Reviewing a calculation

This page is for the engineer on the other side of the table: the one who did not build the model and has to decide whether to sign the check. It assumes you have a PDF calculation package, and possibly read access to the live model.

What you are looking at

An Elementarium report is generated from the solved model, not assembled by hand. Two consequences:

• Nothing in it can disagree with the model as it stood when the report was exported. The combination table, the reactions and the utilisations are the model's own values, printed.
• It is a snapshot. The live model may have moved on since. Check the export date against the revision you are reviewing, and if you have model access, open the named version the report was issued from rather than assuming the live model is the same. Review points can be left as comments pinned to the members they concern, and marked blocking so the next export warns while they are open.

The report's five chapters — project and design basis, structural model, loads, analysis results, design verification — are described in Exporting a report.

Ask for read access

If you can, ask for read access to the project rather than a second PDF. A reader can open every result, table and check, switch combinations, amplify the deflected shape and export tables — and cannot change anything. It removes the "can you send me the model" cycle and the risk of reviewing a revision that is two behind. See Sharing a project.

Sharing requires the author to be on a paid plan. If they are not, the PDF is what you have, and the sections below still apply.

The questions, in the order they pay

1. What was assumed?

Before any number: the design code and national annex, every parameter overridden from the annex default, the combination rules used (6.10, or 6.10a/b), the support idealisations and the reasoning behind them, and the element sizes. All of these should be stated in the report; if they are not, that is the first comment.

The national annex deserves particular attention. The ${\gamma }_{M0}$ it sets divides every cross-section resistance, and a wrong annex moves every utilisation by several percent with nothing looking obviously wrong.

2. Does the model represent the structure?

• Deflected shape. With model access, amplify it and look. Members moving independently of neighbours are not connected; supports that move are not what the report says; asymmetry under symmetric load means the restraints are not what the author thinks. Without model access, the report's deformed-shape figures are the substitute, per combination.
• Supports. Pinned, fixed or spring — and which is conservative for the effect being checked is not always obvious. Supports and releases has the table.
• Releases. Where the drawings show a simple connection, is there a release? Where they show a moment connection, is there not?
• Sections and materials. The member schedule and the materials table are in the model chapter. $E$, density and yield strength are the three to check against the specification.

3. Does the load reach the supports?

The reactions table is in the analysis chapter. Sum the vertical reactions for the self-weight case and compare with the structure's weight; sum them for each load case and compare with the applied load.

Your first analysis describes the same check from the author's side.

4. Are the combinations right?

The combination table in the loads chapter lists every combination with its partial factors. Look for:

• The count is plausible for the number of variable actions and arrangements.
• The favourable permanent variants exist where uplift or overturning could govern.
• Exclusive actions — wind directions, load arrangements — never appear together.
• The SLS set used for deflection is the right one for what is being protected: characteristic, frequent or quasi-permanent.
• The set was regenerated after the last load case was added — a case missing from every combination is the symptom.

Automatic combinations and ASCE 7 combinations list the rules the generator applies.

5. Is the mesh fine enough?

For a model with surfaces, the report should state the element size and that a convergence check was performed, with its result — "halving the element size changed the governing moment by 1.4%". If it does not, ask. A peak moment at a column in a slab is a singularity and should have been read at the column face or integrated over a strip, not at the node.

6. One member by hand

Pick the simplest span — a purlin, a secondary beam — and check it:

You are not checking precision; you are catching a factor of a thousand in a unit, a load over the wrong length, or a section an order of magnitude off. Verifying the solver has the formulas.

7. Only then, the utilisations

Now the design chapter. Every utilisation names the governing combination, the position along the member, the clause and the substituted formula with the intermediate values — section class, resistances, factors. Open the governing one fully and follow the arithmetic. If you disagree, you can now point at the step rather than at the number.

Read two things alongside the ratio:

• The second-worst combination. Where two very different combinations both land near 1.0, the design is sensitive to the load model.
• The messages on the check. Each module raises what it did not verify — load height in $M_{cr}$ and mono-symmetric torsional buckling for the Eurocode module, the net-section, tee, angle, channel and built-up cases for AISC — as a message on the check itself, and in the assumptions and limitations section. Whatever a message names needs that check done somewhere, and the report should say where.
• The check's module. A check created before the member-resistance modules may still run a cross-section-only module, where a utilisation of 0.83 on a slender column says nothing about buckling. The module name is on the check and in the report.

What the software does not do, and the report cannot claim

None of this is hidden: the check messages and the assumptions section state it. The review question is whether the engineer who signed the package dealt with each one.

Getting numbers out for your own check

Result tables copy to the clipboard on any plan, and on Pro export with their filters to CSV or xlsx — Result tables — so the reactions or the member extremes can go straight into your own spreadsheet. For an independent solve, ask the author for the model as a SAF workbook, which SCIA Engineer, RFEM and others read.

Related

Exporting a report describes the package from the author's side, with the pre-issue checklist that this page mirrors.