Comparison study of the effect modeling of flow parameters on the membrane clarification efficiency for pomegranate juice

In recent years, several studies have indicated that modeling techniques based on artificial intelligence can be used for efficient prediction of food industry-related variables. In this study, machine learning methods were used to predict the permeate flux of pomegranate juice in a membrane clarification system based on membrane material, pore size, pressure, flow rate, and processing time. The experimental data were modeled using curve fitting, fuzzy inference system (FIS), artificial neural networks (ANN), and adaptive neuro-fuzzy inference system (ANFIS). Results showed that the permeate flux is a function of time and a power equation can predict the permeate flux with MSE of 0.0136. FIS, ANN and ANFIS models resulted in MSEs equal to 0.0495, 0.0145, and 0.0045 for permeate flux prediction, respectively. According to these findings, ANFIS has resulted in more reliable performance which can be used as an acceptable model in the prediction of permeate flux. The optimum architecture for the ANN was obtained 5-22-1 whilst the architecture of ANFIS models for PVDF and MCE membranes were 3-7-12-12-1 and 4-9-24-24-1, respectively. The results of this study can be used to predict the amount of permeate flux in the absence of experimental data and/or for interpolation and extrapolation of the permeate flux. Practical applications: One of the problems in juice membrane clarification is the accumulation and deposition of rejected compounds on membrane surfaces or inside its pores which results in a membrane fouling. On the other hand, several parameters can have influence on fouling and predictions of juice permeate flux during the membrane processing whereas they are important in industrial applications. Therefore, providing a model which able to predict the permeate flux having the value of effective input parameters seems to be useful. In this regard, several artificial methods can be used.