ABSTRACT
The shield tunnel boring machine (TBM) excavation progress in an urban area was evaluated by monitoring surface settlements near the tunneling face. This study implemented a stage-updated machine learning modeling method to predict the TBM operational parameters and corresponding the maximum surface settlements around the tunnel alignment before being excavated yet. Five TBM machine control parameters, i.e., advance speed, backfill grout injection volume, face pressure, thrust force, and cutter torque, were predicted by learning the preceding excavation records. Subsequently, the predicted machine parameters were utilized to predict the surface settlements in upcoming excavation zones. The settlement database was collected from a subway tunnel project in Hong Kong to establish and verify the developed model. The maximum surface settlement predicting model was validated at 5 locations, demonstrating a root mean squared error of 2.488 mm.
