Options
Einsatz von neuronalen Netzen zur Vorhersage des Materialvolumens von Injektionsbaustellen im Tunnelbau
Publikationstyp
Journal Article
Publikationsdatum
2021-10
Sprache
German
Author
Institut
Enthalten in
Volume
98
Issue
10
Start Page
767-774
Citation
Bautechnik 98 (10): 767-774 (2021-10)
Publisher DOI
Scopus ID
Use of neural networks for the prediction of the material volume of injection sites in tunnel construction. In this paper, a method is presented that uses digital documentation on injection construction sites to calculate automated, construction-accompanying predictions of the injection quantities still to be expected. In the construction project under investigation, waterproofing injections are being made as part of the Stuttgart 21 project for a 3.2 km long, twin-tube railroad tunnel. The presented method uses a particular form of a neural network, the Feed Forward Network. The network is trained with the injection quantities per tunnelmeter of one tunnel tube to predict the other's injection quantities. After a brief introduction to the operation of neural networks, it is shown that the presented method can forecast the total injection quantities with an accuracy > 5 %. Renesco GmbH has collected the data in cooperation with eguana GmbH.
Schlagworte
Building Management
construction management
construction time forecast
Geotechnical engineering
grouting
neural networks
tunneling
Tunnelling