Simulation Does Not Replace Engineering Judgement
Software will happily solve the wrong problem to twelve significant figures. The engineer's job is deciding which problem is worth solving — and knowing when the answer is implausible.
Modern analysis tools are extraordinary. They are also indifferent to whether the problem you set them is the problem you have.
The failure is upstream
A solver applies boundary conditions faithfully. It does not know that your inlet profile came from a different operating point, that the geometry was defeatured past the point of relevance, or that the flow you modelled as steady is shedding vortices in reality.
None of these produce an error message. They produce a result — plausible, well-resolved, and wrong.
Estimate before you compute
The most valuable habit in analysis is forming an independent expectation first. A hand calculation, a scaling argument, a comparison against a similar published case — anything that gives you a number to be surprised by.
Without a prior expectation, every output looks reasonable. With one, an answer off by a factor of three announces itself immediately, and you investigate instead of publishing.
This is not distrust of the tool. It is the only mechanism by which a setup mistake becomes visible.
Interrogate the convenient result
There is a specific moment worth watching for: the result that makes your life easier. The margin that clears by 3%. The design that meets requirement without the change you were dreading.
That is exactly when scrutiny tends to relax, and exactly when it should intensify. Convenient results deserve more checking than inconvenient ones, because nothing prompts you to look again.
State assumptions where they can be found
Every analysis rests on assumptions. Recorded in a report appendix nobody opens, they might as well not exist. The engineer reading your work in two years — possibly you — needs to know what was assumed and where it stops being valid.
The most useful sentence in a technical report is often: "This analysis is not valid if…"
What the tools changed and what they did not
They removed the arithmetic barrier. Problems that were intractable are now routine, and that is a genuine expansion of what engineering can do.
What has not changed is the requirement to decide what to model, what to neglect, and whether to believe the answer. Those judgements were always the substance of the discipline. Faster solvers make them more consequential, not less — because it is now possible to be wrong at far greater scale and with much better graphics.