基于洞壁实测信息的FA-PP岩爆预测模型应用研究

为了给岩爆隧道提供更加符合现场实际规律的岩爆预测结果，针对岩爆预测评价中应力判据界限值的多样性以及单一判据预测结果准确性低的特点，提出采用投影寻踪（PP）与萤火虫算法相结合的方法进行岩爆预测评价，解决了岩爆预测中采用多种应力判据时多项指标的模糊性和不兼容性等问题。此方法根据岩爆等级划分标准，构建了以实测数据、现场岩爆信息为基础的能够根据各项指标判定岩爆等级的投影指标函数，选取σ_θ/R_c，σ_t，σ₁三个岩爆评价中常用的指标，并采用萤火虫算法（FA）优化投影指标函数，寻求投影指标函数的最佳投影方向并求得最佳投影值，保证了此方法的精确性。研究结果表明：采用萤火虫投影寻踪算法（FA-PP）进行岩爆等级预测可避免传统预测方法中由于主观原因造成的误差，判定精度高，也免去了利用多项判定指标与现场实际情况进行对比而选择一个相对接近实际情况判据的繁杂过程；此方法可利用目标隧道的已知岩爆发生信息掌握其岩爆发生特点及对3个评价指标的敏感度，构建出符合目标隧道岩爆发生规律的FA-PP岩爆预测模型；以拉林铁路桑珠岭隧道为研究对象，将此方法应用到岩爆预测中，得到的预测结果与现场实测结果较为符合，验证了该方法的可靠性。

Application Research of FA-PP Rockburst Prediction Modelfor Tunnel Walls

According to the diversity of the stress criterion threshold in rockburst prediction evaluation and low accuracy of the single criterion, it is difficult to obtain accurate rockburst predictions. To obtain rockburst prediction results in line with the actual rules for rockburst tunnels, the method of combining projection pursuit and firefly algorithm was proposed. This method addresses the problems of ambiguity and incompatibility of multiple indicators when multiple stress criteria are applied in rockburst prediction. Furthermore, this method divided the criterion according to the rockburst level, and constructed a projection index function based on the measured data and on-site rockburst information to determine the rockburst level according to three indicators. These indicators, σ_θ/R_c,σ_t and σ₁, used the firefly algorithm to optimize the projection index function to find the best projection direction and projection value to ensure the accuracy of this method. The result shows that predicting the rockburst level using the firefly algorithm-projection pursuit (FA-PP) avoids the error caused by subjective reasons in the traditional prediction method, thereby increasing the determination accuracy. On the other hand, it eliminates the complicated process of selecting a criterion that is relatively close to the actual situation while using multiple judgment indicators to compare with the actual situation on site. This method can use the known rockburst occurrence information of a target tunnel to grasp its rockburst occurrence characteristics and sensitivity of the evaluation indexes to construct a FA-PP rockburst prediction model that meets the rockburst occurrence law of the target tunnel. The Sangzhuling tunnel of Lasa-Linzhi railway was selected as the research object, and this method was applied to the rockburst prediction. The obtained prediction results are in good conformity with the field test results, which verifies the reliability of this method.