Research On System Degradation Modeling And Reliability Assessment Methods

With the continuous development of high technology and constant application of new materials and crafts,high integration,high intelligence and complexity of products/systems are becoming more and more dominant.Under such backdrop,traditional reliability assessment methods based on failure data are faced with new realistic difficult problems such as small-sample,longevity,zero failure and complex correlation.As a key technology in reliability design,test and assessment of highly-reliable and long-life products,reliability modeling and assessment technology based on performance degradation are currently the research focus and hot spot in the reliability engineering area.Compared with failure data,performance degradation data which can be obtained by real-time measuring during product mission phase,are closer to product failure mechanism,contains corresponding relations between physical quantity,performance and lifespan,and provides abundant information sources for reliability assessment.Aiming at specific problems in engineering practice,the dissertation focuses on the dynamic threshold system,single-stage system with calibrations and multi-stage system with calibrations.Basic theories and methods of stochastic processes are adopted in the dissertation to conduct the modeling research in system degradation process,and reliability assessment methods based on performance degradation are put forward.The main research contents of the dissertation are as follows.Firstly,system degradation models based on Wiener diffusion process are built in the situation of three kinds of dynamic thresholds i.e.linear threshold,interval threshold and random threshold,and the corresponding reliability assessment methods and calculation methods for relevant indices are also presented.Essential lemmas for linear and curve threshold reliability calculation are put forward and proved in the situation of linear threshold.Reliability assessment methods and procedures based on Monte Carlo simulation are put forward aiming at special failure intervals such as round,rectangle,oval and multi-dimensional intervals in the situation of interval threshold.In the situation of random threshold,corresponding reliability calculation methods are put forth based on Monte Carlo simulation in terms of the uncertainty of drift parameter and failure threshold.Secondly,degradation process models based on Wiener diffusion process for single-stage system with calibrations are developed and reliability assessment methods are presented.Two degradation models are built in accordance with the classical Kijima model and conversion relationship between models is proved.A smooth function is established to solve the problem of discontinuity of drift function resulted from pre-specified calibrations,new distributed solutions of system reliability with first passage time and traditional solutions of partial differential equations with absorbing boundary conditions are put forward.Functional relationship between indirect observational signals and system direct states is built,and problems of degradation modeling and reliability assessment for system under calibrations with the indirect observational data are solved.Thirdly,multi-stage system with calibrations degradation data models and process models are developed from the point of degradation data statistical analysis in the situation of definitive failure threshold and random failure threshold,and reliability assessment methods of multi-stage system with calibrations based on Bayesian are put forward.Estimated value and posterior distribution of modeling parameters in each system operating stage is solved by adopting MLE and Bayesian methods,and conjugate distribution,modeling parameter variance monotonic decreasing and other theorems are proved.Aiming at random failure threshold,the method of reliability assessment for multi-stage system with calibrations are put forward based on Monte Carlo simulation,and the reliability influence rule of randomicity is validated by comparing and analyzing the reliability assessment results in the two different situiations.Fourthly,the reliability assessment sensitivity of degradation system based on OAT and probability density function is conducted.On the one hand,taking degradation model for single-stage system with calibrations as an example,the sensitivity of calibration degree,drift coefficient,initial degradation value and diffusion coefficient is analyzed based on OAT method.On the other hand,taking Uniform distribution and Beta distribution as an example,the sensitivity of initial degradation value obeying different probability distributions is studied.Finally,an engineering application case study is conducted.With a certain type surface ship Platform Inertial Navigation System as an example,a system degradation process model is built and the reliability of Platform Inertial Navigation System is assessed during characteristic forming stage,delivery stage and service stage by adopting reliability modeling and assessment methods for multi-stage system with calibrations,thus effectively verifying the validity and practicability of the model and the methods.This dissertation is of important theoretical and practical significance and its relevant theories can be applied to solve difficult problems concerning reliability assessment in complex systems such as dynamic threshold system,single-stage system with calibrations,multi-stage system with calibrations and other projects.The study perfects the system degradation process models,enriches reliability assessment theories and methods based on performance degradation and enjoys broad popularization prospect with its great engineering application value.