Jaraiedi, Majid; Creese, Robert; Clark, Robert (1986, ASQC) West Virginia University, Morgantown, WV
The inspection function, whether automated or manual, is an essential component of most quality control systems. This activity has profound effects on the production process and profitability of the enterprise. The typical inspection plan does not consider errors in inspection, which can result in the rejection of conforming parts (Type I error) or the acceptance of non-conforming parts (Type II error).Inspection research has recently focused on evaluation of inspector accuracy using Signal Detection Theory (SDT). Plots of Hit Probability Versus False Alarm Probability are called Receiver Operating Characteristic (ROC) curves. ROC curves can be used effectively to measure inspection performance. An ROC curve is based on a priori inspector's perception of the payoff matrix and the signal probability. A perceived change in either factor results in a shift in the criterion point, which is the cutoff boundary for acceptance and rejection of the item being inspected.This paper employs Bayes' theorem to convert a priori probabilities of detection of non-conformaning items (Hit Probability) and rejection of conforming items (False Alarm Probability) to posterior probabilities of inspection effectiveness. These posterior probabilities give rise to the definition of Receiver Analysis Curves (RAC), which depict the "after the facts" consequences of inspection error. The horizontal axis on the RAC shows the Average Outgoing Quality of the inspection procedure, and the vertical axis is a measure of the accuracy of detection of non-conforming items.
Error,Receiver Operating Characteristic (ROC) curves