Chapter 4. Results from Laboratory and Field Evaluations

Introduction

As described earlier, laboratory and back calculation tests were performed on subgrade soils gathered from primary roads in the State of Wyoming. Laboratory testing consisted of determining the soil's resilient modulus, R-value, water content, optimum water content, plasticity index, soil classification, and group index. Back calculation tests consisted of substituting values obtained from the field deflection tests into three computer programs to obtain a second set of resilient modulus values. This chapter presents the results from the above tests and provides a comprehensive discussion and analysis of the data gathered.

Site Characteristics

The pavement structure for each of the nine sites included in this research consisted of an asphalt concrete (AC) layer, a granular or treated base, and the underlying subgrade soil. Four of the nine sites had a treated base, three with an asphalt treated base (ATB) and one with a cement treated base (CTB). Table 4.1 summarizes the thicknesses of the AC and base layers for all test sections. All of these thicknesses were determined from the pavement cores at each test section. In addition, approximately 610 to 762-mm (24 to 30-in.) of the underlying subgrade soil was removed in each Shelby tube for laboratory testing.

Results from Soil Property Tests

Several fundamental soil property tests were conducted after performing the resilient modulus test on the soil samples. These tests included: sieve analysis, liquid and plastic limits (LL and PL, respectively), and water content determination. Tables 4.2 and 4.3 summarize the values for these soil property tests from the summer of 1992 and spring of 1993, respectively. A high percentage of the sites had water contents below the optimum water content. All of the soils were classified as A-4, A-6, or A-7-6 based on the AASHTO soil classification system. However, one sample from the summer of 1992 and two samples from the spring of 1993 had different soil classifications, A-1-B, A-2-4, and A-2-6, respectively. In addition, several of the sites had large group index (GI) values, indicating poor quality soils. Finally, the plasticity index (PI) values were moderate to high as shown in the tables.

Laboratory Resilient Modulus Values Based on 41.4-kPa (6-Psi) Deviator Stress

Samples for laboratory M_R testing were primarily selected from the middle portion of each Shelby tube because of the visible disturbance on both ends. This selection process provided samples that were rigid and "undisturbed." After extracting each sample from the tube, diameter and height measurements were taken and recorded on laboratory data sheets. Next, a rubber membrane was placed over the soil sample and two porous stones which were on both ends of the sample. After aligning the sample in the testing device, the resilient modulus tests were conducted by following the AASHTO specifications. During each of the fifteen loading conditions, the last five cycles were saved on disk for future retrieval to determine the resilient modulus value. After completing the first set of tests on undisturbed samples, another set of tests were conducted on disturbed samples using a zero confining pressure and deviator stresses ranging from 13.8 to 69.0-kPa (2 to 10-psi) in 13.8-kPa (2-psi) increments. The disturbed samples were prepared by destroying the original undisturbed samples and re-compacting them in five, equal lifts under a static load.

For each of the testing conditions described earlier, the peak deformation and applied load readings were retrieved from the files created during testing. By entering these values into a special spreadsheet, resilient modulus values were calculated from each of the different confining and deviator stress conditions. Figure 4.1 shows an example of the M_R summary spreadsheet used in this research (Note: values in metric units). The upper half of the spreadsheet shows the measured values under different testing conditions. These values include: mean deviator load, mean applied deviator stress, mean recoverable deformation from each LVDT, mean resilient strain, and mean M_R value. A logarithmic plot of resilient modulus versus deviator stress was created using these values as shown in the lower left hand corner. In addition, a simple linear regression analysis was performed to develop a general equation for determining the resilient modulus value as a function of the deviator stress (M_R = f(σ_d)). After obtaining the equation, a deviator stress of 41.4-kPa (6-psi), suggested in the literature, was substituted into the equation to determine a design resilient modulus value. The lower right hand portion of the spreadsheet summarizes other important laboratory information, such as the R-value specimen height, the linear regression equation for M_R, the coefficient of determination (R²), and the condition of the sample. Similar spreadsheets were created for each test site and sample condition (undisturbed ring, undisturbed actuator, disturbed ring, and disturbed actuator). An example of the M_R calculation sheet used to create the summary sheet is presented in Appendix B.

Figure 4.1 M_R Summary Spreadsheet

Tables 4.4 and 4.5 summarize all the laboratory subgrade soil resilient modulus values. Some of the tests were not completed, shown by a blank space, because testing could not be performed on an undisturbed sample or the values obtained from the test were significantly beyond the range of realistic subgrade M_R values.

Back Calculated Resilient Modulus Values

Information obtained from pavement coring and deflection testing were used in this part of the research. First, the deflection measurements were corrected to a standard temperature of 21° Celsius (70° Fahrenheit) using a computer program called TAFFY (1988). This computer program produces temperature adjustment factors for the deflection readings based on an algorithm recommended by the Asphalt Institute. This program requires the following information: thicknesses of asphalt cement and untreated base layers, surface temperature, and previous 5-day air temperature history. Since the previous 5-day air temperature history was not known in this research, an average high and low temperature were entered into the program. These values were obtained by taking the recorded air temperature during the deflection testing and adding or subtracting 8.33 degrees to obtain Celsius temperatures (15 degrees to obtain Fahrenheit temperatures), respectively.

The computer programs, MODULUS, EVERCALC, and BOUSDEF were used in this analysis. All three programs require the following input parameters: magnitude of the load creating the deflection basin, the FWD load plate radius, distance of the sensors from center of the load plate, corrected deflection measurements, layer thicknesses, and the estimated Poisson's ratio values for all layers. Tables 4.6 and 4.7 summarize the back calculated resilient modulus values obtained with these three programs.

Results from R-Value Tests

After completing all resilient modulus tests, the soil samples were re-compacted and prepared for the R-value test. Each soil sample was compacted in a 102-mm (4-in.) diameter and 64-mm (2.5-in.) high mold by using static load compaction. The specifications for Resistance R-Value and Expansion Pressure of Compacted Soils AASHTO T 190 (ASTM D 2844) outline these testing procedures. In addition, the final R-values were corrected for variations in specimen height. Tables 4.8 and 4.9 summarize the results of the R-value tests for both sampling periods.

Statistical Analysis

As mentioned earlier, data were obtained from nine different sites during two different time periods, the summer of 1992 and the spring of 1993. Five of these sites were common to both time periods, one was specific to the summer of 1992, and three were specific to the spring of 1993. Table 4.10 summarizes the sites analyzed in each time period.

As a result of the laboratory and back calculation tests, several measured variables were available for analysis. These variables included: the resilient modulus (measured under four conditions), R-value, and certain soil characteristics (actual and optimum water contents, plasticity index, soil classification, and group index). Because the nine sites had a variety of soil classifications, statistical analyses were completed by taking into account these differences as necessary. In addition, all analyses were based upon log₁₀(M_R), abbreviated as LMR, instead of M_R itself because this minimized the differences between high and low resilient modulus values obtained at each test site.

Relationship Between Resilient Modulus and R-Value

Because the resilient modulus and the R-value provide similar information on a section's subgrade, one would assume that a relationship exists between these two laboratory tests. As a result, correlations were obtained between the measured R-values and the four measured resilient modulus conditions for both time periods. These four conditions were the undisturbed M_R from the ring, the undisturbed M_R from the actuator, the disturbed M_R from the ring, and the disturbed M_R from the actuator. Recall, the ring refers to the LVDT's placed inside of the testing chamber and the actuator refers to the LVDT's placed on the loading piston. Table 4.11 presents the correlations obtained from these laboratory measurements. Comparisons can be made within the rows of this table because they are based on the same soil samples. However, differences in the soil classifications between Periods A and B may distort comparisons between rows. Overall, this table shows that the disturbed soil LMR's were not significantly correlated with the R-value, but that the undisturbed soil LMR's were correlated with the R-value. Correlations between undisturbed and disturbed LMR's (not shown) were modest to nonexistent. Therefore, samples should remain undisturbed if the resilient modulus is to be a meaningful measure for pavement design. Only undisturbed LMR's were used in remaining analyses, unless noted otherwise.

The Effect of Sensor Locations on M_R Measurements

The correlations shown in Table 4.11 also favor the placement of the LVDT's outside the testing chamber on the loading piston (actuator) instead of on the rings inside the chamber. However, observed differences in the correlations with the R-values were not extreme, and placements were also compared on the basis of measurement precision. In order to ensure that all variability measured was attributable to differences in measurement methods, values were adjusted for site, period, and sample tube. The test for differences in variances for paired data (Snedecor & Cochran, 1989) showed the ring variance to be greater than the actuator variance (t=2.238, df=20, p=0.0368). The greater variation in ring measurements can be explained by the fact that it is difficult to obtain good contact between the LVDT's on the ring and the soil sample. Therefore, the remaining analyses were completed using actuator measurements only.

Although measurements at the actuator appear to be preferable, the possible relationship between actuator and ring measures was examined. Table 4.12 shows a high correlation between actuator and ring measurements of LMR. In addition, a t-test of paired differences indicates that ring measurements were on average higher than actuator measurements. For undisturbed samples, a repeated measures analysis indicates a similarity in differences between ring and actuator measurements (p=0.206).

The Effect of Sample Locations on M_R Values

Sample selection from the Shelby tubes is an important issue when determining the resilient modulus value. If the layers within a tube systematically differ from each other, with the upper portion consistently having higher or lower values than the lower portion, one would expect a noticeable difference in the values obtained from the selected samples. However, available data do not yield evidence of such differences (repeated measures analysis F_2,13=1.27, p=0.3126). On the other hand, if one assumes the layers are similar to each other, averaging the LMR values will give more reliable results than using the value from a single layer. Overall, it is not possible with the available data in this research to select one layer over another without an additional reference criterion.

Relationship Between Back Calculated and Laboratory M_R Values

Besides laboratory testing, M_R values can also be determined by back calculations using information from non-destructive tests. As mentioned earlier, the following three back calculation computer programs were utilized in the research: MODULUS (MP), EVERCALC (EP), and BOUSDEF (BP). In order to consider the quality of these programs, logs of back calculated values (designated as LMR-MP, LMR-EP, and LMR-BP, respectively) were compared to laboratory LMR values. The site-by-period mean LMR from undisturbed samples measured on the actuator was used as the best available value for the "true" resilient modulus, the one exception being a single site for which only ring measurements were available in Period A. Because means were calculated from a different number of observations, a weighted analysis was used (weight = sample size). Table 4.13 presents the results of this analysis. Note that the EVERCALC program appears to be slightly superior to the other two back calculation programs. In general, all back calculated values match better with each other than they do with the laboratory measurements.

Assuming constant differences between logs of back calculated and laboratory values, the best estimated differences appear in Table 4.14, along with implied relationships between laboratory and back calculated values of M_R. A 95% confidence interval for the appropriate correction factor (C) for subgrade soils in Wyoming, based on the EVERCALC program, is [0.20, 0.32], where M_R = C * [back calculated M_R value].

Relationship Between M_R Values and Soil Properties

Another important question to consider when selecting a M_R value is the relationship with common soil properties. The possible relationship between LMR and four factors, moisture = (actual % water content - optimum % water content), plasticity index, soil classification, and group index were analyzed. Because the group and plasticity indices were highly correlated, only one was ultimately considered for describing soil-M_R relationships, group index (GI).

Moisture and LMR were related, and their relationship depended on soil type. Similar strengths of the relationship between soil factors and responses were found for both undisturbed and disturbed (remolded) samples, and also for R-values (refer to Table 4.15). All of the test sections had one or more of the following types of AASHTO subgrade soil:  A-4, A-6, and A-7-6. For each of these classifications correlations were developed to determine the effect of moisture on the measured values. Overall, values for undisturbed and remolded M_R values and R-values  from A-4 and A-6 soils decreased as water content increased. The A-7-6 subgrade soils, however, showed very little change in the measured values (refer to Table 4.16).

Chapter Summary

In this chapter, the results from the laboratory tests and back calculation computer programs were presented. Several statistical analyses were also conducted and summarized to evaluate the factors influencing the determination of the M_R value used in designing new pavements or overlays. In general, these analyses indicated that the design resilient modulus value should be chosen based on laboratory tests using undisturbed soil samples and the actuator LVDT's. Multiple M_R values obtained from the same Shelby tube should also be averaged to give a better representation of the subgrade soil. The M_R values calculated from the equations based on the actuator LVDT deformation readings will be used to determine overlay thicknesses at each test site. This analysis will be presented and analyzed in the following chapter.