Rockburst Prediction Method Based on K-Nearest Neighbor Pattern Recognition

摘要

Rockburst is a geological disaster induced by mining at great depth. How to predict rockburst effectively for safety during mining has become an unresolved key problem. Because of poor understanding of the mechanism and influence factors of rockbust, it is very difficult to give accurate prediction using conventional methods. A new method based on k-Nearest Neighbor pattern recognition tech, which is one of the simplest and most effective tools in the field of pattern recognition, is proposed. First, the historical instances with influence factors induced rockbust are collected into database. Then, k historical instances whose influence factors similar to that of new instance are selected through scanning the database based on the neighbor similarity function. Finally, roburst risk of the new instance can be recognized by majority vote among the k nearest historical instances. The method gives accurate rockburst predictions under novel conditions when mining at great depth. The results of case studies at deep gold mines in South African show that this method is scientific, feasible, and promising.