Development Of Fault Prognosis System For Marine Power And INS Components

The Marine power supply and inertial guiding components are important guarantee for the safe and reliable navigation of the ship,which can bring enormous potential safety hazards once they fail.Therefore,it is of great research value and application value to ensure the reliability of Marine power supply and inertial navigation components,to make fault prediction research on them,and to make "maintenance according to the situation" for potential faults.Th is dissertation takes a ship-type power supply and inertial navigational component as the research object,and focuses on its failure prediction technology,and carries out research work including failure mode analysis,feature parameter extraction,status assessment and parameter prediction technology.Based on the joint test platform,developed including data acquisition module,data preprocessing module,state evaluation module,fault prediction module of software fault prediction,The hardware equipment is selected and the complete fault prediction system is built in conjunction with the fault prediction software.Specific research contents are as follows:For marine power,first analyze the failure mode.In order to solve the problem that early failure of power supply is difficult to find,a wavelet analysis is used to extract feature parameters and a method of combining Hidden Markov Model(HMM)and KL distance is used to perform state assessment.Turn early power failures that change slowly into apparent KL distances to discover early power degradation trends and achieve power state assessment.Aiming at the problem that the power supply needs to be predicted when the health state of the power supply deviates to a certain extent,but is not enough for maintenance,a state prediction method based on the combination of limit learning machine(ELM)and HMM is proposed to realize the state prediction of the power supply.Combined with the actual data,the method is proved to be effective.For gyroscope in the Marine inertial navigation system,first of all,analyze its failure mode,and propose a view that put random drift data as fault feature parameters.According to the characteristics of randomness,non-linearity,nonstationarity and weak time-varying of random drift data,a trend prediction method based on grey correlation vector machine is proposed to improve the prediction accuracy of random drift.In order to solve the problem of poor accuracy of long-term trend prediction by correlative vector machin es(RVM),a dynamic updating of RVM model by using grey correlation analysis is proposed to realize long-term trend prediction of random drift.Combined with the actual data,the method is proved to be effective.According to the demand analysis of fault prediction system,the overall structure of fault prediction system is designed.Based on the joint test platform and the fault prediction method proposed in this paper,the components needed for the fault prediction system are designed.Complete component development based on Qt and Matlab.Hardware equipment needs to be selected,and the software integration developed in this paper is integrated into a complete fault prediction system.On the joint test platform,we build a test plan,and we finish the uni t test,system integration test for the failure prediction system.Using the data provided by the cooperative unit of this project to complete the system test,the test results show that the development of the fault prediction system for ship power supply and gyroscope is effective and practical.