Optimal layout and hydraulic design of branched networks using genetic algorithms

Abstract

A genetic algorithm (GA) model for the optimum layout and optimum hydraulic design of a branched pipe network was developed. The primary objectives of the model were to identify the optimum connections of demand nodes to the source node, the optimum pipe sizes and an appropriate pump. A methodology for the application of the GA technique to branched pipe networks for the optimum design was presented. A standard GA with three traditional operators was employed. Any branched pipe network consisting of a single source node and a number of demand nodes may be optimized using this model. The model examined a case study consisting of 11 demand nodes and a source node. The model with tournament selection found the optimum solution with a cost of $107,036 within 788 s (on a DEC station 5000/240 running on Unix) while with proportionate selection found the solution with a cost of $110,840 within 1136 s. In order to find a new set of values of GA parameters that may perform better, sensitivity analysis was carried out. The minimum cost solution with the new values of GA parameters was obtained at a cost of $106,534.