Research on precursory signal identification and strength prediction of granite failure under uniaxial compression

Aiming at the problems of strong suddenness of brittle failure of granite, difficulty in identifying precursor signals, and insufficient generalization ability of strength prediction models, granite from Longchang, Sichuan Province is taken as the research object, and a coupled stress-acoustic monitoring test under uniaxial compression is carried out. Based on the test results, the sensitivity of acoustic emission (AE) characteristic parameters in warning rock failure is quantitatively evaluated by the defined dimensionless parameter precursor response ratio (PRR). Then, the grey relational analysis (GRA) method is used to optimize the strength influencing factors, and compressive strength prediction models are constructed based on LSBoost, Transformer, and KELM algorithms, respectively. The model accuracy, robustness, and generalization ability are evaluated by multiple indicators, and the full-range local dependence method is introduced to conduct parameter sensitivity analysis. The results show that the PRR value of AE absolute energy is the highest (the maximum, average, and cumulative values are 309.77, 246.12, and 349.21, respectively), indicating that absolute energy is the most sensitive precursor indicator for identifying granite failure. A two-level early warning threshold based on sliding-window average PRR is preliminarily established, which provides a quantitative reference for damage state identification of granite specimens. Duration and rise time are precursor indicators with moderate sensitivity and can be used as auxiliary verification parameters, while the signal-to-noise ratios of ringing count and amplitude are low and thus should not be used alone as warning parameters. Under small-sample nonlinear datasets, the LSBoost model shows the best accuracy and robustness, with a test set coefficient of determination ( R ²) of 0.84 and a mean absolute error (MAE) of 3.34. In contrast, the Transformer model demonstrates good anti-overfitting ability, while the KELM model is prone to overfitting. Young’s modulus is identified as the core driving factor determining the uniaxial compressive strength of granite, with a resulting strength fluctuation range of 22.59 MPa. Local dependence analysis shows that Young’s modulus has a positively correlated step effect. The research findings provide methodological reference and theoretical support for rapid strength assessment and failure precursor warning of granite rock masses, and lay a foundation for the transformation from laboratory research to engineering non-destructive testing scenarios.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-19
DOI
https://doi.org/10.1038/s41598-026-81357-5
Primary Topic
Rock Mechanics and Modeling
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Research on precursory signal identification and strength prediction of granite failure under uniaxial compression

Junqiao Hu, 庄亚欣, Changqing Li, Erfeng Zhang et al.
Scientific Reports
Rock Mechanics and Modeling
article

Research on precursory signal identification and strength prediction of granite failure under uniaxial compression

Junqiao Hu, 庄亚欣, Changqing Li, Erfeng Zhang, Enyao Li, Peineng Yan
article en

Abstract

Aiming at the problems of strong suddenness of brittle failure of granite, difficulty in identifying precursor signals, and insufficient generalization ability of strength prediction models, granite from Longchang, Sichuan Province is taken as the research object, and a coupled stress-acoustic monitoring test under uniaxial compression is carried out. Based on the test results, the sensitivity of acoustic emission (AE) characteristic parameters in warning rock failure is quantitatively evaluated by the defined dimensionless parameter precursor response ratio (PRR). Then, the grey relational analysis (GRA) method is used to optimize the strength influencing factors, and compressive strength prediction models are constructed based on LSBoost, Transformer, and KELM algorithms, respectively. The model accuracy, robustness, and generalization ability are evaluated by multiple indicators, and the full-range local dependence method is introduced to conduct parameter sensitivity analysis. The results show that the PRR value of AE absolute energy is the highest (the maximum, average, and cumulative values are 309.77, 246.12, and 349.21, respectively), indicating that absolute energy is the most sensitive precursor indicator for identifying granite failure. A two-level early warning threshold based on sliding-window average PRR is preliminarily established, which provides a quantitative reference for damage state identification of granite specimens. Duration and rise time are precursor indicators with moderate sensitivity and can be used as auxiliary verification parameters, while the signal-to-noise ratios of ringing count and amplitude are low and thus should not be used alone as warning parameters. Under small-sample nonlinear datasets, the LSBoost model shows the best accuracy and robustness, with a test set coefficient of determination ( R ²) of 0.84 and a mean absolute error (MAE) of 3.34. In contrast, the Transformer model demonstrates good anti-overfitting ability, while the KELM model is prone to overfitting. Young’s modulus is identified as the core driving factor determining the uniaxial compressive strength of granite, with a resulting strength fluctuation range of 22.59 MPa. Local dependence analysis shows that Young’s modulus has a positively correlated step effect. The research findings provide methodological reference and theoretical support for rapid strength assessment and failure precursor warning of granite rock masses, and lay a foundation for the transformation from laboratory research to engineering non-destructive testing scenarios.

Scientific Reports
China Coal Technology and Engineering Group Corp (China) (CN), Guangdong Ocean University (CN)
Rock Mechanics and Modeling
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon