Defect detection for aircraft components: An approach using ultrasonic guided waves and neural networks

摘要

Ultrasonic Guided Wave (UGW) techniques have become significant tools for the application of Non-Destructive Testing (NDT). Their ability to propagate within a structure for large distances is a key feature. In the case of structures with relatively simple geometries, such as pipes, the inspector's analysis of received reflections of the excited waves can highlight defective areas of the structure. However, when dealing with complex geometries, such as those found in aircraft components, the occurrence of reflections received also becomes complex, thus making analysis challenging. For the presented technique, neural networks are employed to analyse trends of the ultrasonic signals received over time and provide insight regarding the emergence of defects in the monitored components. Large amounts of data were collected from samples representing critical aircraft components at various temperatures, and defects were introduced artificially whilst monitoring occurred. Neural networks were trained using this data and evaluated against their accuracy in classifying specimens as either defective or defect free. In this paper, the proposed monitoring technique and the results of the experiments are presented.