The Research Of MM-SVM Based Satellite Fault Diagnosis Technology

With the rapid development of China’s comprehensive national strength and aerospace technology development, satellite gets application in different fields. The utilization of satellite orbit satellites can be improved by extending the life of in-orbit satellite. Satellite fault diagnosis technology is one of the important means to achieve that goal,which has been got the attention of the researchers and rapid development in recent years.In this paper, the satellite fault diagnosis technologies were analysised and researched in-depth. The features satellite fault diagnosis includes real-time, high accuracy, low incidence,small sample size and numbers of parameters. Based on these considerations, we propose a satellite fault diagnosis method based on support vector machine(SVM).This paper proposed a multi-model support vector machines(MM-SVM) technology which based on the study of satellite fault diagnosis related technologies and support vector machine. The effectiveness of MM-SVM is tested through the existing real satellite data and it showed that the accuracy of MM-SVM reached 100 %, far higher than other methods. Specific work is as follows :(1)The current research of satellite fault diagnosis technology is analyzed in detail in this paper.We chose data-driven fault diagnosis technology as a research combined with the development of the current operating status of satellite monitoring technology.(2) The satellite fault data gets less with the increasing levels of satellite development, it makes the performance of some ways which based on the existing data-driven fault diagnosis technology being not as well as wished. This paper presents a fault diagnosis model based on multi-model SVM(MM-SVM). At first,we establish the appropriate SVM model fault diagnosis through sample study, then get the last results following the rule of "Most of the vote " for satellite operation fault diagnosis.(3)Ttraditional data warehouse technology is not suitable for satellite fault diagnosis timeliness requirements because of the satellite big data scale.We use the Hadoop platform for large-scale distributed data management and integrate data mining software R, designing and implementation of a distributed data warehouse HR system, and we builded the MM-SVM model fault diagnosis on this system.(4) The use of satellite data accumulated over the years in orbit and fault records state,it were tested based on the HR system, and compared with several other commonly used satellite fault diagnosis technology. The results show that MM- SVM makes higher accuracy and lower false positive rate compared to other fault diagnosis methods.