Skip to content

Add closed-form tests for shear-deformable members - #341

Open
Ayberkrk wants to merge 1 commit into
JWock82:shear_deformationfrom
Ayberkrk:test-shear-deformable-members
Open

Ayberkrk wants to merge 1 commit into
JWock82:shear_deformationfrom
Ayberkrk:test-shear-deformable-members

Conversation

@Ayberkrk

Copy link
Copy Markdown
Contributor

Relates to #250 and the review of #289 on the shear_deformation branch.

What this adds

Testing/test_shear_deformable_members.py checks shear-deformable members against closed-form Timoshenko beam solutions: the Euler-Bernoulli bending term plus the shear term from integrating V/(kGA) along the member. A timber-like section (E/G = 16, 0.2 m x 0.6 m, k = 5/6) is used so the shear term is a large share of each result.

Passing checks (all agree to rel_tol=1e-9):

  • Cantilever tip load, strong axis: PL^3/(3EI) + PL/(kGA), at the node and through member.deflection() at x = L.
  • The same cantilever with shear_deformable=False reduces to Euler-Bernoulli.
  • Minor axis cantilever, confirming that Iy and ksz are paired.
  • Simply supported uniform load, midspan node: 5wL^4/(384EI) + wL^2/(8kGA).
  • Fixed-fixed uniform load: midspan deflection wL^4/(384EI) + wL^2/(8kGA), and end moments wL^2/12.
  • Propped cantilever under uniform load: the prop reaction from compatibility including shear, which is larger than the Euler-Bernoulli 3wL/8.

These cover the modified stiffness matrix, the fixed-end reactions, and a statically indeterminate case, so they should help with the review of #289.

Interior deflections omit the shear term

Two tests are marked xfail(strict=True) because they fail on this branch. Between nodes, member.deflection() returns the Euler-Bernoulli value for a shear-deformable member. For the section above, the midspan deflection of a single simply supported member is underestimated by 38% under uniform load and 43% under a midspan point load; for a fixed-fixed member it is 75%. Deflections at nodes are correct.

The cause appears to be that BeamSegZ.deflection() and BeamSegY.deflection() integrate curvature only, and _segment_member() infers the starting slope theta1 from the Euler-Bernoulli slope-deflection relation (Member3D.py around line 2685). The end conditions are therefore met, but the shape between them is the Euler-Bernoulli shape. For a shear-deformable member the missing term is (M(x) - M1)/(kGA) within a segment, and theta1 would need the matching correction. I did not change this here because it touches the segment classes and the P-Delta path; I am happy to open a follow-up PR if you would like that approach. With strict=True, these two tests will start failing as unexpected passes once the behavior is fixed, as a reminder to remove the markers.

Testing

Full suite on this branch: 91 passed and 2 xfailed (85 passed before this change).

The expected values are written out in the tests. While preparing them I also cross-checked the cantilever and simply supported cases against timoshenko-engine, a small library I maintain, which gave identical values. The tests do not depend on it.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant

⌂ 换网址