Machine learning prediction and optimization of thermodynamic analysis and energy enhancement of a hybrid infrared dryer for onion slices

Abstract

Machine Learning (ML) and Artificial Intelligence (AI) are important tools for modelling drying processes to reduce moisture and preserve food products. This study investigated the drying performance of an industrial infrared conveyor belt drying system on onion slices under different drying conditions. The effects of drying temperature, infrared intensity, and airflow rates were evaluated. The results demonstrated that increasing IR power and air temperature significantly reduced drying time by 44.23%. Effective moisture diffusivity increased from 0.238 × 10⁻¹⁰ to 0.457 × 10⁻¹⁰ m²/s, indicating enhanced internal moisture transport at elevated thermal inputs. The lowest Sect. (10.72 kWh/kg) was achieved at 600 W, 65 °C, and 0.3 m/s, while the highest (22.26 kWh/kg) occurred at low temperature and high airflow conditions. Thermal efficiency improved with increasing temperature and radiation intensity, reaching a maximum of 21.92%. However, the Artificial Neural Network model exhibited excellent predictive capability with a correlation coefficient (R) of 0.999, accurately estimating key drying parameters. Self-Organizing Map (SOM) analysis identified distinct operational clusters, revealing that higher air temperature and IR power reduced drying time and energy consumption, whereas increased airflow increased energy usage. Therefore, the study demonstrates that integrating AI and statistical tools provides a robust framework for optimizing industrial drying systems, enabling reduced energy consumption and improved process efficiency.