Reliability-Based Design Optimization of Flexible Pavements Using Kriging Models

Abstract

In the design of flexible pavements, the target reliability levels can be achieved through a number of structural design alternatives by varying the design parameter (i.e., thickness and resilient moduli) combinations. As each alternative comes with a different initial cost, a quantile-based reliability-based design optimization (RBDO) strategy is presented to design multilayer flexible pavements sections with the lowest initial investment requirements. The reliability constraints are satisfied by casting the performance functions of fatigue and rutting as probabilistic design constraints in the RBDO formulation. As the performance functions require computationally expensive models, which can be quite restrictive in the RBDO setting, Kriging metamodels are constructed through an adaptive refinement technique to predict the pavement responses under loading. The quantile-based RBDO problem has been solved to find the layer thicknesses, given the layer resilient moduli, which satisfy prespecified target reliability levels, and is seen to outperform the conventional design alternatives. The optimal solutions indicate that, for pavement sections that are designed with thin asphalt layers and thicker granular base/subbase layers, the same (target) levels of reliability can be achieved by increasing the asphalt layer thickness, particularly when the pavement failure is governed by fatigue. The increased asphalt thickness is offset by substantial reductions in the base layer thickness, resulting in (overall) thinner sections, as well as lower construction costs. These observations are validated for different cost models, taking into account the variations in cost across regions, subject to the availability of resources. The quantile-based RBDO strategy can be extended to any multilayer pavement structure with multiple failure modes without compromising pavement safety during its design life.