Signal Detection Theory in the Analysis and Optimisation of Industrial Inspection tasks

Abstract

The overall aim of this study was to investigate, the applicability of Signal Detection Theory (SDT) to a number of problems in the area of industrial inspection, including training and selection. Two industrial studies comprising three experiments are presented, together with four laboratory experiments and a correlational study. The first three chapters of the thesis comprised a comprehensive review of SDT and the literature of inspection. Chapter 2 described an industrial case study, the inspection of photographs of nuclear particles, designed to test the applicability of SDT in an applied setting. The variables of auditory noise, defect complexity and time on task were also considered. SDT in its unequal variance form was found to fit the data. The next case study attempted to apply SDT to the inspection of photographic film. A two stage decision making model was proposed to describe performance in this task. The first two laboratory studies investigated the effect on performance of within and between session changes in defect probability. It was found that the subject could adjust his criterion appropriately to between session changes if feedback was provided, and to within session changes if he received prior warning of the change. The final laboratory studies were concerned with training the inspectors ability to modify his criterion, and the enhancement of his sensitivity. The first experiment replicated previous work in perceptual training, and the second utilized a wide range of differing training techniques. It was found that certain combinations of conditions were significantly superior in achieving the training goals. A correlational study was conducted utilizing the results of the previous two experiments and tests of various cognitive skills. Significant correlations were found between certain groups of test scores and performance on the task. These tests were proposed as potential selection techniques for inspectors.