Exploration and Improvement of Fuzzy Evaluation Model for Rockburst

Rockburst is a highly destructive geological hazard that can cause casualties and equipment damage. To achieve high-accuracy discrimination of rockburst intensity, this article proposes an improved model that addresses the inefficient maximum membership principle used in traditional rockburst fuzzy evaluation models. The stress coefficient σθ/σc, brittleness coefficient σc/σt, and elastic deformation energy index Wet are selected as evaluation indicators for rockburst classification. Subjective and objective weights are obtained using the Delphi method and entropy weight method (EWM). Three types of membership function distribution forms are then used to obtain the membership degrees of each indicator to rockburst grades: trapezoidal membership function (TMF), normal membership function (NMF), and quadratic parabolic membership function (QPMF). Finally, six traditional models and six improved models are established using the maximum membership principle (MMP) and weighted average-maximum membership principle combination evaluation principle (WMP), respectively. Based on the analysis of 100 sets of rockburst field data, the accuracy, precision, recall, and F1-score of the improved evaluation model are increased by 11.3%, 0.097, 0.068, and 0.089, respectively, compared to the traditional model. The Delphi-NMF-WMP model is selected as the best model, with four performance indices reaching 97.0%, 0.979, 0.979, and 0.978. The best model is applied to evaluate the rockburst intensity of the Cangling Tunnel, Dongguashan Copper Mine, and Jiangbian Hydropower Station Diversion Tunnel, with evaluation results consistent with the actual situation, demonstrating the reliability and scientificity of the model.