METHOD OF SATELLITE FAULT DIAGNOSIS BASED ON ROUGH SET

The fault diagnosis of satellites is a difficult problem due to the complex and unique of structure and control of satellite and the presence of multi-excite sources. Generally, one satellite fault mode is relative to many symptom variables. These variables comprise a set, called the primary symptom set of the fault mode. If each variable in the primary symptom set is necessary to the fault diagnosis? The answer is "no".That is, some variables in the set are redundancy variables. Current reasoning method of fault diagnosis uses all of the variables of primary symptom set to diagnose fault. Because of the redundancy variables in the set, the reasoning method is usually complicated. The redundancy variables not only give no contributions to the fault diagnosis, but also affect the accuracy of the fault diagnosis. So, the methods that can remove the redundancy variables in the primary set should be found. Rough set is a kind of advanced uncertainty reasoning theory. It has several advantages. One of them is to be used to simplify and optimize the sets, which have redundancy variables. Based on Rough set, a new method of satellite fault diagnosis is proposed in this paper. This method removes the redundancy variables in the primary symptom set of satellite fault mode firstly. The remainder variables comprise a set, called the reduction symptom set of satellite fault mode. Then, on condition that the diagnosis accuracy is guaranteed, the proposed method only uses the variables in the reduction symptom set to diagnose fault. The diagnosis reasoning is relatively simpler, because it uses fewer variables to diagnose fault. The content of this paper is organized as following: Firstly, the shortage of current methods of satellite fault diagnosis and the necessity of applying Rough set to satellite fault diagnosis are analyzed. Secondly, the theoretical method of Rough set for diagnosing satellite fault is briefly studied. Thirdly, the practical application of the method is given. Finally, the proposed method is discussed and the conclusions of this paper are given. It is shown that the method proposed in this paper is effective for satellite fault diagnosing.