DYNAMIC ROOF CRUSH INTRUSION IN INVERTED DROP TESTING Stepen A. Batzer, P.D., P.E. Renfroe Engineering, Inc. United States of America Robert M. Hooker Eliseco Systems United States of America Paper Number 05-0146 ABSTRACT Inverted drop testing of veicles is a destructive determination of roof strengt used by industry, government organizations and independent engineers to determine veicle safety wit respect to rollover collision. In tis paper, te results of numerous inverted drop tests are summarized and analyzed, giving bot te amount of permanent and temporary roof crus tat occurs during impact. Only unmodified production veicles wit sound roofs were tested. Te amount of dynamic roof crus varied from a low of 0 to a maximum of 7.0 cm, te relationsip between elastic and plastic roof crus was not found to be statistically significant, and prediction intervals for A and B-pillar crus were developed. INTRODUCTION Many engineers believe tat strong roofs provide significant protection to occupants during rollover collisions. As of tis writing, te American National Higway Traffic Safety Administration (NHTSA) as opened docket #5572 regarding review of te tecnical metodology for certifying te roof strengt of passenger veicles. In tis docket can be found arguments in support of, and counterarguments dismissive of, te assertion tat stronger roofs (beyond a certain minimal point) are safer roofs. Tis paper addresses te lack of solid data regarding impact-generated dynamic intrusion into te occupant space. Tere is currently a lack of information regarding te dynamic intrusion of te roof structure into te occupant capsule as a result of rollover. Tis information as not been tabulated as a result of Family of Motor Veicle Safety Standards (FMVSS) 216 tests, and cannot be reliably measured from actual rollover collisions. In some cases, evidence of dynamic intrusion is present due to witness marks on eadrests and oter components during rollovers, but te actual intrusion distance still must be estimated rater tan measured. STATIC ROOF CRUSH TESTING Roof strengt is regulated in te United States by te FMVSS 216 standard, Roof Crus Resistance Passenger Cars, and was adopted on September 1, 1973. General Motors developed te procedure at teir researc laboratories. One reason tat it was adopted was for its repeatability, a desirable attribute for expensive, time-consuming tests. Te pre-amble of te FMVSS-216 standard states, Te purpose of tis amendment is to add a new Motor Veicle Safety Standard tat sets minimum strengt requirements for a passenger car roof to reduce te likeliood of roof collapse in a rollover accident (empasis added). As was alluded to in te introduction, tere is significant, ongoing controversy regarding roof crus as it relates to occupant injury. Certainly, if roof crus is an issue in occupant safety, it is immaterial as to weter or not te intrusion tat may or may not ave injured te occupant was temporary or permanent. DYNAMIC ROOF CRUSH TESTING Te quasi-static roof crus test mandated by te FMVSS 216 subjects te veicle to a maximum force significantly less severe tan would be applied to te veicle during a multiple rollover. Te Society of Automotive Engineers (SAE) recommended practice J996, Inverted Drop Test, is also a test of rollover craswortiness, and was developed by SAE in te late 1960s. Since it is a more severe test, numerous engineers prefer it to te quasi-static FMVSS 216 test. Batzer 1
Te SAE J996 test was designed, to obtain as closely as possible deformation of a veicle roof or roll bar structure wic occurs in a veicle rollover. In tis test, te subject veicle is inverted, given a roll angle, pitc angle, and drop eigt tat are representative of te assumed loading at rollover. Te angles present ensure tat te majority of potential energy is transferred directly to te A-pillar structure. Tis standard does not specify any crus measurement metodology, permanent or dynamic. DEVICE DESIGN AND TEST PROCEDURE A reusable telescoping rod assembly was designed and constructed to document dynamic crus. Te two rods are made of cold-rolled 4130 steel, approximately 56 cm in lengt, wit a 2 cm nominal inner diameter of te tin-wall ollow (female) upper rod, and a 2 cm nominal exterior diameter of te solid (male) lower rod. Te rods are not spring loaded, but free to move axially in extension and compression. Te rod ends are capped wit macined gimbals tat fit into bases to allow rapid re-orientation of te rod ends during testing tus preventing binding. Te orientation of te rod assembly inside of te veicle is suc tat it is perpendicular to te test pad as te veicle is inverted and ready to be dropped. Te driver s seat is removed or modified as necessary to accommodate rod mounting. Te top base is riveted into place at te root of te pillar / roof rail interface, and te bottom base is welded to te floor or seat structure. Once te device is in place, its installed lengt is measured, and a rubber o-ring is positioned at te exterior mating rim of te female rod. As te rod compresses during impact, te o-ring is displaced by te female rod. As te roof rebounds, te o-ring remains in place. By measuring te distance between te o-ring and te female rod, te amount of dynamic roof crus is determined. In some configurations of tis dynamic roof crus measurement device, an ink marker is affixed to te female rod and te tip bears against te male rod in order to provide furter visual documentation of relative rod travel. Tese two measurements were always found to be in agreement. Tus, tis simple device documents bot permanent (plastic) and dynamic (elastic) deformation of te roof. Te rod is examined for free travel before and after testing to ensure no binding as occurred. RESULTS AND STATISTICAL ANALYSIS A compilation of te drop testing results is given in Table 1, given in Appendix I. Measurements were made to te nearest sixteent of an inc, but ave been given in SI units to te nearest millimeter. Te amount of plastic intrusion for te A pillar varied from a low of 8.3 cm to a ig of 42.5 cm, wile te elastic varied from 0-6.4 cm. Te amount of plastic intrusion for te B pillar varied from a low of 3.2 cm to a ig of 40.6 cm, wile te elastic varied from 1.3-7.0 cm. Te average dynamic roof crus for bot pillars was found to be approximately 4.4 cm. Figure 1 sows te plastic roof crus plotted against te drop eigt for bot te A and B pillars. As can be seen, te amount of plastic roof crus is not strongly correlated wit drop eigt. Tese figures sow te effect of differing roof strengts across different veicle designs. Figure 2 sows two graps of elastic versus plastic roof crus for bot te A and B-pillars. Importantly, tere is no apparent trend linking te two different crus types. If least-squares regression lines were added to te plots, tey eac would sow only a very modest positive slope. Calculations reveal tat te confidence intervals on tese two slope magnitudes includes zero, meaning tat tere is no statistically significant relationsip between te two types of crus. Figure 3 sows tat A & B pillar plastic crus are strongly correlated. As expected, as te A pillar plastic crus increases, te B pillar residual crus also increases. Te A-pillar plastic crus was always measured to be greater tan tat of te B-pillar plastic crus, altoug sometimes te two values differ only sligtly. Te average difference between te measurement sites was found to be 5.0 cm. Te regression of te B pillar crus on to A-pillar crus is: Bˆ = 1.13A - 3.06 (1) were Bˆ is te predicted residual B-Pillar crus, and A is te measured A-pillar plastic crus. Te regression yields an R 2 = 0.958. Tis equation sows tat tere is an approximate 3 cm crus tresold for te A pillar to induce crus at te B pillar. As was sown in Fig. 2, te elastic and plastic crus intrusions are not strongly correlated. It is, owever, wortwile to construct a prediction interval for te amount of elastic intrusion tat is independent of te Batzer 2
plastic crus. Tat is, if anoter drop test were performed for a randomly selected FMVSS-216 compliant veicle, wat interval of elastic intrusion values would bracket te next measured value wit a 90% success rate? Tus, if it is sensible to model crus measurements from te population of FMVSS- 216 compliant veicles as normally distributed, one may use te sample means and sample standard deviations from Table 1 to state suc intervals predicting next measured values. Figure 4 sows Q- Q plots for te A and B pillar dynamic crus measurements. Te data appears sufficiently well beaved to use a standard prediction limit interval analysis. A 90% prediction interval on te elastic intrusion is made as follows [Vardeman and Jobe, 2001]: x ± t α s ν, 1-2 1 + 1 n (2) were ν = n-1, n = sample size, and α = 0.90. Tis yields two prediction intervals for te dynamic A- Pillar (3) and dynamic B-Pillar intrusion crus: 0 cm A 0.90 8.4 1.3 0.90 cm (3) cm B 8.1 cm (4) Te A-pillar dynamic intrusion is of greater consequence, as te front seats are more likely to be occupied, and te plastic intrusion of te A-pillar is usually greater tan tat of te B-pillar in rollover collisions. Te FMVSS-216 requires tat te veicle does not exceed 12.7 cm plastic intrusion during quasi-static testing. An 8.4 cm dynamic intrusion into te occupant survival space is a significant fraction of tis allowable plastic deformation level. CONCLUSIONS During rollover collisions, energy is dissipated at a relatively low rate, making tese events muc less severe from te point of view of te veicle tan are oter types of collision suc as frontal impact. Francini [1969] discussed te cras survival space wic needs to be maintained for occupant survival. Te volume of interior space enveloping te occupant represents te survival space, and takes into account te size, posture and position of te occupant. It is of principal importance in designing a veicle for craswortiness. An analysis of te testing presented in tis paper seds new insigt into te integrity of te occupant survival space during rollover collisions. It as been sown for te sample set presented tat te measured crus at te A pillar exceed tat at te B pillar, tat te dynamic crus averaged approximately 4.4 cm, and tat te amount of plastic and elastic crus are not strongly enoug correlated for te relationsip to be statistically significant for our sample size. Furter, a 90% prediction interval for te elastic intrusion of FMVSS-216 compliant veicles encompasses 0 8.4 cm at te A-pillar, and 0 8.1 cm at te B-pillar. ACKNOWLEDGEMENTS Te autors express teir sincere gratitude to Dr. Stepen B. Vardeman of te Iowa State University for is work in reviewing te statistical analysis presented in tis manuscript. REFERENCES [1] Federal Motor Veicle Safety Standards; Roof Crus Resistance, Federal Register, Vol. 66, No. 204, pp. 53376-53385. [2] Francini, E., Cras Survival Space is needed in Veicle Passenger Compartments, SAE 690005. [3] Grymes, Cristie L., and Collier Sannon Scott, Product Demonstrations, Doc. 567140, Ver. 1, 6/4/01. [4] SAE J996, Inverted Veicle Drop Test Procedure, August, 1967. [5] Vardeman, S. B., and J. M. Jobe, Basic Engineering Data and Analysis, Duxbury Tomson Learning, 2001. CONTACT INFORMATION Dr. Steve Batzer is te senior forensic engineer of Renfroe Engineering, Inc. He can be reaced at Renfroe Engineering, Inc., 13045 W. Hwy 62, Farmington, P.O. Box 610, AR, 72730, telepone 479-846-8000; fax 479-846-8002; email batzer@renfroe.com. Mr. Robert Hooker is te founder and president of Eliseco Systems, a Micigan-based automotive testing facility. He can be reaced at Eliseco Systems, 2046 Scool Road, Weidman, MI, 48893; telepone 989-644-8999; fax 989-6444-8990. Batzer 3
APPENDIX I DATA Make Model TABLE I: Plastic and dynamic roof crus measurements. Yea r Rol l ( o ) Pitc ( o ) Drop Heig t Plasti c Crus Dynami c Crus Total Crus Locatio n Ford Aerostar 1993 25 5 30.5 19.7 5.1 24.8 Ford Aerostar 1993 25 5 30.5 11.7 5.1 16.8 Ford Bronco 1984 25 5 II 30.5 32.4 6.4 38.7 Ford Bronco 1984 25 5 II 30.5 27.5 5.4 32.9 Ford F-150 1986 25 5 30.5 42.5 48.6 Ford F-150 1986 25 5 30.5 40.6 6.4 47.0 Honda Accord 1988 25 5 45.7 21.3 4.4 25.7 Honda Accord 1988 25 5 45.7 10.8 4.1 14.9 Hyundai Excel 1991 25 5 30.5 21.9 4.8 26.7 Hyundai Excel 1991 25 5 30.5 17.8 5.1 22.9 Isuzu Rodeo 1994 25 5 30.5 12.4 1.7 14.1 Isuzu Rodeo 1994 25 5 30.5 8.9 5.4 14.3 Nissan Pickup 1985 25 5 30.5 34.6 34.6 Nissan Pickup 1985 25 5 30.5 34.5 2.5 37.0 Plymout Laser 1992 25 5 30.5 10.2 6.4 16.5 Plymout Laser 1992 25 5 30.5 3.2 7.0 10.2 Pontiac Fiero 1986 25 5 45.7 8.3 3.2 11.4 Pontiac Fiero 1986 25 5 45.7 3.8 3.2 7.0 Subaru Loyale 1993 25 5 30.5 17.8 1.3 19.1 Subaru Loyale 1993 25 5 30.5 11.4 1.3 12.7 Suzuki Samurai 1988 45 0 91.4 22.9 6.4 29.2 Suzuki Samurai 1988 45 0 91.4 19.1 6.7 25.7 X 22.2 4.1 26.3 s 10.2 2.2 10.4 X 17.2 4.7 21.2 s 11.7 1.7 11.8 Batzer 4
APPENDIX II Statistical Analysis Graps Plastic Crus 5 4 3 2 1 2 4 6 8 10 Drop Heigt Plastic Crus 5 4 3 2 1 2 4 6 8 10 Drop Heigt Figure 1. Plastic crus versus drop eigt for te A-Pillar. Elastic Crus 7.0 5.0 3.0 1.0 1 2 3 4 5 Plastic Crus Elastic Crus 7.0 5.0 3.0 1.0 1 2 3 4 5 Plastic Crus Figure 2. Elastic vs. plastic crus measurements. B-Pillar Plastic Crus 5 4 3 2 1 2 4 6 A-Pillar Plastic Crus Figure 3. B vs. A pillar plastic crus. A-Pillar B-Pillar -0-1.00 0 1.00 0 Normal Quantile Dynamic Crus -0-1.00 0 1.00 0 Normal Quantile Figure 4. Q-Q Plots A & B pillars. Batzer 5