Acoustic emission characteristics of coal failure using automatic speech recognition methodology analysis

Abstract

Monitoring acoustic emissions (AE) is an effective way to identify coal deformation and destruction processes. It is therefore of great significance to analyze the characteristics of AE during coal destruction process. This paper applies the Mel frequency cepstrum coefficient (MFCC) approach of automatic speech recognition (ASR) to analyze the characteristics of the AE of coal. The MFCC of AE within 40 ms during the uniaxial compression failure of 55 coal samples was extracted. The results show that the MFCC changes regularly with increasing stress on the coal sample, which changes from the beginning to the end of loading. The ratio of stress to the compressive strength of the coal sample is defined as the stress state of the coal sample and the correlation between MFCC and the stress state of the coal sample is analyzed. MFCC-3 (the third parameter of MFCC) and MFCC-6 (the sixth parameter of MFCC) match the linear change relationship at the relevant stress state. The distribution characteristics of MFCC-3 of 55 coal samples under the same stress state showed that the parameter value is normally distributed under the same stress state. If MFCC-3 is less than -2.481, the probability that stress will reach 90% of its ultimate strength exceeds 93.8%, and the probability of coal failure exceeds 50%. This study shows that the feature extraction method in the field of ASR can be used for the AE feature analysis of the deformation and destruction processes of coal samples, and the extracted MFCC of AE can be used to evaluate their safety state. These results are of great significance to further advance the analysis of the characteristics of the AE of coal.