VIBRO-ACOUSTIC SIGNATURE COMPARISON AND TIMEWARPING CORRECTION WITH MULTI-SCALE CORRELATION

A vibro-acoustic signature in the context of power transformer tap-changers comprises a burst of transients generated by a sequence of electromechanical events. Comparison of vibro-acoustic signatures lacks precision unless the events in the sequence are synchronised to make juxtaposition of the signatures possible. Time warping offers a solution of this problem in that it temporally reorganises the position of the elements of one of the signatures so that it matches the time frame of the other. Furthermore, the scale of the time warp has a useful diagnostic function. However, the dynamic time-warping (DTW) algorithm used in speech processing is inapplicable in the context of electrical switchgear since some discontinuities are present in the time frame. To overcome this DTW limitation, we propose a multi-scale correlation algorithm which processes these discontinuities adequately and allows us to find the time relationship between the events of two signatures. This paper presents the operating principle of both these algorithms along with results that bear witness to the sharp time-warping correction obtained with multi-scale correlation. This correlation can even correct the order of appearance of events so that the time sequences are juxtaposed and comparison of signatures can be undertaken.