ABSTRACT
Lithology identification is crucial in tunnelling with a tunnel boring machine (TBM) for mitigating the instability risks and for optimizing the time-consuming excavation. However, in most cases, the lithology classes are not adequate due to the low number of drilling holes, anisotropy, and heterogeneity of the rock layers. In this paper, the proposed method is to classify the lithology of the tunnel surrounding layers using automatic machine learning. An earth pressure balance (EPB) machine recorded 18 operational parameters in real time along with the three main lithologies. Python Pycaret was applied to create 14 different classifiers for comparison automatically. According to the Pycaret results, Light Gradient Boosting Machine (LightGBM) model was selected as the best classifier. Due to the imbalanced distribution of tunnel layers, LightGBM performed with five different resampling models to obtain the resampling model for the imbalanced distribution of the classes.
