Chemical Engineering Journal 172 (2011) 267–276 Contents lists available at ScienceDirect Chemical Engineering Journal j ourna l ho mepage: www.elsevier.com/locate/cej Modeling of electrolysis process in wastewater treatment using different types of neural networks Silvia Curteanu a,∗ , Ciprian George Piuleac a , Kazem Godini b , Ghasem Azaryan c a “ Gh. Asachi” Technical University Iasi, Faculty of Chemical Engineering and Environmental Protection, Bd. D. Mangeron, No. 71A, 700050 Iasi, Romania b Ilam University of Medical Sciences, Faculty of Health, Environmental Health Engineering Department, Banganjab Complex, Ilam, Iran c Hamadan University of Medical Sciences, Faculty of Health, Environmental Health Engineering Department, Hamadan, Iran a r t i c l e i n f o Article history: Received 6 February 2011 Received in revised form 29 May 2011 Accepted 30 May 2011 Keywords: Wastewater treatment Electrolysis Neural networks Neural network stack a b s t r a c t Indirect electrolysis has been used for the removal of chlorophyll a (as indicator of algae) from the final effluent of aerated lagoons in the wastewater treatment plant of Bu-Ali Industrial Estate. The efficiency of the process was studied experimentally and by simulation using neural networks. The process analysis was done in different conditions of retention time (5–50 min) and using two types of electrodes based on aluminum and stainless steel, with different distances between electrodes (from 1.0 to 3.5 cm). The electrical current and the average voltage applied were between 5 and 90 A (0.74–12 A dm -3 ) and 50 V, respectively. The influence of the main parameters of the electrolysis process on the final values for chlorophyll a, TSS and COD is evaluated experimentally. On the other hand, predictions of the main system outputs of a treated waste as a function of initial characteristics (initial values of chlorophyll a, TSS, COD) and operation conditions (temperature, electric power, time, electrode distance, and electrode type) were performed using artificial neural networks. The modeling methodologies elaborated in this paper are based on different types of neural networks, used individually or aggregated in stacks. They were developed gradually in the sense of improving the model performance. The neural network results represent accurate predictions, useful for experimental practice. © 2011 Elsevier B.V. All rights reserved. 1. Introduction In the present historical context, where the demanding of high quality water represents one important necessary condition, the wastewater treatment processes became more complex. The high level efficiency of exploitation on this kind of treatment is closely related to the stabilization of lagoons and ponds as a method which implies low cost and simplicity for operation and mainte- nance. On the other hand, the main disadvantage consists in the high concentrations of suspended solids of the final effluent, espe- cially composed from significant amounts of algae. Accordingly with World Health Organization Manual for the reuse of effluent, algae are not a pollutant, but the increase of algae suspended solids from ponds and lagoons is considered a flaw [1]. Another factor to be considered refers to the climatic conditions, being known that tropical countries are favorable to the wastewa- ter treatment processes through stabilization [2]. Algae are formed in each square meter of lagoons at the rate of 10–66 g m -3 per day and species such as Chlamydomonas, Euglena, ∗ Corresponding author. E-mail addresses: silvia curteanu@yahoo.com (S. Curteanu), ciprianpiuleac@yahoo.com (C.G. Piuleac), kgoodini@razi.tums.ac.ir (K. Godini). Chlorella, Scenedesmus, Microactinium, Oscillatoria and Microcystis are responsible for fulfilling photosynthesis in the aerated lagoons [3]. Numerous physical and chemical methods are recommended for improving the quality of the final effluents of the lagoons such as microstraining, rapid sand filtration, chlorination (an objectionable method according to the recent regulations due to high chlorine demand, formation of DBPs , pH drop and increase of dissolved solids), flotation, coagulation and flocculation–filtration and final treatment by the use of natural systems. However, there are various limitations in application of these methods: a high consumption of reagents and not always the water quality parameters are according to the international standards [4]. Neder et al. [2] tested five different natural treatment processes for removing algae from stabilization ponds effluents, respectively: rock filter, sand filter, floating aquatic plants, constructed wetlands, and overland flow. Several evaluation criteria are used in order to relate the capabilities of the post-treatment processes to the mul- tiple objectives. The main mechanisms for algae eradication or removal by indirect electrolysis are: (a) electro-oxidation/disinfection, (b) electro-flocculation, (c) electro-flotation, or any combination of these three mechanisms. When an inert (electrochemically nondis- solving) anode is used, the free chlorine and/or chlorine–oxygen 1385-8947/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.cej.2011.05.104