Data-driven prediction and optimization of performance parameters in rotary wick humidifiers using ANN models

• Rotary wick-type humidifier was developed and validated for effective cooling. • MSS boosted performance up to 70 rpm; higher speeds cut humidification. • Optimal wick thickness was 6 mm; thicker wicks blocked airflow efficiency. • MLP outperformed polynomial regression with higher R 2 and lower errors. • SHAP-backed MLP gave robust, accurate, and physically interpretable insights. Efficient and sustainable humidification is a critical challenge in modern HVAC systems, particularly in hot and dry climates where cooling and moisture control are simultaneously required. This study presents the design, fabrication, and performance evaluation of a rotary wick-type dynamic humidifier developed to deliver cold and humid air with high energy efficiency. Unlike conventional static wick systems, the rotary configuration enhances air–water interaction and enables dynamic performance control. The system was experimentally tested under varying operating conditions—air velocity (2–10 m/s), motor shaft speed (30–90 rpm), and wick thickness (3–6 mm). Maximum performance was achieved with a mass transfer coefficient (MTC) of 38.59 kg/m 2 ·s and a coefficient of performance (COP) of 6.79, demonstrating its effectiveness in passive cooling–humidification. Results showed that performance improved with shaft speed up to 70 rpm and wick thickness up to 6 mm, beyond which efficiency declined due to reduced air–wick contact and airflow resistance. To model thermodynamic metrics (ΔT, ΔW, COP, MTC), a comparative analysis was carried out between a Multi-Layer Perceptron (MLP) neural network and polynomial regression. The MLP consistently achieved higher predictive accuracy, operating within ±3 % error bands, with lower MAE/MSE and higher R 2 values than regression models. SHAP analysis further confirmed air velocity as the most dominant parameter, with motor shaft speed and wick thickness primarily influencing moisture addition (ΔW). Overall, the findings establish the rotary wick-type humidifier as a novel, energy-efficient solution for sustainable cooling applications and demonstrate that MLP-based predictive modeling provides a robust and interpretable tool for system optimization and design.