Verification
Which answers in Structural have been checked against an independent published source, what exactly is compared — and what has not been checked yet.
How the checks run
Every check on this page is an automated test that runs through the same code you use in Structural — the same model checks, the same analysis preparation and the same solver. The app is not deployed unless every one of them passes, so a change that moves an answer away from its source is caught before it reaches you.
Steel member design — EN 1993-1-1
28 worked examples from published design guides — 294 quantities and 209 section properties, each with its page or sheet number recorded beside it:
| Source | Examples | Sections |
|---|---|---|
| SCI P364, Steel Building Design: Worked Examples — Open Sections (2009) | Examples 2–5, 7, 9–13 | UB/UC beams, columns and beam-columns |
| SCI P374, Steel Building Design: Worked Examples — Hollow Sections (2010) | Examples 1–7 | SHS/RHS members, EN 10210 hot finished |
| BCSA/SCI Handbook of Structural Steelwork, Eurocode edition (55/13, 2013) | Examples 2.1, 2.2, 3.1, 5.1–5.4 | UB/UC members |
| Designers’ Guide to Eurocode 3 (Gardner & Nethercot) | Example 6.7 | CHS column |
The app is given only what the book was given — the section’s dimensions, the grade, the lengths between restraints and the design forces. It works out the section properties, the cross-section class and the buckling curve itself, runs the check, and compares:
- Section properties — area, second moments, elastic and plastic moduli, shear areas, torsion and warping constants — within 1.5 % (2 % for shear areas, 4 % for torsion constants, which the tables round more coarsely).
- Classification — the class, exactly.
- Resistances — Nc,Rd, Mc,Rd, Vpl,Rd, Mcr, χ, Nb,Rd, χLT, Mb,Rd, the interaction factors kij and equations 6.61 and 6.62 — within the tolerance the book’s own rounding justifies, normally 1–2 %.
- The buckling curve — the book’s imperfection factor α against the one the app selected, exactly.
Twenty figures in the books are worked by a method the app does not use — SCI P362’s simplified lateral-torsional slenderness, a Table 6.6 correction factor read for a named moment shape, and EN 1993-1-5’s reduced plate slenderness in one class 4 example. Those are printed beside the app’s answer on every run, with the reason, but not asserted.
What the comparison found
When the suite was first run it caught four errors in code that already had passing unit tests: hollow-section corner radii ignored in the section properties; the §6.3.3 “susceptible to torsional deformations” decision taken from χLT instead of the section; the I-section row of Annex B Table B.1 used for rectangular hollow beam-columns; and the cold-formed buckling curve used as the default for hot-finished hollow sections. All four are fixed, and the examples keep them fixed.
Analysis — published verification examples
Each model is built from the published document (no model file is published with them) and analysed exactly as your model is:
| Check | Source | What is compared |
|---|---|---|
| CSI verification example 1-001 | CSI Analysis Verification, Frames, Problem 1-001 — published answers and an independent hand calculation (Cook & Young) | A three-member frame under seven load types — self weight, global uniform, joint force and moment, projected, local uniform, local trapezoidal and local point loads. Load cases 1 to 5 agree with the published displacements to all five printed digits. |
| CSI verification example 2-002 | CSI Analysis Verification, Shells, Problem 2-002 — MacNeal & Harder’s straight cantilever | A cantilever modelled as shell elements under six tip loads, in and out of its plane, against the published answers and beam theory. |
| CSI verification example 2-005 | CSI Analysis Verification, Shells, Problem 2-005 — Timoshenko & Woinowsky-Krieger | Rectangular plate bending: both edge conditions, both aspect ratios, coarse and fine meshes, at a span-to-thickness ratio of 10,000. Every fine-mesh answer is within 0.3 % of the closed form. |
| CSI verification example 2-006 | CSI Analysis Verification, Shells, Problem 2-006 — the Scordelis-Lo barrel vault (MacNeal & Harder) | A curved concrete shell roof under its own weight: −0.3109 ft at the centre of the free edge against the published −0.3086 ft, and the same answer whether the load is applied as self weight or as an area load. |
Analysis — exact solutions
These need no reference program at all: the right answer is a formula, or a balance of forces that must hold whatever the model.
| Check | Source | What is compared |
|---|---|---|
| Buckling and P-Delta | Euler; Timoshenko & Gere | Euler’s critical load for all four ideal end conditions; a pinned-base portal frame buckling in sway; Timoshenko’s amplification under P-Delta; and the P-Delta analysis refusing a load exactly at the buckling solver’s critical load. |
| Modal and response spectrum | Euler–Bernoulli beam theory | The first four frequencies of a simply supported beam and the first three of a cantilever; how the participating mass is distributed between modes (1 : 1/9 : 1/25); base shear per mode under a response spectrum; and mass participation that sums to exactly one over a complete set of modes, including the rotational ratios. |
| Equilibrium checks | Statics | Reactions balance the applied load for springs and area springs, for supports on nodes the mesh adds, for rigid links, for notional loads and for loads applied at joints on a plate’s edge; and a plate’s stiffness modifiers act as the closed form says — a bending factor of 0.25 makes a cantilever strip exactly four times as flexible. |
One solver, three places
The browser, the cloud engine and the desktop app run the same solver, built from the same source. Tests solve the same models with the browser build and with the native build and require the answers to agree to within 10−7 of each quantity’s own scale.
Not covered yet
- AISC 360 steel design has no published-example suite yet; its checks rest on hand-checked tests.
- Concrete design (EC2, ACI 318) and timber design (EC5, NDS) have no published-example suite either.
- Tees are refused by the EC3 member check, so no example covers them.
- The Quick Design and Connection Design calculators and SeismoLab are separate tools and are not part of these suites.
Whatever the checks cover, you remain responsible for the design: check the model and the results as you would any engineering calculation.