Prediction of p-wave velocity and anisotropic property of rock using artificial neural network technique

Abstract

Physico-mechanical properties of rocks are significant in all operational parts in mining activities, starting from exploration to final dispatch of material. Compressional wave velocity (p-wave velocity) and anisotropic behaviour of the rocks are two such properties for understanding the rock response behaviour under stress conditions. They also influence the breakage mechanism of rock. There we some known methods to determine the p-wave velocity an anisotropy in in situ as well as in the laboratory. These methods we cumbersome and time consuming. In the present investigation, artificial neural network (ANN) technique is used for the prediction of p-wave velocity and anisotropy, taking chemical composition and other physico-mechanical properties of rocks as input parameters. Cross-validation technique termed as leaving-one-out is used, as the numbers of data sets are limited in number. Network with six input neurons, one hidden layer with five neurons, and two output neurons is designed for sandstone. Similar network for marble with four hidden neurons is also designed. To deal with the problem of overfitting of data, Bayesian regulation is used and network is trained with 1500 training epochs. The coefficients of correlation among the predicted and observed values are high and encouraging and mean absolute percentage error (MAPE) obtained are very low.