An 80 m high arch-gravity dam has been affected by alkali–aggregate reaction (AAR) since the late 1970s, leading to progressive concrete expansion and significant irreversible upstream displacement. The structural response of the dam is governed by the interaction of thermal effects, stress redistribution, rehabilitation interventions, and nonlinear joint behaviour, making the interpretation of monitoring data particularly challenging. This study focuses on the development of artificial neural networks (ANNs) to support dam safety assessment and improve the interpretation of measured behaviour. A first neural network was designed to evaluate the reliability and internal consistency of concrete thermometer measurements, which constitute a fundamental input for interpretative analyses. The network architecture and input selection were defined based on physical principles governing heat transfer within mass concrete. By learning correlations among temperature sensors and environmental variables, the model proved capable of identifying even minor anomalies in thermal data, increasing confidence in the monitoring data used for safety assessment. A second neural network was developed to reproduce the expected structural behaviour of the dam, characterized by nonlinear mechanisms associated with peripheral joint opening and AAR-induced expansion. The model was trained on the results of a calibrated three-dimensional numerical simulation incorporating AAR kinetics, hydrostatic loading, and temperature effects. This hybrid training strategy enabled the neural network to embed mechanically consistent behaviour while efficiently capturing nonlinear relationships that are difficult to represent using conventional statistical approaches. The strong agreement with monitoring measurements confirms the effectiveness of the proposed approach. The results demonstrate that neural networks, when properly structured and physically informed, can be effectively integrated into established dam monitoring frameworks, strengthening data interpretation and supporting more reliable decision-making in dam safety management.