Article Preview
Top1. Introduction
The input-output data sets are used for modeling and determining the identity of complex systems behavior. The mathematical relationship among input-output data sets must be known. This problem is solved by using soft computing as fuzzy logic, neural network and genetic algorithm that have high ability to identity and control these nonlinear systems.
Drinking water treatment plants have to provide high quality drinking water in shortest possible time with minimal costs (Bratby, 2006). The treatment process used depends on the quality and nature of the raw water. Conventional treatment process, consisting of primary sedimentation, coagulation/flocculation, secondary sedimentation, and followed sedimentation and filtration are still the major process used for drinking water (Robenson, Shukor, & Aziz, 2009).
Most researches were done on such parts of water treatment process as influence of parameters of raw water in water treatment process on drinking water quality and coagulant dosage in coagulation process in water treatment (Robenson et al., 2009; Van Leeuwen et al., 2003).
Temperature of the raw water can have a significant effect on coagulation and flocculation (Kang & Cleasby, 1995). The pH of raw water is also a critical factor in coagulation process, optimal ALUM dosage accrues at pH between 6-8 (Xie et al., 2012). Turbidity (TU) can affect or be affected by the physical, microbiological, chemical and radiological characteristics of water (Apostol, Kouachi, & Constantinescu, 2011). Raw water alkalinity and water temperature are the most important raw water quality variables in terms of selecting a particular coagulant (Kang & Cleasby, 1995; Xie et al., 2012). Electrolytic conductivity also called specific conductance is a useful test in raw water for quick determination of minerals. Other important parameters that influence coagulant dosage in the water treatment process are color and suspended solids (Heddam, Bermad, & Dechemi, 2012).
Generally, jar tests are used to evaluate coagulant dosage in coagulation process of water treatment (Maier, Morgan, & Chow, 2004; Zhao et al., 2012). The coagulant dosage depends on raw water parameters such as temperature, pH, Turbidity, Suspended solids, Electrolytic conductivity and color.
Nowadays, many studies are being performed on coagulant dosage in drinking water treatment. So far, the polynomial equation, neural network models, and fuzzy structures have been used for predicting and modeling of coagulant dosage. For example, Bazer-Bachi et al. developed two separate models based on polynomial equations used to determine coagulant feed rates for the Clairfont water treatment plant in France (Bazer-Bachi, Puech-Coste, Aim, & Probst, 1990). Van Leeuwen et al. and Baxter et al. developed an ANN model for the prediction of optimal alum doses based on jar tests conducted on surface waters collected in southern Australia and Edmonton, Alberta, Canada (Baxter et al., 2001; Van Leeuwen, Daly, & Holmes, 2005). Also Maier et al. developed artificial neural networks for predicting optimal alum doses and treated water quality parameters in Australia (Maier et al., 2004). Adaptive Neuro-Fuzzy Inference System (ANFIS) method has just been reported by Wu, Guan-De, and Lo Shang-Lien; in order to understand the effect of fuzzy theory in ANN, the authors developed and compared the performance of ANN and ANFIS approaches. The study was conducted at the water treatment plant in Taipei County, Taiwan, and the models were used to model poly aluminum chloride dosing. According to the comparisons, it was found that the ANFIS technique could be employed successfully in modeling coagulant dosage from the available raw water data (Wu & Lo, 2008). Recently, a multi-objective uniform-diversity genetic algorithm (MUGA) has been presented in (Jamali, Hajiloo, & Nariman-Zadeh, 2010; Jamali, Nariman-zadeh, & Atashkari, 2008) which will also be used in this work.