Development of genetic-based models for predicting the resilient modulus of cohesive pavement subgrade soils

The accurate determination of resilient modulus (M-r) of pavement subgrade soils is an important factor for the successful design of pavement system. The important soil property M-r is complex in nature as it is dependent on several influential factors, such as soil physical properties, applied stress conditions, and environmental conditions. The aim of this study is to explore the potential of an evolutionary algorithm, i.e., genetic algorithm (GA), and a hybrid intelligent approach combining neural network with GA (ANN-GA), to estimate the M-r of cohesive pavement subgrade soils. To achieve this aim, a reliable database containing the results of repeated load triaxial tests (RLT) and other index properties of subgrade soils was utilized. GA was employed to develop a precise equation for predicting M-r of subgrade soils. In addition, GA was used as a tool for determining the optimal values of the weights and the bias of the ANN-GA approach. The developed ANN-GA model was then transferred to a functional relationship for further application and analyses. Several validation and verification phases were conducted to examine the performance of the developed models. The results indicated that both GA and ANN-GA models could accurately predict the M-r of cohesive subgrade soils, and performed better than other models in the literature. Finally, a sensitivity analysis was conducted to evaluate the effect of the utilized parameters on M-r. (C) 2020 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.