When solving a steady-state model, users may observe significant reductions in solution time by repeatedly saving a converged Snap and then re-solving the model while initializing from that Snap. For example, after obtaining a converged solution, the results are saved to a Snap, the model is re-solved using that Snap as the initialization source, and the process is repeated several times. Does this approach introduce any stability concerns, and why can it result in faster solution times?
No, using a converged Snap to initialize subsequent steady-state solutions is not considered a bad practice and should not introduce stability issues. When the Initialize Steady State from Snap option is enabled, Flownex uses the solution values stored in the selected Snap as the initial guess values for the next steady-state solve.
Because these guess values are already close to the final converged solution, the solver typically requires fewer iterations to reach convergence. This can result in a substantial reduction in solution time and total iteration count compared to starting from the default initial values.
It is therefore normal to observe faster convergence when repeatedly initializing from an updated converged Snap. Additional improvements may diminish after several iterations as the initial guesses become increasingly close to the final solution. In general, using a converged Snap as the starting point for future steady-state calculations is a valid and effective method for improving solver performance.