ESTIMATING FRACTURE TOUGHNESS OF ROCK (KIC) USING ARTIFICIAL NEURAL NETWORKS (ANNS) AND LINEAR MULTIVARIABLE REGRESSION (LMR) MODELS

Mode I fracture toughness or critical stress intensity factor (KIC ) is one of the most important features of rocks to be considered in designing engineering structures in rock such as underground excavations, blasting in rock and slope stability. In this research work, fracture models based on both Artificial Neural Network and Linear Multivariable Regression were applied to estimate mode I fracture toughness,KIC . In these models the standard data of International Society of Rock Mechanics (ISRM), Chevron Bend (CB) was used. By using the neural network and its proper training to create logical relationship between input and output variables, an optimal model for each series were obtained. comparing the results showed that feed- forward multi-layer perceptron Back- Propagation artificial Neural Network (BPNN) with Levenberg-Marquardt training algorithm and with 2 layers which have 5 and 3 neurons in the first and second layer respectively with R=0.99, RMS=0.084 and SSE=0.01419 is the most appropriate neural network and linear multivariable regression with R=0.97, RMS=0.23 and SSE=1.178 isproper for estimation of fracture toughness of rocks.

Keywords:

Artificial Neural Networks (ANNs) , Linear Multivariable Regression (LMR) , Fracture Mechanics , Mode I Fracture Toughness(KIC)