Using RBF Neural Network for Fault Diagnosis in Satellite ADS

In this paper, a new hybrid learning strategy composed of K-means clustering algorithm and Kalman filtering is employed to train radial based function (RBF) neural network for fault diagnosis in satellite attitude determination system. Because Kalman filtering and K-means clustering algorithm both adopt linear update rule, their combination produces a new hybrid training algorithm that can converge quickly. Simulation results demonstrate that the proposed approach is effective for fault diagnosis in satellite attitude determination system.