Research On The Method Of Small Sample Reliability Modeling For CNC Machine Tools Based On Bootstrap-bayes

CNC machine tools are the basal equipments of the modern equipmentmanufacturing industry, the technical level of which always are the direct embodimentof the modernization and comprehensive national strength of one national industrial.Nowadays, the competition in the market is mainly the competition about thereliability. How to improve the reliability of the independent products, and occupy aplace in the fierce competition CNC machine tools market, is the major problem theequipment manufacturing industry in our country must face.Accurate evaluation of the reliability of the CNC machine tools is theprerequisite and basis of the reliability improvement, and the establishment of theminimum error reliability model is the core of the reliability evaluation task. Theexisting reliability modeling method mainly focusing on large sample situation, whichbased on the low of large numbers, can efficiently estimate the situation of more than50of the samples. But while the number of sample is too little, the traditionalmodeling methods become not suitable. This paper in the light of the situation ofsmall sample for the CNC machine tools reliability modeling, studied the followingaspects of modeling method.(1) Taken the scene failure date of the subsystem of five units of machiningcenters from enterprise as the object of study, through the comparative analysis ofeach stage in the course of modeling, estimated the distribution parameters andobtained the original reliability model of the subsystem of machining centerrespectively by using the least square method, the K-S test method and the greyrelational analysis, obtained the evaluation results by using a modeling method oflarge sample.(2) Taken the failure date of the subsystem of one marching center of above asthe object of study, and the failure data of similar subsystems of three marchingcenters as the prior information, established the reliability model of small sample byBayes method. Combined the Bootstrap sampling method and the Bayes calculation method into the improved Bootstrap-Bayes calculation methods, effectively ruled outthe faults about the excessive subjectivity in the process of determination the priordistribution by Bayes and the excessive faults in the process of sampling by Bootstrapsampling method, accurately modeled and estimated the parameters of priordistribution of reliability model.(3) By using the Gibbs sampling method of Monte-Carlo and Winbugs software,got the posterior distribution and expectations of the parameters of reliability model.This method effectively ruled out the faults about the complex calculation. Throughthe comparative analysis of the above three reliability model, the new method wasproved to be correct.(4) Taken the Visual Basic6.0as the development platform, by comprehensiveusing Access database technology and Matlab software, established a reliabilityinformation management and evaluation system for CNC machine tools and itssubsystems, to which added the above research results, in order to standardized thesmall sample modeling process by the new method.