Spacecraft Fault Diagnosis In Space Environment Based On Artificial Intelligence

Space environment is complex and changeable,which is one of the main reasons for causing spacecraft operation failures or anomalies.With the development of space technology,the level of spacecraft automation is getting higher and higher,so failure of spacecraft in operation environment will cause unpredictable loss.Therefore,how to make intelligent and effective fault diagnosis for spacecraft has become a hot topic in this field.However,spacecraft faults in space environment are complex and diverse.The traditional method of spacecraft fault diagnosis is limited to the knowledge and experience of experts and relevant technicians,so it is unlikely to carry out spacecraft fault diagnosis in time and effectively.Therefore,this paper studies the spacecraft fault diagnosis methods based on artificial intelligence technology in space environment.1)To begin with,we propose a Back Propagation(BP)based fault diagnosis method for spacecraft in space environment.This method establishes a complex mapping relationship between the environmental data of spacecraft and the probability of abnormal events in the space environment.The BP neural network is trained by the collected environmental data of the spacecraft,and the trained network is verified by test data.2)Furthermore,we propose a fault diagnosis method for Spacecraft Based on Deep AutoEncoder Network(DAEN)in space environment.First,the method uses pre processed spacecraft environment data to train deep self code neural network,extracts the original environmental data through the trained model,extracts the characteristics of the environmental data to complete the fault classification of the spacecraft,and then makes use of the random forest algorithm to analyze the importance of the spacecraft environmental data,and finds out the space loop.The key environmental impact factors that trigger spacecraft failures are obtained from the new environmental data feature subset,and the fault diagnosis method based on the deep self coding network is used again to complete the spacecraft fault diagnosis.The simulation results are as follows:For the spacecraft fault diagnosis method based on BP neural network in the space environment,this method can achieve higher prediction accuracy in spacecraft fault diagnosis.For the spacecraft fault diagnosis method based on the deep self coding network space environment,on the one hand,it uses the classifier network(Softmax)to diagnose the spacecraft fault,and uses the method.The accuracy of classification results obtained by the method is improved.On the other hand,the fault diagnosis of spacecraft is carried out again by using the feature subset of environmental data obtained from the importance analysis.The performance of the method is further improved.