Artificial neural network and machine learning predictive model for assessing physicochemical properties of garlic slices (Allium sativum L.) during microwave-assisted convective drying process

Abstract

This study evaluates the physicochemical characteristics of garlic slices dried using a microwave-assisted convective dryer controlled by an artificial neural network. The chosen drying conditions included: microwave power (100, 200, and 300 W), air temperatures (45, 55, and 65 °C), and airflow velocity (0.3, 0.5, and 1.0 m/s). Results showed that at 65 °C, 300 W, and 0.3 m/s, the minimum flavor was 4.95 mg/g dry mass, marking a 39.50 % reduction in allicin content. The highest vitamin C content of 0.1751 mg/g with a water activity level of 0.505 was recorded at drying conditions of 1.0 m/s, 45 °C, and 100 W. However, it was observed that increasing power to 300 W at 45 °C and 0.5 m/s improved the rehydration ratio by 15.53 %. This study utilized precise ANN modelling to achieve an excellent fit by clarifying the interactions among drying parameters, time, and physicochemical parameters. PCA highlighted notable similarities between total color changes and rehydration ratios of garlic samples. Integrating an ANN into microwave-convective drying provides advanced tools to optimize food drying processes, thereby enhancing productivity without compromising product quality.