I am working on an RMA2 model in Dallas, TX. The scatter points I used to generate the mesh were supplied in State Plane coordinates. In the area I am working, the (x,y) coordinates are on the order of (2400000,6900000). I was getting a few GFGEN slope rule violations in what seemed to be well-formed elements, and I never could figure out why. Finally, I needed to refine the mesh in a few areas and I started getting RMA2 crashes in the COEFS subroutine, again in what seemed to be well-formed elements. Thanks to this forum, I determined that my problems might have been caused by RMA2 truncating the nodal coordinate values because they were so large.
To solve this problem, I used the node transform option in SMS to offset the nodes so that the coordinate values were on the order of 5 digits instead of 7 digits. This was a simple translation of the nodal coordinates - I subtracted 2400000 from the x coordinates and 6900000 from the y coordinates. There were no negative coordinates generated as a result of this translation. When I ran GFGEN, all the mysterious slope rule violations were gone. I thought my troubles were over.
Unfortunately, I get radically different results for steady-state simulations using the same boundary conditions depending on what the nodal coordinates are. After I moved the nodes, I saw an unusual circulation pattern in the steady-state velocity vectors, and the general direction of flow through part of the mesh was in the opposite direction.
I then experimented with moving the nodes around with different offets, and each time I moved the nodes around the results varied.
My first thought was, it must be caused by Coriolis forces, but the global latitude specification for all materials in all runs was 0.0.
I don't have a tremendous amount of experience with RMA2, and this has got me stumped. It may be an easy solution - any suggestions would be appreciated.