摘要
针对传统故障诊断方法不能有效识别滚动轴承振动信号且难以在振动信号中提取所需特征、预处理步骤过多和难以区分故障类型的问题,提出了一种新型卷积神经网络(CNN)模型。首先对传统CNN的激活函数和池化方式进行改进,然后以卷积长短时记忆(Conv-LSTM)作为卷积层。相比于传统的故障诊断方法,新型CNN模型可直接训练切片后的原始加速度振动信号,并可同时识别和分类出故障类型。SKF6005型轴承实验结果表明,新型CNN模型相对于纯CNN模型具有更好的分类效果和更高的分类准确度。
Aiming at the problems that traditional fault diagnosis methods can not effectively identify the vibration signal of rolling bearing, it is difficult to extract the required features from the vibration signal, there are too many preprocessing steps and it is difficult to distinguish the fault types, a new convolutional neural network(CNN) model was proposed. Firstly, the activation function and pooling method of traditional CNN were improved, and then convolution long short-term memory(Conv-LSTM)network was used as convolution layer. Compared with the traditional fault diagnosis method, the new CNN model can directly train the original acceleration vibration signal after slicing, identify and classify the fault types at the same time. The experimental results of SKF6005 bearing show that the new CNN model has better classification effect and higher classification accuracy than pure CNN model.
作者
毕鹏远
邱超
宋强
Bi Pengyuan;Qiu Chao;Song Qiang(College of Electrical Engineering,Henan University of Technology,Zhengzhou 450001,China;College of Mechanical and Electrical Engineering,Henan University of Technology,Zhengzhou 450001,China)
出处
《煤矿机械》
2021年第8期186-189,共4页
Coal Mine Machinery
基金
国家自然科学基金面上项目(61673160)。
