Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
63 Performance Data Analysis Approach The effects of several factors, including the experimental design factors, on the development of roughness and distress in SPS-1 flexible and SPS-2 rigid pavements are examined in this chapter. The SPS-1 experimental design factors include ⢠AC thickness, ⢠Base thickness, ⢠Base type, ⢠Subdrainage, ⢠Climate, ⢠Subgrade, and ⢠Traffic. For SPS-2 pavements, the experimental factors include ⢠PCC thickness, ⢠Concrete flexural strength, ⢠Base type, ⢠Slab width, ⢠Subdrainage, ⢠Climate, ⢠Subgrade, and ⢠Traffic. The last three factors listed for each pavement type (climate, subgrade, and traffic) are those that would be expected to be most useful in explaining differences in observed performance among pavements of similar design at different locations. The other factors listed are those that would be expected to be most useful in explaining differences in observed performance among different test sections at a given site. The main goal of this study is to assess the effects of subdrainage on the performance of the pavements in the SPS-1 and SPS-2 experiments. The previous phase of this research demonstrated the feasibility of making this assessment by com- paring distress and roughness between drained and undrained pairs of test sections at each site (13). For example, referring back to the SPS-1 experimental design matrix presented in Chapter 2 (Tables 4 and 5), the relative effects of undrained dense aggregate base and drained permeable asphalt-treated base may be assessed by comparing distress and roughness in the following test section pairs at each site: ⢠0101 versus 0110, ⢠0102 versus 0111, ⢠0113 versus 0122, and ⢠0114 versus 0123. In each of these four pairs, the design AC surface thickness and the design base thickness are the same, and the subgrade, traffic, and climate are the same at each site. Thus, analyzing all of the relevant pairs at all of the SPS-1 sites using paired difference t-tests blocks the effects of these other factors and detects any significant differences in performance that can be attributed to some sections having undrained dense aggregate base and others having permeable asphalt-treated base. There are some limitations to this pairwise comparison approach to analysis of the performance data. One limitation is that treating some factors as qualitative rather than quan- titative variables (for example, using design layer thicknesses rather than as-constructed thicknesses) masks the contribu- tion of variation in these factors to the overall variation in observed performance. Another limitation is that blocking out the effects of factors that differ by site (climate, subgrade, and traffic) focuses the analysis only on the one experimental factor analyzed (the base/drainage factor) and precludes assessment of the poten- tial effects of climate, subgrade, and traffic. Blocking for these other factors does not, as some believe, confuse the analysis of the factor of interest by failing to take their influence into account; rather, it clarifies the analysis by taking their influence C H A P T E R 5 Roughness and Distress in SPS-1 Flexible and SPS-2 Rigid Pavements
64 into account in the appropriate manner. This does mean, how- ever, that an analysis of this nature only has something to say about the statistical significance of the one factor analyzed. It cannot yield quantitative conclusions about the relative statis- tical significance of the other factors in the experiment. Another obstacle exists that no choice of statistical test can overcomeânamely, the base type and subdrainage factors are confounded in both the SPS-1 experiment and the SPS-2 experiment. So while it is possible to test for statistically sig- nificant differences in performance between, for example, sections with undrained dense aggregate base and otherwise equivalent sections with permeable asphalt-treated base, it is not possible to determine, on the basis of a statistical test, whether any such differences in performance are due to the two different base types or to the two different drainage situ- ations. The larger question, however, is how much does the base/drainage factor of the SPS-1 and SPS-2 experimental de- signs influence performance, compared with other experi- mental factors and site features? If the base/drainage factor has a strong influence on the performance of the different pavement sections at the SPS-1 and SPS-2 sites, it then becomes necessary to consider the results of the statistical tests of the performance data together with the results of the deflection analysis, flow time testing, and assessment of the natural drainage characteristics of the subgrade soil to make some judgments about how much the stiffness of the base seems to be the influential aspect of that experimental design factor, versus how much the quality of base drainage seems to be the influential aspect. If, on the other hand, the base/ drainage factor does not have a strong influence on the performance of the different pavement sections at the SPS-1 and SPS-2 sites, then it becomes much less important to try to determine whether base stiffness or the quality of drainage is the key aspect of that experimental design factor. To assess the relative importance of the base/drainage factor, compared with the other design and site factors in the SPS-1 and SPS-2 experiments, the available perform- ance (roughness and distress) data were analyzed using an approach that does not rely on test section pair comparisons alone, but rather examines the effects of all of the experimen- tal factors together. This approach employs regression analy- sis to detect which of the factors involved are significant. Regression analysis also overcomes the other limitation to pairwise comparison mentioned earlier. It allows qualitative and quantitative variables to be considered together, which permits quantitative consideration of factors such as layer thickness, subgrade stiffness, climate, and traffic level. This reduces the ânoiseâ due to section-to-section and site-to-site variations in these factors, which might otherwise mask the contribution of variation in these factors to the overall varia- tion in observed performance. Although testing for significance of factor effects applies sta- tistical tests to the linear regression of performance differences with respect to each of the factors, it does not imply a pre- sumption that those relationships are better described by linear, rather than nonlinear, regression. The question of interest is not what is the nature (linear or nonlinear) of the relationship of the factors to the observations, but whether any relationship exists at all between the factors and the observations. Linear regres- sion is a tool for detection of significant factor effects that should be the first step in model buildingâthe step that iden- tifies which variables should be included in any kind of predic- tion model subsequently developed. The next step in the model-building process, which is beyond the scope of this study, would be to identify the model form that yields the best prediction of the observed performance measure as a function of the factors that have been found to be significant. Regression Model for Assessing Effects of SPS-1 Experimental Factors The following regression model was used to assess the significance of subdrainage and other SPS-1 experimental factors to the development of pavement roughness and distress: Y = a0 + a1 YFIRST + a2 HAC + a3 HB + a4 B1 + a5 B2 + a6 B3 + a7 B4 {+ a8 DRN } + a9 TMP + a10 PRECIP + a11 ESUB + a12 HEQUIV + a13 CESAL + a14 TIME where Y = latest available measurement of performance measure of interest (distress or international roughness index [IRI]), or change in perform- ance measure; YFIRST = first available measurement of performance measure of interest; HAC = as-constructed AC surface thickness (in.); HB = total thickness of as-constructed base and sub- base, if any (in.); B1 to B4 = SPS-1 base type variables (defined below); DRN = 1 if drained, 0 if not drained; TMP = average annual temperature (°F); PRECIP = average annual precipitation (in.); TMI = Thornthwaite moisture index; ESUB = backcalculated subgrade modulus (psi) (see Chapter 4); HEQUIV = backcalculated equivalent pavement thickness (in.) (see Chapter 4); and CESAL = accumulated 18-kip ESALs from date of open- ing to traffic to date of Y measurement. The four SPS-1 base type variables (B1 to B4) are dummy variables that identify which of the SPS-1 experimentâs five
base types is present. The values assigned to each of the base variables, as well as to the DRN variable, for each of the five base types are shown in Table 22. Note that the value of DRN can always be determined from the values of B1, B2, B3, and B4. The DRN variable and the base type variables are redun- dant. The DRN variable thus cannot be considered in the regression analysis together with the base type variables, as the regression analysis cannot be run correctly if collinearity is introduced. The DRN variable in the regression model form above is for illustrative purposes only and is thus shown in braces. The DRN variable was not used in the regression analysis; instead, the four base type variables were used. The design, climate, and backcalculation results used in the regression analyses of the SPS-1 performance data are shown by test section in Table C-1 in Appendix C. The data used are shown in Table C-2. The accumulated ESAL and age data used are shown in Table C-3. Regression Model for Assessing Effects of SPS-2 Experimental Factors The following regression model was used to assess the sig- nificance of subdrainage and other SPS-2 experimental factors to the development of pavement roughness and distress: Y = b0 + b1 HPCC + b2 HIGH + b3 WIDE + b4 B1 + b5 B2 {+ b6 DRN } + b7 TMP + b8 PRECIP + b9 K + b10 CESAL + b11 YFIRST + b12 BAR + b13 AC where Y = latest available measurement of performance measure of interest (distress or IRI), or change in performance measure; HPCC = as-constructed concrete slab thickness (in.); HIGH = 1 if design 28-day concrete strength = 900 psi, and 0 if design 28-day concrete strength = 550 psi; WIDE = 1 if outer slab constructed 14 ft wide and 0 if outer slab constructed 12 ft wide; B1, B2 = the SPS-2 base type variables; B3, B4, B5 = 0-1 variables for base types in supplemental SPS-2 test sections (HMAC, none, or CAM, respectively); DRN = 1 if drained, 0 if not drained; TMP = average annual temperature (°F); PRECIP = average annual precipitation (in.); K = estimated static k value from backcalculation (psi/in.); CESAL = accumulated 18-kip ESALs from date of opening to traffic to date of Y measurement; and YFIRST = first available Y measurement. The AC variable is used for two supplemental test sections at the Arizona SPS-2 site: AC = 1 if pavement type is AC and 0 if pavement type is PCC. The BAR variable is used to identify supplemental test sections without dowels at the Arizona, North Dakota, and Washington SPS-2 sites: BAR = 1 if PCC pavement is dowelled and 0 if un- dowelled. The two base variables B1 and B2 are dummy variables that identify which of the SPS-2 experimentâs three base types is present. The values assigned to each of the base variables, as well as to the DRN variable, for each of the three base types, are shown in Table 23. As with the DRN and the base type variables in the SPS-1 experiment, the B1 and B2 base type variables in the SPS-2 experiment indicate the value of, and are redundant with, the DRN variable. There are, however, several supplemental sections in the SPS-2 experiment with a base type other than one of the three base types in the main experiment. There are some sections with a hot-mix asphalt concrete base, one section with no 65 Base Type B1 B2 B3 B4 DRN Dense-graded aggregate 0 0 0 0 0 Asphalt-treated base 1 0 0 0 0 Asphalt-treated base over dense-graded aggregate subbase 0 1 0 0 0 Permeable asphalt-treated base over aggregate subbase 0 0 1 0 1 Asphalt-treated base over permeable asphalt-treated subbase 0 0 0 1 1 Table 22. Values of base and drainage variables in the SPS-1 performance regression equation.
66 base at all, and some sections with a cement-aggregate mix- ture base. The presence of one of these three base types is indicated by a value of 1 for the base variable B3, B4, or B5. Not to include these three supplemental base type variables would erroneously suggest that a section for which B1 = 0 and B2 = 0 had the third type of base in the main experiment (the undrained dense-graded aggregate). Since the type of drainage is not generally known with a great degree of confi- dence for the supplemental test sections, no value (neither 0 nor 1) is assigned to the DRN variable for those sections. The design, climate, and backcalculation results used in the regression analyses of the SPS-2 performance data are shown by test section in Table C-4 in Appendix C. The performance data used are shown in Table C-5. The accumulated ESAL and age data used are shown in Table C-6. Selection of IRI Data for Analysis Roughness in the SPS-1 and SPS-2 pavement sections was analyzed using IRI data extracted from the MON_ PROFILE_MASTER table in the LTPP database. The IRI is a roughness parameter obtained from a mathematical model applied to a measured profile. The model simulates a quarter- car (one-wheel) system traveling at 80 km/h. The IRI value is the cumulative vertical movement of the suspension of the quarter-car system, divided by the traveled distance. The MON_PROFILE_MASTER table reports three IRI values for each profile run: the left wheelpath IRI, the right wheelpath IRI, and the average of the two values. One deci- sion that must be made in the analysis of IRI data is which of these three IRI values to use. If there is no significant differ- ence between left wheelpath and right wheelpath IRI, then they can be presumed to be samples from the same popula- tion. If this is the case, it does not matter which value (the left wheelpath, the right wheelpath, or the average) is used. If, on the other hand, there is a significant difference between left wheelpath and right wheelpath IRI, that means that the two are samples from different populations, and one of the two should be selected and used consistently in the analysis. In this case, the average of the two values is not a better indica- tor of the true IRI than either one of them, nor even as good an indicator as either one of them. To assess this, left and right wheelpath IRI values were compared in paired difference t-tests using the results from all of the profile runs conducted on SPS-1 and SPS-2 test sec- tions from the date of their construction through 2004. The results, shown in Table 24, indicate that the left and right wheelpath IRI values are significantly different in both the SPS-1 and SPS-2 data sets. The mean right wheelpath IRI was greater than the mean left wheelpath IRI in both cases; the right wheelpath IRI was thus selected for use in further analy- ses for this study. For each profile measurement date, the mean right wheelpath IRI was calculated from the right wheelpath IRI values from all of the profile runs on that day (usually five runs, but sometimes as few as one or as many as 15). These mean right wheelpath IRI values were used in the regression analysis. Base Type B1 B2 DRN Dense-graded aggregate 0 Lean concrete base 1 Permeable asphalt-treated base 0 0 0 0 0 1 1 Table 23. Values of base and drainage variables in the SPS-2 performance regression equation. SPS-1 SPS-2 Left Wheelpath Right Wheelpath Left Wheelpath Right Wheelpath Mean IRI (m/km) 0.958 0.992 1.350 1.401 Mean difference 0.034 0.051 Standard deviation sD 0.208 0.253 n 9879 9954 Calculated t 16.05 20.05 t 0.025, n-1 1.96 1.96 Significant difference? yes yes Table 24. Tests of significant difference between left and right wheelpath IRI.
Fluctuations in IRI not Due to Pavement Deterioration The expectation is that IRI will tend to increase over time, as the pavement deteriorates. IRI does not, however, always increase steadily over time. Sometimes the IRI of a test sec- tion is lower than that measured at the same test section a year earlier, a month earlier, or even a day earlier. There are sev- eral reasons why IRI might decrease from one testing date to the next, including the following: ⢠Rehabilitation or maintenance between testing dates; ⢠Seasonal variation; ⢠Measurement in different paths; ⢠Different starting locations; ⢠Spikes in the data caused by reflection of light from the white paint stripe at the start of a test section; and ⢠Problems with the profilometer electronics, sensors, or distance measurement (32). It is not necessarily true, however, that an IRI decrease, or an IRI increase for that matter, always has a physical expla- nation. Some portion of the variation seen in IRI data is random variationâthat is, some fluctuations in IRI, both upward and downward, are not significantly different from no change at all. Initial IRI Values Examination of the first available IRI values for the SPS-1 and SPS-2 test sections reveals a surprisingly large disparity between the two experiments in these initial (or early) levels. This is illustrated by the cumulative frequency distributions shown in Figure 91. The mean first IRI for the SPS-1 test sections was 0.88 m/km, while the mean first IRI for the SPS-2 test sections was 1.30 m/km. This disparity cannot be attributed to the first SPS-1 IRI values being obtained from profile runs conducted sooner after the pavement was opened to traffic than were the SPS-2 IRI values. As Figure 92 shows, there is not much difference in the cumulative frequency distributions of time from the date of opening to traffic to the date of the first available IRI data. Indeed, Figure 92 shows that the first IRI values after the pavement was opened to traffic tended to be obtained sooner 67 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40 2.60 First IRI (m/km) SPS-1 SPS-2 0 10 20 30 40 50 60 70 80 90 100 Cu m ul at iv e pe rc en ta ge Figure 91. Cumulative frequency distributions of first IRI of SPS-1 and SPS-2 test sections.
68 from SPS-2 sections than from SPS-1 sections. Figures 91 and 92 together indicate that, in general, better initial smoothness levels were obtained in the construction of the SPS-1 test sections than in the SPS-2 test sections. The same might not necessarily be true, however, of more conventional paving (as opposed to the 500-ft test sections of varying design, as in the SPS-1 and SPS-2 experiments). Regression Analysis of Factors Affecting IRI In SPS-1 Flexible Pavements The regression models presented previously were used to determine which of the various experimental design factors and site factors in the SPS-1 and SPS-2 experiments have had a significant effect on roughness development. The relative contributions of the SPS-1 factors to the r2 of the regression model for latest observed IRI are summarized in Table 25. The factor most significant to the last observed IRI was initial IRI, which alone contributes 26% of the 38% total r2 possible when all other factors are included in the regression. The latest observed IRI values of the SPS-1 test sections were more strongly related to the initial IRI values than to all of the other factors combined. The latest IRI values of the SPS-1 test sections are plotted against the initial IRI values in Figure 93, along with the linear -10 0 10 20 30 40 50 60 70 80 90 100 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 Time to first IRI (yrs) Cu m ul at iv e pe rc en t f re qu en cy SPS-1 SPS-2 Figure 92. Cumulative frequency distributions of time from date of opening to traffic to date of profile run corresponding to first available IRI values, for SPS-1 and SPS-2 test sections. Independent Variable Combined r2 with this variable added FIRST_IRI 0.26 TIME_LAST_IRI 0.30 H_EQUIV 0.31 TMI 0.33 PRECIP 0.34 ESUB 0.35 B2 0.35 B3 0.36 B4 0.37 HB 0.37 CESAL_LAST_IRI 0.38 TMP 0.38 HAC 0.38 B1 0.38 Table 25. Significance of SPS-1 regression variables to last IRI.
trend line for last IRI versus initial IRI for each base type. It is not surprising that the pavement sections that are rougher soon after construction would also be rougher at a later point in time. What is surprising is that the roughness at the later point in time would be more strongly correlated to the initial roughness than to any and all of the other experimental design factors. The next most influential factor in the regression for last IRI was the time to the measurement of the last IRI (that is, the age of the pavement section). This is also as one would expect, although it might surprise some to see that accumu- lated ESALs are much lower on the list of variables, in order of contribution to r2, than age. In fact, both age and accumu- lated ESALs were about equally well correlated (about 10%) to last IRI when analyzed independently. Because age and accumulated ESALs are themselves fairly strongly correlated, once either of these two terms is in the regression model, the addition of the second one does not do much to improve r2. Similarly, the backcalculated equivalent thickness of the pavement structure was more significant in the regression for last IRI than was either AC surface thickness or base thick- ness, both of which are reflected in the backcalculated equiv- alent thickness. Not surprisingly, once this variable is in the model, the AC surface thickness and base thickness terms do not add anything to r2. After initial IRI and age, the next most influential factors in the regression for last IRI were the backcalculated equiva- lent thickness of the pavement structure, the Thornthwaite moisture index, and the average annual precipitation. Those five factors account for 34% of the 38% total r2 possible. The base type/drainage factors, together with all other factors, increase r2 by only 4%. The cumulative frequency distributions of last IRI for the five different base type/drainage combinations in the SPS-1 experiment are shown in Figure 94. There is an evident sep- aration between the three groups of pavements that have an asphalt-treated base layer (undrained AC/ATB, undrained AC/ATB/DGA, and drained AC/ATB/PATB) and the two groups of pavements that do not have an asphalt-treated base layer (undrained AC/AGG and drained AC/PATB/ AGG). It is reasonable to conclude from this, along with the finding concerning the significance of backcalculated equiv- alent thickness to the regression, that whatever fairly minor effect the base type/drainage factor had on last IRI was due 69 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 First IRI (m/km) La st IR I (m /km ) AC/AGG AC/ATB AC/ATB/AGG AC/PATB AC/ATB/PATB Linear (AC/ATB) Linear (AC/ATB/AGG) Linear (AC/PATB) Linear (AC/ATB/PATB) Linear (AC/AGG) Figure 93. Linear correlations of latest IRI to initial IRI for SPS-1 pavement sections.
70 to differences in base stiffness and not differences in drainage. The relative contributions of the SPS-1 factors to the r2 of the regression model for change in IRI (latest minus initial) are summarized in Table 26. The four most influential variables were age, equivalent thickness, Thornthwaite moisture index, and precipitation. Initial IRI contributed very little to the regression for change in IRI over time. This finding, together with the finding mentioned above concerning the effect of ini- tial IRI on latest observed IRI, indicates that for the SPS-1 experiment, those pavements that were rougher initially were, not surprisingly, rougher later, but not because they increased in roughness at a faster rate. This is illustrated in Figure 95, in which change in IRI is plotted against initial IRI. For some base types, there is a slight upward trend, while for others there is a slight downward trend; the overall trend, indicated by the regression line and equation shown, is a slightly negative slope that is not significantly different from zero. The cumulative frequency distributions of change in IRI for the five different base type/drainage combinations in the SPS-1 experiment are shown in Figure 96. Again, the three groups of pavements with an asphalt-treated base layer (undrained AC/ATB, undrained AC/ATB/DGA, and drained AC/ATB/PATB) exhibit smaller changes in IRI than the two groups of pavements that do not have an asphalt-treated base layer (undrained AC/AGG and drained AC/PATB/AGG). However, the difference between the two sets of distributions is fairly small, at least in the vicinity of the medians (50th per- centiles) of the distributions. The separation of the AC/AGG distribution curve from those of the other four groups at 0 10 20 30 40 50 60 70 80 90 100 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 4.00 4.25 4.50 4.75 5.00 Last IRI (m/km) Pe rc en t o f s ec tio ns AC/AGG AC/ATB AC/ATB/DGA AC/PATB/AGG AC/ATB/PATB Figure 94. Cumulative frequency distributions of last IRI for SPS-1 pavement sections. Independent Variable Combined r2 with this variable added TIME_DELTA_IRI 0.05 H_EQUIV 0.08 TMI 0.10 PRECIP 0.11 TMP 0.12 FIRST_IRI 0.12 HB 0.12 B4 0.13 B3 0.13 B2 0.13 CESAL_DELTA_IRI 0.14 ESUB 0.14 HAC 0.14 B1 0.14 Table 26. Significance of SPS-1 regression variables to change in IRI.
71 y = -0.1544x + 0.3484 R2 = 0.0054 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 First IRI (m/km) Ch an ge in IR I (m /km ) AC/AGG AC/ATB AC/ATB/AGG AC/PATB AC/ATB/PATB All Linear (All) 0 10 20 30 40 50 60 70 80 90 100 -0.50 -0.25 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 Change in IRI (m/km) Pe rc en t o f s ec tio ns 2.50 AC/AGG AC/ATB AC/ATB/DGA AC/PATB/AGG AC/ATB/PATB Figure 95. Linear correlation of change in IRI to initial IRI for SPS-1 pavement sections. Figure 96. Cumulative frequency distributions of change in IRI for SPS-1 pavement sections.
72 higher cumulative percentage levels indicates that the largest changes in IRI tended to occur in this group, and this is con- firmed by Figure 95 as well. The small differences in median change in IRI raise the question of whether different rates of change in IRI over time in service are entirely responsible for the differences seen in last IRI for the different groups. The cumulative frequency distributions of the initial IRI values were examined, and as Figure 97 shows, the pavement sections without ATB (un- drained AG/AGG and drained AC/PATB/AGG sections) tended to have higher initial IRI values than the pavement sections with ATB (undrained AC/ATB, undrained AC/ATB/ DGA, and drained AC/ATB/PATB). It is thus reasonable to conclude that whatever fairly minor effect the base type/drainage factor has had on the latest observed IRI values and rates of change in IRI over time for the SPS-1 pavement sections has been due to differences in base stiffness, not differences in drainage. Furthermore, the differences in IRI by base type are not entirely due to differ- ent rates of change in IRI over the time that the pavement sec- tions have been in service, because there is evidence that the pavements with weaker bases (lower backcalculated effective thickness) also tended to be rougher initially. Regression Analysis of Factors Affecting Rutting in SPS-1 Flexible Pavements The relative contributions of the SPS-1 factors to the r2 of the regression model for rutting are summarized in Table 27. Age was by far the most significant factor in the regression, contributing 25% to the total possible r2 of 51%. However, the correlation of age to rutting in the SPS-1 data is not as positive a correlation as one would expect. As Figure 98 shows, there is actually a negative correlation between change in rutting (latest measurement minus first measurement) and age for three of the base groups (AC/AGG, AC/ATB, and AC/PATB) and essentially no correlation for the other two groups (AC/ATB/PATB and AC/ATB/AGG). The negative correlations for the first three groups are due to some pave- ment sections with unusually high rutting at a young age (less than 6 years). The greater consistency of rutting values in pavement sections older than 6 years suggests that aside from some pavement sections that developed unusually high rut- ting at a young age, the majority of the SPS-1 pavement sections do not appear to have started to develop increased rutting with increasing age (or with accumulated traffic). 0 10 20 30 40 50 60 70 80 90 100 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 Initial IRI (m/km) Pe rc en t o f s ec tio ns AC/AGG AC/ATB AC/ATB/DGA AC/PATB/AGG AC/ATB/PATB Figure 97. Cumulative frequency distribution of initial IRI for SPS-1 pavement sections.
The backcalculated equivalent thickness of the pavement structure was the second most influential factor in the regression for rutting, raising the r2 by another 10%. Rutting values are plotted against the backcalculated equivalent thick- nesses of the SPS-1 pavement sections in Figure 99. Note that the handful of pavement sections with unusually high rutting levels all had fairly low backcalculated equivalent thicknesses, and most of them were AC over undrained aggregate base. Figure 99 illustrates the different ranges in backcalculated equivalent thickness for different base types (the AC/AGG and AC/PATB sections having lower equivalent thicknesses than the AC/ATB, AC/ATB/PATB, and AC/ATB/AGG sections) and that pavement sections with higher equivalent thicknesses tend to have less rutting. The cumulative frequency distributions of rutting for the five different base type/drainage combinations in the SPS-1 experiment are shown in Figure 100. Four of the five groups have very similar distributions; the one that is noticeably dif- ferent is the undrained AC/AGG section, which had a larger percentage of sections with unusually high rutting values. Recall, however, from Figure 98 that most of these unusually high rutting values were measured on younger pavement sections. These higher rutting values were not anomalies observed at any one particular SPS-1 site: the sites with one or more sections with 12 mm or more of rutting were located in Kansas, Nebraska, Ohio, and Virginia. 73 Independent Variable Combined r2 with this variable added TIME_LAST_RUT 0.25 H_EQUIV 0.35 TMP 0.38 ESUB 0.41 TMI 0.43 PRECIP 0.46 HB 0.48 B3 0.49 CESAL_LAST_RUT 0.50 B2 0.50 B4 0.51 HAC 0.51 B1 0.51 -5 0 5 10 15 20 25 30 35 0 1 2 3 4 5 6 7 8 9 10 11 12 Time (years) Ch an ge in ru tti ng (m m) AC/AGG AC/ATB AC/ATB/AGG AC/PATB AC/ATB/PATB Linear (AC/ATB/AGG) Linear (AC/AGG) Linear (AC/PATB) Linear (AC/ATB/PATB) Linear (AC/ATB) Table 27. Significance of SPS-1 regression variables to rutting. Figure 98. Change in rutting versus age for SPS-1 pavement sections.
74 Average annual temperature, backcalculated subgrade modulus, Thornthwaite moisture index, and average annual precipitation were the next most influential factors in the re- gression for rutting. The base type/drainage factors, together with the remaining factors considered (accumulated ESALs, base thickness, and AC surface thickness, all of which are cor- related to other factors already in the model), add only 3% to the total r2. Since backcalculated equivalent thickness was an influ- ential factor in the regression for rutting, but the base type/drainage factors were not, it is reasonable to conclude that it is the stiffness of the base, not the presence or absence of drainage, that was responsible for whatever role the differ- ent base types played in the development of rutting in the SPS-1 pavement sections. This is reinforced by the similarity of the cumulative frequency distributions for rutting among the undrained AC/ATB and AC/ATB/DGA sections and the drained AC/PATB/AGG and AC/ATB/PATB sections. The weakest sections (undrained AC/AGG) were the ones most likely to exhibit more rutting after just a few years in service. Regression Analysis of Factors Affecting Cracking in SPS-1 Flexible Pavements The relative contributions of the SPS-1 factors to the r2of the regression model for cracking are summarized in Table 28. The most influential variable was the Thornthwaite mois- ture index, contributing 19% to the 44% total possible r2. The positive correlation between cracking and Thornthwaite moisture index is illustrated in Figure 101. Recall that a low Thornthwaite moisture index results from a combination of high temperatures and low precipitation, while a high Thorn- thwaite moisture index results from a combination of low temperatures and high precipitation. The next most influential variables were, in decreasing order of importance, the other climatic variables (average annual temperature and average annual precipitation), accumulated ESALs and pavement age, subgrade modulus, equivalent thickness, base type/drainage variables, and sur- face and base thickness. The relatively minor effects of these factors may be due to the fact that most of the SPS-1 sections 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 Equivalent thickness (in) R ut tin g (m m) AC/AGGAC/ATB AC/ATB/AGG AC/PATB AC/ATB/PATB Figure 99. Rutting versus backcalculated equivalent thickness for SPS-1 pavement sections.
have not yet developed much cracking, which is probably because of the relatively low truck traffic levels at most of the SPS-1 sites. The cumulative frequency distributions of cracking for the five different base type/drainage combinations in the SPS-1 experiment are shown in Figure 102. Roughly half of the sections in each group have no cracking. As with the IRI and rutting frequency distributions shown earlier, the weaker pavement sections (undrained AC/AGG group and drained AC/PATB/AGG) have more cracking than the stronger pave- ment sections. These findings suggest that whatever minor effect the base type/drainage factor has had on cracking in the SPS-1 pavement sections, it has been due to differences in base stiffness rather than differences in drainage. Regression Analysis of Factors Affecting IRI in SPS-2 Rigid Pavements The relative contributions of the SPS-2 factors to the r2 of the regression model for latest observed IRI are summarized in Table 29. As was the case for the SPS-1 pavements, the most influential variable was initial IRI. For the SPS-2 pave- ments, however, its influence was not as strong, contributing 11% to the total possible r2 of 41% (compared with con- tributing 26% of the total 38% r2 for the SPS-1 pavements). The latest IRI values of the SPS-2 test sections are plotted against the initial IRI values in Figure 103, along with the lin- ear trend line for last IRI versus initial IRI for each base type. The next most influential variables in the regression for last IRI were age, backcalculated k value, the base type vari- able B2 (indicating the presence or absence of a permeable 75 AC/AGG AC/ATB AC/ATB/DGA AC/PATB/AGG AC/ATB/PATB 0 10 20 30 40 50 60 70 80 90 100 Pe rc en t o f s ec tio ns 0.0 1.5 3.0 4.5 6.0 7.5 9.0 10.5 12.0 13.5 15.0 16.5 18.0 19.5 21.0 22.5 24.0 25.5 27.0 28.5 30.0 Rutting (mm) Figure 100. Cumulative frequency distributions of rutting for SPS-1 pavement sections. Independent Variable Combined r2 with this variable added TMI 0.19 TMP 0.21 PRECIP 0.30 CESAL_LAST_CRACK 0.33 TIME_LAST_CRACK 0.36 ESUB 0.39 H_EQUIV 0.40 B4 0.41 B1 0.41 B3 0.42 HAC 0.44 HB 0.44 B2 0.44 Table 28. Significance of SPS-1 regression variables to cracking.
76 asphalt-treated base), accumulated ESALs, Thornthwaite moisture index, temperature, and the BAR variable (indicat- ing the presence or absence of dowel bars). The other base type variables, along with slab width, concrete strength, as-built slab thickness, backcalculated equivalent slab thickness, and precipitation, contributed very little to the regression. The cu- mulative frequency distribution of last IRI for the different base type/drainage combinations in the SPS-2 experiment is shown in Figure 104. The solid lines shown indicate the three types of drainage in the main SPS-2 experiment: undrained aggregate, undrained lean concrete base, and drained perme- able asphalt-treated base. The dotted lines show the distribu- tions for two base types (both presumably undrained) found in some supplemental sectionsâHMAC (there are seven of these) and CAM (four of these). No distributions are shown for two other situations represented by supplemental SPS-2 sectionsâone section with a concrete slab on grade, without any base (at the Colorado SPS-2 site), and two sections of as- phalt concrete over aggregate base (at the Arizona SPS-2 site). The cumulative frequency distributions in Figure 104 indicate that the SPS-2 sections with undrained aggregate bases tend to have the highest IRI values, followed by the sec- tions with the undrained LCB sections and then by the drained PATB sections. The similarity of the distributions for the two undrained base types compared with the PATB distribution explains why the base type variable for PATB was the most significant variable in the regression. The distribu- tion for undrained HMAC base is between that of the drained PATB and the undrained LCB and AGG distributions, and the undrained CAM distribution is even better (lower IRI val- ues) than that of the drained PATB. Nonetheless, it might not be wise to put too much weight on the findings for the HMAC base and CAM base types, as there are few sections with these base types in the SPS-2 experiment. It is more difficult to determine for the SPS-2 experiment than for the SPS-1 experiment if what distinguishes the drained base sections from the undrained base sections is due to differences in stiffness of the base or to differences in drainage. In the case of flexible pavements, increased base stiffness is expected to increase overall structural capacity and thus improve performance. If pavements with drained asphalt-treated base perform better than pavements with 0 10 20 30 40 50 60 70 80 90 100 -60 -40 -20 -50 -30 -10 0 10 20 30 40 50 60 70 80 90 110 100 120 Thornthwaite moisture index Cr ac ki ng (p er ce nt ar ea ) AC/AGG AC/ATB AC/ATB/AGG AC/PATB AC/ATB/PATB Figure 101. Cracking versus Thornthwaite moisture index for SPS-1 pavement sections.
undrained aggregate base, but not better than pavements with undrained asphalt-treated base, it is reasonable to con- clude that the differences in performance are attributable primarily to the stiffness of the base, and not to the presence or absence of subdrainage. In the case of rigid pavements, however, the base stiffness that optimizes performance by achieving the best balance between load-related stresses and curling-related stresses is one that is neither too weak nor too stiff. Thus, it is not sur- prising that pavements with asphalt-treated base would perform better than pavements with either untreated dense- graded aggregate base or lean concrete base, all other things being equal. The difficulty arises in assessing whether the drainability of the permeable asphalt-treated base influenced performance beyond the influence attributable to the stiffness of the base. This is an assessment that needs to be made for each of the performance aspects considered. At least with respect to the analysis of latest available IRI data, one indication that drainage was not significant to the differences observed is that Thornthwaite moisture index and 77 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 Percent area cracked Pe rc en t o f s ec tio ns AC/AGG AC/ATB AC/ATB/DGA AC/PATB/AGG AC/ATB/PATB Figure 102. Cumulative frequency distributions of cracking for SPS-1 pavement sections. Independent Variable Combined r2 with this variable added FIRST_IRI 0.11 TIME_LAST__IRI 0.20 FIRST_K 0.25 B2 0.28 CESAL_LAST__IRI 0.30 TMP 0.32 TMI 0.35 BAR 0.37 B3 0.38 B1 0.40 B5 0.41 WIDE 0.41 AC 0.42 Hpcc 0.43 HIGH 0.43 FIRST_HEQ 0.43 PRECIP 0.44 B4 0.44 Table 29. Significance of SPS-2 regression variables to IRI.
78 average annual precipitation made very slight contributions (3% and 1%, respectively) to the regression. Another indication is that the pavement sections with one of the undrained base types (CAM) exhibited even lower IRI values than the pave- ment sections with PATB. Again, there are fairly few HMAC and CAM base sections in the experiment. It does make sense, though, that the influence of these base types on concrete slab performance might be comparable to that of PATB since, like PATB, both of them are more rigid than untreated dense- graded aggregate and less rigid than lean concrete. The relative contributions of the SPS-2 factors to the r2 of the regression model for change in IRI are shown in Table 30. The most influential variables were age, backcalculated k value, the B2 (PATB) base type/drainage variable, accumu- lated ESALs, average annual temperature, Thornthwaite moisture index, and dowel bar presence. As was the case in the regression for latest IRI, Thornthwaite moisture index and precipitation contribute little to the regression. The variables that were most influential in the regression for change in IRI are the same, but in a slightly different order, as those that were most influential in the regression for last IRI, except that initial IRI was not influential in the regression for change in IRI. In fact, initial IRI was not even selected by the regression algorithm for inclusion in the model, and thus does not appear in Table 30. Change in IRI is plotted against initial IRI for the SPS-2 sections in Figure 105. There is a slightly negative slope to the overall trend line, but the lack of significance of this slope is indicated by the very low r2 associated with the trend line. Similar results were obtained for change in IRI versus initial IRI for the SPS-1 pavements (see Figure 95). The cumulative frequency distributions of change in IRI for the different base type/drainage combinations in the SPS-2 experiment are shown in Figure 106. As was the case for latest IRI, the largest changes in IRI occurred in the undrained PCC/AGG sections, followed by the undrained PCC/LCB sections and then the drained PATB sections. Though they are few in number, the undrained PCC/HMAC and PCC/CAM sections exhibited even smaller changes in IRI than the PATB sections. The cumulative frequency distributions of initial IRI were examined to assess how much of the differences in latest IRI should be attributed to changes in IRI over time in service. These distributions are plotted in Figure 107. The disparity PCC/AGG PCC/LCB PCC/PATB PCC/HMAC PCC/no base PCC/CAM Linear (PCC/AGG) Linear (PCC/LCB) Linear (PCC/PATB) Linear (PCC/HMAC) Linear (PCC/CAM) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 First IRI (m/km) La st IR I (m /km ) Figure 103. Linear correlations of latest IRI to initial IRI for SPS-2 pavement sections.
among the initial IRI distributions for the different types of SPS-2 pavements is even greater than for the different types of SPS-1 pavements (see Figure 97). This is especially true in the broad middles of the distributions, in the vicinity of the median values. The SPS-2 sections with the highest median initial IRI values were the supplemental sections built on HMAC base. The next highest were the core experiment sections built on LCB, followed by those built on AGG base, followed by those built on PATB. The supplemental sections built on CAM base had the lowest median initial IRI. Figure 107 shows that sections built on PATB tended to be smoother initially than sections built on LCB or AGG, which explains some of the differences in latest IRI values among these base types. Figure 107 also shows that sections built on AGG tended to be smoother initially than sections built on LCB. This seems to have been countered by larger changes in IRI in the AGG sections than in the LCB sections (see Figure 106), resulting in fairly similar median values of latest observed IRI for AGG sections and LCB sections (see Figure 104). At the upper end of these cumulative frequency distri- butions, the pavement sections that exhibited the largest changes in IRI were mostly AGG sections, with some LCB sections (see Figures 103 and 105). From these findings it is reasonable to conclude that what- ever effect the base type/drainage factor has had on the SPS-2 pavement sectionsâ latest observed IRI values and rates of change in IRI over time has been predominantly due to dif- ferences in base stiffness. The potential effect of drainage is not necessarily ruled out, but no particular evidence was detected for the role of drainage, independent of the role of base stiffness, in the development of roughness in the SPS-2 79 PCC/AGG PCC/LCB PCC/PATB PCC/HMAC PCC/CAM 0 10 20 30 40 50 60 70 80 90 100 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 4.00 4.25 4.50 4.75 5.00 Last IRI (m/km) Pe rc en t o f s ec tio ns Figure 104. Cumulative frequency distributions of last IRI for SPS-2 pavement sections. Independent Variable Combined r2 with this Variable Added TIME_DELTA_IRI_ 0.14 FIRST_K 0.20 B2 0.22 CESAL_DELTA_IRI 0.24 TMP 0.26 TMI 0.29 BAR 0.31 B3 0.32 B1 0.34 B5 0.35 WIDE 0.36 Hpcc 0.36 AC 0.38 HIGH 0.38 FIRST_HEQ 0.38 B4 0.38 PRECIP 0.38 FIRST_IRI 0.38 Table 30. Significance of SPS-2 regression variables to change in IRI.
80 pavements. Furthermore, the differences in IRI by base type are not entirely attributable to different rates of change in IRI over the time that the SPS-2 pavements have been in service, because there is evidence of some significant differences in initial IRI values by base type. Of the three main base types in the SPS-2 experiment, the lean concrete base was associated with the highest initial IRI values, while the dense aggregate base was associated with the highest long-term IRI values. Selection of Faulting Data for Analysis In the LTPP studies, as in other pavement performance studies that include PCC pavements, it has been common practice to measure transverse joint faulting at about 1 ft and 2.5 ft from the outer lane edge. The Distress Identification Manual for the Long-Term Pavement Performance Program indicates that faulting at joints in LTPP sections should be measured â0.3 m and 0.75 m from the outside slab edge (approximately the outer wheel path)â (33). The LTPP data- base contains edge and wheelpath faulting measurements for joints and cracks in all of the jointed concrete test sections (for both SPS and general pavement studies sites). It is not obvious why field technicians should spend time measuring and recording faulting at two locations at each joint and crack or why time should be spent entering the measurements at both locations in the wheelpath. A rationale sometimes offered is that measurement at both locations offers compat- ibility with data from other studies in which faulting might have been measured at only one of the two locations. In fact, it does not appear that this has ever really been an issue in analysis of faulting data, since faulting has been measured at both the edge and wheelpath in most if not all of the major PCC pavement performance studies conducted in the United States over the past 30 years. The fact that pavement researchers attach importance to measuring faulting at both edge and wheelpath locations sug- gests that some significant difference is believed to exist be- tween the two. Yet when faulting data are used to develop faulting prediction models, it is common practice not to mention which set of faulting measurements was used in the model development and thus which locationâs faulting the model is presumed to predict (11, 34-36). A review of the literature has found only one comparison of edge and y = -0.2763x + 0.5893 R2 = 0.0215 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 First IRI (m/km) Ch an ge in IR I (m /k m ) . PCC/AGG PCC/LCB PCC/PATB AC/AGG PCC/HMAC PCC/no base PCC/CAM all Linear (all) Figure 105. Linear correlation of change in IRI to initial IRI for SPS-2 pavement sections.
81 PCC/AGG PCC/LCB PCC/PATB PCC/HMAC PCC/CAM 0 10 20 30 40 50 60 70 80 90 100 0.00-0.50 -0.25 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 4.00 4.25 4.50 Change in IRI (m/km) Pe rc en t o f s ec tio ns Figure 106. Cumulative frequency distributions of change in IRI for SPS-2 pavement sections. PCC/AGG PCC/LCB PCC/PATB PCC/HMAC PCC/CAM 0 10 20 30 40 50 60 70 80 90 100 0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 Initial IRI (m/km) Pe rc en t o f s ec tio ns Figure 107. Cumulative frequency distributions of initial IRI for SPS-2 pavement sections.
82 wheelpath faulting: in a 2000 evaluation of LTPP faulting data, in more than 90% of the pairs of edge and wheelpath faulting measurements examined, the difference between the two measurements was found to be between â1 mm and 1 mm (37). Since this range is the same as the precision of the faultmeter used in monitoring LTPP test sections, the researchers conducting that evaluation judged the difference between edge and wheelpath faulting to be insignificant. In this study, all of the edge and wheelpath faulting meas- urements for the SPS-2 rigid pavement test sectionsâmore than 40,000 edge and wheelpath measurement pairs in totalâwere retrieved from the LTPP database and subjected to a paired difference t-test. As the results in Table 31 show, wheelpath faulting exceeded edge faulting by a slight 0.006 mm on average, and this mean difference was not found to be statistically different from zero at the 95% confidence level. This finding reinforces the earlier studyâs conclusion that edge and wheelpath faulting are not significantly different, at least in the LTPP database (37). The wheelpath faulting data set was selected for use in the analysis for this study. Regression Analysis of Factors Affecting Faulting in SPS-2 Rigid Pavements The relative contributions of the SPS-2 factors to the r2 of the regression model for faulting are summarized in Table 32. Not surprisingly, the most influential factor in the regression was the BAR variable indicating the presence or absence of dowels, which contributed 13% to the total possible r2 of 31%. The next most influential factors were the accumulated ESALs, the B2, B1, and B3 base type/drainage variables (indicating the pres- ence of PATB, LCB, and AGG, respectively), and age. Variables related to the slab thickness, slab width, concrete strength, and backcalculated pavement stiffness or subgrade k value did not contribute much to the regression, nor did any of the three climatic variables that were considered. The cumulative frequency distributions of faulting for the different base type/drainage combinations in the SPS-2 experiment are shown in Figure 108. The distributions for undrained LCB and drained PATB are very similar, while the rightward displacement of the distribution for undrained AGG base indicates that pavements with this base type devel- oped more faulting than pavements with either of the other two base types. Plotting the distributions without including the undowelled sections does not change this disparity. Note that two of the undowelled sections had AGG base, four had PATB, one had LCB, and one had HMAC base. Among these eight undowelled sections, the two with the highest faulting levels, 2.9 and 3.0 mm, were the two undowelled sections with aggregate base at the Arizona SPS-2 site. Faulting in the remaining six undowelled sections (at the Arizona, North Dakota, and Washington sites) ranged from 0 to 1.3 mm. These findings, particularly the similarity of results for undrained LCB and drained PATB, suggest that whatever effect the base type/drainage factor has had on the develop- ment of faulting in pavements in the SPS-2 experiment has been due to the stiffness of these bases, compared with the lesser stiffness of the undrained dense-graded aggregate base. This conclusion is reinforced by the observation that the undowelled pavements with aggregate base developed more than twice as much faulting as undowelled pavements with drained or undrained stabilized bases, even those at the same sites. Regression Analysis of Factors Affecting Cracking in SPS-2 Rigid Pavements The relative contributions of the SPS-2 factors to the r2 of the regression model for cracking are summarized in Table 33. Age and Thornthwaite moisture index were the most in- fluential variables, with each contributing 13% to the total Wheelpath Minus Edge Faulting (mm) Mean difference 0.006 Number of pairs, n 41,168 SD 0.692 t calc 1.75 Test at 95% confidence level: 0.05α tα/2, n-1 1.96 Lower limit of confidence interval -0.0007 Upper limit of confidence interval 0.0126 Reject null hypothesis? no Table 31. Significance of difference in wheelpath versus edge faulting in SPS-2 data. Independent Variable Combined r2 with this Variable Added BAR 0.13 CESAL_LAST_FAULT 0.18 B2 0.20 B1 0.23 B3 0.25 TIME_LAST_FAULT 0.26 FIRST_HEQ 0.27 WIDE 0.28 B5 0.28 HIGH 0.29 TMI 0.29 PRECIP 0.30 TMP 0.30 FIRST_K 0.31 B4 0.31 Hpcc 0.31 Table 32. Significance of SPS-2 regression variables to faulting.
possible r2 of 42%. Accumulated ESALs, average annual tem- perature, and average annual precipitation were the next most influential variables. These were followed by the B1 base type variable (indicating the presence of lean concrete base), the backcalculated effective k value, and the B2 and B3 vari- ables (PATB and AGG base, respectively), although the con- tributions of the last three variables, and all of the remaining variables considered, were very slight. The cumulative frequency distributions of cracking for the three different base type/drainage combinations in the main SPS-2 experiment are shown in Figure 109. This plot shows that the LCB sections were much more likely to develop cracking than the AGG and PATB sections. More than 60% of the LCB sections had some cracking, while only about 30% of the AGG sections and the PATB sections had some crack- ing. Of those two, more cracking occurred in the AGG sec- tions than in the PATB sections. Distributions are not shown for the undrained PCC/HMAC and undrained PCC/CAM sections because none of these sections had any cracking. The largest amounts of cracking were observed in the LCB sections at the Nevada SPS-2 site, which, construction records indicate, had problems with excessive concrete shrinkage and slab cracking during the construction of some sections. The disparity between the cumulative frequency dis- tribution curves for cracking in LCB sections and cracking on AGG and PATB sections is not, however, a distortion caused by the rather anomalous performance of the pavements at the Nevada site. The LCB sections developed more cracking than did comparable sections with AGG or PATB base at several SPS-2 sites, including Arkansas, Michigan, North Dakota, 83 PCC/AGG PCC/LCB PCC/PATB PCC/HMAC PCC/CAM 0 10 20 30 40 50 60 70 80 90 100 0.00-1.00 -0.50-0.75 -0.25 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50 2.75 3.00 3.25 3.50 3.75 4.00 Faulting (mm) Pe rc en t o f s ec tio ns Figure 108. Cumulative frequency distributions of faulting for SPS-2 pavement sections. Independent Variable Combined r2 with this Variable Added TIME_LAST_CRACK 0.13 TMI 0.26 CESAL_LAST_CRACK 0.29 TMP 0.33 PRECIP 0.37 B1 0.38 FIRST_K 0.40 B2 0.41 B3 0.41 BAR 0.41 Hpcc 0.41 B5 0.42 FIRST_HEQ 0.42 WIDE 0.42 HIGH 0.42 B4 0.42 Table 33. Significance of SPS-2 regression variables to cracking.
84 Ohio, and Washington. The sections that exhibited the most cracking were the sections with lean concrete base and thin (8-in.) concrete slabs (test section designs 0205, 0206, 0217, and 0218; see Table 6). The stiffest base type in the experiment, lean concrete base, may have been good for performance in terms of roughness and faulting, but it had a pronounced detrimental effect on cracking performance, particularly in the thinner concrete slabs in the experiment. Sections with undrained AGG, the weakest base type, also had more cracking than sections with drained PATB. Sections with undrained HMAC and CAM bases had even less cracking than sections with drained PATB. As with the other SPS-2 performance measures dis- cussed previously, while the design of the experiment makes it difficult to rule out a potential effect of drainage on crack- ing development, the above findings suggest that the differ- ences in cracking observed to date are due not to drainage differences but to differences in base stiffness. PCC/AGG PCC/LCB PCC/PATB 0 10 20 30 40 50 60 70 80 90 100 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 Percent area cracked Pe rc en t o f s ec tio ns Figure 109. Cumulative frequency distributions of cracking in SPS-2 pavement sections.
