SENSITIVITY ANALYSIS OF INJURY PATTERN FOR PASSENGER VEHICLE S DRIVERS IN C2C COLLISIONS AT INTERSECTION

Transcription

1 SENSITIVITY ANALYSIS OF INJURY PATTERN FOR PASSENGER VEHICLE S DRIVERS IN C2C COLLISIONS AT INTERSECTION Chinmoy, Pal Shigeru, Hirayama Nissan Motor Company Ltd Japan Sangolla, Narahari Manoharan, Jeyabharath Gopinath, Prakash Vimalathithan, Kulothungan Renault Nissan Technology Business Centre India India John, Combest Nissan Technical Center North America USA Paper Number ABSTRACT Nearside occupants are at a higher risk for serious injury than far-side occupants in side impact collisions. Accidents where the driver is the only occupant accounts for more than 8% of all side impact injuries. This paper presents the results of study on sensitivity of serious injury outcome for drivers involved in nearside car to car (C2C) collisions, especially at intersection. In total, 865 intersection C2C crashes (NASS-CDS CY ) are analyzed in detail to determine the injury level outcome based on different crash factors, such as deltav, age, gender, striking vehicle type, impact location (F,Y,P,Z,B-regions) and impact angle. Injury sensitivity with respect to impact location was studied in detail. A univariate logistic regression was done to check the probability of a serious injury occurring between the center and end locations. A higher number of serious to fatal injuries (57%) occurred in the Y-region when compared to other impact locations. Additionally, a higher number of serious to fatal injuries (6%) occurred when the direction of impact is 1 o clock. Injury occurrence for L-type offset impacts at both ends of the vehicle s-b and F regions, were quite low. In P and Y regions (Ttype impact), the chances of having AIS3+ injuries were higher for an impact angle of 1 o clock than those of 8 and 9 o clock. The probability of having AIS3+ injuries was higher in senior (age>6) drivers than in younger drivers. When the striking vehicle changed from PV to SUV, a higher number of serious injuries were observed. AIS3+ injuries in head and lower extremity (including pelvis) injuries were increased, when the striking vehicle was a SUV as compared to a PV. But, there is not much change in the thorax region. Finally, the above real world accident results were also verified with barrier to car FE simulations. A certain amount of offset of the maximum intrusion point of the deformation profile from the occupant sitting position in both the forward direction (towards A-pillar) and backward direction (towards C-pillar) influences the outcome of the driver s injury level. Results of the present study provides opportunity for considering new interaction terms between impact locations, impact directions, occupant height, vehicle type and others in improving the accuracy of AACN ISP algorithm injury prediction in nearside lateral collision. PAL 1

2 INTRODUCTION In the National Automotive Sampling System [1], (NASS, US) real world field injury data are collected to obtain different crash and occupant related information such as vehicle/ crash data, impact vector, occupant exposure, incidence, risk, and injuries to different body regions of the occupants inside the vehicle. This information is essential for (i) the development and improvement of new anthropomorphic test devices with the help of various crashworthiness studies, and (ii) introduction of new or modifications of existing vehicle safety and traffic standards to meet the continuous demand of an ever-changing mobile society around the world. Side impact crashes are generally critical even at low impact speeds because of the energy absorbing capability of side structure, [2],[3]. Starting with the introduction of side impact regulation standards, side impact New Car Assessment Program (SINCAP, 1996) and consequent upgrades of new injury criteria and the specification of the side impact ATD device in the front and rear seating positions [4],[5],[6] the crashworthiness and occupant safety performance of new generation vehicles have been improved considerably, in last decade. These are the contributions of various valuable past research works [7],[8],[9] including those of full-scale vehicle MDB and pole tests conducted using different types of advanced internal instrumentation. With the introduction of the present IIHS s PV-vs- SUV C2C tests and future advanced biofidelic ATDs (5 th -ile World SID) for occupant safety evaluation, more studies are necessary to identify the remaining scopes of possible improvements and further reduction of fatalities based on real world accident analysis. A recent study indicates that approximately 4% police-reported rear-end crashes may have been prevented if all vehicles were equipped with FCW with AEB. Cicchino [1] mentioned that AEB systems that perform at a full range of speeds would likely prevent more crashes and injuries. Hence, in the near future, the present AEB market trend will lead to more attention and focus on C2C intersection crashes when the share of AEB in PV will reach 99% by 222 in US [11]. The objective of this study is to focus on the sensitivity analysis of injury pattern in C2C side impact intersection collisions of passenger vehicles to indicate various effects, such as, impact location, impact angle, striking vehicle type, gender, and height of the driver. Using CY NASS- CDS data, Xinghua et.al, 212 [12], indicated the importance of the consideration of the crash configurations beyond the scope of existing sideimpact regulatory tests and stressed the necessity of vehicle crashworthiness and restraint system design to better protect occupants in real-world crash scenarios. However, this present study used more recent crash data (CY NASS-CDS) to capture the current trends of driver injuries in newer vehicles and also verified the accident analysis trend with numerical FE simulations. DATA & METHODS This study used National Automotive Sampling System Crashworthiness Data System (NASS-CDS) accident data from calendar year 24 to 214. Table 1 shows the assumptions used to prepare the input data set in this study. The accident samples are limited to car to car intersection side imact planar collisions (i.e., excluded crashes with primary general area of damage as top or bottom and rollovers). Values with unknowns have been removed. In total, 865 vehicles were extracted using criteria, to perform the accident analysis and logistic regression [13] analysis (details of which are described in later sections). Logistic regression is used to determine the effect of impact location on serious injuies occurring in side impact and results were calculated using XLSTAT software [14]. Table 1. List of criteria for input dataset General Area Damage1=Left Direction of Force DOF=8, 9, 1 Impact Location=F, P, Y, B, D, Z Body Type PV (1-9,17) Model Year>=2 Driver Role=1, (Seat Position=11) Age16+ V2V OBJCTD<=3 Towed Away Vehicles No Ejection No Rollover No Fire Occurrence Excluded AIS7 injury Table 2 shows the final data set extracted from NASS CDS CY using the criteria mentioned in Table1. In total, 865 occupants with 3941 injuries involved in near side impacts were selected with six PAL 2

3 % of AIS3+ injuries % of AIS3+ injuries collision deformation codes( F, P,Y, Z, D, B) and three main impact angles (8, 9, 1 o clock) as shown in Figure 1. A weighted count of the number of AIS injuries in six impact locations and three impact angles were mentioned in Table A1 of Appendix A. It is found from the data that serious AIS3+ and minor AIS1&2 injuries were 14.5% and 85.5%, respectively. Table 2. List of input dataset Weighted data Raw data Occupants 181, Injuries 3,57, (4%), follwed by the head (33%) and the lower extremities including the pelvis region (12%). It is evident that thorax and head are more likely to have AIS3+ injuries among all the regions in side impact collisions. 5% 4% 3% 2% 1% % 4% 33% 12% 8% 5% 2% 1% Figure 2. Distribution of AIS3+ injuries in side impact with respect to each body region Figure 1. NASS-CDS collision deformation code (8, 9, 1 o clock are impact angles) RESULTS This section discuss the injury pattern of occupants involved in car to car collisions at intersection. The parameters considered in this study as folows: delta-v, age, gender, striking vehicle type, impact location, impact angle. Injury sensitivity with respect to impact location is studied in detail and verified with barrier to car FE simulations. A univariate logistic regression was carried out to check the probability of serious injury occurance between the center and end locations. Variation of Serious Injuries in Individual Body Region Figure 2 shows the AIS3+ serious injury distribution with respect to each body region for the occupants involved in near side impacts. It is observed that the thorax region has highest number of AIS3+ injuries Variation of Serious Injuries with Delta-V and Max Crush Value Figure 3 shows the variation of AIS3+ injuries with respect to delta-v. As the delta-v increases, the percentage of AIS3+ injured occupants increases. Figure 4 shows the variation of AIS3+ injuries with respect to the amount of deformation (maximum crush value). It should be noted that generaly the amount of deformation on external body depends on the impact velocity and strength of the vehicle. As the impact velocity increases, the level of AIS3+ injuries also increases and similar relationship exists with amount of external deformation value Delta-V (kmph) Figure 3. Variation of AIS3+ injuries with respect to Delta-V PAL 3

4 % of AIS3+ injuries % of AIS3+ injuries External Deformation of Vehicle (cm) Figure 4. Variation of AIS3+ injuries with respect to deformation (cm) Variation of Serious Injuries in Senior and Non Senior Occupants Table 3 summarizes the details of AIS3+ injuries for senior (age 6) and non-senior (age<6) occupants. Senior occupants are more likely to sustain AIS3+ injuries than non-senior occupants in side impacts. Table 3. AIS3+ injuries in senior and non-senior occupants Age Senior (Age 6) Non Senior(Age<6) Occupant count Within the driver alone population of PV, it is found that 64% of drivers were female and only 36% were male drivers. So the female drivers are more likely to get involved in side impact accidents at intersection than the male drivers in PV. The percentage of AIS3+ injuries corresponding to all six types of defomation locations and that of the combined P & Y regions corresponding to more cabin intrusion, are shown in Figure 6 for both male and female drivers. Female drivers suffered a higher number of AIS3+ injuries (57%) than male drivers (29%) in all regions. It is observed that the percentage of AIS3+ injury is more for occupants at P & Y locations (71%-female, 43%- male) when compared to that of all six regions (57%- female, 29%-male). It is observed that the percentage of AIS3+ injuries for females at P & Y locations when compared to that of all six regions. Percentage of AIS3+ injury is the ratio of number of AIS3+ injuries divided by total number of injuries (AIS1-6) in the specified regions for both male and female drivers. 8% 6% 4% 2% % Female 57% 29% Male 71% 43% All regions P & Y region AIS3+ / AIS1-6 17% 13% Variation of Serious Injuries in Female and Male Occupants Figure 5 shows the distribution of vehicles by the number of occupants. In the side impact accidents analyzed, 8% of the vehicles were driver alone and 2% had more than one occupant. Multiple occupants, 2% Driver alone, 8% Figure 6. Percentage of AIS3+ injuries in all six deformation locations and P+Y regions alone The variation of AIS3+ injuries is plotted with respect to the occupant height for both female and male drivers as shown in Figure 7. Using secondary axis of the right of this Figure 7, individual percentages of each category with respect to the total population (female and male drivers) are also overlayed on it for better visualization. First, the percentage of AIS3+ injuries is higher for shorter female drivers and as the height of the occupant increases, the percentage of AIS3+ injuries start to decrease. So, it is observed that the chances of having AIS3+ injuries are higher for shorter female drivers (corresponding to AF5, the 5 th percentile female population which is around cm in height). Figure 5. Distribution of vehicles with number of occupants present in side impact accidents PAL 4

5 % of AIS3+ injuries % of AIS3+ injuries % of AIS3+ injuries % of total population 1% 75% 5% 25% % Female Male % of population Occupant height (cm) 25% 2% 15% 1% 5% % Figure 7. Variation of AIS3+ injuries with respect to height of occupant for both female and male and percentage of total population by either an SUV or PV. Looking at the percentages of vehicle types in total striking vehcile population of this study, it is also observed that PV has the highest share with 48% followed by the SUV with 23%. The percentage of AIS3+ injuries are given in Figure 8 for PV and SUV striking vehicles. The percentages of AIS3+ injuries are higher when the striking vehicle type is a SUV (7%) than when it is a PV (6%). The results compare the percentage of AIS+3 injuries when impact location is only P & Y regions with that for all the all the regions. It is evident that the chances of having AIS3+ injuries are more at the P & Y regions than that of all the regions for both types, PV and SUV striking vehicles. The percentage of AIS3+ injuries for female and male drivers are plotted in Figure 9 when the striking vehicle changes from PV to SUV. The chances of having higher AIS3+ injuries for female drivers are higher when struck by a SUV than those by a PV. Table 4 summarizes the average height of females and males for PV and SUV struck vehicle. Within the AIS3+ population, the average height of the female drivers (164 cm-pv, 165 cm-suv) is less than that of the male drivers (177 cm-pv, 178 cm-suv). The average hieght difference between female and male is 13 cm. Table 4. Average of height of female and male 8% 6% 4% 2% % All regions 6% 62% PV P& Y region 7% 75% SUV When the PV as struck vehicle AIS3+ population Female (11) 164 Male (83) 177 Avg. Height (cm) When the SUV as struck vehicle AIS3+ population Female (19) 165 Male (11) 178 Avg. Height (cm) Variation of Serious Injuries with Striking Vehicle Type (PV v/s SUV) With increasing market share of SUV vehicles, the effect of the striking vehicle is also an important factor to be studied in side imapcts to check the sensitivity of injuryies. In this analysis, we studied the effect of injury sensitivity when a passanger vehicle was struck Figure 8. Variation of AIS3+ injuries with striking vehicle (PV v/s SUV) for all six regions and P & Y regions 1% 8% 6% 4% 2% % Female 66% 61% PV Male 81% 7% SUV Figure 9. Variation of AIS3+ injuries with striking vehicle (PV v/s SUV) for male and female Figure 1 shows the variation of AIS3+ injuries in four body regions (head, thorax, abdomen, lower extremities) for both PV and SUV striking vehicles. It is observed that the AIS3+ injuries in the head, lower extremity (including pelvis) and abdomen regions are PAL 5

6 % of AIS3+ injuries % of AIS3+ injuries % of AIS3+ Injuries increasing more than that in thorax when the type of striking vehcile is changed from PV to SUV. The following three mechanisms usually observed in IIHS side impact test conditions will explain the change in injury pattern related to SUV and PV striking vehicles[18]. (a) Matching of bumper height of SUV with the pelvis height of small females and less engagement of sill member of the struck vehicle with bumper of the striking vehicle may lead to more lower extremity injuries. (b) Shorter females upper torso, spinal cord and head will laterally bend more towards the window side due to higher input load near the pelvis and abdomen regions. (c) More intrusion at the abdomen region for SUV striking vehicle will lead to more abdomen injuries. 1% 8% 6% 4% 2% % 8% 69% PV SUV 86% 95% 82% 81% 74% Figure 1. Variation of AIS3+ injuries for different body region in P & Y region (PV v/s SUV) 6% Head Thorax Abdomen LE Variation of Serious Injuries with Respect to Impact Location and Impact Angle Figure 11 shows the percentage of AIS3+ serious to fatal injury distribution with respect to the impact location and impact angle, respectively. A higher number of serious to fatal injuries (57%) occurred in Y-region when compared to other impact locations and a higher number of serious to fatal injuries (6%) occurred when the direction of impact is 1 o clock. Injury occurrence for L-type offset impacts at both ends, B and F regions, were the lowest of all locations studied. In P and Y regions (T-type impact), the chances of having AIS3+ injuries were higher for an impact angle of 1 o clock than for angles of 8 and 9 o clock. Figure 12 shows the variation of AIS3+ injuries with respect to impact angle. A 1 o clock impact has a higher number of AIS3+ injuries (6%) than other impact angles Figure 11. Distribution of AIS3+ injuries in side impact with respect to impact location and angle Highest F Y P D Z B Location of Impact and Angle 8% IIHS Test Configuration 32% 6% 8' 9' 1' Impact Angle Figure 12. Variation of AIS3+ injuries with respect to impact angle Though the current IIHS side impact test configuration is mainly focused for P-region to reduce compartment intrusion, from this accident analysis it is observed that the Y-region is having higher number of AIS3+ injuries than the P-region. Logistic Regression Test A statistical test was performed to determine the probability of AIS3+ injuries with respect to impact location. Out of 865 vehicles, 55 vehicles were considered. To understand the sensitivity of injuries with respect to impact location, the side regions of the vehicle area is divided into two separate categories i.e. both ends as one category and the central region as another category, as shown in Figure 13. The cases whose impact location is in front (F region) and back (B region) were combined into one (end region) and the cases whose impact location is other than the front and back are considered as impact happening at the center. It is PAL 6

7 observed that 3% and 9% of AIS3+ injuries were happening at the center and end regions respectively. % of AIS3+injuries Figure 13. Classification of impact location into two: 1) center and 2) end regions A univariate binary logistic regression test was carried out to test the significance of being impacted at center than at the two ends away from A and C pillars. As shown in Table 4, a statistical significance (p<.5, chi square test) is found between the AIS3+ injury (dependent variable) and impact location (independent variable). The results also show that the odds ratio between the impact at center and impact at two end locations is 6.8. Hence, impact at the side center location is 6 times as likely as impact at vehicle side end locations leading to AIS3+ injuries. Impact locations at the side front (away from A pillar) and the side end (away from C pillar) regions may be safer than that at side central region (close to driver seating position) to cause comparatively less serious AIS3+ injury in near side collisions. Table 4. Logistic regression test AIS3+ Wald Value Chiprediction Square *p > Chi² Odds ratio Constant <.1 (reference) Statistically Significant *p<.5 Verification of Serious Injury Sensitivity by FE Simulations This section discuss the results of AIS3+ injury sensitivity obtained from a series of full vehicle FE simulations. The above real world accident results were also verified with barrier to car digital simulation. For the FE simulation, a Movable Deformable Barrier (MDB) is used to hit the passenger vehicle fitted with AM5% dummy in driver seating position. This test is (was) done with an impact angle of 27 degrees (9 o clock) and with an impact speed of 62 km/h. The MDB hits the vehicle at different impact locations as shown in Figure 14. First, it was hit at the center (close to the driver seating position) and injured values were measured. This injury value were used as a reference to compare with those of other locations. Barrier impact position was changed by 25 mm of several increments for both forward and rear side of the center reference position. Figure 14. FE simulation test configurations with center (base): SINCAP As shown in Figure 15, the chest injury values decrease as the closest impact edge of the barrier moves away from driver. There is a significant reduction of injuries (4%) as the impact location moves away from the center to end regions. To determine the probability of AIS3+ injury at different impact locations, the US-NCAP side impact chest injury risk curve for dummies, as mentioned in Equation 1 [15], has been considered here. All of the serious injury probabilities were calculated and shown in Figure 16. It is observed that the probability values of AIS3+ injuries are decreasing rapidly as the impact location moves away from the center, the reference position. But the injuries at 25 mm and 5 mm were increasing because they are very close to B pillar (driver s seating position). Both accident analysis and FE simulation results confirm that higher number of injuries at center and lower number of injuries at end locations. Hence, from driver s seating position and injury occurrence point of view, the impact PAL 7

8 locations in F (front end) and B (back end) regions are safer than those at P and Y (at center) regions. Figure 15. Chest injury sensitivity w.r.t location of impact (base: SINCAP) (AACN). It is discussed in detail in reference [16]. So adding an interaction terms to a model drastically changes the interpretation of all of the coefficients [16, 17]. If there were no interaction terms, the angle of impact would have unique effect of ISP. But the interaction means that the effect of each variable on ISP is different for different values of other independent interacting variables such impact locations, gender and others. LIMITATIONS P (Chest) (AIS3+) = Figure 16. Probability of AIS3+ chest injury w.r.t location of impact (base: SINCAP) 1 (1+exp( max.rib deflection in mm)) (1) Please note that all the above mentioned results were verified for only PV struck vehicles in C2C intersection accidents but not for other vehicle types. A limited number of cases were studied in this research work. However, considering all possible accident scenarios, more detailed verifications are needed by using various combinations of physical C2C experiments and simulations using different dummies and types of vehicles in order to make any generalized statement as stated above. It is also necessary to do similar accident analysis for other countries for verification. CONCLUSION DISCUSSION Similar injury reduction is observed in other body regions (abdomen and pelvis) and the results are mentioned in the Appendix. Injury sensitivity is also verified with 1 and 8 o clock impact angles. All of the injury values decrease quickly as the impact location changes from center to end. Please note that only 9 o clock impact angle FE simulation results are explained in detail in this paper but not the results of other angles. Similar changes in injury patterns with respect to different impact locations were observed when the AM5 was replaced with the AF5 dummy in IIHS test condition. This was done to verify the changes in injury patterns in both average males and shorter females. Since the serious injury outcome is changing with position of impact location, the results of this analysis provides some indications how to consider different interaction terms between impact locations, impact directions, gender, height etc., in improving the Injury Severity Prediction (ISP) algorithm related Advanced Automatic Collision Notification This paper discussed the sensitivity of the injury patterns of C2C side impact accidents at intersection for PV vehicles using NASS CDS CY data. The following specific conclusions can be drawn from this study:- a) More than 8% of C2C side crash intersection accidents are driver-alone cases and within that population, female are more likely to have AIS3+ injuries than male drivers. Shorter females (AF5) are most likely to have AIS3+ injuries. Further detail study with more accident data is necessary to identify the effect of gender difference (male and female) for shorter populations. b) SUVs produces a higher number of AIS3+ injuries than PVs as the striking vehicle. AIS3+ injuries in head and lower extremities (including pelvis) injuries increases when the striking vehicle is an SUV, when compared to PV. But there is not much change observed in thorax region. c) A higher number of AIS3+ injuries occur in side distribution Y location than the side center P location. Crashes which occur at a 1 o clock impact yield a higher number of AIS3+ injuries than at the 9 o clock impact angle in both P and Y regions. Injury PAL 8

9 probability occurance with impact at the center is 6 times more than the end regions(f&b). The level of injuries decreases rapidly as the closest impact edge of the striking vehicle moves away from the side center location to the side-end locations in barrier to car simulations. REFERENCES [1] National Highway Traffic Safety Administration (NHTSA), National Automotive Sampling System (NASS) Crashworthiness Data System Analytical User s Manual 214, National Centre for Statistics and Analysis, U.S. Dept. of Transportation, Washington, DC. [2] Helena, S., et al, 211. "Effect of Side Impact Protection in Reducing Injuries 22th ESV, Paper Number (211). [3] Tomiji, S., et al, 211. "Study on Car-To- Car Side Impact 22th ESV, Paper Number 289 (211). [4] Randa, R.S., et al, 23. "NHTSA Side Impact Research: Motivation for Upgraded Test Procedures ESV, Paper Number 492 (23). [5] Matthew, L Occurrence of Serious Injury in Real-World Side Impacts of Vehicles with Good Side-Impact Protection Ratings IIHS, Ruckersville, Virginia. Traffic Injury Prevention 16.S125-S132, (215). [6] Kuppa S., et al, 23. Development of side impact thoracic injury criteria and their application to the modified ES-2 dummy with rib extensions (ES-2re) Stapp Car Crash J. (23), 47: [7] Narayan, Y., et al, 28. Chest Deflections and Injuries in Oblique Lateral Impacts Traffic Injury Prevention (28) 9: , 28. [8] Pintar, F.A., et al, 29. Test program to define oblique chest loading in side impact Paper presented at: Experimental Safety of Vehicles; June 6 11, 29; Washington, DC. [9] Shaw, J.M., et al, 26. Oblique and lateral impact response of the PMHS thorax Stapp Car Crash J. (26), 5: [1] Cicchino, J. B. et al, 217. Effectiveness of Forward Collision Warning Systems with and without Autonomous Emergency Braking in Reducing Police-Reported Crash Rates Accident Analysis and Prevention, Feb (217). [11] Fildes, B. et al, 215. Effectiveness of low speed autonomous emergency braking in real-world rear-end crashes Accident Analysis and Prevention, Vol. 81, Pages 24-29, (215). [12] Xinghua, L Impact direction effect on serious-to-fatal injuries among drivers in near-side collisions according to impact location: focus on thoracic injuries Accident Analysis Prevention 48, page , (212). [13] David G. Kleinbaum Mitchel Klein, 21. Logistic Regression, A Self Learning Text Third Edition. ISBN: (21). [14] XLSTAT User s Guide, version, (215). [15] Safety Companion 216, CARHS publications [16] Chinmoy Pal, et al Identification of Factors Influencing Injury Severity Prediction (ISP) in Real World Accident Based on NASS-CDS, Int. JSAE Vol.6, pp (215). [17] Chinmoy Pal, et al Improvement of injury severity prediction (ISP) of AACN during on-site triage using vehicle deformation pattern for Car to Car side impact, short communication, AAAM 217. [18] Chinmoy Pal, et al Development of High Efficiency Load Path Structure to Enhance Side Impact Safety Performance, short communication, ESV215 Paper Number NOMENCLATURE: NASS CDS: National Automotive Sampling System Crashworthiness Data System AIS: Abbreviated Injury Scale (1998 version) AM5: 5th percentile American male AACN: Advanced Automatic Collision Notification ISP: Injury Severity Prediction AF5: 5th percentile American female IIHS: International Institute of Highway Safety NCAP: New Car Assessment Program PAL 9

10 PV: Passenger Vehicle SUV: Sport Utility Vehicle C2C: Car to Car Pelvis injury % APPENDIX A. Table A1. Weighted count of AIS injuries with respect to different impact locations and impact angles (Percentage of injuries mentioned in bracket) AIS Count Number (%) Impact Location And Impact Angle B D F P Y Z Total Minor (.2) (.29) (.32) (.52) 81 (.3) 541 (.2) 652 (.2) 221 (.1) (3.3) (1.55) (1.95) (1.43) (1.32) (5.5) (1.4) (3.69) (8.7) (1.5) (11.19) (2.78) (1.67) (4.92) (7.1) (73.35) 554 (.18) 5526 (.18) 369 (.12) 4329 (.14) (.51) (.9) (.97) 7821 (.26) (1.4) (4.52) 1173 (.36) 2259 (.74) (.7) (12.21) 3 81 (.3) (.1) (1.8) 166 (.5) (.15) 232 (.7) 1291 (.42) 242 (.67) (.77) 3471 (.11) 4168 (1.34) (5.37) 4935 (.16) 8617 (.28) 8921 (.29) (1.78) Serious (.4) 2352 (.8) 391 (.1) (.2) (.1) 165 (.1) 2371 (.8) 2371 (.8) 6197 (.2) 439 (.1) (.37) 1373 (.34) 2678 (.9) 848 (.3) 4588 (.15) 771 (.3) (.86) 935 (.31) 1581 (.5) 3463 (.11) 757 (.2) 4199 (.14) 5322 (.17) 236 (.7) 7275 (2.38) (1.19) (.4) (.1) 484 (.2) (.2) 2793 (.9) (.3) (.1) (1.54) 1997 (.6) (.16) 2527 (.9) (.58) 2738 (.88) (1.48) 6997 (.23) (1.52) 2232 (6.55) 1463 (.34) (.44) (.55) (14.55) AIS (.2) (.31) (.34) (.54) (6.12) (2.2) (1.61) (1.61) (5.28) (2.12) (5.47) (1.52) (1.53) (13.75) (31.85) (2.38) (6.1) (8.26) AIS3+: 14.5% injuries, AIS1&2: 85.5% injuries 15% 1% 5% % Base Figure A1. Pelvis injury sensitivity with respect to location of impact PAL 1

11 P(Abdomen) (AIS3+) Abdomen injury % P(pelvis) (AIS3+) 8% 6% 4% 2% % Base Figure A2. Probability of AIS3+ pelvis injury with respect to location of impact 15% 1% 5% % Base Figure A3. Abdomen injury sensitivity with respect to location of impact 4% 3% 2% 1% % Base Figure A4. Probability of AIS3+ abdomen injury with respect to location of impact Equation 2 shows the probability of pelvis AIS3+ injury P pelvis (AIS3+) = 1 (1+exp( pelvis force in N)) (2) Equation 2 shows the probability of abdomen AIS3+ injury P abdomen (AIS3+) = 1 (1+exp( Abdomen force in N)) (3) PAL 11

12 CLARIFICATION OF PRIORITY FACTORS FOR REDUCING TRAFFIC ACCIDENT FATALITIES IN THE U.S. AND BENEFIT ESTIMATION OF AEB SYSTEM FOR ONCOMING VEHICLES Takashi, Hasegawa Hiroyuki, Takahashi Satoshi, Udaka Toyota Motor Corporation Japan Paper Number ABSTRACT Until recently, most joint initiatives between the government and industry in the U.S. to help reduce traffic accident fatalities have focused on enhancing passive safety, such as through the safety assessment ratings of the New Car Assessment Program (NCAP) and the like. In addition, active safety technologies, such as automated emergency braking (AEB) and lane departure warning (LDW) systems, have also started to become more widespread. This paper describes a study that estimated the safety performance of recent vehicle models by analyzing U.S. traffic accident databases. This data was then used to estimate the benefit of the safety systems in these vehicles and to guide the development of the next-generation of safety technology. The fatality rate of each collision mode in recent vehicles was compared by analyzing data from the Fatality Analysis Reporting System (FARS), and the key contributing factors of fatal collisions were analyzed and prioritized using data from the National Automotive Sampling System Crashworthiness Data System (NASS-CDS). The estimated fatality rate reduction in all collision modes was identified by analyzing FARS data. A relatively large reduction in fatalities caused by side collisions and accidents not involving a collision (mainly rollovers) was found. The collision mode with the highest fatality rate for recent vehicles was frontal vehicle-to-vehicle (VTV) collisions. The most significant contributing factors in fatal frontal VTV head-on collisions were the advanced age of the driver, intrusion of the steering post into the cabin, high delta-v, and high occupant weight. The last three factors are particularly seen in high kinetic energy collisions. Although LDW systems are expected to help reduce frontal VTV head-on collisions, many collisions occur while the vehicle is negotiating a curve (not yet evaluated in NCAP tests) or while the other vehicle is traveling in the opposite direction in the same lane as the driver s vehicle. The benefit of LDW systems in such scenarios is difficult to estimate. Therefore, as a supplement to LDW systems, AEB systems for oncoming vehicles are being considered to help reduce kinetic energy in frontal VTV collisions. This study estimated the potential fatality rate reduction using the relationship between velocity and the probability of a fatality occurring in a frontal collision. To enable safe activation of AEB in response to an oncoming vehicle, additional analysis of field data will be required to ensure that the system does not interfere with the normal operation of the driver. However, if this can be achieved, an AEB system for oncoming vehicles may help to reduce fatalities in traffic collisions in the U.S. Consequently, this paper also estimates the approximate benefit of this system. Hasegawa 1

13 INTRODUCTION The U.S. has adopted a dual approach to help reduce fatalities and injuries caused by traffic accidents through both active and passive safety initiatives. While organizations such as the government and the Insurance Institute for Highway Safety (IIHS) are introducing regulations and safety assessment ratings, industries are working to develop and encourage the use of safety technologies. Active safety technologies already in widespread use include antilock brake systems (ABS) and electronic stability control (ESC) systems. More recently, automatic emergency braking (AEB) systems targeting vehicles and pedestrians in front of the driver s vehicle, as well as lane departure warning (LDW) and lane keeping support (LKS) systems that help drivers to maintain lane discipline have also been launched on the market. The objective of these systems is to help mitigate damage caused by a collision (AEB), and to help prevent the vehicle from leaving the lane unsafely or from driving off the road altogether (LDW and LKS). As these systems become more widespread and functionally capable, the number of such accidents, as well as the resulting number of fatalities and injuries, may well decrease. At the same time, advances in passive safety performance are being stimulated by the establishment of injury criteria for frontal and side collisions. These criteria have helped to encourage wider use of airbags and other restraint systems, as well as optimized vehicle body deformation characteristics. In recent years, the IIHS has introduced the small overlap (SOL) crash test, and the National Highway Traffic Safety Administration (NHTSA) is currently considering introducing a new frontal oblique test. The adoption of tests such as these has the potential to help reduce fatalities and injuries under a wider range of real-world collision conditions. Another recent trend is the wider adoption of an integrated approach to enhancing safety by activating and controlling passive safety restraint devices using information from active safety sensors that monitor the situation around the vehicle, which are being installed on more and more vehicles [1][2]. Statistics for 213 showed that passenger vehicle occupants accounted for the highest proportion of traffic accident fatalities in the U.S., and that frontal collisions were responsible for the most fatalities [3]. Research has already estimated the potential benefit of LDW systems in these accidents [4]. Therefore, this paper focuses on estimating the benefit of AEB systems in frontal collisions as a supplement to LDW and LKS systems. First, this study analyzed the fatality rate for recent vehicle models, and estimated the fatality breakdown once these latest models become more widespread based on 213 data in the Fatality Analysis Reporting System (FARS). This approach was used to confirm the proportion of frontal collisions in the total number of fatal accidents. Furthermore, data from the National Automotive Sampling System Crashworthiness Data System (NASS-CDS) was used to confirm the details of these frontal collisions from both active and passive safety standpoints to help analyze the potential benefit of an AEB system for oncoming vehicles. Next, this study estimated the relationship between changes in collision velocity and fatality risk in vehicle-to-vehicle (VTV) frontal collisions. Finally, based on these results, this paper discusses the potential reduction in fatality rate and future benefit of AEB systems for oncoming vehicles. Through these estimations and analyses, this paper identifies a potential relationship between the use of AEB to reduce collision velocity and passive safety performance in a broad sense, as well as to help enhance active safety performance. Further motivation for research was provided by this study, which identified the potential of AEB for oncoming vehicles as a promising system for the future. FRONTAL COLLISION RESEARCH This section discusses the potential benefit of AEB for oncoming vehicles based on accident analysis from the standpoints of passive and active safety. Breakdown of Fatalities after Popularization of Recent Vehicle Models This section uses the fatality rate for each model year (213 FARS data) and the difference between the fatality rates of belted and unbelted occupants in frontal collisions (25 to 213 NASS-CDS data) to estimate the fatality rate breakdown of collisions involving the most recent vehicle models, assuming that the occupants are wearing seatbelts. First, the current fatality breakdown in 213 FARS data was analyzed. Tables 1 to 3 show the data classifications, and Figs. 1 and 2 show the results. Figure 1 indicates that occupants of passenger vehicles or light trucks (LTVs) accounted for 6% of the total number of fatalities over the whole of the U.S. (19,756 out of 32,719), and that frontal collisions were responsible for the most fatalities Hasegawa 2

14 (11,33, 35%). Figure 2 categorizes these frontal collisions into single-vehicle collisions, VTV collisions, collisions fatal for front seat occupants, and collisions fatal for occupants in other seats. The two largest categories accounted for roughly the same number of fatalities: front seat fatalities in singlevehicle collisions (5,284, 16%) and front seat fatalities in VTV collisions (5,12, 16%). FARS 213 N=32719 people Noncollision P.Car/L.Truck, 272, 6% Fr. Collision P.Car/L.Truck, 1133, 35% Frontal Collision Frontal Collision P.Car/L.Truck Frontal Collision Motorcycle Frontal Collision H.Truck/Bus/Unknown Side Collision P.Car/L.Truck Side Collision Motorcycle Side Collision H.Truck/Bus/Unknown Rear Collision P.Car/L.Truck Rear Collision Motorcycle Rear Collision H.Truck/Bus/Unknown Noncollision (Single R/O etc.), P.Car/L.Truck Table 1. Classification codes of occupant and vehicle type PER_TYPE BODY_TYPE 1, 2, 9 Motorist 1-1, 17 Passenger cars , 19-22, Pedestrian 28, 3-32 Light trucks 6, 7 Cyclists 39, 4, 48, 49 3, 4, 8, 6-64, 66, 67, Other 1 72, 78, 79 Large trucks 5-52, 55, 58, 59 Buses 12, 42, 65, 73, 9-99 Other/ unknown 8-89 Motorcycles Table 2. Classification codes of collision type IMPACT 1 1, 11, 12 Frontal collision 2-4, 8-1, 61-63, Side collision 5-7 Rear collision 13, 14, 18, 19 Other Non-collision 98, 99 Unknown Table 3. Classification codes of frontal and side collisions VE_FORMS Frontal collision SEAT_POS 1 Single 11, 13 Fr Seat Side collision SEAT _POS IMPACT VTV Other Other , Near 2-4, side , Far side 8-1, Side Collision P.Car/L.Truck, 5321, 16% Rear Collision P.Car/L.Truck, 16, 3% Noncollision (Single R/O etc.), Motorcycle Noncollision (Single R/O etc.), H.Truck/Bus/Unknown Cyclist Pedestrian Other/Unknown Figure 1. Proportion of fatalities by occupant, vehicle, and collision type. VTV, Fr Seat, Far Side 97 VTV, Fr Seat, Near Side 211 Single, Fr Seat, Far Side 727 Single, Fr Seat 5284 Frontal Collision VTV, Fr Seat 512 Single, Fr Seat, Near Side 971 P.Car/L.Truck, Single, Fr Seat P.Car/L.Truck, Single, Rr Seat/Other P.Car/L.Truck, VTV, Fr Seat P.Car/L.Truck, VTV, Rr Seat/Other P.Car/L.Truck, Single, Fr Seat, Near Side P.Car/L.Truck, Single, Fr Seat, Far Side P.Car/L.Truck, Single, Rr Seat/Other P.Car/L.Truck, VTV, Fr Seat, Near Side P.Car/L.Truck, VTV, Fr Seat, Far Side P.Car/L.Truck, VTV, Rr Seat/Other Figure 2. Proportion of fatalities by singlevehicle or VTV collisions and seat position in frontal and side collisions. Next, the study estimated the fatality rate per model year for the main accident patterns involving passenger vehicle occupants. The fatality rate was calculated by dividing the estimated number of vehicles on the road by the number of fatalities. Figure 3 shows the number of fatalities per model year for frontal and side collisions and fatal accidents that did not involve a collision (non-collision accidents such as single-vehicle rollovers). Figure 4 shows the number of vehicles on the road per model year. This number of vehicles was estimated by multiplying the number of vehicles registered each year by the rate that vehicles tend to remain in use over time, which was identified in previous research [5]. Based on this approach, the fatality rate per 1, vehicles was calculated for each model year. Figure 5 shows the results. The thin lines show the fatality rates for each model year and the bold lines show the moving average over two-year periods. Hasegawa 3

15 The fatality rate for all types of accidents is decreasing for newer model years. This analysis identified the relationship of the fatality rate with each model year and the latest two model years (average of 212 and 213). Fatalities [People] Figure 3. Distribution of fatalities by model year. Estimated Number of Remaining Vehicles [Million Vehicles] Figure 4. Estimation of number of remaining vehicles of each model year. Fatality Rate per 1, Vehicles ¹ Model Year Passenger Cars Light Commercial Vehicles Total Frontal, Single, Fr seat Frontal, VTV, Fr Seat Side, Single, Fr seat, Near side Side, Single, Fr seat, Far side Side, VTV, Fr seat, Near side Side, VTV, Fr seat, Far side Noncollision (Single R/O etc.) Model Year Model Year Frontal VTV Frontal Single Side Single Near Side Single Far Side VTV Near Side VTV Far Unknown Noncollision (single R/O) Next, this study estimated the difference between the fatality rates of belted and unbelted occupants in frontal collisions using 25 to 213 NASS-CDS data. The data was filtered as shown in Table 4 and classified as shown in Table 5. After removing the results in which the belted state of the occupants was unknown, the number of data items (all injuries) was 2,223 people for single-vehicle collisions and 2,63 people for VTV collisions. Figure 6 shows the proportion of fatalities in the total number of injuries. These results indicate that the fatality rate of both single-vehicle and VTV collisions was approximately 8% for unbelted occupants and approximately 2% for belted occupants. Table 4. Filtering codes of frontal single-vehicle and VTV collisions Frontal single Frontal VTV TOWPAR 1 TOWPAR 1 VEHFORMS 1 VEHFORMS 2 EVENTS 1 EVENTS 1 ROLLOVER ROLLOVER 1, 11, 12, 21, 1, 11, 12, 21, 31, 32, 41, 31, 32, 41, DOF1 52, 61, 71, 72, 81, 91, 92 DOF1 GAD1 F GAD1 F SEATPOS 11, 13 Opposite Veh. DOF1 Opposite Veh. F GAD1 SEATPOS 11, 13 52, 61, 71, 72, 81, 91, 92 1, 11, 12, 21, 31, 32, 41, 51 52, 61, 71, 72, 81, 91, 92 Table 5. Classification codes of belted and unbelted occupants MANUSE 4, 14 Belted, 2, 3, 5, 8 Unbelted/Other Figure 5. Fatality rate per 1, vehicles in each model year. Hasegawa 4

16 Fatality Rate (Fatal/All Injuries) 12% 1% 8% 6% 4% 2% % Single Collision VTV Collision 2.5% 2.1% Belted (Lap and Shoulder) 8.3% 7.9% Unbelted/Other VTV, Fr Seat Near Side, Residual 91 Single, Fr Seat, Residual 137 Reduction by improved safety performance of recent models Frontal Collision -398 Reduction by belted -879 VTV, Fr Seat, Residual Reduction by seatbelts -746 Frontal Residual Collsion P.Car/L.Truck, Single, Fr Seat, Residual Frontal Reduction Collsion by seatbelts P.Car/L.Truck, (Fr. collision only) Single, Reduciton by belted Frontal Reduction Collsion by improved P.Car/L.Truck, safety performance Single, of Reduction recent models by penetraion of recent MY The other patterns are the same as Fig. 2. Figure 6. Comparison of fatality rate between belted and unbelted occupants. Finally, the number of fatalities was estimated for the most recent vehicle models assuming that the occupants are wearing seatbelts using 213 FARS data. Figure 7 shows the estimated results for frontal collisions. The reduction rate used the results of Figs. 5 and 6. Additionally, Fig. 8 shows the estimated number of fatalities involving the most recent models for other collision types. These results indicate that frontal collisions are responsible for more collisions than other collision types. According to this data, the number of front occupant fatalities in single-vehicle and VTV collisions was 1,37 and 2,38, respectively. Fatalities [people] Actual Results FARS Assumption of 12/13MY Fatality Rate Frontal, Single Fr Seat Belted ->Unbelted Assumption of 1% Belted Actual Results FARS213 Unbelted/Other Belted (Lap and Shoulder) Unknown Assumption of Assumption of 12/13MY 1% Belted Fatality Rate Frontal, VTV Fr Seat Belted ->Unbelted Figure 8. Proportion of fatalities by occupant type, assuming all vehicles are from most recent model years and all occupants are belted. Situation Immediately before Frontal VTV Collision In this section, 213 FARS data is used to identify the situation immediately before a collision, focusing on frontal VTV collisions based on the results in Fig. 8. The data was filtered as shown in Table 6. Data items were analyzed from a total of 2,88 people. Table 7 lists the pre-collision states and Fig. 9 shows the results. Table 6. Filtering codes of frontal VTV collisions INJ_SEV 4 Fatal injury PER_TYP 1, 2, 9 Motorists BODY_TYP 1-49 P. Cars/LTV IMPACT1 1, 11, 12 Front VE_FROMS 2+ VTV SEAT_POS 11, 13 Front seat MOD_YEAR Later than 98'MY AIR_BAG 1, 8, 9 Deployed REST_USE 3,12 Lap and shoulder belted Figure 7. Estimation of fatality reduction in frontal collisions, assuming all vehicles are from most recent model years and all occupants are belted. Hasegawa 5

17 Table 7. Classification codes of pre-collision behavior P_CRASH1 P_CRASH2 1, 2, , 2, 3, 5 62, 63 5, 51, 52 ALL , 63 6, 15, ALL 54 ALL 1-9 Other Combinations 49, 2.3% FARS 213 N=288 people 369, 17.7% 21, 1.1% 254, 12.2% 48, 2.3% 166, 8.% 397, 19.% 171, 8.2% Going straight and driver's vehicle departs from lane Going straight and other vehicle encroaches into lane Other vehicle in lane and traveling in same direction Negotiating a curve and driver's vehicle departs from lane Negotiating a curve and other vehicle encroaches into lane Overtaking, changing lanes, or merging Other vehicle in lane and traveling in opposite direction Driver's vehicle loses control 424, 2.3% Other Going straight and driver's vehicle departs from lane Going straight and other vehicle encroaches into lane Negotiating a curve and driver's vehicle departs from lane Negotiating a curve and other vehicle encroaches into lane Overtaking, changing lanes, or marging Other vehicle in lane and traveling in opposite direction Other vehicle in lane and traveling same direction Driver's vehicle loses contorol Other/Unknown Figure 9. Proportion of fatalities by pre-collision behavior in frontal VTV collisions. Table 8 adds classifications based on the presence of evasive driving maneuvers by the driver and vehicle stability. Figure 1 shows the results. It should be noted that there is no pre-collision evasive maneuver or stability data when considering cases in which the other vehicle departed from its lane. Table 8. Classification codes of pre-collision evasive maneuvers and vehicle stability P_CRASH3 1 1 Other Combination 26, 1.2% P_CRASH4 FARS 213 N=288 people 294, 14.1% 157, 7.5% 341, 16.4% 185, 145, 8.9% 39, 14, 6.9% 1.9% 6.7% 365, 17.5% No avoidance maneuver and no skidding Other No maneuver and no skidding Other The other patterns are the same as Fig. 9. Figure 1. Proportion of fatalities by pre-collision behavior in frontal VTV collisions. Therefore, it was assumed that these results were the same as for cases in which the driver s vehicle departed from the lane. Current safety assessment ratings of AEB systems simulate rear-end collisions with the vehicle in front. The proportion of rear-end collisions when the driver s vehicle was stable and no evasive maneuvers were carried out was 7.5%. Although safety ratings in the U.S. have yet to introduce assessments of LKS, LDW assessments are currently carried out on straight roads. The proportion of lane departures on straight roads when the driver s vehicle was stable and no evasive maneuvers were carried out was 16.4%. The proportion of lane departures on straight roads when the other vehicle was stable and no evasive maneuvers were carried out was 17.5%. In addition to lane departures, other common cases when the driver s vehicle was stable and no evasive maneuvers were carried out include when the other vehicle is traveling toward the driver s vehicle in the same lane (8.9%), when the driver s vehicle departs from its lane while negotiating a curve (6.9%), and when the other vehicle departs from its lane (6.7%). Hasegawa 6