Supervised learning-guided design and CNC fabrication of Schwarz TPMS with experimental evaluation for atmospheric water generation

Atmospheric Water Generation (AWG) offers a decentralized and sustainable solution to global water scarcity by extracting moisture from the air. This study enhances AWG performance by combining supervised learning–guideddesign optimization and experimental validation of Schwarz Triply Periodic Minimal Surface (TPMS) structures. TPMS geometries improve condensation through a combination of large surface area and enhanced thermal conduction, enabling efficient droplet shedding. Structural and thermal simulations in nTop software generated data for predicting the top surface temperature—critical for condensation—based on parameters such as lattice thickness, cell size, surface area, and solid volume. Among the supervised learning models tested (Support Vector Machine, K-Nearest Neighbor, and Decision Tree), the Decision Tree showed the best accuracy (R² = 0.9676), captuirng both linear and nonlinear relationships. Using optimized parameters, a Schwarz TPMS structure was fabricated from Aluminum 6061-T6 via multi-axis CNC milling through a layer-wise subtractive method. Experimental evaluation under thermoelectric cooling (4–7 V) showed that the TPMS outperformed a solid aluminum block at 5 V and 6 V, achieving up to 33% higher water yield, though performance declined at 7 V due to frost formation. Thermal mapping highlighted non-uniform cooling from interfacial resistance between layers. This work demonstrates a scalable approach that integrates supervised learning–guided design with subtractive manufacturing for efficient AWG systems.