Prediction of Subgrade Resilient Modulus Using Genetic Algorithm and Curve-Shifting Methodology

This paper demonstrates the applicability of the genetic algorithm and curve-shifting methodology to the estimation of the resilient modulus at various stress states for subgrade soils by using the results of triaxial resilient modulus tests. This innovative methodology is proposed as an alternative to conventional nonlinear constitutive relationships. With the genetic algorithm, laboratory curves for different deviator stress levels at different confining pressures are horizontally shifted to form a final gamma distribution curve that can represent the stress–strain behavior of subgrade soils with the corresponding predicted shift factors. Resilient modulus values for a given stress state can be estimated on the basis of this curve and another gamma function that represents the variation of the shift values for different confining stresses. To compare the effectiveness of these two approaches, coefficients for the Uzan constitutive model were also determined for each laboratory test and compared with those determined by the approach described in this paper. Predicted resilient modulus values from each approach are separately compared with artificial neural network (ANN) model predictions to evaluate their efficiency and reliability for resilient response prediction. The results of the analysis indicated that the curve-shifting methodology gave superior estimates and a coefficient of determination 14% higher than the Uzan model predictions when the results were evaluated with the ANN model outputs. Thus, although it is not a constitutive model, use of the genetic algorithm and curve-shifting methodology is proposed as a promising technique for the evaluation of the stress–strain dependency of subgrade soils.