Estimation of the Mode II Fracture Toughness by Artificial Neural Network

There have been some efforts to investigate shear mode of fracture using rock specimens by different methods in the recent years. Most of the fractures occurring in nature are of the shear, tensile or the combination of both. The determination of Mode II fracture toughness is important for understanding and analyses of the failure process in rock masses. The comprehensive investigations in the methods of determination of Mode II fracture toughness of rocks indicated that punch and shear box tests are reasonable, although there may be still some illusions. Artificial neural networks are capable to establish logical relationship between different variable parameters in order to predict the value of Mode II fracture toughness. In an effort, the test results obtained from punch and shear box tests along with some affective parameters are considered to predict Mode II fracture toughness through neural network models. Thus an optimum network was established and the precision was evaluated. A comparison of the neural network results is made with the linear regression. The details of the investigation are presented in this paper.