基于ISCNN-LightGBM的轴承故障诊断

在传统卷积神经网络与分类器相结合的故障诊断方法中, CNN用于故障特征提取时, 存在着提取的特征质量不高与运行时间较长的问题. 针对以上问题, 本文提出了一种基于改进单层卷积神经网络及LightGBM的故障诊断模型. 该模型通过将特征距离函数嵌入CNN的损失函数中, 提升了CNN特征提取的能力, 增强了CNN与后续分类器之间的联系, 从而提升了整体模型的故障诊断能力. 于此同时, 经过改进的单层的卷积神经网络进一步缩短了模型运行的时间, 提升了模型的诊断效率. 通过对两个不同的公共数据集进行对比实验, 其结果表明, 本文所提诊断模型对多种轴承故障的诊断准确率与诊断效率显著高于其他诊断模型.

In traditional fault diagnosis methods combining the convolutional neural network (CNN) and classifier, there are problems of low-quality features and long-running time when the CNN is used to extract fault features. In this paper, to solve the above problems, a fault diagnosis model based on an improved single-layer convolutional neural network and LightGBM is established. By embedding the feature distance function into the loss function of CNN, the model improves the ability of CNN feature extraction and enhances the connection between CNN and subsequent classifiers, thereby improving the fault diagnosis ability of the overall model. At the same time, the improved single-layer convolutional neural network further shortens the running time of the model and improves its diagnostic efficiency of the model. Through comparative experiments on two different public data sets, the results show that the diagnostic accuracy and efficiency of the proposed model are significantly higher than that of other diagnostic models for various bearing faults.