Rock burst risk decision-making utilizing intuitionistic fuzzy Petri nets enhanced by graph attention mechanism

Rock burst prevention has always been a major concern in coal mine safety. Although early-warning technologies have made considerable progress, early-warning information is still underutilized, and on-site decision measures are often monotonous and lack specificity. This paper presents a rock burst risk decision-making method based on an intuitionistic fuzzy Petri net enhanced by a graph attention mechanism. First, authoritative documents, including rock burst accident investigation reports as well as relevant standards and literature, were collected as the basis for constructing the decision-making index system and training the decision-making model. Then, a graph attention mechanism was introduced to develop an intuitionistic fuzzy Petri nets decision-making model driven by early warning information, influence factors, and decision measures, and the model was trained to obtain the weights and credibility of the optimized decision-making information nodes. Finally, early warning information containing the working face advance, early warning duration and warning level was used for reasoning to infer possible influence factors and recommend decision measures, including drilling, blasting, and hydrofracturing, along with their confidence and non-confidence levels. Taking five dominant factors such as one square, igneous rock intrusion and fault structure during the mining period of a coal mine as examples, the feasibility and effectiveness of the proposed method were verified, providing reliable guidance for the prevention and control of rock burst disasters.