Acoustic emission detection and prediction of fatigue crack propagation in composite patch repairs using neural network

Abstract

This article presents the results of acoustic emission (AE) monitoring of crack propagation in 2024-T3 clad aluminum panels repaired with adhesively bonded octagonal and elliptical boron/epoxy composite patches using FM-73 adhesive under tension–tension fatigue loading. Two crack propagation gages and four broadband AE sensors were used to monitor crack initiation and propagation, respectively. The acquired AE signals were processed in time and frequency domain to identify sensor features correlated with fatigue cycle and crack propagation, which were used to train neural networks for predicting crack length. The results show that AE events are correlated with crack propagation, and crack propagation signals can be differentiated from signals due to matrix cracking, fiber breakage, and shear of the composite patch. Three back-propagation cascade feed-forward networks were trained to predict crack length using number of fatigue cycles, number of AE events, and number of fatigue cycles and number of AE events together as inputs, respectively. It was found that network with fatigue cycles as input gave good results, while the network with just AE events as input gave greater error. However, the network using both fatigue cycles and number of AE events as inputs to predict crack length gave much better results.