Ever noticed when you walk a dog it always wants to investigate the area just beyond the range of its leash? Seems the same universal rule is in play configuring Computational Aided Engineering (CAE) simulations, slightly too big to fit the available resources. If only we had a leash to rein in errant simulations.
CAE computing resources are determined by:
- Analysis technique (e.g. surface or volume based)
- Domain discretization or mesh (i.e. number of elements)
Misquoting George Orwell in Animal Farm, some simulation regions are more equal than others. Yet most automated simulation systems ignore this fundamental truth, treating all regions as equal, over resolving mundane features and under resolving crucial features. Such simulations strain at the leash to provide reasonable turnaround times on all but the most massive of massively parallel computing resources. Visualization, swamped by a tsunami of data, turns into a stuttering, disk grinding chore at frame rates more worthy of early cinema. There has to be another way.
Instead of giving in to all regions are equal, let's embrace some regions are more equal than others. Let's trade brute force for economy and elegance. Throughout the element generation process (mesh generation) Caedium has simultaneous access to the true geometry (not a triangulated approximation) and the simulation physics (techniques) from which it can automatically select an appropriate meshing technique and determine appropriate element sizing. In important regions the user can specify higher accuracy directly on the associated geometry, thus simply and elegantly placing elements where needed, minimizing the turnaround time and maximizing the chance the simulation will fit the available computing resources. Less is more.
Better still if the system has a toolbox of techniques and can automatically select the minimal (in resource terms) technique based on the simulation physics selected by the user. The difference in turnaround time of a surface-based numerical technique over a volume-based technique can be an order of magnitude. Using an appropriate numerical technique is another powerful tool to minimize computing resource requirements. Over time we will develop a toolbox of add-ons offering numerical techniques to enable this vision.
What is power without control? Of course all this power is pointless if not packaged in a single, unified simulation environment. Caedium converses with the user using familiar terminology. Simulation physics terminology is preferred over numerical technique terminology. Widely adopted GUI standards such as drag and drop, toolbars and full undo/redo are incorporated to ease learning and maximize productivity. The environment is goal driven such that specifying what you want where drives the meshing and numerical techniques to produce results in the form of 3D visualization or 2D plots.
Caedium and its add-ons are our vision of a simulation environment; we hope it can be yours too.
Recent blog posts
- Sliver Treatment Strategies for CFD
- How to Fix Small Acute Angles for CFD
- Small Feature Removal for CFD
- Fluid Visualization in Nature
- Design is Compromise
- CFD Is Not Enough
- All CFD Should Be Upfront CFD
- Speed Up CFD with Symmetry and Cyclic Conditions
- Airships Rising
- World Cup Stadium Aerodynamics with CFD