Research On Remaining Useful Performance Prognostic Method For Analog Circuits Based On Degradation Feature Analysis

With the development of electronic engineering technology,various analog circuits are playing an increasingly important role in electronic systems.The structures and functions of electronic systems is becoming increasing complexity,and the reliability requirements are also increasing.Therefore,to implement the ideal maintenance based on condition and avoid the waste of logistic resources,there is an urgent need on fault prognostic.At present,prognostics and remaining useful life prediction in analog circuits have become the focus of the PHM(Prognostics and Health Management: PHM)research technology.Due to the limits of physics,chemistry and materials science research,PHM technology research is still in the exploratory stage and facing many challenges.In view of this,this paper addressed the core problems of circuits degradation feature extraction and circuits prognostics in the PHM technology of electronic systems.The studies start from the following aspects: the feature extraction problem from time domain output signal,the real-time reliability prediction problem for individual circuits or electronic products,and the intermittent fault prediction problem of electronic systems.To solve the above-mentioned problems,this paper proposed a series of study and exploration,which include the noise-assisted technology prediction methods based on Kalman filtering,the combination of noise-assisted technology and Bayesian estimation methods for real-time reliability prediction and the spline-based with weighted Hidden Markov prediction methods.This research is trying to promote the practical application of PHM technology in electronic systems.The main innovations of the paper are as follows:1.Facing the difficulty of feature extraction problems of the output signal in analog circuits degradation process,a prognostic method for analog circuits based on noiseassisted technology is proposed.The Gaussian white noise estimation method based on Kalman filter is used to extract features which can characterize the degradation trend of the circuits.The deviation from the normal output signal is used to form the FI(Fault Indicator: FI)of the circuit performance degradation.The particle filter algorithm is used for the RUP(Remaining useful Performance: RUP)prediction of the circuits.Two examples of analog circuit filters validated the prognostic algorithm proposed in the paper.2.For the requirements of real-time reliability prediction of an individual circuit or electronic product,a real-time reliability prediction method based on noise-assisted technology and Bayesian method was proposed.According to the historical data of the similar products,the performance distribution parameters of the electronic products are estimated as the priori information for prediction.The estimated pseudo-remaining useful performance could be extrapolated based on the abovementioned noise-assisted technology.The information fusion technology based on Bayesian estimation method with the data collected on-site could update the timevarying parameters,which is used to predict the real-time reliability.The POF(Physics of Failure: POF)model of the capacitor corresponding to the accelerating degradation experiment of the embedded planar capacitor was introduced here.By replacing the ideal hypothesis data with real data,the effectiveness of this real-time reliability prediction method was verified.The results showed that the more information,the higher accuracy of circuit reliability prediction.3.For the prognostic problem of intermittent faults in electronic systems,the singularity signal was assumed to be a mathematical way to describe the intermittent faults in the degradation of electronic systems.An ensemble method based on spline function and PSO(particle swarm optimization)-based WHMM(Weighted Hidden Markov Model)were proposed in the paper.This ensemble method used cubic nonpolynomial splines as a new surrogate model.For singularity disturbances problem in degradation data,the second derivative of the spline function model could form a series of observation sequences.The PSO-based WHMM was conducted here to predict the corresponding observation sequences,in this prediction process,the updated parameters would rebuild the spline function which would forecast the degradation state.A numerical simulation example and a practical example involving singularity problems demonstrated the effectiveness of this proposed method.4.The noise-assisted technology mentioned in the paper was applied on a real fuse control circuit system.By introducing the FMEA(Failure Mode and Effects Analysis: FMEA)method,an entirely reliability analysis of the delay circuit module in this system was systematically performed on the components.This research completed the system analysis of fault simulation analysis,fault injection and simulation method research,formed a complete electronic circuits functional fault simulation analysis procedure,and established a FMEA information list of the functional module.The actual FMEA analysis results here verified the proposed noise-assisted technology in this paper.The results demonstrated that this noiseassisted method can effectively predict the RUP of the circuit systems.Compare to the traditional methods,the new method could improve the prediction accuracy and the system reliability assurance level.This research method could be also applied to other electronic systems or equipments,it has great potential and can play an important role in PHM technology.