Attribute control charts for effective statistical process control
Salah Haridy Gad Haridy
Date of Issue2014
School of Mechanical and Aerospace Engineering
The control chart is fast becoming a necessity rather than a fashion in different manufacturing processes and service sectors. No tool can capture the voice of a process better than the control chart. It is an effective tool to monitor a process, reduce variation, improve productivity and ensure quality. The applications of the control chart have now moved into engineering, service management, biology, health care and finance. The control chart is considered as one of the most powerful monitoring techniques in Statistical Process Control (SPC). It is basically used to achieve the statistical control of a process and its output. SPC provides the decision maker with the ability to monitor the quality characteristics of the product, evaluate the process performance and take a quick corrective action when out-of-control statuses and abnormal conditions are going to occur in order to avoid damages and serious economic losses. Attribute control charts play a vital role in monitoring the quality characteristics which cannot be conveniently measured in a continuous numerical scale. Nowadays, attribute charts enjoy a wide range of applications in many fields such as manufacturing processes, healthcare systems and service industries. The main objective of this thesis is to develop new attribute control charts with high detection effectiveness. This thesis proposes five new attribute charts, namely, a synthetic & np (Syn-np) chart, an optimal np & Cumulative Sum (np-CUSUM) chart, a CUSUM chart with curtailment (Curt_CUSUM), an optimal Sequential Probability Ratio Test (SPRT) chart for monitoring p, and finally a novel attribute chart (AFV chart) for monitoring the mean and variance of a variable. A second goal is to provide an overall effectiveness evaluation and systematic comparison among the newly developed charts and different attribute charts in the literature under the same false alarm rate for a fair comparison. The results of this evaluation give a clear conclusion on the overall detection effectiveness of the charts and provide a practical guide to both academia and industry. To achieve this goal, several types of commonly used control charts for attributes including np chart, synthetic chart, Cumulative Sum (CUSUM) chart, Exponentially Weighted Moving Average (EWMA) chart, and Sequential Probability Ratio Test (SPRT) chart are studied in this thesis. The results of these evaluation and comparison can be used as guidelines to facilitate the selection of the attribute control charts for different SPC practitioners. In addition, a general model for the optimal design of the attribute control charts is proposed. In this model, all the independent and dependent charting parameters are optimized using an exhaustive search algorithm in order to achieve the best overall performance. This search algorithm is simple and reliable. The Average Number of Defectives (AND) is adopted as an objective function to design and compare the charts subject to the same false alarm rate. Since AND is an overall measure of chart performance, therefore, minimizing AND will ensure that the control chart has an excellent overall performance across the entire shift range of interest. The results of the quantitative comparative study reveal that the new charts developed in this thesis have achieved a significant improvement in detection effectiveness. Specifically, the Syn-np chart stands as the most effective Shewhart-type chart for attributes in current SPC literature, the np-CUSUM chart and Curt_CUSUM charts are considered as the fastest CUSUM charts for detecting p shifts and the optimal SPRT chart is able to double the overall detection speed compared with the basic SPRT chart. Although the design of the Syn-np, np-CUSUM, Curt_CUSUM and optimal SPRT charts is more complicated and their implementation is slightly more difficult than the existing counterpart charts, the application of the new charts developed in this research can be justified by the substantial improvement in performance. Finally, the new AFV chart that employs a simple attribute inspection is found to outperform the variable X-bar&R and X-bar&S charts from an overall viewpoint, under different circumstances. As a limitation, the AFV chart is not able to detect decreasing variance shifts. Moreover, it is not quite suitable to monitor the processes parameters which change frequently. The new charts developed in this thesis may help SPC practitioners elsewhere to make a correct and timely decision in face of critical problems, to substantially reduce damages and quality cost in the long run, and to pave the way for a new cutting-edge research in attribute SPC.
DRNTU::Engineering::Industrial engineering::Quality engineering