Coupled non-conforming finite element and finite difference approximation based on laminate extrapolation to simulate liquid composite molding processes. Part I: isothermal flow

In composite manufacturing by resin injection through a fibrous reinforcement, several phenomena occur involving the flow of resin and heat exchanges by the resin with the fiber bed and the mold. During processing, through-thickness flows commonly appear when the fibrous preform is made out of a stack of plies of different permeabilities. In many situations, simulation of the mold filling in Liquid Composite Molding (LCM) requires performance of a full three-dimensional analysis. In the pre-processing stage, the construction of three-dimensional finite element meshes of complex parts made out of components of small thickness with respect to their length is very tedious. These parts are typically composed of an assembly of shell and flat panels. In such parts, a good aspect ratio of the finite elements must be respected to ensure appropriate simulation results. These conditions result in meshes with a very large number of degrees of freedom, which translates into a too high computational burden. For these reasons, a new numerical approach is presented in this paper for accurate and faster simulation of the filling phase in LCM. Based on the fabric reinforcement layup, a new non-conforming finite element is developed to quickly evaluate the through-thickness flow. Starting with a spatial triangular mesh as input geometry and based on the stacking sequence of the preform, an extrapolation algorithm is used to extrude the mesh in the thickness direction and generate the 3D non-conforming finite elements. To further evaluate the validity of the three-dimensional model, an experimental verification was performed for a typical through-thickness flow. Then, a comparative study is conducted to demonstrate the advantages of the proposed methodology in terms of accuracy. The results are compared with 2D (triangles) and 3D (tetrahedrons) finite element solutions. Finally, the performance of the model is assessed in terms of computer time.