Characterizing the fracture evolution of sandstone by using Mel-frequency cepstral coefficients of acoustic emission

Study of damage and fracture of rocks is important for monitoring and forecasting of dynamic hazards in mines, tunnels, rock bodies and other engineering. Under the conditions of uniaxial loading, stepwise loading, cyclic loading, sandstone loading and destruction experiments were conducted and full waveform monitoring of acoustic emission (AE) was synchronized. The Mel-frequency cepstral coefficient (MFCC) of AE signals was extracted from the whole sandstone loading process. Then, the MFCC was combined with the video camera of sandstone fracture process and the strain analysis using digital image correlation (DIC), to characterize the response of MFCCs for the sandstone fracture process, and to examine the relationship between the MFCC of AE and the fracture of sandstone. In addition, the advantages and underlying mechanism of this coefficient for characterizing sandstone fracture were discussed. The results show that, for different loading conditions, there are fluctuations of MFCC-1 in the loading stage, among which, the fluctuations of MFCC-1 are small and random in the compression and elastic stages, and the fluctuation amplitude significantly increases and shows significant periodicity in the unstable deformation and post-peak failure stages, whereas the MFCC-1 is approximately constant with almost no fluctuations in the load-holding and unloading phases. A large number of new fractures in the sandstone start from the unstable deformation stage at the late loading period and continue throughout the whole failure process, and the new fractures are generated in a regular, intermittent and violent manner, resulting in AEs in the form of intermittent high-amplitude waveform clusters; in the compaction and elastic stage, the new fractures are fewer in number, and the closure friction of the original fracture is dominant, resulting in the random generation of AE in the form of a single waveform. Unlike conventional parameters such as AE energy and ringing counts, which mainly characterise individual waveforms, the MFCC mainly characterises the overall AE waveforms over a period of time, and is less affected by individual waveforms. This can explain why the MFCC exhibits different responses at different loading stages of the sandstone and has an advantage in characterizing sandstone rupture. Among them, the fluctuation of MFCC-1 is reflection of the fracture and damage process of sandstone, the significant periodic fluctuation of MFCC-1 is a precursor of sandstone destabilization and damage, and the significant sudden rise during the fluctuation of MFCC-1 foretells the emergence of macroscopic cracks in sandstone. The MFCC can be used as an important parameter in the analysis of AE. The research results provide a new means of analysis of the evolution of fractures in coal rock.