基于随机森林的阻变型通风网络故障诊断方法*

为了在矿井通风网络发生阻变型故障时，能够快速准确判断出故障位置和故障量，提出1种基于随机森林的通风网络故障位置和故障量诊断方法。利用矿井通风仿真系统IMVS将唐安矿模拟故障生成空间数据集并进行数据预处理，构建基于随机森林的故障诊断模型，并利用该诊断模型对唐安矿矿井通风网络模拟故障位置和故障量进行判断和预测。引用多种方法对模型进行度量，通过唐安矿模拟实验验证基于随机森林的故障诊断模型的有效性。将随机森林和决策树的故障诊断准确率进行对比，研究结果表明:随机森林较决策树故障准确率有进一步的提高，并发现故障地点失误诊断多是相邻巷道，在一定程度上工作人员对故障地点的判断并不受其影响。

Resistance variant fault diagnosis method of ventilation network based on random forest

In order to quickly and accurately determine the fault location and fault quantity when the resistance variant fault occurs in the mine ventilation network,a diagnosis method for the fault location and fault quantity of ventilation network based on the random forest was proposed.The mine ventilation simulation system IMVS was used to generate the spatial data set of simulated faults in Tang’an Mine and conduct the data preprocessing.A fault diagnosis model based on the random forest was constructed,and this diagnosis model was used to judge and predict the simulated fault location and fault quantity of mine ventilation network in Tang’an mine.Multiple methods were used to measure the model,and the effectiveness of the fault diagnosis model based on the random forest was verified by the simulation experiments in Tang’an mine.The comparison of fault diagnosis accuracy between random forest and decision tree showed that the fault diagnosis accuracy of random forest was further improved than that of decision tree.Through the research,it was found that the error diagnosis of fault location were mostly the adjacent roadway,and the judgment of fault location by the staff was not affected to a certain extent.