Fluids

Fluids

Topics related to Fluent, CFX, Turbogrid and more.

Monitoring and judging solution convergence

    • taosif_alam
      Subscriber

      Hi !

      I am doing a transient simulation of solidification and melting of a solid block heated on the top surface by a moving laser beam heat source. I have already checked the mesh quality and its pretty good. I have setup the normalized scaled residuals’ convergence criteria to 1e-6. Also, I am monitoring max velocity and max temperature for solution convergence at each time step and its stop criteria is set to 1e-6. The time step is 1e-6 and iterations per time step is 40. I have provided the photos of all the solver setup I have done. And judging the convergence monitor, I observe the solution convergence does not change after 20 iterations and stays constant at the order of 1e-4 while the residuals are at the order of 1e-2- 1e-3 after 20 iterations. The residuals gradually decrease to the order of 1e-5 to 1e-6 after 40 iterations. Now, my question is, as the solution convergence does not change after 20 iterations, can I decrease my iterations per time step from 40 to 20? And what are the other things should I take note while monitoring convergence? 

    • Rob
      Forum Moderator
      Put the update back to 1 - checking convergence every 10 iterations isn't ideal and that option really ought to be hidden in the more modern releases. If you're not converging within around 15 iterations reduce the time step. You want to be moving though time with as few iterations as possible and 15-ish is generally the most efficient number per time step to converge: set maximum to be higher than that.
      Note the additional convergence criterion check for a CHANGE not a value so aren't widely used.
    • taosif_alam
      Subscriber
      Now, I have updated per iteration convergence check to 1. I have reduced timestep size to 0.5e-6 and set the max iteration per timestep to 15. The residual convergence is not so good now. Here is the picture

      What should I do to increase convergence other reducing timestep size?
    • Rob
      Forum Moderator
      What was the time step? Increase the max number of iterations and make sure you converge every time step. If you don't the results tend to wander off and produce rubbish.
Viewing 3 reply threads
  • The topic ‘Monitoring and judging solution convergence’ is closed to new replies.