Methodology
How we know a simulation is right.
Simulation codes will produce beautiful plots whether or not they are correct. Our verification & validation (V&V) process makes sure yours are both.
- 1
Analytical benchmarking
Before modelling your problem, we reproduce problems with known answers in the same code and configuration: linear dispersion relations, Landau damping rates, two-stream growth rates, sheath and Child–Langmuir solutions.
- 2
Convergence studies
We refine grid spacing, time step and particles per cell until the quantities you care about stop changing, and we report the remaining numerical uncertainty.
- 3
Conservation diagnostics
Total energy, momentum and charge are tracked throughout every run. Unphysical heating, noise or drift is caught early, not after publication.
- 4
Cross-code & cross-method checks
For critical results we compare against a second code, a reduced model (fluid, MHD, Vlasov) or experimental data, to separate physics from numerical artefacts.
- 5
Reproducible reporting
You receive input decks, versions, scripts, notebooks and a verification report, so every figure can be regenerated by you, a reviewer or your future students.
Verification ≠ validation
Verification asks “are we solving the equations correctly?” Validation asks “are these the right equations for reality?” We address both, and we tell you clearly which is which.
The same standard for AI
Surrogate models and AI-generated code go through the same scrutiny: held-out tests, physical constraints and uncertainty estimates.
Open by design
Where clients agree, we contribute benchmarks back to the community through our open V&V benchmark suite.
Want an independent check of your results?
Our V&V audit benchmarks your code or simulations and delivers a signed verification report, typically within 1–3 weeks.