I’ve been on parental leave recently, which apparently means my brain has had time to wander into places it normally doesn’t.
One thing I keep coming back to: what actually happened to structural engineering when FEA software arrived?
From what I can tell, it wasn’t the catastrophe some probably feared. Engineers didn’t disappear. But the job shifted. The premium stopped being on who could grind through a stiffness matrix by hand, and started being on who could build a sensible model, interrogate the outputs, and know when something smelled wrong.
Most of us can’t derive the full FEA formulation from scratch today. I’m not sure that’s a problem. We trust the software, we understand the principles well enough to sanity check the results, and we know what “back of envelope” should roughly give us.
I wonder if AI is just the next version of that same transition, not a fundamentally different one. The workflow might look like: rough hand calc, AI-generated design, engineer sanity checks the output. Faster overall, probably better optimised, but still requiring someone who genuinely understands the physics.
The engineers I’d be more worried about aren’t the ones who understand structures deeply. They’re the ones whose primary value is knowing how to operate the software efficiently. That was probably true after FEA too, we just didn’t talk about it as much.