Shear strength prediction of treated soft clay with sugarcane bagasse ash using artificial intelligence methods

Abstract

Soil shear strength is an essential engineering characteristic used in designing and evaluating geotechnical structures. In this study, we intend to analyse and compare the performance of the Genetic Algorithm - Adaptive Network-based Fuzzy Inference System (GANFIS) and Artificial Neural Networks (ANN) in predicting the strength of soft clay. Case studies of 144 soft clay soil samples from Sarang Buaya, Semerah, Malaysia, were utilised to generate training and testing datasets for developing and validating models. RMSE and R have been employed to validate and compare the models. The GANFIS has the highest prediction capability (RMSE=0.042 and R=0.850), while the ANN has the lowest (RMSE=0.065 and R=0.49). From a comparison of the two models, it can be stated that GANFIS is the most promising technique for predicting the strength of soft clay.