Research On Statistical Process Monitoring Based On The Coefficient Of Variation

As the main tool in statistical process monitoring(SPM),control chart has widely application in industrial engineering,medical service,etc.With the method and technology of mathematical statistics,it realizes the process monitoring during the manufacturing.According to the collected sampling information to judge the system’s status,it prevents the nonconformity production and guarantees the quality of the process.Most traditional control charts are designed for detecting the shifts in the process mean(X)and standard deviation(S).However,in some cases,the change in the process mean or standard deviation fails to demonstrate the process is out of control.When the process population mean(μ)and standard deviation σ varies in a fixed proportional way,only monitoring the process X or S is unable to understand the process and traditional X or S chart is no longer applicable.To deal with the dilemma,the process coefficient of variation(CV)is recommended.Monitoring the process CV can efficiently acquire the status of the process and discovery the assignable causes resulting in the quality fluctuation.Research on monitoring the CV has increased in recent years.How to enhance the sensitivity to the shifts in the CV by constructing new monitoring schemes has become the hot research topic in the academia.The existing CV charts are mostly used to monitor the univariate CV.Nevertheless,with the increasing complexity of the production process,the production data tends to exist in a multidimensional form.In such a situation,existing univariate CV charts cannot satisfy the manufacturing requirements obviously.Therefore,it is more important for the engineers to monitor the process multivariate CV(MCV).Considering foregoing problems,this paper focuses on updating and improving the CV monitoring schemes to detect the shifts in CV quickly from two perspectives,i.e.univariate and multivariate cases.To monitoring the univariate CV,this paper proposes one-sided double exponentially weighted moving average(DEWMA)chart firstly in Chapter 3.The run length(RL)metrics are computed by Monte Carlo simulation and the result shows that the proposed charts improve the charts’performance when detecting the small shifts in CV.Considering the effect of adaptive strategy on the efficiency of traditional control charts,Chapter 4 investigates the performance of the DEWMA CV charts with variable sampling interval(VSI)feature.Using auxiliary information based(AIB)estimator,the AIB DEWMA CV charts are proposed in Chapter 5.By incorporating VSI feature into the above schemes,the VSI AIB DEWMA CV charts shorten the time to make the out of control signal.When the process shift happen in the process start-up period,the fast initial response(FIR)feature is adopted into the MEC CV charts which are constructed by mixing the exponentially weighted moving average(EWMA)and cumulative sum(CUSUM)charts in Chapter 6.It turns out that the FIR feature makes the proposed MEC CV charts respond quickly against the initial shifts in the process.According to the real industrial example from sintering process and other fileds,it is shown that the proposed CV charts have excellent monitoring ability in practice.To complete the monitoring and diagnosis of multivariate data,Chapter 7 gives two one-sided CUSUM charts for monitoring the upward and downward shifts in the process MCV.The RL properties are evaluated based on Markov chain model.On one hand,the optimal design is conducted for deterministic shifts by using Nelder-Mead algorithm.One the other hand,the integration in expected values of the RL metrics is calculated by using Legendre-Gauss quadrature when the shift is unknown.Based on the above study,the effect of the sample size and parameter dimension on the performance of the CUSUM MCV charts is in-depth investigated.The numerical and graphical comparisons with other MCV charts show the superiority of the CUSUM charts.Based on the application to the financial investement industry,the proposed MCV charts are assumed to possess superior performance and be of great value.