Research On Bayesian Reliability Assessment For NC Machine Tools Based On RBF Neural Network And Nonlinear Constraint

Reliability evaluation technology is one of the key technologies of numerical control machine tools,especially high-grade heavy-duty CNC machine tools.The Bayes method using the prior information to supplement the field test information is the most important methods to solve the problem of reliability evaluation for a heavy CNC machine tool,which is difficult to solve because of the lack of field fault data based on the traditional probability statistics theory.In this paper,the blindness of the determination of prior distribution of Heavy CNC machine tools reliability evaluation based on Bayes theory is pointed out is targeted and research on reliability evaluation of CNC machine tools based on Bayes method is developed taking single and small number of heavy-duty CNC machine tools as the research object.The main research work is as follows:(1)In order to solve the problem that similar machine tool fault data distribution type being not unique leads to the traditional Bootatrap method can not determine the prior distribution,by analyzing and pointing out the reasons why the data distribution type of machine tools is not unique,a method based on RBF neural network extending the fault data to determine the distribution type was proposed,the accuracy of which was verified.On this basis,in order to reduce the sampling error,a prior distribution was established by using the parameter Bootstrap method.A method of determining the prior distribution based on RBF neural network and parameter Bootatrap was proposed.(2)In order to solve the problem that the reliability of the prior information calculated by the KL distance method may differ greatly from the actual truth value,the irrationality of KL distance method to determine the probability density of field fault data was analyzed and pointed out and a method,which can reduce random factor interference because of considering randomness of probability density of field fault data,for computing prior information reliability based on nonlinear constraint-KL distance was proposed.The effectiveness of new method was verified by an engineering example.(3)The method of prior distribution construction proposed in this paper was applied to evaluate the reliability of CXK××× milling machining center based on Bayes method and the probability distribution of the CXK××× fault interval and the mean time between failures(MTBF)were obtained.The evaluation results of the Bayes method in a small sample of a machine tool and the results of the probability statistics theory of a large sample of this machine tool were compared to illustrate the accuracy of the method proposed in this paper.