作　　者： ; ; ; ;

机构地区： 华南理工大学汽车摩擦学与故障诊断研究所

出　　处： 《润滑与密封》 2012年第8期7-10,56,共5页

摘　　要： 针对能量耗损的故障诊断方法,提出一种基于流形学习与支持向量机的能量耗损信号分析算法。该算法将采集的能量数据重构到高维空间中,利用扩散映射(Diffusion Maps)的方法提取高维能量信号中的低维流形特征,然后利用支持向量机(SVM)对提取的低维流形特征进行分类,并用分类的准确率作为算法有效性的衡量指标;同时,利用局部切空间排列(LTSA)方法对能量信号进行分析,以比较2种算法对能量信号特征的提取能力。结果表明,基于扩散映射的方法具有更好的低维特征提取效果,从而证明了该算法对于能量耗损信号分析的有效性,为实现基于能量耗损的故障诊断奠定了基础。 Based on the energy-dissipation-based fault diagnosis method, a manifold-learning-based energy dissipation signal analysis algorithm was proposed. The data of energy dissipation was reconstructed to a high-dimensional space, the diffusion map method was used directly to extract the intrinsic low-dimensional manifold feature from the high-dimensional energy signal, the support vector machine （SVM） was adopted to classify the low-dimensional manifold feature, and the effi- ciency of the algorithm was evaluated via the accuracy of such classification. In order to compare the ability of feature ex- traction, another method of local tangent space alignment（LTSA） was introduced to analyze the energy dissipation signal. The experimental results show that the method of diffusion maps realizes better feature extraction, which verifies the feasi-bility of the Diffusion Maps-SVM algorithm, and further lays a foundation to the realization of the energy-dissipation-based fault diagnosis.

关 键 词： 流形学习 扩散映射 能量耗损 支持向量机 故障诊断

领　　域： [自动化与计算机技术] [自动化与计算机技术]