ACOUSTIC EMISSION DETECTION AND PREDICTION OF FATIGUE CRACK PROPAGATION IN COMPOSITE PATCH REPAIRS USING NEURAL NETWORKS

摘要

Composite patch repairs are increasingly used to repair damaged aircraft metallic structures. This calls for a reliable method for assessing the integrity of bonded repairs and detecting crack propagation after patch repair without stripping the repair. This paper presents the results of Acoustic Emission (AE) monitoring of cracked 2024-T3 clad aluminum panels repaired with adhesively bonded octagonal and elliptical single sided boron/epoxy composite patches under tension-tension fatigue loading. Two crack propagation gages and four AE sensors were used to monitor crack initiation and propagation respectively. AE signals were acquired and processed in time and frequency domain to identify sensor features correlated with fatigue cycle and crack propagation. Methods for AE noise reduction, accurate source location, and crack length prediction are presented. The results show that AE events and fatigue cycles are correlated with crack propagation. The identified sensor features were used to train three back-propagation feed forward neural networks to predict crack length based on the number of fatigue cycles, AE event number, and fatigue cycles with AE events together, as inputs. It was found that network with fatigue cycles as input gave better results than network with just AE events as input. However the network using both fatigue cycles and AE event number as inputs gave best results.