Steady simulation component

The steady simulation component monitors the change between consecutive forward fields and freezes the forward fluid solve when the change is below a user-defined tolerance.

In case files, enable it under case.simulation_components:

"simulation_components": [
    {
        "type": "steady",
        "tol": 1e-7,
        "log_frequency": 50,
        "scalar_coupled": false,
        "compute_control": "tsteps",
        "compute_value": 1
    }
]

Convergence measure

For each field, steady computes an energy-like norm of the increment \(\Delta q = q^n - q^{n-1}\),

\[ \lVert \Delta q \rVert = \frac{\sqrt{\int_\Omega (\Delta q)^2 \, d\Omega}}{\Delta t\,|\Omega|}. \]

The component evaluates this for fluid fields (u, v, w, p) and, when present, one scalar field.

Stopping behavior

Fluid convergence is reached when max(u, v, w, p) is below tol.
Scalar convergence is checked separately when a scalar is present.
If scalar_coupled = true, both fluid and scalar must satisfy tol before fluid freezing is applied.
Once converged, the component sets fluid freeze = true.

Note: Scalar freezing is not yet applied by this component. The current implementation only freezes the fluid solver.

Parameters

Parameter	Type	Default	Description
`type`	string	-	Must be `"steady"`.
`tol`	real	`1e-6`	Convergence threshold for the field-increment norm.
`log_frequency`	integer	`50`	Write cadence (in timesteps) for convergence diagnostics.
`scalar_coupled`	bool	`false`	If `true`, require both fluid and scalar convergence before freezing fluid.
`compute_control`	string	inherited	Standard Neko simulation-component scheduling control. Typical value: `"tsteps"`.
`compute_value`	integer/real	inherited	Scheduling interval/value associated with `compute_control`. Typical value: `1`.

Output

The component writes convergence data to steady_state_data.csv with header:

iter,time,u,v,w,p,scalar

where each field column is the normed increment at that logging point.

Limitations

The component currently supports at most one scalar field.
For truly time-dependent optimization workflows, prefer the unsteady configuration described in Unsteady simulations.