Automatic Hit Detection in High-Noise Environments Using Frequency Domain Data of Continuous AE Waveform

Abstract

Acoustic Emission method is one of the few powerful non-destructive evaluation techniques capable of real-time monitoring of microscopic damage occurring during materials manufacturing or processing. Consequently, it holds great potential for application in process monitoring and quality control. However, measurement environments in actual manufacturing and processing sites are generally very noisy. Furthermore, as the process progresses or equipment conditions change, the frequency characteristics and amplitude of the noise often fluctuate discontinuously or periodically. To consistently obtain appropriate waveform analysis results under such measurement environments, manual adjustment of noise filtering conditions and AE event detection threshold voltages are often required. However, this operation demands expertise and compromises real-time performance, ultimately preventing the full utilization of AE method's inherent real-time advantage. Therefore, this research proposes a method to achieve fully automated AE analysis that does not require manual adjustment of analysis parameters even in fluctuating high-noise environments. This is accomplished by employing continuous waveform measurement, converting the continuous waveform into frequency-domain data (i.e., spectrogram, time-frequency-intensity data), and detecting AE hits within this frequency-domain data.