Parameter Adaptive Statistical Quality Control Charts

Statistical process control (SPC) has been widely used to monitor various indus-trial processes by using statistical methods and techniques, to improve and guarantee the quality of production. Recently the methods used in the industrial statistics have reflected changes in the discipline, such as improved computer technology and avail-ability, resulting in highly flexible and computer-intensive approaches to scientific data analysis. Statistical approaches to screening, flexible modeling, pattern characteriza-tion, and change detection that were infeasible 20 years ago are now viable. Many of the problems in industry today concern the analysis of huge and complicated data sets that lead to improved quality or better understanding of the manufacturing or devel-opment process. Thanks to recent advances in sensor and information technologies, automatic data acquisition techniques are commonly used in increasingly complicated industrial processes and dataset. Also, a large amount of data and information related to quality measurements in a process has become available. Thus, statistical approaches to make use of the data and process information regarding control, monitoring and di-agnosis have become possible and beneficial in industrial practice. However, SPC to monitor and control the quality of such data-rich processes remains in many challeng-ing problems.SPC was developed in 1920s and Dr. Walter A. Shewhart from US Bell Telephone Laboratories developed the first sketch of a modern control chart in May 1924. Dur-ing World Warâ…¡, SPC technique was extensively developed both in the UK and in the United States. However, its application and importance lost in these West countries as industries during peacetime production. Meanwhile, this technique was introduced the Japanese by Dr. Deming and leaded a big impact to Japan in the 1950s. Japanese industry widely applied SPC technique to save costs, improve qualities, and thus at-tract customers interests. The successful application of SPC can be verified by Japan's rapid economical development, especially its industrial occupation in worldwide mar-ket. Following Japan, other Asia countries, for example, Korea, Taiwan adopted SPC technique to improve their industry. Both two countries made great success in many kinds areas, such as manufacturing, quality management, social service in 1980s. Chi-nese economy is now growing at the world highest rate, however, its economic foun-dations are mainly depended on some low technical industries. Many important areas are nearly subjected to labor intensive industries. The problem of low quality and reli-ability in Chinese production always needs a long way to improve. In 2009, the global economic crisis told us, China must find its new way to improve its industry structure and reform many areas to get through the crisis and make a new and robust step for the future development. Thus, as a important part to develop our countries industry, quality should be paid enough attention.SPC technique includes many efficient tools. Generally speaking, control charts play a very important role in modern SPC technique. Among various kinds of control charts, there are three fundamental types of control charts. The first one is the well-known Shewhart X chart which is the earliest chart to monitor the process changes (process mean and variance). The advantage of X chart is that it is more efficient in detecting large magnitude of the process shifts than others, but less in small and moderate shifts. The reason is that X chart only uses the information from the current sample readings but ignores the former ones. In order to overcome this shortcoming and increase the sensitivity to the small shifts, Dr. Page proposed the famous cumulative sum (CUSUM chart for short) chart which based on the Wald test in 1954. Another important chart, exponential weighted moving average (EWMA for short) chart, was first proposed by Roberts in 1959. Both CUSUM and EWMA charts have the same character that they will use not only the information of the current sample but also will use the former ones, and they are nearly same efficient in detecting small shifts of the process. Totally saying, it is hardly to say which type of chart performs better than others.Most SPC applications assume that the quality of a process can be adequately represented by the distribution of a quality characteristic. However, in many recent ap-plications, the quality of a process may be better characterized and summarized by the relationship between the response variable and one or more explanatory variables. That is, the focus would be on monitoring the profile that represents such a relationship, in- stead of on monitoring a single characteristic. There can be found much research work on profile monitoring, for example, from linear profile to nonlinear case, from single process stage to multi-stages. Now, these issues are drown more and more attention by the researchers. From another aspect of quality control, the traditional assumption that the process shifts is only assumed fixed (called step shift) no longer seem reasonable for many situations. Instead, many literatures pay more attention to study the case that the process shifts are unknown or with other complicated pattern, not only fixed one. These kind of study makes the control chart become more practical and useful.Moreover, most application of the control chart are heavily based on parameter choice. However, this can be a shortcoming in the real practice since that the engineer may make a choice only by his experience or other subjective views. Many failures in using the control chart are mainly refer to the wrong choice of the parameters. To overcome this shortcoming, many researchers make lots of efforts to design a kind of control chart with parametric-free feature or parametric adaptive feature (called Adap-tive Control Chart). Other problem in application the control chart is refer to the sam-pling schemes, for example, the sampling time intervals and sample sizes. These kinds of issue refer to the economic design of the control chart, which fully concern about the operating fairs of time, cost and.people. It is important to note that these two is-sue (parametric adaptive feature and sampling schemes) draws much attention in our researches, and the structure of this dissertation is demonstrated as follows:In Chapter 2, we propose a single control chart based on the parametrical adaptive feature which integrates variable sampling intervals procedure. The comparison has been made with other novel charts. The results show that our new chart performs better than the existing charts.In Chapter 3, we develop an adaptive CUSUM control chart based on the EWMA scheme mentioned in Chapter 2 to monitoring the multivariate process. We compare this adaptive chart with another adaptive one. The results show that our chart can detect the changes much faster than the existing one.In Chapter 4, we propose a weighted CUSUM control chart which avoids the as-sumption that the process shifts is fixed or pre-known, to monitoring the autocorrelated process. And the results show that when the process is out of control shortly after the process begins, our new chart performs better than that chart.In Chapter 5 and 6, we extend our research range to various type of the process mean shifts. Two omnibus control scheme are proposed based on likelihood ratio meth-ods to monitor any kinds of unknown process mean and variance shifts, respectively. We compare it to the other charts and we find that our charts always performs better than the comparative charts not only for step shifts, but also for various patterned cases.In the last chapter, Chapter 7, we summarize the main results and suggest some interesting problems worthy of further investigation.