Runtime Diagnostics and Simulation Efficiency
Many users have probably seen the list of Runtime Diagnostics available at the left side of the Simulate panel and probably wondered about them. They must be there for a reason, right? Well indeed, they are. The Runtime Diagnostic plots were designed to provide users with a simple indication of how the simulation is progressing. By understanding what these plots represent, users can achieve efficient simulations with minimal effort.
The image on the left shows a typical list of Runtime Diagnostics available during a simulation. The actual list of available diagnostics during any given simulation depends on which physical models are active. For example, if the Solidification model is active, Solid Fraction will be available.
By clicking on a particular item in the list, it will be highlighted, displayed in the box at the top of the list, and also plotted to the right of the list.
The blue progress bar provides an indication of how far along the simulation is relative to the Finish Time. If the simulation termination condition is Fill Fraction, the progress bar will indicate progress relative to the specified Finish Fraction. Similarly, if the termination condition is Solidified Fluid Fraction, the progress bar will indicate progress relative to the specified Finish Fraction.
Below is a brief description of some of the Runtime Diagnostics.
Stability Limit D&T
Ideally these two plots would be identical. The most efficient simulation is one in which the Time Step (Grey line) used to advance the simulation is equal to the Stability Limit (Black line). The time step can never be greater than the stability limit since this would lead to an inaccurate result.
There are many reasonable scenarios in which the time step can be less than the stability limit. In the image at the left, the initial time step is set by the user to a value less than the stability criteria to get the simulation going gradually. Since the stability limit is larger than the time step, the solver gradually increases the time step toward the stability limit during the simulation.
If the time step is much less than the stability limit throughout the simulation (inefficient simulation), the CPU time may be excessive. This is typically due to excessive pressure iteration required to compute a challenging physical problem. To allow a larger time step to be used, a more powerful pressure solver such as GMRES may be necessary. Also, the parameter ITDTMX, can be increased. This parameter is set on the Model Setup, Numerics, Convergence Controls in the box titled, "Maximum number of pressure iterations before time step is reduced."
The default value of ITDTMX for the SOR pressure iteration method is 50. If SOR is used and the number of pressure iterations exceeds 50, the time step will be reduced by 5%. Clearly continued excessive pressure iteration can lead to a very small time step.
EPSI & Max Residual
Another useful runtime diagnostic is EPSI & Max Residual. EPSI represents the convergence criterion that is automatically computed by FLOW-3D. Max Residual represents the level of convergence in the pressure iteration scheme and can be thought of as the maximum difference between the amounts of fluid moved in and out of the computation cells. Pressure iteration will continue during each time step until Max Residual is less than EPSI. If the maximum number of iterations is reached (defaults are set by FLOW-3D) and max residual is still greater than epsi, an error message will be output indicating this.
If the simulation physics are challenging--for example, large flow gradients--the user may see many pressure iteration failures and the EPSI & Max residual diagnostic may look similar to the image on the left.
When the Convergence Criterion (Grey) is greater than Max Residual (Black), there is a possibility of fluid volume gain or loss in the simulation. To improve pressure convergence, the user may:
- Reduce the Initial Time Step on Model Setup, Numerics, Time Step Controls
- Reduce the Maximum Time Step (same panel as #1)
- Increase Maximum Number of Iterations on Model Setup, Numerics, Convergence Controls
The runtime diagnostic Volume Error, shown in the image below, gives an indication of how much fluid volume gain or loss has occurred during the simulation.
A Volume Error less than 0.5% is usually acceptable and would not require any action on the part of the user. If the volume error is large, the user may:
- Reduce iteration failures if they occur
- Consider higher level advection scheme such as IFVOF=6 available in Model Setup, Numerics, Volume-of-Fluid advection.
When running filling simulations, the onset of solidification can be a primary factor in deciding whether to continue the simulation. Perhaps the existence of any solidification before filling is complete would not be acceptable. The runtime diagnostic Solid Fraction, shown in the image below, displays the amount of solidified fluid in the simulation. If any solidified fluid was detected, the user is quickly notified of this.
While there are many other runtime diagnostics available, the ones covered here will give the user a good indication of how their simulation is progressing without any post-processing.