Optimization of olive pomace dehydration process through the integration of computational fluid dynamics and deep learning

Abstract

ABSTRACT This study introduces an innovative approach to optimize the dehydration process of olive pomace by combining computational fluid dynamics (CFD) and deep learning. Through CFD, it identifies the optimal air inlet velocity in a prototype of a passive direct solar dehydrator for olive pomace, which allows for the reduction of its moisture content for subsequent use as biomass. The prototype was simulated in ANSYS software, and this simulation consisted of the following steps: prototype design, meshing, selection of physical models, material assignment, boundary condition simulation, and validation of results with data obtained from the prototype. Following this process, it was concluded that the optimal air inlet velocity to the dehydration chamber is 0.1 m/s. Concurrently, an artificial neural network model was used to analyze data from sensors in the physical prototype, revealing that solar radiation and ambient temperature are the most influential variables on the temperature of the dehydration chamber. This analysis resulted in a predictive model for the optimization of the dehydration process, with a correlation coefficient of 0.9699 for temp_up and 0.9710 for temp_down, and a Willmott coefficient of 0.9999, demonstrating a high concordance between the model’s predictions and the experimental data. The model’s input variables include solar radiation, ambient temperature, and both external and internal air humidity. This integration of CFD and deep learning offers a promising methodology for improving olive pomace dehydration systems and the industry in general.