Enhancing TBM operations under complex geological conditions using data-driven methods

• Main Author: Fu, Xianlei
• Other Authors: Tiong Lee Kong, Robert
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Civil engineering::Construction technology
• Online Access: https://hdl.handle.net/10356/171997
• Institution: Nanyang Technological University

With the rapid pace of urbanization and increasing demand for underground space, tunnel construction gains popularity for its contribution to the subway transportation system. Tunnel Boring Machines (TBM) have been extensively applied in tunnel construction due to its efficiency, safety, and less impact on the ambient environment. During tunnel construction, the TBM monitoring system generates a huge volume of real-time data, which could be beneficial for improving the TBM performance during tunnel construction. However, due to complex geological conditions and knowledge gaps in TBM-soil interaction mechanism, there is a lack of an effective approach to extract useful knowledge from the data. As a result, the industry still heavily relies on human experience for TBM operations, which may impose risk on safety and compromise the performance of TBM. Therefore, the motivation of this Thesis is to establish a reliable approach that can enhance the overall TBM operation during excavation using data-driven methods. To achieve the overall research objectives, 4 major steps are proposed which starts with quantitative analysis of soil-TBM interaction mechanism, followed by geological detection, TBM performance estimation and lastly the TBM operating parameters optimization with digital twin which the details are elaborated in research objectives 1 to 4, respectively. The key findings are summarized as follows: (1) Through the proposed clustering method, TBM’s key operating parameters behave differently under various geological conditions. (2) The proposed ensemble MobileNets (EMNet) method can identify the soil condition using the photos of the excavated mucks with high accuracy. (3) By considering the spatial, temporal, and causal relationships among the TBM operating parameters, the proposed causal temporal graph convolution network (CT-GCN) method predicts TBM’s parameter with high accuracy. And (4) The established digital twin model enables physical-cyber interaction through the Internet of Things (IoT) sensors, which achieves real-time monitoring of overall TBM working status and enhances TBM’s overall performances through the proposed online-optimization approach.