Verification and Validation of Simulation Codes

The distinct disciplines of solving the equations right and solving the right equations, applied to every Kronos workload.

Two questions, not one

Before any simulation result is trusted, it must pass two separate tests. Verification asks whether the code solves the equations correctly. Validation asks whether those equations describe reality. They are different questions, and Kronos treats them as distinct disciplines, because a perfectly verified code can still model the wrong physics, and correct physics can still be coded wrong.

Verification

Verification checks the numerics. It compares code output against analytic solutions where they exist, confirms that error shrinks at the expected rate as the mesh or time step refines, and checks conservation of mass, energy, and charge. A neutronics code is verified against known benchmarks; a finite-element solver is verified against manufactured solutions with a known answer.

• Comparison against analytic and benchmark solutions
• Convergence at the expected order under refinement
• Conservation of mass, energy, momentum, charge
• Code-to-code cross-checks between independent solvers

Validation

Validation checks the physics against experiment. Here Kronos faces a real constraint: the machines are not built, so there is no breeder or burner data to validate against directly until FOAK around 2030. Kronos therefore validates its codes against existing experiments from the wider field and against well-characterized reference cases, establishing credibility on known ground before applying the codes to its own designs.

This limit is stated honestly. Until first tritium, Kronos's machine predictions are extrapolations from validated codes, not validated machine results, and no hardware net-gain is claimed before FOAK. Verification can be complete now; full machine validation cannot, and the architecture is built to fold real pulse data into validation the moment it exists.

V and V close the loop on the whole foundation. They gate what results enter the archive as certified, what data trains twin models, and what surrogates are promoted to real time. Combined with reproducibility and UQ, they are what let Kronos treat its simulations as a serious design basis for both machines rather than as illustrations.