Pavement Structural Evaluation at the Network Level: Final Report

CHAPTER 4. DATA COLLECTION AND ANALYSIS WORK PLAN

Based on a literature review and a survey of device manufacturers, owners, and users, the TSD and the RWD were identified as potentially being capable of meeting the project objectives. The term "TSDD" is used to collectively refer to the two devices.

This chapter details the work plan that was developed and implemented for the field trials and subsequent analyses that were carried out over the remainder of the project to accomplish the following objectives:

• To confirm that the TSDDs met a minimum set of specifications related to the structural evaluation of pavements at the network level.
  
  
• To propose processes to incorporate pavement structural information from the successful TSDDs into network-level PMS applications

These objectives and hence the work plan were driven by the analyses that were conducted over the remainder of the project. The work plan was organized into the following sections: analysis methodologies, field trial locations, testing sequence, schedule, and summary.

4.1 Analysis Methodologies

Rada and Nazarian developed manufacturer-independent specifications required for a number of pavement applications using TSDDs.⁽¹⁹⁾ The authors also provided a comprehensive list of future research needs geared at improving the accuracy and repeatability of the moving pavement deflection testing devices. Those needs were grouped into three major categories: equipment-related, measurement-related, and application-related issues.

The pavement application from the Rada and Nazarian study most relevant to this project is the determination of overall pavement structural capacity indicators/indices. This application serves to assess the overall pavement structural capacity in terms of index values or structural remaining life, and ultimately to support the development of M&R decisions and cost estimates based on structural indicators (comparable to approach used within MicroPAVER^TM and other PMSs).⁽²⁵⁾ However, only the assessment of pavement structural capacity was considered in this work plan, as the development of M&R decisions and costs estimates was beyond the scope of this project. To accomplish the goal of this project, a TSDD must meet the following minimum specifications:

• Accuracy of deflection measurements: The absolute values of deflection parameters are important. As such, the device should be able to make accurate measurements. Accuracy is defined as the closeness of measured deflections to their actual values.
  
  
• Precision (repeatability) of deflection measurements: The relative differences in deflection parameters between different points, and as such, the precision of the measurements is also important. Precision of measurements is defined as obtaining similar results on a given section with multiple surveys within a short period. The more precise the reported deflections are, the more certain the conclusions drawn with respect to the assessment of the uniformity of the pavement and the identification of problem spots and anomalous locations will be.
  
  
• Monitoring of applied load: The magnitude of the load applied by the tire to the pavement as well as the area of the instantaneous contact between the tire and pavement impact the measured deflection parameters. The magnitude of load may change due to a change in parameters such as the roughness of the road, and the contact area may change due to change in tire pressure (e.g., with temperature). Rada and Nazarian recommend that in the short-term, the tire pressure should be monitored during the operation, and in the long-term, appropriate instrumentation that monitors the magnitude of applied load should be added to the devices.⁽¹⁹⁾
  
  
• Operating speed: One of the benefits of TSDDs for network-level operations is that traffic control is not necessary. The device should preferably operate at the posted speed limit on a given road without sacrificing the quality of the data collected. The optimal speed of the operation as a function of the road characteristics should be evaluated.
  
  
• Distance between deflection measurements: One of the major advantages of TSDDs is the denseness of the data collected. A point-to-point density of the measurements of 1 ft (0.305 m) or less is desirable. The feasibility of achieving this density with acceptable quality should also be evaluated.
  
  
• Reporting of measured deflections: Independent of the denseness of the data collection, the use of statistical methods to report a representative deflection over a representative distance is desirable. Ideally, the reported representative deflections should also include the standard deviation for each representative distance in addition to the mean value. The representative distance should ideally be adjustable based on the length of the survey and variability in the representative deflections reported. The collected data should be segmented and compared with the structural and functional condition of the roads using modern techniques.
  
  
• Collection of additional information: Additional information about pavement structure from other rapid field testing data is desirable. For example, the moving device can be fitted with a GPR to estimate the pavement layer thicknesses and changes in the pavement structure along the survey. A high-speed video camera is also valuable for visual inspection of the condition of the pavement sections with anomalous deflection readings.

For practicality, a compromise among the precision, accuracy, loss of details with spatial averaging, and speed (and cost) of operation should be made. An optimized level of accuracy and precision should be achieved over a reasonable spatial averaging. The practical threshold values for the accuracy and spatial averaging should be defined based on the pavement structure, pavement responses of interest, and modes of failure associated with these responses.

Figure 5 is an idealized flowchart for accomplishment of the goal and objectives of this project as related to the determination of overall pavement structural condition indicators. Five major activities are identified in the flowchart. Activities 1–3 are desktop studies/analyses that set up the framework for the fieldwork to be done under activity 4. Activity 5 is another desktop study/analysis subtask where the results from the first four subtasks are integrated to recommend an appropriate network-level pavement structural condition assessment algorithm and procedure.

Figure 5. Flowchart. Idealized approach to successful accomplishment of project's objectives.

The five referenced activities also address the remaining project tasks. Activities 1–3 address the validation and evaluation of devices, activity 4 covers the field data collection, and activity 5 addresses the development of analysis methods and processes for incorporation of results into highway agencies' PMS applications. These five activities are detailed next, and changes to the work plan activities are highlighted in future chapters of the report, as applicable.

Activity 1: Establish Pavement Structural Condition Threshold Values

Since current TSDDs are perceived to be less applicable for rigid pavements, the project team decided, in consultation with FHWA, that the focus of the project would be on flexible pavements. As reflected in figure 5, this first activity consisted in addressing the following issues:

• Define typical categories of pavements.
  
  
• Define modes of failure.
  
  
• Establish ranges of pavement structures.
  
  
• Define structural-related responses.
  
  
• Select candidate deflection basin parameters.
  
  
• Define threshold value for each structural-related response.

Define Typical Categories of Pavements

The magnitude of surface deflection necessary to indicate the potential damage to a pavement structure is controlled by the type of the pavement (rigid versus semi-rigid versus flexible), the characteristics (thickness and stiffness) of the pavement layers above the foundation, the stiffness of the foundation layer, and the traffic volume. Four levels of traffic and representative ranges of pavement structural parameters were proposed (see table 5 and table 6, respectively). Through Monte Carlo simulation, the project team developed an extensive database of pavement sections.⁽²¹⁾ Supplementing this database was the pavement responses from Jacob Uzan Layered Elastic Analysis (JULEA).⁽²⁶⁾

1 ksi = 6.89 MPa

1 inch = 25.4 mm

Note: Subgrades are usually considered as infinite depth. As a result, no data for subgrade thickness are provided.

Define Modes of Failure

Given that the focus of this project was network analysis, the project team proposed to focus on the traditional modes of failure related to fatigue cracking of AC layer and subgrade rutting for flexible pavements.

Establish Ranges of Pavement Structures

The number of 18-kip (1,440-kN) ESALs to failure for each pavement section based on the two failure criteria was established using criteria similar to those recommended by the Asphalt Institute (AI). The AI equation to predict number of repetitions to fatigue cracking is shown in figure 6, while the AI equation for the number of axle loads to cause 0.5 inch (12.7 mm) of rutting is shown in figure 7.⁽²⁷⁾

Figure 6. Equation. AI fatigue prediction.

Where:

N_f = Number of repetitions to fatigue cracking.

V_b = Effective asphalt content in volume (percent).

V_a = Air voids (percent).

εt = Tensile strain at the critical location.

E = Stiffness of the material.

Figure 7. Equation. AI rutting prediction.

Where:

N_d = Number of axle loads to rut depth failure criteria (0.5 inch (12.7 mm)).

ε_c = Vertical compressive strain on top of the subgrade.

More mechanistic approaches such as those proposed in the American Association of State Highway and Transportation Officials (AASHTO) Mechanistic-Empirical Pavement Design Guide (MEPDG) or CalME, a software program developed by Caltrans/University of California Pavement Research Center using the mechanistic-empirical methodologies for analyzing and designing the performance of flexible pavements were used.^(28,29) Pavement structures that preliminarily yielded a design life of 10–40 years for each traffic category were delineated. The distribution of pavement structures along with the mean, median, and standard deviation were used to develop the representative ranges of pavement structures for each traffic category. These results are presented in section 8.4, Relationship Between Indices and Critical Response.

Define Structural-Related Responses

The primary structural responses considered were the tensile strains/stresses at the bottom of the AC and the compressive strains/stresses at the top of subgrade. In addition, the compressive strains/stresses in the middle of the AC and base/subbase were considered as surrogates for the rutting of the AC and base/subbase, respectively, when the rutting failure mode was considered.

Select Candidate Deflection Basin Parameters

Based on the FWD measurements, a number of deflection-basin related parameters that were perceived as strong predictors of the critical structural-related responses and structural conditions of the pavements were proposed. Horak and Emery provided an algorithm for using FWD-derived indices for airfield pavement evaluation, while Thyagarajan et.al. proposed a process for using indices for highway pavements.^(30,21) The FWD-measured deflections were simulated for the pavement sections developed under item 1 of this activity (i.e., define typical categories of pavements) and to estimate these and similar deflection basin parameters. Some of these parameters are presented in figure 8 through figure 18.

Figure 8. Equation. Definition of radius of curvature R1.

Where:

r = Distance from the load.

D₀ = Deflection under the load.

D_r = Deflection at a distance r from load.

Figure 9. Equation. Definition of radius of curvature R2.

Figure 10. Equation. Definition of deflection basin area (A).

Where:

D_i = Deflection at i inches away from load.

Figure 11. Equation. Definition of shape factor F1.

Figure 12. Equation. Definition of shape factor F2

Figure 13. Equation. Definition of SCI.

Figure 14. Equation. Definition of Base Curvature Index (BCI).

Figure 15. Equation. Definition of Base Damage Index (BDI).

Figure 16. Equation. Definition of slope of deflection (SD).

Figure 17. Equation. Definition of area under pavement profile (AUPP).

Figure 18. Equation. Definition of tangent slope (TS).

Where:

dD = Difference in deflection.

dr = Difference in distance.

Define Threshold Value for Each Structural-Related Response

Ideally, TSDDs are able to provide adequate data accurately and precisely enough so that the performance history of a pavement section can be estimated with reasonable accuracy before any functional or structural distresses are evident. In this ideal process, the critical strains (e.g., tensile strain at the bottom of the AC) are small enough, and the model for estimating performance of the pavement with time is accurate enough so that highway agencies can conduct what-if analyses (considering life-cycle cost analysis) to make more informed decisions about the best use of their M&R budgets. The project team strived to reach that level of capability. The following items were addressed to evaluate whether the suggested process could be implemented in the near future:

• Do TSDDs provide enough information to reliably estimate critical strains and stresses?
  
  
• Do TSDDs provide accurate enough and precise enough information?
  
  
• Do highway agencies have enough information about the pavement structure to translate the parameters measured with the TSDDs to critical strains/stresses for network- level studies?
  
  
• Are there accurate and calibrated performance models available to predict the performance of pavements?

The subsequent activities, especially activity 4 (field evaluation of devices in figure 5), were designed to answer the first three questions comprehensively. It was thought that the fourth question could be addressed through close collaboration between FHWA and project teams.

An alternative way of establishing the thresholds considered was to conduct simple structural analyses to estimate the remaining life from the time of testing. The relevant structural responses for the representative pavement structures in each traffic category corresponding to 2, 5, and 10 years^[1] of remaining life were preliminary considered as the thresholds for deteriorated, marginal, and well-performing pavements, respectively. In other words, if the critical strains/stresses exceeded the 2-year remaining life thresholds, the pavement was considered a candidate for reconstruction, and if the critical strains/stresses were less than the 10-year thresholds, the pavement was considered in good condition. Pavements with remaining lives of 2–5 years were considered candidates for major rehabilitation, and pavements with remaining lives between 5 and 10 years were considered candidates for maintenance or light rehabilitation.

These concepts were revisited after field data with TSDDs became available to develop the best strategies for implementing TSDDs (addressed as part of activity 5).

Activity 2: Relate Structural-Related Responses to Deflection Parameters Measured with TSDDs

The analyses proposed under activity 2 were based on the traditional static layered elastic algorithms. The TSDDs used proprietary hardware and software to estimate dynamic surface deflection parameters imparted by the moving tire loads. Since the trucks carrying the devices often traveled at traffic speeds, the resulting dynamic surface responses were affected by inertia and damping of the layered pavement system. The evaluation of the capabilities of these devices was therefore undertaken by a computational model that is capable of modeling moving loads traveling on a layered medium. The five issues addressed under this activity are as follows:

• Using a numerical model to predict structural-related responses and TSDD-measured parameters.
  
  
• Understanding TSDD estimated deflection parameters.
  
  
• Evaluating feasibility of estimating candidate deflection basin parameters from TSDD-measured parameters.
  
  
• Establishing sensitivity of structural-related responses to each TSDD-measured parameter and index.
  
  
• Selecting most sensitive TSDD indices for further considerations.

Using a Numerical Model to Predict Structural-Related Responses and TSDD-Measured Parameters

The computer software 3D-Move is ideally suited to evaluate and compare pavement responses measured with TSDDs. 3D-Move estimates the dynamic pavement responses at any point within the pavement structure using a continuum-based finite-layer approach. The 3D-Move model can account for important pavement response factors such as the moving traffic-induced complex three-dimensional contact stress distributions (normal and shear) of any shape, vehicle speed, and viscoelastic material characterization for the pavement layers. (See references 2 and 31–33.) The pavement surface deflection is affected by many factors that include pavement layer characteristics (thickness and stiffness properties), vehicle speed, and damping. Damping in particular plays a major role in the form of time lag between the loaded tire and the deflection response. The 3D-Move model uses viscoelastic formulation and complex frequency domain analyses, such that damping can be specified as either a single value or as a function of frequency. For each of frequency considered in the fast Fourier transform (FFT), the corresponding damping (frequency dependent, if required) was selected and used to obtain the imaginary part of the modulus.

Since rate-dependent (viscoelastic) material properties can be accommodated by 3D-Move, it was considered an ideal tool to study pavement response as a function of vehicle speed through the direct use of the frequency sweep test data (dynamic modulus and damping) of AC mixture.

Several field validations (e.g., Penn State University test track, MnROAD, and University of Nevada-Reno (UNR) Off-Road Vehicle study) that compared a variety of independently measured pavement responses (stresses, strains, and displacements) with those computed by 3D-Move have been reported in the literature. (See references 34, 35, 31, and 33.) Hajj et al. reported that the responses from 3D-Move were within 6 percent of those estimated by ViscoRoute developed by Chabot et al. for thin and thick pavements.^(36,37) Those studies demonstrated the applicability and versatility of the 3D-Move approach. Further validation of 3D-Move to strengthen the validity of its application in relating device measurements to structural responses was carried out as part of the field evaluation of the devices as discussed under activity 4.

Understanding TSDD Estimated Deflection Parameters

Each TSDD has its own way of calculating deflection parameters. The TSD measures the surface vertical velocity at as many as nine points^[2] (in the front from the mid-point between the dual tires) within the deflection bowl using a Doppler laser technology. The vertical velocity measurements were divided by the vehicle speed to arrive at the slopes of the deflected shape at the measuring locations. These slopes were then fitted to a deflection bowl to estimate the surface deflection at the mid-point between the tires (D₀) and other locations.

The RWD used six spot lasers mounted on a horizontal aluminum beam to measure the deflected pavement surface (longitudinally along the midpoint between the dual tires). Two sensors (sensor D located 7.25 inches (184 mm) behind the axle and sensor F located 7.75 inches (197 mm) in front of the axle in figure 19) were within the deflection bowl, while the other four sensors represent locations within the undeflected pavement surface. The A, B, C, and E sensor readings were used to obtain the load-induced surface deflection at the location of sensors D and F. The following questions were addressed as part of these simulations:

• Are the algorithms used to estimate surface deflection parameters appropriate? Routinely used critical surface deflection parameters by the TSD include SCI₃₀₀ (D₀ – D₃₀₀), SCI₂₀₀ (D₀ – D₂₀₀), and D₀. Conversely, the deflection at sensor D (see figure 19) is used by the RWD. A subset of pavement structure database from each traffic category was used as input to 3D-Move to fully explore the strength of the algorithms used to estimate the deflection parameters. The subset was selected based on the results of the linear elastic (LE) analyses performed under activity 1. For example, the appropriateness of the location of sensor D of the RWD needed to be evaluated, as it was not readily apparent whether the maximum deflection occurred 7.25 inches (184.15 mm) behind the axle for all pavement structures and vehicle speed. It was anticipated that the damping characteristics along with stiffness properties (soft versus stiff) governed where the maximum pavement deflection occurred.
  
  
• What are the ideal locations for sensors? The sensor locations on the TSD are changeable, while they are fixed on the RWD. Accordingly, it was worthwhile to explore whether other sensor locations could better predict pavement layer conditions. An experiment plan as shown in table 7 was adopted to explore the optimal selection of sensor locations. It was also considered possible that the optimal sensor locations may change as the speed of the test vehicle changes and thickness of the AC layer; this matter is addressed next.
  
  
• How does the speed of test vehicle impact the measured deflection parameters? Field studies revealed that vehicle speed played a significant role in pavement response. Though the devices under consideration were designed to operate at or close to posted traffic speeds, the vehicles may operate at lower speeds for a variety of reasons. The role of vehicle speed on the measured pavement deflection is important, especially if a comparison is to be made between the data from the same TSDD traveling at different speeds. The speed of the vehicle may also impact the optimal location of the sensors.
  
  
• Table 8 provides the experiment plan that was adopted with 3D-Move to investigate the role of vehicle speed. If the speed of the vehicle impacted the optimal location of the sensors, a strategy to recommend the best compromise on the location of the sensors could be developed.
  
  
• How does the seasonal change in material properties of pavement layers (e.g., temperature for AC and moisture for base and subgrade) impact the response? It was considered important to quantify the variability in the surface deflection (and hence candidate indices or parameters relating to pavement structural condition) caused as a direct consequence of the change in material properties brought on by the seasonal changes. This variation can be used to compare the TSDD measurements taken on a given pavement section at different seasons of the year. Table 9 provides the experiment plan adopted with 3D-Move to investigate the role of seasonal changes. Since the seasonal changes in material properties are not explicitly considered in 3D-Move (i.e., are simulated by increasing or decreasing the properties of different layers), the results from the previous two questions were used to address this question. It was also decided that if a gap in the trends was observed, additional 3D-Move cases would be executed.

1 inch = 25.4 mm

Figure 19. Illustration. RWD sensor locations.

1 inch = 25.4 mm

Evaluating Feasibility of Estimating Candidate Deflection-Basin Parameters from TSDD-Measured Parameters:

For practical implementation, it was considered highly desirable to simulate the TSDD-measured parameters and to estimate the candidate deflection-basin parameters established under activity 1 for estimating the pavement structural conditions.

Establishing Sensitivity of Structural-Related Responses to Each TSDD-Measured Parameter and Index

The pavement structure database or its subset was used as input to 3D-Move to estimate the relevant deflection parameters and corresponding indices selected in activity 2 at three vehicle speeds: slow (20 mi/h (32.2 km/h)), intermediate (40 mi/h (64.4 km/h)), and fast (60 mi/h (96.6 km/h)). The two databases were merged for further statistical analyses to address the following:

• How well the surface deflection from static analyses under activity 1 correlated to the surface deflection parameters simulated with 3D-Move for each TSDD.
  
  
• How well the critical strains/stresses from static analyses under activity 1 correlated to the maximum critical strains/stresses simulated with 3D-Move.
  
  
• How well the critical strains/stresses established earlier under activity 2 correlated to the surface deflection parameters simulated with 3D-Move for each TSDD.
  
  
• How well the critical strains/stresses established earlier under item 3 (i.e., evaluating feasibility of estimating candidate deflection basin parameters from TSDD-measured parameters) of activity 2 correlated to candidate indices from activity 2.
  
  
• Which candidate indices from item 3 (i.e., evaluating feasibility of estimating candidate deflection basin parameters from TSDD-measured parameters) under activity 2 were more sensitive to the surface deflection parameters simulated with 3D-Move and the critical strains/stresses established for each TSDD.

Selecting Most Sensitive TSDD Indices for Further Considerations

Through correlation and sensitivity analyses explained in the previous item, the most representative indices associated with different modes of failure were identified for further field validation.

Activity 3: Establish Ideal Measurement Characteristics for TSDDs

The outcomes of activities 1 and 2 were used to establish the ideal measurement characteristics. The five issues addressed under this activity were as follows:

1. Establish the minimum and maximum likely values of deflection parameters anticipated from each TSDD: The database developed under activity 2 contained the minimum and maximum likely values of deflection parameters for each pavement category selected under activity 1. The distribution of the surface deflection parameter for each TSDD was therefore established. Based on a reasonable confidence level (80–90 percent), the minimum and maximum likely deflection parameters were also established.
   
   
2. Compare the minimum and maximum deflection parameters with each TSDD sensor specifications: Such statistical comparisons were carried out to establish the likelihood of the success of each device in providing useful data for each pavement category selected in activity 1.
   
   
3. Establish the minimum and maximum likely values of critical strains/stresses anticipated from each TSDD: The database developed under activity 2 contained the minimum and maximum likely values of critical strains and stresses for each pavement category selected under activity 1. The distributions of the critical strains/stresses for each TSDD were established. Based on a reasonable confidence level (80–90 percent), the minimum and maximum likely critical strains/stresses were also established.
   
   
4. Establish the desired precision of the TSDD raw measurements: Using the correlation analysis from activity 2 and the distributions from items 1 and 3 under this activity, the sensitivity of the critical strains/stresses to TSDD deflection measurements were established. The sensitivities were then mapped to the desired precision of the measurements for each pavement category.
   
   
5. Establish the desired accuracy of the TSDD parameters: Using the correlation analysis from activity 2 and the distributions from items 1 and 3 under this activity and the thresholds for weak and strong pavements, the confidence levels for delineating the weak and strong pavements for each pavement category were established. Based on these confidence intervals, the desired accuracy of the devices was also established. The established preliminary target precisions and accuracies were compared with those obtained from the devices during field study as discussed later under activity 5.

Activity 4: Field Evaluation of Devices

The deflection measurement that defines the minimum requirements for the capable devices include the accuracy of measurements, precision of measurements, and a number of other items that are categorized under operational limitations of devices. To summarize this information, the parameters that support or validate the interpretation of the data collected with a device are not straightforward. This is because the response parameters cannot be measured directly; the raw data collected with a TSDD have to be combined with the pavement structure and pavement conditions through either empirical, analytical, or numerical algorithms to estimate the critical strains/stresses within or at the interfaces of pavement layers. Accordingly, a rigorous field study was required to evaluate, validate, and improve the numerical results and suggested criteria for estimating the structural conditions of the pavements from the parameters measured with the TSDDs. The bulk of the field study activities were carried out at the Minnesota Department of Transportation (MnDOT) MnROAD facility located north of Minneapolis, MN; the basis for selecting this facility for use in the study is provided in section 4.2, Field Trial Locations. The specific issues addressed as part of field study activity include the following:

• Accuracy of deflection measurements.
  
  
• Precision of deflection measurements.
  
  
• Operational limitations of devices.
  
  
• Validation of 3D-Move using MnROAD data.

Accuracy of Deflection Measurements

The accuracy of the measurements was determined by comparing the reported values from each device with the same values from an external sensor. The main focus of the determination of the accuracy was the deflection parameters measured by the devices. The RWD estimated the surface deflection using a spot laser, and the TSD measured the surface velocity using Doppler laser technology. Table 10 summarizes of the objectives and other factors related to this experiment. A comparison of the deflection parameters measured at the densest interval with each device with the deflection parameters measured with embedded sensors at the surface of the pavement was proposed.

Based on the status report of the MnROAD sensors that measure displacement parameters (i.e., embedded linear variable differential transformers (LVDTs) and accelerometers), they did not function reliably. As such, the project team decided to retrofit pavement test sections with appropriate surface sensors for this activity. Three alternative sensors (LVDTs, geophones, and accelerometers) were considered as possible candidates for embedded sensors, as outlined in  table 11. It was decided to primarily use geophones for accuracy purposes since they are the least expensive, can be easily ruggedized in a steel casing, and have one-to-one correspondence to the deflection parameters measured by the TSD. In addition, one accelerometer was used at each accuracy test section to verify the responses of the retrofitted geophones.

A secondary parameter related to the accuracy of the measured deflection parameters is the instantaneous applied load to the pavement. The impact of pavement roughness on the instantaneous load applied to the pavement is documented in the literature.^(38,39) The state of the practice in the analyses of the TSDD deflection data is usually based on the assumption that the instantaneous load is equal to the static load, but in this project, the impact of the variation in the instantaneous load on the performance of the devices was studied. The TSD was equipped with instrumentation to measure the instantaneous applied loads concurrent with the deflection measurements. Such measurement is currently lacking from the RWD. Instrumenting the test sections for directly measuring the instantaneous load seemed impractical. However, since the TSD was equipped to estimate the load applied to the pavement, an evaluation of the potential benefits of normalizing measured TSD deflection parameters with loads reported by the device was carried out.

Table 12 summarizes the objectives and other factors related to this experiment.

Precision of Deflection Measurements

The precision of the measurements is estimated by comparing the reported values from replicate measurements in a short period of time. Table 13 summarizes the objectives and other factors related to this experiment.

Operational Limitations of Devices

Although the major technical issues can be conceptually addressed with the established levels of accuracy and precision, many other practical parameters can also impact how well the condition of the pavement can be assessed and were therefore considered in this project. These practical parameters can be optimized to ensure that the maximum information can be robustly extracted from the devices. Marginal additional data collection was needed for this purpose; the data collected under the previous two items of this activity (i.e., accuracy of deflection measurements and precision of deflection measurements) were processed differently to address most of these issues. Those issues were as follows:

• Desirable reporting interval: Academically speaking, it would be desirable to collect and store data as densely as possible. In practical terms, however, the data storage requirements for such a dense dataset are prohibitive. Since the goal of this project was network-level analyses, it was suitable to statistically process the data to reasonable reporting intervals. Typically, arithmetic averaging of the data over a certain distance (e.g., 0.1 mi (0.161 km)) is carried out for this purpose. While 0.1 mi (0.161 km) is typical for the network level and is the state of the practice with most other network-level PMS data elements, there are also cases of localized weak sections within the 0.1 mi (0.161 km) that should be identified. This optimum interval is controlled by the size of the network to be tested, the uncertainty of the device measurements (including the precision and accuracy), and a number of other institutional requirements and practices of local highway agencies.⁽¹³⁾ The goals of this effort were to improve the method of statistical processing of the data and to optimize the reporting scheme so that with minimal volume of data the significant areas with structural defect could be identified.
  
  Some of the simple options that were considered include reporting the coefficient of variation (COV) in addition to the average value per 0.1 mi (0.161 km) or the moving average with a base length of 0.1 mi (161 m). Some more advanced improvements considered include supplanting the simple averaging with the more modern geospatial methods or adjusting the reporting interval dynamically based on the COV of a section relative to a threshold COV (set based on the precision of the device). The results of the study were validated on long pavement sections by comparing the lengths of the abnormally high deflection zones from the device and the corresponding lengths of pavements with high FWD deflections or extensive distress. This is addressed in chapter 10.
  
  
• Desirable operating speed: To expedite the collection of data safely, operating TSDDs at the posted highway speed was preferred. The quality of the data collected at a given speed may also be impacted by the pavement structure, roughness, texture, and geometrical design. It may be possible to relate the optimal operating speed of the devices to these parameters. To that end, statistical analyses were conducted to relate the quality of the deflection parameters collected at the three speeds under items 1 and 2 (i.e., accuracy of deflection measurements and precision of deflection measurements) of this activity to pavement structure. Roughness and geometry of the sections selected were only considered in the context of explaining anomalies. The results of these analyses were used to make some preliminary recommendations on the operating speeds. Where possible and applicable, the data were also correlated to texture to explain the anomalies in the measured data.
  
  
• Temperature adjustment: The variation in the deflection parameters with temperature, especially for pavement structures with reasonably thick AC layers, is well documented in the literature. Even though several procedures for adjusting the FWD deflections with temperature are proposed (e.g., ASTM D7228, "Standard Test Method for Prediction of Asphalt-Bound Pavement Layer Temperatures"), their applicability to the highway speed devices is not known.⁽⁴⁰⁾ Accordingly, a preliminary study was made to see whether the potential differences in deflections measured at the two temperatures for flexible pavements tested under item 1 of this activity could be explained by the changes in temperature and whether the existing FWD-based relationships could be used with the TSDD data. Supplementary data collected at other temperatures during the day were also considered to study the impact of temperature more rigorously. Moreover, the 3D-Move program has been previously used to model temperature variations by treating the AC layer as comprised of many individual sublayers and assigning each sublayer an individual set of frequency-dependent modulus, E*. E* can vary as a function of the pavement temperature. Accordingly, a master curve approach was ideally suited to get these temperature-dependent E* versus frequency curves for the AC sublayers.
  
  
• Compatibility of devices: Even though the RWD and TSD collect different deflection parameters, ideally they should yield similar structural conditions for the pavement network. To assess whether the devices were interchangeable, the interpreted pavement condition data (not raw measured deflection parameters) was compared. The assessment consisted of relating the TSDDs' structural condition rating from all sections tested to one another to quantify the linearity and bias. Student t-tests and F-tests were used to determine whether the data from the two devices were from the same population.

Validation of 3D-Move Using MnROAD Data

Existing pavement response measurements on flexible pavements at MnROAD facility include strain responses in longitudinal (vehicle direction) and transverse directions and vertical pressure histories in base and subgrade layers (see table 10). MnROAD contains more than 90 reliably operating longitudinal and transverse dynamic strain gauges in flexible pavement cells and more than 40 dynamic pressure gauges in the foundation layers. These measurements were directly compared with those computed by 3D-Move. Since the project focus was on AC layer condition, attention was given mainly to AC strain measurements. The 3D-Move modeling requires pavement layer configurations and properties and traffic loading. An existing MnROAD database of material properties, which include FWD data and also viscoelastic characterization (dynamic modulus and damping) of AC properties (e.g., master curve of frequency sweep data), were used. The tire load measurements provided the information on the tire-pavement interaction load.

In addition, data collected from the supplementary embedded surface geophones were also used in the validation. The durations of the time histories were variable (based on the vehicle speed) to cover a distance of ±15 ft (4.58 m) from the embedded sensor. Based on preliminary analysis using 3D-Move, the spacing between geophones was determined to ideally be about 5–6 inches (127–152 mm) at a number of locations. The measured velocity time histories from the geophones along with the estimated displacement time histories were relevant since they were the basic measurements that were used by the TSDD under consideration.

Activity 5: Best Strategies for Implementing TSDDs in Network-level Evaluation

The main goal of this activity was to integrate the outcomes from the previous four activities into a coherent set of practical guidelines and protocols for the successful implementation of the TSDDs in network-level structural condition assessments for use in State transportation department PMS applications. Additional data analyses and alternative data interpretation algorithms were considered. Based on the outcomes of the 3D-Move validations, additional simulations were also carried out. The four issues addressed under this activity include the following:

1. Appropriate TSDD indices: The main effort associated with this issue was to subject the most promising structural indices selected under activity 2 to the data obtained from relevant MnROAD sections with TSDDs and to evaluate what indices explain the structural conditions of each section as judged by the severity of the distresses, IRI, and FWD deflections. To that end, applicable test sections were subdivided into weak, marginal, and strong (when appropriate) to compare with the predicted structural conditions based on the TSDD measurements. The outcomes were then applied to the longer pavement test sections as discussed in the next chapter for validation purposes.
   
   
2. Optimal operational parameters for the TSDDs: The main activity associated with this issue was to evaluate and supplement the numerical results and experimental analyses to recommend the most practical means of utilizing the TSDDs. As part of this item, the following information was provided for each TSDD:
   
   
   • Ideal pavement types and those that are not likely to lend themselves to conclusive structural conditions.
     
     
   • Speed of operation range for valid measurements as a function of pavement type.
     
     
   • Possible limitations due to the road geometry and functional conditions of the pavements.
     
     
   • Possible recommendations about the sensor locations.
     
     
   • Recommended additional features in the potential TSDDs for better interpretation of structural conditions (like GPR, temperature sensor, etc.).
     
     
3. Most appropriate algorithm for structural condition assessment: The goal of this effort was to integrate the best means of systematically processing the raw deflection parameters from the TSDDs to obtain the appropriate indices and the preferred statistical or geospatial methods to summarize the processed results into representative structural condition categories for network-level pavement management.
   
   
4. Recommended protocols: The goal of this effort was to integrate the outcomes from each step into a straightforward generic protocol so that different highway agencies can utilize them for their evaluation and modification based on their specific needs.

4.2 Field Trials Location

To support the analysis methodologies detailed in the previous section of the report, the field trial location(s) should provide the following pavement factorial parameters:

• Pavement type (flexible, rigid, or composite).
  
  
• Pavement surface texture (smooth or rough).
  
  
• Pavement thickness (thin to thick).
  
  
• Pavement condition (excellent to poor in terms of ride quality and/or distress).
  
  
• Horizontal curves.
  
  
• Vertical gradients.
  
  
• Different unbound layer strength.

Other testing considerations required by the analysis methodologies include varying temperatures and device speeds. These considerations could be taken into account in determining the field location(s) but could also be controlled once the field locations were selected by varying the time of day the testing occurred to control temperature or the speed of the device, provided varying the speed of the device did not pose a safety concern.

Several potential field trial locations were considered to fulfill the requirements mentioned, including the MnROAD facility; instrumented test sections in Kansas, Ohio, and New York; and test sections previously used in the evaluation of the RWD in Louisiana.

4.3 Summary

This chapter presented a detailed work plan developed for the remainder of the project, including analysis methodologies and field trial locations. The details provided in this chapter were developed to accomplish the following two objectives:

• Confirm that the TSDDs met a minimum set of specifications related to the structural evaluation of pavements at the network level.
  
  
• Propose processes to incorporate pavement structural information from the successful TSDDs into network-level PMS applications.

The decision was made to hold the field trials at the MnROAD facility because it provides a multitude of test sections in one location and contains readily available information, including environmental and dynamic load response data. In addition to the MnROAD test sections, additional field trial testing was planned on an 18-mi (29-km) loop located in Wright County, MN, near the MnROAD facility.

¹ These values were considered preliminary, for the purposes of the work plan, and are subject to change during the analyses.

² The TSD that was used in the field trials discussed in chapter 5 only had six points.