Excavation damage zone fracture modelling for seismic tomography : a comparison of explicit fractures and effective medium approaches

We model the full wavefield produced by a seismic velocity survey and optimise the representation of the fracture zone to best match field waveforms. The velocity survey was part of a mapping study on fractures in the Excavation Damage Zone (EDZ) of ONKALO underground research facility at Olkiluoto. The EDZ results from excavation of the rock mass, which modifies stress conditions changing the nature and behaviour of pre-existing fractures and generating new fracturing. These fractures act as the main transport pathways for contaminants both in and out of a geological disposal facility (GDF). Our goal is to test different representations of the fracture zone and to determine which models most successfully improve the interpretation of the fracture zone, producing estimates of a key unknown parameter, fracture stiffness, in addition to fracture sizes, fracture geometry, fracture density and crack density. We use modelling techniques previously tested in theoretical and laboratory studies and assess their performance on a real engineering problem. The paper introduces the field experiment and relevant information from the GDF in Finland. It describes the methodologies used for representing the fracture networks in the models — Explicit Fracture models with two approximations called Pixelised Fracture Model (PFM) and Equivalent Discrete Fracture Medium (EDFM), the Effective Medium (EM) model, and two versions of the Localised Effective Medium (LEM) model (LEM fine, LEM thick). These alternative representations were used within models of the field experiment and the calculated waveforms were used in an iterative inversion for fracture stiffness. Results show that the EM model and the EDFM model were unsuccessful in matching recorded waveforms. The fine LEM model and the explicit PFM model produced the best results especially after iterative optimisation of the fracture stiffness, giving confidence that further optimisation will lead to improved characterisation of the fracturing from the full waveform data