基于CNN的矿井外因火灾火源定位方法研究

Study on localization method of mine exogenous fire source based on CNN

为了在矿井发生火灾时能够及时准确定位火灾发生位置,解决当前矿井火灾监测方法中存在的局部监测方法覆盖区域较为有限、整体监测定位系统泛化性能不强等问题,提出1种基于深度卷积神经网络(CNN)的矿井外因火灾火源位置识别模型,阐述该模型的基本框架及数据处理方法,使用Ventfire程序生成火灾模拟数据进行训练及优化。研究结果表明:对案例数据测试中,定位发生火灾巷道的准确率为85.3%,定位火灾发生巷道所在范围(5条巷道)的准确率为92.6%。研究结果可为提高矿井外因火灾识别定位的智能化水平提供参考。

In order to locate the fire location in time and accurately when a fire occurs in the mine,and solve the problems of limited coverage of local monitoring methods and weak generalization performance of the overall monitoring and localization system in current mine fire monitoring methods,a location recognition model of mine exogenous fire source based on deep convolutional neural network(CNN)was proposed.The basic framework and data processing methods of the model were elaborated,and the Ventfire program was used to generate fire simulation data for training and optimization.The results show that in the data testing of the case study,the accuracy of locating the fire prone roadway is 85.3%,and the accuracy of locating the range of the fire prone roadways(5 roadways)is 92.6%.The research results can provide reference for improving the intelligent level of identifying and locating the mine exogenous fires.