A Fault Detection for a Correlated Process with the Use of SPC/EPC/NN Scheme

The statistical process control (SPC) chart is effective in detecting process shifts. One important assumption for using the traditional SPC charts requires that the plotted observations are independent to each other. Otherwise, the so called "false alarm" would be increased, and these improper signals result in the wrong interpretation. However, this independent assumption is often not the case in practice. The use of engineering process control (EPC) has been therefore proposed to overcome this problem. Although EPC is able to compensate for the effects of disturbances, it also decreases the monitoring capability of SPC. This study proposes a combination of SPC, EPC and neural network (NN) mechanism to solve this difficulty. Using SPC/EPC/NN scheme, this study introduces a useful technique to identify the starting time of a process disturbance based on the execution of a binomial random experiment.