Geology prediction based on operation data of TBM: comparison between deep neural network and statistical learning methods

TL;DR

This paper proposes a deep neural network approach for predicting geological conditions ahead of tunnel boring machines using operation data, demonstrating improved accuracy over traditional statistical methods in a Chinese tunnel case study.

Contribution

It introduces a novel deep learning method for TBM geology prediction, outperforming existing statistical models and enhancing tunnel construction safety and efficiency.

Findings

Deep neural network achieves higher prediction accuracy.

The approach effectively estimates geology ahead of excavation.

Potential to complement geophysical prospecting methods.

Abstract

Tunnel boring machine (TBM) is a complex engineering system widely used for tunnel construction. In view of the complicated construction environments, it is necessary to predict geology conditions prior to excavation. In recent years, massive operation data of TBM has been recorded, and mining these data can provide important references and useful information for designers and operators of TBM. In this work, a geology prediction approach is proposed based on deep neural network and operation data. It can provide relatively accurate geology prediction results ahead of the tunnel face compared with the other prediction models based on statistical learning methods. The application case study on a tunnel in China shows that the proposed approach can accurately estimate the geological conditions prior to excavation, especially for the short range ahead of training data. This work can be regarded as a good complement to the geophysical prospecting approach during the construction of tunnels, and also highlights the applicability and potential of deep neural networks for other data mining tasks of TBMs.