Computational Engineering / Diagnosing computational models
The checks that come first
Four things to do before reading any result, none of which takes longer than a minute.
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.
- Model Doctor lab case banks are separate from these lesson blocks and need their own full technical audit.
- Confirm mesh-quality limits and punching checks for each adopted element and design standard.
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
Start with independent estimates, equilibrium, the displaced shape and solver diagnostics. These checks are quick ways to identify discrepancies before relying on detailed results. Record the expected behaviour before interpreting the output, so that the check does not simply repeat the model’s assumptions.
By the end of this lesson you should be able to
- Run the equilibrium check and interpret the residual correctly
- Use the deflected shape as a qualitative test rather than as reassurance
- Compare quantities against the plausible bands
- Read a solver's warnings rather than dismissing them
What you should already know
- The stiffness method and where reactions come from (Module 9)
- Deflected shapes and load paths — Structural Behaviour teaches the qualitative reasoning this lesson uses
Equilibrium
Compare all applied forces and moments with all support and constraint reactions for the same load case, using consistent units and a stated normalised tolerance. A discrepancy needs investigation: load extraction, omitted reactions, solver convergence, conditioning and implementation can all be relevant. A balanced model can still have incorrect connectivity, loads or support conditions.
Displaced shape
Compare with an independently sketched expectation. Check direction, continuity, magnitude and, where the problem has a unique symmetric response, symmetry. Movement at a flexible or settling support may be intended. Read the display magnification before judging the apparent size.
Scale
Use independent calculations and comparable structures as screening checks. Typical building-floor frequencies or span/deflection ratios are not universal acceptance bands. Material stiffness depends on the material, temperature, cracking, creep and adopted design basis; steel values of 200–210 GPa and concrete values appropriate to its grade are examples, not automatic diagnoses. Check unit weight separately from mass density.
Solver diagnostics
Read every relevant warning and record how it was resolved. Severity labels and stopping behaviour are software-specific. A clean run does not establish validity, and a warning should be assessed against the actual formulation and model.
Worked example
Four checks on a model that looks fine
Given
- A six-storey frame, 32 m × 18 m on plan
- Design permanent plus imposed load at serviceability: 9.5 kN/m²
- Model reports total vertical reaction 32 780 kN
- Maximum floor deflection reported: 11 mm on a 9 m span
- First natural frequency reported: 6.8 Hz
- One severe warning: 'element 4471 length below tolerance'
Find
Whether the results can be used
Practice
A model of a 24 m × 15 m floor plate applies 7.5 kN/m² and reports a total vertical reaction of 2 610 kN. What percentage of the applied load is unaccounted for?
Check yourself
A model's total loads and reactions balance to within 10⁻¹¹ of the applied load, and the deflected shape looks entirely sensible. What has been established?
Worked example
Three minutes on a model you have never seen
Given
- A colleague sends a model and asks whether the results look reasonable
Find
What to do in the first three minutes
Check yourself
Why is looking at the deflected shape a genuine engineering check rather than a formality?
Summary
- Total loads against total reactions, every model, every time, before reading anything
- Predict the deflected shape before looking, or it is not a check
- Treat an unexpected magnitude as a reason to investigate the assumptions, inputs and response
- Read the warnings before the results, for the same reason
- Each check covers different issues; passing them does not establish complete model validity