Statistical Process Control For Job Shop Manufacturing With Bayesian Approach

Process distribution parameters are usually unknown in practical. Traditional statistical process control is based on maximum likelihood estimators of process parameters. In order to eliminate the defect of traditional statistical process control based on maximum likelihood estimation for job shop manufacturing, the conjugate Bayesian approach is chosen to estimate the process distribution parameters. Process mean and process variance is respectively set as normal distributed and inversed-gamma distributed. Prior information is selected from previous batches according to group technology, Lilliefors test, Bartlett’s test and ANOVA analysis, and then prior parameters can be obtained. After combining the sample information of current batch, posterior distribution and its bayes estimator is presented. The calculation of control limits and process capability indices based on conjugate Bayesian approach is proposed for job shop manufacturing to eliminate the defect of traditional statistical process control. With the increase of sample size, the weight of sample information will be automatically larger in bayes estimators. When sample size is large enough, statistical process control based on maximum likelihood approach is a special example of statistical process control based on conjugate Bayesian approach. Simulations and a practical example verify the effectiveness and practicability of the method...