Análise do efeito da composição química na secagem de pasta em leito de jorro

Abstract

In the process of paste drying in spouted bed, the paste is submitted to the system, changes in rheological properties occur, a powder is formed and collected by a cyclone. Currently, various pastes have been tested and their chemical composition may indicate the behavior of the process. When it comes to modeling, many challenges are found, mainly the ones related to the incorporation of the pastes in the models. Therefore, this work was developed in order to obtain more detailed information about the drying process of pastes in a spouted bed, considering the effect of their chemical composition. Based on the effects of the presence of pastes, a hybrid CST / Neural model was verified experimentally to describe the heat and mass transfer phenomena. Besides, a neural model capable of predicting variations in the fluid dynamic behavior of the process was proposed. Drying experiments were carried out with water, homogenized egg, propolis, sweet whey, whole milk, skimmed milk, semi-skimmed milk and low lactose content milk in a spouted bed of cylindrical column of 0.5 m of diameter and 1 m of height, inferior conical base (angle of 60°) with an entrance nozzle of 0.03 m of diameter, top cone (angle of 60°) and outlet diameter of 0.05m. Polyethylene particles of 4.38mm were used as the inert part. During the experiments, changes in pressure drop, inlet air temperature, particle bed temperature at three positions axially from the annular region, outlet temperature, temperature of dry and wet bulb at the cyclone exhaust, as well as the moisture of the powder produced and covering formed on the inert particles were determined. From the information obtained, it is clear that the presence of sugar and fat cause different changes in the behavior of the spouted bed, especially when it comes to relative moisture and pressure drop. Regarding the modeling, the models proposed presented simplicity as for implementation and flexibility to interpolating values. These might be used in the future for other types of pastes by including other input data in the neural network, for example, physical properties of pastes.