Reliability Evaluating Technology Of CNC Machine Tools Based On Repair Quality

CNC machine tools is the key fundamental equipment for realization ofmanufacturing technology and modernization of manufacturing equipment, whoseperformance, quality and ownership is a measure of a country’s level of industrialmodernization and an important symbol of the comprehensive national strength. Thereliability of CNC machine tools is the guarantee to achieve its performance and createeconomic benefits. Quantitative reliability assessment of the machine tools is the majoractivity in reliability engineering. The machine reliability assessment technique hasmade significant achievements over recent years, but the problem persists. Mainlyreflected in:(1) Methods for non-repairable products and repairable products such as lifedistribution and stochastic process, cannot be distinguished and often misused;(2)Alarge number of studies have focused on the solution of the model parameters and buildnew models, while ignoring the study of the application in engineering practice for theexisting models;(3)The calculation and interval estimation of reliability assessmentindices are still shortage;(4) There is little intensive research dig into the impact onmachine reliability from repair quality, and existing models are too mathematicalcomplex.Aiming at the problem mentioned above, the following work is done, which basedon the support of National Major Scientific&Technological Special Project and fromin-depth research on CNC machine tools reliability evaluation oriented to repair quality:(1) Study is carried out on the basis concept of the reliability assessment and itsmetrics. A comparative analysis of the inherent reliability and field reliability, missionreliability and basic reliability is made, which help to set a clearly defined scope of"reliability" for this thesis. Meanwhile, this paper explored and discussed the basicmeaning of MTBF and failure rate, made an intensive analysis in MTBF, gives a moresuitable definition for machine tools (repairable products).(2)A methodology for reliability estimation of multiple machine tools withminimal repair based on power law process is proposed, and failure data functionestablished. CBT and CLT two-step-test is developed to check the homogeneity ofmultiple machines in assessing. Point and confidence interval estimators are given bymaximum likelihood method and Fish information matrix, which of reliability are alsoderived through Delta method. (3) In order to solve reliability evaluation with minimum repair for small sample, aBayes method based on Gibbs theory to estimate PLP parameters is proposed. TheGelman-Rubin method is adopted to confirm the convergence. Monte Carlo-Error isalso calculated. By comparing the length of Confidence interval and value of standarddeviation, we concluded that Bayes method outperforms MLE for small sample.(4) Based “virtual age” concept, Kijima generalized renewal process is proposed toassessing machine reliability with imperfect repair. Nonlinear optimization is used tosolve model parameters estimating. MC simulation method is applied to obtain thereliability index point estimates and confidence intervals, since there are no closedexpressions for these metrics. Derivation and examples analysis demonstrate thatKijima model is a generalization of Weibull renewal process and power law process. Byintroducing a degree of repair factor, Kijima generalized renewal process is able toconsider repair quality more effective, thus the assessment results more closer to thereal work of CNC machine tools and their repair activities.(5) An approach based on generalized proportional intensity model (GPIM) isproposed for the first time to overcome difficulties in reliability estimating underimperfect repair using Kijima models, and a GPIM function for failure times is built.Point and confidence bounds estimates of model parameters are given by Maximumlikelihood and Fisher information Matrix method, which of reliability metrics are alsoderived and offered through Delta method. Likelihood ratio tests are adopted to checkthe time trend and repair effect.(6) In order to overcome the limit for GPIM that is repair effect index0, anapproach based on Log-linear proportional intensity model (LPIM) is proposed. Thispaper investigated the meanings of parameters and the properties of the model. A LPIMfunction for failure times is built. Point and confidence bounds estimates of modelparameters are given by Maximum likelihood method and Fisher information Matrixmethod, which of reliability metrics are also derived and offered through Delta method.A real example analysis demonstrates that, LPIM approach outperforms perfect repairand minimal repair methods considering the effects of repair activity, capable to capturequantitatively repair effects on machine tools reliability, hence the instantaneous andcumulative reliability properties obtained more practical.