Scholarlink

<p>As tunnel engineering advances toward greater depths, longer distances, and increasingly complex geological conditions, the efficient and reliable maintenance of tunnel boring machine (TBM) cutterheads has become a critical factor affecting construction safety and cost. To address uncertainties in rock mass parameters and enhance the scientific basis of cutterhead maintenance decisions, this study proposes a novel cutter wear prediction method. First, a disc cutter wear prediction model is developed based on the sliding distance generated during the rock-breaking process, incorporating key geotechnical parameters and cutter movement trajectory characteristics. The model fully accounts for uncertainties in rock mass properties and uses Monte Carlo simulations to quantify the spatial variability of rock parameters, thereby characterizing the evolving failure probability of cutters as excavation progresses. Second, to reduce model uncertainty, a Bayesian updating approach is introduced to dynamically correct model bias using field monitoring data. As observational data accumulate, the posterior standard deviation decreases significantly, resulting in progressively improved prediction reliability. Finally, a novel machine learning-based rock mass classification model is proposed, using TBM operational parameters to estimate the fundamental inputs for the Monte Carlo simulations. Through Bayesian optimization of a multilayer perceptron (MLP) and an XGBoost classifier, and by employing a weighted ensemble strategy to integrate the classification results, the final model achieves an accuracy of 96.31% on the test set. It also outperforms the individual models in terms of precision, recall, and F1-score, demonstrating strong classification performance and generalization capability. Overall, the proposed framework enables more accurate cutter wear prediction and risk assessment under conditions of incomplete rock mass information and uncertain parameters. It provides a scientific basis for cutterhead maintenance strategies in TBM construction and holds significant practical value for improving construction safety and reducing cutter replacement costs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bayesian Calibration of TBM Cutter Wear Under Geological Uncertainty

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • Yaoqi Nie,
  • Qian Zhang,
  • Lijie Du,
  • Yanliang Du

摘要

As tunnel engineering advances toward greater depths, longer distances, and increasingly complex geological conditions, the efficient and reliable maintenance of tunnel boring machine (TBM) cutterheads has become a critical factor affecting construction safety and cost. To address uncertainties in rock mass parameters and enhance the scientific basis of cutterhead maintenance decisions, this study proposes a novel cutter wear prediction method. First, a disc cutter wear prediction model is developed based on the sliding distance generated during the rock-breaking process, incorporating key geotechnical parameters and cutter movement trajectory characteristics. The model fully accounts for uncertainties in rock mass properties and uses Monte Carlo simulations to quantify the spatial variability of rock parameters, thereby characterizing the evolving failure probability of cutters as excavation progresses. Second, to reduce model uncertainty, a Bayesian updating approach is introduced to dynamically correct model bias using field monitoring data. As observational data accumulate, the posterior standard deviation decreases significantly, resulting in progressively improved prediction reliability. Finally, a novel machine learning-based rock mass classification model is proposed, using TBM operational parameters to estimate the fundamental inputs for the Monte Carlo simulations. Through Bayesian optimization of a multilayer perceptron (MLP) and an XGBoost classifier, and by employing a weighted ensemble strategy to integrate the classification results, the final model achieves an accuracy of 96.31% on the test set. It also outperforms the individual models in terms of precision, recall, and F1-score, demonstrating strong classification performance and generalization capability. Overall, the proposed framework enables more accurate cutter wear prediction and risk assessment under conditions of incomplete rock mass information and uncertain parameters. It provides a scientific basis for cutterhead maintenance strategies in TBM construction and holds significant practical value for improving construction safety and reducing cutter replacement costs.