<![CDATA[Micro‐vibrations in HVAC air‐duct structures can compromise energy efficiency and structural integrity, yet conventional contact based vibrometry is constrained by installation complexity and environmental sensitivity. This study develops a non‐contact micro‐vibration detection system by integrating Phase‐Based Motion Magnification (PBMM) with Fast Fourier Transform (FFT) analysis, and validates its performance against a Micro-Electro-Mechanical Systems (MEMS)‐based ADXL345 accelerometer. High‐resolution 4K video (60 fps) of a galvanized, rigid square duct (0.8 m × 0.8 m) was processed through a steerable complex pyramid, ideal band‐pass temporal filtering, and phase amplification (α = 20). The magnified phase sequences were reconstructed and subjected to Hann‐windowed, detrended FFT, with dominant frequency peaks refined via parabolic interpolation. Validation with the ADXL345 time‐series yielded a mean absolute frequency error of 2.53 Hz (σ = 1.06 Hz) across five measurement points. Temporal band‐pass settings effectively suppressed spectral artifacts, and phase gain influenced magnification clarity. Computational performance averaged around 10 s/frame on CPU, indicating real‐time feasibility with GPU acceleration. The proposed PBMM+FFT approach offers a practical, non‐invasive alternative for structural vibration monitoring in HVAC systems.]]>
Copyrights © 2026
