This paper presents a methodology to link the ground surface displacements to a tunnel boring machine (TBM) operation parameters. The approach consists at first in the exploration and description of the TBM operation parameters, recorded during the construction of the Toulouse (France) subway line B tunnel, using principal component analysis and agglomerative hierarchical clustering (AHC). Then, partial least squares (PLS) regression models were developed for the prediction of the displacements for the two groups of observations, identified by the AHC (pure settlements and heave at the center), as a function of the ten selected TBM operation parameters and the geometry parameter (X 2/H 2). The developed models take into account the nonlinear relationship between displacements and the operation parameters and the collinearity that exists in the parameters. In addition, the PLS regression allows identifying the parameters that affect the displacement of each group. The comparison of the predicted displacements with the measurements for the validation samples of the two groups shows good agreement.
Read full abstract