FLOW 3D version comparison
 

Have recently gone back to comparing results obtained from different versions of the FLOW3D code we have on hand. We have discovered that there appears to be quite a bit of difference in the results produced when taking (all other inputs equal) a model in version 8.1 and running in 9.1 and later versions. Does anyone know of any major changes to the solver that would produce these discrepancies?
The problem involves an outflow of fluid from a tank. In version 8.1, there is a lot of sloshing at the free surface as the tank is drained, where in 9.1 and later versions, the free surface remains relatively flat. There are some forces put on the tank in the non-inertial reference frame model, which should result in some free surface movement, however the simulated differences are on an order of 10 times greater in magnitude in 8.1 compared to 9.1. First order momentum advection is used in all simulations, along with a SOR pressure solver, a laminar, viscous flow, and a non-inertial reference frame specified by a movin.inp file.
I know 8.1 is an old version of the code, however I would think that a simple moving, draining tank should be easy enough to model such that results should be similar no matter which version is used. Therefore, it is a little discouraging to see such a difference and tough to justify which version simulates the model more “realistically” without a better understanding of what changes may have been made to the solver to produce these differences.

Sam - you should have already been contacted by Flow Science about this. In short, there were models and new inputs added since 8.1 that require more input flags. This especially creates differences when the same prepin file is used for both runs (8.1 and 9.3).

Two major changes affected the viscous flow model and the GMRES pressure solver. The latter did not exist in 8.1 and became the default in 9.3. The viscous flow model requires activation flag in 9.1 which is off by default, while in 8.1 viscous flow option was activated if a positive viscosity coefficient was provided. There were two major releases in between – 8.2 and 9.0, so a bunch of other fixes could have been made in those releases too.