Time-accurate simulations and restart

The current URANS workflow starts from an owned conservative restart artifact. A display field or arbitrary exported visualization file is not a substitute for that restart.

Set simulation_type="urans", architecture="fluxcore", max_iterations=null, a positive physical time_step, and n_physical_steps. Supply restart_artifact_id on the outer simulation request. Match the mesh, physical model and restart compatibility required by validation.

Physical and inner time scales

Choose the physical time step from the flow time scales and the output you need to resolve. The inner iterations converge each physical step; increasing them does not increase the simulated physical duration. Assess both time-step sensitivity and within-step convergence.

The contract supports first- or second-order backward differentiation. The resolved default is BDF2 with at least one first-order startup step. The default inner iteration ceiling is 100, with a chunk size no greater than 25 or that ceiling. Discover current defaults and constraints before overriding them.

Washout and statistics

Use transient_washout_steps to exclude initial adjustment from sampling and transient_sample_steps for the retained period. Their sum must equal the total physical steps. A longer sample should cover the phenomenon of interest; a visually periodic short trace is not proof of statistical convergence.

Set iterations_per_output deliberately. Zero disables intermediate field output while leaving final export controlled separately. More frequent fields increase storage and data-handling costs.

See time-integration API for predictors, inner convergence controls and a validated request shape.