Data fusion approach to layup characterisation of CFRP laminates using ECT, UT, and CT

This work introduces a data fusion (DF) approach to quantifying layup sequences in carbon fibre reinforced polymer (CFRP) laminates using rotational eddy current testing (ECT), pulse-echo ultrasonic testing (UT), x-ray computed tomography (CT), and prior knowledge. The approach involves generating cost functions for each scanning modality and the prior information, according to the Bayesian inference model. Finally, a multi-objective genetic algorithm searches the parameter space to obtain a Pareto front of possible solutions and their trade-offs. The desired solution is then selected by a pretrained weighting of the cost functions. The results show an increase in accuracy and reliability over optimisation approaches that use a single cost function, with the data fusion result giving a solution mean error of between 0.3 and 1.1 degrees.