Application of artificial intelligence methods for predicting transient response of foundation

The present work focuses on analysing the displacement of a shallow foundation due to sudden change in state of loading. When a dynamic load hits the foundation, before attaining steady state, the foundation undergoes sudden displacement. This displacement is a function of soil properties viz. modulus of elasticity, Poisson's ratio, loading conditions and the static load already applied on the foundation. A Finite Element based numerical model that considers soil structure interaction is simulated to generate a data set. After verification of model results, the dataset is analysed using five different methods including Levenberg Marquardt Neural Network (LMNN), Bayesian Regularization Neural network (BRNN), Support Vector Machines (SVM), Multivariate Adaptive Regression Splines (MARS) and Multi Gene Genetic Programming (MGGP). A comparative analysis of all the methods has been presented to find out the most effective method. In addition to these, sensitivity analysis is performed to find out the most influencing input parameter. The BRNN method is found to be the most efficient method and the static load on the foundation is the most influencing parameter as revealed from the rigorous statistical analysis. The outcomes will be helpful in quick analysis of shallow strip foundations.