【摘要】 针对现有一维卷积网络和残差网络在故障诊断方面的不足,本文将一维卷积网络与残差网络相结合,提出了一种基于改进一维残差网络的轴承故障诊断方法。该方法通过添加一条残差连接通道的方式,增加残差网络宽度,以学习更丰富的特征,提高故障诊断准确率。利用6种轴承状态对所提方法的分类效果进行了测试。实验结果表明,所提方法能直接利用振动信号,在较小训练与测试样本比的情况下实现故障诊断,当训练样本为90,测试样本为810(训练与测试样本比为1:9)时,驱动端故障诊断的正确率为99.6%;当训练样本为270,测试样本为630(训练与测试样本比为3:7)时,风机端故障的正确率为97.3%。更多还原

【Abstract】 Aiming at the shortcomings of the existing one-dimensional convolutional network and residual network in fault diagnosis, combining the one-dimensional convolutional network with the residual network, a bearing fault diagnosis method based on the improved one-dimensional residual network is proposed. This method increases the width of the residual network by adding a residual connection channel to learn richer features and improve the accuracy of fault diagnosis. Six bearing states have been used to test the classification effect of the proposed method. The experimental results show that the proposed method can directly use the vibration signal to realize fault diagnosis in the case of a small training-to-test sample ratio. When the training sample is 90 and the test sample is 810(the training-to-test sample ratio is 1:9), the correct rate of fault diagnosis on the drive end is 99.6%. When the training sample is 270 and the test sample is 630(the ratio of training to test samples is 3:7), the correct rate of fault diagnosis on the fan end is 97.3%.更多还原