Fuzzy Logic-Based Model to Predict the Impact of Flow Rate and Turbidity on the Performance of Multimedia Filters

Abstract

This paper uses fuzzy logic–based models to predict and evaluate the performance of multimedia filters utilized in wastewater treatment. A fuzzy logic–based model is constructed and trained to predict the operating time (i.e., treated volume of water) of a multimedia filter. A preset acceptable turbidity value of 5 nephelometric turbidity units (NTU) is used as the breakthrough point. The model is based on a set of experimental data with variable flow rates and influent turbidity. The results from the fuzzy-based model indicate that the simulated treated volume at different inputs of turbidity and flow rate fits the experimental results with a coefficient of multiple determination (R2) of 91.6%. To examine the efficiency of the developed model predicting treated volume, the results obtained from the model are compared with the results obtained from a multiple linear regression model. The accuracy of prediction of both models are examined using the mean absolute error (MSE), root-mean-square error (RMSE), and R2. The MSE, RMSE, and R2 for the fuzzy-based model are 5,318, 72.92, and 98%, respectively, whereas for the regression model they are 3,302, 57.46, and 99%, respectively. Although the regression model appears to be more accurate, the fuzzy-based model is deemed to be more advantageous because it can incorporate the uncertainties in inputs as a result of human judgments and can indicate the errors in the outputs.