Verification and Validation

Verification asks whether the code solves the equations right; validation asks whether the equations describe reality.

Two questions, deliberately separated

Verification and validation (V&V) are distinct disciplines that are easy to blur. Verification checks that software correctly solves the mathematics it claims to solve — via convergence studies, method-of-manufactured-solutions, and code-to-code comparison. Validation checks that the mathematics matches physical reality — via comparison to experiment. A code can be perfectly verified and completely invalid, and vice versa.

How Kronos applies it

• Every reported public number is reproducible from committed code and inputs before it becomes a claim.
• Solvers are verified for convergence and, where possible, against analytic or manufactured solutions.
• Models are validated against published experimental benchmarks from the wider fusion community.
• Results are deposited with a permanent identifier so a third party can re-run them.

Reproducibility as the floor

Kronos treats reproducibility as the minimum bar for a technical claim: a two-tier repro discipline (byte-for-byte where deterministic, tolerance-bounded where not) means an outside reviewer can obtain the same figure from the same inputs. This is what lets the program publish a physics basis openly and invite scrutiny rather than ask for trust.

V&V does not turn a simulation into a hardware result. It makes the simulation credible enough to justify building the hardware that will test it.