The research presented in this paper aims at the support tool for generation of multiple networks with preset or randomised properties. To explore particular phenomena, water distribution analysis may require a coherent set of cases. Readily available in the literature are simple synthetic networks used for benchmarking, either real-life cases that are too diverse in size and configuration. The network generation tool (NGT) developed on the principles of graph theory connects any seed of nodes prepared in EPANET modelling software, by avoiding pipe crossings or unnecessary duplications. The pipe properties can be assigned by specifying a range of arbitrary lengths and diameters, by using coordinates to calculate the lengths, or by genetic algorithm optimisation of initial diameters. Equally, the nodal elevations and demands are arbitrarily assigned when not predefined in EPANET. Several sets of networks have been generated, up to 200 junctions. To test robustness of the tool, 13,000 layouts of a 50-junction seed have been generated using different settings. NGT has been proven to be capable of executing this task mostly within a few minutes, producing network layouts that resemble those from practice.