Module 17 · Course notes
Modelling and checking: is the result even possible?
An analysis is only as good as the model behind it. A program will solve exactly the structure you gave it — including the wrong one, perfectly. The skill this module builds is the audit: a short list of independent checks that catch an impossible result before it becomes a design, no matter how cleanly it converged.
A model is a set of choices: which supports, which releases, how members connect, what element types. Each choice can be wrong, and a solver cannot tell you it solved the wrong structure. So every result is a claim to be tested, and the tests are the ones this whole course has built — read independently, not from the plotted numbers.
Start with the cheap, powerful checks: do the reactions balance the loads? Is the moment zero at every hinge? Does a symmetric structure under symmetric load give a symmetric result? Is there frame action where you expect it? Is the deflected shape credible, at a sensible scale?
The common modelling errors each leave a signature you can learn to read:
- A missing restraint or a wrong release — moment at a hinge, or a base moment where the detail is a pin. - A frame modelled as a truss (all joints pinned) — no moment transfer and excessive sway; the members show axial force only where you expected bending. - Broken symmetry — a symmetric model giving asymmetric reactions points to a stray release, a local-axis slip, or an uneven mesh. - A misleading deformation scale — auto-scaling makes a hairline movement look like a collapse, or hides a real one; always check the magnitude, not just the picture.
Worked example
Worked example — auditing a suspicious result
A plotted result looks tidy. Run it past the audit before trusting it.
Step 1 — totals and hinges
First check the reactions sum to the applied load and that the moment is zero at every hinge. A failure here is decisive: the model is wrong, whatever the plot looks like.
Step 2 — symmetry and behaviour
Then check behaviour: a symmetric frame under symmetric gravity should not sway, and its columns should bend as rigid joints transfer moment. If the result sways, or shows axial-only columns, suspect a release or a frame-as-truss error, and explain it before accepting the numbers.
In the exercises, run the audit: reactions, hinges, symmetry, frame action and scale — and decide whether each result can be believed.
How to read these problems
The three-step method
- 1Points of certainty. The deflected curve must pass through every support and deflect downward under the load. Mark what each support prevents before drawing anything.
- 2Deflected shape and reaction directions. Sketch the compatible deflected shape. To find a reaction's direction, imagine removing that support: the direction that pushes the structure back to its place is the reaction's sense (it may be a hold-down).
- 3Bending moment and contraflexure. Draw the bending-moment diagram on the tension side and check it against the shape: hogging where the curve is convex-up, sagging where convex-down, zero at pins and at every contraflexure.
Rules that must always hold
- 1.The bending moment is zero at a simple support and at an internal pin or hinge.
- 2.A bending-moment diagram crosses the baseline exactly at a point of contraflexure.
- 3.Under a distributed load the bending-moment diagram is curved; under point loads alone it is straight lines.
- 4.At a fully fixed support the deflected shape leaves the support with zero rotation (tangent along the member).
- 5.If a part of the structure stays straight after loading, it carries no bending moment there.
- 6.The moment is drawn on the tension side: sagging below the member, hogging above it.
Now predict for yourself
Modelling and software checking exercises
5 exercises on reactions that do not balance, moment at a hinge, a frame modelled as a truss, broken symmetry and misleading deformation scales. Predict, then reveal an explanation.
Start the exercises →This lesson is educational material. It uses simplified examples to teach principles, and must not be relied on for real design or safety-critical decisions.