Intelligent Prediction and Adaptive Control of Cutter Wear for Pipe Jacking Machines in Complex Geological Conditions

Pipe jacking is widely used in urban underground pipe networks and cross-region pipeline projects. Conventional inspection and replacement practices based on fixed distances or experience cannot track changing ground conditions and tool status in real time, particularly under complex ground conditions. Cutting tools can experience abnormal wear or fail, resulting in unplanned stoppages, reduced tunnelling efficiency, and increased project risk. This study proposes and validates a framework for intelligent management of cutting-tool wear in pipe jacking machines by combining multi-source monitoring, machine learning, and adaptive decision-making. The proposed system adopts an offline retrospective validation mode, in which prediction performance was evaluated on historical project data and control strategies were verified in a simulation environment. First, a data acquisition system was established to integrate machine operating parameters, geotechnical and mechanical properties, and slurry-system information. An attention-based long short-term memory (LSTM) model was developed to learn from these data and achieve accurate time-series prediction of tool wear. Then, an adaptive control strategy was designed to dynamically optimize tunneling parameters and to plan the tool-replacement window effectively. The proposed framework was validated on a natural gas pipeline pipe-jacking project (DN2200, jacking length 268 m) containing fully weathered, strongly weathered, and moderately weathered granite strata. The results show that the proposed system outperforms traditional methods in tool-wear prediction (RMSE reduced to 1.12 mm and R² increased to 0.915), tunneling efficiency (increased by about 14.7%), and risk management. This work presents a model and a practical technical approach for more intelligent management of pipe-jacking construction.

Hali tarjima qilinmagan