Identifying Rock Loads on TBM Shields During Standstills (Non-Advance-Periods)

Paul Johannes Unterlaß*, Georg H. Erharter, Thomas Marcher

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


Tunnel boring machine (TBM) operational data is mostly analysed with respect to data that was recorded during the advance of the TBM. Focusing on data that was recorded during standstills of a gripper TBM, we analyse rock loads that were passively recorded in the cylinders of a small roof support shield. These roof support cylinders are situated beneath the TBM’s shield – extending it against the rock mass during non-advance periods. Equipped with pressure sensors, they enable the unique opportunity of logging rock load variations throughout the tunnel. Hence due to the big amount of resulting data, techniques of unsupervised machine learning (i.e., cluster analysis) are used to automatically pre-process the TBM operational data. Furthermore, regression analysis is used to determine sections of the tunnel where rock loads are mainly occurring on the left or right side respectively. The data driven analysis shows that the main rock loads are occurring on the right side of the TBM which is in good accordance with observation from the construction site, as well as numerical models from literature. This paper contributes towards the understanding of rock load conditions in anisotropic rock masses recorded during the drive of a deep hard rock tunnel.
Original languageEnglish
JournalGeotechnical and Geological Engineering
Early online date21 Aug 2022
Publication statusE-pub ahead of print - 21 Aug 2022


  • Rock Load
  • Machine Learning
  • Anisotropy
  • Stress redistribution
  • TBM non-advance-data
  • Load on TBM shields
  • Rock load
  • Machine learning

ASJC Scopus subject areas

  • Geology
  • Geotechnical Engineering and Engineering Geology
  • Soil Science
  • Architecture


Dive into the research topics of 'Identifying Rock Loads on TBM Shields During Standstills (Non-Advance-Periods)'. Together they form a unique fingerprint.

Cite this