Experimental and neural network modeling of micellar enhanced ultrafiltration for arsenic removal from aqueous solution
Muhammad Yaqub, Seung Hwan Lee
Abstract
The optimization of micellar-enhanced ultrafiltration (MEUF) of arsenic (As) contaminated aqueous solution using cetylpyridinium chloride (CPC) as surfactant was studied through experimental and artificial neural network (ANN) modeling. Experimental studies were carried out by varying operational conditions such as time, pressure, molar ratio of CPC to As, concentration of As and pH of feed solution. Root mean square error (RMSE) and coefficient of determination (R<sup>2</sup>) were considered as performance criterion to evaluate the predicted results of ANN model. The experimental studies provided optimum operating parameters such as pressure 1.8 bar, molar ratio of CPC to As was 5:1, As concentration 1 mM and pH 8.0 of feed solution. ANN model presented reliable results with RMSE values 0.259, 0.553 and 0.623 for training, validation and testing datasets, respectively, while R<sup>2</sup> values for training, validation and testing dataset were noted as 0.962, 0.942 and 0.932, respectively. The proposed ANN model traced input-output relationship to predict As removal efficiency (RE) of MEUF process. Therefore, ANN model can be considered as a competitive, powerful and fast alternate because of its high computational speed, accuracy and economics in MEUF process optimization without doing laborious experimental work.