Summary
The CELLSIUS team at ETH Zürich is building the H2-Sling, a student-designed hydrogen-electric aircraft. Hydrogen fuel cell powertrains promise clean flight, but they hand engineers a hard thermal problem: they run cool, around 80 °C, so very little waste heat leaves through the exhaust, and the rest has to be pushed out through dedicated radiators. On the H2-Sling, that means moving up to 100 kW across five heat exchangers, in hardware that weighs 40 kg and drives close to a third of the aircraft’s drag.
For a bachelor’s thesis, Tim Helfenstein, Head of Flight Testing & Module Coach at CELLSIUS – Project LH2, set out to redesign the fuel cell stack heat exchanger to be lighter and to flow better, all while keeping it printable. They used ToffeeX physics-driven generative design and aimed it where the physics said it would pay off: the coolant inlet and outlet manifolds. The result was a combined manifold mass cut from 2.90 kg to 1.84 kg, far more uniform coolant distribution into the core, and lower pressure losses. The final manifolds were shelled, lattice-reinforced, validated in Ansys, and printed in metal.
Cooling a fuel cell stack at minimum weight
In a hydrogen-electric aircraft, the fuel cell stack is one of the most thermally demanding components of the entire aircraft. The H2-Sling powertrain operates at a high total system efficiency of 45%-51% but at a low temperature of roughly 80 °C. That low temperature is the core of the problem. A combustion engine dumps most of its waste heat straight out of the exhaust, hot; a fuel cell cannot. The small temperature difference between coolant and ambient air means the thermal management system has to work far harder to reject the same energy, and on this aircraft that adds up to as much as 100 kW across five heat exchangers.
The most punishing moment is the take-off climb, when the aircraft holds full power for an extended stretch while flying slowly at a high angle of attack. Heat demand peaks exactly when the airflow available for cooling is at its worst.
Weight and drag make the target harder still. The thermal management components weigh 40 kg, and cooling drag makes up around 30% of the aircraft’s total drag. Trimming mass off the heat exchangers and cutting the pressure the pump has to fight both feed directly into range and payload. The thesis brief was clear: redesign the fuel cell stack heat exchanger to weigh less, flow easier, and reject more heat, all inside these tight limits, and keep it printable.
Focusing ToffeeX where it pays off: the coolant manifolds
ToffeeX is a physics-driven, multi-objective generative design platform. Given a design domain, boundary conditions, and weighted objectives for heat transfer and pressure loss, it adds or removes material across the domain at each iteration, placing structure only where the physics says it earns its place.
The team’s early work confirmed that the aircraft’s fast, pressure-drop-limited airstream was best served by a conventional plate-fin core, so they kept it. The real opportunity was on the coolant side: the inlet and outlet manifolds that feed the core channels are a 3D flow problem, exactly what ToffeeX is built for. Using the 3D single-fluid model, the team optimized the manifolds against three objectives at once: minimize pressure loss, distribute coolant mass flow evenly across the core channels, and minimize the internal fluid volume.
Minimizing the internal fluid volume drives the weight down, both in structure and in the coolant carried on board. Evening out the flow makes the core work harder, because a core fed unevenly wastes channels that run too hot or too cold. ToffeeX solved for both together rather than trading one against the other after the fact.
From optimized flow to a printable part
A great flow field is only useful if it can be built, so the team wrapped ToffeeX in a multi-software workflow. The design domain was first built in nTop, shaped to the interfaces of the heat exchanger core, then optimized in ToffeeX with targeted mesh refinement on the critical outlet regions. The optimized fluid geometry went back into nTop to become a physical part: shelled with a variable wall thickness, given permanent internal supports in the severe-overhang regions so the sealed channels could still print, and wrapped in a Voronoi surface lattice to stay stiff at minimum weight. Finally, the complete manifold was validated in Ansys under coolant pressure loads.
Results
Applying ToffeeX to the manifolds paid off across every objective the team set.
Table 1: Baseline vs. ToffeeX-optimized coolant manifolds
| Metric | Baseline manifolds | ToffeeX-optimized | Change |
|---|---|---|---|
| Combined manifold mass | 2.90 kg | 1.84 kg | 1.06 kg (roughly −37%) |
| Flow distribution into core channels | Uneven | Uniform/Targeted | Improved |
| Coolant-side pressure drop | Reference | Lower | Reduced pump power |
The combined inlet and outlet manifolds went from 2.90 kg to 1.84 kg, a saving of more than a kilogram on a single subsystem of an aircraft where every kilogram costs range. The generative geometry also feeds each core channel its fair share of flow, which raises the effective heat transfer of the whole exchanger, and its smooth internal paths lower the coolant-side pressure drop so the pump draws less parasitic power.
Manufacturing: printed in metal at ETH
The manifolds and heat exchanger were printed in metal via Selective Laser Melting (SLM) at Pilatus Aircraft, a Swiss aerospace manufacturer renowned for versatile, high-performance aircraft, and are shown below still attached to the build plate.
SLM is what makes this approach possible: the optimized manifolds carry organic internal channels and a Voronoi outer skin that no casting or machining process could reproduce as a single sealed part, and printing them in one piece removes the joints and potential leak paths a conventional assembly would need.
Conclusion
The CELLSIUS heat exchanger shows generative design used with judgment: keep the conventional plate-fin core where it wins, and aim ToffeeX at the coolant manifolds where a 3D flow problem plays to its strengths. The payoff was a manifold set roughly 37% lighter that distributes coolant evenly across the core, drops the pressure the pump has to fight, and prints in metal as a single sealed part.
For a student team working to a thesis deadline on a hydrogen-electric aircraft, that is a meaningful gain on a subsystem where weight and pumping power translate straight into flight performance. Looking ahead, CELLSIUS plans to use ToffeeX across more of the aircraft, including fully coupled 3D workflows that optimize heat transfer, pressure loss, and structural integrity together.