Hybrid Formulation of Resilient Modulus for Cohesive Subgrade Soils Utilizing CPT Test Parameters

In the present study, a novel model is introduced for the prediction of a resilient modulus (M-R) of cohesive subgrade soils considering cone-penetration test parameters to establish correlations with the M-R. A reliable previously published database composed of 124 datasets was utilized for the development of the proposed model, which incorporates both cone penetration test (CPT) parameters and laboratory indices. In order to generate the predictive model, a hybrid algorithm combining a firefly algorithm with a multilayer perceptron neural network (FA-MLP) is proposed. The FA algorithm is employed in the MLP network structure to adjust the weights and the bias of the network and, hence, improve the overall performance of the network. The proposed FA-MLP formulation was found to have the capacity to predict, satisfactorily, the M-R of cohesive subgrade soils using the results of the CPT test.