Dynamic early warning of rockburst using microseismic multi-parameters based on Bayesian network

Abstract

Rockburst is a phenomenon where the elastic deformation may be emitted suddenly because of rock fragmentation, ejection, launching, and even earthquake. This may result in casualties, failure or deformation of the support structure, and damage to field equipment. Therefore, early warning of rockburst is significant. In this paper, a dynamic early warning model of rockburst using microseismic multi-parameters based on Bayesian network is proposed. Taking the moment magnitude, seismic energy, source radius, apparent stress, and dynamic stress drop as Bayesian network parameters input. 114 sets of parameters required for Bayesian network structure learning are obtained by pre-processing the historical data of rockburst, and belief update is performed by the Junction Tree algorithm. Besides, the model passes self-validation, 6-fold cross-validation, ROC curve analysis and using new historical data to real-time early warning analysis. The results indicate that the proposed model has good precision and can effectively realize early warning of rockbursts. Furthermore, through the strength of the influence of parent–child nodes and sensitivity analysis, it can find that moment magnitude and seismic energy are the most influential parameters. They can as a significant reference for the early warning of rockburst.

Keywords

• Microseismic monitoring
• Rockburst
• Early warning
• Multi-parameter
• Bayesian network