How good is your mesh anyway?

By

This week I gave an overview of some simulation best practices to some of our Nordic partners, and I thought I’d share one of their tips with you this week.

How do you know if your mesh is sufficiently refined? There are a number of checks and balances that can be used to assess the quality of your mesh in particular areas, but what about the general level of mesh refinement? One under-utilised technique is the convergence check.

image             image

Firstly create an initial mesh on your model and plot the number of degrees of freedom (DoF) against a result of interest (e.g. max displacement). Then remesh with a finer mesh (try to at least double the DoF in your second attempt) and again plot the result of interest. Keep doing this until you see the resultant parameter converge to a solution. When you are happy you have a converged result, you will also have a suitable global mesh.

image

Two additional points to think of:

  • if you choose stress as your function of interest ensure you don’t have a stress concentrating region in your model, otherwise you will find the value increasing indefinitely.
  • if your ‘ideal’ mesh takes forever to solve, you can pick a courser mesh size from the graph that solves quicker and gives you an understanding of the error when compared to the converged solution. If 160,000 DoF takes too long but the result at 80,000 DoF is within 5% error, this may be sufficient.

Try it next time you develop your simulation.

Matthew West

SolidWorks alumnus. I like plate reverb, Rat pedals, Thai curry, New Weird fiction, my kids, Vespas, Jazzmasters, my wife & Raiders of the Lost Ark. Not necessarily in that order.