Structural health monitoring using time-delay embedding and phase-space warping

This paper presents a method to utilize Time-Delay Embedding (TDE) to perform structural health monitoring and obtain a damage metric that tracks deterioration in structural health over time. Proof-of-concept is demonstrated through an experimental investigation. The experimental setup consisted of high-rise building model, made of aluminium, mounted on a digitally-controlled shaker table. Floor accelerations are recorded using accel-erometers on each elevation to capture the dynamic response during the base excitation. Natural modes of vibration and their corresponding frequencies were determined experimentally using a frequency sweep of low-amplitude sinusoidal support excitations. The model building was excited at its natural frequency with moderate-amplitude sinusoidal base excitation to establish its "pristine" condition baseline. Subsequently, damage is introduced to columns on the first floor by cutting two notches at mid-height, significantly reducing their cross-sectional areas. The same base motion protocol was re-applied and acceleration data was collected and processed to detect the pre-known damage patterns as well as the total failure using changes in the oscillation orbits. The damage metric is shown to adequately detect structural damage.