摘要
针对滚动轴承故障诊断问题,提出了一种基于图像形状特征和线性局部切空间排列(LLTSA)的故障诊断方法。首先采用SDP(Symmetrized Dot Pattern)方法对时域信号进行变换,得到极坐标空间下的雪花图像,在分析图像特点的基础上,从图像处理的角度初步提取出图像的形状特征;然后利用LLTSA对初步提取的特征进行维数约简以提取低维特征;最后采用支持向量机(SVM)对低维特征进行分类评估。滚动轴承的故障诊断实验表明图像形状特征能够表征轴承的状态,经LLTSA约简后特征数据的复杂度得到降低,且具有更好的聚类效果,而SVM对轴承4种状态的识别率也达到了100%,验证了该方法的有效性。
Aiming at fault diagnosis problems of rolling bearing,a fault diagnosis method based on image shape features and linear local tangent space alignment (LLTSA)was proposed.Firstly,a time waveform was transformed into a snowflake image in polar coordinate space with the symmetrized dot pattern (SDP)method.The image shape features were initially extracted on the basis of analyzing the characteristics of the image.Secondly,LLTSA was introduced to compress the initial high-dimensional features into low-dimensional ones with a better discrimination.Finally,a support vector machine (SVM)was employed to classify and evaluate low-dimensional features.The test results of rolling bearing fault diagnosis showed that the image shape features can characterize bearing states,and LLTSA can reduce the complexity of feature data and enhance their clustering effect;furthermore,a relatively higher identification rate of four bearing states reaches 100% with SVM,the effectiveness of the proposed method was verified.
出处
《振动与冲击》
EI
CSCD
北大核心
2016年第9期172-177,共6页
Journal of Vibration and Shock
基金
军内科研项目
关键词
SDP
形状特征
线性局部切空间排列
支持向量机
故障诊断
symmetrized dot pattern (SDP)
shape feature
linear local tangent space alignment (LLTSA)
support vector machine (SVM)
fault diagnosis