Settlement Prediction of Shallow Foundations on Granular Soils Using Multi Expression Programming (MEP)

Due to the non-homogeneous nature of cohesionless soils, and the complexity of the parameters related to settlement, generally the exact estimation of the foundation settlement involves many difficulties. This problem has been studied by many researchers and some equations have been proposed for estimating the settlement of shallow foundations. Because of the low accuracy of these equations, researchers are using more modern techniques like finite elements, finite difference, soft computing, and so on, for calculating the settlement. In the current study, a new model is proposed based on artificial intelligence to predict the settlement. The purpose of developing such models is to achieve more accurate and possibly simpler equations. In this paper, the performance of multi expression programming (MEP) method to predict the settlement of shallow foundations on granular soils is evaluated and the values obtained from the developed models are compared with those from the most accurate previous models. Then, the performance analysis, validation, parametric study and sensitivity analysis are performed. As can be seen from the results, the comparison between the MEP model and the most accurate models including ANN, GP, GEP and EPR in the technical literature showed that the model proposed in this study with correlation coefficient of 93.45 percent, performs better than most previously developed models.