Sparse approximation method in the analysis of acoustic emission signal morphological features

• download

   Acoustic emission is the phenomenon of elastic wave radiation by an object
   under examination during its nonlinear transformation. The purpose of the
   investigation of acoustic emission signals is to obtain information on
   solid body restructuring. The acoustic emission signal contains acoustic
   emission single pulse flow distorted by wave propagation channel and
   carries information on dynamic restructuring in the object, a generation
   source. There is a significant distortion of the original pulse, when it
   passes through the environment and receiving channel of a recording
   device, a characteristic signal is recorded. During the activation of a
   plastic process, many elementary sources are formed, that generate
   acoustic emission single pulses, which, in their turn, blending and
   distorting additively, form the resulting signal registered by receivers.
   It is important to detect the pulse morphological structure according to
   the data recorded by a receiver. The authors suggest to apply the matching
   pursuit method providing the sparse signal model. This method sequentially
   extracts functions, which minimize signal approximation error, from the
   family of characteristic wave forms (dictionary). The basis of this family
   is the dictionary, consisting of Berlage pulses, which form correlate with
   the signal most accurately. The paper analyzes signals by the proposed
   method, specific classes of acoustic emission pulses are distinguished,
   the behavior of approximation error is analyzed, and the model for a
   signal is suggested.