为实现煤矿地质透明化，促进煤矿智能化发展进程，解决传统岩层界面识别方法速度慢、成本高以及精度差的难题，围绕岩层界面智能识别进行研究。首先，在实验室内浇筑地层模型，通过自主研制的随钻装置开展位移、转速、扭矩和声压级等参数的实时采集；其次，采用指数加权损失函数自动过滤位移数据异常值，提出钻速计算方法；然后，用变点检测算法、RStudio软件的Strucchange模型和决策树算法分析钻速、转速、声压级和扭矩等参数，对比分析其识别岩层界面的准确性；最后，在2种典型的地质条件下进行岩层界面随钻识别效果分析。结果表明：以钻速为输入参数，决策树算法是快速准确识别岩层界面的最佳方式；现场试验中煤岩界面位置预测的平均误差为0.04 m，但对于复合顶板岩层界面识别准确性相对较低。

This paper investigates measurement while drilling for rock strata interface to achieve geological transparency, promote intelligence development, and address the limitations of traditional rock detection methods in coal mines, including slow speed, high costs, and poor accuracy. Firstly, this study proposes a laboratory-level drilling device for real-time data acquisition of displacement, revolutions per minute, torque, and sound pressure level during drilling the formation model poured in the laboratory. Secondly, an exponentially weighted loss function was used for automatic screening of anomalies in displacement data and obtained a penetration rate that better reflect variations in drilling. Then, the accuracy of rock interface identification was analyzed using parameters such as penetration rate, revolution per minute, sound pressure level, and torque using the application of the change point detection algorithm, the strucchange model in RStudio software, and the decision tree algorithm. Finally, the performance of rock interface identification during drilling was evaluated under 2 typical geological conditions. Rresults indicate that the decision tree algorithm is the most effective method for quick and accurate identification of rock interfaces with penetration rate as the input parameter. On-site tests yielded 0.04m of average error in predicting the position of the coal rock interface. Yet, the method produces relatively low accuracy in rock interface identification of composite roof rock strata.