Abstract
HIGHLIGHTS: What are the main findings? Sub-zero temperatures (−20 °C) intensify ice–rock coupling, causing severe high-frequency volatility and non-linear baseline shifts in Measurement While Drilling (MWD) sensor signals, specifically increasing torque and feed pressure while decreasing rotational speed. A proposed dual-mechanism change-point detection algorithm, integrated with Z-score normalization, successfully filters out temperature-induced “pseudo-interfaces,” achieving a rock layer interface prediction error of less than 1.5 mm. What are the implications of the main findings? The study provides a robust signal-processing framework that effectively compensates for extreme-temperature data drift, significantly enhancing the anti-noise capability and reliability of MWD monitoring in cold-region geotechnical engineering. By delivering highly accurate, real-time stratigraphic profiles, this method establishes a crucial technical foundation for optimizing differentiated explosive charging, thereby reducing hazardous blasting effects and promoting energy-efficient, green mining operations. ABSTRACT: To address the critical challenges of lithology acquisition and low blasting refinement under extreme low temperatures and varying thermal conditions in high-altitude environments, this study develops a real-time dynamic identification method for rock-like interfaces using Measurement While Drilling (MWD) technology. The scope of this research involves the use of a self-developed indoor digital drilling experimental platform to simulate both ambient and freezing (−20 °C) conditions. Procedures included conducting comprehensive comparative drilling experiments on various rock-like materials with distinct strength levels to evaluate their mechanical responses during penetration. The major findings reveal a significant influence of low-temperature hardening effects on MWD parameters; specifically, the frozen state notably increases drilling torque and feed pressure while simultaneously decreasing the stable rotational speed of the drill bit. To resolve the feature parameter drift induced by temperature variations, a novel interface recognition algorithm is proposed that integrates Z-score normalization, change-point detection, and multi-dimensional spatial clustering. Through a dual-detection mechanism involving both single-point and cumulative features, the algorithm effectively captures precise mutation information during rock layer transitions. It further incorporates multi-dimensional indicators, such as consistency, change intensity, and point density, to perform comprehensive weighted scoring. Experimental results demonstrate that the proposed algorithm effectively eliminates the systematic offset of parameters caused by temperature fluctuations. The prediction error at both “strong-weak” and “weak-strong” transition interfaces is maintained within 1.5 mm, which significantly improves the accuracy and robustness of interface recognition under complex and varying working conditions. These key conclusions provide essential technical support for the implementation of differentiated charging and green refined mining operations, ensuring greater energy efficiency and environmental protection in cold-region engineering.