Verification, Refinement, and Applicability of Long-Term Pavement Performance Vehicle Classification Rules

Chapter 10. Summary Conclusions and Recommendations

This project examined the performance of the rule set developed by the LTPP Traffic ETG for classifying vehicles at WIM sites. This chapter summarizes the findings and recommendations of the project.

Vehicle Class Rule set Comparison

The LTPP WIM rule set was compared with 10 different classification rule sets from eight States: California, Florida, Michigan, Washington, Wisconsin, Missouri, Ohio, and Virginia. Some of the State rule sets were designed to operate only as automatic vehicle classifiers (i.e., without access to axle weight data), and others were designed to work with WIM systems (i.e., allowing the use of axle weight data in the vehicle classification process). California and Florida both supplied one of each type of rule set.

At one level, all the rule sets tested had many similar characteristics. However, at the same time, the LTPP WIM rule set differed to at least some degree from each State rule set. Those differences focus on the following four areas:

• Use by the State rule set of axle or gross vehicle weight information (the LTPP rule set does), and if so, what boundaries are used to separate vehicle classifications.

• Use by the State rule set of different boundary conditions for separating two classes of vehicles with similar numbers of axles.

• Consideration by the State rule set of specific vehicles that are common to that State (and that are not explicitly considered in the LTPP rule set).

• Processing of axle weight and spacing rules in a specific order by the State rule set. This usually means that the rule set has vehicle classification definitions that overlap for two or more vehicle classes. When this occurs, the processing rules are designed so that a vehicle falling into an area where definitions overlap is always assigned to one specific classification. The original LTPP WIM rule set does not require processing rules in a specific order.

The details of each of these types of differences, as they occur in each of the tested State rule sets, are discussed in chapter 3 of this report.

Because of these differences in classifying vehicles based on their axle configurations, no two classification rule sets produce the same volume estimate for all 13 FHWA vehicle classes. However, there is no simple answer to the question, "How does truck volume by class and the associated load spectra for those vehicles vary if a classification rule set other than the LTPP rule set is used?"

When compared with the LTPP classification rule set, different State classification rule sets shift vehicles of different characteristics from one FHWA vehicle class to another. Thus, without specifically examining the State classification system used to collect the traffic volume (by class) count, it is not possible to predict which truck classes will gain or lose volume if a State's classification rule set is used instead of the LTPP rule set.

However, even understanding the differences in how two classification rule sets are designed does not allow accurate prediction of the magnitude of truck volume changes, nor how those volume changes would affect the load spectra that apply to those trucks. This is because traffic characteristics tend to vary enough from site to site (even across multiple sites within a single State) that the percentage of vehicles that change FHWA classes when different classification rule sets are applied also varies considerably from site to site because of how fleet characteristics change from site to site, even within a single State.

The wide variety of changes observed in truck volumes given the application of any specific State classification rule set is further reflected in the axle weight spectra produced from those truck volumes. When these differences in load spectra are combined with the volume changes observed in individual truck classes, it appears that significant differences in predicted pavement loads for some specific vehicle classes may occur. However, these differences are likely to affect pavement design only when the vehicles in question are heavy (in at least one of the two rule sets) and constitute a moderately large percentage of the total load at that site.

Nature and Size of Loading Errors Resulting From Use of Different Classification Rule sets

Project analyses show that the use of two different classification rule sets to develop a pavement loading estimate (one rule set for producing the traffic volume by vehicle classification estimate and the other rule set to produce the NALS used in the pavement analysis traffic load computation) can result in significant error in the pavement loading estimate. Load errors occur mostly in vehicle Classes 4–8, 10, and 13. The size and significance of that error is highly variable, depending not only on which classification rule sets are used but on which site is considered. However, the sites that have a high percentage of heavy Class 9 vehicles or light Class 5 vehicles are not likely to be significantly affected by these errors.

The effect of using the LTPP load spectra and State-specific classification count data on the accuracy of the total traffic load estimate is nearly impossible to predict without performing a detailed analysis for each site and each State classification system. To illustrate the complexity of making this estimate, assume that the State rule set in use classifies all cars pulling trailers as trucks, and the LTPP rule set classifies no cars pulling trailers as trucks. If a road is restricted to heavy trucks and experiences no cars pulling trailers, then this difference in rule design has no effect on the truck volume estimate. On the other hand, if the road experiences very heavy recreational vehicle traffic and little heavy truck traffic, there will be many such vehicles, and the State rule set will badly overestimate the volume of trucks. This overestimation of truck volume when multiplied by the LTPP load spectra will overestimate the traffic load created by that class of vehicles.

Sensitivity of the MEPDG Pavement Design Models

Project analysis results indicate that, for pavement designs typical for RI designs subjected to a high volume of heavy truck traffic (and which typically consists of 75 percent or more of Class 9 vehicles), differences in vehicle classification rule sets, compared with the LTPP classification rule set, are likely to have very little practical impact. For these cases, a combination of vehicle classification data collected using the non-LTPP vehicle classification rule set with load spectra obtained from data collected using the LTPP classification rule set should not result in significant errors in either MEPDG or AASHTO 93 design and analysis outcomes, provided that the selected load spectra (surrogate from nearby sites or defaults) accurately describe the expected traffic loading at the site.

On the other hand, for roads that experience lower truck volumes and low percentages of Class 9 trucks, using different classification rule sets for the collection of truck volumes and load spectra can result in moderate differences in pavement performance predictions.

Because of the complexity of the relationships being explored, the researchers highly recommend that readers examine the details of chapters 5 and 6 before using multiple vehicle classification systems for volume and load spectra data when performing pavement analysis on roads with thin pavement designs and low truck volumes.

This study found the following answers to the questions stated in the research objective:

• Can LTPP GPS sites use WIM data from the SPS sites using the LTPP classification rule set?

  • Based on the analysis of worst-case scenarios of classification rule set differences, it can be concluded that LTPP GPS sites can use WIM data from the SPS TPF sites, with the exception of PCC pavements constructed on low truck volume roads that have unusually heavy Class 9 axle load spectra. It is not likely that many LTPP PCC sites will fall in this category.

  • Prior to applying load spectra based on SPS TPF sites to a GPS site, it is recommended that the loading condition at the GPS site be well understood (at least in a descriptive form, such as typical, lighter than typical, heavier than typical, etc.). Once the loading condition is identified, the appropriate default or surrogate load spectra from SPS TPF sites can be used.

• Does it make any practical difference whether the various class rule sets used at the GPS sites differ from the LTPP classification rule set?

  • The impact of potential differences in classification rule sets with regard to Class 5 and 8 vehicles was investigated. The analysis results indicate that, for traffic conditions observed at LTPP GPS sites, no practical difference in pavement design outcomes is likely to result from inclusion of lightweight Class 5 and 8 vehicles using non-LTPP classification rule sets. Pavement design sensitivity results indicate that Class 5 vehicles are too light to produce any significant damage due to additional Class 5 volume. Class 8 volumes and percentages observed at LTPP sites are not enough to make significant contributions to pavement damage. A couple possible exceptions are GPS sites 16-3023 and 6-3010. Vehicle classification data and expected loading conditions for these sites should be further investigated to determine whether high Class 8 volumes are attributed to lightweight vehicles.

Disclaimer

The analyses presented in this report cover a limited number of traffic scenarios (volume, class, load spectra) and pavement design scenarios. Based on the traffic data obtained from SPS TPF sites, it is expected that the analyses represent the worst-case scenarios in terms of predicted pavement design differences. However, the percentage of misclassified vehicles depends greatly on a specific vehicle stream observed at a site, creating an indefinite number of misclassification scenarios. It is not practical to test the MEPDG sensitivity to all possible combinations of site-specific vehicle streams and vehicle classification rule sets.

Refinement of the LTPP WIM Rule set

In general, the LTPP rule set does an excellent job of classifying vehicles at all of the 18 TPF sites included in this analysis. It is particularly good at differentiating between passenger vehicles pulling trailers and real trucks. The few errors it does make are caused primarily by the inability of current WIM systems to detect differences in the number of tires on the rear axles of single-unit trucks. This factor is often the only distinguishing characteristic between some passenger vehicles pulling trailers and light trucks pulling trailers because these vehicles share similar axle spacing characteristics.

Because these vehicles have similar axle spacings, changes to the LTPP rules are unlikely to improve on the current performance of the rule set because shifts in axle spacing parameters or axle weight parameters will likely create as many errors as they resolve.

The project team identified three minor limitations in the LTPP rule set:

• The Class 7 definition needs to be extended to allow inclusion of Class 7 vehicles with six or seven axles.

• The Class 10 definition needs to be extended to allow inclusion of large double-unit resource hauling vehicles and tractors pulling very heavy low-boy trailers, with seven and eight axles.

• The Class 13 definition needs to be extended to allow inclusion of Class 13 vehicles with 10, 11, or 12 axles.

Although not all of these vehicle types are present at all TPF test sites, at least some of these vehicles were present at each test site. In addition, all of these vehicles have the potential to be extremely heavy and thus contribute significantly to estimated pavement damage, even if they are not present in large numbers. As a result, additional rule definitions were implemented within the LTPP WIM rule set.

The vehicle classification algorithm for the refined LTPP WIM system was tested at three SPS test sites in Pennsylvania, Maryland, and Tennessee. The evaluation found that the refined WIM algorithm works as intended. No errors were observed with the vehicles tested. The researchers recommend that the new algorithm should be installed in LTPP WIM sites.

The refined LTPP WIM rule set is shown in table 35. In the revised rule set, there is now some overlap in the vehicle definitions for Classes 7, 10, and 13. As a result, the LTPP rules should be applied in the order shown in the table (lower numbered classes first). That is, if a vehicle fits into a lower numbered classification, that vehicle should be classified in that lower class of vehicles. For example, if a vehicle could be classified as either a Class 10 or a Class 13 vehicle, it should be classified as Class 10.

It is further recommended to revise MEPDG axle loading defaults (NALS and axle per class values) based on LTPP TPF data once all data for Classes 7, 10, and 13 are reprocessed using the new rules. The LTPP database should also provide means for recording weights of axles groups with five or more axles.

This classification rule set is applicable nationwide for any WIM scale used for LTPP data collection. The rule set can be modified by States to also collect data on State-specific vehicle classifications (e.g., triple trailer trucks in Oregon), in which case it can also be used for State-specific WIM data collection. This rule set is not applicable for use in AVC equipment that does not collect axle or wheel weights.

Table 35. Refined LTPP WIM Rule Set.

- Indicates not applicable

Min = Minimum

Max = Maximum