LAUSR

Soil collapse is defined as a considerable reduction in soil volume when inundated under constantly applied pressure is known to be responsible for the failure of geotechnical structures such as highway/railway embankments and earth dams. Gene expression programming (GEP) is used as an artificial intelligence (AI) for the formulations of collapse potential in terms of the difference between the sand and clay percentages or the coefficient of uniformity, initial water content, initial dry unit weight, and wetting pressure in this paper. The experimental data available in the literature have been gathered for predicting collapse potential with the empirical formulations developed in the training and test sets of GEP-based models. Besides, additional experimental data derived from different literature are obtained to confirm the applicability and generalizability of the developed GEP-based formulations. The prediction performances of GEP models are compared to the experimental results and regression-based formulations proposed in the literature. These comparisons and statistical values obtained from analyses show that the GEP-based models are detected to be more effective methods to estimate the collapse potential. Moreover, a series of parametric analysis is conducted to perceive influences of input parameters on collapse potential by using GEP-based formulations.