Deviation Prediction And Control For Multistage Manufacturing System

With the constantly intensified market competition, the enterprises have moreexpectations of the product quality. Generally，product quality forms along with the designprocess, manufacturing process and sale process. For manufacturing enterprises, themanufacturing process quality is the key effect on the product quality. In this paper, wefirstly discuss the analysis and modeling of the simplest manufacturing process, andgradually extend the model to universal situations, which provide a reliable qualityanalysis method for enterprises.First of all, three factors which affect the manufacturing deviation in seriesmanufacturing system are analyzed: fixture installation deviation, location datum drift,and cutting tool deviation. In this process, State Space Model is used to model themanufacturing deviation propagation of the simple series manufacturing system. And thenapplies the model to the specific production instance to verify the feasibility of the model.In order to make up for State Space Model can only be applied to the simple seriesmanufacturing system, two kinds of quality diagnosis theories are introduced in themethod to construct the following deviations: independent deviations of single station, andtransmission deviations between stations. Independent deviations of single station ismodeled by State Space Model, and transmission deviations between stations is modeledby least squares support vector machine. Comprehensive the above two kinds ofdeviations, we can get the manufacturing deviation model of the parts. Then the feasibilityand superiority of the model is verified by examples. In addition, the problem of modelaccuracy was analyzed.Finally, according to the quality of the predicted data to conduct statistical processanalysis and diagnosis: in view of the single factor attribute variables，using two kinds ofcontrol chart diagnosis theory and two kinds of process capability index of diagnosistheory to analyze and diagnose; in view of the multiple factors attribute variables, usingMultivariate T2control chart and multivariate statistical process control based on principalcomponent analysis (PCA) to analyze and diagnose. Meanwhile, the paper illustrates theabove method using detail examples.