@article{Ding2026, 
author = {Ziwei Ding and Chenchen Zhang and Wenliang Sun and Yunjun Dong and Maoqing Yu},
title = {Failure Mode and Failure Precursor Identification of Layered Sandstone Based on Cluster Analysis},
year = {2026},
journal = {Chinese Journal of Underground Space and Engineering},
volume = {22},
number = {3},
pages = {1090-1102},
keywords = {failure precursor, clustering analysis, acoustic emission monitoring, cyclic loading and unloading, layered sandstone},
url = {https://www.sciopen.com/article/10.20174/j.JUSE.2026.03.34},
doi = {10.20174/j.JUSE.2026.03.34},
abstract = {Layered sandstone is widely distributed in the surrounding rock environment of underground space engineering. The method based on cluster analysis is used to study the failure modes of layered sandstone under cyclic loading and identify failure signals, which can analyze the failure behavior of rock mass under complex stress conditions. Cyclic loading experiments were conducted on layered sandstone samples to monitor the acoustic emission signals of rock failure. The signals were classified using K-means++clustering algorithm, revealing the failure mode of layered sandstone under cyclic loading and identifying precursor signals related to rock failure. The results show that: The macroscopic failure of uniaxial loaded specimens is mainly shear failure, with a mixed shear tensile failure observed in specimens loaded three times per level, and tensile failure observed in specimens loaded five times per level; When the sample is loaded three times per level, micro cracks and large-scale cracks develop together in the early and middle stages. In the later stage, the stable expansion of large-scale cracks is the main trend, and ultimately the main cracks are formed by the penetration of micro cracks. When the sample is loaded five times per level, the large-scale cracks in the sample develop stably and quickly penetrate when approaching failure; The contour coefficients of cluster analysis under different loading conditions were 0.81 and 0.93, respectively, and three types of signals were classified: Type Ⅰ signal was mainly tensile signal, accounting for 39.2% and 37.9% of the two groups of samples; Type Ⅱ signal was mainly mixed failure, accounting for 43.3% and 57.3% of the two groups of samples; Type Ⅲ signal was mainly shear failure, accounting for 17.4% and 4.8% of the two groups of samples, and the frequency of occurrence increased with the failure of adjacent rock samples. The research provides a scientific basis for the failure mode and failure prediction of layered sandstone.}
}