基于二维图像和CNN-BiGRU网络的滚动轴承故障模式识别

Rolling bearing fault mode recognition based on 2D image and CNN-BiGRU

为确保对滚动轴承故障诊断的有效性,结合卷积神经网络(CNN)在图像特征提取与分类识别的优势,利用格拉姆角场(GAF)将滚动轴承一维振动信号转换为二维图像数据,既保留了信号完整的信息,也保持着信号对于时间的依赖性。并由此提出基于卷积神经网络与双向门控循环单元(BiGRU)的诊断模型。首先将二维图像作为模型的输入数据,通过卷积神经网络提取图像的空间特征,再由双向门控循环单元筛选其时间特征,最终由分类器完成模式识别。通过对滚动轴承不同故障程度以及不同故障位置的诊断试验,准确率分别达到99.63%以及99.28%,其效果均优于其他常用算法,证明了所提方法的可行性。

Here,to ensure the effectiveness of rolling bearing fault diagnosis,combined with advantages of convolutional neural network(CNN)in image feature extraction and classification recognition,1D vibration signals of rolling bearing were converted into 2D image data using Gram angle field(GAF)to not only retain complete information of signals,but also maintain the dependence of signals upon time.Furthermore,a diagnosis model based on CNN and bidirectional gated recurrent unit(BiGRU)was proposed.Firstly,2D image data were taken as input data of the model,then spatial features of image data were extracted with CNN,and temporal features were screened with BiGRU.Finally,the mode recognition was finished with a classifier.It was shown that through diagnosis testing for rolling bearings with different fault degrees and different fault positions,the correct rates reach 99.63%and 99.28%,respectively,the effect is superior to other common algorithms to prove the feasibility of the proposed method.