Rollover Scenarios in Europe

Transcription

1 77 J. Gugler, H. Steffan Graz University f Technlgy, Institute n Mechanics, Graz, Austria G. Lutter, S. Fleischer Delphi Deutschland GmbH, Electrnics & Safety, Vehicle Safety Labratry, Berlin, Germany Rllver Scenaris in Eurpe Abstract Rllver accidents seem t be a rising prblem in Eurpe and therefre the systematic f this accident scenari shuld be investigated. Based n statistical investigatins n majr Eurpean accident databases fr different cuntries a series f 73 real wrld rllver accidents was analysed. These cases were recnstructed using PC-Crash and preliminary categrised using a mdified USbased rllver classificatin. In a first step, the rllver events were recnstructed frm the pint f cnflict t the vehicle s rest psitin. The vehicles kinematics as well as its linear and rtatinal velcities were derived. In a secnd step typical velcity characteristics as well as kinematics were identified and the events categrised accrding t these criteria. Based n these results fur main categries were defined, cvering all recnstructed accidents. This categrisatin was based n mechanical parameters (rtatry and translatry kinematical data f the vehicle). Significant differences can be seen fr different scenaris fr the first phase f rllver. Ntatin ν change in velcity due t an impact ϕ rll angle ϕ& rll rate Intrductin Rllver accidents are ne scenari happening t passenger vehicles. This scenari was nt investigated n an Eurpean level. Investigatins mainly in the US have been dne n this accident scenari and a classificatin based n these investigatins was derived. Fr the US classificatin eight different scenaris are knwn by ASIC [1], based n typical sequences in the case f rllver. This scenaris are: Trip-ver when the lateral mtin f the vehicle is suddenly slwed r stpped inducing a rllver. The ppsing frce may be prduced by a curb, pthles, r pavement dug int vehicle wheels. Flipver when the vehicle is rtated alng its lngitudinal axis by a ramp-like bject such as a turned dwn guardrail r the back slpe f a ditch. The vehicle may be in yaw when it cmes in cntact with a ramp-like bject. Bunce-ver when a vehicle rebunds ff a fixed bject and verturns as a cnsequence. The rllver must ccur in clse prximity t the bject frm which it is deflected. Turn-ver when centrifugal frces frm a sharp turn r vehicle rtatin are resisted by nrmal surface frictin (mst cmmn fr vehicle with higher centre f gravity (COG)). The surface includes pavement surface and gravel, grass, dirt, etc. There is n furrwing r guging at the pint f impact. Nte that if rtatin and/r surface frictin causes a trip, then the rllver is classified as a turn-ver. Fall-ver when the surface n which the vehicle is traversing slpes dwnward in the directin f mvement f the vehicle COG such that the COG becmes utbard f its wheels (Nte: The distinctin between this cde and flipver includes a negative slpe.). Climb-ver when the vehicle climbs up and ver a fixed bject (e.g. guardrail, barrier) that is high enugh t lift the vehicle cmpletely ff the grund. The vehicle must rll in the ppsite side frm which it apprached the bject. Cllisin with anther vehicle when an impact with anther vehicle causes the rllver. The rllver must be the immediate result f the impact between the vehicles. Fr example, this culd ccur at an intersectin where a vehicle is struck in the side and the mmentum f the struck vehicle results in a rllver. End-ver-end when a vehicle rlls primarily abut its lateral axis. Based n this scenaris it was investigated, if they are applicable fr Eurpean rllvers t. Fr chsing real wrld accidents fr in-depth studies, basic studies f the statistics were analysed and resulted with the fllwing characteristics fr rllver accidents. Statistical Analysis SFERCO et al. [2] fund ut in the German In- Depth Accident Study (GIDAS) and the Cperative Crash Injury Study (CCIS) in the UK had shwn, that rllver cunt fr 5-15% f all

2 78 accidents. Single rllver events, withut any multiple impact are events in Eurpe up t 5% f all accidents. This is third f all rllver accidents. Fr multiple rllver accidents the first event is the impact rather the rllver. S a rllver can be regarded as a cnsequence f an impact rather than an initiatr. Mst vehicle rllvers invlve ne cmplete rll r less and they ccur abut the lngitudinal axis f the vehicle, apprximately half in each directin. When an impact fllws an initial rll, it is frequently against a fixed bject (rather than a vehicle) and appears t randmly invlve all parts f the vehicle. In cases where rllver fllws an initial impact, the impacts are split between thse against cars and thse against fixed bjects. A disprprtinate number f the initial impacts is against the sides f the vehicle that rlls ver (rather than the frnts). Analysis frm the British natinal accident data (STATS 19) frm KIRK [3] shwed that 6% f all car casualties were injured in cars with an element f rllver and 12% fr killed and severe injured car ccupants (KSI). Of all cars that have a fatal ccupant r ccupants, 15.1% have an element f rllver. Fr cars that have an element f rllver, accidents that ccur whilst negtiating a bend are far mre cmmn than fr nn-rllver cars, althugh verall nrmal ging ahead accidents are mst cmmn. Fr cars that have an element f rllver, 77% are single vehicle events. Fr single vehicle crashes frm crashes with anther vehicle the mst cmmnly vehicle struck is anther car. Of all cars, 3.9% that have an injured ccupant have a rllver and d nt impact anther vehicle. Fr cars with killed and severe injured ccupants, frntal impacts are clearly the mst cmmn. A higher prprtin f vehicles that have an element f rllver leave the carriageway, fr KSI cars 81.9%. This als crrelates with an increased prprtin f bjects hitting ff the carriageway fr cars with an element f rllver, fr KSI, 67.7%. An increase in KSI rate is evident when the car leaves the carriageway. Fr cars with an element f rllver, the mst cmmnly struck bject ff the carriageway is a tree fllwed by entering ditches. The mst cmmn car rllver accident scenari is fr the vehicle nt t impact any ther vehicle r vehicles and t hit a fixed bject ff the carriageway and n bject in the carriageway, accunting fr 45.5% f all vehicles that have any element f rllver and an injured ccupant. Of all severity rllver cars, 18.9% have n ther vehicle impact r any cdeable impact with an bject n r ff the carriageway. SFERCO et al. [4] were lking n differences f rllver data fr US and Eurpe and it was shwn that rllvers, as a single event (rllvers withut the ccurrence f any impact) are rare events in Eurpe. FAY [5] fund that rllvers ccur mre frequently as a part f mre cmplex accident sequences invlving multiple impacts. In mst f these multiple impact cases, the first event in the sequence is an impact rather than a rllver. In the US, rllvers have been identified as a significant safety issue, because a rllver crash is far mre likely t result in fatalities than a nn-rllver. Althugh nly 3 percent f all passenger vehicles invlved in crashes in 2000 experienced rllver, 20 percent f passenger vehicles invlved in fatal crashes rlled. In particular, Sprt Utility Vehicles (SUV), Multi Purpse Vehicles (MPV) and ther light trucks are ver-represented in rllver accidents. Real Wrld Accidents fr In-Depth Studies Fr this investigatin a database cntaining abut 150 real wrld passenger vehicle rllver accidents was used. The strategy fr chsing these cases fr in-depth studies is based n the results f the statistical investigatins as well as the quality f dcumentatin f the cases. This cases were recnstructed numerically using the accident recnstructin sftware tl PC-Crash [6]. Methd Relevant Mechanical Parameters The PC-Crash recnstructin files f the recnstructins prvided mechanical data n tire side frces f all fur wheels, velcities in x-, y- and z-directin, the rll angle, the rll rate and the angular acceleratin as these seemed t be f imprtance fr detecting a rllver. After studying the prvided data it seemed prmising t further assess the imprtance f the rll rate fr categrisatin as it is used in state f the art technlgies. Fr the general analysis it was fcused n the rll rate and the rll angle as the relevant parameters. This fcus seemed plausible as a high rll rate at a lw rll angle might nt lead t a rllver whereas even a lw rll rate at a large rll angle with the centre f gravity nearly abve the wheels will cause

3 79 a rllver. There shuld be a direct interrelatin between the parameters rll rate and rll angle and a rllver case. In-depth Analysis f Relevant Cases General Analysis frm Pint f Cnflict t End f Rllver The first step f analysing the PC-Crash data was t plt the rll rate [deg/s] as a functin f the rll angle [deg]. As the different rllver cases analysed vary widely in their rll angle, the rll rate rll angle graphs shw very different patterns in the latter stages f the rll. The vehicle behaviur during the rlling phase (i.e. after the initial event) seems t happen at randm. As many graphs shwed similarities at lwer rll angles, the rll rate rll angle graphs were pltted frm the rll angle at the initial event (ϕ t=0 = 0 ) t a rll angle f ϕ = 90. This range als includes the relevant phase fr detecting a pssible rllver and fr triggering pssible safety systems. When cmpared, grups f these graphs (ϕ = 0 K 90 ) shwed distinctive similarities and culd be srted int categries. The mst bvius grup is frmed by cases with an impact. The graphs shw distinctive differences between cases withut impact r with impact. Categry 1: Rllver caused by sme kind f impact (ther vehicle, tree, r ther) The initial rll rate jumps t frm a high peak and rapidly decreases afterwards befre increasing again at rll angles f apprximately 45. If the impact is preceded by yawing and/r a sideway skid the graph may shw a γ -frm r encircle the centre f the crdinate system befre it shws the characteristic mentined abve. Categry 2: Rllver caused by ramp like bject (e.g. flat car, guard rail, slpe) The rll rate quickly rises t a high level but des nt decrease as significantly afterwards as in categry 1. The yaw angle remains at lw levels (less than 30 ). Fig. 1: Example fr basic analyse f whle rllver event up t 90 rll angle Fig. 2: Example fr categry 1

4 80 Categry 3: Rllver caused by yawing and skidding sideways with vehicle being affected by a ditch r slpe The yaw angle at the start f the rll actin differs widely (0 t ver 200 ) but n average seems t be lwer than in categry 4. Categry 4: Rllver caused by yawing and skidding sideways n an even surface The rll rate builds up and the rll angle increases a little until it reaches a cnstant value. The rll rate then decreases again as far as zer deg/s r belw. When the vehicle starts t rll the rll rate rises t a high level. The graph shws a picture resembling the Greek letter γ. Due t strng yawing the graph may encircle the centre f the crdinate system. The increase in the rll rate is slwer than in cases with impact. The yaw angle at the start f the rll actin is mstly in the range between 70 and 90. Fig. 3: Example fr categry 2 Fig. 4: Example fr categry 3 Fig. 5: Example fr categry 4

5 81 Categry 5: Rllver caused by ther causes The graphs f pitch-vers shw a very different characteristic in the rll rate rll angle diagram which des nt seem t be cmparable t the previus cases In-Depth Analysis fr First Phase f Rllver In a secnd phase the recnstructed cases were analysed frm the start f the rllver (as described belw) up t 90 degree rll angle. Fr this analysis the fllwing time-depending mechanical parameters were available fr the recnstructed cases: Rtatinal mtin: angle, angular velcity and angular acceleratin fr rlling, yawing and pitching (referenced t centre f gravity in a reference crdinate system) Linear mtin: linear mvement, linear velcity and linear acceleratin in x-, y-, z-directin (referenced t centre f gravity in a lcal crdinate system) Additinal: tire frces (side and nrmal) Fr rllvers different parameters were analyzed n their characteristics ver the rll angle. The fllwing parameters shw the mst significant influence fr categrisatin fr the first phase f rllver: Rll rate vs. rll angle Lateral velcity vs. rll angle (in a glbal crdinate system) Lngitudinal velcity vs. rll angle Due t the lng duratin f a rllver the characteristic becmes mre and mre randmized if the whle rlling phase is used. S the rllver is divided int 4 phases (see figure 6): 1. Pre-rll phase 2. Pint f n return 3. First phase f rll 4. Rlling phase Pre-Rll Phase The pre-rll phase is the phase when the vehicle is cming int a destabilized driving mde till the pint f n return where the rllver cannt be avided. In this phase active safety can be used t stabilize the vehicle and avid exceeding the pint f n return. The mre it seems t be unavidable t stabilize the car passive safety devices can als be preactivated in this phase. If pssible estimatin n the severity f the impending rllver shuld be dne. Pint f n Return This is nt really a time pint. It is mre a shrt time interval when the rllver cannt be avided and passive safety devices have t be activated t reduce the risk f injuries t ccupants. First Phase f Rll The first phase f rll starts frm the pint f n return and cvers apprximately the first 90 degrees f rll angle. It ends with the first impact f the vehicle structure with the grund. The car can always be in cntact with the grund r lse the cntact (flying phase). Rlling Phase The rlling phase is the phase frm the end f the first phase f rll until the vehicle's rest psitin. The mst imprtant parameter fr this phase is the number f turns. Fr the detailed analysis f the rllvers the first phase f rll was used and defined in a little Fig. 6: Phases f a rllver Fig. 7: Determinatin f first phase f rllver fr a 360 leftside rllver

6 82 Fig. 8: First phase f rllver fr a 360 left-side rllver mdified way. The start f this phase the pint f n return was defined as last significant zer f the rll rate vs. the rll angle. Due t the numeric data a threshld fr the zer f the rll rate f 0.05 rad/sec was used. Fr the end f the fist phase f rll the 90 degree rll angle criteria was used due t the difficulties in finding the first impact in the rlling phase. Results Based n the prcedure described all cases were analysed in detail fr their significant rll angle rll rate behaviur fr the first phase f rllver. Als the lngitudinal and lateral velcity characteristics were investigated and the fllwing results were gained. The main rllvers can be classified by the fllwing categries: Impact induced rllvers ν <30kph ν >30kph Ramp-like bject induced rllvers Skidding and Yawing Trip induced rllver Turning and rllver Others Impact Induced Rllvers Rllver accidents induced by any kind f impact (mstly with anther vehicle but als with ther bject). This type f rllver scenari is divided int tw sub-categries depending n the change f velcity ( ν) during the impact. Fr high ν values Fig. 9: Characteristics f rllver induced by an impact with ν <30kph fr different real wrld accidents the impact inducing the rllver is cnsidered as mre harmful event than the fllwing rllver. ν < 30kph Fr this scenari the ν fr the rlling vehicle is less than 30kph. This is based n the analysis f the real wrld accidents. figure 9 shws the characteristics fr this type f rllver. The initial rll rate jumps t frm a high peak caused by the initial impact. Fr increasing rll-angles the rll-rate decreases fast and then increases mderate. delta ν>30kph Rllver accidents with an ν higher than 30kph are cnsidered t have a severe frnt r side impact. Therefre it is necessary t activate the passive safety system fr this kind f impact. The fllwing rllver is nt as harmful as the initial impact. Ramp-like Object Induced Rllvers This type f rllver is induced by any kind f ramp-like bject. This culd be a guardrail, the end f a cncrete barrier as well as an embankment, slpe r the hd f an ppsite car acts like a ramp. As can be seen in figure 10 the rll rate rises quickly t a high level and stays nearly cnstant fr the increasing rll-angle. The analyse shws als that the lngitudinal velcity is high and the lateral velcity is n a lw level. Skidding and Yawing - Trip Induced Rllver This scenari happens when a car is tripping e.g. the tires are digging int gravel r sil. This is equal t a higher frictin acting in the tire-grund cntact and therefre a higher lateral frce can be

7 83 vehicle the car gets int a unstable mde and a rllver results. The frictin in the tire-grund cntact is nt increased as in case f tripping. Figure 12 shws that there is a significant initial scillatin in rll-rate caused by the unstable driving mde. The all ver behaviur is the same like in case f a tripped rllver. Others/Special Others Fig. 10: Characteristics f rllver induced by ramp-like bject Fr the remaining rllvers its nt easy t categrize them particularly as they are very rare events e.g. the end-ver-end rllver, where the rll-axis is lateral. Sme ther special cases are the free fall f a vehicle e.g. dwn frm a bridge r cases where the car is yawing and the back f the car is tripped when cntacting the sil n the rad side. Discussin The rllver cases chsen fr recnstructin represent the statistical results frm the survey laid ut in the intrductin as fllws: Fig. 11: Characteristics f a rllver induced by tripping Fig. 12: Characteristics f a rllver induced by turning btained. figure 11 shws that the rll rate increase mderate t a cnstant value. The lngitudinal velcity decreases t a cnstant value and the lateral velcity decreases rapid t a cnstant value. Skidding and Yawing - Turning and Rllver In this type f rllver the vehicle is nrmally driven n a rdinary surface. Due t the driven maneuvres and the dynamic characteristics f the Accrding t the STATS % f all rllvers are single vehicle events. Frm the 73 recnstructed cases 59 (81%) were single vehicle event cases which is the same prprtin as fund ut by SFERCO et al. [2] and KIRK [3]. 43% f the recnstructed cases have an initial impact befre the rllver which is a little less than FAY's [5] (58%) findings. In 12 recnstructed cases (16%) the vehicle impacts an bject ff the carriageway. Accrding t KIRK this prprtin is 46%. In mst rllver cases the vehicle turns arund its lngitudinal axis and makes 4/4 turns r less. The prprtin f rlls t the right r t the left is half/half. Accrdingly, f the recnstructed cases nly 5 vehicles (7%) turned arund their lateral axis. Of the thers 49 vehicles turned 4/4 r less (67%). In 16 cases (22%) the vehicles turned mre than ne turn (5/4 t 30/4). Fr 7 cases it was nt pssible t accunt fr the number f turns arund the lngitudinal axis as they were either pitch-vers r the recnstructin file did nt give enugh infrmatin. Of the vehicles turning arund their lngitudinal axis 32 turned t the right and 31 t the left. Accrding t GIDAS and CCIS analysis by SFERCO et al. [2] nly arund ne third f all rllvers ccur as single, islated events in the UK

8 84 and Germany. The remainders ccur during mre cmplex multiple impact crash sequences. 43 f the recnstructed cases had n impact and can be regarded as single islated events. This prprtin (59%) is abut twice as high as stated by Frd. Cnclusins The rllver categrisatin defined in this wrk can be used als by nn-prfessinal analysts fr precategrisatin f an accident. The rllver can be cmpared easily t the fur main categries: impact induced, ramp-like bject induced, skidding and yawing r thers. When recnstructing a real wrld accident a final classificatin can be dne when analysing the vehicle trajectry and its kinematical data. The selected real wrld accidents represent the statistical findings frm ther authrs. Prject 5 th Framewrk-Prgramme Grwth Prject Rllver G3RD-CT , P. FAY, R. SFERCO, R. FRAMPTON: Vehicle Rllver - An Imprtant Element in Multiple Impact Crashes, ESV 2003, Japan, PC-Crash User Manual: Dr. Steffan Datentechik Linz, Austria, 2003 Ackwledgement This study was perfrmed within the prject Imprvement f Rllver Safety fr Passenger Vehicles funded under the R&TD prgram Grwth by the Eurpean Cmmissin ( References 1 S. ASIC: Field Cnditins Definitins - NASS- CDS Classificatin, Presentatin, Delphi, Eurpean Cmmunity R&TD-Prject 5 th Framewrk-Prgramme Grwth Prject Rllver G3RD-CT , R. SFERCO, P. FAY: Vehicle Rllver in Eurpe - A Real Wrld Overview, Presentatin, Frd Mtr Cmpany, Eurpean Cmmunity R&TD-Prject 5 th Framewrk-Prgramme Grwth Prject Rllver G3RD-CT , A. KIRK: Great Britain Statistics, Reprt Wrk Package 1 Task 1.1, Vehicle Safety Research Centre, Eurpean Cmmunity R&TD-Prject 5 th Framewrk-Prgramme Grwth Prject Rllver G3RD-CT , R. SFERCO, P. FAY, S. ASIC: Cmparisn f US and Eurpean Rllver Data, Reprt Wrk Package 1 Task 1.1, Frd Mtr Cmpany and Delphi, Eurpean Cmmunity R&TD-