Noise Diagnostics at AE Monitoring of Hazardous Industrial Assets

Abstract

This chapter considers the acoustic emission (AE) signal-noise discrimination problem that arises at AE monitoring of polymerization reactor. Different kinds of AE data including AE hit sequences, waveform set, and noise RMS (root mean square) samples were recorded and processed both in time and frequency domains to solve the problem. It was found that at a signal-to-noise ratio exceeding 6–10 dB most of the operation noises can be discriminated from AE data by means of AE parameter prefilters and rather simple post-processing recognition procedures. It was shown, however, that in the case of distant AE sources, when the characteristics of true and noise signals are similar, additional information can be successfully extracted from RMS data, particularly when a random wideband noise is modulated by the low frequencies related to the working equipment. Such RMS-based information not only helps to interpret results but also can be used for the purposes of diagnostics of asset-operating condition. Alternative method is proposed, in which the correlation analysis of AE hit sequences is used to determine the operating noise resonances responsible for the formation of false AE hits; that is, in this case relevant information is extracted directly from AE hit data.