Prediction of the Undrained Shear Strength of the Remolded Soil with Non-Linear Regression, Fuzzy Logic and Artificial Neural Network

Abstract

The aim of this study is to predict the undrained shear strength (Cu) of the remolded soil samples and for this purpose, non-linear regression (NLR) analyses, fuzzy logic (FL) and artificial neural network (ANN) modelling were used to assess. Total 1306 undrained shear strength results of soil types of CH, CL, MH and ML from 230 different remolded soil test settings on 21 publications were collected while six different measurement devices were used by researchers. Although water content, plastic limit and liquid limit were used as input parameters for FL and ANN modelling, liquidity index or water content ratio were considered as input parameter for NLR analyses. In NLR analyses, 12 different regression equations were derived for prediction of Cu. Feed-Forward backpropagation and TANSIG transfer function were used for ANN modelling while Mamdani inference system was preferred with trapezoidal and triangular membership function for FL modelling. The experimental results of 914 tests for training of the ANN models, 196 for validation and 196 for testing were used. It was observed that the accuracy of the ANN and FL modellings were higher than NRL analyses. Furthermore, the simple and reliable regression equation was proposed for assessments of Cu values having higher coefficient of determination values (R2).