Chapter 8 - ANN prognostication and GA optimization of municipal solid waste leachate treatment using aluminum electrodes via electrocoagulation-flocculation method

Abstract

This chapter presents the results of artificial neural network (ANN) modeling of turbidity reduction in municipal waste leachate using aluminum (Al) electrodes. The network was trained for different numbers of iterations to obtain the best number of neurons for the hidden layer (Nbest). The optimum turbidity removal and the optimum parameters that influence the treatment process (initial pH, current density, electrolysis time, settling time, and temperature) were also obtained using the genetic algorithm (GA) tool. The data were also statistically analyzed. The Nbest was observed as 10, having the lowest root mean square error (0.0053), sum of the squares error (0.00279), and absolute average deviation (0.00106), and highest R2 value (0.9886); these values indicate that the predicted and experimental responses are close and the ANN can be used to model the process. The optimal solutions for the input process parameters for pH, current dosage, electrolysis time, settling time, and temperature were 3.6004, 1.3502A, 3.6004minutes, 9.001minutes, and 36.004°C, respectively, with an optimal solution of 99.98%. The GA optimization tool was efficient as the experimental value of 99.84% was close to 99.98%. A main effect analysis shows that the parameters have an influence on the response variable and are significant to effectively and efficiently reveal the results of the experiments (P values less than 0.005). The Pearson correlation results revealed that pH, current dosage, electrolysis time, and settling time are the independent variables that are strongly significant for the response variable. Meanwhile, temperature is not significant for the response variable. The theoretical electrode consumption and electrical power consumption at the optimum conditions were 0.594088gAl/m3 and 1.51038×10−5Wh/m3.