Preforming of non-crimp fabrics with distributed magnetic clamping and Bayesian optimisation
Rajan Jagpal, Evangelos Evangelou, Richard Butler, Evripides G. Loukaides
Abstract
A novel preforming process was developed for non-crimp fabric (NCF) materials that generated in-plane tension through discontinuous blank boundary conditions. The method employed magnetic clamps and was designed to be both flexible and scalable, with clear routes to industrialisation. The capability of the process was explored in physical trials for a hemispherical and a cubic geometry. Characterisation of a biaxial veiled NCF showed the veil had a dominant effect on the bending mechanics. Subsequently a macroscale finite element model was developed to include an efficient bending idealisation and non-orthogonal in-plane material behaviour. Finally, global process optimisation of the preforming process was demonstrated. The optimisation approach used Gaussian process modelling with a periodic kernel to estimate the wrinkle size for untested clamping arrangements and then deployed Bayesian optimisation to find the optimal configuration. Results indicated that distributed magnetic clamping was effective and amenable to surrogate modelling.