Quantification of rock heterogeneity and application in predicting rock mechanical properties

摘要

Heterogeneity is an essential feature of rock, while most rock mechanics models only adopt conventional rock physical properties and ignore the influence of heterogeneity. Quantifying rock heterogeneity is of great significance for accurately predicting rock mechanical properties. This paper introduced homogeneous coefficient to quantify rock heterogeneity and designed a model for automatically evaluating rock homogeneous coefficients by combining computer tomography and image recognition technology. In addition, this paper optimized the model parameters by correlation analysis and sensitivity analysis to predict various rock mechanical properties. When using rock homogeneity coefficient as a model parameter, the predicted loss value of rock strength is 2.95, which is 25.7 lower than that of the model without rock homogeneity coefficient. The predicted loss value of elastic modulus is reduced by 3.9, while it is almost unchanged for Poisson’s ratio. Empirical equations are also proposed adopting optimized model parameters, providing reliable predicted results as accurate as artificial neural network model. The Pearson correlation coefficient for rock strength could reach 0.9537. In this paper, the mature technology combination in the computer field is applied to predict rock mechanical properties, which shows great application prospects.