Full Waveform Inversion for Advance Exploration of Ground Properties in Mechanized Tunneling

Abstract

We present a new approach for exploration of small- and large-scale geological structures in mechanized tunneling environments by analyzing actively excited seismic waves illuminating the soil in front of the tunnel-boring machine. Applying the concept of full waveform inversion, we exploit all information contained in the seismic recordings to image unknown disturbances of soil properties along the tunnel track. Owing to the lack of field data, we explore two-dimensional synthetic test models containing exemplary disturbances of the soil, where real recordings are replaced by synthetic ones obtained by solving the (an)elastic wave equation with a Nodal Discontinuous Galerkin method. An extension to three dimensions is straight forward but computationally expensive. Disturbances in these test models are reconstructed with an iterative procedure guided by minimization of the misfit between measured and predicted records. Our method is validated by means of a simple test case and different minimization procedures including a quasi-Newton method are compared. The presented method is able to reconstruct the spatial distribution of seismic wave velocities (P and S waves) and can detect large and small-scale objects ahead of the tunnel face characterized by wave velocities differing from that of the surrounding bedrock. The imaging process is stabilized by a multiscale approach, where the frequency content of the recordings is successively extended with the number of iterations. Incorporating viscoelastic attenuation of the surrounding bedrock does not compromise the capability of detecting large-scale objects as long as low-frequency waves are used.