NARXNN Modeling of Ultrafiltration Process for Drinking Water Treatment

Ultrafiltration (UF) process has gained attention over times, particularly in treating drinking water treatment. A major challenge in achieving high quality of drinking water in UF process is membrane fouling. Membrane fouling has great effects on the performance of filtration process. To overcome fouling accurately, a prediction model is necessary. With prediction model, membrane fouling can be handled in a right way, so that, efficiency of filtration process can be maximize. This paper presents a study on modeling based on non-linear autoregressive with exogenous input neural network (NARXNN) of UF pilot plant specifically from treating drinking surface water. The NARXNN was used to model the permeate flux and transmembrane pressure (TMP). In this work, LM training algorithm was employed. The performance of the model was measured based on mean square error (MSE), root mean square error (RMSE) and correlation of coefficient (R). The simulation results demonstrate that proposed NARXNN modeling able to give high prediction rate with R value of 0.91743. It shows, the prediction values agree well with the actual values. With this model, membrane fouling can be successfully simulated and monitor accordingly.