基于贝叶斯优化与CBAM-ResNet的乏燃料剪切机故障诊断方法

乏燃料剪切机是动力堆乏燃料后处理首端的重要设备，状态监测与故障诊断对于保证乏燃料剪切机的安全运行、避免重大事故、减少其维修时间和费用有着重要的作用。针对目前我国针对乏燃料剪切机的故障诊断研究少、数据获取难度大、故障诊断的准确率低等问题，构建基于贝叶斯优化与卷积块注意力模块CBAM的残差神经网络模型。首先在利用双声道差分法对噪声降噪，将其转化为梅尔频谱图并进行数据增强；其次引入CBAM对残差网络进行改进，提高网络的深层次特征提取能力，并利用贝叶斯优化算法训练优化器等超参数，得到最优超参数后重新训练网络模型。最后，通过实验结果显示所构建模型的诊断准确率为93.67%，对比其他方法有显著的提高。

Fault diagnosis of shearing machines based on improved residual network

The spent fuel shearing machine is an important equipment at the head end of the power reactor spent fuel reprocessing. Condition monitoring and fault diagnosis play an important role in ensuring the safe operation of the spent fuel shear, avoiding major accidents, and reducing its maintenance time and cost. In view of the problems such as the lack of research on fault diagnosis of spent fuel shears in China, the difficulty of data acquisition, and the low accuracy of fault diagnosis, a residual network model based on Bayesian optimization and CBAM is constructed. First of all, the noise is reduced by using the dual-channel difference method, and it is converted into mel spectrograms. Secondly, CBAM is introduced to improve the residual network and improve the deep feature extraction ability of the network. Bayesian optimization algorithm is used to train the super parameters such as the optimizer, and the network model is retrained after the optimal super parameters are obtained. Finally, the experimental results show that the diagnostic accuracy of the model is 93.67%, which is significantly improved compared with other methods.