CAE & automation / Field guide 01
A simulation validation checklist
A useful model review separates three questions: did we solve the equations as intended, do those equations represent the physical system, and is the result adequate for the decision?
1. Record the decision
Write down the output, operating range, acceptable error and consequences of a wrong answer. A model meant to rank heat exchanger concepts needs different evidence from one used to certify a maximum component temperature.
2. Fix the inputs and assumptions
Record geometry revision, material properties, boundary conditions, correlations and control logic. Identify measurements used for calibration separately from measurements held back for validation.
3. Check numerical behavior
Track residuals and conservation errors. Repeat key cases with tighter tolerances or a refined mesh and time step. Look at the engineering quantity of interest, not only solver convergence.
4. Compare with independent evidence
Plot predicted versus measured outputs across the operating envelope. Include measurement uncertainty and report bias and outliers. A match at one condition is not evidence of broad predictive performance.
5. Test sensitivity
Vary uncertain inputs within defensible ranges. Document which assumptions change the decision and which do not. Flag extrapolation outside the validated range.
Review record
Decision and acceptance criterion\nModel version and input sources\nNumerical settings and independence check\nCalibration cases and held-out validation cases\nError and uncertainty by operating condition\nKnown limitations and approval
This checklist is a starting point for a reviewable report. Adapt thresholds and validation cases to the risk of your application.
Get future workflows