With the advent of 5G and the future 6G networks, the increasing number of systems interconnected to the network, generates extremely huge amount of data. The autonomous structural health monitoring (SHM) of many structures and bridges represent an important sensor network application that generate a considerable amount of data that must be elaborated and managed. In such scenario, this paper proposes a set of machine learning (ML) tools to perform automatic feature selection and detection of anomalies in a bridge from vibrational data. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric measurements in both standard and dam- aged conditions. The proposed framework starts from the first two fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by density-based time-domain tracking algorithm. The fundamental frequencies extracted are then fed to one class classifier (OCC) algorithms that perform anomaly detection. Finally the effect of reducing the number of sensors used to monitor the network, the number of bits used to quantize the accelerometric measurements, and the observation time is reported, with the purpose to reduce the amount of data stored without deteriorate the damage detection capability of the system. In numerical results a widely comparison of OCC algorithms is reported, more in detail, principal component analysis (PCA), kernel principal component analysis (KPCA), auto-associative neural networks (ANN) and one class classifier neural network (OCCNN2) are tested and their robustness evaluated. In many cases, OCCNN2 algorithm increases the performance with respect to classical anomaly detection algorithms in terms of F1 score and accuracy. Moreover it is observed that only three sensors are enough to accomplish the anomaly detection task and also a reduced number of bit and observation time can be used without affect the algorithms performance.