Reducing operations and maintenance costs of water mains is a strategic goal. In particular, reducing the costs incurred due to water pipe failures can lead to a significant reduction in time, monetary costs, and overall improved quality of service. These costs include the costs of diagnosing, and repairing failures, the costs for the lost water as well as the costs related to the downtime of the faulty sub-system. The goal of this research is to develop algorithms for predicting, diagnosing, and repairing failure such as leaks and bursts in steel water mains. In particular, we intend to develop algorithms able to perform risk analysis for leaks and bursts, outputting the probability that a given pipe will leak or burst in the near future. A significant challenge in developing the algorithms is that network condition and failure data is only available for a certain percentage of the pipes network. Thus, the developed algorithms have to be designed to be robust to noisy data and missing information.