PSI - Issue 77

Francisco Afonso et al. / Procedia Structural Integrity 77 (2026) 584–592 F. Afonso et al. / Structural Integrity Procedia 00 (2026) 000–000

588

5

Finally, a PCB Piezotronics 356A17 tri-axial accelerometer was also used with a NI 9234 DAQ module, for data validation, with an acquisition rate of 1000 samples per second. Using the previously described components, it is possible to summarize the tested configurations as: • Neuromorphic camera with standard lens and unbalanced fan. • Accelerometer with unbalanced fan. • Neuromorphic camera with macro lens and balanced fan. • Frame camera with macro lens and balanced fan, using Image Tracking or Digital Image Correlation.

3. Results and discussion

3.1. Neuromorphic camera with standard lens and unbalanced fan

Bands containing printed parallel lines were positioned both on the side and top of the fan, enabling measurements in two orthogonal planes. With the standard lens, the vibration amplitudes were too small to be clearly observed and, therefore, an M4 nut was attached to one blade, unbalancing the rotor and increasing the vibration amplitude. This test was conducted with an earlier version of the fan holders, which may a ff ect the final measurements. As the fan vibrates, the stripes will appear to move primarily in the direction perpendicular to their orientation, since this motion drives brightness changes at the stripes’ edges, triggering events in the neuromorphic camera. Figure 4 shows the fan’s side with the horizontal and vertical bands, as well as their corresponding event activity through Prophesee’s Metavision Studio. White points indicate positive events (brightness increase) and blue points indicate negative events (brightness decrease). Due to the oscillatory nature of the motion, the polarity of events alternates over time.

(a) Bands on fan’s side.

(b) Fan during operation, as seen from Prophesee’s Metavision Studio.

Fig. 4: Neuromorphic camera and standard lens setup, measuring the fan’s side.

Both views yielded agreeing results: the horizontal band produced clearly visible motion-related events, while the vertical band was barely distinct, indicating much smaller vibration amplitudes in that direction. The dominant frequency was approximately 31.6 Hz, which, although measured with a di ff erent holder configuration, is close to the 31.8 rps obtained with the tachometer for the unbalanced fan. This procedure was repeated with the camera observing the top of the fan (Figure 5), alongside the vibration measurement from Prophesee’s Metavision SDK 4. The horizontal band is highlighted in red and the vertical band in blue.