A system response persistence measure for use in ambient data monitoring

To aid in the real-time monitoring of system stability, it is useful to assess the relative damping of an underlying impulse response from noisy, ambient data. This is typically done using modal decomposition, procedures that accurately and consistently estimate natural frequencies of oscillation but struggle with estimates of damping. Here we introduce a measure of “persistence” as a robust measure of the duration of an impulse response. The longer a natural response persists the less damping it is perceived to have. The proposed measure scales linearly with the duration of an impulse response. It is easy to calculate as the energy in a normalized autocorrelation signal, and it serves as a measure of (inverse) damping.