关键词: k-means, 故障诊断, 谱聚类, 图分割, 并联机构 闭环结构 运动学分析 性能分析 轨迹规划

Abstract: The network model of fault diagnosis is put forward, so the data clustering is transformed to network segmentation. And the min-max cut criterion is taken as objective function of segmentation. In view of the disadvantage of the higher computation complexity in the traditional min-max cut criterion optimization algorithm, an algorithm using k-means to improve the process of searching optimal segmentation point is introduced. The applications such as benchmark data and four-stage piston compressor diagnosis problem show that the new algorithm has no strict requirements on data distribution, and can achieve feature extraction and diagnosis fast and effectively.

Key words: Fault diagnosis, Graph segmentation, k-means, Spectral clustering, parallel mechanism closed-loop structure kinematics analysis performance analysis trajectory planning