Okay, I’ll ramble for a bit. My thesis leans toward the physics side, although the case study has implications for biology. As we know, nuclear energy has been available for a few decades now; however, there is still no real consensus on where to dispose of nuclear waste. One of the most popular ideas is to dispose of it in deep geological formations, considering it a safe and effective long-term solution for managing high-level and long-lived radioactive waste.

One of the alternatives being studied as a potential host for future repositories is clayey materials, since they possess favorable properties like low hydraulic conductivity, small molecular diffusion, and a significant retention capacity against radionuclide migration. Hence, multiple underground research facilities have been constructed to study the behavior of these materials under real conditions, and one of them is the subject of my thesis. Among the findings, researchers have noted that damage is induced around the excavated galleries in the form of fracture networks, and this excavated damaged zone (EDZ) controls the near-field hydromechanical response of the host rock. These fractures also serve as preferential pathways for water flow, which can be interpreted as an overall increase in permeability around the tunnel, and consequently, they can act as pathways for radionuclide migration.

Sooo, what am I actually investigating? Well, first of all, for my case study, there have been multiple in-field studies that measured pore water pressure before, during, and after excavations, as well as wall convergence. Another critical factor measured on-site is the extent of the induced damaged zone. Previously, work has been done to numerically reproduce the hydromechanical behavior of this case study. In terms of deformations, everything is A-OK: wall convergence, the extension of the damaged zone, shear bands (it all aligns). The tricky part is the hydraulic side.

Some of the pore water pressure sensors were installed within the boundaries of the damaged zone, and for those, the numerical simulations show close agreement with the measured behavior. However, other sensors were installed outside the damaged zone, and the numerical models underestimate the drop in pore water pressure after the excavation front has passed the section. A particular quirk of this case study is that the rock exhibits anisotropy; the damaged zone extends very little in one direction, whereas it extends significantly in the other. This means some sensors are at the exact same distance from the tunnel, but one might be inside the damaged zone while the other is not. Crucially, the previous simulations were carried out in 2D, while the problem is inherently 3D, so some discrepancies are to be expected, although not as large as the ones observed.

Now, the original simulations established that permeability was a function of a variable that maps permanent deformation due to shearing; in other words, it’s a variable that is only non-zero within the damaged zone. So, the hypothesis is that the change in permeability is not confined to the damaged zone, but rather extends further out. To test how this hypothesis works, a fellow grad student under the same advisor (he’s a PhD candidate while I’m a Master’s student) ran some simulations using a variable that tracks elastic deformation, which does extend outside the damaged zone. He got a closer approximation to the measured behavior, and his work is still in progress.

What I’m doing is approaching the problem from a different perspective. Previous works were all transient and hydromechanically coupled. I, on the other hand, am strictly interested in the long-term picture, since long-term changes in permeability are our primary concern (they act as a proxy for how likely radionuclide migration is based on the extension of the change in the permeability field).

So, I took the now near-steady readings from the pore water pressure sensors and constructed my own finite element model to represent the analyzed section. Then, I selected some optimization algorithms and objective functions to measure the difference between the simulations and the observed data. I established multiple permeability laws that are functions of different state parameters (stress, deformations, confined to the damaged zone, extending outside of it), carried out steady-state/long-term water flow analyses, and evaluated the objective function using the pressure values at the nodes closest to the sensor locations to measure the discrepancy with the observed data. The optimization algorithm automates the process of adjusting the variables that scale the impact of these state variables on the permeability change.

For the state variables, I’m using the fields resulting from the original simulations as well as the ones generated by my ‘academic brother.’ I know those fields depend on the change in permeability and don’t strictly correspond to the laws I’m testing. Nonetheless, this analysis helps point out that there indeed needs to be a change in the permeability field outside the damaged zone. So far, I’ve seen that permeability laws producing a change outside the damaged zone, and using the state variable field from the simulation that already assumed an external change, show a much closer agreement with the measured data. Ideally, at the end, I would perform a fully coupled hydromechanical simulation to test the best-performing permeability law and see how it behaves, but that might be reserved for a future paper.