Typical Standard Approach to check for results
- Put an initial course mesh ~ 0.1 –0.2 grid size everywhere
- Use constant viscosity method
- Run for 10 iterations check the flows and all heat sources
- Run for a further 100 iterations check the solution for areas where lots of things are happening and refine mesh if necessary
- Change turbulence model to k-ε method
- Refine mesh by increasing local mesh density and overall density by ~ 10 -20%
- Refine mesh until solution is independent of mesh (given available time)
- Use a grid of 3*3*3 cell monitors to check independence
- If cell monitors are levelling out and the convergence is flattening use either relaxation or false time steps to speed up convergence
Judging the completion of simulations
The two main ways of judging whether the run has completed are as follows:
- Check the residual history printed on screen. The sum of the normalized absolute residuals should diminish steadily.
- Check the monitoring cell location for the dependent variables at the user set location within the fluid domain. These should stabilise to the converged solution.
It is very important that checks are made during the initial stages of the analysis to monitor progress of the solution run. In cases of solution divergence, the run should be terminated and appropriate adjustments made to the relevant control parameters such as relaxation factors, false time steps etc.
Neglecting this can result in lost time and unproductive simulations. Note, however, that increases in residuals and oscillations in the computed variables during the early stages of a run are not uncommon and should disappear after a few iterations. The run should therefore be given sufficient time to stabilise before any judgement is made on its progress.
Difficulty in achieving convergence
Some common issues are:
|
Problem
|
Cause
|
Remedy
|
|
Residuals continuously increasing
|
False time steps set too high
|
Reduce false time steps for velocities and possibly temperature
|
|
Residuals fail to reduce
|
Oscillating flow pattern
|
Reduce false time steps for velocity
|
|
Unrealistic initial values
|
Check initial values under Settings
|
|
Internal heat source without sink, e.g. radiator in room with adiabatic surfaces
|
Check that the problem is realistic. Set cell monitor points and check for continuously increasing/decreasing values, which would indicate imbalance
|
|
Erratic convergence
|
Unstable flows, e.g. strong jets or buoyancy driven plumes
|
Reduce false time steps for velocity.
|
|
Mass residual reduces very slowly
|
Various causes
|
Increase number of inner iterations for pressure
|
|
Residuals all reducing steadily but very slowly
|
False time step set too low
|
Increase false time steps
|
If the MicroFlo solver experiences numerical difficulties during the calculation run the following items should be checked and may need to be changed or altered:
- Physical model shape
- Boundary Conditions
- Make use of sensible initial values
- Quality of the grid
- Solver parameters