Improvements to strain computation and reliability analysis of flexible pavements in the mechanistic-empirical pavement design guide

The first part of this study evaluates the effect of load-strain linear proportionality assumption employed in the Mechanistic-Empirical Pavement Design Guide (MEPDG) design of flexible pavements. In the design procedure, the strains computed through Jacob Uzan Layered Elastic Analysis (JULEA) are used to accumulate the damage and distresses over the design period. To minimize computing time, the MEPDG makes the assumption that the computed strains are linearly proportional to the applied load and exploits this assumption to extrapolate strains from an 18 kip single axle load and the specified tire pressure to the entire load spectrum in the traffic composition. However, in reality, for truck loads of interest, the contact (tire inflation) pressure remains within a narrow range whereas the contact area changes with axle load. The study showed that this assumption can lead to significant error in the prediction of Hot Mix Asphalt (HMA) rutting. A method that computes strains at three reference axle loads, and extrapolates or interpolates the strain values for other axle loads is shown to provide a suitable alternative to the MEPDG assumption.The second phase of the study is on the improvements to the reliability analysis currently used by the MEPDG. This procedure relies on the variability of the measured output (distress) obtained from a data base. This does not directly account for the variability of the input parameters that induce such variability in distress in reliability predictions. This study developed a reliability procedure that directly evaluates the effect of uncertainties in the model input parameters of HMA materials on the reliability of flexible pavements using Monte Carlo and Latin Hypercube simulation and Rosenblueth's 2K+1 point estimate method. The proposed reliability procedure also uses some techniques to reduce the extensive computational time involved in simulations. The Latin Hypercube simulation method is found to be an efficient alternative to the computationally intensive Monte Carlo simulation. Rosenblueth's 2K+1 point estimate method is not capable to capture the output distribution type. Robust sensitivity analyses through Tornado plots and extreme tail analyses are used to identify the relative importance of the input variables on the predicted distress.