PSI - Issue 78

Laura Gioiella et al. / Procedia Structural Integrity 78 (2026) 1436–1442

1439

upsampling factor for the UCC algorithm (set to 100); ( iv ) expected displacement magnitudes; ( v ) size and number of targets required to meet the above constraints; ( vi ) processing capacity of the laptop used to handle the data stream from the camera sensor. Two Teledyne FLIR BLACKFLY S BFS-U3-23S3M-C video cameras were employed, each connected to a laptop via USB 3.0. The roof camera was fitted with a 16 mm focal length lens and configured with a FOV of 800 × 700 pixels (width × height) to capture the displacement of targets located at: L0 (reaction mass, 200 × 200 mm); L1 (shake table, 200 × 200 mm); L2 (+4.32 m, 100 × 100 mm); L3 (+7.67 m, 100 × 100 mm); L5 (+14.38 m, 50 × 50 mm). In this setup, L0 and L1 represent, respectively, the roof displacement with and without the shake table motion. The ground camera, designed solely to measure the lateral (y) roof displacement for redundancy and noise control, was equipped with a 50 mm focal length lens. Its FOV was set to 900 × 300 pixels, capturing a single target (T7, 100 × 100 mm) placed within the roof floor thickness (+21.08 m). Figures 2(b) and 2(c) show the FOV and targets used for the roof and ground cameras, respectively. All recordings were acquired at 100 frames per second (FPS), chosen as an optimal balance between hardware limitations, video file size, and expected measurement performance, given the selected parameters (FOV, ROIs, and target dimensions).

b

a

c

Fig. 2. (a) testbed building; (b) FOV and targets acquired by the roof camera; (c) FOV and targets acquired by the ground camera.

3.2. Seismic input and results More than fifty playbacks were conducted during Phase II testing. Among them, this study focuses on selected results for the Niigata ground motion, applied in three dimensions at the intensity level of the Risk-Targeted Maximum Considered Earthquake (MCEr), corresponding to a peak ground acceleration (PGA) of 0.607 g in the lateral (y) direction of interest. Fig. 3 compares the roof lateral displacement obtained from the roof camera (target T0) and the ground camera (target T7). The lower close-up, showing the first seconds of motion, highlights that the signal from the roof camera is significantly affected by high-frequency disturbances induced by the shake-table, whereas the ground camera signal appears cleaner. Nevertheless, the ground camera data are still noisier than expected, likely due to the sensor’s installation on a tripod approximately 26 m from the platen. The upper close-up presents the final portion of the recordings, corresponding to the building ’ s free vibration phase (fundamental period ≈ 1.05 s). In this phase, with the shake-table no longer operating, the noise level in both signals is greatly reduced, revealing a much clearer correspondence between the two measurements.