【摘要】 针对滚动轴承在多种工况下多种故障共存导致的智能故障诊断模型准确性下降问题，本文提出了一种二维卷积神经网络（2D-CNN）和长短期记忆网络（LSTM）集成的滚动轴承多故障诊断模型（MF2D-CNN-LSTM）。用2D-CNN代替以往的1D-CNN作为空间特征提取器，以获取信号多种局部特征和全局特征，并引入批量归一层（BN）来优化模型。使用平均池化层代替最大池化层，全局平均池化层代替Flatten，将LSTM层作为信号时序信息特征提取器，对该模型处理数据的特征提取过程进行可视化。综合信号的空间和时序信息特征，通过分类层输出分类结果并将其可视化。通过西安交通大学XJTU-SY实验数据，验证了所提方法在多种工况下的多种轴承故障共存分类方面具有出色的效果。更多还原

【Abstract】 This paper proposes a rolling bearing multi fault diagnosis model(MF2D-CNN-LSTM) that integrates 2D convolutional neural network(2D-CNN) and long short term memory network(LSTM) to address the issue of decreased accuracy of intelligent fault diagnosis models caused by the coexistence of multiple faults in rolling bearings under various working conditions. 2D-CNN is used instead of the previous 1D-CNN as the spatial feature extractor to obtain multiple local and global features of the signal, and a batch normalization(BN) layer is introduced to optimize the model. The average pooling layer is used instead of the maximum pooling layer, the global average pooling layer is used instead of the Flatten layer, and the LSTM layer is used as the signal temporal information feature extractor. Visualize the feature extraction process of the data processed by the model. By integrating the spatial and temporal information characteristics of the signal, the classification results are output through the classification layer and visualized. The XJTU-SY experimental data from Xi’an Jiaotong University has verified that the proposed method has excellent performance in classifying multiple bearing faults under various working conditions.更多还原