During selection of optimum sewerage and wastewater treatment systems in rural settlements, often a large number of potential technical solutions is generated. The economic criterion is most frequently assigned the greatest importance which finally results that the solutions involving the lowest total costs are preferred. The conventional approach considerably complicates the choice of the optimum solution, because it requires great efforts in determining the size of considered solutions and preparing of corresponding cost estimates.
The paper analyzes the possibility of using of neural networks in economic analyses of wastewater systems. The neural network NENECOS (NEural Network for approximate Estimation of COsts of wastewater Systems) has been created, which allows simple, fast and adequately accurate estimation of total or unit costs of construction, operation and maintenance of sewerage systems, without the need for prior sizing and preparing of cost estimates. This allows simple and more efficient economic comparison of a greater number of alternative solutions. The limitation of the neural network NENECOS is the possibility of approximate estimation only for smaller rural settlements up to 500 population equivalents.
Selection of the optimum wastewater collection and treatment system in rural areas is a complex and demanding process. This statement is the result of the fact that nowadays there is a large number of available technical solutions of wastewater collection and treatment applicable in small rural settlements. So far, practice has shown that in most cases it is possible to generate a large number of potential combinations of wastewater collection and treatment procedures, which makes the entire process of selection of the optimum solution more difficult (Hamilton, 2004, Pinkham et al., 2004, Vouk, 2006, Vouk and Malus, 2007 and Vouk et al., 2008).
Likewise, the practice has shown that in selection of the optimum solution the greatest importance is assigned to the economic criterion. Therefore, in discussing of a larger number of alternative solutions, determining of total costs of their construction, operation and maintenance is considered important. The conventional method of conducting of economic analyses is based on previous dimensioning of all considered wastewater collection and treatment technical solutions. Dimensioning of considered solutions represents the basis for working out of approximate cost estimates per individual items (construction of gravity sewers, construction of pressure sewers, number of pumping stations, operation costs, maintenance costs, etc.). Results of approximate cost estimates represent the basis for economic comparison of alternative solutions. Obviously, such procedure in most cases (larger number of alternative solutions) is very time-consuming, demanding larger efforts and a higher level of expertise.
The purpose of this paper is to offer the possibility of simple and quick economic analysis for different technical and technological solutions of wastewater collection and treatment. In this context, the authors consider the use of neural networks a useful and adequate solution.
Reviewing of recent worldwide experience and practice shows that in the few recent years the number of professional and scientific papers discussing the use of neural networks in the field of sanitary engineering has been constantly growing (Baruch et al., 2005, Bowers and Shedrow, 2000, Chen et al., 2003, Dandy et al., 1998, El-Din and Smith, 2002, Khalil et al., 2006, Kunwar et al., 2009, Shukla et al., 1996 and Sperac, 2004). However, the absence has been noticed of auxiliary tools that might be used for economic analyses of wastewater collection and treatment systems.
In the given framework of efforts, the neural network NENECOS (NEural Network for approximate Estimation of COsts of wastewater Systems) was created, which allows simple, fast and adequately accurate estimation of total or unit costs of construction, operation and maintenance of different wastewater collection and treatment systems, without the need for prior dimensioning and preparing of cost estimates. The procedure will enable the users to get, in a simple and quick manner, the insight into economic feasibility of individual solutions. This will enable the users to take an elimination step, i.e. to eliminate the solutions clearly deviating from the rest in the economic aspect.
The paper will describe the method of creating of the NENECOS neural network, with description of characteristic steps – preparing of the database, network training and testing. Also, the method of the network use, i.e. estimation of output parameters, will be shown.
The use of the described neural network would considerably simplify and accelerate conducting of economic analyses as the basis for comparison of a larger number of alternative solutions and final selection of the optimum alternative.
The neural network NENECOS is intended for all stakeholders participating in the decision-making process, not only for professionals (investors and designers) but also to other interest groups (local authorities and end users). The designers would be able, already in the preliminary design phase to carry out more easily the entire analysis of a larger number of alternatives and select the optimum one. The commissioners of design documentation would also have a simpler access and control of design solutions, with selection of the optimum solution as the final objective.
