Optimization of desalinationwastewater treatment unit performance; Experimental investigation accompaniedwith artificial neural network and adaptive neural fuzzy interferences modelingAuthor(s): Farshad Farahbod, Narges Bagheri, Mahdi Asadi
Almost high salinity effluent streamof desalination units is drained into the sea or dispersed on soil breaking the aqua salt concentration balance and also increase salt content of sea ecosystem or soil. So it is concerned with environmental engineering, corrosion engineering, control engineering, chemical engineering and etc. So zero discharge desalination (ZDD) plants have been proposed with a view of reaching salt and water instead of hazardous saline wastewater. Predicting wastewater pretreatment performance (effluent total hardness, CO2content and electrical conductivity) as the first step in ZDD plants is considered in thiswork both experimentally (on a pilot plant) andmathematically (modelingwith artificial neural network and adaptive neuro fuzzy inference system). So, optimum operating conditions (150 ccAl2(SO4)3 as coagulant, mixing rate in first pretreatment reactor= 110 rpm, 600 cc NaOH and 450 cc Na2CO3 as additives) are recognized, then optimal NN architecture (three layer feed forward back propagation networkwith 10 neurons in hidden layer, Levenberg-Marquardt algorithmis as network training function, tangent sigmoid transfer function) and also optimal ANFIS architecture (five layers with six neurons in two hidden layers, two Bellmembership functions and four rules) are determined. The results confirmpredictivemodeling byANNismost efficient comparing withANFIS in prediction of performance.