Changing Mindsets for Large-scale Modelling

Solving large-scale problems is substantially different from solving small-scale problems, and a change in approach is often required. Communicating this idea to end use practitioners can be difficult, yet important.

We demonstrate with an example showing the need to move away from direct solvers in order to solve large solid mechanics problems in O($n)$ time. Whilst this may be uncontroversial, it does result in new areas of focus. Rather than trying to reduce the number of cells in a mesh, there is now be a stronger emphasis on good cell quality. Additionally, iterative solvers are more difficult to drive than a black-box direct solver, so more experience is required to get good results.

By demonstrating significant speed-ups on real world problems, we can make a good case for the investment of time and training which is required.