Secagem convectiva de folhas de hortelã : análise baseada no ajuste de correlações empíricas, superfícies de respostas e redes neurais
Costa, Ariany Binda Silva
MetadataMostrar registro completo
Mentha x villosa H. is a species of mint popularly known as regular mint. Besides being widely used as a spice, their leaves yield an essential oil with many healthy compounds that are commonly used in formulation of medicinal and cosmetic products. Convective drying is an interesting alternative to preserve such compounds, but because the oils are rich in aromatic constituents, and the leaves are fragile material, heating must be moderate and proper drying techniques have to be used. In the present work, the drying of mint leaves was investigated in two types of equipment, namely a natural convection oven and a forced convection horizontal dryer. The fresh leaves were initially characterized by measuring their dimensions, bulk density and moisture content. The influence of air temperature, size and growing stage of leaves on the drying kinetic curves was investigated initially for drying in the natural convection oven. The results showed that only the temperature affected the drying rates, while the size or growing stage had no significantly effect on the process. For forced convective drying, the influence of air temperature, air velocity and mass of the samples was investigated. The temperature was the variable with stronger influence on the drying curves. The mass of the sample affected the drying kinetics only at the highest temperature (60 oC), while the influence of air velocity was weak for most conditions. Empirical and semi-empirical models were fitted to experimental data of dimensionless moisture content versus time. The models from Page and Henderson-Pabis provided the best fittings to the experimental data, depending on the mass of the sample. To complement the analysis, multiple regression models based on surface responses were fitted to experimental data. Also, a neural network model was built to simulate experimental data. The resulting network needed seven neurons to represent satisfactorily the drying kinetic data, and provided excellent estimates of experimental results.