【摘要】 针对旋转机械齿轮箱的齿轮与轴承故障,提出一种融合对称点图案SDP(Symmetrized Dot Pattern,SDP)和卷积神经网络CNN(Convolutional Neural Network)的故障诊断方法。首先,以模拟实验台MCDS获取大量故障实验数据,经过预处理产生的一维振动信号再经SDP转化为特征信息丰富的二维雪花图像;然后将SDP图像输入至CNN自动提取特征,再用分类器识别故障特征。实验证明,该方法能够有效和准确地识别齿轮箱的故障,各类故障的识别正确率在96%以上。更多还原

【Abstract】 A novel fault diagnosis method combining Symmetrized Dot Pattern(SDP) and convolutional neural network(CNN)is proposed for gear and bearing faults of a rotating machinery gearbox in this paper.Firstly,a large number of fault experimental data are obtained by simulated experimental platform MCDS,and one-dimensional vibration signals generated by preprocessing are transformed into two-dimensional snowflake images with rich feature information by SDP.Then the SDP image is input to CNN to automatically extract the features,and the classifier is used to identify the fault features.Experiments show that this method can identify the faults of gearbox effectively and accurately,and the correct rate of all kinds of faults is above 96%.更多还原