Reliability Evaluation Of CNC Machine Tool Based On Bayesian Network

Computerized numerical control(CNC)machine tool is a complex overall system.There are many electronic components and mechanical components in its subsystem,and the subsystem must maintain high coordination and stability when working.Due to the properties and functional characteristics of the above machine tools,the corresponding reliability evaluation is faced with the characteristics of complex fault manifestations,difficult statistics of fault information and small sample data.Based on the above problems,this paper sorts out and analyzes the main fault manifestations of CNC machine tools,and then carries out some reliability tests.According to the experimental results,Bayesian theory and related improved algorithms are used to complete the reliability evaluation of CNC machine tools.The main research contents are as follows:(1)According to the characteristics of numerous subsystems and complex fault manifestations of CNC machine tools,the fault modes of CNC machine tools are statistically classified.Then the failure mode and effects analysis(FMEA)analysis method is used to analyze the failure of the whole machine of CNC machine tools.Then the Fault Tree Analysis(FTA)analysis method is used to establish the relevant model of the main drive system of CNC machine tools.Considering the small probability of the basic event is difficult to calculate,the fuzzy mathematics theory is introduced to estimate the probability of the basic event determined by the minimum cut set,and the fuzzy number of the occurrence probability of the top event is further determined.Finally,the importance of each event was analyzed.(2)Aiming at the problem of reliability data collection of CNC machine tools,a set of reliability test process is designed,which includes sampling method design;Fault determination and counting criterion design;Statistics and calculation of spindle load information;Air rotation test,static stiffness test of spindle system,spindle torque and cutting load test,etc.Aiming at the problem of small sample data collected by tracking test,Bayesian theory is introduced.Based on Bayesian prior information and test data,the compatibility of samples is tested by JB test and rank sum test.(3)When using Bayesian network to evaluate the reliability of CNC machine tools,determining a priori model is the premise of reliability evaluation.Based on the regularity analysis of prior information,the maximum likelihood method and the least square method,the point estimation and interval estimation of prior model parameters are determined.The AIC information criterion is used to determine the possible distribution types of the prior model.The KS test shows the feasibility of the determined model.Finally,according to the principle of parameter optimization,an improved adaptive average optimal particle swarm-support vector machine(AMPSO-SVM)algorithm is proposed to optimize the determined model.(4)When the distribution of the prior model is determined,the edge distribution of the model parameters is determined based on the bootstrap method.Then,based on the idea of markov chain conte carlo(MCMC),the reliability posterior model of the CNC machine tool is iteratively obtained by using the Gibbs sampling technology through the Open BUGS software.According to the above model,the reliability of the CNC machine tool is evaluated,and the calculated value of the reliability index is obtained.