中国安全科学学报 ›› 2023, Vol. 33 ›› Issue (1): 95-104.doi: 10.16265/j.cnki.issn1003-3033.2023.01.0846

• 安全工程技术 • 上一篇    下一篇

基于DWT与SVM的风门开闭阶段识别方法

邓立军1,2(), 尚文天1,2,**(), 刘剑1,2, 周煜凯1,2, 宋莹3   

  1. 1 辽宁工程技术大学 安全科学与工程学院,辽宁 葫芦岛 125105
    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
    3 山东工商学院 管理科学与工程学院,山东 烟台 264005
  • 收稿日期:2022-09-12 修回日期:2022-12-10 出版日期:2023-01-28 发布日期:2023-07-28
  • 通讯作者: ** 尚文天(1998—),男,内蒙古赤峰人,硕士研究生,研究方向为矿井智能通风与异常识别。E-mail:
  • 作者简介:

    邓立军 (1985—),男,湖北京山人,博士,副教授,主要从事通风网络解算、矿井智能通风等方面的研究。E-mail:

    刘剑, 教授

  • 基金资助:
    国家自然科学基金资助(51904143); 山东省自然科学基金资助(ZR2020QE125)

Identification of air door opening and closing stage based on DWT and SVM

DENG Lijun1,2(), SHANG Wentian1,2,**(), LIU Jian1,2, ZHOU Yukai1,2, SONG Ying3   

  1. 1 School of Safety Science and Engineering, Liaoning Technical University, Huludao Liaoning 125105, China
    2 Key Laboratory of Mine Thermo-motive Disaster and Prevention, Ministry of Education, Huludao Liaoning 125105, China
    3 School of Management Science and Engineering, Shandong Institute of Business and Technology, Yantai Shandong 264005, China
  • Received:2022-09-12 Revised:2022-12-10 Online:2023-01-28 Published:2023-07-28

摘要:

为解决因风门开闭导致的风速传感器数据异常波动与误报警问题,提出一种基于离散小波变换(DWT)与支持向量机(SVM)的风门开闭阶段识别方法。使用多尺度滑动窗口将传感器风速监测数据离散化为若干段不同尺度的子时间序列数据,利用统计方法与DWT,提取各尺度子时间序列数据中的统计特征与隐含的波动特征,建立SVM风门开闭阶段识别分类模型。为进一步优化识别结果,基于重叠度(IoU)规则合并、修正、组合、取优分类识别结果,再根据相似准则建立长度方向取变率为2、整体相似比为1∶16的相似试验模型,开展风门开闭扰动试验,验证方法的可行性。结果表明:在测试集上的识别准确率较高,对于风门开闭时间的识别准确率可达到90.08%,风门开闭阶段的划分准确率可达到71.05%,优化滑动窗口尺度数量,可继续增加方法识别的准确率。

关键词: 离散小波变换(DWT), 支持向量机(SVM), 风门开闭, 阶段识别, 多尺度滑动窗口, 重叠度(IoU)

Abstract:

In order to solve the false alarm problem of wind speed sensor and abnormal fluctuation of wind speed sensor data caused by air door opening and closing, a recognition method of air door opening and closing stage based on DWT and SVM was proposed. A multi-scale sliding window was used to discretize the continuous wind speed monitoring data into several sub-time series of different scales. Statistical methods and DWT were used to extract statistical features and hidden wave features in discrete data of sub-time series data of various scales, and SVM air door opening and closing stage recognition and classification models were established. In order to further optimize the recognition results, the classification recognition results were merged, modified, combined and optimized based on the IoU rule. Finally, according to the similarity criterion, the similarity experimental model with the length direction variable rate of 2 and the overall similarity ratio of 1∶16 was established, and the feasibility of the method was verified by the similarity test of the air door opening and closing disturbance. The results show that the recognition accuracy of this method is high in the test set, the recognition accuracy of air door opening and closing time can reach 90.08%, while the classification accuracy of air door opening and closing stage can reach 71.05%. Hence the optimization on the number of sliding window scales can increase the recognition accuracy.

Key words: discrete wavelet transform (DWT), support vector machine (SVM), air door opening and closing, stage identification, multiple-scale sliding window, intersection over union (IoU)