EVALUATION OF LABORATORY, CONSTRUCTION ANDPERFORMANCE VARIABILITY BY BOOTSTRAPPING ANDMONTE CARLO METHODS FOR RUTTINGPERFORMANCE PREDICTION

摘要

This paper demonstrates an innovative reliability analysis approach for prediction of asphaltrutting performance. In this approach, reliability can be evaluated by considering thevariability in laboratory test results, layer thicknesses and stiffnesses, and measured in-situperformance. The effects of input design parameters variability on predicted performancewere determined using the calculated distributions of calibration coefficients. To assess thecontribution of each input parameter’s variability to variability of calculated calibrationcoefficients, various cases were created by including and excluding the variability in theseparameters in the calibration process. This permitted comparison of calculated variabilityconsidering different sources of variability, with measured performance variability. Thesedistributions were also used for rutting performance prediction and reliability evaluation ofhighway sections. In this way, rut depths for different reliability levels can be predictedwithout performing computationally intensive calculations within the design software. Theresults indicated that distributions of calibration coefficients calculated by using measured rutdepths (performance variability) are very similar to calibration coefficient distributionscalculated by using thickness and stiffness (construction) variability. This result suggests thatvariability in performance can be effectively predicted by using the variability in thicknessand stiffness for heavy vehicle simulator (HVS) test sections because variability in thicknessand stiffness were found to be the major factors that control the variability in measured rutdepths. The effect of laboratory test results variability on calibration coefficient distributionswas found to be negligible when compared to the effects of stiffness and thicknessvariability. Although the reliability approach proposed in this study was developed using theresults of a specific laboratory test and rutting model, the general procedure can be applied toany pavement design software for any type of distress.