A potable water supply network design methodology was implemented using a Genetic Algorithm (GA). Information about location and network connectivity, pipe materials and minor losses coefficients is required. The GA searches for designs (diameter combinations) optimal from the point of view of cost of the networks' pipes while satisfying supply for demanded volumes and minimum pressures in the network junctions. Adjustment of GAs to this problem is shown: functions of adaptation (construction cost), recombination (crossover) and mutation are modified to suit it. A scheme (Cost-Restriction function) is introduced to induce the algorithm to generate designs that fulfill the required hydraulic restrictions. The variation of the cost of resulting design based on probability of mutation, generation size, and coefficients specific to the adaptation function are studied. Special attention is given to the influence of the factors that vary the diversity within each generation of the algorithm (represented by the probability of mutation and one of the variables of the reproduction function). Finally the designs obtained through the algorithm are compared with the designs obtained according to Featherstone criterion.