A comprehensive review of intelligent machine learning based predicting methods in long-term and short-term rock burst prediction

Abstract

Rockburst is a geological hazard frequently encountered in deep underground engineering projects that threaten workers’ safety and causes damage to an excavation. The occurrence of rockburst has motivated researchers to intensely investigate different methods to predict its severity. Consequently, scholars applied Machine Learning (ML) methods in rockburst prediction to address the complex, nonlinear relationship between rockburst and its impacting constituents intelligently. Although some past reviews attempted to provide an overview of ML methods in rockburst prediction, there has been no systematic study that details the significance of ML over other methods, insights related to ML types and model description, and a detailed comparative study between different ML methods in terms of performance, technical information and merits and demerits of each method in existing research, etc. Hence, aiming to provide a comprehensive and resourceful review of long-term and short-term rockburst prediction, this work initially defines rockburst, highlights the limitations of several previous prediction approaches, and explores the significance of ML over them. Secondly, a brief description of each predicting model is provided. After, achievements and advancements in the existing field over the past two decades are surveyed and categorised with technical information and rockburst databases are established. Furthermore, the merits, demerits and performances of all methods in the current application are discussed and suggestions to handle rockburst data are presented. Finally, future work remaining in this area of research is identified and overall conclusion is drawn.