Evaluating Onset Times of Acoustic Emission Signals Using Histogram Distances

Determining the onset of a transient signal, for example in seismograms, acoustic emission (AE) signals, or ultrasonic signals is very important in non-destructive process monitoring and geophysics. In some cases, the AE from a malfunction is relatively weak, with a low signal-to-noise ratio. Thus, the signals are often hidden, making it rather difficult to separate them from the background noise. The present work proposes a new method of evaluating onset times based on bin-to-bin histogram distances measured with the Bhattacharyya coefficient. A criterion and a standard procedure for determining an onset are formulated. A key parameter, window length, was discussed in detail. Tests on AE signals from a pencil-lead break, single-grit scratching, and filament breakage reveal the feasibility and effectiveness of the proposed method. This method is believed to be especially appropriate when the emission signal strength of the target malfunction is lower than environmental AE signals generated by other stationary sources. It can also be used as an alternative method for identifying onsets in AE signals in process monitoring and other fields.