Change-point Detection Approach For Multistage Processes With Binary Data

With the continuous development of technology and industrial upgrading,whether manufacturing or service industries,the importance of product and service quality is increasing.Statistical process control technology as one of the important tools of quality management and control,the central idea is to ensure product quality and process stability to the greatest extent.The early traditional statistical process control technology mainly uses the mathematical statistics method to monitor the production process in real time.These statistical process control techniques are mainly for the control of a single quality characteristic.With the increasing complexity of manufacturing and service processes,production and service generally need to go through multiple stages.The output quality of the previous stage is often the input quality of the next stage.Therefore,the quality level of the current stage is not only affected by this current stage,but also by the output quality of the previous stage.Most of the researches on multi-stage process quality control assume that the variables obey normal distribution.However,in many production and service processes,due to the influence of time or cost and other factors,response variables often cannot be described by measurement data,but can only be represented by discrete data.In this thesis,this thesis focuses on the Phase Ⅰ change-point identification method research when the quality characteristics are binomial data in the multi-stage process.The generalized linear model is used to model the multi-stage process,and the transfer mechanism of influencing factors in the multistage process is constructed.On the basis of modeling binomial multi-stage process,a control diagram based on likelihood ratio test is proposed to identify the position of sample change points in the multistage process of binomial data.By introducing the direction vector into the likelihood ratio control graph to expand the control graph,it can not only identify the position of the sample change point,but also further determine the position of the change stage.Finally,the detection performance of the change point identification method is compared and analyzed with the other control chart.Simulation results show that when the parameters of binomial multi-stage process have different degrees of deviation,the method proposed in this thesis is significantly superior to T~2 control chart in sample detection probability and accuracy in phase Ⅰ study of the binomial data multistage process.Finally,this thesis applies the method to a commercial video broadcasting(CVB)service system,and it is verified that the method has good detection performance in practical application of binomial multistage process.