Run

What this step does

Setup check. Before a run, the step lists findings about the case without meshing:

• inlet Reynolds number $Re = U D_h/\nu$: a warning below 2300, because the turbulence model then adds a viscosity that a laminar flow does not have;
• inlet Mach number: a warning above 0.3 (incompressible), an error above 0.6 (compressible);
• the straight length of each outlet pipe: a warning below 2 diameters;
• cells across a passage, transient Courant number, heat-transfer walls, and perforated-wall parameters outside their model's range.

Each finding links to the step that sets it.

Pipeline. Start run solves on the saved mesh of the Mesh step. A mesh is generated first when there is none; when the saved mesh's patches no longer match the boundary conditions, the step asks before meshing again. The stages are:

1. mesh; 2. decomposePar; 3. The potential-flow solver (initial field); 4. solver phase 1: first-order upwind for the turbulence quantities (Upwind iterations before the switch); 5. solver phase 2: the configured schemes; 6. reconstructPar.

Every stage writes its own log. While phase 2 runs, the program checks every few seconds whether the mean of the pressure drop has converged (Δp mean-stability tolerance) and, once it has, ends the solver with its fields written, before Maximum iterations. A steady compressible case that diverges in the start-up phase restarts once with conservative relaxation. Acoustics, when selected, runs after the flow.

Quality gates. Gate 1 judges the mesh (non-orthogonality, skewness, layer coverage, and that the meshed domain is the fluid). Gate 2 judges the solution: the pressure-drop mean has converged, the residuals are below Residual target (initial residuals, OpenFOAM's convention), the global mass imbalance is below 0.1 % of the inlet flow, the wall $y^+$ is in range and, with heat transfer, the energy balance closes within 1 %. Several further rows are advisory and never fail the gate.

Outcome. One label summarises a finished run:

• Completed: every gate passed.
• Completed with warnings: the run finished and the pressure-drop mean converged, but a gate (typically the residual target) or an advisory check needs attention.
• Not converged: the pressure-drop mean has not converged, or there were too few iterations.
• Failed: gate 1 failed, the solver crashed or a step returned an error.
• Stopped: stopped by the user.

The same label appears on the run badge, the status line, Open Recent and the Results step.

Pause, Stop, Continue. Pause makes the solver write its fields at the current iteration and exit, so no CPU or memory remains in use and the run survives a restart of the program. Continue resumes from those fields with a continuous iteration count (identical results on serial and 2-rank runs). Pause is not available in the start-up phase or for transient runs. Stop (two clicks) ends the run at once; the partial result is still evaluated and can be continued. An acoustic sweep can also be paused: finished frequencies are kept and reused when the assembled system is unchanged.

CFD and acoustics are two runs. Each physics selected on the Physics step has a run container on this page, CFD first, then Acoustics; a physics that is off is not shown. With fluid flow and acoustics both selected, "Start CFD run" meshes and solves the flow only; Pause, Continue and Stop in the CFD container act on the CFD. The Acoustics container becomes available once the CFD has written a solution, and a strip between the two says what the acoustics use of it (the mapped medium, or only the saved solution). "Start acoustic run" builds the acoustic model on the saved CFD fields (grazing and through-flow speeds, the mapped medium) with the acoustic settings as they are at that moment, and has its own progress bar, Pause, Continue and Stop. So the CFD can be continued for more iterations first, or acoustic-only settings (frequency range, ports, impedance models) changed, without solving the flow again; "Run acoustics again" repeats the acoustic run on the current CFD. When the CFD solution is not converged, the acoustic run still starts and its QA table says so. A run from the queue or the command line still runs both parts one after the other. "Settings changed since this run" in Results keeps the CFD's settings from the CFD run and the acoustic settings from the latest acoustic run.

After a solve each run container opens its key numbers: for the CFD the total pressure loss, the loss coefficient, the iteration count and the mass imbalance; for the acoustics the quality gates, the number of frequencies, the degrees of freedom and the sweep time. The residuals, the pressure-loss history, the two QA gates and the acoustic sweep and mesh tables are in the results containers below.

Cores. With Parallel cores at 0 (auto), the program meshes serially and solves on min(physical cores − 1, cells / 15 000) processes.

Queue. "Add this case" copies the case, the CAD and the exported surfaces into a new folder inside the project's _files folder (Where projects and results are stored), so later edits do not affect it; the Results step's run selector shows their results. One job runs at a time, and a direct Start run waits for the queue. "Add mesh-independence study" queues three meshes of Mesh-independence study parameter with a GCI estimate.

Source

Validation results of the program:

• Core count, on a 4-core / 8-thread machine with 184k cells (seconds per iteration): 1 core 0.909, 2: 0.572, 3: 0.516, 4: 0.481, 7 processes 0.457.
• Pause / Continue on a 46k-cell pipe (paused at iterations 142 and 170): the resumed pressure drop is identical to that of an uninterrupted run to 6 digits.

The convergence rule and the mass-imbalance limit (ten times tighter than the usual 1 % guidance) are described in Δp mean-stability tolerance.

Valid range

Fields in this step 1

Related 9

• Mesh-independence study parameterqueue.mesh_study.parameter
• Parallel coressimulation_control.parallel_cores
• Δp mean-stability tolerancesimulation_control.dp_plateau.tolerance_pct
• Δp windowsimulation_control.dp_plateau.window_iterations
• Maximum iterationssimulation_control.max_iterations
• Residual targetsimulation_control.residual_target
• Upwind iterations before the switchnumerics.turbulence_upwind_iterations
• Meshstep.mesh
• Resultsstep.results