Summary: | 碩士 === 國防大學理工學院 === 兵器系統工程碩士班 === 100 === The current weapon systems in military mostly rely on foreign procurement. Establishing the preventive maintenance strategy mainly relies on the maintenance handbook from manufacturer. Due to deviations of season and climate factors from foreign country and operating habit of users, generally, the handbook from foreign country is irrelevant to use and results in inaccurately predicting the availability of these weapon systems. Therefore, collecting the real failure data is essential to establish a preventive maintenance strategy of weapon systems. However, obtaining complete failure data is impractical, mainly because performing acceleration test is impossible or consumes numerous amount of cost. Most of the lifetime data obtained from maintenance belongs to the right-censored type. Therefore, establishing a procedure of parameters estimation for right-censored lifetime data of weapon systems can really benefit prediction of their availability in establishing the preventive maintenance strategy. This thesis applies the Monte Carlo simulation to generate failure data from the specific lifetime distribution initially. Then, via setting different observation times, various degrees of right-censored lifetime data are simulated using Matlab program. Accordingly, this thesis employs maximum likelihood estimate (MLE) to estimate the distribution parameters of various degrees of right-censored lifetime data. Furthermore, several simulated experiments with considerations of different distribution parameters and observation times are obtained to perform sensitivity analysis. From sensitivity analysis, the effects of various degrees of right-censored lifetime data on the estimation error of distribution parameters can be determined form which we can estimate the distribution parameters of weapon systems more accurately.
Keywords: reliability, Monte Carlo simulation, maximum likelihood estimate, response surface methodology, sensitivity analysis
|