Use of Neural Networks in Design of Coastal Sewage Systems

The collector network of a coastal sewage system is normally combined and evacuates domestic and industrial runoff wastewaters in rainy weather. This means that overflows can arise at certain points in the network in rainy weather. The volume and duration of these overflow points and, consequently, the capacity of the coastal interceptor-collectors, are conditioned by the environmental regulations governing receptor waters, and more specifically, in most cases, by those governing bathing waters. The most widely accepted criteria on the quality of these waters are those established by the corresponding directive in terms of exceedance time of fecal coliform concentration limits [nonfulfillment time (NFT)]. This random variable is itself conditioned by a series of factors, and can be studied by means of complex models with high computing costs. This paper briefly presents a method with a probabilistic focus for studying coastal sewage systems, in which the use of neural networks is proposed as a means of reducing computing costs. More specifically, neural networks with optimum functional link are used, in which optimization is carried out genetically.