Introducing our comprehensive design platform: NuPlant

The Fusion Igniter Meeting, Santa Fe New Mexico – 28^th July 2025

At the Fusion Igniter Meeting, nTtau Digital’s CSO Dr. Muhammad Omer and Caminno’s CTO Dr. Vignesh Perumal showcased a breakthrough: the NuPlant design platform that compresses months of fusion power plant design into just hours.

NuPlant unifies system-level generative design, physics-informed simulation, multi-objective optimization, and immersive visualization down to manufacturable component detail.

Muhammad opened with the end state: a VR scene showing an integrated IFE plant where both the site layout (turbine hall, heat exchangers, hot cells, roads, parking) and the reactor hall (first wall, target, blanket, shielding, bioshield, containment) have been algorithmically optimized. The VR walk through is presented in the video below:

The NuPlant platform spans pre-conceptual designs through detailed engineering, with three design goals: faster, more efficient, and fully automated.

Deep Dive into the NuPlant Platform

Our platform is modular and package-agnostic: it can drive mechanical, electrical, and building layouts and plug into SmartPlant 3D, plus open tools like MOOSE, OpenMC, and CAD/meshing stacks. 

Users supply three things—design parameters, constraints, and objectives (e.g., minimize cost, maximize efficiency or Tritium Breeding Ratio (TBR))—and the engine explores the feasible design space at both macro (plant) and micro (component) scales. Instead of traditional repetitive engineering loops (assume → calculate → iterate), their automated Galaxy workflows transform a single input into:

• Parametric geometry

• Meshing

• Neutronics, thermal-fluid, and structural mechanics simulations

• Costing and performance outputs

No fragile script chain, no manual handoffs, just streamlined, end-to-end automation.

To push speed even further and explore the design space in-depth, they train lightweight surrogate models on the generated data to predict neutronics and system KPIs nearly instantly. 

NuPlant in Action- Faster, Smarter and Fully Automated

1. Pre-Conceptual Design of Blanket and First Wall

During the demo, a neutronics surrogate model predicted TBR, first-wall heating, and cost in a “flash of a second” from slider-controlled geometry and material selection—versus ~45 minutes for a comparable Monte Carlo run. Accuracy loss was bounded (≈15%), which is acceptable for steering pre-conceptual choices and determining which cases merit high-fidelity reruns. Figure 1 shows a snapshot of the surrogate model predicting the TBR, first wall heating and cost of the reactor for given parameters. 

Fig 1. A snapshot of surrogate model predicting TBR, first wall heating and cost of reactor for given parameters

2. Multi-Objective Optimization – Finding the Sweet Spot

With surrogates in place, they ran multi-objective optimization by specifying design requirements such as: 

• Maximize TBR, 

• Minimize heating, 

• Minimize cost and, 

• Parameter bounds for each design variable (e.g., first-wall 0.005–0.05 m, blanket 0.5–1.5 m), 

The optimizer evaluated tens of thousands of combinations to return Pareto-optimal sets. Figure 2 shows the parameter bounds for each variable and the corresponding design objectives.  

In one case, the optimizer converged to a blanket thickness of 1.47 m and a first-wall thickness of 0.0081 m yielding TBR ≈1.27. Figure 3 shows the optimized geometry parameters. A major change in reactor geometry is noticeable compared to the geometry in figure 1, as the optimizer embeds manufacturing constraints directly into the design process. Re-weighting priorities (cheapest first) produced a different optimum (TBR ≈1.14) in seconds. 

Fig 2. Parameters bounds for each design variable and objectives to perform automated design iterations

Fig 3. The sweet spot. Optimal parameters to meet design targets.

3. From Pre-Conceptual to Detailed Component Design Using Gen-6 Optimization

Perumal then zoomed into the component level with Caminno’s Gen-6 optimization of an IFE target: refining it for manufacturability and service loads based on scientific machine learning models. The demo target was required to survive inertial launch and withstand residual heat long enough to reach chamber centre, while remaining compressible enough to avoid delaying ignition. Gen-6 performed two-scale optimization: at the macro level it reshaped the object and redistributed material; at the micro level it selected lattice topology, feature size, and functional grading to hit strength, thermal compliance, and cost targets. The scientific ML model operated at the differential-equation level, capturing coupled physics while reducing data requirements, and output a functionally graded lattice target ready for downstream manufacturing workflows. Figure 4 shows optimization of target using Gen-6.

Fig 4. Micro level optimization of target using Gen-6 (image credit: CamInno)

Finally, NuPlant extended optimization to site-level planning. With the reactor hall fixed at the centre for safety, the engine reorganized the overall layout according to design objectives. A “constructability” mode prioritized clear access and staging; a “cost” mode arranged interdependent systems along a compact spine to minimize corridors and piping; a “balanced” mode co-optimized cost, buildability, and functionality. All variants satisfied baseline constraints, and the selected layouts could be exported to SP3D, Revit, or client toolchains. Figure 5 shows a balanced powerplant layout solution.

Fig 5. Balanced powerplant solution which gives equal importance to cost, functionality and constructability

Conclusion:

The NuPlant design platform demonstrated transformative benefits across all stages of fusion power plant development - from pre-conceptual design of components to detailed component optimization and site-level planning. By integrating physics-informed simulations, surrogate modelling, and multi-objective optimization into an automated workflow, the tool compressed months of effort into hours, while maintaining engineering fidelity.

Compared to traditional simulation-based workflows, NuPlant offers:

• Drastic reductions in time (months to hours).

• Significant cost savings (cloud compute cost reduced by orders of magnitude).

• Improved accuracy-to-speed trade-offs (bounded error ~15% acceptable for early-stage design).

• Scalability across system, component, and site-level optimization.

• Modularity and adaptability, enabling seamless integration with diverse workflows and rapid extension to new use cases

• Manufacturability-aware outputs, ensuring designs are practical and build-ready.

NuPlant marks a step change in fusion engineering, delivering designs that are economically viable, code-compliant, and aligned with safety regulations—while satisfying physics and engineering design requirements at unprecedented speed and scale.