BIOMIMETIC PASSIVE COOLING ARCHITECTURES FOR HOT-HUMID CLIMATES USING MACHINE LEARNING-DRIVEN COMPUTATIONAL DESIGN

Hot-humid climates pose persistent challenges to building thermal comfort due to high ambient temperatures, elevated relative humidity, and limited diurnal temperature variation. Conventional mechanical cooling systems dominate these regions but significantly contribute to energy consumption and carbon emissions. Passive cooling strategies provide a sustainable alternative; however, their effectiveness in humid environments is constrained by moisture control limitations. This study investigates an interdisciplinary approach that integrates biomimetic architectural principles with machine learning (ML)-driven computational design to develop optimized passive cooling architectures suitable for hot-humid climates. Drawing inspiration from naturally evolved biological systems—such as termite mounds, which regulate temperature through self-organized ventilation networks, and elephant skin, whose wrinkled morphology enhances convective and evaporative heat dissipation—this research translates biological strategies into parametric architectural models. These models are subsequently optimized using genetic algorithms (GAs) and multi-objective ML frameworks to minimize thermal discomfort while reducing cooling energy demand. A structured framework is proposed that integrates biomimetic abstraction, parametric modeling, environmental simulation, and evolutionary optimization. Performance evaluation is conducted using adaptive thermal comfort indices under representative hot-humid climate datasets. Simulated case studies demonstrate that ML-optimized biomimetic ventilation systems and façade morphologies can reduce indoor thermal discomfort by 20–40% compared to conventional passive designs, while also decreasing cooling loads by up to 35%. The findings highlight the capacity of ML-driven biomimetic design to address the dual challenges of heat and humidity through geometry, airflow control, and surface adaptation rather than mechanical dependence. Despite challenges related to computational intensity and biological abstraction fidelity, the study establishes a robust foundation for next-generation, climate-responsive architecture. This work contributes to sustainable building design discourse and aligns with global energy-efficiency and climate resilience goals.