基于多核监督流形学习的旋转机械故障诊断

Fault diagnosis of rotating machinery based on multi-kernel supervised manifold learning

为了准确地对旋转机械进行故障诊断,提出了一种多核监督流形学习算法(multi-kernel supervised manifold learning,MKSML)。MKSML算法可以有效地对高维故障数据进行特征选择,筛选出区分度高的低维故障特征。借助监督学习的思想,增强了同类样本的聚集性和不同类样本之间的差异性;同时基于所设计的多核函数提出了加权邻域图构建方法,能够保留近邻点之间的距离信息和角度信息,有效地抑制故障特征选择时样本中的异常值和噪声的干扰。通过灰狼优化算法调整MKSML算法相应的参数,使算法能够应用于不同类型的旋转机械故障诊断。在此基础上,建立了一种基于MKSML算法的旋转机械故障诊断模型,并进行了轴承故障诊断实验以及齿轮故障诊断实验。

In order to accurately perform fault diagnosis for rotating machinery,a multi-kernel supervised manifold learning(MKSML)algorithm was proposed.More specifically,MKSML algorithm allowed to effectively select the features of high-dimensional fault data,and extract the low-dimensional fault features with better discrimination.Through the idea of supervised learning,the clustering of similar samples and the differences between various samples have been enhanced.A novel weighted neighborhood graph was proposed by constructing multi-kernel function.The distance information and angle information between adjacent points were retained.And the interference of outliers and noise in the sample was suppressed.Through the gray wolf optimization algorithm to adjust the MKSML parameters,the algorithm could be applied to various types of rotating machinery fault diagnosis.The fault diagnosis model of rotating machinery based on MKSML was proposed,and bearing fault diagnosis experiments and gear fault diagnosis experiments were conducted.