Model-Assisted Probability of Detection of Cracks Under Fastener Head

Abstract

After the Aloha Airlines Flight 243 accident in 1988, the importance of inspecting for widespread fatigue damage was emphasized. Managing the risk of multi-site damage (MSD), where multiple flaws occur in a single structure, became a key factor in maintaining the structural safety of aircraft. It is essential to demonstrate the reliability of nondestructive inspection (NDI) when detecting structural damage due to MSD. Structures affected by MSD often require inspection without being disassembled due to various reasons, such as increased inspection time, inefficient manpower and material consumption, and increased probability of MSD due to human error. NDI reliability is strongly influenced by the shape of the test specimen, and the uncertainty about crack conditions after inspection is greater when inspection is performed on a structure without disassembly than when it is assembled. To estimate the reliability of NDI in structures without disassembly, a model-assisted probability of detection (MAPOD) experimental design was used, combining physical modeling and machine learning methods. The test specimens simulated flaws such as discharge machining notches and fatigue cracks, and the detection probability was estimated based on the size of the detection signal. This was used in risk assessment to determine the inspection interval for the structure.