LAUSR

Flywheel bearing is a key mechanical part of a satellite. Its health plays an important role in the fatigue life of the satellite. However, it is rather difficult to diagnose the health state of the bearings due to the complex satellite system. This paper attempts to propose a three-direction correlation dimension method to diagnose the bearings of a satellite flywheel at three typical states based on K-medoids clustering technology. A set of spatial spheres representing different bearings are modeled to recognize these three states of the bearings. To avoid misdiagnosis or loss of the bearing state, a twice-cluster scheme is employed. A series of tests is carried out to observe the effectiveness of the proposed method. The result shows that the proposed method is capable of diagnosing the different states of the bearings and its accuracy is higher than 99% at given conditions.