Study For Reliability Evaluation Of NC Machine Tool Based On Survival Analysis And Likelihood Theory

This paper studies reliability evaluation on a class of censored life test, namely the MTBF(mean time between failures) studies. The study finds that NC machine tool failure is mostly subjected to the Weibull distribution, and quite a part of failure data are censored data. Furthermore, there is no mature method and standard to evaluate the reliability of the censored data that is subjected to Weibull distribution, at present at home and abroad.Firstly, the paper discusses censored ways of NC machine tool reliability test, combines with operation mechanism of NC machine tool reliability test, draws run charts of NC machine tool reliability test, researches simulation method of NC machine tool reliability censored life test with stochastic simulation method(Monte Carlo method). The method lay the foundation for reliability evaluation based on Monte Carlo method.Then, Fixed Time Censoring with No Replacement test model(number of samples from 50 to 3000) and Fixed Time Censoring with Replacement test model(10 test machine tools, test time from 2000 hours- 12000 hours), applies the maximum likelihood method to MTBF value point estimation and interval estimation based on Monte Carlo method, and studies the evaluation method of "rejecting true probability" and "accepting false probability". According to Professor Chen Jiading of consistency theory with censored data, analyzes the results of the study, combines with NC machine tool reliability assessment of practical engineering, probes into the approximate range of "rejecting true probability" and "accepting false probability" in current assessment methods for reliability engineering.Third, Fixed Time Censoring with Replacement test model(50 test machine tools, test time from 1000 hours- 6000 hours), applies the maximum likelihood method to MTBF value point estimation and interval estimation based on Monte Carlo method, and studies the evaluation method of "rejecting true probability" and "accepting false probability". And then likelihood method and survival analysis method for comparative study, interesting and meaningful results are obtained. On the one hand, it confirms Professor Chen Jiading’s conclusion in small sample(sample size between 10 to 30), also has some supplement for large sample(sample size thousands).Fourth, in order to the further application of rich survival analysis theory, this paper discusses model transformation between Random Time Censoring test model in survival analysis theory and Fixed Time Censoring with Replacement test model closer to engineering practice.