A prediction method of stress intensity factor for mode-I crack in coal rock based on deep learning

Abstract

The stress intensity factor (SIF: K) analysis is of utmost importance in the evaluation of the coal-rock mass damage. However, the existing calculation processes seem cumbersome, even failing to meet the remote real-time determination of the SIF. Against this background, this paper introduces a prediction method of the SIF for the mode-I crack in the coal rock based on deep learning (DL). As soon as the model training is completed, the SIF can be acquired only by inputting the coal-rock image. To obtain extensive images of the mode-I crack of coal rock during the three-point bending test, a high-speed camera is used. The digital image correlation (DIC) and a crack-tracking algorithm (CTA) are further utilized to calculate the SIF corresponding to each image. Afterward, the model is trained and validated based on the self-designed convolutional neural network (CNN) architecture. The final results demonstrate that the fitting coefficient between the true value and the predicted one reaches 0.961, which indicates that the trained model is able to accurately predict the SIF.