基于半监督拉普拉斯特征映射的压缩机故障辨识

Compressor Fault Recognition Based on Semi-Supervised Laplacian Eigenmaps

旋转机械在现代生产体系中具有不可替代的作用,其故障诊断技术对避免恶性损坏事故的发生显得尤为重要。如何选择和提取有效的故障特征,将直接影响故障辨识的诊断精度。针对旋转机械故障诊断的非线性、非平稳性等特点,结合半监督学习和流形学习思想,提出了一种半监督拉普拉斯特征映射(SSLE)算法,并将其应用于空气压缩机的故障辨识。该方法充分利用少量标签样本和大量无标签样本信息,提取有利于分类的故障样本低维流形特征,并利用最小二乘支持向量机(LS-SVM)分类器进行了故障分类与辨识。采用非线性的特征学习方式,有效提取了故障信号中的敏感特征信息,增强了故障模式识别的分类性能。压缩机故障辨识试验结果表明,与主成分分析(PCA)算法和拉普拉斯特征映射(LE)算法相比,基于SSLE算法的故障辨识性能更好。

Rotating machinery plays an irreplaceable role in modem production system,and its fault diagnosis technology is very important to avoid the occurrence of vicious damage accidents. How to select and extract the effective fault features will directly affect the diagnosis accuracy of fault identification. Aiming at the characteristics of non - linearity and non - stationary in fault diagnosis of rotating machinery, and combined with semi - supervised learning and manifold learning, a semi - supervised laplacian eigenmaps algorithm（ SSLE） is proposed,and is applied to the fault identification of air compressor. This method takes information of a small number of labeled samples and a large number of unlabeled samples, to extract the low - dimensional manifold features of faulty samples which are helpful for classification ; and the least square support vector machine （ LS - SVM） classifier is used for the fault classification and identification. By adopting nonlinear feature learning mode, the proposed method effectively extracts the sensitive feature information of the fault signals, and enhances the classification performance of the fault pattern recognition. The test results of compressor fault identification show that compared with the principal component analysisc （PCA） method and Laplacian eigenmaps（LE） method,the SSLE method has better fault identification performance.