China Mechanical Engineering ›› 2026, Vol. 37 ›› Issue (3): 656-667.DOI: 10.3969/j.issn.1004-132X.2026.03.015

Previous Articles    

Fault Diagnosis Method of Belt Conveyor Roller Bearings Based on NSST4-SVD-DBN

HU Kun1,2(), CHEN Zhuo2, HAN Xin3, JIANG Hao2, NIU Jie2   

  1. 1.State Key Laboratory of Digital Intelligent Technology for Unmanned Coal Mining,Anhui University of Science and Technology,Huainan,Anhui,232000
    2.Faculty of Mechanical and Electrical Engineering,Anhui University of Science and Technology,Huainan,Anhui,232000
    3.Huainan Mining Company Limited Guqiao Coal Mine,Huainan,Anhui,232000
  • Received:2025-04-25 Online:2026-03-25 Published:2026-04-08
  • Contact: HU Kun

基于NSST4-SVD-DBN的带式输送机托辊轴承故障诊断方法

胡坤1,2(), 陈卓2, 韩信3, 蒋浩2, 牛杰2   

  1. 1.安徽理工大学深部煤炭无人化开采数智技术全国重点实验室, 淮南, 232000
    2.安徽理工大学机电工程学院, 淮南, 232000
    3.淮南矿业有限责任公司顾桥煤矿, 淮南, 232000
  • 通讯作者: 胡坤
  • 作者简介:胡 坤*(通信作者),男,1981年生,教授、博士研究生导师。研究方向为带式输送机故障诊断。E-mail:hk924@126.com
  • 基金资助:
    国家自然科学基金(52274153);国家自然科学基金(51874004);芜湖研究院研发专项基金(ALW2021YF10)

Abstract:

Aiming at the problems of difficulty in extracting feature information generated by belt conveyor roller bearing faults, as well as low accuracy and poor robustness of fault diagnosis and identification, NSST4, SVD and DBN methods were combined to propose a suitable method for belt conveyor roller bearing acoustic signal fault diagnosis. Firstly, sequential variational mode decomposition (SVMD) was used to process the acoustic signals to enhance the recognizability of fault features. Second, the processed one-dimensional signals were converted to a two-dimensional time-frequency matrix by NSST4, which was used as the inputs of the feature matrix. Subsequently, the feature matrix was downsized using SVD technique to extract the key singular value vectors that might characterize the status of the roll bearings. These singular value vectors were then input into DBN, and the DBN core parameters were optimized by the improved sparrow search algorithm (ISSA) to improve the recognition performance of the model. Finally, in order to further validate the effectiveness of the proposed method, it was tested by simulated fault experiments and field experiments. In the simulated fault experiments of the roller bearings, the accuracy rate of the proposed method reaches 97.91%. Compared with other 5 methods, the accuracy of the proposed method is the highest, and the mean absolute error (MAE) is the lowest. In the field experiments, the recognition accuracy reaches 96.57%.

Key words: fault diagnosis, acoustic signal, double fourth-order synchronous compression transform (NSST4), singular value decomposition (SVD), deep belief network(DBN)

摘要:

针对带式输送机托辊轴承故障所产生的特征信息难以提取,以及故障诊断识别准确率低、鲁棒性差的问题,将二重四阶同步压缩变换(NSST4)、奇异值分解(SVD)与深度置信网络(DBN)相结合,提出一种带式输送机托辊轴承声信号故障诊断方法。利用逐次变分模态分解(SVMD)对声信号进行处理以增强故障特征的可辨识度。通过NSST4将处理后的一维信号转换为二维时频矩阵,并将该矩阵作为特征矩阵输入。采用SVD技术对特征矩阵进行降维处理,提取出能够表征托辊轴承状态的关键奇异值向量。这些奇异值向量随后被输入DBN中,DBN核心参数通过改进的麻雀搜索算法(ISSA)进行优化,以提高模型的识别性能。通过模拟故障实验和现场实验进行了测试,验证了所提方法的有效性。在托辊轴承的模拟故障实验中,所提方法实现了97.91%的准确率。与其他5种方法对比发现,所提方法准确率最高,且平均绝对误差(MAE)最低。在现场实验中,识别准确率可达96.57%。

关键词: 故障诊断, 声信号, 二重四阶同步压缩变换, 奇异值分解, 深度置信网络

CLC Number: