ABSTRACT
ABSTRACT: Pipe sizing is the most common task in the design of water networks carried out for enforcing water distribution systems, e.g. by adding parallel pipes, adding new connections or replacing old pipes. A typical objective is to minimise capital cost whilst satisfying hydraulic performance of the network. Pipes are manufactured in discrete diameters, where available sizes depend on the pipe material. Approaches to the pipe-sizing problem can be broadly grouped into the three categories. The trial and error method using a hydraulic simulator, the optimal design using nonlinear programming and the optimal design using discrete search methods such as genetic algorithms. The trial and error method is time consuming and requires a significant experience from the user. The nonlinear programming approach may lead to finding a local rather than the global solution. The discrete search methods can potentially find the global optimum but they require long calculation time. The paper proposes a hybrid approach by combining the nonlinear programming and the evolutionary approaches. The method is efficient in finding the global optimum in a comparatively short time and can facilitate significantly the interactive design procedure.
