Novel nonlinear feature identification in vibration-based damage detection using local attractor variance
Recent research in the broad scope of structural health monitoring has introduced a somewhat new paradigm within the context of vibration-based damage detection. The incorporation of statistical-based analysis has shown great promise in more accurately assessing the appearance, the location, and the scope (Levels I-III) of damage in structures from measured global vibration properties. One of the most important and least-developed aspects of the newer paradigm is the problem of "feature extraction", or identifying the most appropriate measurements, whether direct or indirect, for sensitively assessing damage. In this paper, we propose a feature extracted from a nonlinear time series involving attractor variance and test it on an eight-degree-of-freedom structure subject to a linear damage model.
