Decision tree analysis of cutter selection for tunnel boring machines: A study of geological conditions and machine types in high-performing TBM projects

Abstract

The selection of appropriate cutter types and configurations is crucial for effective Tunnel Boring Machine (TBM) operations. This study introduces a novel decision tree methodology to develop a model for selecting primary cutter types (PCT) and cutter configurations based on actual data from 112 high-performing TBM projects. Key parameters, including geological conditions (Soft Soil, Hard Soil, Coarse Soil, Bouldery Ground, Soft Mixed, Hard Mixed), machine type, diameter, and cutter type, were analyzed. Various decision tree algorithms, including C4.5, CART, SVM, Random Forest, and ensemble methods such as Bagging, AdaBoost, XGBoost, and LightGBM, were applied and evaluated using performance metrics (accuracy, precision, recall, F1-score, and ROC-AUC). Results indicate that CART and KNN algorithms are the best performers, with accuracies of 89.3 % and 88.4 %, respectively, while AdaBoost was the least effective. Decision rules from the CART model reveal geological conditions as the most significant predictor of PCT, followed by machine type and diameter. This study provides a systematic framework for cutter configuration selection in mechanized tunneling, offering practical guidelines for the industry.