As-Encountered Prediction of Tunnel Boring Machine Performance Parameters using Recurrent Neural Networks. Issue 10 (October 2020)

• Record Type: Journal Article
• Authors: Nagrecha, Kabir; Fisher, Luis; Mooney, Michael; Rodriguez-Nikl, Tonatiuh; Mazari, Mehran; Pourhomayoun, Mohammad
• Is Part Of: Transportation research record. Volume 2674:Issue 10(2020)
• Page Start: 241
• Page End: 249
• Publication Date: 2020-10
• Subjects: Transportation -- Periodicals; Roads; Transport -- Périodiques; Routes -- Périodiques; Routes -- Conception et construction -- Périodiques; Roads; Transportation; 388.05
• DOI: 10.1177/0361198120934796
• Languages: English
• ISSNs: 0361-1981

Abstract

The earth pressure balance tunnel boring machine (TBM) is advanced excavation machinery used to efficiently drill through subsurface ground layers while placing precast concrete tunnel segments. They have become prevalent in tunneling projects because of their adaptability, speed, and safety. Optimal usage of these machines requires information and data about the soil of the worksite that the TBM is drilling through. This paper proposes the utilization of artificial intelligence and machine learning, particularly recurrent neural networks, to predict the operational parameters of the TBM. The proposed model utilizes only performance data from excavation segments before the location of the machine as well as its current operating parameters to predict the as-encountered parameters. The proposed method is evaluated on a dataset collected during a tunneling project in North America. The results demonstrate that the model is effective in predicting operation parameters. To address the potential issue of gathering sufficient data to retrain the model, the possibility of transferring the trained model from one tunnel to another is tested. The results suggest that the model is capable of performing accurately with minimal or even no re-training.