Dam owners gather monitoring measurements for surveillance and safety purposes, sometimes since the first impoundment. These measurements are often topographic surveys, displacements, piezometric measurements, or even flow measurements. These time series have a rich history, and therefore, today, it is logical to learn about the behaviour of structures from this rich database especially by using machine learning techniques. Among many of these powerful techniques, the use of artificial neural networks (ANN) has proven to be one of the most appropriate machine learning technique when applied to the analysis of data monitoring measurements. The paper briefly sums up the principle of artificial neural networks and their advantages when compared with the historical model HST (Hydrostatic-Season-Time) which is based on the multiple linear regression. Then, the home-made software NOVAE (Neural NetwOrks for Valuable Analysis and Expertise) is presented. NOVAE enables all engineers responsible for dam surveillance to use the technique of ANN when analysing monitoring measurements. The software is directly linked with the database which stores the monitoring data collected from the 320 dams monitored by EDF. Therefore NOVAE makes it easy to get interesting results without the need of specific skills in the field of machine learning techniques, and by simply using the default parameters which are preset. The advantages of using ANN are then illustrated on a case study. In that case, the HST model has a low quality of prediction and is outperformed by ANN. As exhibited by NOVAE, ANN not only improve the numerical performance, but also the interpretation of the behaviour of the dam. Indeed, they allow to connect the loads (hydrostatic, thermal …) acting on the structure and their induced effects, even if their interactions are complex. Finally, the use of the ANN in an industrial context leads to a better assessment of the dams’ safety.