Identification of Vehicular Impact Conditions Associated with Serious Ran-off-Road Crashes (2010)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

4A detailed literature review was conducted to identify stud- ies relevant to the identification of impact conditions for ran-off-road crashes. The review identified numerous studies pertaining to ran-off-road crashes that could have some bear- ing on this project. However, upon review, most of these studies utilized only police-level crash data, which do not have the required details or information to assess the impact con- ditions of ran-off-road crashes. An annotated bibliography is shown as Appendix A, and a summary of related ongoing research studies are presented in Appendix B. Only a summary of results of the literature review is presented in this chapter. The literature review is presented under four general headings: 1. In-depth crash data collection 2. Impact conditions of ran-off-road crashes 3. Data needs for study of ran-off-road crashes 4. Reconstruction of ran-off-road crashes 2.1 In-Depth Crash Data Collection Crash data collection can be grouped into three general levels of detail: 1. Police-reported level 2. Enhanced police-reported level 3. In-depth level More detailed discussions on these three categories of crash data collection are presented in this section with examples. It should be noted, however, that these examples are intended as illustrations only and are by no means all inclusive. There have been so many studies using crash data over the years that it would not be feasible to include even a fraction of the studies in this review. The police-reported level is the most common type of crash data available. State and local police officers are required by law to investigate all reportable crashes and complete police acci- dent reports on these crashes. The data are then coded and entered into state crash data files. Crash data on the police- reported level are generally very limited in detail. Occasionally, more detailed data are collected on selected crashes, such as those resulting in fatalities and severe injuries, but such detailed investigations constitute only a small fraction of crashes. Most of the collected data elements are intended for identi- fication and record-keeping purposes—such as date, time, and location of crash; vehicle(s) and driver(s) involved; damage to the involved vehicle(s) and other property; injury sustained by driver(s) and occupant(s) of vehicle(s); and a brief description of what happened in the crash. The crash data may be merged with other data files for additional information. For example, the Highway Safety Information System (HSIS) combines crash data with other roadway and vehicle-related data files, such as roadway inventory, traffic, alignment, bridge inventory, vehicle identification and registration, etc., to expand the information database for use in various analyses. Even with the merged files, crash data on the police-reported level still lack the detail needed for analysis beyond problem identification and are of little use from the standpoint of estimating impact conditions of single-vehicle, ran-off-road crashes or evaluating the impact performance of roadside safety features. Thus, studies pertaining to police-reported level crash data are not included in the literature review. Crash data on the enhanced police-reported level are used in selected research studies in which additional data elements are collected to supplement the police-reported data. The sup- plemental data collected vary from study to study depending on the objective(s) of the study. Most of the supplemental data pertain to items of specific interest to the studies, such as details of roadside conditions, inventory of a particular road- side object(s), etc. However, there have been a few studies in which the investigating officers were asked to provide infor- mation on departure and impact conditions. In a study by Garrett and Tharp, the investigating officers were asked to provide estimates on impact speed and angle on C H A P T E R 2 Literature Review

324 crashes that occurred on the Ohio Turnpike over a period of five months during the summer and fall of 1967 (8). Simi- larly, in a study by Perchonok et al. to assess the relationships between single-vehicle, ran-off-road crash frequency, severity, and roadway and roadside features, data on over 9,000 crashes were collected from six states (2). The investigating police offi- cers were asked to complete supplemental field forms, includ- ing data pertaining to impact conditions, such as impact speed and angle. While enhanced police-reported level crash data provide more detailed information, its utility on estimating impact conditions is limited by a number of factors: 1. Expertise and experience of the investigating police officers. Most police officers receive some basic training in crash investigation, but only a small proportion of the officers receive the highly specialized training in crash reconstruc- tion needed to accurately estimate impact conditions. The quality of data collected by police officers without the spe- cialized training may be questionable. 2. Knowledge of the impact performance of roadside safety features. Even for trained officers, reconstruction of single- vehicle, ran-off-road crashes pose special problems unless the person is also knowledgeable of the impact performance of roadside features. Most reconstructions are based on energy dissipation and balance. For many ran-off-road crashes, energy dissipated by the struck object constitutes a significant portion of the energy equation and must be properly accounted for. This in turn will require knowledge on the impact performance of roadside features, which is beyond the training received by police officers. 3. Time and effort required. To properly reconstruct a crash to estimate its impact conditions would require time and effort beyond those available to an investigating officer. Thus, it is reasonable to expect that estimates of impact conditions would be based mostly on the judgment of the officers and less so on step-by-step reconstruction of the crashes. In summary, enhanced police-reported level crash data, which uses investigating officers to collect supplemental data, could provide more detailed information on the impact con- ditions of single-vehicle, ran-off-road crashes. However, as discussed above, there are serious limitations to this approach that could not be easily overcome. Thus, the use of enhanced police investigation to estimate impact conditions is not recommended. To properly estimate the impact conditions of single-vehicle, ran-off-road crashes, an in-depth level of crash investigation is required. The required data would include detailed data on the roadway, vehicle trajectory, object(s) struck and damage sus- tained, vehicle and damage measurements, and driver and occupant injury levels. The cost associated with in-depth crash investigation is, as may be expected, very high and there have only been a few ad hoc studies that incorporated such in-depth crash data, i.e., the data collection was designed specifically for the study. The most notable study involving in-depth crash data is per- haps the study on crashes involving pole support structures (9). A stratified random sample of over 1,000 crashes involving utility poles, breakaway and nonbreakaway luminaires, and sign supports were investigated in-depth, and the crashes were reconstructed to estimate the impact conditions. The in-depth crash data were then analyzed in conjunction with police- reported level data on all crashes and all pole crashes, enhanced police-reported level data on unreported crashes, and pole inventory data to address the study objectives. The results of the study include the extent of the pole crash problem; the characteristics of pole crash sites, vehicle damage, and occu- pant injuries; assessments on the performance of various pole types; and a cost-effectiveness evaluation of the break- away modification as a safety treatment. Another study of crashes on highway narrow bridges involved in-depth investigation of 124 crashes that occurred on bridges (10). Again, the in-depth crash data were analyzed in conjunction with police-reported level data on crashes that occurred on 11,880 bridges from five states and supplemen- tal field data on a sample of 1,989 bridges to address the study objectives. The results of the study include extent of the narrow bridge crash problem and the associated crash frequencies and rates; relationships between various bridge physical and oper- ational characteristics to crash rates and severities; and the char- acteristics and relationships between crash and injury severity for crashes at bridges. Other studies have utilized data from various in-depth crash investigation programs conducted by the National Highway Traffic Safety Administration (NHTSA). Since its inception in late 1960, NHTSA has sponsored numerous programs to col- lect in-depth crash data. The programs changed over the years, from the multidisciplinary accident investigation (MDAI) pro- gram in the late 1960s in which a small convenient sample of crashes were studied in great detail to the current NASS CDS that investigates a nationally representative stratified random sample of crashes in lesser detail. However, these in-depth data collection programs are designed to meet the data needs of NHTSA and the emphasis is, therefore, on data pertaining to the vehicle, occupant, and injury severity. Unfortunately, data pertaining to roadway and roadside characteristics are mostly lacking, which limits the use of the data for highway-related research, such as the current study. In order to make use of the NASS CDS data, supplemental data collection is necessary to gather information required for the specific study. The supplemental data collection can be prospective or retrospective in nature. The NASS program 5

6has a special study subsystem that allows for prospective col- lection of supplemental data in addition to the standard data elements collected under CDS. For instance, three special studies were designed to collect in-depth crash data on longi- tudinal barriers, pole support structures, and crash cushions (11, 12, 13). These special studies were met with different degrees of success. Nearly 1,200 cases were collected under the Longitudinal Barrier Special Study (LBSS) while only a negligible number of cases were collected under the pole and crash cushion special studies. The LBSS cases were sub- sequently reconstructed using the conservation of energy approach and the data were analyzed to examine the severity of barrier length-of-need (LON) crashes versus barrier-end impacts. Cases involving failure of the barrier system were reviewed clinically (14). Crashes involving concrete barriers were selected from the LBSS data file for use with an FHWA study on rollovers caused by concrete barriers (15). Of the 130 crashes involving concrete barriers, 31 resulted in rollovers. In addition to comparing the characteristics of crashes resulting in rollovers to those of non- rollovers, the rollover crashes were also clinically analyzed to identify potential causes for the rollovers. These studies illustrated the potential application of the spe- cial studies as well as the problems associated with their con- duct. This special study approach was not again utilized until the recent Large Truck Crash Causation Special Study, spon- sored by the Federal Motor Carrier Safety Administration (FMCSA). The purpose of this study was to determine specific causes of large truck (trucks with gross vehicle weight rating of over 10,000 lbs) crashes. These crash causation data will help to identify crash countermeasures the FMCSA can undertake with regard to interstate motor carriers, their drivers, and their vehicles; and in cooperation with other DOT agencies and state governments with regard to the non-commercial vehicles, pedestrians, and pedal cycles involved in the crashes. Another approach is to supplement the NASS CDS data ret- rospectively with additional field data collection. Data ele- ments of specific interest to the study, but not covered under NASS CDS, are identified and collected using supplemental field data collection. The key limitation of this approach is that the supplemental data elements should not change over time since the supplemental data are collected one to two years sub- sequent to the occurrence of the crashes. This is not a bad assumption for most data elements pertaining to highway and roadside characteristics since they typically do not change except during major construction or reconstruction. This retrospective approach was utilized in ongoing NCHRP Project 17-11, “Recovery-Area Distance Relation- ships for Highway Roadside” (16). The objective of the study is to develop relationships between recovery-area distance, roadway and roadside features, vehicle factors, encroachment parameters, and traffic conditions for the full range of high- way functional classes and design speeds. Part of the research involved clinical analysis of 338 NASS CDS cases from 1997 and 1998. Field data on roadway and roadside characteristics of crash sites were collected to supplement the standard NASS CDS data elements. These sampled cases (e.g., police accident reports, field forms, scaled diagrams, and photographs) were then manually reviewed to glean additional information beyond the comput- erized data elements. The crashes were then reconstructed to estimate impact conditions and vehicle trajectories from the manual review such as impact sequence, pre- and post-impact vehicle trajectories, impact angle, etc. The same retrospective approach and data collection pro- tocol used in NCHRP Project 17-11 were used in the rollover study (17) sponsored by FHWA, except that the cases were sampled from the 1999 NASS CDS data file. The objectives of this study were to determine the specific causes of rollover events associated with the full range of passenger vehicle col- lisions in which such an event occurred. In fact, the data from NCHRP Project 17-11 were utilized in this study with addi- tional in-depth clinical reconstruction on the 180 rollover crashes contained in the database. In addition, new data from 175 NASS CDS cases from 1999 were added to the database. However, NHTSA recently changed its privacy policy to discard police accident reports after only one year. This policy change effectively eliminates this retrospective approach since the only means of identifying the crash sites was from the police accident reports. The prospective special study is the only viable approach for future studies using the NASS CDS program. A new emerging technology may provide a totally new and better source of data on impact conditions. Automobile man- ufacturers have installed Event Data Recorders (EDRs) in selected vehicle lines in recent years. The EDR is designed as a controller for monitoring airbag deployment and seatbelt usage and recording data pertaining to the crash event in case of a crash. Data elements recorded include crash pulse, seat- belt usage, and pre-crash information, such as speedometer reading and engine performance parameters. In the future, EDR data may simplify reconstruction of crashes to estimate impact conditions and could become and an invaluable sup- plement to in-depth crash investigation. NHTSA is currently collecting available EDR data under its NASS CDS and Special Investigations (SCI) programs and compiling the data into a national database. While the EDR technology is relatively new and little actual data are currently available, its potential in the future is very promising: • EDRs are now deployed in all vehicle lines, so more data should become available. • The number of data elements and the length of recording period are somewhat limited now. However, with rapid

7advances in electronics, many more data elements could be incorporated into the EDRs and the recording period could increase significantly. • In addition to the interest of NHTSA, the highway roadside safety community has also shown great interest in the EDR data. A study, NCHRP Project 17-24, “Use of Event Data Recorder (EDR) Technology for Roadside Crash Data Analy- sis,” was conducted to review and recommend a minimum set of EDR data elements for roadside safety analysis as well as procedures to retrieve, store, and use the data (18). While the EDR technology is very exciting and promising, there is still much development to be done and impediments to overcome before it can reach its potential, including: • Engineering issues. There are no current standards govern- ing the design and use of EDRs, such as data elements to be included, data format, data retrieval, etc. Such standards are needed if data are to be collected on a large scale. Also, cur- rent EDR data elements are, as expected, focused on vehicle parameters with no specific consideration for information pertaining to ran-off-road crashes. • Institutional barriers. EDR data are intended for the data needs of vehicle manufacturers, which may be reluctant to share their proprietary designs for competitive and legal con- siderations. Inputs from governmental agencies and research institutions are needed in the early planning and design stages if the EDR data are to be expanded into the roadside safety area. • Legal consideration. There are still questions pertaining to ownership of the EDR data, privacy issues, use of EDR data in tort claims, etc. Until such concerns are addressed and resolved, large-scale collection of EDR data appears unlikely. 2.2 Impact Conditions of Ran-Off-Road Crashes Despite the large number of studies on ran-off-road crashes, there are relatively few studies that actually attempted to esti- mate the impact conditions. The main reason for the lack of such effort is that, in order to estimate the impact conditions, an in-depth level of crash investigation is required, including detailed data on the roadway, vehicle trajectory, object(s) struck and damage sustained, vehicle and damage measure- ments, and driver and occupant injury levels. The costs associ- ated with in-depth crash investigation are, as may be expected, very high and there have only been a few studies that incorpo- rated such in-depth crash data. Another limitation is that some of the studies, such as the LBSS data, were not based on a rep- resentative sample and the resulting distributions of impact conditions could be biased, probably toward the more severe crashes. Some earlier work relied on reconstruction of impact speed and angle by the investigating officers, such as the studies by Garrett and Tharp (8), Perchonok et al. (2), and Lampela and Yang (1). As discussed previously, the use of enhanced police- reported level crash data to estimate impact conditions is lim- ited by a number of factors, such as expertise and experience of the investigating police officers, availability of time for the offi- cers, and lack of officers’ knowledge on the impact perfor- mance of roadside safety features. Thus, while the results from these studies provide some insights into impact conditions, their accuracy is somewhat questionable. Under the pole and narrow bridge studies (9, 10), impact conditions were estimated from in-depth investigations and presented in the reports. Mak et al. took the data from these studies and developed statistical models for the distributions of impact speeds and angles (3). After screening, a total of 596 cases were available for analysis. The authors found that the gamma function provides the best fit for univariate impact speed and impact angle distributions. Statistical models for impact speed and angle distributions were then developed using the gamma function for the following five functional classes: • Freeway • Urban arterial • Urban collector/local road • Rural arterial • Rural collector/local road For some roadside features, such as longitudinal barriers, impact conditions are defined by both impact speed and angle. However, there is no known means of mathematically express- ing a joint gamma distribution. The authors tested various known joint (bivariate) distributions, but with no success. They then proceeded by assuming that the impact speed and impact angle are independent of each other and estimated combined probability distributions for impact speed and angle stratified by functional class and based on the gamma distribution. These impact speed and angle distributions were used in some of the cost-effectiveness analysis procedures, including the Texas Transportation Institute (TTI) ABC model (19). The distribu- tions were adjusted to reflect the current higher speed limits under NCHRP Project 22-14 (20). The revised impact condi- tion distributions were used with the Roadside Safety Analysis Program (RSAP) (21). Other sources of impact conditions include data from on- going NCHRP Project 17-11 and the FHWA Rollover Study (16, 17). A total of 559 NASS CDS cases from 1997 through 1999 were selected under these two studies. Supplemental field data were collected on these cases, which were then recon- structed to estimate the impact conditions. The impact speed and angle distributions developed under these two studies

8were significantly different from previous findings. However, it was later found that the scales on some of the diagrams used for the impact angle reconstructions might be distorted. In order to fit the scale diagrams onto a web page, the longitudi- nal and lateral scales were compressed differently, thus lead- ing to incorrect impact angle estimates. Plans are underway to reconstruct these cases again to correct the errors and reana- lyze the revised data. It should be mentioned that in order to properly establish the distribution of impact conditions, the data source needs to be either the population (i.e., all ran-off-road crashes) or a rep- resentative sample. Some databases, such as the LBSS, are sam- pled on the basis of a comparative analysis and are not suitable for determining impact condition distributions. 2.3 Data Needs for Study of Ran-off-Road Crashes There have been a number of studies that looked into the data needs for studying ran-off-road crashes. A study by Mak and Sicking identified issues and gaps in the state of the knowl- edge needed to improve the cost-effectiveness analysis proce- dure and to develop data collection plans for those issues and gaps that could be addressed with crash data. The research pro- posed five studies and developed data collection plans for those studies: • Validation of encroachment frequency/rate • Determination of encroachment frequency/rate • Effect of roadside conditions on impact probability and severity • Distributions of impact conditions • Relationships of impact conditions, performance limits, and injury probability and severity These study plans were reviewed by a panel of experts and their comments taken into consideration. The recommended study on the distributions of impact conditions focuses on impact speed, angle, and vehicle orientation in addition to vehicle size, weight, and the nature of the roadside object/ feature. The plan for this study includes the following tasks: • Select sample roadway segments for each of the six highway types • Set up data collection protocol, including sampling plan, accident notification scheme, data collection forms, etc., and familiarize and train investigators with the protocol through a small pilot study • Investigate in-depth a representative sample of single- vehicle, ran-off-road accidents on these selected roadway segments • Reconstruct the sampled accidents to determine impact conditions • Compile descriptive statistics on vehicle trajectory and impact conditions • Develop mathematical models for the distributions of impact speeds and angles These proposed studies and data collection plans are over 10 years old, but they still are applicable today and of great interest to the current study. Miaou proposed a method to estimate vehicle roadside encroachment rates using accident-based models (22). Miaou concluded that the proposed method could be a viable approach to estimating roadside encroachment rates without actually collecting the encroachment data in the field, which can be expensive and technically difficult. A pilot study was conducted by Daily et al. (23) to examine the feasibility of this approach. Data were collected on 56 km (35 mi) of tangent sec- tions of rural two-lane highways in Idaho, including detailed roadside, crash, and traffic data. Encroachment rates were esti- mated from the collected crash data and found to be in the same order of magnitude as previous research. It was con- cluded that this approach is feasible, although it is limited by the current state of knowledge with respect to data on the tra- jectories of vehicles involved in ran-off-road, fixed-object acci- dents. An experimental plan for future research that would produce improved estimates of encroachment rates was devel- oped, but not recommended for immediate implementation. While this study has no direct bearing on the current study, it could be of interest in future data collection efforts. Data on encroachment rates are over 25 years old and may be outdated in light of the significantly changed conditions in the interven- ing years, including improvements made to the safety design of highways (e.g., clear zone concept and improved barriers and terminals) and vehicles (e.g., front and side airbags, antilock brakes, and crush management), and other safety counter- measures (e.g., mandatory seatbelt law, tightened blood- alcohol-content law). If a major data collection effort is to be implemented in the future, encroachment data may be one of the objectives. A list of suggested data elements for use with the current NASS CDS program was proposed by Eskandarian et al. in a study to assess the compatibility between vehicle design characteristics and roadside safety hardware (24). These data elements pertain to the design characteristics, pre-impact con- ditions, and impact conditions of struck features and assess- ment of impact performance of features. While the suggested data needs pertain mostly to the issue of compatibility between vehicle design and roadside safety features, the information would be helpful to establishing the data needs for the data collection effort under the current study. Under the recently completed NCHRP Project 17-24 on the potential use of EDR data for roadside safety evaluation, the authors examined the data needs for roadside safety analysis and assessed whether the data needs can be satisfied with EDR

9mation of semi-rigid barriers is estimated from a series of computer simulations that correlate impact severity to maxi- mum barrier deflection. The impact severity (IS), calculated using the following equation, has been shown to be a good indicator of the degree of loading and maximum deflection of a barrier during an impact. where: IS = Impact Severity M = Vehicle mass V = Vehicle velocity θ = Impact angle The IS value, in conjunction with the impact angle, can then yield a direct estimate of impact speed. The impact speed cal- culated from barrier deflection should be verified by energy loss calculations to make sure that the estimates using both approaches are consistent. Another procedure was developed for reconstructing rigid barrier impacts under the study to assess rollovers on concrete barriers (15). For impacts involving concrete barriers, there is typically no deformation/damage to the barrier. However, it was found that vehicle/barrier friction was a major source of energy dissipation during a crash. Thus, energy loss due to deformation/damage to the barrier is replaced by vehicle/ barrier friction, which is estimated as a function of the length of barrier contact. Total energy loss is then calculated as the sum of energy losses due to vehicle crush, vehicle/barrier fric- tion, post-impact vehicle trajectory, and the impact speed calculated accordingly. As a means of verification, the vehicle crush energy is matched to the energy associated with the lateral velocity of the impacting vehicle. If both energy estimates are compara- ble, the procedure is believed to be reasonably accurate. If not, the vehicle crush energy would be adjusted appropriately and a new estimate of the impact speed generated. This iterative procedure was found to give reasonably good estimates of impact speed when used to evaluate findings from full-scale crash tests. Another computerized reconstruction procedure was devel- oped for ran-off-road crashes involving pole support struc- tures, including breakaway and non-breakaway utility poles, luminaire supports, and sign supports (25). Energy losses due to vehicle crush and post-impact vehicle trajectory are esti- mated using the CRASH3 program. Energy loss associated with breaking or fracture of the pole is estimated based on empir- ical test data. Impact speed is then calculated from the total energy loss. IS M V= ( )1 2    sinθ data. A list of new data elements for EDR was proposed. As mentioned previously, the EDR technology is very exciting and promising. However, until such time that these new EDR data elements become available, in-depth crash investigation will remain the primary means of obtaining such detailed crash data. 2.4 Reconstruction of Ran-off-Road Crashes There are a number of existing procedures that have been developed for reconstructing special types of ran-off-road, fixed-object crashes (14, 15, 25), including: • Semi-rigid and flexible barrier • Rigid barrier • Pole support structure These reconstruction procedures are based on the general principle of identifying the energy loss parameters during the collision and summing the total to determine the change in velocity from point of impact to point of final rest. The com- ponents of the energy loss in a typical crash include: • Vehicle crush • Deformation/damage of roadside feature • Vehicle trajectory Energy due to vehicle crush can be estimated manually using equations from Campbell (26) or using a computerized recon- struction procedure, such as CRASH3. Energy loss due to post- impact vehicle trajectory is estimated using equations of motion. Adjustments are made to account for skidding and sliding. For rotating vehicles, the distance traveled is based on the angle of rotation and the radius and the energy loss calcu- lated accordingly. Energy loss due to vehicle trajectory can also be estimated using a computerized reconstruction procedure, such as CRASH3. These two energy loss items can be stan- dardized and incorporated into a single reconstruction proce- dure. Unfortunately, energy loss due to deformation/damage of the roadside feature varies greatly among the roadside fea- tures and impact configurations, e.g., barrier length-of-need versus barrier end impact. Thus, there is not a single proce- dure that can be used to reconstruct all ran-off-road crashes. Instead, different reconstruction procedures are needed to accommodate the wide variety of roadside features. A reconstruction procedure for semi-rigid and flexible bar- riers was developed for the LBSS data (14). The procedure utilizes similar techniques for estimating vehicle crush and trajectory energy losses. Energy loss associated with the defor-