Swelling Prediction in Compacted Soils Using Adaptive Neuro-Fuzzy Inference System

Abstract

Swelling in compacted soils may lead to some damages to structures and buildings. For the sake of reducing such damages, soil swelling should be determined, so as to make the structures exhibit adequate resistance against such a phenomenon. For most cases, fully non-linear relations have been observed between soil swelling and the parameters contributing to swelling in compacted soil. As such, soil swelling should be determined via either experimentations or prediction models. However, being extremely timely, swelling tests require special expensive equipment. Accordingly, there is a need for models which can use available data to theoretically give swelling estimations of a relatively high accuracy without getting busy with swelling tests and associated issues. Investigated and evaluated in this research are the ability and application of an adaptive neuro-fuzzy interference system (ANFIS) developed by subtractive clustering and fuzzy c-mean clustering to determine and predict swelling in compacted soils. The results along with the obtained values of root mean squared error (RMSE), mean absolute error (MAE) and coefficient of correlation (R) indicated that the proposed ANFIS model succeeded to predict swelling in compacted soils at a good level of accuracy. Therefore, ANFIS models can be used to predict swelling without getting busy with swelling tests and associated issues. KEYWORDS: Swelling of compacted soil, Subtractive clustering, Fuzzy c-mean clustering, ANFIS, Prediction.