Modeling fibre orientation in short fibre suspensions using the neural network-based orthotropic closure

Industrial methods for modeling the fibre orientation within short-fibre reinforced polymer composites require the use of a closure approximation, whereby the equation of motion for an even-ordered orientation tensor depends on the next higher even-ordered orientation tensor. The orientation within the processed part correlates to the material stiffness tensor; therefore it is essential to have confidence in the accuracy of the selected closure. This paper suggests a novel methodology in formulating a closure by employing an artificial neural network (ANN) training algorithm, and presents the coordinate frame invariant Neural Network-Based Orthotropic Closure (NNORT). Results demonstrate that, for most of the flows investigated, the NNORT closure offers accuracies representing the orientation and the stiffness tensors, equal to or greater than the current industrially employed closures, without any computational increases.