You don’t need to run an FEA to know whether the one you’ve been handed is any good. You need to know what’s supposed to be in it, and what its absence means. This is written for the person on the receiving end of an analysis report — procurement, project management, a client’s engineer who didn’t run the model — who needs to sanity-check a vendor’s work without redoing it.
Start with the assumptions, not the conclusion
Every FEA report should open by stating, explicitly, what it modeled: the geometry (and any simplifications made to it), the material properties used and their source, the boundary conditions, and every load case checked. This section is often skimmed past to get to the result — that’s backwards. A perfect-looking stress plot built on the wrong boundary conditions or an unstated load case is a confident wrong answer. If you can’t tell from this section exactly what was and wasn’t checked, that’s the first red flag, before you’ve looked at a single number.
Check for a mesh convergence study
FEA results depend on mesh density — too coarse a mesh in a high-stress region and the reported peak stress can be meaningfully wrong, usually understated. A convergence study re-runs the analysis at progressively finer mesh density in the regions that matter and shows the result stabilizing, proving the reported stress isn’t just an artifact of how the part happened to be meshed. No convergence evidence means you’re trusting that the mesh was fine enough, with nothing to check that trust against. This is a specific, checkable thing to ask for if it isn’t already in the report.
Read the margin summary, not just the color plot
A contour plot showing mostly blue with a small patch of red looks reassuring, but color scales are relative to whatever maximum the software picked for that image — they’re not a substitute for numbers. What you actually want is a tabulated summary: calculated stress (or strain, or fatigue damage) against the allowable, for every load case, with the margin stated as a number. If a report only shows plots and never states “calculated X against allowable Y, margin Z%” anywhere, ask for the table. It’s normally already been produced internally to reach the conclusion — it just hasn’t made it into what you were handed.
Confirm the result is checked against a named code clause
“The stress is acceptable” isn’t a conclusion — it’s a claim that needs a citation. A defensible report states which code and which specific clause the result is being checked against: ASME BPVC Section VIII Division 2 Part 5’s protection-against-plastic-collapse criteria, a specific API 6X allowable, a named fatigue curve. If the acceptance criteria aren’t traceable to a specific code reference, there’s no way for a reviewer — you, your own engineers, or a certifying body — to independently check whether “acceptable” was applied correctly.
The short version
A report you can trust states its assumptions plainly, shows its mesh is fine enough to believe, gives you numbers instead of just colors, and ties its conclusion to a named code clause. If any one of those is missing, that’s not necessarily proof the analysis is wrong — but it is proof you can’t currently verify that it’s right, and that’s worth pushing back on before the design moves forward on the strength of it.
Every analysis we deliver includes the full trail: stated assumptions, mesh convergence evidence, a tabulated margin summary, and results checked against named code clauses. See our FEA & Simulation services, or talk to an engineer if you have a report you’d like a second opinion on.