Research And Application Of Online Monitoring Method Based On Censored Dat

As the complexity of production processes increases,the existence of censored information in the data collection process has received increasing attention from scholars.Censored data is different from the processing of complete data set,so for both the industrial production process of product quality,or health care,financial services and other fields of information censoring data has been a hot issue in the research.Statistical process control is one of the most important methods in the field of quality control.As the main means of statistical process control,control charts are widely used in industrial production,medical health,traffic flow monitoring and other fields,and play an important role.However,in the process of data collection,limited by factors such as time and cost,information will be censoring.Most of the traditional statistical process control chart research is based on historical sample observations and parameter estimation of complete data.However,the data in need of monitoring has the characteristics of censoring,and the control chart designed according to the complete data can no longer meet the monitoring needs of actual production and other fields.Therefore,it is urgent to improve the existing fixed monitoring schemes and build control charts that adapt to new data forms and new application scenarios.This paper will study the methods and applications of online monitoring based on censored data,and propose online monitoring strategies suitable for various censored data scenarios.For data with high censoring rate in various distribution forms,window censoring and progressive censoring data types,corresponding control charts are designed to meet the needs of online monitoring in different scenarios.Overall,this paper is divided into three main parts.The first part starts from the monitoring of basic censored data and considers the robust monitoring of censored data under unknown distribution in the case of high censoring rate;the second part starts from the scenario of window censoring and investigates the construction of control charts for the alternate renewal process in the case of window truncation;the third part starts from the type II progressive censored data and analyzes the robust monitoring in the case of inadequate sample size due to censoring.The first chapter introduces the research background and significance of this paper,the literature review and the overall framework of the paper.In Chapter 2,some basic knowledge required for this paper is introduced,including the basic theoretical ideas and construction methods of control charts.The three main contributions of this paper are in Chapters 3 to 5.Finally,in Chapter 6,the research content and future research directions of this paper are summarized.The first part of the study is based on a semi-parametric approach for monitoring data with high censoring rates.For time and cost considerations,high censoring rate is quite common in life tests,which is a critical issue in lifetime monitoring.Conventional control charts designed for highly censored data are commonly based on the Weibull distribution.However,the distribution assumption may not be valid in practice,which brings challenges to the monitoring procedures.Motivated by this,a semi-parametric exponential weighted moving average(EWMA)control charting procedure is developed for highly censored lifetime data of any distribution.The control scheme uses the Kaplan-Meier estimator to construct the cumulative distribution function(CDF)and generalized Pareto distribution to improve the tail estimation.Then a Kolmogorov-Smirnov statistic defined by the differences between the in-control CDF and the empirical CDF is integrated into an EWMA charting scheme to monitor the Type I right-censored sample.We use simulation studies and a real-data analysis to show the efficiency of the proposed control chart.The second part of the study is a new scenario application based on windowcensored data.In the case of failure,repair,and replacement of industrial components or physical infrastructure,the working and out-of-service times form an alternating renewal process(ARP).Typically only events that occur within a specific time period are recorded,resulting in a window-censored situation.A window-censored observation of an ARP can be considered as a stationary ergodic stochastic process with alternating values of 0 or 1 in a short time interval.Thus,censored observations may be left-censored,right-censored or doubly-censored.In this paper,we propose a likelihood-ratio-based cumulative sum(CUSUM)control chart,along with a modified conditional expected value exponentially weighted moving average control chart to monitor window-censored observation.The construction and implementation of this control scheme is studied by simulation and its properties are evaluated.Finally,this paper shows by example that it has a good monitoring effect in practical applications.The third part of the study is based on the scene monitoring of progressive typeII censored data.Progressive censoring is essentially an extension of the traditional censoring mechanism,but its censoring mechanism is more flexible,so it has received more and more attention and research.Progressive censoring is characterized by allowing a subset of observations to be removed within two failure times,which leads to increased complexity in monitoring censored data.It is precisely because it is difficult to ensure that sufficient samples are observed with progressive censored data,and it is urgent to solve the update of monitoring strategies under progressive censored data.In this part,based on the goodness-of-fit test,the parameter distribution is estimated by using the data in phase I,and the empirical likelihood function of the censored data is used in phase II to construct monitoring statistics.Through numerical simulation,it is verified that the strategy has higher accuracy and robustness than the original method at the monitoring level,especially for the case of high censoring rate.Finally,the superiority of the new monitoring scheme is verified again through case analysis.Finally,we summarize this paper and discuss some potential research directions.