基于STFT同步压缩变换和ResNet参数迁移的滚动轴承智能诊断  

Intelligent Diagnosis of Rolling Bearing Based on STFT Synchrosqueezing Transform and ResNet Parameter Migration

作　　者：傅顺军[1] 罗强[1] 李江[1] 马相龙 孙博闻 FU Shunjun;LUO Qiang;LI Jiang;MA Xianglong;SUN Bowen(Shanghai Marine Equipment Research Institute,Shanghai 200030,China;Research Institute of Advanced Equipment,School of Energy Engineering,Zhejiang University,Hangzhou 310058,China)

机构地区：[1]上海船舶设备研究所,上海200030 [2]浙江大学能源工程学院特种装备研究所,杭州310058

出　　处：《船舶工程》2024年第12期58-67,74,共11页Ship Engineering

摘　　要：为提升舰船用泵的滚动轴承的故障诊断准确率,提出一种基于短时傅里叶变换(STFT)同步压缩变换和ResNet参数迁移的智能诊断方法。采用基于STFT的同步压缩变换算法,将轴承的一维振动信号变换为二维时频图,再将时频图输入至基于参数迁移的深度残差卷积神经网络(ResNet)中进行特征提取和故障识别。利用实验室的轴承故障模拟数据对所提方法的有效性进行验证,并将所提的网络模型与其他经典模型进行对比。所提方法具有较高的诊断准确率、抗噪性和泛化性。研究结果可为舰船用泵的滚动轴承的智能诊断提供参考。In order to improve the fault diagnosis accuracy of rolling bearings for marine pumps,an intelligent diagnosis method based on short-time fourier transform(STFT)synchrosqueezing transform and ResNet parameter migration is proposed.A synchrosqueezing transform algorithm based on STFT is used to transform one-dimensional vibration signals of bearings into two-dimensional time-frequency maps,which are then input into the deep residual convolutional neural network(ResNet)based on parameter migration for feature extraction and fault identification.The effectiveness of the proposed method is verified by using the fault simulation data of bearings in the laboratory,and the proposed network model is compared with other classical models.The proposed method has high diagnostic accuracy,anti-noise performance and generalization.The research results can provide a reference for the intelligent diagnosis of rolling bearings for marine pumps.

关 键 词：短时傅里叶变换 同步压缩变换 残差网络 迁移学习 

分 类 号：U671.99[交通运输工程—船舶及航道工程]