Fault diagnosability evaluation for spacecraft control systems via direction similarity

To improve the fault diagnosis performance in the system design stage, a method for fault diagnosability evaluation applied to spacecraft control systems is proposed in this paper. The method can be used to guide fault diagnosis algorithm design and optimize sensor placement. The existing problems for quantifying fault detectability and fault isolability based on Kullack-Leibler divergence (K-L divergence) are presented through rigorous proofs. To solve these problems, induced norm and cosine similarity are adopted to design the novel quantitative indexes of evaluating detectability and isolability, respectively. And then a detailed evaluation procedure is provided. Finally, a spacecraft system, which is composed of infrared earth sensors and gyroscopes, is exemplified to show the superiority of the method mentioned in this paper over the method utilizing K-L divergence.