Inspection of Sheet Materials — Model and Data

Abstract

In the inspection of sheet materials in industry, two interrelated tasks are involved: Visual search and decisionmaking. When these tasks have been studied separately, it has been found that both exhibit a speed/error trade-off. As more time is allowed for the task, errors usually decrease. However the speed/error trade-off in industrial inspection studies shows that, as time per item is increased, Type 2 errors (accepting faulty items) decrease while Type 1 (rejecting good items) errors actually increase. A model combining visual search and decision theory is postulated and tested on the inspection of flat glass for discrete faults. Separate experiments are performed to test the ability of inspectors to (1) search for and locate known defects in sheets of glass, and (2) decide whether a small fixated circular dot is above or below the standard required. The results of these two experiments are used to derive model parameters and the model used to predict performance in the factory situation. The model predicts the same change in errors with increasing time per item as results obtained previously in flat-glass inspection.