Strategy: Fixing Convergence Problems
When a study does not converge, you should create a
convergence graph result window. If the graph shows that the study almost converged during the run, the results are likely to be accurate.
If the graph shows that the quantity did not come close to converging, you should take one or more of these steps, and then run the study again:
• Define and show a fringe plot of the p-level. This will show the polynomial level to which Creo Simulate calculated to reach convergence for each edge. If one or two elements went to much higher p-levels than the rest, you should try dividing those elements by adding datum points in that location to seed the mesh.
• If an element did not reach convergence at the location where a load, heat load, constraint, convection condition, or prescribed temperature is applied on a point, or at a small feature of significance, you should be sure you have the appropriate point options selected in the area of the
AutoGEM Settings dialog box. Proper use of these options enables AutoGEM to create small elements surrounding the point or near the small feature. Alternatively, for more precise control over AutoGEM, you can define
Isolate for Exclusion AutoGEM controls.
You can edit the analysis to specify those small elements as
excluded elements and direct
Creo Simulate to ignore these excluded elements when calculating convergence.
• Check the convergence value you entered when you defined the analysis. If that value is too tight, especially if it is below 1%, you should try loosening it.
• Check the polynomial order you entered when you defined the analysis. If the maximum is lower than 9, you should try increasing it.
• If you are primarily interested in a quantity other than the convergence quantity, check if that quantity converged. You only need to rerun the study if the quantity of interest did not converge or come close to converging.