Recent advances in artificial intelligence (AI) and machine learning (ML) have enabled significant progress in pipeline failure prediction. However, research focusing on critical hydraulic components -as particularly valves are- remains surprisingly limited despite their operational importance.

Based on the foregoing, this study proposes a predictive model for hydraulic valve failures using artificial neural networks (ANNs) to mitigate operational risks and costs associated with malfunctions. Considering that hydraulic valves are critical components in water supply and distribution systems, implementing an AI-based failure prediction system will enable early fault detection, optimize maintenance planning, and reduce costs linked to emergency repairs.

The methodological framework includes collecting operational data—such as pressure, flow rate, temperature, and vibration measurements—from a set of valves. These data will be pre-processed to train an ANNs capable of identifying patterns associated with potential failures. Different ANNs architectures will be evaluated, including multilayer perceptrons (MLPs) and recurrent neural networks (RNNs), to determine the most accurate approach for early anomaly detection. The model will be validated using real-world operational data, with performance comparisons against statistical models and other techniques documented in the literature.

The expected outcomes include precise fault identification and, where feasible, estimation of the valves remaining useful life (RUL). These results will support a transition toward predictive and proactive maintenance strategies. Furthermore, this approach could be extended to other hydraulic components, thereby improving the efficiency of water supply and distribution systems.