流体机械

2026, v.54;No.649(07) 91-101

[打印本页] [关闭]
本期目录(Current Issue) | 过刊浏览(Archive) | 高级检索(Advanced Search)

基于振动信号特征提取的煤矿通风机故障诊断
Fault diagnosis of coal mine fans based on vibration signal feature extraction

董仕涛,张军,秦子健,尹伟,王磊
Dong Shitao,Zhang Jun,Qin Zijian,Yin Wei,Wang Lei

摘要(Abstract):

为提升煤矿主通风机的运行可靠性,提出一种基于小波包能量谱和一维卷积神经网络(1D-CNN)的故障诊断方法。以通风机6类典型工况的振动信号为研究对象,通过基于fk22小波基函数与rigrsure阈值规则的硬阈值算法对信号进行降噪处理,再经过3层小波包分解提取八维子频带能量特征,最后输入优化后的1D-CNN模型实现故障分类。实验台测试结果表明,该模型在6类典型工况识别中的验证准确率达到94.44%,较基准1D-CNN模型提高了11.11个百分点。矿区现场测试进一步表明,进风口轴承测点在敏感频带附近呈现出较明显的异常能量集聚现象,其特征与实验台故障样本具有较高的一致性,说明所提方法在现场复杂工况下具有一定的工程应用潜力。研究可为煤矿主通风机运行状态监测与故障预警提供参考。
To improve the operational reliability of coal mine main fans,a fault diagnosis method based on wavelet packet energy spectrum and 1D convolutional neural network(1D-CNN) is proposed.Taking vibration signals from six typical operating conditions of the fan as the research object,a hard thresholding algorithm based on the fk22 wavelet basis function and the rigrsure threshold rule is used for signal denoising.Then three-level wavelet packet decomposition is performed to extract eight-dimensional sub-band energy features,which are finally input into an optimized 1D-CNN model for fault classification.Test bench results show that the validation accuracy of the model in identifying the six typical operating conditions reaches 94.44%,which is 11.11 percentage points higher than that of the baseline 1D-CNN model.Further field tests in the mining area show that the measurement point at the air inlet bearing exhibits obvious abnormal energy concentration near the sensitive frequency band,and its characteristics are highly consistent with the fault samples from the test bench,indicating that the proposed method has certain engineering application potential under complex field conditions.This study can provide a reference for operating condition monitoring and fault early warning of coal mine main fans.

关键词(KeyWords): 煤矿通风机;小波阈值降噪;小波包能量谱;卷积神经网络;故障识别
coal mine fan;wavelet threshold denoising;wavelet packet energy spectrum;convolutional neural network;fault identification

Abstract:

Keywords:

基金项目(Foundation): 潞安化工集团科技项目(23A8103107C);; 国家自然科学基金项目(51175005)

作者(Author): 董仕涛,张军,秦子健,尹伟,王磊
Dong Shitao,Zhang Jun,Qin Zijian,Yin Wei,Wang Lei

参考文献(References):

扩展功能
本文信息
服务与反馈
本文关键词相关文章
本文作者相关文章
中国知网
分享