Introduction
Once the environment, case settings, and mesh have been checked, you can run the simulation.
The next responsibility is to determine whether the numerical solution has converged sufficiently for the quantities you want to interpret.
Learning objective
Assess a simulation using residual histories, conservation checks, and monitored engineering quantities.
Key idea
Judge convergence using several indicators together. A completed run or a low residual value alone is insufficient.
Understand the idea
Residuals describe the remaining imbalance in the discretized equations during the solution process. Their histories help reveal progress, stagnation, and instability.
Their definition and scaling depend on the solver, so avoid treating one numerical threshold as universally sufficient.
Also inspect conservation. For a steady system without mass sources, the net mass flow through all external boundaries should approach zero.
Finally, monitor quantities relevant to the engineering question. Depending on the case, these may include pressure difference, flow rate, or torque.
A steady calculation should produce sufficiently stable quantities of interest. Persistent oscillations require investigation; they can arise from numerical problems or from flow behaviour that is poorly represented by the chosen steady model.
Engineering example
Suppose the residuals decrease while the predicted pressure difference continues drifting.
The calculation has not yet demonstrated a stable pressure-difference result. Running longer may help, but the mesh, boundary conditions, and numerical settings may also require review.
Keep the evidence together rather than judging the run from one plot.
Practical task
Run the workflow documented in the starter and retain the solver log.
Prepare:
- Residual histories.
- A mass-balance assessment.
- A history of at least one relevant engineering quantity.
Explain why you consider the run sufficiently converged—or why further investigation is needed.
Checkpoint
If the solver reaches the configured final iteration, has convergence necessarily been achieved?
No. It may simply have reached its stopping limit. Convergence must be assessed from the numerical evidence and the stability of the quantities being evaluated.
References
- The starter’s execution instructions and solver documentation.
- Tutorial on CFD Verification and Validation — NASA