回采巷道围岩稳定性预测模型及其成熟度

Prediction Model of Surrounding Rock Stability of Roadway and Its Maturity Tolerance

针对我国煤矿井下回采巷道围岩稳定性预测方法不够成熟、准确率较低的问题,引入支持向量机算法,将回采巷道围岩稳定性的评价指标作为输入向量,稳定性等级作为输出向量,利用粒子群算法不断调整惩罚因子c、核参数g与单核系数λ1和λ2,建立了基于PSO-MKSVM的回采巷道围岩稳定性预测模型。定义出衡量回采巷道围岩稳定性预测模型的精度度量P和非均等代价下的代价度量E,并以成熟度度量M作为评价模型性能的准则。结果表明有多组λ1和λ2使得回采巷道围岩稳定性预测模型的成熟度度量达到最大值M=1,性能较稳定。

For the problems of immaturity and low accuracy of prediction methods of surrounding rock of roadway, we introduced support vector machine（SVM） algorithm and took the evaluation indexes of surrounding rock stability of roadway as the input and its stability levels as the output. A prediction model based on PSO-MKSVM was established by applying particle swarm optimization（PSO） algorithm to adjust the values of penalty factor c, kernel parameter g, and single kernel coefficient λ1 and λ2.The precision tolerance P and cost tolerance E under the unequal prices were defined to weigh the prediction results. Then the maturity tolerance M was confirmed as a criterion for assessing model performance. The experimental results showed that multiple sets of λ1 and λ2 values make the maturity tolerance M of prediction model of roadway surrounding rock stability reach the maximum value of 1, and the performance is stable.