基于对称点模式GoogLeNet红外光谱的矿井有害气体监测技术研究  

Research on Monitoring Technology of Harmful Gas in Mine Based on Symmetrical Point Mode GoogLeNet Infrared Spectrum

作　　者：任旭 盛敏敏 刘雅君 REN Xu;SHENG Minmin;LIU Yajun(Xiaojiawa Coal Mine of Shanxi Jinxing Energy Co.,Ltd.,Lvliang,Shanxi 035300,China;Shanghai Co.,Ltd.of China Coal Science and Industry Group,Shanghai 200030,China;Shenyang Research Institute Co.,Ltd.of China Coal Science and Industry Group,Fushun,Liaoning 113122,China)

机构地区：[1]山西锦兴能源有限公司肖家洼煤矿,山西吕梁035300 [2]中煤科工集团上海有限公司,上海200030 [3]中煤科工集团沈阳研究院有限公司,辽宁抚顺113122

出　　处：《自动化应用》2025年第2期224-228,共5页Automation Application

基　　金：辽宁省教育厅面上项目(JYTMS20231448)。

摘　　要：在煤矿生产过程中,瓦斯及有害气体长期对煤矿生产造成威胁,因此一种在线直观的监测手段尤为重要。在对红外光谱和深度学习的研究基础上,提出利用深度学习与红外光谱监测矿井有害气体。采用GoogLeNet模型分析矿井内有害气体的红外光谱,提出一种对称点式GoogLeNet红外光谱矿井有害气体监测方案。在对比实验中,综合考虑所有性能量化指标可以看出,GoogLeNet在针对井下有害气体的任务中具有较大优势,其mAP和准确率相较于其他模型更为理想。对数据集扩充后再次对GoogLeNet模型进行训练,能够使其性能进一步提升,以满足对于井下有害气体检测任务的基本要求。In the process of coal mine production,gas and harmful gases pose a threat to coal mine production for a long-time,so an online intuitive monitoring method is particularly important.Based on the study of infrared spectrum and deep learning,this paper puts forward the use of deep learning and infrared spectrum to monitor harmful gases in mines.The infrared spectrum of harmful gas in mine is analyzed by using GoogLeNet model,and a symmetric GoogLeNet infrared spectrum monitoring scheme for harmful gas in mine is proposed.In the comparative experiment,considering all the performance quantitative indicators comprehensively,it can be seen that GoogLeNet has great advantages in the task of detecting harmful gases underground,and its mAP and Accuracy are more ideal than other models.Training GoogLeNet model again after expanding the data set can further improve its performance and meet the basic requirements of this paper for detecting harmful gases underground.

关 键 词：矿井 有害气体 红外光谱 深度学习 

分 类 号：TD712[矿业工程—矿井通风与安全]