The results from a water supply network model simulated on a computer should ideally match the values recorded in the field. Such a validation is known as calibration. Good calibration of a supply network can lead to higher standards of service for public health and safety. This study uses genetic algorithms (GAs), a relatively new approach to optimization, to refine the demand values associated with nodes in a network. This approach is used to obtain a more precise hydraulic calibration of an existing network. Suitable tracers, such as fluoride supplied to the network of pipes, may help to make the calibration procedure significantly easier. To accomplish this purpose, a GA-based computer code was generated. The methodology introduced here uses fluoride concentrations from an existing, network in the U.S. as the basis for the calculation of nodal demands. In order to check the validity of the model further, a fictitious network is also analyzed.