MTBF Evaluation For CNC System Based On Small Sample Data

Traditional reliability evaluation method needs a large sample of failure data to ensure the accurate of evaluation results, but it is difficult to meet the requirements through the reliability test actually. Therefore, this paper will study reliability evaluation method of CNC system based on small sample of failure data. In fact, the basic purpose of the study is estimating the mean time between failures (MTBF) by small reliability sample data of CNC system. The VMC850E CNC system is chosen as the evaluation object, and the reliability data has been collected already.By analyzing the field data of the VMC series CNC systems, there are some problems as following that make the data can’t used to evaluate the MTBF directly. Firstly, some records are not objective and accurate because of irresponsible behaviors of data recorders. Secondly, some failure records in the list are not related with the machine at all, and some records even can’t be seen as failures. The last but the most important problem is that the field data were all recorded by calendar time, they cannot be used directly in evaluation no matter from statistics or unit, and a further transformation processing for it is very essential.For the problems shown above this paper gives specific solutions, and completes the data preprocessing of the VMC850E and VMC650E CNC systems. What’s more, the total failure time method is used to process the data of VMC650E system with lots of censored data. After the data pretreatment, the distribution model of CNC system failure interval time can be obtained, and some methods are put forward to solve the problem of reliability evaluation with small sample data based on the model.In the process of the evaluation based on the small sample data, this paper raises two different methods. The first one is on the basis of regression convert method that converts the failure data of the similar products with the one to be evaluated. And then analysis the distribution model of the time between failures by the data expanded. Based on the above research, the work of reliability evaluation could be completed with ease. The second method that brought forward to solve the problem of reliability evaluation with small sample data is based on the Bayes method. Firstly, the distribution model parameters should be put as random variables, and choose appropriate prior distributions for them by their physical meaning or experiences. Secondly, determine the super parameters of the prior distributions with their priori information, and derive the posterior distribution with the small sample data. Hence, the MTBF could be calculated.In this paper, the two methods above is used to evaluate the reliability of the VMC850E CNC system, and the deviation of MTBF obtained respectively is less than5%which meets engineering requirements. At the same time, the evaluation results also prove the effectiveness of each method. These studies provide effective methods for the reliability evaluation of CNC system based on small sample data, these methods could be good reference for the evaluation in other areas.