Impact damage detection in carbon fibre composites using neural networks and acoustic emission

During drop-weight impact testing of Carbon Fibre Reinforced Polymer (CFRP) composite laminates, transient Acoustic Emission (AE) waveforms were detected using a broadband piezo sensor firmly attached to the surface of the specimens. Two types of CFRP were investigated. The first material, reinforced by T300 carbon fibres, was subjected to impact energies ranging from 2 J to 15 J, The second material, reinforced with T800 carbon fibres was subjected to impacts of 5.4 J to 20 J. Resultant damage areas were measured using ultrasonic C-scan. A Back Propagation Artificial Neural Network (BPANN) model was trained to reproduce the AE transient waveforms produced by the least damaging events. Differences between these AE outputs and those produced by more damaging impacts were represented by Means Square Errors (MSE), The MSE and traditional AE energy and count parameters were correlated with impact damage area, The best indicator of the level of damage produced during impact was the impact energy, The MSE values obtained from the neural networks appears to be a better correlation to damage area than traditional AE parameters in T300 CFRP, while a poorer correlation was observed for T800 CFRP, This result may be attributed to either the arbitrary character of AE detection or scatter in material properties. The results are discussed in terms of impact and AE criteria which may be used as indicators of damage occurrence and damage severity.