A fuzzy rule-based algorithm to improve the performance of statistical process control in quality systems

Statistical process control (SPC) is an important part of quality control s ystems in industrial applications. It is widely used to monitor parameters in production processes and detect abnormal parameter values that indicate a fault in the process. Measurements of controlled parameters commonly exhi bit random variations that arise from either environmental changes or rando m variations in the measuring instrument itself. SPC uses control charts to determine whether variations in measurements are due only to random change s within the range expected or whether they indicate a real process fault. Inevitably, traditional control charts sometimes generate Type I errors (fa lse alarms), indicating a process fault when none actually exists, and caus ing an unnecessary stoppage of the plant. In other cases, Type II errors ar e generated, where real faults are either not detected at all, or are detec ted only after some time delay during which product quality has been impair ed. This paper describes an investigation into the use of fuzzy logic to mo dify SPC rules, with the aim of reducing the generation of false alarms and also improving the detection and detection-speed of real faults.