多域信息融合结合改进残差密集网络的轴承故障诊断

针对滚动轴承原始时域信号信息单一、深度卷积神经网络提取的特征对信息的传递存在差异等问题，该研究提出了一种多域信息融合与改进残差密集网络的轴承故障诊断方法。为了获取故障的多方面信息，先对原始数据进行多域变换，再将融合信息输入经卷积注意力改进的残差密集网络进行深度学习。经注意力机制改进的网络能够实现对提取特征的重要性区分，提高网络的训练速度、改善识别准确率。试验结果及对比分析表明该算法可以提取较为全面的特征，较传统方法具有更好的识别效果。

Bearing fault diagnosis based on information fusion and improved residual dense networks

Aiming at the problem that the original time signal of a rolling bearing is relatively simple and the features extracted by convolution neural network are different during useful information transmission,a bearing fault diagnosis method was proposed based on multi-domain information fusion and an improved residual dense network.In order to obtain the multifaceted information of fault signal,the original data are transformed in multiple domains.Then the multi-domain information is input into the residual dense network improved by convolution attention mechanism for deep learning.Discrimination is realized according to the importance of the extracted features.The training speed and efficiency of the neural network were improved.Experimental results and analysis show that the proposed method can extract more comprehensive features and has higher recognition accuracy than traditional methods.