SHM is generally performed using continuous vibration or deformation monitoring. These signals correspond then to time series, whose evolution can be predicted based on their past history. This paper presents such a study on the Austerlitz bridge in Paris: it is a steel bridge, built during the 19th century, and carries the Parisian metro since then. The monitoring is therefore important to detect anomalies in its behavior. For that, it has been instrumented by 24 strains sensors (optic strands) and 4 temperature sensors, which record the corresponding signals during train passages at a 100Hz frequency. The individual passage signals have been extracted, the corresponding data has been gathered (data fusion) and cleaned. Finally, the evolution of the monitoring signals has been predicted based on their evolution with time and temperature. Machine Learning algorithms of type VARMAX have been used to realize the time prediction. The anomalies, defined as differences between the prediction and measured signals, have been identified. This paper will explain the context and the research questions, the work that has been realized and to finish the still-open points.