Verification and Validation Workflows
Verification checks that a code solves its equations correctly; validation checks that those equations describe reality. Both are needed to trust a prediction.
Two distinct questions
Trust in a simulation rests on two separate questions, often confused. Verification asks whether the code correctly solves the equations it claims to solve, a mathematics-and-software question. Validation asks whether those equations correctly describe the physical world, a physics question answered by comparison with experiment. A code can be perfectly verified and still invalid if its physics model is wrong.
A disciplined workflow keeps these separate and documents both, because a prediction is only as trustworthy as the weaker of the two.
Verification methods
Verification uses convergence studies, confirming the error shrinks at the expected rate as the mesh or timestep refines, comparison with analytic solutions, and the method of manufactured solutions, where a source term is engineered so a chosen exact solution satisfies the equations and the code's error can be measured directly. Code-to-code benchmarks add cross-verification.
Validation methods
Validation compares code output to experiment through synthetic diagnostics, on equal footing, and quantifies the agreement with metrics that account for both measurement and model uncertainty. A validated model states its domain of validity; extrapolation beyond it is flagged.
Design relevance
For the Hyperion breeder, verified codes validated against existing experiments produce the design predictions, while the honest caveat remains that final validation for this specific machine awaits its operation. No hardware net-gain claim is made before FOAK first tritium around 2030.
- Verification: solving the equations right
- Validation: the equations describe reality
- Convergence, manufactured solutions, benchmarks
- Synthetic diagnostics enable fair validation