Utilização do método bayesiano na descrição da cinética de secagem da polpa de jabuticaba por modelos de regressão não linear

Abstract

Jabuticaba is a native tree from to the Atlantic Forest in Southern Brazil, and its fruit is widely consumed in the fresh form, but it is highly perishable, requiring conservation techniques. The drying process is very convenient, as well as reduces post-harvest losses, adding value to the final product. The aim of this study was to describe the drying kinetics of Jabuticaba pulp at temperatures of 50 and 60°C. By using nonlinear regression models wich where evaluated regarding to the structure of autoregressive dependence erros of first order, using the methods of Bayesian Inference with informative and non-informative priors and getting the complete conditional posterior densities of paramethers. We evaluated the models: Henderson and Pabis (HP), Simple Three-Parameter Exponential (ESTP), Lewis (L), Diffusion Approach (AD), Two Terms (DT), Page (P) and Midilli (M), finding the diagnosis of convergence, point estimates of mean and fashion and intervals (HPD), as well as selection criteria through software R. The M, ESTP and AD models, presented at least one problem regarding to the convergence, validity of the assumptions of the regression models and/or suitability of the fit the data. The models L, HP and DT with structure of first order errors were adequate to describe the data of Jabuticaba pulp on both temperatures; the HP was more indicated at 50 °C and DT at 60 °C. In 50 °C the drying rate was 0.0078g water/min ranging from 0.0046 to 0.0122 g water/min in the process and at 60°C the drying rate was observed in a first stage of 0.0143 g of water/min, changing between 0.0132 and 0.0155 g of water/min, in the second stage the rate was 0.0689 g of water/min ranging from 0.0690 to 0.0710 g of water/min. With the drying temperature increase there was a decrease of about 50% in the drying period.