Study On The ALT And Reliability Assessment Of The CNC System

CNC(computer numerical control) system reliability is the key issues of the CNC system development, that is also the main factors of domestically CNC system market share. The starting time of researching CNC system in China is later than the western developed countries, and the homebred CNC system reliability is also lower. The gaps between the homebred CNC system and the abroad CNC system is in gradually narrowing, but that is still no widely accepted and approbated by society. The researching of CNC system reliability, especially the reliability design, reliability growth and reliability evaluation is very urgent. Reliability assessment is the basis and premise for reliability ascension. Only finding out and verify the reliability level, it is possible to further improve reliability. So that it has great theory significance and utility value for researching of CNC system reliability assessment technique.Based on the Major national CNC special project "Research of CNC system reliability technology and heavy machine tool operation reliability" and the national natural science funds "Study on the CNC system reliability evaluation method applying Bayesian theory", the small sample and rapid assessment technique is discussed in this dissertation to seek the CNC system reliability assessment techniques and methods. The main content is as follows:1. With the collected failure data of a type of CNC system within a few years, the traditional reliability statistics analysis method was used to fit the fault interval time distribution model of the CNC system and hypothesis testing, which get a conclusion that using Weibull distribution can reasonable description the CNC system fault interval time distribution; Statistical analysis was done to the fault frequency and histogram was drawn, also the failure criticality analysis was done, so the weak links of the CNC system and the environment and working stress factors of influence the CNC system reliability were found out.2. The Bayesian theory was introduced to the CNC system small sample reliability assessment. The Bayesian sample evaluation model was established with using the failure rate as the random variables. Using the collected lots of the CNC system failure data to construct the prior distribution of Bayesian inference, and obtained corresponding parametric of the prior distribution. According to the likelihood function for the testing scheme, the posterior distribution of failure rate was deduced, further the other reliability parametric were calculated; A new statistical method based on the distribution type conversion was proposed and used to the Bayesian computation for Weibull distribution. At frist Weibull distribution was transformed into exponential distribution, then the exponential distribution Bayesian computation was done using the exponential distribution characteristics, thus the method avoided large integral computation, and the calculation efficiency was increased.3. Because the starting stress level is low, the efficiency of step-up-stress accelerated life testing(SUS-ALT) still not high enough, and the efficiency of step-down-stress accelerated life testing(SDS-ALT) although is raised, but the failure mechanism would be changed by high stress, so the self-adaption accelerated life testing(SA-ALT) method was putted forward based on SUS-ALT and SUS-ALT. Then the stress loading criterion and implementation steps of SA-ALT were introduced. The double side censored data elimination method was putted forward based on distribution type conversion. According to the method, the Weibull distribution would be converted to exponential distribution, and double side censored data eliminate method was adopted to avoid the ALT data conversion, so the statistical analysis efficiency was improved.4. Aiming at the accuracy and experimental cost, the optimization design model of SA-ALT test scheme called half traversal and half search method was putted forward using the variance of characteristic life and the failure sample size. In the method, some parameters values would be traveled through and some parameters values would be iteration searched, thus it would find out a optimum of the minimum characteristic life variance and the minimum failure sample size, and this parameters combination would be the optimal test scheme.5. The function and construction scheme of the CNC system reliability testing platform were introduced. The CNC system SA-ALT with a single stress of temperature was designed, and certain failure data of a kind of CNC system were got through a large number of accelerated life tests. the traditional method and the double side censored data elimination method based on distribution type conversion were adopted to the statistical analysis of those data, and the statistical inference results was correct and practical.