Fuzzy inference system to modeling of crossflow milk ultrafiltration

Fuzzy inference systems (FISs) have been adopted as a powerful tool to adequately model and simulate the crossflow ultrafiltration of milk. It primary aim of this research was to predict the permeate flux, total hydraulic resistance and the milk components rejection as a function of different physicochemical properties. Dynamic modeling of ultrafiltration performance of colloidal systems (such as milk) is an important criterion in designing of a new process, due to the complex nature of milk in a dynamic manner. Such processes exhibit complex nonlinear behavior due to unknown interactions between compounds and colloidal system, thus the theoretical approaches were not being able do successfully model and simulate the processes. In addition, it has been attempted to test the FIS ability to predict new data that is not normally available. To obtain this objective, fuzzy output with a threshold of 1.2 was constructed in order to predict rejections from the limited number of accessible experimental data. The findings of this work also reveal that there exist an excellent agreement between the experimental data and predicted values. Furthermore, the experimental results shows that the total hydraulic resistance and solutes rejection (except for protein) increased significantly with time for every corresponding value of the hydrodynamic parameters. On the other hand, the permeate flux decreased sharply. However, the fat rejection does not change with time for every value of the pH. (C) 2007 Elsevier B. V. All rights reserved.