A comparative study to estimate the mode I fracture toughness of rocks using several soft computing techniques

Abstract

Fracture toughness is an important phenomenon to reveal the actual strength of fractured rock materials. It is, therefore, crucial to use the fracture toughness models principally for simulating the performance of fractured rock medium. In this study, the mode-I fracture toughness (KIC) was investigated using several soft computing techniques. For this purpose, an extensive literature survey was carried out to obtain a comprehensive database that includes simple and widely used mechanical rock parameters such as uniaxial compressive strength (UCS) and Brazilian tensile strength (BTS). Several soft computing techniques such as artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), gene expression programming (GEP), and multivariate adaptive regression spline (MARS) were attempted to reveal the availability of these methods to estimate the KIC. Among these techniques, it was determined that ANN presents the best prediction capability. The correlation of determination value (R2) for the proposed ANN model is 0.90, showing its relative success. In this manner, the present study can be declared a case study, indicating the applicability of several soft computing techniques for the evaluation of KIC. However, the number of samples for different rock types should be increased to improve the established predictive models in future studies.