Planning the rehabilitation and strengthening of distribution networks using genetic algorithm software

Over the last decade, computer techniques known as Genetic Algorithms (GAs) have been developed which are capable of optimizing very large complex non-linear problems with multiple local optima. Based on a model of the mechanisms by which species evolve in nature, GAs have rapidly gained acceptance among the international research community as perhaps the best numerical search technique for optimal decision making with very large engineering systems. This paper demonstrates how these techniques are now being applied to real engineering planning and design in the area of piped water distribution systems. The Genetic Algorithm is presented in outline, the reader being referred to previously published works for more detail. The methods by which these techniques can be applied to the planning and design of improvements to water distribution networks are explained in detail, in the context of two commercial projects recently undertaken by the authors. The first was to identify the cheapest combination of alternatives for replacing over 500 old cast iron pipes in a large distribution network. The second was to identify the cheapest way of reinforcing an existing 700 pipe network with extra pipes to meet increased demands. Other previous and current areas of application are summarized, and the potential for Genetic Algorithms to assist engineers and planners in future projects is discussed.