基于NRS-ACPSO-SVM的冲击地压危险性预测模型

发布时间：2018-03-29 19:45

本文选题：冲击地压　切入点：危险性预测　出处：《中国安全科学学报》2017年10期

【摘要】：为快速、准确地预测冲击地压危险性,提出基于NRS-ACPSO-SVM的冲击地压危险性预测模型。首先,在综合分析冲击地压危险性影响因素的基础上,以重庆砚石台煤矿为例,选取煤层厚度、倾角、埋深等10个影响因素作为冲击地压危险性的特征指标;然后,基于邻域粗糙集(NRS)理论对特征指标进行降维,提取出影响冲击地压危险性的关键属性构成约简集;最后,为避免支持向量机(SVM)模型受惩罚因子C和核函数参数σ随机性影响,采用自适应混沌粒子群算法(ACPSO)优化SVM模型参数,将约简集作为ACPSO-SVM模型的输入进行训练,利用训练好的ACPSO-SVM模型预测样本,并对比其他模型的预测结果。研究表明:NRS能有效地约简属性,简化模型结构,模型预测精度与运行效率均有明显提高;利用ACPSO优化SVM模型能避免结果陷入局部极值,提高收敛速度及预测精度,用该模型可有效地预测冲击地压危险性等级,其预测错误率为0。
[Abstract]:In order to predict the impact ground pressure risk quickly and accurately, a prediction model of impact ground pressure risk based on NRS-ACPSO-SVM is put forward. Firstly, on the basis of comprehensive analysis of the influencing factors of impact ground pressure risk, taking Yanshitai coal mine in Chongqing as an example, the thickness of coal seam is selected. Ten influencing factors, such as dip angle and depth of burying, are used as characteristic indexes of rock burst hazard. Then, based on the theory of neighborhood rough set (NRS), the key attributes which affect the risk of impact ground pressure are extracted to form the reduction set, which is based on the theory of neighborhood rough set (NRS) to reduce the dimension of the characteristic index. In order to avoid the influence of penalty factor C and kernel function parameter 蟽 randomness on the support vector machine (SVM) model, the adaptive chaotic particle swarm optimization algorithm (ACPSO) is used to optimize the parameters of the SVM model, and the reduction set is trained as the input of the ACPSO-SVM model. The trained ACPSO-SVM model is used to predict the samples, and the prediction results of other models are compared. The results show that the proposed model can effectively reduce the attributes and simplify the model structure. The prediction accuracy and operation efficiency of the model are improved obviously. Using ACPSO to optimize the SVM model can avoid the result falling into local extremum and improve the convergence speed and prediction accuracy. The model can effectively predict the hazard grade of rock burst, and the prediction error rate is 0.
【作者单位】：辽宁工程技术大学系统工程研究所;
【基金】：国家自然科学基金资助(71371091) 辽宁省教育厅基金资助(L14BTJ004)
【分类号】：TD324;TP18

【相似文献】