Drying prediction of banana slices using Kriging surrogate model based on genetic algorithm

Abstract

This work presents the prediction of a Kriging surrogate model based on a genetic algorithm optimization search. Using optimal Latin hypercube sampling, a sequencing optimization based on simulated annealing was applied to the design of experiments of the drying of banana slices using a hot air technique. The independent variables were the temperature, banana slice thickness and drying time. The dependent variables were the final moisture content, shrinkage and drying energy. Under the relationship between independent variables and dependent variables, Kriging models approximated the objective function.
A genetic algorithm optimization was applied to search for the optimal independent variables. Numerical results showed that the Kriging surrogate model with the relationship between the Gaussian function and the second−order polynomial function given by the function weights of W₁ W₂ and W₃ at 0.80, 0.15 and 0.05, respectively, was the best function model. The optimal values of the independent variables were temperature of 70°C, banana slice thickness of 4 mm, and drying time of 155 min. The minimum average error of the multi−objective function was about 0.52% by comparing the prediction results with the experimental results. This optimization approach was to predict the drying of banana slices using a hot air technique.