Here are some potential checkpoints for this analysis:
- Since ParaView doesn’t have a second derivative function, the Hessian matrix is calculated by applying the first-order accurate gradient (Gradient Filter) twice.
- Because it’s a first-order accurate gradient, it will be constant within a cell. However, the Gradient Filter maintains continuity by taking the average of the gradient values of the cells sharing a point as the derivative value at that point. Nevertheless, a loss of accuracy is likely unavoidable when applying such a derivative twice.
- It’s currently unclear if the observed artifacts are a result of applying the first-order accurate gradient twice. We need to check if using a higher-order accurate derivative from VMTK improves the situation.
Other Concerns:
- Has cell data been converted to point data? (It’s advisable to enable this conversion in the Fluent to EnSight format export options.)
- Has the issue of internal boundaries being outputted been resolved? (See: https://discourse.paraview.org/t/fluent-dat-file-export-on-gold-ensight-paraview-format-creating-mesh-lines/16500)
- Was the Gradient Filter applied before cutting the tube?
- Are there any ghost cell issues? (Given that parallel visualization doesn’t appear to be used, this is likely not an issue.)
- Are there multiple blocks? If so, applying the Merge Blocks filter first might lead to improvements.