Abstract: A new fault diagnosis method based on orbit manifold topological space is put forward. Firstly, vibration signals from the working is reduced the noise by using EEMD for vibration signal noise reduction. Restored signal constitutes an orbit. An orbit of the orbit track points is defined as a dimension feature of the high-dimensional space; the high-dimensional characteristics space is reduced into lower-dimensional space based on LTSA algorithm, then orbit manifold sensitive characteristics in the topological space are obtained. It provides a new method for fault diagnosis of electrical system. Application of the method to test-bed of normal, misalignment, rotor rub-impact fault diagnosis of three types, and fault diagnosis of NC lathe spindle and eccentric properly, the correct rate is up to 99%. The validity of the algorithm is verified.

Key words: orbit;LTSA algorithm;sensitive feature recognition;fault diagnosis