NPSS Convergence Limits for Stalled or Oscillating Residuals

Direct Answer

NPSS does not publish a deterministic rule that always prioritizes relaxation tuning over element reordering. The documented workflow is diagnostic: inspect the residual history, classify the failure as oscillation or stalling, and then choose the remedy. Oscillating residuals are typically addressed first by reducing relaxation or damping. Stalled residuals usually indicate poor initial guesses, inconsistent constraints, or weak information propagation, which may require improving starting values or reordering elements.

Confirmed Facts vs. Practical Inference

Confirmed: NPSS uses a modified Newton-Raphson solver with residual monitors and solver-side parameters. It enforces a maximum iteration count and a residual tolerance. Reaching the iteration limit without meeting tolerance is the formal signal that the deck requires structural or initial-condition changes.

Practical inference (version-sensitive): Exact namelist keywords, default iteration limits, and relaxation ranges vary by release. Confirm parameter names in the installed NPSS User Guide before editing decks.

Step-by-Step Diagnostic

1. Record the residual history. Enable iteration history printing or plotting. Capture whether residuals oscillate, decay monotonically, or stall across successive iterations.
2. Classify the failure. If residuals swing between positive and negative values or grow in amplitude, treat the case as oscillatory. If they plateau or move very slowly toward tolerance, treat it as stalled.
3. For oscillation: Reduce the relaxation or damping factor in small increments. Compare iteration count and residual trajectory against the baseline run. Avoid excessive damping, which can mask physically inconsistent constraints.
4. For stalling: Back up the deck. Improve initial guesses for tightly coupled stations first. If stalling persists, reorder two tightly coupled elements to change solver coupling and rerun to test whether residual propagation improves.
5. Check the limit. If the solver reaches the maximum iteration count without meeting tolerance, the deck is formally non-convergent under the current structure and requires structural or initial-condition changes.

Steady-State vs. Transient Behavior

In steady-state mode, the solver drives cycle residuals to a global tolerance. In transient mode, convergence is assessed per time step. Consequently, transient decks respond better to time-step reduction and transient damping than to element reordering. Applying steady-state remedies to a transient deck may destabilize time integration.

Missing Diagnostic Detail

The single detail that changes the recommendation is the residual pattern: do the residuals oscillate or stall? Without that history, relaxation tuning and reordering are equally plausible starting points.

Verification Workflow

Because exact solver keywords are version-sensitive, verify them in the installed documentation. A safe workflow is:

# 1. Enable iteration history (check exact keyword in your NPSS version)
#    e.g., set solver print/plot options in the deck or run control

# 2. Baseline run
npss your_deck.inp > baseline.log

# 3. After adjusting relaxation or reordering, rerun and compare
npss your_deck_modified.inp > modified.log

Compare residual trajectories and iteration counts between logs. If stalling persists after relaxation reduction and initial-guess improvement, reorder elements only after backing up the working deck.

Note: This guidance reflects general NPSS solver behavior. Verify parameter names and defaults against the installed release before applying changes.