The possibilities of using artificial neural networks as a support tool in failure rate analysis were shown. On the basis of failure rate the reliability of water pipes could be established. In unusual and very frequent situations (epidemic, war, floods, droughts) it is necessary to make a quick reaction when damages occur in water-pipe networks. Modern IT tools seem to be an alternative method in the analysis of current situation in management of critical infrastructure. On the basis of exploitation data of one water system the investigations were carried out. Totally 40 models of artificial neural networks were built. Sensitivity analysis and assessment of usefulness of chosen models for dependent variable prediction using quality and quantity predictors were performed. The obtained results point out that further investigations are necessary. The changing exploitation situation forces improvement and using mathematical tools in new conditions of water-pipe networks.