Search & filter

Search for a publication

Search & filter

Year

Author

City

Country

Classification of surface settlement of stacked twin tunnels by machine-learning

Classification of surface settlement of stacked twin tunnels by machine-learning

Classification_of_Surface_Settle

H. Choi / D. Kim / S. Park / J.-Y. Oh / K. Pham

To ensure the safe operation of tunnel boring machines (TBMs) in urban areas, surface settlements should be monitored at several levels such as the alert, action, and alarm level, to prevent any damage to nearby infrastructures. In this study, the random forest (RF) machine learning model was implemented to predict the surface settlement levels using a database collected from a Hong Kong subway tunnel site. The predicted surface settlements were categorized into four settlement classes as following: (i) heaving, (ii) surface settlement below 5 mm, (iii) surface settlement between 5-10 mm and (iv) surface settlement more than 10 mm. The RF model successfully predicted and classified the surface settlements for the test data sets with 0.618 average accuracy. The performance of classification result was further evaluated using the F1 and AUC (area under the receiver operating characteristic curve) scores. In addition, the significance of each settlement influence factor was analysed with the feature importance value obtained from the optimal RF model.

More details

0,00 €

Year 2022
City Copenhagen
Country Denmark
ISBN 978-2-9701436-7-3