Application Of Fault Prediction And Diagnosis Technology In The Satellite Based On Rough Set

Fault prediction and diagnosis technology play an indelible role in spacesystems engineering, guaranteing spacecraft in the running stage with highreliability and safty, and lengthening the service of the aircraft. With satellite as thediagnosis object, rough set theory as the main line, grey model and support vectormachine as the forecasting model, case-based reasoning, fault tree analysis andbayesian networks as the diagnosis model, a variety of computer language as tools,this paper studies and develops a set of fault prediction and diagnosis systemapplicable for satellite.Before the systematic application of (neighborhood) rough set theory, thispaperexperimentally investigated the determining method of the neighborhoodradius of neighborhood rough set, and obtains the new determining principles andmethods. Finally, through comparative experiments, the validity of the principlesand methods is proved.This paper puts forward a forecasting method–combining with gray model,rough set and support vector machine. With the contrastive analysis of a variety ofthe grey system forecasting models, more suitability for satellite telemetry dataprediction of the metabolism grey forecasting model is validated. By using thecombined method of rough set and support vector machine (SVM), the predicteddata is categorized. By contrast, grid optimization method is verified to be mareapplicable to the prediction system.Case retrieval and similarity calculation and other key problems of case-basedreasoning are studied. In this paper, progressive retrieval strategies is selected, theanalyzation of which lead to the result that the application of delete-data approach insimilarity calculation is more accurate than others to deal with the data-missingsituation. A new computing method for attribute weight is defined by using roughset and information entropy, the results of which show that the method in this paperis more objective, universal and accurate.This paper carries out the contrast analysis of the qualitative analysis andquantitative calculation between the fault tree and the bayesian network model, andputs forward a fault diagnosis plan-using FTA to analyze qualitatively and BN tocalculate quantitatively. In order to achieve a more ideal diagnostic effect, this paperputs forward a kind of reduction way based on neighborhood rough set at the time oftranslating form fault tree to bayesian network model. The experiment indicates thatthe diagnostic effect of the fault diagnosis model of the network which is simplifiedwith this method is better. Combined with the actual demand, we develop a fault prediction and diagnosissystem for satellite by a variety of computer languages, and realize visualization,schematization, and friendly human-machine interaction of the diagnosis system. Inthe fault tree diagnosis module, this paper puts forward a new method of databasearchitecture which effectively reduces the redundancy of the fault tree databasebuilding, and implements a graphical mapping function of the fault tree through thesecondary development of Visio.