Satellite Attitude Control System Mutation Features Extraction Based On Support Vector Machine

At present, space technology is becoming one of the science and technologies that develop rapidly in today’s cutting-edge technology. Now it has played an increasing important role in many areas including political, economic, military, and science and technology. At the same time, the rapid development of space technology also makes the structure and function of the desired type of spacecraft systems become increasingly complex. Some spacecraft systems are very expensive; having a high reliability of the entire spacecraft system is the basic need. Satellite system is an important part.Once a part of satellite fails, the function may direct loss or even lead to some catastrophic event.At last it will lead to a waste of country property. Until now, the satellite system is usually to ensure its high reliability using hardware and software.This paper focuses on the satellite downlink data which is to extract the information of the current state in order to determine the satellite’s operational status.This paper analyzes the main part of the satellite attitude control system, dynamics and kinematics modeling, and the use of quaternions and Euler angles of two parallel ways to describe satellite state in detail. This paper uses the eigenvalues of empirical mode decomposition method to extract the signal, energy entropy signal threshold which is set to determine whether there is an exception occurs, and then it describes the main principles of support vector machine and how to optimize the two more effective parameters. The support vector machine is used in two ways to extract the satellite status information, the first one is a to obtain residual information from the observed values and use this information to analyze the state.This paper puts forward an optimized support vector machine for complex cases which effectively improve fitting effect and use less support vector number.This method effectively avoid over-learning. The second one is to extract symbol entropy and use the entropy values as classification symbol. Sign entropy can effectively differentiate among different states, and then classify them according tothe support vector machine algorithm. Comparing to existing methods, this algorithm can improve the classification accuracy.