Development of the technique of clustering by acoustic emission signal parameters

The technique of clustering by a set of parameters of acoustic emission (AE) signals, which allows reducing the time of processing of the recorded data, is considered. A comparative analysis of the stability of three clustering techniques (by shape, leading edge rise rate, and a set of AE signal parameters) with respect to the effect of random noise distributed according to a normal law is performed. A possibility of using the considered clustering techniques for distinguishing different stages of duralumin specimen fractures and during welding of steel specimens is experimentally demonstrated.