Research On Modeling Key Technology Of Machine Learning Methods For Dynamical Adaptation Of Satellite Fault

The satellite is an high-complex system.The satellite is expensive and the maintence cost is high,which is of great strategic importance to the country.Because of its special working environment,satellites are easily disturbed by various factors.If the satellite fails to deal with the problem in time,it may cause a satellite accident and cause irreparable damage.When satellite failure occurs,it is frequently along with the change of the parameters.Therefore,it is very important to carry out relevant fault diagnosis and modeling research on satellite data,so as to realize the monitoring of satellite status and reduce the possibility of satellite failure.The main research contents and contributions of this thesis are as follows:(1)The main research object of this thesis is the measured data from satellite which provided by the partner.Through the analysis of the characteristics of the telemetry data briefly,this thesis puts forward the data preprocessing steps,which includs removal of outliers,interpolation compensation.Besides,the quantitative determination index of model precision is illustrated.(2)In the present thesis,an improved particle swarm optimization algorithm is been proposed to optimize the wavelet neural network for fault diagnosis of satellite telemetry data.The model can diagnose faults accurately.In order to control particle swarm optimization's disadvantages,an improved particle swarm optimization algorithm is proposed in order to optimize the parameters in the wavelet neural network.This optimization is able to achieve self-adaptive diagnosis and more accurate.(3)In this thesis,an improved ant colony algorithm is used to optimize the radial basis neural network for fault diagnosis of satellite telemetry data.The model can diagnose faults well.For ant colony algorithm,a new off state is proposed for ants to improve the ant colony algorithm can get into local optimal situation.Through adaptive adjustment of parameters,it can make the diagnosis more accurate.(4)The data is processed and revelent information is extracted which is provided by the data provider.According to the algorithm designed in this thesis,the diagnosis of satellite fault line is realized and analyze the test results.(5)According to the algorithm used in this thesis,I develop a set of verification software for satellite fault diagnosis.Above all the aspects,based on the simulation of satellite telemetry data,this thesis proposes two methods of satellite fault diagnosis,which can realize the diagnosis of satellite fault.