Prediction of TBM Tunneling Parameters through an LSTM Neural Network

Abstract

With the wide application of tunnel boring machines (TBMs) in tunnel construction, the adaptive tuning of the TBM tunneling parameters has become a research focus. Nowadays, since complicated geological conditions are still challenging to predict, the fine-tuning of tunneling parameters mainly relies on operational experience. Artificial intelligence provides a convenient solution for predicting tunneling parameters through data mining and machine learning. Based on in-situ data, this work proposed a novel method for segmenting the original data in tunneling cycles and analyzing the parameter correlation for data size reduction. Subsequently, a model was established based on an LSTM to predict tunneling parameters in the steady phase based on the data in the rising phase. The results demonstrated that the model is capable of predicting torque and thrust accurately. This makes it possible to adjust the TBM tunneling parameters according to current geological conditions in real time. The present study is of great significance for the tunneling efficiency and construction safety in the actual TBM construction, since it can improve its scientific and intelligent level.