Enhancing exergy and sustainability performance in solar drying of Stevia rebaudiana using IoT ‐regulated airflow reversal and machine‐learning‐based prediction

Abstract Solar drying is a sustainable alternative to conventional thermal drying; however, non‐uniform airflow can generate temperature gradients, uneven moisture removal, and increased thermodynamic irreversibility. This study developed and evaluated an IoT‐controlled direct solar dryer equipped with an automatic airflow‐reversal system (ARS) for drying Stevia rebaudiana leaves under Egyptian climatic conditions. ARS was compared with a conventional unidirectional airflow system (CUAS) at airflow rates of 0.08 and 0.15 m 3 s −1 . The assessment included drying kinetics, thermal energy efficiency, exergy performance, tray‐to‐tray moisture uniformity, exergy‐based sustainability indicators, and energy‐payback and avoided‐carbon analyses. Nine machine‐learning algorithms were also evaluated using 88 observations, of which 70 were used for training and internal validation and 18 for independent testing under a leakage‐controlled framework. ARS substantially reduced tray‐to‐tray moisture variability; the mean moisture‐content coefficient of variation decreased from 28.58% under CUAS to 5.45% under ARS at 0.08 m 3 s −1 and from 20.65% to 3.79% at 0.15 m 3 s −1 . Peak drying efficiencies reached 47.1% and 42.0% under ARS at 0.08 and 0.15 m 3 s −1 , respectively, compared with 32.8% and 22.6% under CUAS. Gaussian Process Regression provided the best prediction of exergy efficiency, IP, WER, and SI (test R 2 = 0.998769–0.999957), whereas Multilayer Perceptron performed best for EDC, EIF, and EEF (R 2 = 0.968795–0.973909). The results demonstrate that IoT‐regulated airflow reversal can improve drying uniformity and thermodynamic sustainability while providing a reliable data‐driven framework for performance prediction within the investigated experimental domain.