Analysis of Real-Time Vibration Data

摘要

In recent years, a few structures have been instrumented to provide continuousrnvibration data in real time, recording not only large-amplitude motions generated byrnextreme loads, but also small-amplitude motions generated by ambient loads. The mainrnobjective in continuous recording is to track any changes in structural characteristics, andrnto detect damage after an extreme event, such as an earthquake or explosion.rnThe Fourier-based spectral analysis methods have been the primary tool to analyzernvibration data from structures. In general, such methods do not work well for real-timerndata, because real-time data are mainly composed of ambient vibrations with very lowrnamplitudes and signal-to-noise ratios. The long duration, linearity, and the stationarity ofrnambient data, however, allow us to utilize statistical signal processing tools, which canrncompensate for the adverse effects of low amplitudes and high noise. The analysis ofrnreal-time data requires tools and techniques that can be applied in real-time; I.e., data arernprocessed and analyzed while being acquired.rnThis paper presents some of the basic tools and techniques for processing and analyzingrnreal-time vibration data. The topics discussed include utilization of running timernwindows, tracking mean and mean-square values, filtering, system identification, andrndamage detection.