Viscoelastic fluids appear in various industrial applications, including adhesive application, additive manufacturing, seam sealing and parts assembly with adhesive. These processes are characterized by complex geometry, moving objects and transient multiphase flow, making them inherently difficult to simulate numerically. Furthermore, substantial amount of work is typically necessary to setup simulations and the simulation times are often unfeasible for practical use.
In this thesis a new Lagrangian-Eulerian numerical method for viscoelastic flow is proposed. The viscoelastic constitutive equation is solved in the Lagrangian frame of reference, while the momentum and continuity equations are solved on an adaptive octree grid with the finite volume method. Interior objects are modeled with implicit immersed boundary conditions.
The framework handles multiphase flows with complex geometry with minimal manual effort. Furthermore, compared to other Lagrangian methods, no re-meshing due to grid deformation is necessary and a relatively small amount of Lagrangian nodes are required for accurate and stable results. No other stabilization method than both sides diffusion is found necessary.
The new method is validated by numerical benchmarks which are compared to analytic solutions as well as numerical and experimental data from the literature. The method is implemented both for CPU computation and in a hybrid CPU-GPU version. A substantial increase in simulation speed is found for the CPU-GPU implementation. Finally, an industrially suitable model for swirl adhesive application is proposed and evaluated. The results are found to be in good agreement with experimental adhesive geometries.