Evaluating the Performance of Roadside Hardware

Transcription

1 Evaluating the Performance of Roadside Hardware Christine E. Carrigan Phone: Malcolm H. Ray Phone: Archie M. Ray Phone: RoadSafe, LLC Box Main Street Canton, Maine Submitted July 28, 2016 Resubmitted October 31, 2016 Word count Text = 5,700 Figures & Tables: 250 words each = 1,500 Total number of words= 7,200 Paper prepared for consideration of presentation and publication at the 96th Annual Meeting of the Transportation Research Board, January 2017

2 Carrigan, Ray, and Ray 1 Evaluating the Performance of Roadside Hardware Christine E. Carrigan, Malcolm H. Ray, and Archie M. Ray ABSTRACT The performance of roadside hardware is evaluated during the design phase using fullscale crashes. Crash testing has evolved over the last 40 years into a uniform evaluation standard currently presented in the Manual for Assessing Safety Hardware. The Roadside Design Guide provides guidance on how to select, locate and place roadside hardware which has been successfully crash tested. Ultimately, however, it is the performance of the hardware under real field conditions that should be used to evaluate the performance of designs. Crashes must occur and be observed in order to evaluate the in-service performance of roadside hardware. A review of crash frequency or probability of crashes occurring with particular roadside hardware evaluates the guidance for the application of hardware presented in the Roadside Design Guide since it recommends placement, location and layout parameters. Evaluation of the performance of the hardware itself, however, is based only on the outcome of crash that actually occurs. This paper presents a summary of crash testing evolution, confounding factors which should be acknowledged when evaluating any roadside hardware and the strengths and weaknesses of different study designs to estimate the effectiveness of the hardware. A discussion of data analysis techniques would have far exceeded the space limitations imposed, therefore, are excluded from this paper. An illustration of the suggested techniques is presented using flared guardrail terminal data from Maine. It is recommended that prior to any ISPE the evaluators first become knowledgeable in: (1) the development history; and (2) the acceptable uses of the hardware.

3 Carrigan, Ray, and Ray INTRODUCTION The performance of roadside hardware such as longitudinal barriers, sign supports, guardrail terminals, and work zone devices is evaluated during the design phase using full-scale crash tests. The hardware must pass the currently adopted evaluation criteria to be eligible for federal-aid funding on the National Highway System. (1, 2, 3, 4) The AASHTO Roadside Design Guide provides guidance on how to select, locate and place roadside hardware which has been successfully crash tested. (5) The crash test evaluation criteria are limited by their very nature to a small number of impact speeds and angles and particular vehicles. After successful crash-test performance has been demonstrated in crash tests, the roadside hardware is installed in the field according to the guidelines of the AASHTO Roadside Design Guide and the appropriate State design standards. Ideally, the field performance should be assessed using an in-service performance evaluation (ISPE) as recommended and described in NCHRP Report 490.(11) ISPE refers to the process of examining real-world collisions to determine how effectively roadside hardware works when exposed to the full spectrum of vehicle types, impact conditions, maintenance, weather and traffic conditions that the roadside hardware is exposed to in the field. ISPEs have been recommended in the crash test guidelines since at least NCHRP Report 153 was published in Report 153 notes the purpose of crash tests are to screen out those candidate systems with functional deficiencies. The final evaluation of an appurtenance must be based on carefully documented in-service use. (1) The expressed purpose of NCHRP Report 230, published in 1981, was to present uniform procedures to highway agencies, researchers, private companies, and others for crash testing and in-service evaluation as a basis for determining safety performance of candidate appurtenances. (2) Ross et al. in NCHRP Report 350 presented procedures for...conducting vehicle crash tests and in-service evaluation of roadside safety features or appurtenances. (3) NCHRP Report 490, published in 2003, provided specific recommendations for setting up, performing and analyzing the data for roadside safety ISPEs. (11) In 2005, the FHWA issued an information memorandum stating NCHRP Report 490, In-Service Performance of Traffic Barriers, published in 2003, provided a model methodology that could be used by hardware manufacturers and transportation agencies to monitor the performance of their hardware. (6) AASHTO published the Manual for Assessing Safety Hardware (MASH) in The stated purpose of MASH, like the previous roadside hardware assessment documents, is to present uniform guidelines for crash testing and in-service performance evaluation. MASH provided guidance for ISPEs through reference to NCHRP Report 490. Crashes must occur and be observed in order to evaluate the in-service performance of roadside hardware. The frequency or probability of these crashes occurring will depend on a variety of factors including highway speed, horizontal curvature, grade, lane width and many other highway characteristics. The AASHTO Roadside Design Guide (RDG) provides guidance on the placement, location, and layout of hardware. (5) A review of crash frequency or probability of crashes occurring with a particular roadside hardware evaluates the guidance presented in the RDG, not the performance of the hardware per se. Evaluation of the performance of the hardware itself is based on the outcome of observed crashes that actually occur involving the hardware of interest. In other words, hardware performance is a conditional probability; given a crash has occurred, what is the crash outcome and performance of the roadside hardware?

4 Carrigan, Ray, and Ray The need to quantify the in-service performance of roadside hardware has been recognized for decades. In the last 40 years, the crash testing of roadside hardware has evolved into a set of detailed, precise recommendations that apply a uniform evaluation standard to test results by reputable and certified testing laboratories. ISPEs, however, have not enjoyed the same level of evolution and uniform application. This paper presents a summary of crash testing evolution, confounding factors which should be acknowledged when evaluating any roadside hardware in the field and the strengths and weaknesses of different study designs to estimate the effectiveness of the hardware and control for confounding factors. An illustration of the suggested techniques is presented using the roadside hardware category of flared guardrail terminals. The discussion of data analysis techniques would have far exceeded the space limitations imposed, therefore, is excluded from this paper with the exception of a brief background section BACKGROUND This paper discusses the statistical design of an ISPE. Reference is made to many data analysis techniques. While a detailed discussion of each technique is not the intent of this paper, a very brief summary and references for additional reading are presented. Case-control and cohort studies compare the relative effect of a treatment to a non-treatment. A case-control study establishes the population of interest from existing crashes, then distinguishes between the cases and controls. These types of studies are done retrospectively and are referred to by Report 490 as retrospective ISPEs. The cohort study establishes a study area and collects cases and controls as the crashes occur. These types of studies are done prospectively and are referred to by Report 490 as prospective ISPEs. These and other statistical methods are further documented in the Federal Highway Administration technical report FHWA-SA HARDWARE DESIGN The recommended crash test matrices are different for each type of roadside hardware (e.g., longitudinal barriers, sign supports, luminaires, workzone devices). Each type of hardware is evaluated using a recommended testing matrix unique to that group. These tests are conducted at specific impact conditions (i.e., speed and angle) and with a specific vehicle as outlined in the crash test recommendations. The crash test recommendation allows for consistency and uniformity between the testing laboratories such that all hardware is evaluated according to the same standard. While the test impact conditions are intended to represent the practical worst case crash scenario there are obviously many more types of impact scenarios, impact conditions and vehicle types than can be practically evaluated using full-scale crash tests. There is, therefore, a need to examine the performance of roadside hardware under the full range of conditions that are experienced in the field. Neither NCHRP Report 153 nor 230 explicitly included performance or test levels. (1, 2) The crash test matrix included small, medium and large passenger cars. Supplemental tests for heavier vehicles such as utility buses (i.e., school buses), small and large intercity buses, tractor trailer trucks, and tanker trailer trucks were included in NCHRP Report 230. (2) NCHRP Report 239 included four service levels for bridge railings and attempted to establish the service levels based on the capacity of the bridge railings based on the NCHRP Report 230 supplemental tests. The AASHTO Guide Specification for Bridge Railings (7) further extended the concept of multiple performance levels for bridge railings. NCHRP Report

5 Carrigan, Ray, and Ray (3) was published in 1993 and expanded the concept of performance levels, specifying as many as six different test levels for different types of roadside hardware. Changes in vehicle fleet characteristics were one of the issues that prompted the update from Report 350 to the Manual for Assessing Safety Hardware (MASH) in (4) MASH includes essentially the same test level approach with some changes to the vehicles and impact conditions. The selection of test vehicles for test levels 1-3 (passenger vehicle tests), for example, were established by sales data which represented a major change from Report 350. The impact speeds and angles for the length of need (LON) test matrix for longitudinal barriers tested under NCHRP Report 350 and MASH are shown in Table 1. The changes between NCHRP Report 350 and MASH are highlighted by the bold-italic font. Notice that the impact speed and angle did not change for the pick-up truck, tractor van-trailer or tractor tankertrailer tests. The impact speed did not change for the small car, but the angle was increased from 20 degrees to 25 degrees to match the pick-up truck test. The impact angle did not change for the SUT, but the speed was increased from 50 mph to 56 mph. (3, 4) The masses of the small car, pickup truck and SUT were all increased and some of the geometric dimensions of the test vehicles were also changed to reflect more modern vehicles. Updates to the MASH TL3 test matrix for median barriers in a 4H:1V v-ditch have also been proposed and will likely appear in the next edition of MASH.

6 Carrigan, Ray, and Ray Table 1. Recommended Longitudinal Barrier LON Impact Speed and Angle. TL1 TL2 TL3 TL4 TL5 TL6 NCHRP Report 350 (3) Small car 820C Small car 31mph/20 44mph/20 62mph/20 700C pick-up 2000P 31mph/25 44mph/25 62mph/25 SUT 8000S 50mph/15 Tractor Van-Trailer 50mph/ V Tractor Tank- Trailer 50mph/ T MASH (4) Small Car 1100C 31mph/25 44mph/25 62mph/25 Pick-Up Truck 31mph/25 44mph/25 62mph/ P SUT 10000S 56mph/15 Tractor Van-Trailer 50mph/ V Tractor Tank- Trailer 50mph/ T NOTE: Content in bold-italic font highlight changes from NCHRP Report 350 to MASH ISPE STUDY DESIGN Once new hardware has been designed and successfully crash tested it is installed in the field. The design and test procedures provide a best assessment of the hardware performance prior to field deployment, but the field performance should be monitored to ensure that the hardware actually performs as desired under real-world conditions. In-service evaluation of the hardware is based on the observation of real crashes under real-world conditions rather than the idealized conditions used in crash testing. When evaluating hardware performance in the field, the frequency or probability of having a crash is not usually the issue under evaluation. The issue under evaluation is the performance of the hardware in a crash. Defining the question to be evaluated (e.g., crash severity, redirection, rollover after

7 Carrigan, Ray, and Ray impact, etc.) is necessary. Assuming the crash has occurred and limiting the analysis to hardware performance is appropriate in order to compare the field performance with the expected performance based on the crash test evaluations. When conducting an ISPE, assessing the hardware performance is best thought of as one component of a conditional probability. There is some probability that the hardware will be struck by an encroaching vehicle and there is a second conditional probability of a poor outcome given a crash has occurred. The first probability is independent from the performance of the hardware. The hardware specified and installed at the time of construction may not have been evaluated to the most current testing specification. For example, the MASH implementation agreement between the FHWA and AASHTO allowed for a staged implementation of MASH hardware through 2019 while MASH was published in 2009.(8) Currently, there are both NCHRP Report 350 and MASH systems available for use and many systems tested under older criteria remain on the roadways. Direct comparison of the field performance of any roadside hardware system should acknowledge these nuances. One may learn from a review of the study area design and/or construction specifications all or some of these factors and appropriately limit the study population. A study design which identifies the hardware type to be included in the population and the question to be studied will allow for consideration of the relevancy of the crash testing specification under which the hardware was developed. The analyst designing a study for the in-service performance evaluation of roadside hardware should be familiar with the testing criteria evaluation objectives for the hardware being evaluated and design the study to assess the relevance of the hardware testing criteria based on the conditions it is exposed to in-service. The following sections discuss some confounding factors and restriction or matching approaches which can be used to ensure the study design is addressing the question being asked. Impact conditions If the impact conditions are different than those adopted in the crash testing standard, the hardware may not perform as crash tested. Impact conditions are not included in typical policelevel crash datasets, therefore, field data collection and crash reconstruction would be necessary at each crash scene to measure the impact speed and angle, vehicle orientation, impact on the device, etc. with high confidence. A cohort study would be an appropriate study design for gathering these data and estimating the effect of impact conditions outside of the crash test criteria. Sometimes surrogates for impact speed like posted speed limit and highway type can be used to at least restrict the sample to similar ranges of impact conditions. A case-control study could be used if one were to use posted speed limit as a surrogate for impact speed. Scene diagrams or other police-reported information may be used to determine the vehicle orientation at impact relative to the hardware as a surrogate to impact angle. Typically, a cohort (i.e., prospective) study provides a smaller quantity of higher precision data whereas a case-control (i.e., retrospective) study provides a larger quantity of lower precision data. In addition to the impact conditions, the impacting vehicle should be considered. The hardware test levels are limited to certain particular vehicle types. For example, guardrail terminals are typically designed to test level three (TL3) which only involves passenger vehicles such as pick-up trucks and sedans. Impacts with single-unit trucks or tractor-trailer trucks are outside of the TL3 design conditions.

8 Carrigan, Ray, and Ray Site Design, Maintenance and Repair The roadside hardware in crash tests are installed in new condition by well-trained individuals on idealized terrain. The hardware is correctly installed on a site in accordance with the hardware designers guidance and/or the guidance provided in the AASHTO Roadside Design Guide and MASH (5, 8). There is no need for maintenance or repair of the system since new hardware is installed on a closed test track and struck by only the single test vehicle. Each subsequent test is conducted using a new installation of new hardware. Neither the crash test nor field performance of improperly installed or maintained hardware or hardware in disrepair is known based on typical crash tests. A field inventory of a designated study area could be used to establish the existing maintenance and/or repair status of the hardware if maintenance and repair records are available. Trends in maintenance or repair issues may be observed and the more prevalent trends may be crash tested in a laboratory setting to establish the crash test performance. Inspection of hardware, after the damage has occurred, can be informative when conducted by a trained data collector. A cohort or prospective study would best serve this instance. If detailed maintenance and repair records are available these may be useful in a case-controlled (retrospective) study as well. As a surrogate to field data collection, the reported sequence of events can often be used to determine events prior to or subsequent to the impact with the hardware. Guardrail terminals, for example, are designed to allow the vehicle to gate behind the system and the area behind the terminal should be clear of hazards and the terrain should be flat. A review of impacts with guardrail terminal where the secondary impacts included rollover in ditch, impact with tree, impact with pole, impact with shrub, to name several possibilities, would indicate that the area in question may have issues with the proper site design and location of the hardware. In this instance, a case-control or retrospective study design could be used. Data Restricting or Matching The goal of roadside safety and the focus of the AASTHO Roadside Design Guide, as described in the preface, is on safety treatments that minimize the likelihood of serious injuries when a driver does run off the road. (5) Data analysis, therefore, should progress is such a way that the analyst can measure the success toward achieving this goal. A singular focus on only fatal or serious injuries does not assess this goal. Crash outcomes of all severities must be collected in order to assess the probability of serious and fatal injuries. Filtering for the crash sequence of events is an effective way of using available crash data through a retrospective or case-control study approach. Different filters may be used to address slightly different questions. First and only harmful events (FOHE) with roadside hardware (i.e., the only harm was from the impact with the hardware) are the most similar to the crash testing evaluation criteria and subsequently the performance of the hardware itself when using crash severity to measure acceptable field performance. FOHE cases, however, may eliminate cases where the hardware was not located correctly on the site and cases where other impacts after the initial impact with the hardware occurred and caused harm. For example, a FOHE filter is inappropriate for addressing whether or not the study is using appropriate site design techniques or if additional impacts are possible after impact with the hardware. FOHE filters are also inappropriate when trying to study the redirective properties of the hardware because the second impacts will not be captured. First harmful event (FHE) mixes the severity of the impact with the hardware with the harm caused by other subsequent events in the sequence, but can be used in evaluation of other

9 Carrigan, Ray, and Ray in-service questions. For example, how frequently is longitudinal barrier breeched? Which types of vehicles breech the barriers most frequency? Do certain vehicles roll over more frequently than others with particular hardware? Any harmful event (AHE) and most harmful event (MHE) present the most challenges. These filtering strategies confuse the harm done and limit the ability to address other ISPE questions. While this information is certainly useful to capture and report, it does not generally address specific hardware performance questions. For example, a crash with a guardrail may result in typical redirection and after redirection into the roadway the vehicle may strike another vehicle. Is the harm observed in the crash more a result of the vehicle-to-vehicle crash or the guardrail crash? Does this particular hardware have a propensity for redirecting vehicles into traffic? Summary Each of these confounding factors can result in a real-world observable event where the roadside hardware does not perform as observed in the crash tests or events where the data analysis misinterprets how the hardware was originally designed and evaluates the in-service performance against this misconception. A study design which considers which question is being addressed at the onset is necessary to capture the effect of these factors on field performance ILLUSTRATION USING FLARED GUARDRAIL TERMINALS Crash Testing Evolution of Terminals The evolution of the recommended crash test matrix for guardrail terminals is summarized and compared in Table 2 for end-on impact evaluations, Table 3 for the length of need (LON) impact evaluations, and Table 4 for the reverse direction impact evaluation. The only substantial changes to crash testing criteria for terminals over the last 20 years have been changes to the test vehicle weights and the angle for the gating tests. Maine Study of Flared Terminals Generally, flared guardrail terminals have been preferred in Maine because they tend to be less prone to nuisance damage from snow plowing activities. A typical flared terminal is located such that the end is four feet offset from the tangent section of the guardrail so the end a positioned farther from the travel lanes. The 2004 Maine Department of Transportation Highway Design Guide instructs designers to use tangent guardrail terminals only [w]here there is physically not space available for an approved flared terminal. (9) This guidance was restated in a 2014 policy memo which noted that flared terminals are preferred. (10) It would be appropriate, therefore, to limit a study of in-service performance of guardrail terminals in Maine to flared terminals since tangent terminals are used only in places that have particular placement constraints. In other words, flared terminals should only be compared to other flared terminals and tangent terminals should only be compared to other tangent terminals. Using the Maine State Police crash database, guardrail terminal crashes were identified using the object struck code 18 Impact Attenuator / Crash Cushion and 28 Guardrail End. The geographic reference for each crash record was used to identify the location on Google Earth. The terminal installed at the time the GoogleEarth photo was taken was noted and used to identify the type of guardrail terminal likely at the site at the time of the crash.

10 Carrigan, Ray, and Ray 9 Table 2. Evolution of Center of Nose Minimum Test Matrix (1, 2, 3, 4) Y Veh Y = offset Impact angle = 0 degrees Impact angle =α degrees Specificaiton NCHRP 153 NCHRP 230 NCHRP 350 MASH Test Impact Speed (mph) Test Vehicle 4500lb sedan 2250lb subcompact S S 4500S 0 Y Specificaiton Test Impact Speed (mph) Test Vehicle 0 NCHRP feet NCHRP *-30 d 820C TL1=31 veh *-32 d 820C 15 TL1=31 S*-30 d TL2=44 700C a width/4 NCHRP 350 S*-32 d TL2=44 700C a 15 *-31 d 2000P 0 *-33 d 2000P 15 *-30 d veh 1100C *-32 TL1=31 width/4 d 1100C 5-15 TL1=31 *-31 d TL2= P 0 MASH *-33 d TL2= P 5-15 *-38 d 1500A * Replace star with the test level (e.g., TL1 would be 1-test number) a Test is optional b Gating terminal and redirective crash cushion; c Non-gating terminal and redirective crash cushion; d both gating and non-gating terminals and redirective crash cushions. α

11 Carrigan, Ray, and Ray 10 Table 3. Evolution of Length of Need (LON) and Critical Impact Point (CIP) Minimum Test Matrix. (1, 2, 3, 4) BLON=Beginning of LON L= distance from nose to BLON. Impact angle = α degrees IP=Impact point Impact Specificaiton Test Speed (mph) NCHRP 153 NCHRP 230 NCHRP Test Vehicle 4500lb sedan 25 BLON 2250lb subcompact 15 L/2 4500S 25 BLON S 15 L/2 α IP Specificaiton Test Impact angle = α degrees IP=Critical Impact Point Impact Speed Test Vehicle NCHRP NCHRP S 15 L/2 S*-36 c 700C a 15 BLON S*-34 b 700C a 15 *-36 c TL1=31 820C 15 BLON *-34 b TL1=31 820C 15 *-35 b TL2= P 20 BLON NCHRP 350 *-38 c TL2= P 20 *-37 c 2000P 20 BLON TL1=31 *-34 d TL1= C 15 MASH *-35 d TL2= P 25 BLON MASH TL2=44 *-36 d 2270P 25 * Replace star with the test level (e.g., TL1 would be 1-test number) b Gating terminal and redirective crash cushion; c Non-gating terminal and redirective crash cushion; d both gating and non-gating terminals and redirective crash cushions. α

12 Carrigan, Ray, and Ray 11 Table 4. Evolution of Reverse Direction Test Matrix. (1, 2, 3, 4) Specificaiton Test Reverse direction Impact Speed Test Vehicle Location NCHRP NCHRP NCHRP 350 MASH *-39 d *37 d TL1=31 TL2=44 TL1=31 TL2= P midlength α P CIP 25

13 Carrigan, Ray, and Ray Ten types of guardrail terminals and attenuators were identified in the data. Three of the guardrail terminals are installed using a nominal four-foot flare: the BCT, FLEAT and SRT. These terminals dominated the Maine dataset. While these three types of terminals are used with a nominal four-foot flare, the BCT was originally developed under NCHRP Report 153 and failed the NCHRP Report 230 evaluations (1, 2) while the FLEAT and SRT were tested and evaluated using NCHRP Report 350. (3) The NCHRP Report 350 tested terminals were originally implemented to address field-observed issues with the BCT, however, the BCT remains in use on some Maine roadways like in many other States since States usually only upgrade hardware after it has been extensively damaged or requires extensive maintenance. Maine also uses the MELT extensively. While the MELT was not identified in the dataset, it is believed the chosen use of Google Earth may have presented challenges distinguishing between the BCT and the MELT. The MELT passed the NCHRP 230 evaluations and the NCHRP Report 350 TL2, but is not an NCHRP TL3 w-beam terminal. The BCT and the MELT are referenced collectively herein as the BCT/MELT to remove any errors introduced in data collection in distinguishing between the two. One cannot be absolutely sure without a detailed review of maintenance and construction records that the terminal shown on GoogleEarth is the same as what was in place at the time of the crash. This case-control approach provides a reasonable first step in matching crash records to terminal types for States without photologs or detailed maintenance and repair records. It is important to recognize these data were collected from 2012 through 2014 for each of the terminals in the study so the performance of each terminal was assessed using the same vehicle population, under the same traffic conditions even though the terminals themselves were designed using different crash testing specifications. It is also important to note that no effort was made to control for highway type or posted speed limit in this study to avoid over-dividing the sample. That being said, however, Maine is heavily dominated by two-lane rural roadways having only two divided highways with a combined length of just over 350 miles. There were 616 crashes where the guardrail end or crash cushion was struck in any sequence of events identified during the study period. After reviewing locations using GoogleEarth, it was determined that 416 records where a code 18 or 28 was specified did have a guardrail end of some type at the location specified (i.e., 200 were eliminated from further consideration because there was no apparent guardrail terminal at the location indicating that either the location or harmful event was mis-coded). Crashes were identified with the following w-beam terminals: BCT/MELT; Buried in back slope; CAT; ET Plus; FLEAT; and SRT; Crashes where also identified with crash cushions and non-crashworthy ends (e.g., trailing ends). In some situations there was insufficient data in the crash report to determine which terminal was present at the specified location (e.g., before or after an access point) was involved in the crash. The absolute risk of observing a fatal or severe injury crash involving a flared terminal in Maine was found and is shown in Table 5. The best measure of hardware performance is when the hardware is the first and only object struck since the resulting injury can be confidently associated with the terminal collision. While the SRT and FLEAT both experienced about the

14 Carrigan, Ray, and Ray same number of first and only harmful events, there were no incapacitating or fatal injury crashes involving the SRT whereas there were two with the FLEAT. The incapacitating and fatal injury percentage for the SRT was zero, the FLEAT was percent and the BCT/MELT was 3.03 percent. The FLEAT, which is a TL3 Report 350 crash tested system, had an incapacitating and fatal injury percentage that was three times higher than the older BCT/MELT. When the evaluation is limited to Report 350 test level three vehicles (i.e., passenger vehicles as identified on the police report), the BCT/MELT still had a lower risk than the FLEAT,3.23 and 5.26 percent, respectively. Table 5. Percent of Fatal or Incapacitating Injury in Police-Reported Crashes Involving Flared Guardrail Terminals. FLEAT BCT/MELT SRT Any Harmful Event Absolute Risk (%) % C.I. ±4.90 ±3.39 ±0.00 Cases Most Harmful Event Absolute Risk (%) % C.I. ±6.83 ±4.22 ±0.00 Cases First and Only Harmful Event Absolute Risk (%) % C.I. ±11.03 ±4.91 ±0.00 Cases First and Only Harmful Event with TL3 vehicle Absolute Risk (%) % C.I. ±8.43 ±5.22 ±0.00 Cases Hard copies of the police reports with the scene diagrams and crash narratives were requested for all of the 15 severe and fatal crashes where terminals were involved in any sequence of the events from the Maine State Police to review the impact conditions. While flared terminals are typically 37.5-feet or more in length, much of the terminal is designed to function as w-beam guardrail would function. This section which functions as w-beam guardrail is generally characterized as the portion of the terminal beyond the length of need (BLON). Crash reports, however, do not typically recognize the BLON as part of the terminal. The BLON collisions, therefore, are rare within most datasets because they are generally coded as guardrail collisions even though damage to the BLON many times requires terminal components to repair the roadside hardware. This retrospective review was able to determine the impact scenario for twelve of these crashes. The results are shown in Table 6. A larger sample is needed to form conclusions, however, even within this small sample, side impacts with the ends of the terminals were observed. Side-impacts are outside of the testing criteria, therefore, the hardware is not designed to function during these types of impacts.

15 Carrigan, Ray, and Ray Table 6. Impact Conditions by Terminal Type for Severe and Fatal Crashes. Noncrashworthy Impact Location FLEAT BCT/MELT Total No. % No. % No. % No. % End On Impact Redirected behind Redirected in front Side Impact All End On Beyond Length of Need Penetrated Redirected Reverse direction Hit From Behind All BLON DISCUSSION AND CONCLUSIONS The preceding analysis of the 2012 through 2014 Maine crash records indicates that the FLEAT results in a higher risk of severe and fatal injury than the BCT/MELT, however, these findings were not statistically significant and the sample size is small. On the other hand, the size affect (i.e., 5.26 versus 3.23) is large and could be important. The FLEAT, which is a Report 350 TL3 crash tested system, had a severe and fatal injury percentage that was almost three times higher than the older BCT/MELT. Conclusive results would require a larger sample and accounting for possibly confounding effects of posted speed limit, highway type and other variables. Each State database, as demonstrated above, certainly have limitations but also present the minimum variables to provide an opportunity for at least a rudimentary review of the inservice performance of roadside hardware. This paper has summarized the evolving crash testing standards as well as factors which should be considered in the assessment of field performance of roadside hardware. It is strongly recommended that prior to the study of the in-service performance of any roadside hardware that the evaluators first become knowledgeable in: (1) the development history; (2) the acceptable uses within the greater jurisdictional area (e.g., USA and Canada); (3) the specified uses within the study area ACKNOWLEDGMENTS The authors wish to thank the Maine Department of Transportation and Maine State Police for providing the data for this demonstration.

16 Carrigan, Ray, and Ray REFERENCES 1 Bronstad, M.E. and Michie, J.D., "Recommended Procedures for Vehicle Crash Testing of Highway Appurtenances," NCHRP Report 153, Transportation Research Board, Washington, D.C., Michie, J.D., "Recommended Procedures for the Safety Performance Evaluation of Highway Appurtenances," NCHRP Report 230, Transportation Research Board, Washington, D.C., H. E. Ross, Jr., D. L. Sicking, R. A. Zimmer and J.D. Michie, Recommended Procedures for the Safety Performance Evaluation of Highway Features, Report 350, National Cooperative Highway Research Program, Transportation Research Board, Washington, D.C., Technical Committee for Roadside Safety, Manual for Assessing Safety Hardware, American Association of State Highway and Transportation Officials, Washington, D.C., Task Force for Roadside Safety, Roadside Design Guide, American Association of State Highway and Transportation Officials, Washington, D.C., Baxter, John R., (November 17, 2005). INFORMATION: In-service Performance Evaluation and Continuous Monitoring of Roadside Safety Features [Memorandum HSA-10]. Washington, DC: Federal Highway Administration, mo111705/ 7 AASHTO, Guide Specifications for Bridge Railings, American Association of State Highway and Transportation Officials, Washington, D.C., Everett, Thomas, (January 7, 2016). INFORMATION: AASHTO/FHWA Joint Implementation Agreement for manual for Assessing Safety Hardware (MASH), Washington, DC: Federal Highway Administration, implementation_agmt.pdf 9 Maine Department of Transportation Highway Design Guide, Volume 1, December 2004, accessed online accessed January Maine Department of Transportation Engineering Instruction, Guardrail Terminal Policy, August, 2014, accessed online uardrail%20policy% doc, accessed January, Ray, M. H., J. A. Weir, J. A. Hopp, In-Service Performance of Traffic Barriers, National Cooperative Highway Research Program Report No. 490, National Academy of Sciences, Washington, D.C., 2003