岩体结构破裂是严重制约矿山、地铁、隧道等地下空间工程建设及其安全运行的重要因素。实现对岩体结构破裂状态的识别是当下研究的热点与重点之一。为此，开展了不同条件的砂岩加载破坏实验，提取了加载全程的声发射梅尔倒谱系数及其波动差，研究了系数及其波动差在砂岩受载破坏全程的变化规律，分析了1号系数(一组声发射梅尔倒谱系数包括12个，1号系数指第1个声发射梅尔倒谱系数)及其波动差与砂岩破裂状态的相关性特征，基于此提出了砂岩破裂状态声发射梅尔倒谱系数判识方法，构建了判识准则并进行判识效果检验。结果表明：随载荷增加，1号系数整体上增大，系数及其离散性在破坏阶段显著增大并表现出显著的规律波动性特征；1号系数波动差具有阶段性变化特征，波动差的大小及其起伏变化可表征砂岩的破裂，波动差整体增大及突增的变化可反映砂岩非稳定变形和峰后破坏阶段的宏观破裂，波动差的突增幅度可反映砂岩破裂程度；声发射梅尔倒谱系数及其波动差对砂岩破裂表现出良好的响应特征，该特征受不同加载条件的影响较小，说明声发射梅尔倒谱系数在反映砂岩破裂上具有适用性；1号系数及其波动差与砂岩破裂状态具有较好相关性，该相关性可分为3个阶段，即1号系数及其波动差在砂岩微破裂阶段分布集中，在临近失稳破坏阶段分布范围急剧增大、整体值升高且出现高异常值，在峰后破坏阶段分布范围进一步增大、整体值更高、高异常值更多；利用1号系数的75%位点值和异常值、1号系数波动差的75%位点值和异常值构建了砂岩破裂状态判识准则，采用三分类模型混淆矩阵对判识准则的效果进行了检验，判识准确度和精准度分别为90.43%、94.45%。该成果可为其他种类煤岩的破裂状态识别提供借鉴，为煤岩失稳监测预警提供参考。

Rock mass structure rupture is an important factor that seriously restricts the construction and safe operation of underground space engineering projects such as mines, subways and tunnels. Realizing the identification of rock mass fracture state is one of the hotspots and emphases of current research. In this study, some experiments of sandstone loading failure under different conditions were carried out, the acoustic emission Mel-frequency cepstral coefficient (MFCC) and its fluctuation difference during the whole loading process were extracted, the variation law of the coefficient and its fluctuation difference during the whole loading failure process was studied, and the correlation characteristics of coefficient No. 1 (According to the calculation of acoustic emission Mel-frequency cepstral coefficients, it can be seen that a set of acoustic emission Mel-frequency cepstral coefficient includes 12, and coefficient No. 1 refers to the first Mel-frequency cepstral coefficient) and its fluctuation difference with the fracture state of sandstone were analyzed. Based on this, a method to identify sandstone fracture state using the Mel-frequency cepstral of acoustic emission was proposed, and the identification criteria was constructed. The identification effect was finally verified. The results show that with loading increase, the coefficient No. 1 increases as a whole, and the coefficient value and its discreteness increase significantly in the failure stage and show significant regular fluctuations. The fluctuation difference of the coefficient has the characteristics of periodic variation. The size of the fluctuation difference and its fluctuation can characterize the fracture of sandstone. The overall increase and sudden increase of the fluctuation difference can reflect the macroscopic fracture of sandstone in the unstable deformation and post-peak failure stage, and the sudden increase level of the fluctuation difference can reflect the fracture degree of sandstone. The acoustic emission Mel-frequency cepstral coefficient and its fluctuation difference show good response characteristics to sandstone fracture, which is less affected by different loading conditions, thus they have applicability in reflecting sandstone fracture. The coefficient No.1 and its fluctuation difference have a good correlation with the fracture state of sandstone. The correlation can be divided into three stages as: in the micro-fracture stage of sandstone, the coefficient No.1 and its fluctuation difference are intensively distributed; in the unstable deformation stage just prior to the peak load, the distribution range increases sharply, the overall value increases and the high abnormal value appears; in the post-peak failure stage, the distribution range further increases, the overall value is higher, and more high abnormal values appear. The identification method and criteria of sandstone fracture state were constructed by using the 75% site value and outliers of coefficient No. 1 and the 75% site value and outliers of the fluctuation difference of coefficient No. 1. The effect of the identification criteria was tested by the confusion matrix of the three-classification model. The accuracy and precision of identification are 90.43% and 94.45%, respectively, which indicate the identification effect is good. The results can provide a reference for the identification of the fracture state of other types of coal and rocks, and for the monitoring and early warning of coal rock instability.