03 / SAMPLE REPORT
One planted error. One second opinion.
A bolted end-plate beam-to-column connection modelled in a commercial solver. One node set was left incomplete on purpose. This is the report feaX produces after rebuilding the model, writing it for a second solver, running both on the same laptop and comparing.
Download both input filesFIGURES ILLUSTRATIVE UNTIL THE BENCHMARK SET IS PUBLISHED
1 / TRANSLATION
31keywords translated
1not translated · *CONTROLS, solver tuning, no effect on the model
0unsupported element types
2 / COMPARISON · TOLERANCE 5%
| QUANTITY | YOUR SOLVER | SECOND SOLVER | Δ | PASS |
|---|---|---|---|---|
| Max displacement, mm | 4.21 | 4.29 | +1.8% | yes |
| Sum of vertical reactions, kN | 96.0 | 160.3 | +40% | no |
| Max bolt force, kN | 118.2 | 121.7 | +3.0% | yes |
| Peak von Mises, end plate, MPa | 312 | 324 | +3.8% | yes |
| Initial rotational stiffness, kNm/rad | 41 200 | 40 100 | -2.7% | yes |
3 / MODEL CHECKS
OKUnits. Consistent N, mm, MPa. E = 210 000 MPa.
OKEquilibrium, second solve. Applied 160.0 kN; the second solver's reactions sum to 160.3 kN.
!Support set incomplete. NBASE holds 4 of the 16 nodes on the base face (line 212). The original model's reactions sum to 96.0 kN against 160.0 kN applied; the missing load is carried by an unintended path. Suggested fix: regenerate NBASE from the base face.
!Stiffness classification. S_j,ini = 41 200 kNm/rad classifies as semi-rigid under the boundary you named (EN 1993-1-8 §5.2.2, braced frame). The frame model assumes rigid.
OKMesh convergence. Peak stress moved 2.1% on one refinement around the bolt holes.
OKHand calculation. Component method: end-plate utilisation 0.85. FE: 0.87.
endplate_M20.inporiginal · line 208–216
208*NSET, NSET=NTOP
209113, 114, 115, 116, 117, 118, 119, 120,
210121, 122, 123, 124, 125, 126, 127, 128
211**
212*NSET, NSET=NBASE
2131, 2, 3, 4
214**
215*BOUNDARY
216NBASE, 1, 3, 0.0
endplate_M20_b.inpsecond solver · generated
1** feaX 0.x · from endplate_M20.inp (original solver)
2** 31 keywords translated · 1 skipped: *CONTROLS
3*NODE, NSET=NALL
…
187*NSET, NSET=NBASE
1881, 2, 3, 4
189** feaX check: NBASE covers 4/16 nodes of face BASE
190*BOUNDARY
191NBASE, 1, 3, 0.0
4 / BOTH INPUT FILES, YOURS TO RERUN
Nothing in this report requires feaX to reproduce. Run either file on your own installation and you get the numbers above.
WHAT A CHECK CAN AND CANNOT TELL YOU
CATCHESUnit mismatches. Incomplete or wrong supports. Mismatched node and element sets. Unconverged meshes. Solver-specific behaviour. Keywords that do not survive translation.
DOES NOT CATCHA wrong load case, a wrong material, a structure idealised the wrong way. It lists the assumptions it found so the engineer can judge them. Two solvers agreeing means the model was solved as written, not that it was the right model. That judgement, and the signature, stay with the engineer.