基于一维卷积神经网络的滚动轴承自适应故障诊断算法

Adaptive fault diagnosis algorithm for rolling bearings based on one-dimensional convolutional neural network

现有的滚动轴承故障诊断算法依赖于人工特征提取和专家知识,然而滚动轴承复杂多变的工作环境使得传统的智能故障诊断算法缺乏自适应性。针对此问题,提出了基于＂端到端＂的自适应一维卷积神经网络（ACNN-FD）故障诊断算法。首先,将各类故障状态的原始振动信号进行有重叠分段预处理用于构建训练样本和测试样本;然后,将每个训练样本以某一尺度的＂时间步＂进行划分作为所建立的一维卷积神经网络模型的输入,利用深度网络结构实现对原始振动信号特征的自适应层级化提取;最后在输出端利用Softmax分类器输出诊断结果。通过轴承数据库实验表明算法能够实现高达99%以上的故障识别准确率,同时在不同负载下良好的泛化性能,具备实际应用的可行性。

Current intelligent fault diagnosis methods largely depend on manual feature extraction and expertise knowledge. However,the complex and changeable working environment of bearings make traditional fault diagnosis algorithms lack of adaptability. To solve this problem,an ＂end-to-end＂adaptive fault diagnosis algorithm based on one-dimensional convolutional neural network called ADCNN-FD is proposed. Firstly,different fault signals of bearings are segmented into training and testing datasets. Then,all training samples divided by a certain ＂time steps＂are utilized as input of the model,and the features are extracted from the raw temporal vibration signal by deep architecture adaptively and hierarchically. Finally,a Softmax classifier is utilized to output diagnosis results at the top of the model. Experiments of rolling bearing datasets demonstrate that the proposed method can not only achieve more than 99% fault recognition accuracy,but also obtain good generalization performance under variable loads,which is applicable in practice.