In this study we have modelled a seismic velocity survey which took place in ONKALO underground research facility at Olkiluoto, Finland as part of a study for mapping the fractures in the Excavation Damage Zone (EDZ). The purpose of this work was to test the performance of the models used previously to model laboratory experiments on a real-world, complex engineering problem.
The fractures of the studied area were mapped and were implemented into the models. We aim for simple geometries in the modelling and since the geometry of the fractures was complex, we tested five different approximate approaches to represent the fracturing in the models. For explicit fracture representation, we create two subcategories called the Pixelised Fractured Model (PFM) and the Equivalent Discrete Fracture Medium (EDFM). A Localised effective medium (LEM) model was also used and again there were two versions, the first using a thick LEM layer and the second using very fine LEM layers. The fifth approach used an Effective medium (EM) for the entire fractured region.
A wave source was derived for use in the models by creating an inversion which used waveforms from the deepest propagation paths only where there was little or no presence of fractures. The initial stiffness used in the fractures was based on the size of the fracture according to scaling laws. We then selected specific waveforms for comparison of the model and data based on the raypath relative to specific fractures.
Waveforms from the EM model and the EDFM model were not comparable with the data. The EM model gave a much higher amplitude and significantly late arrival (~0.1 ms) while the EDFM model was more far too attenuated with late arrivals. The PFM model and the two LEM models produced more comparable results and further work was applied to these models to optimise the fracture stiffness to achieve better correspondence with the data for specific ray-paths (see figure below). The work indicates that creating such fracture models and then applying an optimisation may yield better interpretations of fracture stiffness.
The waveforms are the result of the manual iterative optimisation for fracture stiffness or LEM fine model against the experiment data and the initial model results.