An approach for decisions on revision of control charts' limits

Improvement of product and process quality has gained importance in order to provide competitive advantage and reduce production costs in today's rapidly developing market conditions. The main objective of process improvement is to stabilize a process by eliminating systematic errors rapidly and to prevent their recurrence. Stability of a process implies that it produces output in predictable ranges over time. Control charts are one of the most important tools to ensure process stability and are used for process monitoring to detect out of control conditions. A process is considered to be in control if a control chart does not give a signal out of control. If the process is in control, it is assumed that the mean of the monitored control statistic is constant and those assignable causes, which could cause additional variation in the process, are not present. Within the scope of this thesis, however, it is assumed that the mean of the control statistic for an in-control process is not a constant but a random variable. Due to the dynamic nature of processes, this is a more realistic approach for process monitoring problems where the mean of the control statistic may indicate small deviations from time to time because of minor undetectable causes. While designing control charts, historical observations from a relatively short period of time are used and control limits are determined accordingly. These control limits are a model of an in-control process, and out-of-control conditions that may occur in the process are detected by deviations from this model. However, over time, there may be natural variability in a process that is considered to be in control. Therefore, it is necessary to revise control limits of a control chart from time to time by using a systematic method for revisions. In this thesis, an integrated approach is proposed for monitoring processes and revision of control limits when necessary. The proposed approach is evaluated by simulations and performance evaluations are presented to indicate benefits.