Pavement Structural Evaluation at the Network Level: Final Report

CHAPTER 10. RECOMMENDATIONS

The ultimate goal of the project was to establish a reliable measure of the structural condition of all bound pavement layers above the unbound base layer as it deteriorates over time under traffic and environmental loading based on moving pavement deflection technology and measured at posted traffic speeds. Towards achieving this goal, the study focused on the following two activities:

• Evaluating the capabilities of the available TSDDs.
  
  
• Developing compatible analysis methodologies for enabling the quantification of pavement deterioration over time under traffic and environmental loading.

A system approach was followed so that the TSDDs could be effectively used as a tool for network-level PMS applications and decisionmaking. The flowchart of an idealized system is shown in figure 219. There are several alternative approaches for implementation. In the suggested approach, the decisions based on the traditional condition metrics are confirmed and/or adjusted based on the pavement structural condition as derived from the TSDD measurements. In this manner, the pavement community has the opportunity to gradually implement the proposed changes while avoiding abrupt changes to their institutional approaches.

Figure 219. Flowchart. Idealized PMS containing TSDD structural evaluation component.

To deliver a robust system, the level of analysis sophistication should be balanced with the uncertainties of the TSDDs. Since the manufacturers are actively improving the TSDDs, the interpretation and analyses software should also be progressively improved. The following factors were considered in proposing areas of opportunity for improvements to the system in an integrated and balanced manner:

• Accuracy of measurements as defined by the closeness of measured deflection parameters to their true values.
  
  
• Precision of measurements as defined by obtaining similar results on a given section with multiple surveys within a short period.
  
  
• Operational parameters of the devices such as the speed of the vehicle, the spatial frequency of data reporting by the device, environmental conditions, and road geometry without sacrificing the quality of the data collected.
  
  
• Availability of additional information such as pavement structure, surface condition of the road, and applied loading (i.e., traffic) to the pavement.
  
  
• Data analyses in terms of minimum input data necessary to characterize the structural condition of the road, the deflection parameter(s) that best relate to the structural condition, and the optimal algorithm to estimate the structural condition.
  
  
• Data interpretation in terms of the best use of statistical methods to report a representative structural condition over a representative distance with the spatial variability for each representative distance in addition to the mean value.

Within the context of the above factors, the two devices (TSD and RWD) studied in this project were found to be capable of providing reasonably accurate and precise response measurements for flexible pavements. Since adequate data were not available for proper evaluation, the research team is unable to comment on the applicability of the two devices to rigid pavements.

Based on the system approach discussed so far, practical recommendations are provided in the following subsections in terms of the equipment, data collection, and data analysis to optimize the operation of these devices. Both short- to long-term recommendations for improving the robustness of the system as a network-level PMS tool are provided.

10.1 Equipment Recommendations

One of the most desirable attributes of the ideal TSDD for use on flexible pavements is for the device to provide deflection parameter basins (i.e., deflection parameters at two or more points). Due to the many years of legacy research with the FWD, the current preference of the pavement community is to use lessons learned from using vertical deflection basins as opposed to other deflection parameters.

Based on the numerical analyses conducted in this study, at least three sensors located between the center and 18 inches (457.2 mm) from the wheels are desirable for quantifying fatigue cracking of the AC layer. Aside from the spot directly between the two wheels, the desirable specific locations for these sensors depend on the pavement structure. For thinner AC layers (less than 6 inches (152.4 mm)), it is desirable for the sensors to be within 8 inches (203.2 mm) from the center of the wheels, while for the thicker pavements, sensors located between 8 and 18 inches (203.2 and 457.2 mm) from the center of the wheels seem best suited for capturing the fatigue cracking potential of the AC layer.

For subgrade rutting, on the other hand, sensors that are 24 inches (609.6 mm) or farther from the center of the wheels are considered appropriate. From the results of the field studies conducted as part of this project, the uncertainty (precision and/or accuracy) associated with the TSD sensors positioned between the wheels and farther than 24 inches (609.6 mm) may require further improvements. It is difficult to comment on the accuracy of the sensors farther than 24 inches (609.6 mm) since the deflections are too small (i.e., less than the sensitivity of the embedded sensors used in the project). Given the available data, it might be reasonable to assume that this statement is also applicable to the TSDD sensors.

In the short term, the two RWD sensors could be repositioned, perhaps at 4 and 12 inches (304.8 mm) in front of the wheel center, to provide deflections that can be used more readily to assess the fatigue cracking of the AC layer. In the medium term, it would be desirable to decrease the uncertainty associated with the reported deflections so that more advanced analyses can be performed. In the long term, the installation of more sensors should be considered.

In the case of the TSD, deflection parameter measurements are provided at multiple locations (up to nine as of the date of this report). Aside from increasing the number of sensors, the best compromise for the TSD evaluated in this project is to place the sensors at distances of 4, 8, 12, and 18 inches (101.6, 203.2, 304.8, and 457.2 mm) from the center of the wheels for evaluating fatigue cracking and 24 and 48 inches (609.6 and 1,219.2 mm) for quantifying subgrade rutting. In addition, a sensor placed behind the rear axle (e.g., -12 inches (-304.8 mm)) would be useful in capturing the viscous lag in the deflection basin due to the moving load.

Aside from repositioning and/or adding sensors, it would also be desirable in the short term to decrease the uncertainty associated with the models for computing the deflection basin from deflection slopes and to improve the precision of the measurements at sensor spacings longer than 24 inches (609.6 mm). In the medium term, and until the pavement community migrates to developing indices based on deflection slope, the algorithm to convert deflection slopes to deflection basin may be improved. The long-term focus of the pavement community should be in implementing an algorithm that can directly use deflection slopes.

It is also helpful to know the load characteristics applied to the pavement during testing, especially as the analyses techniques become more advanced. The load characteristics can change not only due to the vehicle loading and tire pressure, but also due to roughness of the road and strong cross winds during the operation. The reliability of the analysis in the proposed system approached could improve if the load magnitudes are reported along with the deflection parameters. Even though the dynamic load characteristics were not available to the project research team at the time of field testing, the TSD device is currently capable of collecting load information. The RWD developers should consider providing such information.

The issue of the relative and absolute calibrations of the device should also be addressed by the manufacturers. In the short term, it would be useful to have a set of straightforward instructions to document that the sensors are aligned properly, that they collect data correctly, and that other components (e.g., the GPS unit and the temperature sensors) are functioning and collecting data properly. In the medium to long term, a set of measurement protocols should be devised to facilitate the consideration of the impact of the seasonal variations in the analyses as much as possible.

Finally, it would be advantageous to equip the TSDDs with auxiliary devices for measuring the pavement structure (e.g., GPR), pavement smoothness (e.g., IRI), and pavement surface condition (e.g., high-definition cameras). This additional information can not only be used to make the structural analysis more conclusive, but with proper planning, it can be used for other purposes such as asset management, which could make the surveys more affordable.

10.2 Data Collection Recommendations

The two TSDDs considered in the project collect data densely. However, the data reported are averaged over a certain distance. Theoretically, the shorter the averaging distance, the better the certainty of the analysis will be. However, there is no question that averaging is necessary for State transportation departments to work with a manageable amount of data. In the case of the RWD, averaging is done over 0.1-mi (0.161-km) intervals, while for the TSD, averaging is done over 32.8-ft (10-m) intervals.

Averaging is an effective way of minimizing the amount of random noise in the raw data, but excessive averaging can also mask changes in the signal due to changes in the structural condition of the pavement. The TSDD developers have carefully studied the raw signals from their devices to propose the optimal averaging distances. However, the level of sophistication of the data analysis in the suggested balanced system should be set based on the capabilities of the TSDDs. As such, the uncertainties in the deflection measurements should be delineated from the spatial variation in the measurements due to changes in the pavement structure or condition.

In the short term, manufacturers would report not only the mean deflection parameters but also additional statistical information (such as the standard deviations or coefficients of variation) associated with those mean values and their variability. In that manner, the analyst can judge the level of uncertainties associated with each deflection parameter reported. In the medium term, the level of uncertainties in the measured deflection parameters should be verified through independent research. Such analysis requires access to the raw or spatially averaged data over a short distance (e.g., 3.28 ft (1 m) or less). In the long term, manufacturers would improve their devices to a level that the averaging can be done as part of the analysis and not data collection.

The data collection should ideally be done at the posted speed limit. Based on this study, it seems that the variability in the collected data decreases as the vehicle speed decreases. The best strategy is to collect the data at the lowest practical speed possible without requiring lane closures or causing safety issues. For example, if the maximum posted speed limit is 65 mi/h (104.65 km/h), then data collection at 40 to 45 mi/h (64.4 to 72.45 km/h) may be desirable.

The manufacturer's recommendations concerning calibrations, vehicle warm-up, tire pressure checks, and other vehicle readiness elements should be carefully followed. Also, based on the project findings, the following additional suggestions are provided:

• The impact of the seasonal environmental variations on the TSDD measurements was not evaluated in this project. However, until such study is carried out, the repeat survey of the sites should be carried out during approximately the same time of the year to minimize seasonal effects.
  
  
• The short-term variations in ambient temperature were studied at the MnROAD facility as part of this project. Until a formal study is carried out to establish the allowable range of testing temperatures, a conservative operational temperature range of 45 to 85 °F (7.22 to 29.44 °C) is suggested. The lower limit takes into consideration that the uncertainty in the study measurements were lower for less stiff pavement structure. The upper limit is proposed based on the observation that the precision of the devices (at least for the TSD) became worse as the ambient temperature increased.
  
  
• Testing is not recommended during precipitation or when the pavement surface is wet due to the current sensors used to measure the pavement responses, which are affected by moisture.
  
  
• The impact of the wind was not explicitly studied. However, testing should be carried out with caution in windy conditions, especially when the wind is predominantly perpendicular to the direction of travel. This is especially true if the TSDD-applied load is not being measured on the same side where the deflection measurements are taken.
  
  
• The horizontal and vertical curves did not seem to significantly affect the TSD results. Nonetheless, it would be a good practice to carefully review data collected near extreme curves.
  
  
• As a quality check, repeat data collection is recommended on a small portion (i.e., 5 to 10 percent) of the network under consideration to ensure consistent data are being collected.
  
  
• The frequency of data collection should be further studied by individual State transportation departments by taking into consideration the effective lives of their pavements and their current PMS experiences and practices. Until then, TSDD testing once every 3 to 5 years on average appears reasonable.
  
  
• Based on the initial investment, the daily cost of the operation of the TSDDs is significantly greater than testing with the FWD at present. However, based on the daily productively of the two devices, the costs per mile associated with TSDDs are substantially less than the FWDs. The cost associated with the TSDDs may be further reduced as State transportation departments embrace their use, incentivizing more service providers to become available and the analysis algorithms to become more automated.

10.3 Recommended Data Analysis for Network-level PMS Applications

The suggested TSDD data analysis for incorporation into network-level PMS applications can be summarized in the following four steps:

1. Calculating representative indices for estimating structural condition of pavement: Based on this study, the most feasible parameters are DSI or SCI considering the fatigue cracking of the AC layer as the critical parameter.
   
   
2. Estimating horizontal strains at bottom of AC layer using recommended or user selected index: Based on the available information, the feasible models are recommended in table 80.
   
   
3. Adjusting the estimated strains to a standard temperature: The strains computed in step 2 need to be corrected to a standard reference temperature for consistent evaluation and tracking of the deflection parameters over time (section 8.7). Assuming a standard reference temperature of 70 °F (21.11 °C), the recommended approach entails the following computations:
   
   
   • Compute temperature correction factor based on the temperature at time of the TSDD field measurements and the reference temperature of 70 °F (21.11 °C).
     
     
   • Compute the AC dynamic modulus based on the strains computed in step 2 and the AC layer thickness.
     
     
   • Compute the AC dynamic modulus at the reference temperature of 70 °F (21.11 °C).
     
     
   • Compute the temperature corrected strains using the AC modulus at the reference temperature of 70 °F (21.11 °C).
     
     
4. Establishing structural adequacy of pavements using temperature-corrected strain: Ideally, it would be desirable for the pavement analyst to be able to determine whether different segments of the pavement network are candidates for preservation, maintenance, rehabilitation, or reconstruction or whether they are adequate as they are. As a minimum, the algorithm associated with this item needs to be able to provide information on whether the pavement is structurally sound for the anticipated traffic and whether a lower level treatment can be used to correct any functional deficiencies or require structural treatment. The initial work toward this goal can be found in Thyagarajan et al. amongst others.^(21,66) Abdallah et al. suggests a probabilistic method for this purpose as applied to FWD using artificial neural networks. Similar and/or follow-up work should be pursued by the pavement community.⁽⁶⁷⁾

One of the most critical factors related to data analysis is the spatial averaging of the data. For an effective utilization, spatial statistical analysis and segmentation of the TSDD data are necessary. An effective analysis tool should allow pavement engineers to distinguish the changes in the road segments due to either changes in the pavement structure or the deterioration of the pavement sections. The statistical analysis for segmentation should be done considering the capabilities of the TSDD, and the condition and nature of the pavement structures.

Assuming that the recommendations related to the averaging during data collection presented previously can be addressed, in the short term, a probabilistic structural analysis could be incorporated for a robust segmentation of the roads at the network level. In the medium and long terms, a dynamic and adaptive optimization could be added to assist the decisionmakers.

As stated in the previous two chapters, a single robust universal index that can predict the structural performance of all pavements at the network level could not be identified. For a robust system approach, information about the pavement layer thicknesses and especially the AC layer is required to use the indices identified. Also, as reflected in the previous section of this chapter, additional data are needed to understand measurement variability and to predict future treatment requirements. These data include the following:

• Surface condition of the pavements, including such parameters as IRI and extent of cracking, to check and/or to help rationalize the results.
  
  
• Best estimates of the present traffic and future traffic projection.
  
  
• Ambient condition at the time of testing.

It would be highly useful for the manufacturers and/or the owners of the TSDDs to work towards using multi-function survey vehicles by incorporating systems that are currently part of the automated pavement condition survey vehicles.

10.4 Recommended Future Research

This report represents the first step toward the eventual implementation of a robust system approach for the structural evaluation of pavements at the network level. The previous sections in this chapter provide several specific technical recommendations to the pavement community to collectively improve the TSDD equipment, data collection, and analyses for network-level applications. The following general items are logical follow-up activities to this study.

The first important follow-on activity should focus on further evaluating the results of this project. The precisions reported for the TSDDs are comparable with those reported by others.^(3,19,16) However, the accuracies of the devices have not been reported extensively in the literature. Also, the structural models proposed in this report have not been evaluated using actual field data. The use of data collected by the RWD and TSD for highway agencies throughout the country can be used for this purpose. The data collected by or for international highway agencies (e.g., Denmark, Poland, South Africa, Australia, and Italy) are valuable, too. Of particular interest is the production-level data collected in the United Kingdom for their network-level evaluations over the past few years. In addition to evaluation of the findings, these data will enable the extension of the project results to a broader spectra of conditions, including different pavement structures, environments, and subgrade soils.

Given that the TSDDs have only been commercially available in the recent past, it would be beneficial to monitor the changes in the TSDD deflections seasonally and with the growth of distress over time. For this purpose, the devices should be utilized repeatedly (seasonally) at sites that are being used for LTPP monitoring (similar to or in conjunction with the LTPP program sites). It is acknowledged that this will be a long term project. However, the data collected in that manner, at several diverse sites selected through a careful experimental design, can provide a wealth of information about the frequency of network-level data collection and manner to use the TSDD data. Moreover, making an initial measurement immediately after the completion of the construction of the test sections is not only an efficient way to evaluate the construction quality but also provides a structural capacity datum for the pavement.

Beyond the above recommendations, other potential future research areas under controlled conditions may be necessary to perform the following:

• Confirm the predictive power of the recommended deflection indices through the use of measurements taken by strain gauges at the bottom of the AC layer during TSDD loadings. While the data collected as part of this study served as a first step, there were issues with the functioning of strain gauges, and the data were limited to AC layers of around 5 inches (127 mm) in thickness. Accordingly, additional data for thinner and thicker pavements structures would be preferred.
  
  
• Expand and validate the prediction of subgrade compressive strain to complement horizontal tensile strains at the bottom of AC layer.
  
  
• Validate the utility of the deflection indices through investigation and implementation of their reasonableness for the network-level PMS application proposed in section 10.3.
  
  
• Apply TSDDs to PCC pavements to evaluate their applicability to network-level evaluation of this pavement type, including estimating load transfer and joint condition evaluations.
  
  
• Expand the use of the data from TSDDs with multiple sensors for backcalculation of layer moduli and/or assessment of the condition of individual pavement layers as well as overall pavement structure.
  
  
• Conduct a more indepth study of the impact of ambient conditions as well as pavement structures/conditions on the deflection measurements for finer adjustments of the operational guidelines as suggested in section 10.2.
  
  
• Consider the impact of the non-linear response of the unbound materials in thin pavement structures on the TSDD measured responses.
  
  
• Explore methodologies for the development of structural performance curves based on TSDD derived structural indices measured over time for use in conjunction with current performance curves for predicting future condition and programming projects.

Structural capacity is important and the TSDD technology should be in place within the next 5 years. Some institutional issues should also be considered for the smooth transition of the incorporation of TSDDs into network-level PMS applications. Some of the key issues include the following:

• How to gradually incorporate TSDD data into network-level pavement evaluations of the following:
  
  
  • Transportation agencies that solely make decisions based on the functional aspects of the pavements (e.g., IRI, cracking, rutting, and faulting) without structural testing.
    
    
  • Agencies that utilize FWD for structural testing in addition to functional testing to make decisions.
    
    
• Possible incorporation in the near future of TSDD measurements into the pavement rulemaking of the Moving Ahead for Progress in the 21st Century Act legislation as a pavement condition metric in addition to IRI, cracking, rutting and faulting, and its use in defining pavement performance measures.⁽⁶⁸⁾