Optimization of Distribution Parameters for Estimating Probability of Crack Detection

摘要: In the inspection of aircraft structures, probability detection has typically been determined based on size damage alone. However, process involves randomness due to variability in conditions, including inspector competence, difficulties associated with location, and type damage. To account for these other factors, we develop a simple model from assumption that each combination crack location there is threshold size, such larger cracks will be detected smaller ones missed. The proposed fits sizes 2603 events reported 43 panels inspected by 62 inspectors U.S. Air Force study. values are obtained maximizing matching between data. First, when thresholds used, matches 78 % events. It further increased 81 both employed. For comparison, percentage using alone only 55%. Cross-validation fitting indicates part improvement represents noise data, about 72% repeatable. extended represent inconsistent behavior inspectors. Replicating observed inconsistency same reduces 72 %. Confidence intervals presented. may used assess effect better training inspectors, this illustrated study shows that, if poorer half replaced top half, identified 50% can reduced half.