Prediction of compressive strains at the top of subgrade based on BP neural network

Based on layered elastic theory, the compressive strains at the top of subgrade were predicted using BP neural network. According to the types of pavement structure in common use, the database of surface deflections with their corresponding structural parameters based on layered elastic theory was established. BP neural network was developed using the established database and was used to predict the compressive strains at the top of subgrade. The predictive effect of compressive strains at the top of subgrade backcalculated by theoretical deflection basins was tested. At the same time, generalization ability of the developed BP neural network was verified. It indicated that error of the compressive strains at the top of subgrade predicted by the developed BP neural network and the theoretical values calculated by layered elastic theory program was within 6%. It would provide the references with the model of BP neural network to estimate the health conditions of subgrade.