ANFIS Models with Subtractive Clustering and Fuzzy C-Mean Clustering Techniques for Predicting Swelling Percentage of Expansive Soils

Abstract

Civil engineering faces significant challenges from expansive soils, which can lead to struc-tural damage. This study aims to optimize subtractive clustering and Fuzzy C-Mean Cluster-ing (FCM) models for the most accurate prediction of swelling percentage in expansive soils. Two ANFIS models were developed, namely the FIS1S model using subtractive clustering and the FIS2S model utilizing the Fuzzy C-Mean Clustering (FCM) algorithm. Due to the MATLAB graphical user interface's limitation on the number of membership functions, the coding approach was employed to develop the ANFIS models for optimal prediction accuracy and problem-solving time. So, two programs were created to determine the optimal in-fluence radius for the FIS1S model and the number of membership functions for the FIS2S model to achieve the highest prediction accuracy. It is feasible to make relatively accurate predictions about swelling percentage by utilizing complex models that analyze available data and theoretically predict swelling percentages. This can be accomplished without the need for swelling tests and laboratory-related issues.

Keywords

• expansive soils
• swelling
• subtractive clustering
• fuzzy c-mean clustering (fcm)
• sensitivity analysis