Statistical signal parameters of acoustic emission for process monitoring

Abstract

Many technical processes, e.g. in mechanical engineering, are causing acoustic emission. Acoustic emission (AE) consists of elastic waves, generated by stress changes in a solid. These waves can be detected at the surface of the solid by piezoelectric sensors. Classical methods to characterize acoustic emission signals include detecting and counting single events, describing their energy and frequency properties. The spreading conditions for acoustic waves in solids and the interference of a large number of AE sources lead to quasi-continuous signals from which no individual AE event can be extracted. This is also typical for wire sawing. If AE signals shall be used for online process monitoring, it is necessary to extract signal properties that are correlated with process changes. A common feature is the RMS value of the signal, which is correlated with the energy of AE and was found to be very sensitive to changing process conditions. Other features used are the peak values of the signal and the number of zero crossings. To get more information about the actual state of the observed process, parameters of the statistical distribution of short-time RMS like mean value, variation coefficient and skewness have been tested and their sensitivity to process changes have been investigated. An online monitor has been developed based on a hard- and software concept, adapted to process continuous acoustic emission data, with fast acquisition rates and signal processing