This paper presents the results from multivariate regression analyses on rut data from the SPS-1 experiment in the Long Term Pavement Performance (LTPP) program to develop models for predicting permanent deformation parameters ?? and Ôê× for a three-layer pavement system (asphalt concrete, base and subgrade). All available material, structural and climatic data used to explain rutting were extracted from the LTPP database. Using simple linear regression, ?? and Ôê×-were regressed against these independent variables. The variables that have reasonable relationships (relatively higher R2) were introduced into the multiple linear regression models. The backward regression analysis was used to select the statistically significant variables for the final models. The variables selected for AC rutting included the strain at the middle of the AC layer, % passing sieve number 10 and % voids filled with asphalt of the most upper AC layer, the average of daily maximum air temperatures for the year, and the freezing index. A total of 15 out of 109 sections were used for predicting ?? AC and Ôê× AC. This is due to the limited amount of available data to calculate VTM, VFA, and VMA, which are important for explaining the rate of the AC rutting. The variables selected for base rutting included the backcalculated base modulus, thickness of the base layer, % passing sieve number 200, a newly developed weighted average gradation index, and the strain at the middle of the base layer. A total of 27 out of 109 sections were used for predicting ?? base and Ôê×-base. The variables selected for subgrade rutting included the strain at the middle of the first 40 inches of subgrade, a weighted gradation index and the plasticity index of the subgrade, the number of days above 32.2 oC, the number of days with more than 0.25 mm precipitation, and the backcalculated subgrade modulus. A total of 17 out of 109 sections were used for predicting ?? subgrade and Ôê× subgrade. In general, the ??- prediction models for all layers are more precise than those for Ôê×. This could be due to the fact that the ?? and Ôê×-values were backcalculated from time-series data, which show the rate of growth in rut depth over time. Also it should be noted that Ôê×-values for the AC and base layers were significantly affected (positively) by their corresponding ??-values. This implies that pavements with lower Ôê×-values (lower initial rutting) will show lower ??-values (higher rut growth with time) and vice-versa.
Samenvatting