Exploring the 25,000+ Configuration Space
Fusion design is a high-dimensional search; computing lets a small team examine tens of thousands of candidate machines instead of a handful.
Why the space is so large
A fusion device is defined by dozens of coupled choices: magnetic field, plasma current, aspect ratio, elongation, triangularity, density and temperature profiles, wall materials, blanket geometry, and heating mix. Each choice interacts with the others through plasma physics, neutronics, heat transport, and structural limits. The number of physically distinct combinations worth examining runs well past 25,000.
The old way and the new way
Historically each candidate meant a hand-built model and days of expert time, so programs evaluated a handful of points and defended them. That approach cannot cover a space this size. Automated model generation, batch simulation, and surrogate models let the same physics be applied uniformly to every candidate, removing the human bottleneck and the inconsistency that comes with it.
How a sweep is organized
- Parameterize the machine so a single vector describes one candidate.
- Sample the space with a space-filling or adaptive design of experiments.
- Run each candidate through the same validated physics chain.
- Score against constraints and objectives, then refine near promising regions.
What it buys Kronos
For the Hyperion breeder and the Aegis and MetroVolt burner housings, a broad sweep does two things. It finds configurations a human search would never have proposed, and it maps the boundaries of what is feasible, so the eventual choice is defensible against every neighbor rather than asserted in isolation.
The output is not a single answer but a landscape: which constraints bind, where the trade-offs are steep, and which regions are robust to uncertainty. That landscape is the input to closing the design point.
Every candidate is stored with its inputs and outputs so the sweep is reproducible and auditable, which matters for later licensing evidence.