Evaluation of Machine Learning Algorithms in Tunnel Boring Machine Applications: A Case Study in Mashhad Metro Line 3

Abstract

Abstract This research explores the prediction of Tunnel Boring Machine (TBM) performance in the excavation of Mashhad Metro Line 3 using machine learning techniques. The study leverages a robust dataset comprising 113 features recorded over 305 working days. Multiple Linear Regression, Decision Trees, and Multi-Layer Perceptron Neural Network models are employed to analyze TBM performance, with a specific focus on the penetration rate. The results reveal comparable performance among the models, indicative of a potentially linear relationship between selected features and the penetration rate. Feature importance analyses provide valuable insights into key parameters, contributing to a better understanding of the excavation process. The discussion addresses the interpretability of the Multiple Linear Regression model and potential overfitting concerns, emphasizing the impact of dataset quality on model consistency. The study contributes to the advancement of accurate predictions in TBM performance during tunneling projects, with a particular application to Mashhad Metro Line 3. The findings and methodologies presented in this research offer insights into the field of tunnel construction and excavation.