基于多传感器信号和卷积神经网络的滚动轴承故障诊断

针对滚动轴承故障信号非平稳非线性且易受背景噪声干扰的特点,结合深度学习的优势,提出了一种基于卷积神经网络(CNN)的滚动轴承故障诊断法。将不同故障下多个传感器测得的1维(1D)振动信号转化为2维(2D)灰度图像作为网络输入,并将其分为训练集和测试集;将训练集输入卷积神经网络进行训练,自动提取其中的特征;测试集被用于验证学习完毕的网络的有效性,实现滚动轴承故障识别。该方法不依赖于人为经验和信号处理技术进行预先的信号特征提取,实验数据分析表明,相比于经典的支持向量机和概率神经网络方法,提出的方法识别准确率更高且更稳定。

Fault diagnosis for rolling element bearings based on multi-sensor signals and CNN

Aiming at the non-stationary and nonlinear characteristics of rolling element bearings’ fault signals which are easily interfered by background noise, a fault diagnosis method was proposed in this paper based on convolutional neural network (CNN) due to the advantages of deep learning methods.First, the one dimension (1D) bearing vibration signals collected by multi-sensors under different faults were converted to two dimension (2D) gray images as the input of CNN which are divided into training set and testing set.Then, the CNN was trained by the training set and the representative features can be extracted automatically.Finally, the effectiveness of the trained CNN was verified by the testing set to identify the fault types of bearings.The proposed method does not rely on human experience and signal processing techniques for the pre-extraction of fault features.The analysis results using experimental signals show that the proposed method has higher and more stable prediction accuracy compared with the traditional support vector machine and probabilistic neural network method.