基于融合注意力的卷积神经网络滚动轴承故障诊断模型

Convolution Neural Network Fault Diagnosis Model of Rolling Bearing Based on Fusion Attention

摘要: 滚动轴承在工作中面临着高速运行、工况变化大等挑战，其振动信号具有复杂的形态成分，为了增强捕捉轴承故障信号的特征信息的能力，提出一种融合注意力机制和卷积神经网络的轴承故障诊断模型。首先，用格拉姆角场图像编码技术(GAF)将轴承振动信号转化为二维图像，再将这些二维图像输入融合注意力机制的卷积神经网络(GAF-CNN-SA)自动进行故障特征提取及分类。试验和对比结果表明，本文所提出的故障诊断模型能够针对不同种负载条件下的不同故障位置进行有效识别，并且在轴承故障诊断方面的效果优于其他智能算法。

Abstract: Rolling bearings are faced with challenges such as high-speed operation and great changes in working conditions, and their vibration signals have complex morphological components. In order to enhance the ability to capture the characteristic information of bearing fault signals, a bearing fault diagnosis model combining attention mechanism and convolutional neural network was proposed. Firstly, the bearing vibration signals are converted into two-dimensional images by Gram angular field image coding (GAF), and then these two-dimensional images are input into the Convolution neural network (GAF-CNN-SA) with integrated attention mechanism for automatic fault feature extraction and classification. The test and comparison results show that the fault diagnosis model proposed in this paper can effectively identify different fault locations under different load conditions, and the effect of bearing fault diagnosis is better than other intelligent algorithms.