My last post discussed meshing, more specifically Standard versus Curvature based meshing, so I’ll continue from there. We can continue the mesh discussion with mesh quantity or mesh quality. I’ll go with quantity.
More is not always better with mesh element quantity. You may recall from the patent-pending one-minute calculus post that we are approximating an infinite number of mesh elements, so in theory we can have that many. Like the choice between Standard and Curvature based mesh, we need to ask ourselves why SolidWorks Simulation doesn’t just use “a lot” of elements. The answer is efficiency, or in the FEA world the term is convergence. Each new element introduces up to 10 new calculations - nodes. Each new calculation increases the resources needed to make the calculation and the time it takes to run. Each time a calculation is made, error is introduced. After a certain number of elements are used, the return on investment is very small and in some cases, wrong. Convergence is a separate topic, so I’ll just note that more mesh elements is not always better and we need to minimize the number of elements to the best of our ability to get the best results.
One way to minimize the number of mesh elements is to use defeaturing. It’s right at the front of the Simulation training manual, but on our side of the support inbox, we see it over and over again – models that are too detailed. Overly detailed models create issues for meshing. Detailed models require very small mesh elements or at least mesh controls in localized areas where the geometry isn’t suited for the global mesh type.
The problem begins with the thought that we want to simulate the exact product in the real-world environment. Physicists use things like frictionless ice and perfect vacuums for a reason – that’s the only way the math works. Another example is modeling water in a tank to find the stresses in the tank. There’s no need at all to model the water – we apply those forces to the tank with Simulation.
What we are creating is an approximation of the real world, so there’s no need to use a model with all of the details needed for use in the real world. Using an overly detailed model will not improve the outcome. Besides, the results – regardless of how detailed our model and forces are – will always need to be correlated. No one builds a product without physical testing (no one that stays in business). The key is to create a Simulation setup that closely predicts the physical testing outcome.
There is a Defeature tool in Simulation, but defeaturing can be done in other ways. Some people create simplified configurations for their Simulation runs. Suppressing features is another way. Suppressing or removing parts of an assembly is also a good choice. For example, the vanes in a water pump are needed for a flow analysis, but not to simulate the mounting point stresses on the housing.
So, idealize models, take out what you don’t need, and then work on the mesh.