基于支持向量机的矿井巷道摩擦阻力系数预测研究

发布时间：2018-02-15 01:31

本文关键词： 巷道摩擦阻力系数支持向量机径向基核函数 BP神经网络　出处：《西安科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：矿井巷道通风摩擦阻力系数是反映矿井井巷阻力特性最重要的技术参数之一,它是矿井通风设计、通风系统改造以及加强通风技术管理等工作的技术基础。通常巷道摩擦阻力系数主要通过查表选取和现场实测得到,一方面,我国使用的矿井巷道摩擦阻力系数表为上世纪八十年代所制定,已不能满足目前复杂矿井通风所用;另一方面,现场实测工作较为复杂且工作量大。因此矿井巷道摩擦阻力系数的预测研究对矿井通风设计和煤矿通风安全管理有着重要的理论意义和实际应用价值。本文引入支持向量机对矿井巷道摩擦阻力系数进行预测研究。理论分析得到,采用支持向量机对矿井巷道摩擦阻力系数进行预测是可行的。通过分析对比支持向量机常用核函数的性质,基于径向基核函数相对于其他几种核函数的优点,最终选取径向基核函数作为支持向量机预测矿井巷道摩擦阻力系数模型的核函数。在MATLAB环境下使用LIBSVM,以凉水井煤矿锚喷巷道、锚杆支护巷道、锚网索支护巷道的摩擦阻力系数为研究对象,分别用网格搜索方法、遗传算法方法、粒子群算法方法对训练集进行训练,得到最优的径向基核函数参数和惩罚参数组合,利用选取的参数组合建立预测模型,对验证集进行预测。预测结果表明,这三种方法选取的参数组合建立的预测模型均可有效预测矿井巷道摩擦阻力系数,验证了支持向量机预测矿井巷道摩擦阻力系数的可行性。其中,遗传算法选取参数组合的模型预测结果综合性能表现最好,并和矿井实测结果进行了验证对比。利用支持向量机和常用的矿井巷道摩擦阻力系数预测方法—BP神经网络分别对凉水井煤矿巷道摩擦阻力系数进行预测,结果表明:支持向量机预测矿井巷道摩擦阻力系数的精度更高、效果更好,较BP神经网络预测矿井巷道摩擦阻力系数更具优越性。
[Abstract]:The friction resistance coefficient of mine roadway ventilation is one of the most important technical parameters to reflect the resistance characteristics of mine roadway, and it is the mine ventilation design. The technical foundation of ventilation system reform and strengthening ventilation technical management. Usually, the friction resistance coefficient of roadway is mainly obtained by selecting the table and measuring on the spot. On the one hand, The table of friction resistance coefficient of mine roadway used in our country was drawn up in -20s, which can not meet the needs of ventilation in complex mines at present; on the other hand, Therefore, the prediction of friction resistance coefficient of mine roadway has important theoretical significance and practical application value for mine ventilation design and mine ventilation safety management. The prediction of friction resistance coefficient of mine roadway by support vector machine (SVM) is studied. It is feasible to predict the friction resistance coefficient of mine roadway by using support vector machine. By analyzing and comparing the properties of kernel function commonly used in support vector machine, based on the advantages of radial basis kernel function compared with other kernel functions, Finally, the radial basis function is selected as the kernel function of the support vector machine to predict the friction coefficient of mine roadway. Under the environment of MATLAB, LIBSVM is used to anchor shotcreting roadway and bolt supporting roadway in Liangshui coal mine. The friction resistance coefficient of bolt-mesh-cable support roadway is studied. The training set is trained by grid search method, genetic algorithm method and particle swarm optimization method, and the optimal combination of radial basis function parameters and penalty parameters is obtained. The prediction model is established by using the selected parameter combination to predict the verification set. The prediction results show that the prediction model established by the combination of the three methods can effectively predict the friction resistance coefficient of mine roadway. The feasibility of predicting friction resistance coefficient of mine roadway by support vector machine is verified. By using support vector machine (SVM) and the commonly used prediction method of roadway friction coefficient-BP neural network, the friction resistance coefficient of coal mine roadway in Liangshui coal mine is predicted respectively. The results show that the accuracy and effect of predicting the friction resistance coefficient of mine roadway by support vector machine are higher than that of BP neural network.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD72

【参考文献】