Numerical analysis of techniques to extract bridge dynamic features from short records of acceleration

The use of drones in Structural Health Monitoring (SHM) to charge sensors mounted on a bridge and download their data has gathered interest over the last years. This approach presents the advantage of avoiding the need for long cables running over the bridge or for permanent access to electric power on site. Nonetheless, limitations exist regarding the amount of data that can be transmitted through this method. In contrast to traditional approaches to SHM relying on long records to assess the condition of a structure, the scenario envisioned here deals with short amounts of data. In this paper, specific methodologies for extraction of dynamic features from short data bursts of acceleration signal are tested through numerical simulations. The bridge is modelled as a simply supported finite element beam model that is excited by a series of moving concentrated forces, which represent a random traffic load. Initial conditions are varied allowing for scenarios in which the acceleration record may start once the vehicle is already on the bridge, finish before its exit or combine periods of free and forced vibration. The theoretical acceleration response is obtained for healthy and damage conditions of the bridge, and then corrupted with noise. Focus is placed on how effective these techniques are in overcoming the shortcuts derived from noise and from the short duration of the signal. Therefore, techniques to mitigate common problems such as mode-mixing and edge effects are investigated.