Computational Engineering / Validation, verification and model assurance
The three-stage check
Look and think; check another way; check independently. In that order, because they cost different amounts and catch different things.
Lesson review: expert verification remains open
This lesson has received a full-context editorial review and a check of its technical explanations. Selected arithmetic has been checked separately. This is not independent engineering sign-off.
Checks needed for project use
- A computational engineering reviewer must check algorithms, convergence claims, statistical assumptions and numerical examples beyond the cases explicitly recorded.
- Confirm truss web layout and diagonal angle in the 24 m worked example; 63.4° is not established by the stated eight panels and 3 m depth.
- Formal checking categories and project acceptance criteria require the responsible engineer.
These are educational models and exercises. Check their assumptions before using the results. The full technical content and every diagram have not yet been independently verified.
Concrete and steel calculations retain their stated first-generation Eurocode teaching basis. Use the standards and amendments specified for your project; do not mix generations without an appropriate design basis.
Read the lesson register and outstanding review notesWhy this matters
Checking effort should be proportionate, and the way to make it proportionate is to order it by cost. Stage 1 is nearly free and catches most faults. Stage 2 costs an hour. Stage 3 costs days and is reserved for what warrants it.
Doing them out of order wastes the expensive ones on faults the cheap ones would have found.
By the end of this lesson you should be able to
- Carry out a Stage 1 qualitative check on an unfamiliar model
- Design a Stage 2 check that is genuinely a different route
- Say what Stage 3 requires to count as independent
- Match the depth of checking to the consequence of being wrong
Stage 1 — look and think
Costs minutes. Finds most faults.
- Displaced shape: right direction, right symmetry, right order of magnitude? Note the software's magnification factor — a small one means large real deflections.
- Moment or stress distribution: sensible pattern? Hogging over supports, sagging in spans, peaks where you expect them?
- Discontinuities: a jump in a quantity that should be continuous means a connectivity fault.
- Symmetry: a symmetric structure under symmetric load must respond symmetrically.
- Equilibrium: total loads against total reactions.
If something is wrong and you cannot find it, break the model into parts. Analyse a single bay, a single frame, a single floor. The fault is in one of them, and halving the model halves the search.
Stage 2 — check it another way
Costs an hour or two. The essential requirement is another way:
- A hand calculation on a simplified idealisation.
- A deliberately cruder model — two dimensions instead of three, a beam instead of a truss, a stick instead of a shell.
- A limiting case with a known answer: set a stiffness to zero, or to infinity, and check the model does what theory says it must.
- An energy or equilibrium check independent of the one already run.
A finite element model checked by another finite element model built by the same person from the same assumptions has been checked against nothing. The assumptions are what needed checking, and they were carried across intact.
The limiting-case check deserves more use than it gets. Make the columns infinitely stiff and the beam moments must approach the fixed-end values. Make the connection stiffness zero and the frame must become a set of simply supported beams. If the model does not do that, it is not modelling what you think.
Stage 3 — check independently
Costs days. Reserved for structures where being wrong is expensive or dangerous.
Independence has three components, and a check with only one of them is weaker than it appears:
- A different person, who has not seen the first model's assumptions.
- A different model, built from the brief rather than from the first model.
- A different tool, where the analysis method itself is in question.
The most valuable of the three is usually the first. A second engineer building from the brief will make different assumptions, and the places where the two models disagree are exactly the assumptions that were never explicit.
When two independent models disagree, the disagreement is the point. It has surfaced an assumption. Both engineers may be right about different questions — one used the short-term concrete modulus and one the long-term, and both were defensible until someone had to choose.
Proportionality
A scheme-stage sizing needs Stage 1 and a quick Stage 2. A transfer structure carrying six storeys needs all three. Deciding which, and recording the decision, is itself part of the assurance — because 'we did not do an independent check' is a legitimate position, and 'nobody considered whether to' is not.
Try it
Three-stage model check
Record what each check actually was. The workspace will not accept a stage marked done with no method recorded.
- Stages recorded
- 0 of 3
- Recorded without a method
- 0
Checking effort should be proportionate. A scheme-stage sizing needs Stage 1 and a quick Stage 2. A transfer structure carrying six storeys needs all three — and deciding which is itself part of the assurance.
Stage 1 — look and think
Minutes. Finds most faults.What did you look at? Displaced shape, moment distribution, discontinuities, symmetry, equilibrium.
Stage 2 — check it another way
An hour or two.What different route did you use? A hand calculation, a cruder model, a limiting case, an energy check.
Stage 3 — check independently
Days. Reserve it for what warrants it.Who or what was independent? A different person, a different model built from the brief, a different tool.
What makes each stage count
- Stage 1 must be predicted first. Looking at the result and then judging it is rationalisation, not checking.
- Stage 2 must be a DIFFERENT route. A finite element model checked by another finite element model from the same assumptions has been checked against nothing.
- Stage 3's independence is mostly about the person. A second engineer building from the brief makes different assumptions, and the disagreements are the assumptions nobody had made explicit.
What this shows: 'Checked' is not a record of anything. A check that cannot be named cannot be reviewed, and a reviewer cannot disagree with it.
Worked example
A limiting-case check on a frame model
Given
- A two-storey frame model, beams 8 m spanning between columns
- Uniform load of 25 kN/m on the beams
- The beam is prismatic and horizontal, under the stated UDL, with equal hogging end moments of 96 kN·m and a sagging midspan moment of 104 kN·m
Find
Whether the moment distribution is credible, by checking two limiting cases
Practice
A horizontal beam of 9 m span carries a full-span UDL of 20 kN/m and equal hogging end moments of 110 kN·m. What sagging moment should it have at midspan, in kN·m?
Check yourself
Which of these counts as a genuinely independent Stage 3 check of a transfer structure?
Check yourself
Why is the order of the three stages fixed?
Summary
- Stage 1 costs minutes and finds most faults — do it first, always
- Stage 2 must be a different route, not the same route repeated
- Limiting cases are underused and independent of every stiffness assumption
- Stage 3 independence means a different person, model and sometimes tool
- When two independent models disagree, an assumption has surfaced — that is the value
- Deciding how far to check is itself part of the assurance, and should be recorded