中国安全科学学报 ›› 2022, Vol. 32 ›› Issue (10): 76-82.doi: 10.16265/j.cnki.issn1003-3033.2022.10.1766

• 安全工程技术 • 上一篇    下一篇

基于OCISVM的矿井通风系统在线故障诊断

赵丹1,2(), 沈志远1,2,**(), 刘晓青1,2   

  1. 1 辽宁工程技术大学 安全科学与工程学院,辽宁 阜新 123000
    2 辽宁工程技术大学 矿山热动力灾害与防治教育部重点实验室,辽宁 葫芦岛 125105
  • 收稿日期:2022-04-20 修回日期:2022-08-15 出版日期:2022-10-28 发布日期:2023-04-28
  • 通讯作者: 沈志远
  • 作者简介:

    赵丹 (1982—),女,辽宁阜新人,博士,教授,主要从事矿井智能通风、煤矿粉尘防治等方面的研究。E-mail:

  • 基金资助:
    辽宁省教育厅基金资助(LJ2019JL025)

Online fault diagnosis of mine ventilation system based on OCISVM

ZHAO Dan1,2(), SHEN Zhiyuan1,2,**(), LIU Xiaoqing1,2   

  1. 1 College of Safety Science & Engineering, Liaoning Technical University, Fuxin Liaoning 123000, China
    2 Key Laboratory of Mine Thermo-motive Disaster and Prevention, Ministry of Education, Liaoning Technical University, Huludao Liaoning 125105, China
  • Received:2022-04-20 Revised:2022-08-15 Online:2022-10-28 Published:2023-04-28
  • Contact: SHEN Zhiyuan

摘要:

为解决矿井通风系统故障样本获取困难以及在线故障诊断研究相对匮乏的问题,填补应用传感器实时监测数据进行故障分支诊断的空白,构造一分类支持向量机(OC-SVM)与增量学习(IL)相结合的OCISVM模型。首先,在离线阶段,运用传感器监测到的正常样本数据构造分类超平面;然后,在线检测阶段,依据IL的思想,通过引入德尔塔函数更新分类超平面;最后,利用东山矿通风系统数据库验证并分析OCISVM模型。结果表明:该模型的故障分支诊断准确率可达96.5%,诊断时间开销在毫秒级,在处理不平衡数据时稳定性更高。

关键词: 矿井通风系统, 一分类支持向量机(OC-SVM), 增量学习(IL), 故障诊断, 监测数据

Abstract:

In order to address the difficulty in obtaining failure samples of mine ventilation system and the lack of research on online detection to fill the blank in fault branch diagnosis based on real-time monitoring data of sensors, OC-SVM and IL method were combined to construct OCISVM model. Firstly, monitoring data of normal samples were used to construct classification hyperplane at offline phase. Then, at online detection stage, classification hyperplane was updated by introducing delta function according to incremental learning, and online fault branch diagnosis was achieved based on threshold criterion. Finally, the proposed model was applied to Dongshan coal mine ventilation system. The results show that the model's fault diagnosis accuracy can reach as high as 96.5% while running in milliseconds. Moreover, it demonstrates higher stability when dealing with unbalanced data.

Key words: mine ventilation system, one-class support vector machine (OC-SVM), incremental learning (IL), fault diagnosis, monitoring data