基于正态隶属度—属性区间识别模型的岩爆倾向等级预测

Propensity Grade Prediction of Rockburst Based on Normal Membership-Attribute Interval Recognition Model

岩爆是地下工程开挖面临的关键问题之一,为了准确预测深埋隧洞中岩爆烈度倾向等级,提出了正态隶属度—属性区间识别模型的岩爆预测方法。针对岩爆倾向等级属于典型的多属性有序分割类问题,构建了属性区间识别模型,并将岩爆倾向等级划分为4个等级进行预测。根据岩爆发生的成因和机理,选取应力系数、脆性系数、弹性应变指数和岩石完整性系数作为预测指标,考虑各指标之间、指标与标准等级之间的交互关系,采用正态隶属度函数和Jousselme距离计算评价指标权重。结合13个深埋隧洞工程对该预测模型进行准确性测试,并以双江口水电站SPD9厂房为例进行工程实例验证,该模型预测结果与实际相吻合,证明该模型用于具体工程实践中是可行且有效的,研究结果可为类似深埋隧洞岩爆倾向等级预测提供新的思路。

Rockburst is one of the key problems in underground engineering excavation,in order to accurately predict the grade of rockburst propensity in deep-buried tunnels,a rockburst prediction method based on the normal membership degree-attribute interval recognition model was proposed.Due to the rockburst propensity is a typical multi-attribute orderly segmentation problem,the attribute interval recognition model was constructed to divide the rockburst propensity into four grades for prediction. The occurrence of rockburst is affected by engineering geology,and the geological parameter is usually an interval value.The attribute interval recognition model can be better applied to the problem that each grade index value is an interval value. According to the mechanism of rockburst,the stress coefficient,brittleness coefficient,elastic strain index and rock integrity coefficient were selected as predictive indicators,considering the interaction between the indicators and the standard grade,the normal membership function and Jousselme distance were used to calculate the weight of the index. The method had different weights for the measured values of different indexes under the same index system,which can avoid the drawback that the traditional method didn’t consider the interaction relationship between the indexes,and improve the robustness of the model.The accuracy of the prediction model was tested with 13 deep-buried tunnel projects. Since the value of the averaging coefficient had a greater impact on the prediction accuracy of the model,in order to obtain the optimal value of the averaging coefficient,the step size was 0.1 and selected within the interval[0.05,0.95].The analysis shows that when the averaging coefficient is0.65,the prediction accuracy of the model is the highest,which is 92.31%. The left bank of SPD9 in the Huangjiangkou hydropower project was used to verify the model.The prediction results are consistent with the actual rockburst propensity grade.The prediction results prove that the model is feasible and effective in specific engineering practice,and it can provide new ideas for predicting the rockburst propensity grade in similar deepburied tunnels.