Quantitative Evaluation of Tunnel Rock Mass Integrity Based on MWD Technology

In tunnel construction, the quantitative evaluation of rock mass integrity heavily relies on information from the exposed face, and there are challenges when drilling data is used for integrity evaluation. To this end, this study introduced a novel method for quantitative evaluation of rock mass integrity during drilling, integrating numerical statistics with machine learning. A substantial dataset of digital drilling data was collected, covering three common types of rock mass integrity: relatively intact, relatively fractured, and fractured. Subsequently, a high-performance random forest model for the classification of rock mass integrity was developed through data preprocessing and hyperparameter optimization. The interpretability of the model’s predictive results was enhanced using Shapley additive explanations (SHAP) value theory. Additionally, the instability index and its exponents from the multivariable instability index analysis method were selected and quantified, and the quantitative evaluation model for “interval rock fracture index (IIRFI)” was created. The application of the model in actual tunnel engineering demonstrates an approximate 90% accuracy rate in evaluating the three types of rock mass integrity. The model provides more efficient, accurate, and detailed information on rock mass integrity compared to conventional methods. The study offers a new and effective quantitative approach for assessing rock mass integrity in tunnels, which contributes to improved construction safety and project efficiency.