Repository navigation
BUG: Incorrect vector rotation in GenericSurface.compute_forces_and_moments #932
Description
Describe the bug
Incorrect vector rotation in GenericSurface.compute_forces_and_moments().
Here, we want to represent the black L, D, Q vectors in the blue body frame.
The M_BA matrix should be direction cosine matrix (DCM), however it is given in vector rotation matrices. Vector rotation matrices rotate the vector without changing/rotating reference frame.
Current implementation rotates the force vectors to match body axis, but results are however presented in aero frame.
If it is intended to use rotation matrix instead of DCM, the rotation should be:
- Start with the
Q,L,Dvectors matching+X,-Y,-Zin body frame (i.e. Vector([side, -lift, -drag]) ) - Rotate the vector by
-alphaon X axis andbetaon Y axis. - The results will match the black
Q,L,Dvectors and presented in body frame
Another POV:
- Only focus on blue Zb axis and body frame.
- Rotate the Zb axis by
-alphaon X axis andbetaon Y axis. - the result will match the black V vector
To Reproduce & Expected behavior & Screenshots
If applicable, add screenshots to help explain your problem.
Additional context
Add any other context about the problem here.
Activity
ting-hong-shieh commentedon Aug 14, 2026
I reduced this to three unit force vectors at
alpha = 0.2 radandbeta = 0.1 rad. Currentdevelopiscb6106a717207dd8fc2dfe1446d80ff75022f21b.For the aerodynamic vector
[Q, -L, -D], the current code applies the positive-alpha/negative-beta rotation. Applying the direction-cosine mappingRx(-alpha) @ Ry(+beta)gives:Q=1)[0.995004165, -0.019833838, 0.097843395][0.995004165, -0.019833838, -0.097843395]L=1)[0, -0.980066578, -0.198669331][0, -0.980066578, 0.198669331]D=1)[0.099833417, 0.197676812, -0.975170327][-0.099833417, -0.197676812, -0.975170327]Draft PR #1064, head
84d3b6a9f0b4a8074d764130dc3d189bdc402036, produces the DCM column for all three cases through its wind-to-body coefficient conversion. That makes it an implementation candidate for this issue, although #1064 contains a much broader aerodynamic refactor.Would you prefer to keep this fix and its three basis-vector regression cases in #1064, or split the rotation correction into a smaller PR against
develop? I have not opened a duplicate while #932 is assigned and #1064 is active.Environment: Python 3.12.6; NumPy 2.5.2; macOS 26.5.2 arm64.
Gui-FernandesBR commentedon Aug 14, 2026
@ting-hong-shieh let's wait until #1064 is merged please
stephanLouw commentedon Sep 17, 2026
@ting-hong-shieh Does this bug mean that using Generic surface in the current PyPi version (1.13.0), will lead to a wrong flight prediction?