The following error occurs in the CFX-5 Solver. What does it mean and how do I get around the problem?

| ERROR #001100279 has occurred in subroutine ErrAction. |
| Message: |
| A circular dependency was detected during gradient extrapolation. |
| This error message may be bypassed by setting the expert paramet- |
| er 'stop if gradvxb error=f'. |
| |
| |
| |

**** Entered By: mptooley @ 03/03/2005 11:06 AM ****

The error occurs in Boundary Node Gradient Extrapolation. Gradients on boundary vertices at inflows, outflows, walls, etc. need to be extrapolated from neighboring interior nodes. The gradient calculation cannot be closed on boundaries without interior neighbors (e.g. a tetrahedral element who`s edges straddle all three sides of a corner will result in the corner node not having any internal neighboring nodes). If for a given boundary vertex all neighboring nodes are also boundary nodes, then the solver tries to recursively find a connected interior node. If the solver fails to find an interior node in reasonable distance (10 edges), then it stops with the above message.

Often this error occurs on Free Slip Walls, which are supposed to be Symmetry Planes. In this case the Boundary Condition type should be changed to 'Symmetry'. For symmetry Boundary Conditions zero gradients apply normal to the plane, and correct gradients tangential to the plane. Some other situation in which this can occur are:

- In a region where the mesh is only one element thick.
- If a periodic condition is using GGI rather than 1:1 periodicity and the two interfaces are 'close together' (ie, no interior nodes). In this case switch to 1:1 periodicity or make sure there is an interior node.
- If two wall nodes are separated by a gap of 1 element, the boundary extrapolation may be circular.

If you are sure that only a few nodes are affected, or that those nodes are in areas of low interest, then you may override by setting the expert parameter 'stop if gradvxb error=f'. In this case, zero gradients (in all directions!) will be applied for vertex based gradients on the nodes in question (but element based gradients will be correct), resulting in an upwind scheme. If the effected area is large, it will detrimentally effect the results, as the solution accuracy will be locally first order.

You can set expert parameter using the Definition File Editor, accessed from Tools> Edit Definition File in the Solver Manager. The help documentation describes how to do this.

Show Form
No comments yet. Be the first to add a comment!