Towards a Reliability-Based Pavement Design using Response Surface Methods

Reliability has been incorporated in many pavement design procedures to account for the effects of inputs variabilities and uncertainties on predicted performance. The American Association of State Highway and Transportation Officials (AASHTO) mechanistic empirical pavement design guide (MEPDG) computes the reliability of pavement sections with the assumption that the variability of predicted distresses follows normal distribution. This approach does not account for the systematic contribution of each design input variability on the overall output variance. This paper evaluates a two-component reliability analysis methodology for pavement application. The two-component reliability analysis methodology uses a response surface method (RSM) for a surrogate model generation and the first order reliability method (FORM) for reliability computation. Three different response surface methods (central composite, Box-Behnken and Doehlert designs) were implemented and statistically verified for their suitability for surrogate model generation. The two-component reliability analysis methodology was further utilized for the generation of partial safety factors for the development of a load and resistance factor design (LRFD) procedure for pavement applications. Field pavement sections with a wide range in design inputs and target reliabilities were used to evaluate the proposed reliability analysis methodology. The results have shown that the three RSM can be used effectively for pavement reliability problems.