aTest numbers refer to last two digits in Test Designation for each Test Level unless otherwise noted.

FIGURE 1(a) Vehicle underride of three beam - w-beam transition.

FIGURE 1(b) G1 cable guardrail crash test.

FIGURE 2 Examples of poor performance of breakaway/yielding supports.

FIGURE 3 Gating/non-gating considerations(2).

FIGURE 4 Gating/non-gating considerations.

FIGURE 5 Debris after crash tests.

FIGURE 6(a) 20 mph call box support crash test.

FIGURE 6(b) Two other call box support crash tests.

FIGURE 7 NHTSA passenger compartment measurements from crash tests.

Figure 7(b) Dummy Measurement Data For Front Seat Occupants.

FIGURE 8 Passenger compartment deformation measurement.

FIGURE 9 Barricade crash test resulting in driver vision problem.

FIGURE 10 Test of G9 guardrail with 8000-lb. van at 60 mph and 25 deg.

FIGURE 11 UMTRI fatal truck accident data (4).

FIGURE 12 MB3 box beam median barrier tests using 1800-lb. cars at 60 mph.

FIGURE 13 Unsatisfactory test results, but the longitudinal occupant risk criteria values were not exceeded.

FIGURE 14(a) Occupant risk test of MB3 (20-deg angle).

FIGURE 14(b) Occupant risk test of MB3 (20-deg angle) continued.

ALTERNATIVE TECHNIQUES FOR EVALUATING ROADSIDE SAFETY SYSTEMS
Dean Sicking
University of Nebraska

The performance of roadside safety devices has traditionally been evaluated through full-scale crash testing. Although full-scale crash testing has proven to be an effective method of evaluating the impact performance of roadside safety systems, the associated high costs of research have severely limited development of new systems. A number of alternative evaluation procedures have been developed in an effort to reduce costs. The alternate procedures include using scale models, surrogate vehicle testing, and computer simulation. Each of these techniques is described below and examples showing successful application are included.

Scale Models

Scale modeling involves building and testing small-scale models of a device to estimate the response of a full-scale system subjected to similar conditions. Forces and accelerations measured during these tests must be scaled up to evaluate the behavior of full-scale systems. Developing the appropriate scaling factor is the greatest obstacle that must be overcome in order to make practical use of scale model testing. This process involves first identifying all system characteristics that control forces on the device to be studied. As an example, consider scale model tests of a cylindrical rubber energy absorber. The energy absorbed by a rubber cylinder is controlled by the element diameter, wall thickness, length, and the rubber's modulus of elasticity. The scaling factor relating model tests to full-scale prototype performance was found to be:

where:

D = outer diameter

d = inner diameter

L = length

E_o = initial Modulus of Elasticity

Scale model testing of rubber cylinders proved to be a very valuable tool for designing crash cushion energy absorbing elements.

Scale modeling can be used effectively for this and many other design problems where the number of design parameters is relatively limited and vehicle characteristics are not a consideration. However, when vehicle characteristics are important or when the number of important design parameters becomes increases, scale modeling becomes very difficult and as a result very costly. Consider for example, scale model testing to evaluate guardrail design. A list of factors that can have an important effect on guardrail performance during a vehicular impact are shown in Table 1. Developing scaling factors for such a large number of factors is functionally impossible. Therefore, scale model testing cannot be used as a substitute for compliance testing. However, as described previously, scale model testing remains a valuable tool for use in development of new safety devices.

Surrogate vehicle testing involves impacting a safety device with a pendulum or bogie vehicle. Pendulums and bogie vehicles incorporate a rigid frame with a crushable nose as shown in Figure 1. The crushable nose is constructed from honeycomb aluminum and is designed to replicate the front stiffness of a small car. As shown in Figure 1, the crushable nose systems are designed only for frontal impacts with relatively narrow objects, such as breakaway sign and luminaire supports. Thus, these vehicles cannot be used to measure the safety performance of longitudinal barriers and other safety devices subjected to oblique impacts. However, a serious question still remains regarding the appropriateness of using surrogate vehicles to test breakaway structures.

The crushable nose system currently used on breakaway devices was developed to replicate a 1981 Volkswagen Rabbit. The design of mini-size automobiles has changed dramatically during the last 16 years, especially with regard to energy management in the frontal structure. Whereas the Volkswagen Rabbit had a steel bumper and steel frontal structure, most modern mini-size cars are constructed with Styrofoam filled plastic bumpers and plastic frontal structures. The extensive use of plastic and Styrofoam makes these vehicles much softer than older designs.

Note that surrogate vehicles are most commonly used to test breakaway devices. These devices are designed to remain essentially rigid during the initial stages of impact. As the front of the impacting vehicle is crushed, forces on the breakaway mechanism increase. When impact forces reach sufficient levels to activate the breakaway mechanism, the structural support is knocked out of the path of the vehicle.

The performance of a breakaway device is normally measured in terms of velocity change during the impact. Vehicle crush is the primary source of energy dissipation during these impacts and has a strong effect on the velocity change during such a test. If the vehicle crush force versus depth of penetration is fairly linear, the vehicle crush energy can be expressed in terms of the activation force for the breakaway mechanism:

Where:

As shown in this equation, vehicle energy dissipation is inversely proportional to vehicle stiffness. Soft vehicles have much higher crush energy dissipation than do stiff vehicles. Thus, a softer frontal structure would have a much higher velocity change during an impact with a breakaway device than a vehicle with a stiffer frontal structure. Therefore it is imperative that if surrogate vehicles are to be used to evaluate the performance of breakaway structures, the frontal crush stiffness needs to accurately reflect the design of modern small cars. It is unacceptable to use the current crushable nose which replicates a 16 year old vehicle.

Although surrogate vehicle testing can accurately predict vehicle velocity change if an appropriate frontal crush stiffness is used, there are still other concerns that should be addressed. Many breakaway systems, such as breakaway luminaries, utility poles, and call boxes, have the propensity for falling on top of impacting vehicles. Serious injuries have been reported when heavy supports crush the top of a vehicle occupant compartment. Since bogies and pendulums do not have a roof structure, it is impossible to evaluate the seriousness of a structural support falling on to one of these vehicles. Therefore, when a structural support falls onto a surrogate vehicle, a full-scale crash test must be conducted to accurately assess the safety of the breakaway device.

Another significant shortcoming of surrogate vehicles is the absence of a true suspension. Some full-scale tests of breakaway devices have resulted in vehicles rolling over due to snagging on the post stub or inappropriate bending of the structural support. These rollovers cannot be accurately replicated by either a pendulum or most bogie vehicles in use today. Pendulums are suspended from a structure and therefore cannot rollover. Although bogie vehicles can be used to replicate high speed impacts, the suspensions used on most of these devices to not attempt to replicate an automobile. As a result, bogie vehicles cannot assess the propensity for rollover either.

Computer simulation of roadside safety devices involves using nonlinear finite element analysis programs to model ran-off-road accidents. There a three primary applications of this technology, component design, test extrapolation, and replacement of full-scale crash testing. As summarized below these alternatives are listed in order of increasing difficulty and cost.

Component design using computer simulation involves modeling critical elements of a roadside safety structure. One example of this technique was a recently completed study to develop a modified fuse plate design. A model of the fuse plate was developed as shown in Figure 2. LS-DYNA was then used to model the force deflection characteristics of the fuse plate under typical impact conditions. The computer simulation allowed researchers to model many combinations of materials, plate thicknesses, and perforation geometries to select an optimal design. The candidate design was then tested dynamically to verify simulation findings. After completion of the dynamic testing, full-scale crash tests were conducted to verify the overall performance of the new breakaway mechanism. In this application, computer simulation increased the number of design alternatives that could be evaluated and greatly reduced the cost of the evaluation for each alternative.

Development of the SKT-350 guardrail terminal is another example of using computer simulation in the design of roadside safety devices. The impact head for this energy absorbing guardrail terminal was designed using LS-DYNA. The simulation effort involved designing the deflector system to control the energy dissipation rate. For this simulation, a model of the deflector system being pushed down the guardrail was used to predict the force deflection characteristics of the new energy absorbing head. Deflector system geometry was modified until the energy absorption rate was tuned to a desirable level. Static testing was used to verify the energy dissipation rate prior to the start of full-scale crash testing. In this case, simulation replaced an extensive static test program and thereby greatly reduced the time and cost involved in the development effort.

Bogie testing was undertaken to verify the full-scale impact performance of the impact head and although results verified simulated energy dissipation rates, these tests also indicated that the structure of the impact head needed reinforcement. LS-DYNA was again used to redesign the impact head. In this case, a model of the impact head, shown in Figure 3, was subject to high speed impact with a model of a 3/4 ton pickup. The size and configurations of impact head reinforcing plates were then modified until the simulation indicated that no damage would develop during high speed impacts. Full-scale crash testing of the new terminal design was then undertaken with excellent results. Only six full-scale crash tests were required to meet NCHRP Report 350 safety performance standards. The computer simulation was instrumental in achieving the development goals at a greatly reduced cost compared to conventional design procedures using component and full-scale testing. Thus, computer simulation is a very valuable tool for designing components of roadside safety devices.

Another application of computer simulation is in the area of extrapolation of full-scale crash test findings. The high cost of full-scale crash tests greatly limits the number of tests that can be conducted to investigate the safety of any given device. Further, crash testing procedures are limited to tracking type of impacts, while accident data indicates that 50 percent of all ran-of-road accidents involve non-tracking vehicles. Finally, roadside safety devices are seldom installed in the field under the same ideal conditions used in testing. For example, roadside slopes placed in front or behind a safety device can greatly affect its safety performance. Thus, full-scale crash testing cannot explore all of the impact scenarios that may lead to injuries or fatalities in real worlds accidents. In these cases, computer simulation is the only available method for fully exploring the performance of a roadside geometry or a safety device.

One of the best examples of this application of computer simulation was a study of vehicle stability on roadside slopes. In this effort, a relatively small number of full-scale crash tests were conducted to explore vehicle stability on roadside slopes. These test results were then used to validate the HVOSM simulation model. This program was then used to explore the stability of a wide range of vehicles encountering many different slope configurations. Findings from this simulation effort were implemented to establish existing roadside slope design criteria.

As computer simulation tools, such as LS-DYNA, become more powerful, it is possible to explore the off-tracking impact performance of roadside safety hardware. Although no major study has been completed to date, a number of efforts are currently underway. In this case, questions about the validity of the simulation programs are less critical, since simulation is the only available method for examining these issues.

Using computer simulation as an alternative for full-scale crash testing is the ultimate dream of safety hardware developers. Extensive validation of the computer simulation tools must be completed, carefully analyzed, and widely accepted before this dream can be realized. Unfortunately, the computer simulation community is still struggling with the question of what constitutes acceptable validation. Thus, using computer simulation as an alternate to full-scale crash testing is still out of reach. Nevertheless, as researchers gain skill, simulation codes are refined, and the speed of computer hardware increases, the dream of using this technology to reduce the need for full-scale crash testing begins to appear feasible.

Three alternatives to full-scale crash testing, scale model testing, surrogate vehicle vehicle testing, and computer simulation have been described. Scale model testing has proven to be a valuable tool in the design of specific components for roadside safety devices. However, developing accurate scale models of a vehicle is too costly to ever offer a practical alternative to full-scale crash tests.

Surrogate vehicle testing, involving pendulum and bogie vehicles, has been accepted as an alternative to full-scale testing of breakaway devices. Unfortunately, the implementation of these surrogate vehicle tests still has some major short comings. The crushable nose system used on all surrogate vehicles is out of date and must be revised to more accurately reflect the structure of modern vehicles. Further, since surrogate vehicles do not have a roof structure, the potential for occupant compartment intrusion cannot be evaluated. When surrogate vehicle testing indicates that a breakaway device will fall on top of an impacting vehicle, full-scale testing is still the only method for verify the system's safety performance. Finally, since bogie vehicles do not have accurate suspension models and pendulums are limited to low speed impacts, there is a concern that surrogate vehicle testing cannot accurately measure the potential for vehicle rollover during high speed impacts.

Computer simulation is now a very valuable tool for design roadside safety devices. Further, it is the only available method for examining many types of impacts that cannot be tested and it can be a valuable tool for extrapolating crash test results to other vehicles or impact conditions. Although computer simulation cannot now be used as an alternative to full-scale crash testing, technological advancements are bringing this alternative within the foreseeable future.

NCHRP Project 22-11 began in June, 1994 and is scheduled to be completed in July, 1998. Project objectives are:

• To evaluate current information on the safety performance of roadside features for each subclass of light trucks,

• To assess the significance of gaps in safety performance information, and

• To recommend priorities for future research, testing, and development needed to ensure that roadside features accommodate light trucks.

The project originated due to the recognition of the significant increases in the use of light trucks on US highways, and the mandates of the Intermodal Surface Transportation Efficiency Act (ISTEA) enacted in 1991 by the US Congress. The ISTEA instructed transportation agencies to design roadside features to accommodate light trucks.

The project is structured to undertake three basic efforts, namely, (1) collection of sales, crash test, and accident data for the light truck class of vehicles, (2) evaluate the effectiveness of the 2000P test vehicles, and (3) examine the impact performance of roadside features and light truck class vehicles through computer simulation, accident analysis, and crash testing. In the initial phase an effort was made to identify and acquire information on the light truck subclasses, including dimensional and inertial properties, sales and sales projections, accident data, and crash test data. A modest simulation study was conducted in the initial phase of the study to examine the performance of light trucks when impacting roadside barriers or when traversing roadside geometric features. Comparisons were also made between the properties of the larger light truck subclasses and the 2000P test vehicle, which is a 3/4-ton pickup truck. Results of the initial phase of the project were presented in an interim report.

A summary of the tentative findings of the project to date are as follows:

• Light truck sales now comprise approximately 40 percent of all passenger vehicle sales in the US. In this project seven subclasses are used to categorize light trucks: mini-vans, large vans (1/2-ton and 3/4-ton), small pickup trucks, full-size pickup trucks (1/2-ton and 3/4-ton), small utility vehicles, mid-size utility vehicles, and large utility vehicles. Light trucks should obviously be a major factor in the design of highway safety features.

• In comparison with the larger light trucks, the 2000P test vehicle appears to be a good, representative design vehicle. Key vehicular properties that influence the impact performance of a safety feature include bumper height, front overhang, total mass, the center of mass location, and mass moments of inertia. Sales of the large pickup truck subclass is also greater than any of the other light truck subclasses.

  · A number of crash tests have been conducted with the 2000P vehicle for test level 3 of NCHRP Report 350. Tests of the widely used median barriers and roadside barriers have been conducted. Most guardrail end treatments and crash cushions have been tested with the 2000P vehicle also. The following features have met test level 3 requirements:

• New Jersey safety shaped concrete barrier

• Single slope concrete barrier

• Various bridge rails of state DOTs

• G1 steel cable guardrail system

• steel box beam guardrail system

• G4(2W) W-beam guardrail system

• ET-2000 (proprietary system)

• BEST (proprietary system)

• Slotted Rail Terminal (proprietary system)

• REACT 350 (proprietary system)

• ADIEM (proprietary system)

• QUADGUARD (proprietary system)

• Sand-filled plastic drums

Flexible barriers

• G2 W-beam guardrail system

• G4(1S) W-beam guardrail system

• G9 Thrie beam guardrail system

Marginal performance was observed in limited testing of segmental, precast concrete barriers having the New Jersey shape. Further testing is needed to evaluate the effects of segment length and joint design on impact performance.

The Highway-Vehicle-Object-Simulation-Model (HVOSM) and the BARRIER VII computer programs have been used to study the behavior of light trucks traversing roadside geometric features (embankments and driveways) and the behavior of light trucks when impacting longitudinal barriers. Results of these studies indicate that the smaller light truck subclasses have a greater propensity for overturning than the larger subclasses of light trucks.

Analysis of the FARS data for the 1991-95 period was made and a summary of the findings are as presented. These findings show that (1) light trucks are more prone to overturn in fatal crashes that cars, (2) there is a high incidence of overturn for all passenger vehicles in fatal crashes involving longitudinal barriers, (3) light trucks are more prone to overturn than cars in fatal crashes involving longitudinal barriers, and (4) there is a high incidence of overturn for all passenger vehicles in fatal crashes involving ditches and embankments.

Ongoing Project activities include the following:

• Additional accident studies-- Analysis of several years of data from the Highway Safety Information System (HSIS) accident data files is being conducted to determine if anything definitive can be derived therefrom relative to the impact performance of roadside features in accidents involving light trucks. The researchers are also examining sample data from the National Accident Sampling System (NASS) General Estimates Systems (GES) to determine if this data base can provide additional information on light truck behavior in accidents involving roadside features.

• Crash testing-- Additional funding has been provided for the project to include a series of tests involving four different light trucks impacting the widely used G4(1S) W-beam guardrail system. In each test the barrier will be impacted at 100 km/h, and at an impact angle of 20 degrees. Vehicles to be tested will include a small utility vehicle, a mid-size utility vehicle, a large pickup truck (the 2000P vehicle), and a large passenger van.

It is expected that the research effort will generate a report that will provide an assessment of the ability of the various roadside hardware to accommodate the light truck class of vehicles. This report will identify those vehicle and hardware combinations where significant incompatibility exists and it research plans to address these needs will be provided. The report is expected to be published in early 1999.

Most of the issues to be mentioned here have already been raised by others. The comments in this paper are intended to add perspective and give emphasis to the issues. At the outset it should be noted that I believe that it is important to involve the American Association of State Highway and Transportation Officials (AASHTO) in any update to Report 350. In a recent meeting with people from the national headquarters of the AASHTO, a major concern was expressed about the FHWA’s taking Report 350, a research report created for AASHTO, and making it a standard. This is what the FHWA effectively did. It is important that AASHTO be involved very heavily from the beginning. Members of the AASHTO Task Force for Roadside Safety, which is responsible for the formulation and oversight of AASHTO policies on roadside safety design, need to play a role in the formulation and review of any changes proposed to Report 350.

One of the questions posed has been what should the tolerances be for test installations. These need to be reasonable, but I don't know what the exact numbers should be. For certain, there is a need to accurately and rather precisely report what was tested and what the test conditions were at the test site when testing was conducted. The test installation tolerances should be related to what can realistically be expected real world installations. It is believed that you can get much closer to most nominal dimensions than two inches (50 mm) on most features. An eighth of an inch (3 mm), might be a little tight for most layout dimensions. Designers, suppliers, users, and testers need to be cognizant of the importance of specifying material and installation tolerances. Those who conduct a test should accurately and rather precisely report the actual properties and dimensions of all elements of the test installation, as well as all other aspects of the conducting of a test, that might be expected to influence the test results. It is also important that the specified service tolerances in the materials and installation of a feature be considered in determining if a test installation of a feature adequately represents what might be delivered and constructed under the specifications. It could be that specified service tolerances will need to be tightened to ensure that service installations of a feature will perform as tested or it may be that tests at the limits of the specifications will be needed. For example, the metals currently used in w-beam guardrails are far better than the specification requirements. On the other hand, the thicknesses of the w-beams are very close to the minimum allowed under the specifications. It is obvious that problems could result if tests are conducted on w-beams of typical metal quality and above minimum thickness and service installations are built with w-beams of either minimum allowable metal quality or thickness or both.

Soil and foundation conditions, which, in testing of the MELT terminal during the past year, have repeatedly been shown to be influential in test results, certainly must be adequately addressed in the design and testing. These must be reported and evaluated in assessing the acceptability of hardware after testing. Currently, there are no prescribed tests for determining test site foundation conditions. Therefore, a way to test the strength of test installation soils and to make assessments of the relevance of the test conditions to expected service conditions needs to be prescribed.

On July 25, 1997, the FHWA issued a new memorandum on the use of Report 350, which replaces a similar, November 12, 1993, memorandum. The title of the memorandum is "Action: Identifying Acceptable Highway Safety Features." Attached to the memorandum is a document, "Background and Guidance on Requesting Federal Highway Administration Acceptance of Highway Safety Features," that describes acceptance procedures and how the FHWA believes they should be applied.

Some surprises might be found in this new memo. One might be the requirement that the wiring in any feature must be similar to the in-service condition, so that the influence of the wiring on the break-away characteristics can be assessed and the potentials of an electrical hazard determined if the feature is knocked down. Another item that might be a surprise is the inclusion of explicit discussion of work zone traffic control devices in the body of the memorandum and its attachment.

Because of the deficiencies in many test reports submitted to the FHWA, the attachment to the memo also includes guidance on what should be recorded and reported when testing features, which is considerably expanded in comparison to that in the 1993 memo cited earlier. All testers and producers should pay particular attention to the emphasis on reporting given in the memorandum’s attachment.

Since Report 350 only gives slight recognition to the use of pendulum testing, it is important to point out that this new memorandum shows FHWA’s continued acceptance of this test method for determining the acceptability of some types of features. The conditions for the use of pendulum tests and the procedures for their interpretation are about as were set forth in the 1993 memorandum.

Report 350 covers reporting requirements and touches on pendulum tests. Nevertheless, it is suggested that the reviewers of Report 350 consider whether the language in Report 350 is adequate to meet the needs indicated in this new FHWA memo. It would be appropriate to add more guidance on pendulum testing if it does not.

Some people have suggested that Report 350 calls for too many tests. On this matter, consideration should be given to how extensively a feature is likely to be used. If there are truly unique conditions were a special feature is needed and it will only be used a few times, it might be reasonable to consider special analyses and abbreviated testing procedures or the excusing of testing altogether, if on the basis of experience reasonably safe performance can be assured under the most severe service requirements. Most of the complaining about the Report 350 testing requirements heard relates to the cost of bringing a new feature to market, one which they hope will have widespread use. Based on the surprises encountered in test programs, it is hard to agree that too much testing is required. On the other hand, there may be more Report 350 tests, for some features, that should be waved because there is a very low probability they will reveal a problem with a specific feature. Nevertheless, because of the surprises often encountered in test programs, efforts should be made to err on the side of safety when considering excusing tests. What we need to do is to perform enough tests to fully assess the likely field performance of a feature. If several thousand copies of a feature, probably with long service lives, are to be installed a year, the cost of testing the feature should be considered primarily on the basis of the relevance of the testing to long term, systemwide safe performance.

A minor issue somewhat related to the number of tests required under Report 350 is gating versus non-gating terminals and crash cushions. This is considered a spurious distinction and very serious consideration should be given to dropping it in any update of Report 350.

Some have questioned the test speeds in Report 350. There are valid concerns about test speeds relative to the new speed limits and the actual speeds on the highways. There is a need for a serious look at this issue in conjunction with the speed-angle combinations used in testing.

With regard to the speed-angle combination, there is a need to be cautious about the "1.2 percent" that King Mak talked about. If all the fatalities occur in that 1.2 percent of the occurrences, that is where all the work should be focused. That percentile has to be looked at a little more carefully because it represents a low percentage of all encroachments or even all reported crashes.

As Dick Powers implied, test vehicles are of concern in the FHWA. They create problems for the replication of test results if the same vehicle isn't used. At least, different models of test vehicles should be expected to produce identical test results. To get comparable answers, there is a need to more specifically specify test vehicles.

On the issue of international compatibility, it is hoped that some vehicle in the pick-up category (the light truck category) could be found that is common to a number of nations. The mini-van is getting to be popular in Europe. If this is true, it might be a good universal test vehicle. It is certainly very popular here—just look in any parking lot. However, it doesn't have characteristics similar to a lot of the sport and utility vehicles and this could be a problem. The basic criterion for the selection of test vehicles and test conditions should, of course, be that they ensure acceptable feature performance on the highways where the tested features are to be used. This could mean, because of the vehicle mixes in various locations around the world, that a universal set of test vehicles will not be possible. However, it would be desirable to start by arraying all the vehicles used in the world to see if there are enough different vehicles common to all the world’s highways to permit the development of a universally acceptable set of test vehicles.

In working on the issue of international testing compatibility the issue of laboratory certification has arisen. However, this issue also has domestic implications. To ensure confidence in all testing laboratories, a lab certification procedure is needed. If the Report 350 update could make some suggestions in this direction, it would be helpful.

It is believed that the Report 350 evaluation criteria acceptance levels should be set as close to what can technically be achieved and can be justified. Report 350 does not come as close to following this philosophy for the design and acceptance of safety features as I would have it. There has been some discussion about increasing the complexity of the evaluation values by using a performance measure that combines longitudinal and lateral vehicle motion to obtain a single performance value. This is believed to be partly the result of purists not liking the simplistic, segregated, bi-directional approach used in Report 350 and partly because computers now make a more complex model easy to manage. A more complex approach is acceptable so long as it does not reduce safety or the understanding of what happens in tests. On the other hand, there can be strong opposition to a sometimes mentioned approach that calls for basing acceptance criteria on the assumed use of available occupant restraint systems, unless there is a related reduction in acceptance values. First of all, to recognize the use of restrain systems at this point in time would effectively mean retreating from a standard of performance that many are fairly confident is achievable within our current technical and economic constraints. In addition, the occupant restraint systems are developed to work at much lower speed ranges than have been chosen for the development of highway safety features. Thus, the efficacy of these restraint systems at current highway design speeds is not known.

The recommended acceptance procedures for highway safety features in NCHRP Report 230, developed by Jarvis Michie, presented "limit" occupant impact velocities and ridedown accelerations, which were adjusted by a factor of safety. The proposed factors of safety were 1.33 for both occupant impact velocities (OIV) and ridedown accelerations for all cited features, except the OIV for sign and luminaire supports, for which a factor of safety of 2.67 was proposed, and the lateral OIV for redirected impacts into traffic barriers, for which a factor of safety of 1.5 was proposed. These factors of safety were selected for the safety they would provide and because they were considered achievable. The limit values were set at levels estimated to produce serious but not life threatening injuries (an abbreviated injury scale 3 injury). In truth, the limit values can almost certainly be expected to produce some life threatening injuries at or below the prescribed test impacts and, as already implied, the test conditions don’t cover even all the legal highway speeds. With the adoption of Report 350, the factor of safety concept was jettisoned and the acceptance levels for all features (except sign and luminaire supports) were pushed to the Report 230 limit values, albeit with suggested lower "preferred" values that were almost immediately forgotten. It is strongly recommended that the safety factor concept be reconsidered in the review and update of Report 350. However, this issue might become moot if a serious attempt is made to design roadside features to address the estimated 25 percent of the fatal crashes into fixed objects that involve vehicles sliding side-on into such objects. This is an important consideration in the update of Report 350.

A number of people have found the go-no go nature of the acceptance values in Report 350 troubling. As Dick Powers pointed out, the FHWA has added some precision to the numbers in Report 350 by accepting an occupant impact velocity of 12.2 m/s, where 12 m/s is shown in the report, and a ridedown acceleration of 20.4 gs, rounded to the 20 gs shown in the report. While this might be considered expanding the envelope of acceptance, it does nothing to remove the sharp demarcation between acceptable and unacceptable, probably just the opposite.

Currently there is no way to score the overall crash test performance of a feature to balance good performance results against other results slightly over the specified limit, nor is there a way to factor in expected costs, constructibility, maintainability, or durability of a feature. It may not be possible to come up with a better acceptance procedure than is currently in Report 350. Making it more subjective would open the door to a variety of abuses. Nevertheless, an effort should be made in the update to ensure that the acceptance procedures do not exclude good, highly cost-beneficial safety features.