Enhanced Analysis of Falling Weight Deflectometer Data for Use With Mechanistic-Empirical Flexible Pavement Design and Analysis and Recommendations for Improvements to Falling Weight Deflectometers

CHAPTER 3. LTPP DATA ANALYSIS

This chapter describes the results of the statistical analysis performed on a relatively large sample of FWD data from the LTPP database to assess the following: (1) prevalence of dynamic effects, (2) prevalence of nonlinear behavior; and (3) measurement issues based on evidently erroneous deflection sensor time histories. The data cover all climatic zones, seasons, and temperature ranges.

PRELIMINARY STATISTICAL ANALYSIS OF LTPP FWD LOADING HISTORIES

For this analysis, the research team randomly selected 1,224 tests (17 States, 6 sections per State, 3 stations per section, and 4 load levels). Table 4 summarizes the data extracted from the LTPP database. The time-history plots for each test were visually reviewed, and then the sections were classified using the following criteria:

• Dynamic behavior: Vibrations in the deflection histories.
• Nonlinear behavior: The load-to-deflection ratio shows a consistent trend with load level.

Table 4. LTPP sections used in the statistical analysis.

Figure 13 and figure 14 show examples of FWD time histories exhibiting free vibrations (dynamic behavior) and no dynamic behavior, respectively. Figure 15 and figure 16 show examples of nonlinear behavior with stress stiffening and softening, respectively.

The research team observed that dynamics were present in about 65 percent of the cases while nonlinearity could be prevalent in a range of as low as 24 percent of the cases to as high as 65 percent of the cases, depending on severity level and sensor location. Nonlinearity was more prevalent for the sensors that were far from the center of the load. A more detailed analysis was performed to identify when dynamics and nonlinearity were prevalent.

Figure 13. Graph. Example of time histories showing dynamic behavior for LTPP section 161020, station 1.

Figure 14. Graph. Example of time histories showing no dynamic behavior for LTPP section169034, station 3.

Figure 15. Graph. Example of stiffening behavior for LTPP section 81053, station 3.

Figure 16. Graph. Example of softening behavior for LTPP section 87781 station 3.

DETAILED STATISTICAL ANALYSIS

Dynamic Behavior

As explained above, the time-histories plots for each test were visually reviewed, and then the sections were classified as follows: (1) season, (2) air temperature, (3) wet/dry, or (4) freeze/no freeze. Figure 17 shows the percentage of sections per category where dynamics were prevalent.

The results showed no particular trend with season and temperature, and the sections in the dry and freeze climate zones appeared to be more prone to dynamic behavior. t-tests were performed to assess whether the means of the two groups (wet/dry and freeze/no freeze) were statistically different from each other. Table 5 shows the results of the t-tests. The analysis showed that the difference between dry and wet as well as freeze and no-freeze was statistically significant, with dynamics statistically more prevalent in dry freeze sections (figure 18).

Figure 17. Graphs. Preliminary results—evidence of dynamic behavior by climatic
information: classification by season (top), temperature (middle), and climate zone
(bottom).

Table 5. LTPP sections used in the statistical analysis.

Figure 18. Graph. Mean and standard deviation of percent of sections with dynamics for wet/dry and freeze/no freeze.

Nonlinear Behavior

The following two separate analyses were considered to determine when nonlinearity was prevalent: (1) using all the sections that showed a nonlinear behavior and (2) using only the sites that exhibited no dynamics. The purpose of the second analysis was to investigate the interaction between dynamic and nonlinear effects because both affect the farther sensors. However, the number of sections for the second analysis was small (91 sections) and did not cover all climatic zones, seasons, and temperature ranges, which made the results from the second analysis not reliable. Therefore, only the results from the analysis using all the data are reported. In the analysis, the sections were classified as follows according to the slope of the load-to-deflection ratio trend:

• Softening: The slope is negative.
• Stiffening: The slope is positive.

Table 6 and figure 19 show the distributions of load-to-deflection slope for all sensors. It can be seen that the slope was mainly within ±20-percent range and that it shifted from more positive (stiffening) to more negative (softening) with increasing sensor number, i.e., increasing sensor distance from the load. This means that, as expected, stress softening was more prevalent in the lower pavement layers.

Table 6. Distribution of load-to-deflection slope by sensor.

Figure 19. Graphs. Distribution of load-to-deflection slope by sensor.

To further examine the degree of nonlinearity observed, the research team filtered the data by varying the minimum threshold on the slope for defining when nonlinear behavior was observed. For example, if the threshold for nonlinear behavior was ±5 percent, then any section that exhibited a load-to-deflection slope of more than ±5 percent would be considered as exhibiting nonlinear behavior (stiffening if the slope was positive and softening if the slope was negative). The thresholds were set at 5, 6, 7, 8, 9, and 10 percent. Table 7, figure 20, and figure 21 show the percent of sections showing linear versus nonlinear behavior for the various thresholds and for each sensor. The table and figures also show the split between stiffening versus softening behavior (within those exhibiting nonlinear behavior).

Table 7. Distribution of linear versus nonlinear behavior by sensor and percent slope.

Figure 20. Graphs. Distribution of linear versus nonlinear behavior for 5- to 7-percent threshold load-to-deflection slope.

Figure 21. Graphs. Distribution of linear versus nonlinear behavior for 8- to 10-percent threshold load-to-deflection slope.

The results showed a trend of increasing nonlinearity and more softening as sensor number increased and as the sensor distance from the load increased. This means that, as expected, nonlinearity was more prevalent in the farther sensors, which reflected the behavior of the underlying deeper layers, and that these materials showed much more stress softening than stress stiffening behavior.

Assuming a minimum slope of ±5 percent as a threshold for defining nonlinear behavior, the lowest percent of sections with nonlinear behavior was 52 percent (sensor 1) while the highest was 65 percent (sensor 6). Increasing the threshold to ±10 percent reduced these percentages by about half; the lowest percent of sections with nonlinear behavior became 24 percent, and the highest became 33 percent.

Within the sections exhibiting nonlinear behavior, the percent of sections showing stress softening versus stress stiffening was fairly similar irrespective of the threshold on the slope. The split was about 50-percent softening and 50-percent stiffening for the close sensors and gradually shifted to about 85- to 90-percent softening and 10- to 15-percent stiffening for the intermediate and most distant sensors.

The sections were also classified by: (1) season, (2) temperature, (3) wet/dry, or (4) freeze/no freeze for each sensor, with the nonlinearity threshold slope set at ±5 percent. Figure 22 through figure 24 show the percentage of sections by season, temperature, and climatic zone, respectively, for each sensor where nonlinearity was prevalent.

Figure 22 shows that the percent of sections that exhibited nonlinearity for sensors 1 through 5 was generally highest for summer, followed by fall and spring, and lowest in winter. Sensors 6 through 8 exhibited more nonlinear behavior in the fall. The trend with seasons suggests that nonlinearity is more prevalent when the pavement system is less stiff, as expected. This trend was also generally true with temperature, as shown in figure 23, although there was more variability in the data. Figure 24 shows that there was no particular trend with climatic zone, however. Nonlinear and linear behavior seems to have been evenly split between wet and dry climates and freeze and no-freeze climates. t-tests were performed to assess whether the means of the two groups (wet/dry and freeze/no freeze) were statistically different from each other for all the sensors. Table 8 shows the results of the t-tests. It appears that climate had no effect on the number of sections that exhibited nonlinearity. Table 9 shows the distribution of load-deflection slope by sensors in sections that exhibited nonlinearity.

Figure 22. Graphs. Percent of sections by season where nonlinear behavior was prevalent.

Figure 23. Graphs. Percent of sections by temperature where nonlinear behavior was prevalent.

Figure 24. Graphs. Percent of sections by climate zone where nonlinear behavior was prevalent.

Table 8. Results of the t-tests on nonlinearity for wet/dry and freeze/no freeze conditions for all sensors.

Table 9. Distribution of load-deflection slope by sensor.

MEASUREMENT ISSUES

During the analysis of the LTPP FWD data, signs of measurement issues were also encountered in some tests. Figure 25 presents some samples of these errors. These issues included erroneous deflection sensors (middle row: section 130508 station 1 and section 130566 station 1) or drift (bottom row, left: section 260116 station 1), or data truncation (top row: section 220125 station 3 and section 220125 station 6; and bottom row, left: section 482108 station 1).

Figure 25. Graphs. Examples of measurement issues.

CONCLUSION

In this chapter, detailed exploratory analyses on a relatively large sample of FWD test results from the LTPP database were conducted to assess (1) prevalence of dynamics, (2) prevalence of nonlinear behavior, and (3) measurement issues based on apparently erroneous deflection sensor time histories. The data covered all climatic zones, seasons, and temperature ranges. It was observed that dynamics were present in about 65 percent of the cases, while nonlinearity could be prevalent in a range as low as 24 percent of the cases to as high as 65 percent of the cases, depending on severity level and sensor location. Nonlinearity was more prevalent for the sensors that were far from the center of the load. Because of the prevalence of dynamic behavior (in the form of free vibrations of deflection sensor time histories) observed in the large sample of LTPP FWD test data, it was hypothesized that in the great majority of the cases, the stiff layer condition might not correspond to the presence of shallow bedrock. Such bedrock would be highly unlikely given that it typically lies at much greater depths. Instead, the stiff layer condition could be manifested anytime the soils below the subgrade layer are stiffer than the subgrade layer itself. This could be caused by increased confinement with depth, overconsolidation, or existence of a shallow groundwater table, for example; these situations are very common in any soil profile. This would explain the high percentage of sections from the LTPP database that showed dynamic behavior.