Ultimate bearing capacity prediction of shallow foundations on cohesionless soils using neurofuzzy models

Abstract

This study explores the potential of neurofuzzy computing paradigm to predict the ultimate bearing capacity of shallow foundations on cohesionless soils. The neurofuzzy models combine the transparent, linguistic representation of a fuzzy system with the learning ability of Artificial Neural Networks (ANNs). The data from 97 load tests on footings (with sizes corresponding to those of real footings and smaller sized model footings) were used to calibrate and test the model. Performance of neurofuzzy model was comprehensively evaluated with that of independent fuzzy and ANN models developed using the same data. The values of the performance evaluation measures such as coefficient of correlation, root mean square error, coefficient of efficiency, mean bias error, relative error and mean absolute relative error obtained through the neurofuzzy model are found to be good, which reveals that the neurofuzzy model can be effectively used for the bearing capacity prediction. The values of performance measures obtained for ANN and fuzzy models indicate that the neurofuzzy model significantly outperforms both fuzzy and ANN models. The predicted bearing capacity values obtained through the developed neurofuzzy, ANN and fuzzy models are compared with the values predicted by most commonly used bearing capacity theories. The results indicate that all the three models (i.e., neurofuzzy, ANN, fuzzy) perform better than the theoretical methods.