Online frequency and amplitude tracking in structural vibrations under environment using APES spectrum postprocessing and Kalman filtering

Identifying time-varying frequency and amplitude online in real-life structural vibrations is an essential topic of data processing in structural health monitoring. This paper proposes a novel method for this task. We assume that structural vibration signals are stationary in a short time, thus a spectral analysis method, called amplitude and phase estimation, is conducted to obtain the amplitude spectrum at corresponding time window, and a postprocessing technique is proposed to extract modal frequency and amplitude from the spectrum automatically. The extracted frequency and amplitude could be regarded as the average value of instantaneous frequency and the average value of instantaneous amplitude during the window, respectively. Due to the instability of measured structural vibrations and the uncertainty of spectral shapes under ambient excitation, Kalman filtering is introduced by taking the signal that reconstructed from the identified frequencies and amplitudes as the prediction to enhance the reliability and quality (signal-to-noise ratio) of the next measured signals. Numerical study is performed to inspect the performance of the proposed method. It is also employed to analyze the vibration signals of actual structures, i.e., a cable of a cable-stayed bridge, a hanger of an arch bridge and the main girder of a suspension bridge. The results show its potential to track frequency and amplitude in structural vibrations under environmental measurements. The method is supposed to provide fundamental support for further information obtaining and high-level decision making for structural health monitoring systems. •A novel method for automatic frequency and amplitude extraction in real-time is proposed.•Kalman filtering is introduced for better tracking of frequency and amplitude under ambient excitation.•The method proposed exhibits better tracking performance in the case of low SNR.•Three long-time continuous vibration signals generated by actual structures are investigated.