ANN and Neuro-Fuzzy Modeling for Shear Strength Characterization of Soils

Abstract

We examine the outcome of popular artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS) for estimating the shear strength parameters of c − φ soil. A matrix of one hundred twelve datasets collected using in situ and laboratory tests to train and test the ANN and ANFIS models. Standard penetration test number of blows value along with the soil properties taken as input vectors, whereas shear strength parameters like cohesion (c) and angle of internal friction (ϕ) used as target vectors. The minimum validation error has been employed as the stopping criterion to avoid over fitting in the analysis. Out of four developed models, predicted values through two ANN models were close to actual value in comparison to ANFIS models. Statistical parameters such as coefficient of correlation, root mean square error and average absolute error were used as performance evaluation measures. Based on statistical measures it was observed that performances of ANN and ANFIS models were in accordance with the experimental results and it could substitute tedious laboratory work provided sufficient and reliable data source are offered. The results through performance evaluation measures also reveal that ANN and ANFIS models are effective, versatile and useful way to measure the shear strength parameters of soils.