基于短时傅立叶变换和改进Vision Transformer的滚动轴承故障诊断方法  

Fault diagnosis method of rolling bearing based on short-time Fourier transform and improved Vision Transformer

作　　者：袁新杰 孙飞越 Yuan Xinjie;Sun Feiyue

机构地区：[1]武汉理工大学交通与物流工程学院,武汉430000

出　　处：《起重运输机械》2024年第16期70-75,共6页Hoisting and Conveying Machinery

摘　　要：针对传统故障诊断技术在精确与高效地诊断减速器滚动轴承故障信号方面所面临的挑战,文中提出了一种基于短时傅里叶变换与改进Vision Transformer模型的故障诊断新方法。此方法有效融合了短时傅里叶变换在处理非线性和非平稳信号上的优势以及Vision Transformer在图像分类任务上的卓越性能。通过短时傅里叶变换将一维的振动信号转化为包含时域和频域信息的二维图像数据,进而利用改进的Vision Transformer模型对这些图像数据进行处理,以实现对滚动轴承故障状态的精准诊断。在公开数据集上的实验结果验证了该方法的稳定性与高识别精度,展示了其在滚动轴承故障诊断领域的应用潜力。Considering the challenge of traditional fault diagnosis technology in accurately and efficiently diagnosing the fault signal of reducer rolling bearing,a new fault diagnosis method based on short-time Fourier transform and improved Vision Transformer model is proposed.This method effectively combines the advantages of short-time Fourier transform in processing nonlinear and non-stationary signals and the excellent performance of Vision Transformer in image classification tasks.One-dimensional vibration signal was transformed into two-dimensional image data containing time domain and frequency domain information by short-time Fourier transform,and then these image data were processed by improved Vision Transformer model in order to accurately diagnose rolling bearing faults.The experimental results on public data sets prove the stability and high recognition accuracy of this method,and show its application potential in the field of rolling bearing fault diagnosis.

关 键 词：短时傅里叶变换 Vision Transformer 深度学习 故障诊断 滚动轴承 

分 类 号：TH133.33[机械工程—机械制造及自动化]