Abstract:

In order to solve the bolt corrosion failure in underground mines, a mathematical model based on predicting the bolt corrosion failure rate was proposed. By collecting the relevant data of mine corrosion failure bolts, the principal component analysis (PCA) method and gradient boosting tree (GTB) method were used to extract useful information from the original data. Combined with the SVM model, the corrosion failure probability of bolts in complex mine environments was predicted, and the model was tested by using datasets with continuous features and spatial data. The impact weight of various environmental factors on corrosion failure was analyzed and compared with SVM and GTB-SVM models. The research results indicate that feature transformation based on PCA is a reliable risk prediction method. The PCA-SVM model performs superior in terms of prediction accuracy and robustness of results, with AUC values reaching 0.84 in the training set and 0.83 in the test set. This model will serve as a useful online tool to support the security and digitization of process systems. The PCA-SVM model proposed in the article can accurately predict the corrosion failure probability of bolts.

Key words: deep mines, bolts, stress corrosion cracking, support vector machines (SVM), feature learning