-DYNA). Innovative production methods were developed and demonstrated for this multimaterial part that included infusion and thermoforming tailored blanks with the potential to meet 2 minute cycle times. Adhesive bonding methods were developed and automated for dissimilar thermoplastics and steel interfaces. Automated trimming of the thermoplastic components was developed and demonstrated without damage to the composite structure. A test fixture was designed and integrated into UD-CCM high-energy impact tower simulating the crash behavior during sideimpact crash without using a full vehicle structure. Multiple full-scale B-pillar assemblies (incorporating steel roof and frame rail) were successfully impact tested under 100% equivalent energy of FMVSS No The composite B-pillar response in the vehicle sub-component configuration satisfies all of the intrusion safety requirements to meet the requirements of FMVSS No All composite B-pillars exhibited rebound and post-impact structural integrity in terms of fully supporting the impactor dead weight of kg. The impact test was simulated and compared to the experimental data (deflection, load, and others), validating the predictive engineering approach. The goals of the project, validating the predictive engineering tools and demonstrating equal or better occupant safety performance at reduced weight as equivalent steel vehicle components, have been successfully accomplished. REFERENCES [1] Saito H., Chirwa E.C, Inai R., Hamada H. 2002: Energy absorption of braiding pultrusion process composite rods. Compos Struct 55/4, [2] Jacob G.C., Fellers J.F., Simunovic S., Starbuck J.M. 2002: Energy Absorption in Polymer Composites for Automotive Crashworthiness. J Compos Mater 36/7, [3] Engel S., Boegle C., Lukaszewicz D. 2013: Crushing of Composite Structures and Parameter Identification for Model Development. In Proceedings of the 19 th International Conference on Composite Materials (Montréal, Canada, Jul. 28 Aug. 2) [4] Ferenczi I., Kerscher S., Möller F. 2013: Energy Dissipation and Structural Integrity in Frontal Impact. In Proceedings of the 23 rd Enhanced Safety of Vehicles Conference (Seoul, Republic of Korea, May 27-30) Keuthage 9

90 APPENDIX Appendix 1. Larger depiction of Figure 1: Design Process for carbon composite component in vehicle crash application. Keuthage 10

91 PASSIVE SAFETY STRATEGY FOR ELECTRIC LIGHTWEIGHT VEHICLES WITH MULTI- MATERIAL BODY AND CENTERED DRIVER POSITION OPPORTUNITIES AND LIMITATIONS Raúl, Molinero Fresnillo Emiliano, Core Almarza ZF TRW Spain Lothar, Zink ZF TRW Germany Mervyn J., Edwards TRL ltd United Kingdom Johannes, Holtz Technische Universität Berlin (TUB) Germany Mikkel, Steen Pedersen Insero E-mobility (IeM) Denmark Rubén, Torres Heras Grupo Antolín Spain Paper Number ABSTRACT The growing market share of electric lightweight vehicles requires new passive safety strategies as these vehicles have different behavior in accidents compared to conventional vehicles. Due to their low weight they could experience high deceleration pulses and intrusion levels. The main objective of this study was to develop a passive safety strategy for a light weight electric vehicle to give best in class occupant protection. Challenges involved in this study include, mainly focused on side impact: The use of alternative materials for the body structure (mainly sandwich panels and foam) due to their failure mechanism A novel seating layout with a centrally positioned driver, especially challenging for side impact Within this study two baseline and four final prototypes were built. The development of the vehicle was accompanied by FE-simulation. Two of the baseline prototypes were subjected to Euro NCAP MDB side and ODB frontal impact crash tests. These baseline crash tests served as benchmark for the development of the passive safety strategy and the validation of the FE-model. For side impact two critical issues have been taken into account, namely a high v and a centered driver position which reproduces the current challenge for far side protection. Firstly, FE simulations have been done to develop the restraint systems, followed by sled test development loops in the main Euro NCAP load cases (MDB, Pole and ODB). With the final prototypes a full Euro NCAP crash assessment was performed using the year 2013 rating protocols to allow comparison with the baseline crash tests. The MDB side impact led to very high pulses, that couldn t be addressed structurally due to the high mass ratio between barrier and vehicle. However, the results show that with the proposed restraint system using airbags and a four point seatbelt an adequate protection level could be reached for the centered driver for both MDB and pole side impacts, compared to standard vehicles. For frontal impact, the results showed that, using an approach of a strong compartment built from novel composite reinforced glass fiber / foam panels, Molinero 1

92 combined with a specially designed energy absorption module, an innovative four-point seatbelt and a conventional driver airbag, resulting intrusion could be minimized and adequate protection could be offered to the driver. Overall occupant protection equivalent to a best in class Euro NCAP year 2013 rating was achieved for the vehicle. The strategy developed demonstrated equivalent protection levels based on the Euro NCAP year 2013 suite of tests. Since then, the Euro NCAP assessment has been further improved in terms of representativeness of real-world accidents. These improvements include a heavier barrier for the MDB test and the addition of a full width frontal test. With respect to growing market share of electric lightweight vehicles a passive safety strategy was developed for such vehicles based on Euro NCAP crash tests to give best in class occupant protection. Because of the centrally positioned driver, some challenges have been faced and solved, especially for side impact configurations. Molinero 2

93 INTRODUCTION Current pollution issues in big cities in combination with mobility matters of conventional vehicles with petrol engines has resulted in an increment of electric lightweight cars on the roads, which have to coexist with the traditional kind of vehicles. However, quadricycle versions of these vehicles do not need to fulfill the same legislation as conventional cars to be sold for road use, so basic, low level occupant restraint systems are in general developed for them. Euro NCAP expected that quadricycles would show a very poor performance when they were tested using regular procedures for conventional cars, so Euro NCAP developed special protocols for testing heavy quadricycles through two crash tests, less stringent than protocols for passenger cars: - A full-width frontal impact at 50km/h against a deformable element - And a side impact test, also at 50km/h, in which a deformable barrier is driven into the side of the vehicle. The assessment of 8 low weight vehicles according to these specific procedures [1], 4 of them assessed in 2014 and 4 additional ones assessed in 2016, as it can be seen in Figure 1 and in Figure 2, has shown the low protection capability of these vehicles, which are not able to exceed a 2 star rating. Club Car Villager Ligier Ixio JS Line4 Renault Twizy 80 TAZZARI Zero 2014 FWDB MDB Figure 1. Quadricycles assessed in 2014 according to Euro NCAP quadricycle protocols. Aixam Crossover GTR Bajaj Qute Chatenet CH30 Microcar M.GO Family Figure 2. Quadricycles assessed in 2016 according to Euro NCAP quadricycle protocols. This paper describes the development and the results of a heavy quadricycle, included inside a Project of the Seventh Framework Programme developed by the European Commission called BEHICLE: BEst in class vehicle: Safe urban mobility in a sustainable transport value-chain ; which has the target of fulfilling a rating of at least 4 stars according to Euro NCAP protocols for conventional cars of year In addition to the activities performed inside the BEHICLE project, specific Finite Element crash simulations have been performed to evaluate BEHICLE against Euro NCAP 2016 protocols and Euro NCAP protocols for heavy quadricycles. Although this paper is focused in side impact for a driver occupant in a Far Side configuration, results in frontal impact load cases have been also included in order to have a global view of the behavior of BEHICLE in Euro NCAP tests. METHODS 2016 FWDB MDB In the BEHICLE program, the used method has been to perform as a first step preliminary crash tests with the non-optimized original BEHICLE vehicle according to Euro NCAP 2013 regular car protocols. These results have been used for the FE Model correlation of the vehicle structure and as reference for the development of the passive safety strategy. Afterwards, restraint systems (belt and airbags) have been included in the FE Model with a dummy and intensive simulation runs have been performed. The best configuration has been evaluated via sled tests with real vehicle environment according to Euro NCAP 2013 regular car protocols (ODB, MDB and Pole). Molinero 3

94 Finally, three complete prototypes of BEHICLE have been crash tested according to the three load cases evaluated by Euro NCAP 2013 (ODB, MDB and Pole). Nevertheless, assessment of BEHICLE would be incomplete if updated Euro NCAP protocols were not taken into account, so FE-Models have been also performed according to the Euro NCAP 2016 assessment (FW, ODB, AE-MDB and Pole), as well as Euro NCAP protocols for heavy quadricycles. BEHICLE framework BEHICLE has been designed as a 100% electric vehicle without petrol engine. Instead, it is powered by electric engines placed in the Wheel axis fed by a set of electric batteries placed in a false floor under the cabin ground [2]. As one requirement of the program is the low weight that BEHICLE needs to achieve, the materials in which it is mainly made is a light weight composite panel with a Core of hard foam and an external cover of glass fiber. These 3D composite panels, which mainly composes the lower platform, the firewall, the roof, the three seats and the doors, have all the advantages of standard sandwich panels, but posses enhanced properties since it is a 3D reinforced composite panel such as high strength-to-weight ratio, high buckling and impact resistance, absence of delaminating and high blast energy absorption capability. Another light weight material is structural aluminum. Reinforcing elements between composite panels, like greenhouse structure or the door beams, are made of this material, which also offers an appropriate corrosion resistance, a good weldability and a proper cold formability. The third main material in which BEHICLE is built up is black colored EPP foams (Expanded PolyPropilene) with a mass density of 58 g/l to 66 g/l, located in the front end cover, side sill covers, seats and interior parts. Finally, Polycarbonate (PC) panels are used as glazing due to its thermal insulation properties and high resistance to impact. The outer shell is made of a combination of EPP parts and plastic panels, prototyped in BEHICLE by EPP parts. Other important matters concerning vehicle stiffness in comparison with conventional vehicles is the lack of a B-Pillar, which makes BEHICLE more sensitive to lateral crash impacts; and the lack of petrol engine placed in the front of the car, which could cause different frontal crash pulses and intrusions. A general view of BEHICLE car can be observed in Figure 3. Figure 3. Main view of BEHICLE. Geometrically, as it is shown in Figure 4, BEHICLE has places for three adult occupants, in which the driver is centrally seated, while the two passengers are positioned on the rear seats behind the driver side by side making a triangular layout. Because of this configuration and the reduced space in the compartment, legs for rear occupants are placed on the right and the left of the driver. Figure 4. Rear passenger placed with legs on the side of frontal driver in BEHICLE. Assessment and load cases The driver safety performance in BEHICLE has been evaluated in accordance with the Euro NCAP procedures for conventional cars in 2013 and 2016 and the special protocols developed for testing heavy quadricycles [3 and 4]. Euro NCAP 2013 evaluates driver injuries according to three load cases (see Figure 5): - A frontal impact test, in which the vehicle drives at 64km/h towards a Deformable Barrier, with 40% offset (ODB) and a HIII 50th dummy in the driver position. - A side impact test, in which a Mobile Deformable Barrier of 950kg (MDB) is driven into the side of the vehicle at 50km/h, with an ES-2 dummy in the driver position. - A side impact test, in which the vehicle is moved at 29km/h towards a rigid pole, with an ES-2 dummy in the driver position. Molinero 4

95 MDB ODB ODB Pole Full Width Heavy Quadricycles Euro NCAP evaluates driver injuries according to two load cases (see Figure 7): - A full-width frontal impact at 50km/h against a deformable barrier (FWDB), with a HIII 50th dummy in the driver position - And a side impact test, also at 50km/h, in which a deformable barrier of 950kg (MDB) is driven into the side of the vehicle, with an ES-2 dummy in the driver position. FWDB MDB Figure 5. Euro NCAP 2013 load cases. Euro NCAP 2016 evaluates driver injuries according to four load cases (see Figure 6): - A frontal impact test, in which the vehicle drives at 64km/h towards a Deformable Barrier with 40% offset (ODB), and a HIII 50th dummy in the driver position. - A frontal impact test, in which the vehicle drives at 50km/h towards a Full Width Wall, with a HIII 05th dummy in the driver position. - A side impact test, in which an Advanced European Mobile Deformable Barrier of 1300kg (AE-MDB) is driven into the side of the vehicle at 50km/h, with a World SID 50th dummy in the driver position. - A side impact test, in which the vehicle is moved at 32km/h towards a rigid pole with an angle of 75 degrees with respect to the lateral side of the vehicle, with a World SID 50 th dummy in the driver position. ODB AE-MDB Full Width Pole Figure 6. Euro NCAP 2016 load cases. Figure 7. Euro NCAP tests for Heavy Quadricycle. Devices and tools FE-Models. The finite element simulation model used for this vehicle consisted of 1 to 1.5 million of degrees of freedom, depending on the load case [2]. The simulation was carried out with the finite element software LS-Dyna (Livermore Software Technology Corporation (LSTC), Livermore, CA). The way to proceed has been firstly to transfer the linear and angular accelerations (crash pulse) of a point from the Full Crash FE-Models of BEHICLE into a BEHICLE substructure FEM to reproduce the global motion of the vehicle. The best area to get the pulse is one with high stiffness and low deformation and placed close to the occupants. In conventional vehicles, this point usually comes from the centre tunnel or the base of the opposite B-Pillar, but as BEHICLE has none of them, and because of the fact that the study will be focused on the driver and the floor of BEHICLE does not suffer high deformations, the optimal point to get the pulse is under the front seat close to the floor. In addition, and according to the preliminary information, door intrusions will be important, so in addition to the pulse, it will be necessary to include in the model the intrusions of elements of the door as the inner door panel and the aluminum door beam, depending on the type of lateral load case: - In MDB full crash simulation models, motion of the aluminum door beam and the motion of the door panel under the door beam have been included in the substructure models. - In Pole full crash CAE simulation models, motion of the aluminum door beam, the motion of the door panel under the door beam and the motion of the roof have been included in the substructure models. Molinero 5

96 Sled Tests. As an intermediate step to perform crash tests, performance of passive restraint systems has been evaluated via sled test. In these sled tests, real parts of BEHICLE which have a significant interaction with the dummy have been included in the set up to reproduce the internal environment of the car. In this way, and focused on lateral MDB and Pole load cases, frontal seat, Green House (roof), and left door with the aluminum beam, the EPP cover and the composite door panel have been implemented in the sled tests. In addition, the acceleration (pulse) of a point under the frontal seat has been reproduced in Y direction; and the motion of two points of the door has been replicated. In the case of sled tests with MDB configuration, it is possible to reproduce separately the deformation of the upper and lower part of the door due to the impact of the MDB barrier. Deformation of a selected point of the door beam has been used to reproduce the motion of the upper door, while deformation of a selected point of the composite door panel has been selected for the motion of the lower door. In the case of sled tests with Pole configuration, it is possible to reproduce the pole intrusion into the cabin, directly via a real pole through the window, and the deformation of the door due to the intrusion of the pole. In this case, only one point of aluminum door beam will be used to generate the sled door pulse. In addition, sled test can reproduce the V-shape deformation of the door caused by the pole penetration. Crash Tests. Final step of BEHICLE program has been to perform Crash Tests according to Euro NCAP 2013 test protocols. RESULTS In the BEHICLE project, passive restraint systems have been developed focused on Euro NCAP 2013 test protocols (Frontal ODB, Lateral MDB and lateral Pole load cases), and the assessment has been done via FE-simulation models, sled tests and crash tests. However, although it is expected to perform three crashes according to each configuration with three final prototypes, only the results of the lateral crashes (MDB and Pole) have been included in this paper, as the planned ODB frontal crash test was not performed at the time this paper was written. Additionally to the Euro NCAP 2013 assessment, further FE simulations have been performed to assess BEHICLE against the Euro NCAP 2016 protocols and the Euro NCAP special protocols for testing heavy quadricycles. Preliminary results Prior to the integration of any restraint system, a lateral crash test according to MDB Euro NCAP 2013 procedures was performed with the original, non-improved BEHICLE vehicle as a baseline to get knowledge of the structural behavior and to define the strategy for the definition of the restraint systems. Main conclusion of this first crash test was the good integrity and stability of the BEHICLE compartment. The only significant intrusions have been observed in the door beam. However, due to the low weight of BEHICLE, as shown in Figure 8, a high velocity of the car after the impact (delta-v) was observed (35 km/h). This high delta-v of the BEHICLE indicates a higher crash severity than typical for Euro NCAP side impact (22 km/h to 28 km/h). Figure 8. Velocity comparison with a Supermini vehicle. Concerning dummy kinematics, the dummy s pelvis impacted the door glazing (which was pushed into the vehicle by the barrier). Also dummy head ejection outside the cabin was observed. BEHICLE improvement Previously to the inclusion of the passive restraint systems into the crash simulation models, the correlation of the BEHICLE simulation model to the preliminary MDB crash tests has been performed. As it was observed in the preliminary MDB crash test, important facts to be improved were the high intrusion of different door elements into the cabin and the control of the door bending. In this way, the door area was redesigned in order to get an acceptable behavior of the door. The final solution to improve these matters, schematized in Figure 9, was based on three main modifications of the side structure: - Firstly, the material of the door panel has been replaced from glazing and / or EPP foam to a composite door panel. Molinero 6

97 - Secondly, a redesign of the sill area, in which also a composite panel has been included to increase the coupling of the door to the sill. - Thirdly, the door beam has been slightly curved outwards to abosrob the energy of the impact in a better way. Glazing windscreen Inner EPP door beam cover Driver four-point seat belt (Top Belt). Due to the particular characteristic of BEHICLE (without B- Pillars and with the driver in a central position) a new and innovative four point seat belt (Top Belt), with two retractors provided with load limiters and pyrotechnical pretensioners has been integrated in the car. Retractors are located at the roof in the rear part of the car. Two buckles are attached to both sides of the seat with the purpose of getting fastened to the latch plates. The two latch plates are stored at the roof in front of the driver in the unbuckled rest position. Figure 11 shows a dummy buckled with the four point seat belt system. Aluminum door beam Composite door panel Outer door cover EPP Sill Composite sill and ground panel Figure 9. BEHICLE section in door area. Simulation runs with this door modification have shown an improvement in the intrusion, avoiding that the door panel intrudes into the cabin in MDB and Pole configuration, and reduced the maximal door beam intrusion to a value of 140mm in MDB configuration and 115mm in pole configuration. Figure 11. Top belt Four Point Seat Belt. Side airbag integrated in door (SAB). Due to the very thin BEHICLE seat backrest, it was not possible to integrate the side airbag in the front seat like in conventional cars; therefore it was integrated in the door and fixed to the aluminum door beam. The specific BEHICLE seat layout with the driver seat in the centre offers a new scenario for innovative side airbags because of the increased space between the side structure and the occupant. The airbag shape has been designed to cover all the different seat positions corresponding to different occupant sizes. Figure 10. Maximal Aluminum door beam intrusion. Restraint Systems Innovative restraint systems were developed and integrated in BEHICLE [5]. These restraint systems were adapted to the specific BEHICLE architecture and crash behavior. Concerning side impact, the following restraint systems have been selected to be integrated in BEHICLE and adapted to the particular BEHICLE environment. Figure 12. Side Airbag Module. Figure 13. Side Airbag Module placed in door. Molinero 7

98 First Row Curtain airbag (CAB). (Figure 14). Main contribution of this curtain airbag is to avoid the front occupant s head contact against an external obstacle in pole test collisions, and to reduce the occupant s head excursion in barrier test collisions. The shape of the bag has been designed to cover the different occupant sizes. close to the maximal Euro NCAP rating is provided (15.21 out of 16 points). Rating was only penalized in the back plate area due to the specific design of the frontal seat. Head HIC Head 36 36Res. Acc. Acc. 36 3ms Chest 3ms Acc. Comp Chest Comp 3msVC Comp T12 Fy T12 Mx Backlate Fy Total Abdomen Abdomen Pelvis Abdomen 36 36Acc. 3ms Acc. 3ms Comp Comp Abdomen Abdomen No No Passive Passive Restraint No Restraint Passive Systems Restraint Systems Systems Top Belt Top Belt Top Belt Top + Belt SAB + Top + SAB CABelt + CAB + SAB + CAB No Passive No Passive Restraint Restraint Systems Systems Top Belt Top BeltTop Belt Top + SAB Belt + CAB SAB + CAB Figure 14. First row curtain airbag module. Euro NCAP 2013 assessment Three load cases according to the Euro NCAP 2013 procedures (ODB, MDB and Pole) have been evaluated in the BEHICLE program, including results of FE-models, sled tests and crash tests. FE-models. FE-models according to Euro NCAP MDB load case have shown that in the case of no occupant restraint systems the ES-2 dummy s head contacts the composite Roof panel and the lower rib and abdomen area contact the BEHICLE door beam, see Figure A-1 of Annex A. In a second step, it has been included a Top Belt with retractor pretensioners activated at a Time To Fire (TTF) of 8ms. In this case, the abdomen did not contact the door beam, showing the effectiveness of the Top Belt in terms of Pelvis and Abdomen restraint. However, the head still contacted the roof and the lower rib contacted the door beam. Finally, in addition to a Top Belt, it has been included in the simulation models a side airbag placed in the door beam and a CAB placed in the roof. The TTF of both airbags, SAB and CAB, have been determined in order to be in position and filled with gas at the right time before the dummy contacts them. The optimal Time to Fire of the SAB was 10ms and of the CAB 40ms. As shown in Figure A-1 of Annex A, the simulation showed that the head contact has been avoided as well as any dummy impact against the door beam thanks to the passive restraint systems. In addition, the CAB was able to avoid the head excursion outside the BEHICLE cabin and to provide a low neck bending. As a summary, described in Figure 15, whereas the case without any passive restraint system provides a severe head contact to the roof metal plate and a high abdomen contact to the door, and the case with Top Belt only results in a contact between head and roof, in the case with all passive restraint systems these contacts are avoided and a good level of protection Head assessment Chest assessment Abdomen assessment Pelvis assessment Total Rating Figure 15. Dummy assessment in MDB configuration models without Passive Restraint Systems, with only Top Belt and with all Passive Restraint Systems. FE-models according to Euro NCAP pole load case have shown similar results as FE-models with MDB configuration. It can be seen in Figure A-2 of Annex A that a contact of the dummy head, chest and abdomen to the BEHICLE interior parts is noticed in the baseline models. The Top Belt is able to reduce the dummy values in the abdomen area, and the inclusion of SAB and CAB reduces additionally the dummy values in head and chest area. Therefore also in the pole load case, a good protection level is achieved close to the maximal Euro NCAP rating (15.03 out of 16 points) with the passive restraint system, only penalized by the influence of the frontal seat in the Back plate area of the dummy (see Figure 16). Head HIC Head 36 36Res. Acc. Acc. 3ms Chest 3ms Comp Chest CompVC T12 Fy T12 Mx Backlate Fy Total Abdomen Abdomen Pelvis 36 36Acc. 3ms Acc. 3ms Comp Comp Abdomen Abdomen No No Passive Passive Restraint No Restraint Passive Systems Restraint Systems Systems Top Belt Top Belt Top Belt Top + Belt SAB + Top + SAB CABelt + CAB + SAB + CAB No Passive No Passive Restraint Restraint Systems Systems Top Belt Top BeltTop Belt Top + SAB Belt + CAB SAB + CAB Head assessment Chest assessment Abdomen assessment Pelvis assessment Total Rating Figure 16. Dummy assessment in Pole configuration models without Passive Restraint Systems, with only Top Belt and with all Passive Restraint Systems. In case of Far Side impact, the integration of a 4 point seat belt system with pretension function helps significantly to control the pelvis and abdomen motion, improving the occupant kinematics. In addition, a thick SAB in combination with a CAB offer a good protection of the chest and head area due to its early contact with the occupant and large thickness. As a conclusion, restraint systems integrated in Molinero 8

99 BEHICLE allows to provide a good occupant protection of the centered driver in the case of far side impacts. Sled tests have confirmed similar results than in previous FE-models. Apart from getting an acceptable reproduction of the BEHICLE behavior in both MDB load case, illustrated in Figure B-1 of Annex B, and Pole load case, illustrated in Figure B- 2 of Annex B, the dummy injury values have reached similar results. All passive restraint systems have worked in a proper way: dummy is well coupled to the frontal seat thanks to the Top Belt System, the side airbag avoids any contact of the dummy torso to the door and the curtain airbag protects the dummy head. In both MDB and pole cases back plate forces have been reduced in comparison with FE-models thanks to the smoothing of the frontal seat section, minimizing the interaction of the ES-2 dummy back plate with the lateral part of the seat and achieving the maximal rating of 16 points in the MDB and Pole load cases (see Figures 17 and 18). Sled Tests results Head assessment assessment Chest assessment Abdomen assessment Pelvis assessment Pelvis assessment Total Rating = Total Rating = Figure 17. Dummy Assessment in MDB sled tests. Sled Tests results Head assessment assessment Chest assessment Abdomen assessment Pelvis assessment Pelvis assessment Total Rating = Total Rating = Figure 18. Dummy Assessment in Pole sled tests. Crash Tests. Similar to the results of FE-Models and sled tests, final crash tests according to lateral MDB and pole Euro NCAP 2013 protocols achieved good dummy injury results, reaching a total of points out of 16 in MDB configuration, and points out of 16 in Pole load case, only penalized in both cases by the back plate of the dummy. This fact points out the necessity of smoothing the profile of the backrest of the frontal seat in order to improve back plate values. Results according to MDB load case are shown in Figures 19 and C-1 (Annex C); and results according to pole load case are shown in Figures 20 and C-2 (Annex C). Head assessment Chest Chest assessment assessment Abdomen assessment 4.00 Abdomen assessment 4.00 Pelvis assessment 4.00 Pelvis assessment 4.00 Total Rating = Total Rating = Figure 19. Dummy Assessment in MDB crash test. Head assessment Chest Chest assessment assessment Abdomen assessment 4.00 Abdomen assessment 4.00 Pelvis assessment 4.00 Pelvis assessment 4.00 Total Rating = Total Rating = Figure 20. Dummy Assessment in Pole crash test. Frontal ODB Assessment. In addition to lateral MDB and pole load cases, it has been assessed the BEHICLE performance in frontal impact according to Euro NCAP 2013 protocols in order to get the complete occupant protection assessment. Results in sled tests provided a rating of points out of 16 points, which are shown in Figure 21. In a further step in the project a frontal ODB crash test will be done to confirm those results. Sled Tests results (ongoing) Head Head and and Neck Neck assessment assessment Chest assessment Knee, Femur and and Pelvis Pelvis assessment Lower Leg, Foot Foot and and Ankle Ankle Assessment Assessment Total Rating = Total Rating = Figure 21. Dummy Assessment in ODB sled test. Sled test were performed with a rigid steering column without collapsibility and energy absorption function; therefore, results in chest area could be significantly improved by implementing a collapsible steering column, absorbing occupant energy and increasing the distance to chest. Euro NCAP 2016 assessment Crash tests in BEHICLE program were performed according to the 2013 Euro NCAP protocols. Additionally, in order to complete and update the investigation, the BEHICLE performance was also evaluated according to Euro NCAP 2016 protocols via FE-Models. Figure 22 presents the Assessment according to the four Euro NCAP 2016 load cases. Driver assessment reached the full score of 16 points in lateral load cases (AEMDB and pole), whereas it reached points out of 16 on Frontal ODB load case and points out of 16 points in Frontal FW load case. Molinero 9

100 Euro NCAP rating Frontal ODB FWDB Frontal FWDB MDB Head and Neck assessment 4.00 Chest assessment 1.78 Knee, Femur and Pelvis assessment 4.00 Lower Leg, Foot and Ankle Assessment 3.78 Total Rating = Frontal FW Head assessment 4.00 Neck assessment 4.00 Chest assessment 0.26 Knee, Femur and Pelvis assessment 4.00 Total Rating = Lateral AEMDB Head assessment Chest assessment Abdomen assessment Pelvis assessment Total Rating = Lateral Pole Head assessment Chest assessment Abdomen assessment Pelvis assessment Total Rating = Figure 22. BEHICLE s Driver Assessment according to Euro NCAP 2016 protocols. Euro NCAP quadricycle assessment BEHICLE was originally conceived as a lightweight, subcompact urban electric car, aiming at balanced energetic performance whilst ensuring top-notch safety performance. According to the car classification standards it would fall within the supermini category. But with lower engine power it would fall into the L7e category; therefore, the BEHICLE was also additionally assessed according to the Heavy Quadricycle rating protocol by FE simulation: Figure 23 illustrates the dummy assessment, reaching 12 out of 16 points in FWDB load case and 14 points out of 16 in MDB load case. Head assessment 2.00 Neck assessment 4.00 Chest assessment 2.00 Knee, Femur and Pelvis assessment 4.00 Total Rating = points 14 points MDB 14 points Lateral MDB Head assessment Chest assessment Abdomen assessment Pelvis assessment Total Rating = Figure 23. BEHICLE s Driver Assessment according to Euro NCAP protocols for quadricycles. Benchmarking BEHICLE has reached, according to Euro NCAP 2013 protocols, a rating of points out of 16 points in ODB load case, 7.68 points out of 8 points in MDB load case and 7.83 points out of 8 points in pole load case, with a total rating of points out of 32 points. BEHICLE is 0.91 points over the average of M1 Supermini vehicles evaluated in 2013 according to Euro NCAP 2013 rating protocols (average rating points) [1]. This is shown in Figure Figure 24. ODB + MDB + Pole Euro NCAP rating of M1 supermini vehicles and BEHICLE in ODB MDB Pole Comparing ODB, MDB and Pole rating assessment of BEHICLE s driver occupant with the average of M1 Supermini vehicles evaluated in 2014 according to Euro NCAP 2013 protocols [1], which reached an average rating in the three load cases of points out of 32 points, also illustrates that Molinero 10

101 Euro NCAP rating Euro NCAP rating Euro NCAP rating Euro NCAP rating Euro NCAP rating BEHICLE is close to two points above the average (see Figure 25) ODB MDB Pole FW max. possible rating (16 points) Figure 25. ODB + MDB + Pole Euro NCAP rating of M1 supermini vehicles and BEHICLE in Regarding Euro NCAP 2016 [1], BEHICLE s driver has reached points out of 16 points according to ODB load case, points out of 16 points according to FW load case and 16 points out of 16 points according to AEMDB and Pole load cases. Making a comparison with all M1 - Supermini vehicles assessed according to this Euro NCAP protocols, it can be observed that the driver BEHICLE rating in ODB and FW load cases is placed inside the range of the cars evaluated by Euro NCAP (see Figures 26 and 27) ODB max. possible rating (16 points) Figure 26. ODB Euro NCAP 2016 rating of all evaluated vehicles and BEHICLE. Figure 27. FW Euro NCAP 2016 rating of all evaluated vehicles and BEHICLE. Concerning lateral load cases, BEHICLE has achieved the highest possible score in AE-MDB and Pole load cases (see Figures 28 and 29) Figure 28. AE-MDB Euro NCAP 2016 rating of all evaluated vehicles and BEHICLE AE-MDB Pole max. possible rating (16 points) max. possible rating (16 points) Figure 29. Pole Euro NCAP 2016 rating of all evaluated vehicles and BEHICLE. Molinero 11

102 % of rating % (over 16 points) % (over 16 points) Finally, according to the special protocols for testing heavy quadricycles [1], BEHICLE is able to achieve a total rating of 12 points out of 16 points in FWDB assessment, twice as high as the best quadricycle tested by Euro NCAP (see Figure 30: vehicles assessed in 2014 in blue, vehicles assessed in 2016 in green and BEHICLE in orange). In MDB assessment, BEHICLE has achieved 14 points out of 16 points, 4 points more than the best tested quadricycle, as indicated in Figure 31. With these results, as shown in Figure 32, BEHICLE is able to reach a total rating of 5 stars, far higher than the best quadricycle evaluated by Euro NCAP, which only reached 2 stars % 80.0% 60.0% 40.0% 20.0% 0.0% Figure 30. FWDB quadricycle assessment % 80.0% 60.0% 40.0% 20.0% 0.0% FWDB Figure 31. MDB quadricycle assessment % 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% MDB Total Rating Figure 32. Total quadricycle assessment. DISCUSSION AND LIMITATION The presented development has demonstrated an equivalent protection level of BEHICLE in comparison with conventional Supermini cars based on the Euro NCAP year 2013 suite of tests, via FE models, sled tests and crash tests. Since then, the Euro NCAP assessment has been further improved in terms of representativeness of real-world accidents. These improvements include a heavier barrier for the MDB test and the addition of a full width frontal test. BEHICLE was also evaluated according to Euro NCAP 2016 assessment, but only with FE simulations. In a similar way, BEHICLE evaluation according to protocols for Heavy Quadricycle was performed by FE Models, not being validated in crash tests during this investigation. Even if the BEHICLE was not developed neither against the Euro NCAP 2016 assessment nor the Heavy Quadricycle assessment, it also has achieved a very good rating. Occupant restraint strategy (in special concerning lateral impact) has been developed for the specific BEHICLE occupant seating layout with only one centred occupant in the first seat row. The case of light weight vehicles with a different seating layout with two occupants in the first seat row would need to be object of a specific study, due to the limited space from the occupant to the door. BEHICLE offers a similar side protection level independently of the side of impact (near side or far side) due to the centred driver position, in combination with a symmetrical four point seat belt with pretensioning function, a thick side airbag and a curtain airbag. Therefore, results in terms of occupant safety for far side impact will be equivalent to the results for near side impact. Finally, other matter to be taken into account is the absence of collapsible steering column in BEHICLE. Experience in conventional vehicles has demonstrated that a collapsible Steering Column reduces the loads on the chest in the case of frontal impacts of vehicles, so it offers possibilities to improve the actual BEHICLE results in frontal impact load cases. CONCLUSIONS Despite the fact that light weight cars are limited by their low weight, composite materials have evolved to achieve a high resistance and reduced weight in comparison to traditional materials. This fact, in combination with a good passive restraint system strategy, as in BEHICLE, can offer a good level of occupant safety for the driver, similar to conventional Molinero 12

103 Supermini vehicles that are now on the streets, assessed according to Euro NCAP 2013 and Euro NCAP 2016 procedures, and much better than heavy quadricycles driving along the cities. Also good occupant protection in the case of far side impact was demostrated. Germany, Sep ). IRCOBI Secretariat, Zurich (Switzerland). APPENDIX A. Dummy kinematics in FE- Models ACKNOWLEDGEMENTS The authors would like to express gratitude to the European Commision for its financial support of the project BEHICLE: Best in class vehicle: Safe urban mobility in a sustainable transport value-chain. (Grant Agreement Nº605292), a Seventh Framework Programme Project, THEME [GC.SST ] within which the work presented here has been carried out. The authors also extend their thanks to all the BEHICLE project consortium members for their help, cooperation, open discussion and continued support. REFERENCES [1] European New Car Assessment Programme (Euro NCAP). (2016). Vehicle rating and rewards. Available from [2] Holtz, J Crashworthiness enhancement of a composite intensive, multimaterial fully electric urban car. In proceedings of the 6 th Hybrid and Electric Vehicle Conferences 2016 (HECV 2016) (London, UK, Nov ). Institution of Engineering and Technology (IET), London (UK) [3] European New Car Assessment Programme (Euro NCAP). (2017). Adult Occupant Protection, Rating Explained. Available from [4] Carhs gmbh Safety Companion Aschaffenburg, Germany. Carhs.training gmbh. Figure A-1. Dummy kinematics in MDB configuration models without Passive Restraint Systems, with only Top Belt and with all Passive Restraint Systems. [5] Stein M Concept for lateral Impact Protection of a centered Driver in a light Electrical Vehicle. In Proceedings of the 2014 International Research Council on Biomechanics of Injury conferences (Berlin, Molinero 13

104 APPENDIX B. Dummy kinematics in sled tests 20ms 1 20ms 40ms 1 100ms 80ms Sled T 100ms 120ms Figure A-2. Dummy kinematics in Pole configuration models without Passive Restraint Systems, with only Top Belt and with all Passive Restraint Systems. Figure B-1. Dummy Sled Tests kinematics resultsin MDB sled tests. Molinero 14

105 20ms APPENDIX C. Dummy kinematics in crash tests 20ms 40ms 0ms 80ms 40ms 80ms 120ms Figure C-1. Dummy kinematics in MDB crash test. 120ms Figure B-2. Dummy Sled Tests kinematics resultsand Assessment in Pole sled tests. 0ms 40ms 80ms Figure C-2. Dummy kinematics in Pole crash test. 120ms Molinero 15

106 A DEVELOPMENT OF PANORAMIC SUNROOF AIRBAG Byungho, Min Garam, Jeong Jiwoon, Song Hae Kwon, Park Kyu Sang, Lee Jong Seob, Lee Hyundai Mobis Co., Ltd Republic of Korea Yuji Son Hyundai Motor Co., Ltd. Republic of Korea Paper Number ABSTRACT NHTSA released the Standard FMVSS No. 226 final rule in January 2011 for the protection of passenger from ejection through side windows during rollovers or side impact events. However there is no safety device to protecting the occupants from the roof ejection. Furthermore, the sunroof market size is increasing every year. For these reasons, the ejection to the roof is exposed to great danger. Therefore, in this study, the panoramic sunroof airbag was developed for the passenger protection from the ejection. Based on a vehicle rollover behaviour, the TTF and deployment times were derived. And the cushion structure was designed to prevent the ejection of passengers from a confined space within the roof. MIN <#>

107 INTRODUCTION The sunroof has a market size of $4,924.5 million in And the average growth rate was projected to grow by 10.9 % by ) In particular, the growth rate in the premium automotive market in China, India, and Korea has become more pronounced. result, there is a very high level of ejection in the center and border of the panoramic sunroof. (b) Ejection area in rollover Figure 2. Ejection in rollover (NASS-CDS 2000 to 2015) Figure 1. Global sunroof market size Depending on these mounting rates, the proportion of the vehicles with the sunroof is gradually increasing, resulting in an increase in the number of accidents driven by the sunroof. According to NASS-CDS data 2) analysis in 2000 to 2015, the total accidents caused by ejections from the overall crash rate are 15%, and the departure of the roof is up to 11%. In case of panoramic sunroof, there is mush higher risk in the ejection than conventional sunroof cause of its larger size of the window. Futhermore, the roof is exposed to a very high degree of risk when it has no protection against any rollover incidents in the event of a field crash. Heudorfer et al. had conducted research the airbag which is installed and deployed from the vehicle seatback. 4) And airbags are positioned between the passenger's head and ceiling, reducing the injuries both head and neck. However, due to the structural limitations of the cushion, it was not possible to fulfil the role of the occupants from the ejection of the passengers. In this study, we developed the panoramic sunroof airbag that protects the passenger from the ejection, which increasing risk of accidents with continuously growth of mounting rates. Based on a vehicle rollover behaviour, TTF and deployment times were derived. A cushion structure was designed to prevent the ejection of passengers from a confined space within the roof. MODULE DESIGN CONCEPT (a) Ejection in rollover In the case of the ejection through the it side window, the regulation of the FMVSS No. 226 is protected, but there is no safety device to protecting the occupants from the roof ejection. Thus, the ejection to the roof is exposed to great danger. To identify these risks, NHTSA conducted an evaluation of the risk of occupant ejection from the sunroof. 3) As a The panoramic sunroof airbag is an integral structure that is mounted on the rear of the inside panoramic sunroof panel and is deployed from the rear to the front. In Figure 3, the airbag module consists of an inflator, cushion, mounting bracket, cover and deployment guider, and the deployment guiders are a bar-shaped steel structure with the moved mounting tabs along it. In addition, the mounting tab has plurality of the steel annular sturctures, and the mounting tabs are slid along guider. In this case, the deployment guiders play a role that the panoramic sunroof airbag can be deployed smoothly to the front in full deployment and at the same time controls the upward or downward movement of the cushion in the full deployment, thereby preventing occupant MIN <#>

108 ejection because of the coverage of cushion. The concept of full deployed cushion is deployed inside the panoramic sunroof as shown in Figure 4, and is positioned between the roll blind and panoramic sunroof glass to prevent the occupant ejection. Moreover, the zigzag folding was applied to induce the sequential deployment of the cushion and to reduce the friction between the roll blind and the panoramic sunroof glass. In the present study, three types of cushion patterns were proposed as follow. Figure 5 shows each cushion pattern. (a) Parallel chamber type During the deployment, the gas can be rapidly distributed from the diffuser to each horizontality chamber. (b) Edge chamber type During the deployment, gas is flowed from the diffuser into the left/right outter chambers to induce rapid coverage. (c) Dual step chamber type It is concept that 1 st row chamber is deployed after 2 nd row chamber is fully deployed. The 1 st row chamber and 2 nd row chamber are independently separated from center pillar of the panoramic sun roof to avoid interference with the center pillar of the panormic sun roof. (a) Parallel chamber type (b) Edge chamber type Figure 3. Panoramic sunroof airbag module (c) Dual step chamber type Figure 5. Panoramic sunroof airbag cushion pattern Figure 4. Panoramic sunroof airbag concept DEPLOYMENT CHARACTERRISTIC TTF and deployment time MIN <#>

109 In this study, the TTF and depolyment time was set according to the movement of dummy to the roof area at the rollover test (Test No. 6088) by NHTSA which the test mode was the unbelted condition of FMVSS No. 208 Dolly rollover.5) In this test, the driver dummy was moved to the roof area at 90ms after the full deployed time of curtain airbag. Therefore, the TTF was set by the full deployed time of the curtain airbag. And, the deployment time was set by 60ms in consideration of the marginal time about 30ms due to the movement the driver dummy to the roof area. Cushion patterns The static deployment test was performed to observe the deployment characterictics according to the proposed 3 cushion patterns. In the tests, to reflect the real world accident situation, airbag modules are installed in the panoramic sunroof system which is under mass production. Figure 6 (a), (b), (c) shows the deployment test results of proposed 3 cushion concepts, respectively. In Figure 6 (a), parallel chamber concept is shown to be deployed abnormally due to an interception between the center pillar of the panoramic sun roof and the airbag cushion. As mentioned above, the gas flow uniformly to each chamber from rearward to forward direction. However, the tab velocity VTAB is slower than the gas velocity VGAS because of the friction induced between the tab and the guider pipe. It is observed that this velocity difference between the VGAS and VTAB result in the cushion to be twisted during its deployment. Similar trend is also observed in the edge chamber concept as shown in figure 6 (b). In case of edge chamber concept, the speed of VTAB is expected to be increased since the gas initially flow to the edge of the panoramic roof airbag cushion. However, still the gas velocity VGAS is somehow larger than the tab velocity VTAB and therefore the airbag cushion is twisted during the deployment. The only difference is that the time when the cushion be twisted is delayed compare to parallel chamber concept. Figure 6 (c) shows the deployment test result of the dual step chamber concept. In the figure, it is shown that the cushion deploy normally without twisting and satisfy the coverage requirement to protect the occupants. Actually, in the dual step chamber concept, the gas velocity VGAS should be very similar compare to edge chamber concept at initial stage since it flows gas to the edge of the airbag cushion together with the edge chamber concept. However, as mentioned, there exists an gas delaying region at the center of the airbag cushion in the dual step chamber concept. Therefore, the average of the VGAS should be much slower than that of the previous proposed concepts and finally the gas velocity VGAS and the tab velocity VTAB become almost equal. (a) Parallel chamber type (b) Edge Chamber Type (c) Dual step chamber type Figure 6. Deployment test result of component level Among the proposed cushion patterns, the dual step chamber pattern was finally chosen since it showed MIN <#>

110 stable and robust deployment performance. The full deployment time of the chosen dual step chamber time was 70ms but the full deployment time of the airbag which seems to be insufficient to protect the occupant at proper time. As previously stated about deploymnet time, the airbag should be deployed within 60ms. An effort to increase the airbag deployment speed was needed for the completeness of the product. Therefore, the design concept of the tab was modified. The material of the tab was changed as steel material on behalf of previous fabric material (See Figure 6). The time for full deployment was reduced from 70 to 60 ms when adopting steel tab to the developed airbag system. Deployment characteristic : System level Additional tests were performed in order to guarantee the robust deployment characteristic in the real-world conditions. In general, awning screen is installed inside the panoramic sunroof device. It is easily expected that the position of the screens could affect the deployment characteristics of the panoramic sunroof airbag in the real-world rollover situation. Therefore, additional deployment tests were performed for the rigorous validation of the proposed airbag concept. The tests were performed in real vehicle level including all related parts of the panoramic sunroof devices. the cases the panoramic sunroof airbag deployed properly. EJECTION MITIGATION TEST RESULT In this study, the ejection mitigation(ejm) test of the panoramic sunroof was evaluated on NHTSA roof ejection research which was announced in 2016 SAE 3) and FMVSS No.226 6) which the curtain airbag is evaluated for the prvention of occupant regarding the occupant protection in a rollover accident. As shown in Figure 8, the impacted target was selected as 2 point by 1 point of the corner area and center area, which are expected to have high possibility of the ejection. The impactor weight is 18kg, time is 1.5sec, and the speed is set as 20kph. The ejection mitigation evaluation of panoramic sunroof proceeded the dual stage chamber type cushion which is guaranteed. As shown in Figure 9 and 10, the corner area was 20% higher than the excursion value of the MOBIS standard due to the lack of thickness and absence of the support member, and the center area which has an additional hinge effect is 80% higher than the excursion value of the corner area. Figure 8. EJM test target positions Figure 7. Deployment test result of System level Figure 7 shows the test results. In the Figure 7, it is shown that the airbag deployed stably for both the screen-opened/closed conditions. However, time for full deployment is slightly increased in the closed conition. The tests were also performed in other conditions such as half closed condition and for all In this study, in order to solve excessive excursion value, the concept of the double cushion structure was applied to this issue as shown in Figure 11. The upper and lower cushion chamber patterns were designed to be orthogonal to each other in order to increase the thickness of the cushion and alleviate the reduction of the bending according to the hinge. As shown in Figure 12, the evaluation result of the improved cushion improved 15% at the corner area and 90% at the center area compared with the primary evaluation, and it achieved similar performance to the curtain airbag of the current mass production level. The reason for the lack of improvement in the corner area compared with the center area is that the cushion thickness is increased but the support member with the correlation MIN <#>

111 componets is insufficient. Therefore, the double cushion structure added supporting chamber like tongue shape. This shape will be supported on headlining which area is front vehicle part for the reduction excursion value of corner area as shown in Figure 13. The verification of this improved cushion will be performed within the first half of this year. Figure 9. EJM result of dual step chamber cushion type Figure 13. Additional improved type about double cushion structure for EJM Figure 10. Cause analysis about excessive value CONCLUSIONS In this study, the panoramic roof airbag was developed to protect from the occupant ejection to the panoramic sunroof, which is highly dangerous in the field due to the increase of installation rate. The package and performance were developed on the basis of the current product of the panoramic sunroof module, and the conclusions are as follows. Figure 11. Improved cushion type for EJM Figure 12. EJM result of double cushion structure type 1) The panoramic sunroof airbag was designed as an integrataed mounting module inside the panoramic sunroof module, that is deployed from the rear to the front along the development wire type guider. 2) The optimum cushion pattern was selected through the deployment test and also evaluated the worst condition that was the screen-closed. Through this, the robustness of selected optimum cushion pattern is secured. 3) In order to develop the ejection mitigation performance, the double cushion structure type of delay cushion type was applied to the optimum cushion type to secure the performance of the present level of curtain airbag. MIN <#>

112 4) In the future, the study will be performed on an improvement of the ejection mitigation performance with applying a cushion of the overlap structure to the headlining edge of the front. 5) When the panoramic roof airbag is applied, it is necessary to set up the test procedure and the test specification standard about the ejection mitigation through the study of the occupant ejection from the panoramic sunroof. REFERENCES [1] P&S Market Research Global Automotive Sunroof Market Size, Share, Development, Growth and Demand Forecast to 2022 Industry Insights by Material Type [2] National Highway Traffic Safety Administration, U.S. Department of Transportation [3] Prasad, A. and Duffy, S Update on NHTSA Roof Ejection Research. SAE Government Industry Meeting, January [4] Heudorfer, B. and Breuninger, M ROOF BAG-A CONCEPT STUDY TO PROVIDE ENHANCED PROTECTION FOR HEAD AND NECK IN CASE OF ROLLOVER. 19 th ESV Conference, June [5] Prasad, A. and Duffy, S Update on NHTSA Roof Ejection Research. SAE Government Industry Meeting, January [6] National Highway Traffic Safety Administration, U.S. Department of Transportation ort=y&r_tstno=6068&existvideo=y&v_tstno=6068 &database=v&tstno=6068 MIN <#>

113 Side Crash Detection Using Vehicle Behavior Change Kenyu, Okamura Kazuhiro, Daido Honda R&D Co., Ltd. Automobile R&D Center Japan Paper Number ABSTRACT The conventional detection method of a side crash is using either a pressure sensor located on the door or an acceleration sensor, also referred to as G sensor. These sensors detect body intrusion in a side crash. This paper focused not only on intrusion of body but also on vehicle behavior change, which is detected simultaneously with body intrusion in a side crash. Using intrusion and behavior change of vehicle, an investigation of side crash detection performance was conducted. Two methods were devised to detect vehicle behavior change in a side crash. One method is using yaw-rate sensor located at the center of the vehicle, and the second method is using a G sensor, which has a sensitivity axis in the longitudinal direction of the vehicle and located on the body side. A side crash detection algorithm was also devised which combined G sensor of lateral direction, which detects lateral accelerations in a side crash, and a yaw-rate sensor or G sensor of longitudinal vehicle direction, which detects other changes to the impacted vehicle other than lateral accelerations, referred to in this study as vehicle behavior. This research sought to determine whether crash detection performance can be satisfied for various crash modes using numerical simulations. The results of these numerical simulations indicate that G sensor response time is fast which makes it effective in detecting a high speed crash. The results also showed that yaw-rate data is stable, which implies that data is reliable, allowing the use of the developed crash detection algorithm for predicting vehicle behavior changes, within certain speed limits. Moreover, a side crash test using a test vehicle, also referred to in this paper as Complete Body Unit or CBU, CBU was also completed and confirmed that body intrusion and vehicle behavior change occur simultaneously and can be reasonably detected a side crash using this paper s crash detection algorithm. This could potentially transform side crash detection in the automotive industry. Okamura 1

114 INTRODUCTION The conventional method of detecting impact forces during a side crash is using an accelerometer or a pressure sensor [1] [2]. Accelerometers detect velocity changes while pressure sensors detect door cavity volume changes which indicate a body intrusion in a crash. These methods are effective only when the impact directly hits the sensors. Such limitation requires either the strategic placement of sensors or the use of multiple ones in a vehicle. The latter would increase the complexity of car development and would increase costs. To avoid both development complexity and cost increases, as well as to attain a more inclusive analysis of side crash investigations, the research focused on overall vehicle motion changes during an impact, which is referred to in this research as vehicle behavior change. The purpose of this research is whether crash detection using vehicle behavior change is possible. In the case of a side crash, when the impact point is near the vehicle s center of gravity (COG), the energy is absorbed mainly by body deformation and there is little vehicle behavior change. But when a crash point is farther from the vehicle s center of gravity (For example, in a vehicle of FF layout, in the case of a second row side pole impact), vehicle behavior change occurs simultaneously with body intrusion. Accelerators and pressure sensors are effective in detecting a side crash when the impact is near COG. However, when the impact is away from COG, it may be useful to deploy a system that can calculate vehicle behavior change. Vehicle behavior change is a physical quantity that can be calculated from crash velocity and can be theoretically used for detecting occurrence of crashes. Figure 2 illustrates middle speed and high speed side pole crashes. Both speeds require the deployment of the side curtain airbag. The picture indicates that there is considerable vehicle behavior change. Furthermore, the picture shows that vehicle behavior change is more apparent in the high speed crash than in the middle speed crash. It was observed that there is insignificant vehicle behavior change at low speed crashes. However, at crash speeds that required side curtain airbag deployment, vehicle behavior changes are observable and can be calculated as a physical quantity. This physical quantity can be used for crash detection. Figure 3 shows a middle speed pole crash at 150 msec. after impact. The picture on the left in Figure 3 shows the pole crashing into the first row side. This impact point is close to the center of gravity of the vehicle. Energy is mainly absorbed by the deformation of the body and vehicle behavior change is again, insignificant. In this case, accelerometers and pressure sensors can easily detect body intrusion. It is not necessary to use vehicle behavior change for crash with impact points that are close to the vehicle s center of gravity. The picture on the right of Figure 3, show the impact at the second row side of the vehicle and far from the center of gravity. The picture also show that there is both intrusion of body and vehicle behavior change (Yaw of vehicle) occurring simultaneously at 150 msec. after impact. There is a significant vehicle behavior change, which can be quantified and can be used as an indicator of a crash. METHODOLOGY Principle of detection Figure 1 and Figure 2 indicate a difference of vehicle behavior changes due to a difference in crash speeds with nearly the same crash point at the second row. Figure 1 show the situation at 150 msec. of the CBU test where the pole collides with second row side at low speed and middle speed. Low speed is a crash speed in which the side curtain airbag should not deploy. The middle speed is a crash speed in which the side curtain airbag must deploy. The picture on the left side of Figure 1 show that at the low speed crash, energy is absorbed by the deformation of the vehicle s body and there is insignificant post-impact vehicle velocity change. On the other hand, the right side of Figure 1 show that energy was not absorbed by body deformation and there is considerable post-impact vehicle velocity changes or vehicle behavior change. Figure 1. Vehicle positions at 150 msec. after impact for low and middle speeds. Okamura 2

115 For example, when a side pole crash occurs near a second seat, a yaw-rate sensor detects yaw-rate in counterclockwise direction as shown in Figure 4. Another method is using a satellite G sensor which has a sensitivity axis of vehicle longitudinal direction. The satellite G sensor is located on body side. This sensor is called a satellite impact sensor and is abbreviated as SIS in this paper. For example, when a side pole crash occurs near a second seat, SIS detects G which direction is indicated with green arrows in Figure 4. Figure 2. Vehicle positions at 150msec. after impact for middle and high speeds. Figure 3. Vehicle positions at 150msec. after impact for front and second row side crashes. Detection methods of vehicle behavior change. Two methods were devised as detection methods of vehicle behavior change during a side crash beside second seat. One method is to detect vehicle yaw-rate directly using a yaw-rate sensor. Since a vehicle has inertia, when vehicle behavior change will once occur, data of yaw-rate is stable. So we expected that performance of crash detection has stability. How to use the two detection methods The characteristics of data are different between G and yaw-rate sensors. So we use two methods. Since G is transmitted along a body member during a crash, there is a characteristic that the response is fast. Therefore, it seems that it is suitable for a high speed crash. Because a high speed crash requires a fast response time. The response of the yaw-rate sensor is not as fast as the accelerometer but it is stable which makes it suitable for detecting middle speed impacts. The characteristics of the yaw-rate data are shown by CBU test data in Figure 5. The CBU test data of a side pole crash into the second row at middle speed is the data shown in Figure 5. In this paper, in a case of a high speed side crash, we use accelerometer sensor data for detection of vehicle behavior change as we expect fast response and in a case of a middle speed side crash we use data of yaw-rate sensor to detect of vehicle behavior change as we expect a stable response. In this paper, a timing of side crash detection for establishing passenger protection performance is hereinafter referred to as T-TTF (Target Time to fire) and we researched whether the sensing system can detect the crash at the desired T-TTF. Figure 5. Data of each sensor at middle speed crash. Figure 4. Detection direction of each sensor at a time of a side pole crash. Layout of sensors used for this research Since the behavior of the entire vehicle changes during a side crash event the area that can be utilized for sensor application is large. We aimed at a simple sensor system. So, we tried to integrate as much as possible with a conventional sensor system. For that Okamura 3

116 reason, we have conducted this study based on a simple sensor system shown in Figure 6 and Figure 7. Yaw-rate sensor data was acquired at the SRS-unit located on the front floor. Detection of vehicle behavior change by acceleration used a sensor located on B-PLR LWR as shown in Figure 6 and 7. Conventionally, in order to detect vehicle acceleration, this sensor has a sensitivity axis in lateral direction of vehicle. In this study, the sensor also detects vehicle behavior change, this sensor also incorporates an accelerometer in the longitudinal direction. In this paper, X, Y, Z coordinates are defined as shown in Figure 7. Longitudinal direction of a vehicle is X, lateral direction is Y, and vertical direction is Z. In addition, when it is written as Gy, it indicates that it is G in the vehicle Y direction and when it is written as Gx, it indicates that it is G in the vehicle X direction. Figure 6. Sensor layout. Figure 7. Sensor layout. Crash detection algorithm In a pole crash impacting the second row, there are two features as explained. One is that G and yaw-rate sensors respond for a short time. Another is that Gy data and Gx and yaw-rate data related to vehicle motion change respond are available at the same time. An algorithm to detect these features is needed. We devised a crash detection algorithm based on a two-dimensional map of vehicle motion change. This algorithm shows features which body intrusion and vehicle behavior change occur. The two-dimensional map which we devised is as shown in Figure 8. In this algorithm, the horizontal axis is a value calculated based on data of yaw-rate sensor or SIS Gx. The vertical axis is a value calculated based on data of SIS Gy. Crash detection is carried out based on a path of these data on this map. In order to capturing features of the crash data occur over very short time duration we focus on a point which a value is calculated based on yaw-rate and G. In order to capture a short time event, it is effective to use a difference value of about several tens of msec. for yaw-rate sensor data and a definite integral value of about several tens of msec. for G data. By looking at these data for a certain time in this way, these data becomes large in a short duration event and small in a long duration event. This makes it possible to easily separate a crash event from a misuse event during normal driving. For a certain time of several tens of msec. optimum value differs depending on body, so adjustment is required depending on body. In this paper, in a case of high speed crash, a definite integral value of SIS Gy is used as the vertical axis and a definite integral value of SIS Gx is used on the horizontal axis. In a middle speed crash, a difference value of SIS Gy is used on the vertical axis and a difference value of yaw-rate sensor is used on the horizontal axis. We investigated crash detection performance using these physical quantities in the algorithm. In the map of this algorithm, intrusion of body and vehicle behavior change also come out as a physical quantity corresponding to crash speed, so that distance from the origin has a meaning corresponding to crash speed. We do not want to deploy airbag in case of a low speed crash and want to deploy it in case of a middle speed or more. Therefore, by setting a threshold value to be larger than data of a low speed crash, it is possible to make a judgment that airbag is not deployed, because it does not exceed a threshold value in case of a low speed crash. Since data exceeds a threshold value in case of a middle speed crash or more, it is possible to make a judgment to deploy airbag. Also, when a side crash occurs near center of gravity of vehicle (for example, crash of front seat side), intrusion due to crash comes out large and vehicle behavior change is small, so a value on the horizontal axis is small and a value on the vertical axis is large. So a path of the data on the map of algorithm extends for upward shown as Figure 8. On the contrary, when a crash occurs at far from center of gravity of vehicle, intrusion and behavior come out, so that a path of the data on the map of algorithm Okamura 4

117 extends for diagonally upward. Approximate crash position can be estimated by a direction of the data. Table 1. Conditions of simulation Figure 8. Crash detection algorithm using two-dimensional map. RESULTS Verification on various crashes using simulation For various crashes, we confirmed whether the algorithm we devised can be possible to respond. Simulation used LS-DYNA as a solver and used a model correlated with CBU test. It was verified with a model based on Honda Accord. The verified modes are as shown in table.1. In this time, we selected modes which pole impacts the second row side. We verified by changing diameter of pole, crash speed, crash angle and crash position. The crash angle is as shown in Figure 9. The crash position is as shown in Figure 10, and the second seat side is defined as mid. Front and mid, mid and rear have a distance of 400 mm. The simulation data was acquired at the place shown in Figure 6. Figure 11 and Figure 13 show all the simulation data of the low speed crash. Figure 11 uses SIS Gx and Gy, and Figure 13 uses yaw-rate and SIS Gy. A threshold value for deployment of airbag is set larger than the data of low speed crashes. Figure 12 and Figure 14 represent part of the simulation data. The high speed crashes was detected by SIS Gx and SIS Gy as shown in Figure 12. The middle speed crashes was detected by yaw-rate and SIS Gy as shown in Figure 14.The data in Figure 12 and Figure 14 were plotted up to the T-TTF and could be judged exceeding the threshold until the T-TTF. So it found that the required performance is satisfied. Figure 12 and Figure 14 show the data of a part of the simulation, but we confirmed that the crash detection performance is satisfied in all data of the simulation which we conducted. We found that the crash detection performance on the various side pole crashes was sufficiently satisfied even by the method of detecting vehicle behavior change. Figure 9. Image of side pole crash of angular difference. Figure 10. Image of side pole crash of position difference. Figure 11. Simulation data of the low speed crash and this map is drawn by SIS Gx and SIS Gy. Okamura 5

118 Figure 12. Simulation data of the high speed crash and this map is drawn by SIS Gx and SIS Gy. Figure 14. Simulation data of the middle speed crash and this map is drawn by yaw-rate and SIS Gy. Verification in CBU test We also carried out CBU tests which modes are shown as Table 2. Figure 15 and Figure 16 show the data of the CBU tests. Figure 15 uses SIS Gx and Gy, and Figure 16 uses yaw-rate and SIS Gy. A threshold value for deployment of airbag is set larger than the data of the low speed crashes. For middle speed and high speed crashes data, they were plotted up to the T-TTF. So the data could be judged exceeding the threshold until the T-TTF. So the required performance is satisfied. Table 2. CBU test Modes Figure 13. Simulation data of the low speed crash and this map is drawn by yaw-rate and SIS Gy. Okamura 6

119 Therefore, as shown in Figure 17, the data stays close to the origin. So the misuse during normal driving is not a problem on the algorithm. Moreover, the vertical axis of the algorithm is a physical quantity of intrusion. Since intrusion of body does not occur during normal driving, the data does not come out on the vertical axis. Therefore, it was found that the toughness is high for the misuse. Figure 15. CBU data drawn by SIS Gx and SIS Gy. Figure 16. Yaw-rate data during spin on the low μ road. Figure 16. CBU data drawn by yaw-rate and SIS Gy. DISCUSSION Verification of misuse during normal driving Vehicle behavior change also comes out during normal driving. So, we confirmed performance of the algorithm by CBU test. We considered six modes. These test modes are driving circle, turn, lane change, spin on low μ road, riding on a curb of rear wheel and passing on split μ road. In these modes, the yaw-rate comes out most as spin on low μ road. Therefore, the data of spin on low μ road is represented in Figure 16. The duration of the data of the vehicle behavior change at the time of spinning is several seconds. However, the crash detection algorithm uses a difference value of several tens of msec. It can be seen that the time scale is completely different. Figure 17. CBU data on the algorithm during spin on the low μ road. Verification of influence on vehicle behavior change when center of gravity changes Since vehicle behavior is used, we can consider that an influence will come out when center of gravity of vehicle changes. In vehicle with FF layout, as a case where position of center of gravity changes, it is conceivable that a heavy weight is fixed to the Okamura 7

120 trunk room. We verified this condition using simulation. The simulation was carried out in Accord based car with 300 kg weight fixed to trunk room. As the weight was fixed, the center of gravity of the vehicle moved about 300 mm rearward. The results of simulation show in Figure 19. As the center of gravity changed, the data changed slightly, but the crash detection performance was influenced little. When the crash position changes to a position close to the center of gravity of the vehicle, in principle, intrusion becomes larger and vehicle behavior change becomes smaller, so it is expected that the path on the algorithm will change to upper left in the algorithm. Although slight changes were observed according to this principle it was confirmed that the influence is quite small. shown in Figure 6.The CBU test was carried out using 1-box vehicle with a third seat. The CBU test modes are shown as Table 3. Figure 20 and Figure 21 show the data of the CBU test. The plot of the data was until T-TTF except the low speed crash data. In the crash of middle speed or more, the data exceeds the threshold. So, we confirmed the crash detection performance is satisfied. As for the reason why the detection performance was satisfied with the side pole crashes of third seat side, when the crashes occurred farther from the center of gravity of the vehicle, the value of the vertical axis becomes smaller, but the value of the horizontal axis becomes larger. As a result, the data could be extended on the algorithm to a position far from the origin. Although the data of SIS Gy is small, G sensor can detect. It can be considered that this is due to the fact that the member of body exists up to the third seat and the G data can be transmitted through this member. If the body member does not exist up to the third seat, the data does not extend on the vertical axis of the algorithm. If the data only extend on the horizontal axis alone, it is difficult to judge the crash. This is because the vehicle structure is not compatible with the detection concept. Table 3. CBU test modes Figure 18. Image of the change of COG when heavy weight fixed to trunk room. Figure 19. Data changed on algorithm when the vehicle center of gravity changed. Detection performance of vehicles with third seat Even in the case where the vehicle has a third seat, we also researched in the CBU test whether crash detection is possible with the sensor configuration Figure 20. CBU data on the algorithm using SIS Gx and SIS Gy. Okamura 8