- A run log will be printed. Inspect the log for messages of inappropriate or missing data. Note that the name of a surfaces cannot be empty, although the string "Untitled" will suffice.
- Make sure that the surfaces are defined in a way so that the traling vortices from one panel do not pass through the control points of another, or that their bound vorticies overlap.
- Inspect the residual history. If the residual appears to be converging, but just did not have enough time to reach the specified tolerance, increase the maximum number of iterations, or decrease the tolerance.
- If the Run fails at the very first angle of attack, try choosing an intial angle of attack where your configuration will not be producing much lift (this might be negative).
- If the residual appears to increase or stay level, try reducing the relaxation factor while simultaneously increasing the maximum number of iterations. If this does not help, adjust the dissipation values. If you are in a post-stall region, the amount of k2 dissipation may have to be relatively large. Otherwise, the dissipation should be as small as possible, as it can adversly affect the accuracy of all solutions, and the convergence of pre-stall solutions.
- Typically you will need less dissipation and more under-relaxation the smaller your panels become. This can lead to long run times. Be sure to compare results on coarse and medium-fine panel distributions to see if you really need very high levels of refinement.
- Often good results may be obtained at what seems like relatively large values of the tolerance value. Inspect the Final Spanwise Gamma distribution. The change in Gamma over the last interation is also plotted (in green), which gives an indication of the error in Gamma. If this is at an acceptable level, then the solution can be considered to be valid.
Panel Distributions In general, uniform panels, or smooth changes in panel sizes are best. Avoid placing too many panels on low aspect-ratio wings.
Airfoil Data Irregular airfoil ClCdCm data can inroduce convergence difficulties, especially if they introduce large negative slopes in the lift curve. Consider smoothing your section data if the oscillations are due to measurement or numerical problems rather than representing physical effects which should be resolved.
Multi-Segmented Wings
In order to approximate an elliptical lift distribution, you may have chosen to build a single wing out of multiple surfaces with discontinuous taper changes. For such cases you should still have a continuous Gamma distribution, but your spanwise distributions will have strong kinks at the taper changes. If your computed Gamma distribution is discontinous, first double check the position of the surfaces to be sure they join. Then try increasing the number of panels near the taper change. While doing so, try to make sure that the panel sizes on both sides of the taper change are approximately equal. If the discontinuity in the Gamma distribution persists, try lowering the values of dissipation. Starting from a (low) angle of attack with many alpha steps can also be effective.
Partial-span flaps
At the spanwise position where the partial-spanned flap ends, you will have a very sharp gradient in Gamma, which may lead to a numerical overshoot. If the overshoot is small and in a limited spanwise region, the accuracy of your integrated coefficients should not be significantly affected. You can reduce overshoots using k2 and k4, but be careful not to use so much dissipation that the complete spanwise distribution is affected.