U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000


Skip to content
Facebook iconYouTube iconTwitter iconFlickr iconLinkedInInstagram

Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations

Report
This report is an archived publication and may contain dated technical, contact, and link information
Publication Number: FHWA-HRT-06-121
Date: November 2006

Long-Term Pavement Performance (LTPP) Data Analysis Support: National Pooled Fund Study Tpf-5(013)

Chapter 13. Summary and Conclusions

To meet the objectives of this study, the research was separated into three basic steps. The first step was to develop models to quantify the effect of the environment, particularly related to deep frost or multiple FTCs, on pavement performance in terms of pavement distress. The second measure was to look at pavement design and materials standards that compensate for or mitigate the effect of seasonal frost. The third step was to evaluate costs associated with pavement design elements considering frost-related effects. Supplementary to these efforts, the application to mechanistic-based pavement design was also addressed.

Findings from phase one of this study were incorporated into the work performed in phase 2. As part of phase 1, a literature review was conducted to guide and supplement the analysis. It was also discovered in phase 1 that the dataset necessary to represent the number of variables required in the analysis needed to be expanded from only SMP sites to include other LTPP sections. GPS-1, GPS-2, GPS-6, SPS-1, and SPS-8 experiments were used in the dataset for flexible pavements, while the rigid pavement dataset consisted of GPS-3, SPS-2, and SPS-8 projects. Because frost penetration data are not directly measured at all of these sites (as is the case for SMP projects), surrogate factors were investigated. FI was found to represent relative frost conditions in the dataset, and that information was used in place of frost penetration. The analysis database developed for phase 2 included data to represent the following factors:

The dataset included test sections located in the deep-freeze wet region, deep-freeze dry region, and the no-freeze wet region. The data were not separated by climatic setting, and they were analyzed independently to reduce the number of observations per grouping and because the information is largely dependent on the method used to group the data. Rather, all data for one pavement type were combined. Climatic differences were accounted for through the use of explanatory variables in the regression analysis. Test sections with surface treatments were excluded from the analysis. This resulted in more than 520 and 280 test sections for flexible and rigid pavement structures, respectively.

This study considered the following performance data:

Some of the distress types were excluded from further analysis because of the limited number of nonzero observations or because the variation at one site over time was relatively large. This included raveling and durability cracking.

Models were developed using multivariate regression analysis through an iterative process. All explanatory variables were investigated through the use of P-values to determine if their contribution to the prediction was significant. Variables that were marginally significant were included in the model only if they improved the prediction capability of the regression. In addition, various transformations and relationships between the performance measure and pavement age were investigated. The models with the smallest deviation were selected for use.

Some of the models resulted in predictions that were not logical based on general engineering experience. For example, the strain model estimated a reduction in strain as the pavement age increased. Models exhibiting these types of errors were excluded from the performance comparison evaluation. Table 35 provides an overview of the selected models including R-squared data.

Table 36. Overview of developed performance models.
ModelPavementModel TypeR-squaredObservations
RoughnessFlexibleregression (shifted)0.784,544
RoughnessRigidregression0.782,652
Rut DepthFlexibleregression0.451,966
FaultingRigidregression0.471,384
Fatigue and Wheelpath CrackingFlexible-deductlogisticNA1,977
regression0.631,486
Fatigue and Wheelpath CrackingFlexible-percentlogisticNA1,977
regression0.491,481
Transverse CrackingFlexiblelogisticNA1,920
regression0.711,077
Longitudinal CrackingRigidlogisticNA475
regression0.38240
Transverse CrackingRigidlogisticNA489
regression0.54228

Using these models, performance comparisons were made for the following five climatic scenarios. Performance curves, as a function of pavement age, were evaluated and mean comparisons were made using 95 percent confidence intervals at select pavement ages, generally between 20 and 25 years. These comparisons were used to identify statistically significant performance differences.

In addition, performance predictions in each of the PFS were evaluated using typical climates found in each SHA. It was also observed that regional environmental conditions can vary widely in one state. Climatic differences in each PFS were presented as well.

The PFSs were asked to provide information on the pavement design they would use for a standard primary and interstate highway with set design parameters as well as the material specifications, test procedures, and costs associated with those designs.

The questionnaire responses produced a wide variation in pavement design sections for essentially the same design parameters. The PFSs did not identify any particular treatment in their designs that addressed frost effects other than one reference to the frost heave design procedure contained in the AASHTO Guide for Design of Pavement Structures.(1) In a secondary query, it was found that several of the northern SHAs did require that frost-susceptible subgrade soils must be removed as part of their construction specifications.

The project also called for the review of practices in adjacent States to see if any treatments could be of use; however, only limited contacts were provided. To augment this process, existing research reports or ongoing research for all SHA were reviewed from Web sites for anything available for this process. This query produces nothing to add to that already found for the PFS and the earlier literature review.

Many of the northern SHAs add additional untreated frost-free surfacing as part of their pavement designs based on the maximum measured frost depth.(12) For some SHAs, frost-susceptible subgrade soils are removed and replaced with frost-free material for depths ranging from 0.61 to 1.22 m (2 to 4 ft) as part of their normal construction requirements to eliminate the need to consider frost depth in the design procedures. There is no real way to show the relative value of the extra depth of frost-free material other than to note its widely accepted use. The extra surfacing depth is probably already accounted for in the pavement performance models developed considering that many SHAs follow that practice, and GPS test sections represent standard SHA design procedures. It may be one of the reasons for the longer service lives observed in fatigue cracking on flexible pavements in the moderate-freeze environments, which would have the extra surfacing compared to the wet no-freeze environments that would not have the extra surfacing.

A possible consideration was noted relative to the use of a uniform surfacing section across the full roadway prism. For those SHAs that use frost-susceptible subgrade soils in part of their shoulder section or outside their shoulder surfacing materials, there is the potential for differential frost heaving. The extent to which this is a problem or the effect it has on pavement performance could not be quantified. Any SHA that has experienced differential frost heave may wish to look more closely at the capillary flow of moisture into the roadway section during the advancement of the freezing front and onset of frost.

The possible use of surface seals was also noted in the northern tier States where highway volumes were low enough to allow their use. This treatment did not seem to be practical in SHAs where chip sealing is not usually done, particularly on higher volume roadways.

Little was noted in the area of material specifications. Prior to Superpave, the researchers would have expected to find different (softer) binders used in the more Northern States as well as the use of a finer, lower-void mix compared to the Southern States. In the response provided by the PFSs, most SHAs have or are in the process of switching to Superpave. The use of the Superpave binder specifications should improve cold weather performance, a major consideration in the development of those specifications. The use of Superpave mix design procedures has to a large extent eliminated local adaptations in mix design procedures and specifications that might have provided improved performance in areas of deep frost penetration or numerous FTCs.

Very little was found in the SHA specifications that indicated a trend in specifications that would have had an impact on mitigating frost effects. What variation there was seemed most likely due to the availability or type of aggregate found in the State.

An economic evaluation was conducted, which consisted of computing equivalent uniform annual costs and present worth costs using deterministic and probabilistic LCCA. Standard cross sections were developed for interstate and primary highways using the 1993 AASHTO Pavement Design Guide.(1) LCCA was performed using these standard sections for all of the five regions. However, because many northern SHAs use additional depths of frost-free material to mitigate frost effects, an additional LCCA was performed. In this analysis, the roadway sections for the deep- and moderate-freeze regions included additional unbound base course thickness while the no-freeze region remained unchanged from the standard AASHTO design.

The differences between the costs using the standard sections were relatively small and well within one standard deviation when considering the distribution of the data. Using the mitigated section, which is more representative, resulted in the no-freeze region having life cycle costs that were less than the costs of the other regions and fell outside the range of one standard deviation.

The application of the results of this study to M-E design procedures was also explored. Potential uses of the models in implementing the NCHRP 1-37A Guide design procedure were also discussed. The use of additional frost-free material incorporated either in the pavement design or as a specific subgrade treatment can be accommodated in mechanistic design procedures. In most mechanistic design procedures there is a process that accounts for changing subgrade stiffness experienced throughout the year. In those areas with significant thaw weakening, the amount of time that the subgrade is weaker as well as the amount of weakening that occurs during the spring thaw is input into the program. In these cases, the added frost-free layers would be considered as either an improved subgrade or added subbase in the program. For instance, in the basic pavement design used in the economic analysis in the previous chapter, the AASHTO design procedures produced a pavement section consisting of 150 mm (6 inches) of ACP over 205 mm (8 inches) of untreated base for the design considerations included in the PFS design questionnaire.

In an area where there is 1 m (3 ft) or more of frost penetration during the winter, that subgrade would weaken significantly during spring thaw, possibly as much as 50 percent. That weakening is addressed in the 1993 AASHTO guide by using the effective modulus to account for the reduced springtime stiffness. In a mechanistic-based design procedure, the increased damage accrued during the spring weakening of the subgrade would be accounted for in the program by an increase in stresses incurred during that period as a result of the lower stiffness. If the subgrade soil was replaced with 0.61 m (2 ft) or more of frost-free material, as is done in several of the PFSs, then the stiffness of the subgrade can be increased because of the replaced layer, and the reduced spring weakening effects can be eliminated or significantly reduced. Combined, these inputs would reduce the tensile stresses on the bottom of the pavement and reduce the compressive stresses on the top of the subbase as well as the subgrade. The results should be demonstrated in longer service life for the pavement section being analyzed.

Probably of greater significance is the potential use of the models from this study to help provide a regional calibration for the 1-37A mechanistic-based design procedure. The 1-37A M-E design program uses nationally calibrated damage trends, but the NCHRP 1-37A Guide recommends that the users consider a regional calibration- verification procedure. For those SHAs that do not have the regional data to support the activity, the potential use of the models developed in this project as well as the general procedures to develop regional calibration factors were described.

There is also a potential for using the models developed in this study to augment SHA-collected data for the development of a family of curves for regional use in PMSs. Where an agency does not have sufficient regional data to develop project-specific pavement deterioration curves, the models from this study could be used to develop a family of curves that would fit the local environment. It is not envisioned that these models could be used in place of SHA-specific data but, in the beginning—during the implementation of a PMS—the models could be used to make an initial trial of developing the pavement deterioration trends used in the PMS. As with the implementation of any PMS, as data become available, the SHA should refine the performance trends based on its own actual experience, but these models would provide a very good starting point.

Previous | Contents | Next

Federal Highway Administration | 1200 New Jersey Avenue, SE | Washington, DC 20590 | 202-366-4000
Turner-Fairbank Highway Research Center | 6300 Georgetown Pike | McLean, VA | 22101