A practical checklist for reviewing someone else's FEA report

Divya DlaminiRegisteredAircraft Structural Engineering · 20 yrs · Gothenburg, Sweden

1 week ago3.1K views15 replys

139

Being asked to review an FEA report you did not produce is common and uncomfortable. Pretty contour plots tell you almost nothing. Here is the order I work through, roughly by how often each one turns up a real problem.

First, is there a mesh convergence study? If the report shows one mesh and one answer, you have a number, not a result. Second, do the boundary conditions represent the physical constraint or the convenient constraint? A fully fixed face where the real part is bolted through three holes will under-predict deflection and over-predict stress at the wrong place.

Third, check the reaction forces against the applied load. If they do not balance, something is wrong with the load application and nothing downstream matters. Fourth, look at where the peak stress is. If it is at a singularity — a sharp re-entrant corner, a point load, a constraint node — the peak value is meaningless and the report should say so.

Fifth, is the material model appropriate? Linear elastic results above yield are not wrong exactly, but they are being read wrong if the report draws a conclusion from a stress above the yield strength without saying so. Sixth, does the conclusion follow from the results, or does it follow from what the client wanted to hear?

Standards cited

Clauses are referenced, not reproduced. Consult the current published edition of each standard before relying on any interpretation in this thread.

3 Replies

Mei-Ling FaroukDie Engineering · 10 yrs

Good list, and the ordering by hit rate is the right way to present it. I would add one check near the top, ahead of the mesh study: does the report state what question the analysis was supposed to answer, and does the model type match it? A linear static analysis submitted to justify a part against a vibration durability requirement is wrong before any mesh consideration applies, and it is a surprisingly common submission because the static model already existed. Establishing the intended question first also tells you which of your later checks actually matter — mesh convergence on peak stress is critical for a fatigue assessment and close to irrelevant for a global stiffness check.

On your fourth point about singularities, the useful follow-up for a reviewer is to ask for the stress at a defined distance from the singularity rather than to reject the result. If the analyst can produce stress at a consistent offset across two mesh densities and the numbers agree, the result is usable even with a singular peak in the plot. If they cannot, the peak is all they have and the conclusion does not stand. That reframes the comment from an argument about meshing philosophy into a specific, answerable request.

Omar WatanabePredictive Maintenance Engineer · 19 yrs

Verified expertise in Predictive Maintenance Engineer

Adding the check that has caught the most real errors in my review work: units and the magnitude of the answer. Before looking at anything sophisticated, take the applied load, divide by an approximate load-carrying area, and compare against the reported stress. If they differ by an order of magnitude, you have found the problem and it is usually a unit system inconsistency — density in the wrong units is the classic, because it only shows up in gravity or dynamic results and leaves static results looking plausible. It takes two minutes and it precedes every other check on your list.

The related one is a deflection sanity check against a hand calculation. Almost every real structure has a beam, plate or ring idealisation that gets you within a factor of two of the deflection. If the model and the hand calculation disagree by much more than that, either the boundary conditions are not what the reviewer thinks or the idealisation is wrong, and you want to know which before reading a stress plot. A report that includes its own hand-calculation comparison is, in my experience, close to a reliable indicator that the rest of the work was done carefully.

Aisha MehtaManufacturing Engineering · 25 yrs

Your sixth point deserves to be first rather than last, because it changes how you read everything else. When a conclusion has been reached before the analysis, the tells are consistent and visible: a mesh study that stops at the density which gives the acceptable answer, a boundary condition chosen without justification that happens to relieve the critical location, a material allowable quoted without a source. None of these is individually damning. Together they tell you the analysis was built to support a conclusion, and the review comment has to address that rather than the individual items.

The constructive way to write that comment, since a reviewer accusing an analyst of motivated reasoning goes nowhere useful, is to ask for the sensitivity. Request the result with the next mesh refinement, with the bolted boundary condition rather than the fixed face, and with the minimum specified material property rather than the typical. If the conclusion survives all three, it was sound and the presentation was just thin. If it does not survive any of them, the sensitivity study has made the point for you without anyone having to make an accusation.

Your answer

State your assumptions and cite the clause you are relying on. Answers that show the reasoning get endorsed by engineers in the discipline; answers that only state a conclusion do not.

Do not paste client-confidential drawings or data.