Classification of Acoustic Emission data from buckling test of carbon fiber panel using neural networks

Acoustic Emission is a well-established structural health monitoring technique used to assess damage in structures under load. The identification of signals coming from different damage sources is key for assisting and helping the operator of the structure under monitoring in the maintenance decision making process (e.g. to decide if a repair is necessary due to a potentially dangerous failure or if the structure is still safe). This type of analysis has been a challenge due to the complex nature of Acoustic Emission signals and their parameters, which are historically believed to be different based on the source type (and therefore the damage mode). In this paper we propose the application of a neural network classification algorithm to a compression buckling test of a carbon fiber panel. The classification is supported by real-time strain data measurements with digital image correlation, and the damage type is assessed by pre- and post-test ultrasonic C-scanning and visual inspection of the panel. The results obtained allowed a validation of a neural network classification technique and to categorize the acoustic emission signals, by getting reliable indications of different damage modes (i.e. delamination and matrix cracking) taking place in different regions of the panel.