Neural and neurofuzzy techniques applied to modelling settlement of shallow foundations on granular soils

Abstract

This paper describes two modelling techniques applied to a case study of settlement prediction of shallow foundations on granular soils. The first technique uses multi-layer perceptrons (MLPs) that are trained with the back-propagation algorithm, whereas the second technique uses B-spline neurofuzzy networks that are trained with the adaptive spline modelling of observation data (ASMOD) algorithm. The performance of the models obtained using both techniques is assessed in terms of prediction accuracy, model parsimony and model transparency. The results indicate that both the back-propagation MLP and the Bspline neurofuzzy models are comparable in terms of prediction accuracy, although the back-propagation MLP model is found to perform slightly better than the B-spline neurofuzzy model. In terms of model parsimony, the B-spline neurofuzzy model is found to be more parsimonious than the back-propagation MLP model. In terms of model transparency, the B-spline neurofuzzy model is found to provide a more explicit interpretation of the relationships between the model inputs and the corresponding outputs.