Strength prediction for textile composites using artificial neural network, principlecomponent analysisand unit cells

Pan, Q., Sitnikova, E. and Li, Shuguang (2015) Strength prediction for textile composites using artificial neural network, principlecomponent analysisand unit cells. 20th International Conference on Composite Materials, ICCM 2015, , Copenhagen, Denmark. 19 - 24 Jul 2015.

Abstract

There is no rational failure criterion that can be used for 3D textile composites as they have very complicated mesoscopic architectures and failure mechanisms. An advanced interpolation method called an Artificial Neural Network (ANN) has been employed to solve this problem numerically. However, the ANN based on mesoscale unit cell models requires definition of a large number of input parameters. This substantially increases ANN training time and therefore makes this approach computationally expensive. Therefore principal component analysis (PCA) has been resorted to in order to reduce the number of input parameters of ANN. The ANN has been coded into Abaqus Umat and Vumat user subroutines to describe the progressive failure behaviours of textile composites from virgin state to final failure. This methodology has then been verified by comparing the results from ANN and those from analyses of the unit cell model directly. The application of this methodology has been demonstrated through the simulation of a flat 3D braided textile composite plate impacted by a rigid spherical projectile. Experimental validation of the methodology is under way.

This record has no associated files available for download.

Contributors

Author: E. Sitnikova

Download statistics