Artificial neural networks as a basis for new generation of rock failure criteria

作者：Rafiai Hosein^*; Jafari Ahmad

来源：International Journal of Rock Mechanics and Mining Sciences, 2011, 48(7): 1153-1159.

DOI：10.1016/j.ijrmms.2011.06.001

• Rock strength
• Artificial neural network
• Bieniawski-Yudhbir criterion
• Hoek-Brown criterion
• Ductile failure

Conventional methods for prediction of rock strength are based on using classical failure criteria. In this study, artificial neural networks were regarded as new tools for considering the strength of intact rock in a wide range of loading condition from uniaxial tension to triaxial compression. A comprehensive data set of the values of major and minor principal stresses at failure from 1638 laboratory tests on seven rock types was collected. For each rock type, data were randomly divided into two subsets, training and test sets. Neural networks were trained using training sets to predict the value of major principal stress at failure from uniaxial compressive stress and minor principal stress. Small architecture and regularization method were adopted to avoid over-fitting problems. The same training sets were used in determining the constants of two popular empirical failure criteria, namely Bieniawski-Yudhbir and Hoek-Brown. Then, the test sets were used to examine the accuracy and generalization of the model sin predicting the strength in new situations. Comparison of the results of the neural network models with those of the empirical criteria showed that the former approach always lead to less root mean squared error and higher coefficient of determination. On average, for different rock types, using ANN models led to about 30% decrease in prediction error relative to best empirical models. These models also showed better flexibility in the prediction of major principal stress at failure in both brittle and ductile failure regimes.

• 出版日期2011-10