Rack power densities in data centres approach 140 kW, and prevailing strategies trade one scarcity against another: evaporative towers spend water to save energy, and closed-loop systems spend energy to save water. We report an elephant pinna-inspired heat exchanger whose novelty resides in its body rather than its network, a compliant elastomer casting embedding a bifurcating vascular tree, actively flapped. Bifurcating microchannel sinks are mature and governed by constructal theory, yet a compliant body with an actuated air side remains unrepresented. The African elephant (Loxodonta africana) furnishes an architectural precedent rather than a load-matched one, its pinna operating three to four orders of magnitude below processor die heat flux. Our bioinspired design adopts the branching network and the flapping, leaving vasomotor modulation to future work. The transfer is one-directional.
Four geometries were screened by thermal-resistance analysis in ANSYS Fluent. Leaf venation is optimised for hydraulic supply, the shark rete mirabile for heat conservation, and the fennec pinna for passive radiation, whereas only the elephant pinna couples an actively perfused branching network to an actuated convective surface. Channel radii of 1.40, 1.11 and 0.88 mm satisfy Murray’s law to within 1%. Casting LSR-225 silicone around a sacrificial core yields vascular topology and body compliance in one 10 mm section. Two SG90 servos drive a 180° stroke at 1.3 Hz. Water circulated through a closed 315 cm³ loop, and an uninsulated aluminium chamber heated by a PTC element rated 100 W, unmetered and self-regulating below 60 °C. DS18B20 probes at the chamber and return reservoir were sampled at 2 s across five 300 s trials at 25°C ambient, flapping active throughout.
Timed collection yields 1.04 g/s and Reynolds numbers of 470 at 20 °C rising to 865 at 50 °C, falling 0.63 per generation as Murray dimensioning requires. Flow is therefore laminar throughout, monotonically decelerating, and everywhere within the thermal entrance region. Against a 35.2 °C inlet-to-ambient potential the exchanger sustained 12.7 °C, corresponding to 55 ± 3W over 0.0496 m², or 1118 W/m², a conductance of 1.96 W/K and a thermal resistance of 0.51 K/W. The loop had not attained steady state at 300 s. Therefore, the above represents an instantaneous rate under transient conditions rather than a steady capacity. Residual drift corresponds to storage in the loop water, closing the energy balance at 76 W, consistent with a heater at setpoint. Simulated drops of 1.4 and 3.9 °C fall below measurement, the simulations assuming static walls and still air, closure requiring an air-side coefficient near 75 W/m²K.
Rigid metal exchangers have optimised the network and left the body unexamined, whereas the pinna treats both as a single object, and it is that unity this work attempts to fabricate as a biomimetic construction. A 3 mm elastomer path contributes 47% of total resistance, rendering the body co-dominant with the air side and wall thinning the first design iteration. The closed architecture eliminates on-site evaporative water consumption by design. A static-ear control to isolate the flapping contribution, pressure-drop measurement to test the Murray premise, per-branch flow measurement, power metering and a mesh-independent laminar conjugate re-run are underway.