Artificial Neural Networks Applied to Drying Curves of Capsicum chinense
Year: 2025
Type: article
Abstract: Objective: The objective of this study was to apply Artificial Neural Networks (ANNs) to predict the drying curves of Capsicum chinense at different temperatures, aiming to optimize the drying process and ensure the quality of the final product. Theoretical Framework: Artificial Neural Networks (ANNs) have been widely used in modeling and predicting industrial processes, including the drying of agricultural products. Method: An analogy was drawn between the chemical kinetics equations applied to the drying unit operation and the determination of the parameters: kinetic constant (k) and reaction order (n) for various temperatures. The parameters obtained in this analogy fed the neural network modeling. With the data extracted from the drying curves and the aid of GeoGebra software, the drying kinetics parameters, n (reaction order) and k (reaction rate constant), were determined. An RNA (Recurrent Neural Network) was developed in MATLAB to model the drying curves of the 'pimenta-de-cheiro' pepper (Capsicum chinense) at different drying temperatures (40°C to 70°C). Results and Discussion: ANN training, performed with 245 data points and employing the Levenberg-Marquardt algorithm, resulted in high correlation coefficients (R = 1.0000 for training, 0.9999 for validation and 0.99995 for the overall set) and low mean squared error values (minimum MSE of 9.94 × 10⁻¹¹). Furthermore, the ANN demonstrated generalization capacity when predicting drying curves for temperatures between 30°C and 80°C, confirming its applicability in modeling and optimizing the drying process of 'pimenta-de-cheiro' pepper. The neural network obtained was used to predict the drying curves of 'pimenta-de-cheiro' pepper (Capsicum chinense) at different drying temperatures (40°C to 70°C). Three-dimensional figures were obtained correlating the parameters temperature, kinetic constant (k) and reaction order (n). Due to the analogy made between the drying operation and chemical kinetics, in the drying operation, parameter (k) could be called drying kinetic constant, and parameter (n) could be called drying order. Research Implications: ANN modeling is effective in predicting and optimizing industrial processes, especially in the analysis of large volumes of data and complex systems, contributing to effective quality control in food production and improvement of industrial processes. Originality/Value: Artificial Neural Networks (ANNs) stand out as innovative and effective tools in optimizing industrial processes, especially in the drying of agricultural products. Inspired by the functioning of human neurons, ANNs allow for precise and adaptive predictions to solve complex and dynamic problems.
Authors: Melissa Marques de Sousa, Christiano Pedro Guirlanda, Ana Clara Cassin Cordeiro, Helen Martins Marques, Natália Resende Pinheiro Leite, Juan Canellas Bosch Neto
Language: English
Topic: Food Drying and Modeling
Subfield: Food Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: hybrid
OpenAlex ID: W4408474137