So, I have a reference simulation meshing that's 100% reliable but it's costly in terms of time so I'm looking for optimizing it by finding the minimum mesh allowing at most 1% of error with respect to that reference meshing. What's the best "algorithm" I can do? It's 2D so I thought of doubling number of elements in both directions if error > 1% after initialization with a random meshing, and if error < 1%, I lower both by 1/4. If it's <1%, I lower again, if >1%, I choose the average between it and the previous one.

Is there any more efficient technique?

  • 1
    $\begingroup$ How do you evaluate your error? $\endgroup$ Apr 2 at 15:21
  • $\begingroup$ Absolute value gap between a quantity of a random meshing and a reliable highly-meshed configuration divided by this latter $\endgroup$ Apr 2 at 17:36
  • $\begingroup$ If you already solved the problem to better than 1% accuracy, why would you want to solve it again on a coarser mesh? $\endgroup$ Apr 2 at 22:47
  • $\begingroup$ To see if I can solve it keeping the 1% accuracy but with less CPU computation time $\endgroup$ Apr 4 at 19:06


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.