A hierarchical fuzzy approach for flux prediction and optimization of milk microfiltration

Abstract

A Hierarchical Fuzzy System (HFS), due to the dimensionality problem of ordinary fuzzy systems, was utilized for prediction and optimization of the permeation flux during the milk microfiltration process. A data-driven hybrid identification method based on a genetic and sequential quadratic programming (SQP) algorithm was applied to determine the fuzzy system parameters. The developed model was identified as a powerful tool for highly accurate flux prediction with the mean relative error of 1.5%. Comparison of HFS versus the well-known mathematical models revealed the robustness of the developed model. The optimum operating conditions for maximizing flux were evaluated by HFS and a genetic algorithm. The influences of trans-membrane pressure, temperature, cross-flow velocity and presence of fat globules (skim or whole milk) on membrane performance were elucidated at optimum conditions.