基于多通道时频域信号的卷积神经网络智能故障诊断技术

在滚动轴承故障诊断中，算法难以学习所有负载下的健康状态特征，为有效诊断滚动轴承在变负载下的健康状态，算法需要较强的负载域适应能力。针对上述问题，提出了基于多通道时频域信号的卷积神经网络算法。不同的小波提取不同的特征，算法采用多种小波可以提供多样的健康状态特征。并且全局最大池化替换每一空洞卷积之后的最大池化，从全局范围内提取最大激活。因此，算法只需在源域下训练，即可在目标域下得到良好的诊断效果。为验证该算法的有效性，利用公共数据集进行实验。实验结果表明，该算法在不同负载下的分类精度较其他算法有明显提高，从而可以有效识别滚动轴承的健康状态。

In the field of rolling bearing fault diagnosis, existing algorithms struggle to learn health state features across all load conditions. To effectively diagnose the health states of rolling bearings under variable loads, algorithms must possess strong load domain adaptation capabilities. To address this issue, this study proposes a convolutional neural network algorithm based on multi-channel time-frequency domain signals. Different wavelets extract distinct features, and adopting multiple wavelets allows the algorithm to acquire diverse health state features. Moreover, global max pooling replaces the standard max pooling following each atrous convolution, enabling the extraction of maximum activations from the global scope. As a result, the algorithm only needs to be trained on the source domain to achieve favorable diagnostic performance on the target domain. To validate the effectiveness of the proposed algorithm, experiments are carried out using public datasets. The experimental results demonstrate that the classification accuracy of the proposed algorithm under different load conditions is notably improved compared with other algorithms, thereby enabling effective identification of the health states of rolling bearings.

搜集汇总

背景与挑战

背景概述

该数据集是一篇关于智能故障诊断技术的学术论文，主要研究基于多通道时频域信号的卷积神经网络算法，用于提高滚动轴承在变负载条件下的健康状态诊断精度。论文通过引入多种小波特征提取和全局最大池化技术，增强了算法的负载域适应能力，实验证明其在不同负载下的分类性能优于其他方法。