The Test Design And Evaluation Method Research Of Reliability Project For CNC Machine Tools

Reliability test is an important foundation for reliability study that includes obtaining fault information, conducting reliability analysis and evaluation, and implementing reliability improvement. Reliability experimental designs, whether reasonable or not, will greatly affect the smoothness of follow-up reliability work. Therefore, research on reliability tests and evaluation techniques of CNC machine tools has important theoretical and practical significance for promptly and accurately determining the weak links of CNC machine tools, assessing their reliability levels, and selecting reasonable measures to implement reliability improvement.Based on the results available on reliability test research and the characteristics of CNC machine tools, this paper studies reliability test designs in the following four areas: fixed-time censoring scheme, fixed-number of failures censoring scheme, sequential progressive experiments, and Bayesian reliability tests. Considering manufacturer-side risks, user-side risks, identification ratio, MTBF expectation value, and other information, we conduct reliability experimental designs and propose data analysis and reliability assessment methods for different testing designs.In this paper, based on the characteristics of CNC machine tools'life following the Weibull distribution, we first transfer the Weibull distribution to the exponential distribution through a parameter transformation, combine with the exponential distribution decision-making risk equations, determine the fixed-time censoringand fixed-number test designs for CNC machine tools, and implement the designs in the manufacturers. By considering the impact of truncated time on assessment results, the paper analyzes the fault data using the total failure time method, conducts abnormal data testing using the "average and gradually take over test method" , solves empirical distribution functions based on the "residual ratio method", builds reliability models according to the trend in the fault data, verifies whether the proposed designs are reasonable or not and assesses their reliability levels by comparing the estimated shape parameter in the reliability models with the shape parameters selected initially in the fixed-time design and in the fixed-number design. Second, based on the probability of number for CNC machine tool failures under the Weibull distribution and combining with the traditional sequential experimental design approach, we determine sequential experimental designs for the CNC machine tools. As a decisionin a sequential experimental design may be delayed for long time because the failure points may still be in the testing period, we design truncated sequential experiments for CNC machine tools by increasing truncated failure number and truncated testing time, and modify the acceptation line and rejection line to meet the constraints of decision-making risks. Since a sequential test may be ended too early and result in insufficient failure information for assessing the true reliability value accurately, in this paper, we use the dynamic censoring method for reliability assessment of CNC machine tools based on the idea of sequential experiments, dynamically collect and assess fault data, and stop the experiment when the change of the reliability assessment index is within the tolerable error range.Third, based on the prior information for the CNC machine tool failure rate, we introduce the Bayesian theory into the reliability test research and determine Bayesian fixed-time censoring experimental designs by calculating the Bayesian posterior risks. The fault data collected in the fixed-time censoring experiment in chapter 2 will be applied to the Bayesian experimental design. Since the testing time of a Bayesian experimental design is relatively shorter and results in smaller number of fault data points, we adopt the parameters bias correction method to modify and verify the reliability model built based on the fault data, and further evaluate the reliability level.Fixed-time censoring and fixed-number censoring experiments are able to make the right decision, obtain enough fault data information in the experiment, and accurately assess the reliability index. In addition, fixed-time censoring scheme determines the total test time and number of machine tools before the experiment, which is easy to control the trial process and estimate the cost and resources of the experiment; fixed-number censoring experiment determines the maximum number of failures before the experiment, which gathers enough failure information and provides a reasonable basis for the failure analysis. The average test time of a truncated sequential test is shorter than that of fixed-time censoring and fixed-number censoring experiments. When we only need to decide whether the reliability level of CNC machine tools is satisfied or not and do not need to assess the reliability index, selecting the truncated sequential experimental design is more appropriate. If the manufacturers need to make judgments urgently and do not want to increase the decision risks, fixed-time censoring experimental designs based on the Bayesian method should be selected if prior information of the failure rate could be obtained accurately.