Dynamic Prediction on Driving Attitude of Tunnel Boring Machine (TBM): An Automated Deep Learning Approach

This paper proposes an automated deep learning (AutoDL) framework for the dynamic driving attitude prediction of the tunnel boring machine (TBM) in tunnel construction. A standard process is proposed for the automated deep learning framework, including data preprocessing, optimal hyperparameter tuning, algorithm selection, and model performance evaluation, which can intelligently learn the time-varying monitoring data collected by the smart sensors installed on TBM. Moreover, a prediction interval is carefully designed to well consider the uncertainty, which is stable enough to raise the reliability of the prediction results. To validate the effectiveness and practicability of the proposed approach, it is applied in a TBM project for a natural gas pipeline network in Chongming Island–Changxing Island–Pudong New Area in Shanghai. Results indicate that the proposed automated deep learning with particle swarm optimization (PSO) algorithm greatly reduces the requirements of model training evaluation on the professional knowledge of developers, which can return not only high-accuracy predictions, but also interval-based predictions considering uncertainties. As for the engineering practice value, the deviation of TBM driving attitude in each segment ring can be dynamically and reliably estimated, which provides valuable evidence for managers to timely adjust the TBM operation to ensure the project quality.