FraNetG: Fracture Network Growth       

PI: Thomas Driesner (ETH Zurich), Rolf Krause (Università della Svizzera italiana)

Co-PIs: Patrick Zulian 

July 1, 2021 – June 30, 2024

Project Summary

The project aims at fast, large, and realistic simulations of growing fractures that form complex networks, which is of broad interest in engineering, material sciences and geosciences. In terms of HPC this translates to portable exascale non-smooth methods for non-convex problems. Following our PASC experience in the ongoing FASTER (Forecasting and Assessing Seismicity and Thermal Evolution in geothermal Reservoirs) project we intend to synergistically and successfully combine the science case experience (ETHZ), and the competence in algorithmic development (USI) to deliver real-world relevant HPC applications.   

Simulation of growing large fracture networks has so far been limited to proof-of-concept studies or rather small systems, limited by costly methods such as adaptive re-meshing etc., poor scaling of numerical methods, and often poor convergence. However, real-world applications often require simulating large fracture networks with two main application cases: (a) the growth of large fracture networks is a scientific problem on its own right with non-linear feedback between stress distribution and its dissipation by the growth process, potentially leading to self-organization and emergent behavior that is of yet poorly understood; and (b) the need of "geomechanically plausible/realistic" fracture networks that serve as input to simulations of seismicity, engineering of underground heat exchangers for geothermal energy exploitation and many other applications. For the latter, current approaches mostly use stochastically generated fracture networks that do not - or only partially - honor the laws of mechanics. This causes unphysical responses to stresses etc. and leads to inaccurate to explicitly wrong simulations results, which is unacceptable for safety-critical assessments such as the prediction of induced seismicity, material stability etc. We illustrate the application relevance of our proposed software development at the example of reconstruction of geologic fracture growth histories and geothermal energy technologies and point out future software extension directions for application, e.g., in ore deposits research on natural copper resources needed for greening the economy.  

For the FraNetG project we intend to employ recent advances in simulation methodology and algorithmic approaches to develop a HPC platform that allows growing physically correct, large virtual fracture networks under diverse initial and boundary conditions as the core of diverse scientific or engineering applications.