| The numerically controlled system, as the main control center of CNC machine, was called the soul and cerebrum of the machine. Its reliability relates the reliable level of CNC machine directly. Enhancing technical performance and the reliability of the numerically controlled system and speeding up the industry advancement play a vital role in competitive ability and expanding market share of domestic CNC products in home and the international market.This thesis is one part of national high-tech research and development plan (Plan 863) project "the strut research of appraisal for the reliability of CNC machine and the numerically controlled system ". By analyzing the failure data from the user, we try to establish one distribution model for the failure time, determine indicators for the reliability, find the weak points in existed CNC system and finally to provide the basis for the improvement research of reliability. A suitable comprehensive method for actual situation will be established to solve the problem that is only single indicator is used in evaluation of reliability. This method will be used as a criterion to improve the quality of products for the producer and also help the users to understand the quality of products well.We studied two kinds of the domestically produced numerically controlled systems. An analysis and evaluation method for reliability of CNC system was established by analyzing the data from failure of the machine at fixed time during using. This method is suitable to analyze the reliability of CNC system with smaller samples and lower failure rate. Based on the collected data of reliability in the CNC system and the established plan for the evaluation of reliability in the test, SAS statistics software was firstly used to analyze the distribution type of the data by drawing histogram and the QQ chart. According to small samples with or without non-failure data, we separately utilized maximum likelihood method or maximum likelihood method of improving method (IMML law) to estimate the parameters. Kolmogorov-Smirnov test was also used for testing normality. We found the first failure time of the two kind of CNC system conformed to exponential distribution (Values are 1/879.09 and 1/2426.267).Actual failure situation in the experiments of CNC system was simplified as one model, which may calculate. We proposed for the first time in the consideration of the censored data in the experiments. For the censored and uncensored data, the accumulation failure rate chart was used for primary screening of the distribution mode. Then maximum likelihood method was performed to estimate parameters of the distribution model. Finally, Hollander-Proschan test was used for the fitting examination. By the analysis, we determined the failure time interval of the two CNC systems obeyed the exponential distribution (Values are 1/1186.5 and 1/2307.14).For the first time, we used the method suitable for small samples, which is more precise in interval estimation. The interval estimation was performed on the MTTFF (Mean Time To First Failure) and MTBF (Mean Time Between Failure) of reliability based on Bootstrap's Monte Carlo simulation method.We statistically analyzed failure spots, failure patterns, failure reasons and classifications of failure reasons. The analysis indicated that the failure spots of hardware were often at numerically controlled panel, the detection unit and the motherboard, but the failures'spots of software were often at the real-time administration module, the initialization module and the pretreatment module. According to different characteristics of hardware and software, FMEA (Failure Mode Effects Analysis) and RPN (Risk Priority Number) were performed on the failure modes of hardware and software. The analysis results suggested the highest risk hardware failure pattern is the PLC unit detuning, but the software failure pattern is to zero error caused by the failure of pretreatment module. The results of PCA (Principal Component Analysis) indicated that the dominant failure reasons were wear, aging, primary device damage, sudden power failure and virus infection.For the first time, we used the entropy weight and fuzzy comprehensive evaluation method to comprehensively evaluate the indicators of reliability in two kinds of CNC systems. We chose MTTFF, MTBF and equivalent failure rate as indicators of CNC systems'reliability. All these indicators can reflect inherent reliability, use reliability as well as the failure influence of the system. We also scientifically determined indexes weight by using comprehensive evaluation obtained from combining a subjective expert's evaluation weight with objective entropy weight, thus comprehensively evaluated the reliability level of two studied CNC systems. The reliability evaluation results of two CNC systems indicate that the reliability level of domestically produced CNC systems needs improvements.The work in this thesis can effectively solve the problem existed in current modeling and evaluation of reliability and has the great practical value in the application.
|
PDF Full Text Request