摘要
传统的变分模态分解(VMD)方法在分解复杂信号时存在局限,难以彻底分解出不同信号分量,影响故障特征的准确提取。文章提出了一种特征模态分解(FMD)与多传感信息相融合并结合多尺度卷积神经网络(MCNN)的故障诊断方法。首先,采集在直升机传动系统试验台和主减速器试验台的轴承故障振动数据;其次,采用FMD技术将轴承振动加速信号进行分解,并对不同传感器分解后的各个信号进行重构与归化;最后,将重构后的信号分量作为不同尺度输入至MCNN网络,实现多尺度特征同时提取与融合,经SoftMax分类器对故障进行分类。结果表明,该方法更能有效提取出故障特征,在给定工况下,发动机轴承振动故障诊断的准确率可达100%。
To address the limitations of conventiona Variational Mode Decomposition(VMD)method in processing complex signals,which makes it difficult to completely decompose different signal components and affects the accurate extraction of fault features,it is proposed in this paper a novel fault diagnosis method integrating Feature Mode Decomposition(FMD)with multi-sensor information through a Multi-Scale Convolutional Neural Network(MCNN).The methodology involves:1)collecting vibration data from bearing faults in helicopter transmission system test bench and main reducer test bench;2)decomposing bearing vibration acceleration signals using FMD,follwed by reconstruction and normalization of decomposed signals from different sensors;3)feeding reconstructed signal components as multi-scale inputs to MCNN for concurrent feature extraction/fusion,with final faults classification via SoftMax.Experimental results demonstrate superior fault feature extraction capability,achieving 100%diagnostic accuracy for engine bearing vibration faults under specified operating conditions.
作者
朱曾祯
万安平
程晓民
纪晓声
蒋俊杰
王景霖
单添敏
ZHU Zengzhen;WAN Anping;CHENG Xiaomin;JI Xiaosheng;JIANG Junjie;WANG Jinglin;SHAN Tianmin(School of Mechanical and Electrical Engineering,Anhui University of Science and Technology,Huainan Anhui 232001;Department of Mechanical Engineering,Hangzhou City University,Hangzhou Zhejiang 310015;Key Laboratory of Fault Diagnosis and Health Management Technology,Aerospace Science and Technology,Shanghai 201601)
出处
《湖北理工学院学报》
2025年第3期8-14,63,共8页
Journal of Hubei Polytechnic University
基金
国家自然科学基金资助项目(项目编号:52372420)
浙江省科技计划项目(项目编号:2025C02242,2024C01039)
宁波市科创甬江2035关键技术突破计划项目(项目编号:2024Z177)
宁波市重大专项(项目编号:2022Z071,2023Z032)。
关键词
特征模态分解
多尺度卷积网络
振动故障诊断
航空发动机
轴承
feature mode decomposition
multi-scale convolutional neural network
vibration fault diagnosis
aircraft engine
bearing
