Multi-Parameter Nonlinear Acoustic Emission Precursors of Failure in Coal with Different Burst Tendencies

Acoustic emission (AE) monitoring is widely used to characterize coal failure, but specimens with different burst tendencies cannot be distinguished reliably using a single count, energy, b-value, or fractal indicator. This study reanalyzed archived Vallen AE data from uniaxial-compression tests on five strong-burst, five weak-burst, and three specimen-matched non-burst coal specimens. Thirteen VisualAE event tables were verified against independently decoded primary-data files; hit counts were identical and cumulative-energy differences were below 1%. Vallen C and c records were identified as transmitted and received calibration pulses and were excluded consistently from the physical-AE analysis. Calibration records contributed mean energy shares of 11.1%, 4.3%, and 92.5% in the strong-, weak-, and non-burst groups, respectively. After exclusion, the top 1% of retained events contributed 96.6%, 97.9%, and 66.6% of the AE energy; mean b-values at Ht + 5 dB were 0.788, 0.793, and 1.650; and raw-energy multifractal widths were 1.933, 1.729, and 1.219. The correlation dimension depended strongly on embedding, delay, and scaling-range choices and did not show a universal late-sequence decrease. An exploratory six-component AE multi-parameter index yielded group means of 0.659, 0.798, and 0.150. The results support complementary, explicitly parameterized AE sequence descriptors, while the non-burst sample size (n = 3), absence of strict machine-AE time synchronization, and field-scale transfer requirements limit generalization.