Geometrical modelling of woven composites

• Main Author: Potter, Emily
• Other Authors: Robinson, Paul ; Pinho, Silvestre
• Published: Imperial College London 2013
• Subjects: 629.13
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.572270

Geometrical modelling of 3D woven composites represents a key part in both material and component design. Using current modelling techniques it can be difficult to correctly mesh unit cell (UC) models and apply the necessary periodic boundary conditions (PBCs), often with sacrifices made in idealisation of the weave architecture. An automated voxel meshing technique suitable for modelling woven composite UCs has been developed, which is generic in nature and allows incorporation of architectural weave deformities, including tow rotations/misalignments. The model requires node points with five independent variables to define the UC geometry, and is capable of representing a range of complex material options. PBC application is simplified due to the grid based nature of nodes on the unit cell surfaces. Both simple PBCs and more complex varieties (incorporating offset translations/rotations) are possible, enabling UC domain reduction (depending on weave architecture and types of deformation modelled). Models of 2D and 3D woven composite UCs have shown that good elastic stiffness predictions can be made using voxel models, with agreement between existing finite element models, and correlation with analytical rule of mixtures based solutions. The use of an existing smoothing technique to improve the tow/matrix interface by moving tow surface nodes has provided mixed results. In some cases (dependent on both mesh refinement and geometry), this technique has reduced the artificial peak stresses caused by 90° voxel ‘steps’ at the interface. However at other times, there has been limited improvement from use of this smoothing technique. A new surface improvement algorithm has therefore been developed, using node movements and splitting of elements to improve tow geometry matching. This has removed artificial peak stresses at voxel ‘steps’; however, in its current form, it produces spurious stress concentrations due to element surface incompatibilities. A range of suggestions are made to combat these issues.