Research On The Method Of Storage Reliability For Torpedo

One of the characteristics of torpedo is long storage time before use for only one time.On the whole life cycle, most of the time is in the storage state. The torpedo storage reliabilityis one of the torpedo reliability index.The traditional reliability theory must be based on the failure data of the product.However, for the torpedo system, it is difficult to obtain enough failure data for the most partsof torpedo, and even there is no failure data. It is difficult to obtain the valid results based onthe traditional reliability theory in this situation. With the deepening of the research onproduct failure mechanism. Many failures of products are caused by performance degradinggradually can be found. Performance degradation information obtained in storage/operationalprocess contains a great deal of useful reliability information. No matter whether there isfailure or not, the reliability of the product can be analyzed according to the performancedegradation data. Therefore, the use of the performance degradation information provides us anew way to evaluate the storage reliability of torpedo. This dissertation studies degradationanalysis approach systematically combined with the engineering background, and builds somegeneral and strong adaptive degradation models which are applied in a generalizedcomponent’s reliability evaluation. The main contributions of the dissertation are summarizedas follows:(1) Based on the engineering background of torpedo, the models of monotounous partsand non-monotonous parts are builded.First, the characteristics of torpedo storage reliability, effect factors and the analysisprocess of torpedo system are given. And then, the model of performance degradation part isbuilded, which include monotounous parts and non-monotonous parts. At last, according tothe real example, it is show that the assessment method based on degradation data is moreefficient than that based on the failure data.(2) Research on the storage reliability evaluation method of torpedo parts based onperformance degradation method.According to the performance degradation informance of torpedo parts in the storageprocess and the stochastic process, the degradation process was modeled. At the same time, inthe case of inverse Gaussian process, based on the different prior information, the real-timereliability evaluation framework combine with Bayesian theory is be obtained.The Linex asymmetric loss function can flexible to weigh the overestimation and underestimation is adopted. Therefore, it is more appropriate to evaluate the torpedoreliability than quadratic loss function.(3) Research on the development of torpedo parts storage reliability model by combiningmeasurement errors with fuzzy failure thresholdConsidering both the performance data with detecting error and the fuzzy failurethreshold, we establish the storage reliability model. And then, an efficient method forestimating the prarameters of model based on a novel use of the Genz transform andquasi-Monte Carlo method in the maximum likelihood estimation is presented. Examplespresented show that the proposed method is very efficient compared with the Monte Carlomethod.(4) Research on the development of two-phases degradation model for the torpedo partsin the storage situationBased on the phenomenon that some parts have different degradation rates in thedifferent phase, we model the two-phases degradation model. And then, in the case of inverseGaussian process, the maximum likelihood estimation and real-time Bayesian estimation areobtained. At the same time, the evaluation of change-point based on the Schwarz InformationCriterion (SIC) and the criterion of the residual sum of squares is obtained.(5) Develop a s-dependent competing risk model for torpedo parts subject to multipledegradation processes and random shocks.First, the multiple degradation processes model is developed; the random shockdegration is be considered further. And then, a s-dependent competing risk model isdeveloped. At last, the maximum likelihood estimation according to Markov chain MonteCarlo (MCMC) and three steps maximum likehood estimation are obtained.