You’ve got a floor slab in your building model. Do you use membrane elements? Plate elements? Shell elements? The answer isn’t “whatever the software defaults to” — it depends entirely on what you’re trying to analyse. Pick the wrong element type and you’ll either miss critical behaviour or drown in unnecessary output.

A buried 24-inch diameter pipeline running through a landfill on soft ground. Differential settlements measured at monitoring points spaced 15 to 25 metres apart. A dispute over who’s responsible for the damage. This case eventually made it to court, and the engineering analysis at the centre of it all contained fundamental errors that any engineer working on buried infrastructure needs to understand.

You’ve got a complex structure on your desk, a looming deadline, and powerful analysis software at your fingertips. The temptation? Import the CAD geometry, auto-mesh everything, hit run, and let the computer do the heavy lifting.

You’ve built the model, applied the loads, and the deflected shape looks reasonable on screen. The animations are smooth, and there are no obvious instabilities. But a “pretty” model isn’t necessarily a correct one. In finite element analysis (FEA), the quality of your mesh directly dictates the quality of your design forces. A mesh that is adequate for checking drift might be dangerously unconservative for checking shear in a coupling beam.

Most of the time, our building beams are slender enough that shear deformation is an afterthought. We worry about flexural stiffness, slap on a deflection limit, and move on. But as soon as you start working with deep transfer girders, perimeter spandrels, or short plate girders, ignoring shear deformation can quietly under-predict drift and member demand.

How many times have you been up against a deadline and just idealized a connection as “pinned” or “fixed” to get the analysis to run? For decades, this simplification has been the bedrock of frame analysis, a necessary shortcut when we were armed with little more than slope-deflection and moment distribution. But in today’s world of performance-based design and powerful FEA software, clinging to this binary view isn’t just conservative—it can be inaccurate and uneconomical.