基于STFT-ECA-ResNet18网络模型的滚动轴承变负载故障诊断

Rolling Bearing Variable Load Fault Diagnosis Based on STFT-ECAResNet18 Network Model

针对传统方法处理变负载轴承故障诊断时存在的自适应能力弱,模型泛化性差的问题,提出了一种改进的基于深度残差网络的故障诊断方法。首先,将采集到的一维时间序列信号进行短时傅里叶变换得到二维时频数据,再利用二维卷积神经网络从变换后的数据中提取特征。然后,通过高效通道注意力机制获取通道全局信息并对其权值进行调整,以增强改进网络模型的泛化能力,使其在变负载工况下分类效果得到提高。最后,通过仿真对所提方法进行了验证,结果表明相比传统方法诊断效果改进明显。

Aiming at the problems of weak adaptive ability and poor model generalization of variable load bearing fault diagnosis by traditional methods,an improved fault diagnosis method based on deep residual network is proposed.Firstly,the collected one-dimensional time series signals are converted into two-dimensional time-frequency data by short-time Fourier transform,and features are extracted from the transformed data by using two-dimensional convolutional neural network.Then,the efficient channel attention mechanism is used to obtain the channel global information and adjust its weight,so as to enhance the generalization ability of the improved network model and improve the classification effect under variable load conditions.Finally,the proposed method is verified by simulation,and the results show that the diagnosis effect is improved significantly compared with the traditional method.