Application of Pattern Recognition to the Acoustic Emission Signals of Carbon Steel

Hydrogen Induced Cracking (HIC) is one of failure types of in-service pressure vessels. Acoustic emission (AE) is a good method to monitor HIC. In order to investigate the characteristic of AE signals produced by HIC of carbon steel used in pressure vessel, the hydrogen charging progress of 20R steel was monitored using AE technique. There were three useful AE sources during the test, corrosion and charging, FeS film forming and breaking, and HIC growth, which all can be detected through AE instrument. The amount of big amplitude AE signals were increased obviously with cracking growth and microscopic examinations provided good confirmation. The peak frequency ranges of these three kinds of AE waveform signals were similar, but their distribution of the spectrum magnitude to the corresponding frequency were different and can be used as the discriminating acoustic parameters.Therefore, a pattern recognition method based on frequency feature extraction, fisher ratio and trained back-propagation (BP) network was developed for the sources´ signals class decision. Finally some testing data study was given to show the high efficiency the proposed method. These results provide a useful way of monitoring HIC development in pressure vessels for practical inspection.