Abstract
Rockburst disasters generally occur within specific geological structures. Consequently, predicting and evaluating the potential rockbursts necessitates more focuses on the "geological carrier". Considering the geological structure effects of rockbursts, a novel prediction method based on the key geological structures was proposed by combining numerical simulation with neural network. This study systematically investigated typical geological structures and geomechanical modes of rockbursts. An evaluation index for the relative energy release effect of rockbursts was established through the numerical simulation, which can be adopted to analyze the sensitivity of the key structural parameters. Subsequently, a surrogate model was developed using GA-BP neural network combined with Latin hypercube sampling to accurately represent the relationship between the key structural parameters and rockburst effects. Using the rockburst intensity classification scheme and the Monte Carlo method, the samples within the dangerous range of rockbursts were identified. By analyzing the rockburst sample points of various grades, the confidence intervals for the key structural parameters were interpreted, with the prediction method developed for identifying the key geological structures. This innovative method facilitated the rapid "dictionary-style" prediction of rockbursts in underground engineering. It comprehensively considered the impacts of geological structures on rockbursts and could offer a new pathway for precise rockburst disaster prediction.