Geomechanical modeling for thermal spallation drilling

Wells for Engineered Geothermal Systems (EGS) typically occur in conditions presenting significant challenges for conventional rotary and percussive drilling technologies: granitic rocks that reduce drilling speeds and cause substantial equipment wear. Thermal spallation drilling, in which rock is fragmented by high temperature rather than mechanical means, offers a potential solution to these problems. However, much of the knowledge surrounding this drilling technique is empirical - based on laboratory experiments that may or may not represent field conditions. There is a lack of understanding coupled with large uncertainties in the phenomenology of the process which could be resolved with computer modeling. System-scale simulations of thermal spallation employ phenomenological models of rock damage due to the thermal gradient and erosion of spalls by the fluid. This allows rapid evaluation of mass and energy balances near the drill head and parametric studies of the relationship between drilling performance and the parameters of the jet. However, large-scale modeling is not suitable for resolving rock grains and inhomogeneities, and the damage model must account for the temperature gradient - not typically used as an input parameter in such material models. Thus, the parameters governing such phenomenological damage models require calibration by mesoscale simulations that fully resolve rock grains. This paper outlines a new numerical modeling effort investigating the grain-scale processes governing thermal spallation drilling. Several factors affect spall production at the mesoscale, including grain size and size distribution, surface temperatures and material heterogeneity. To investigate the relative influence of these factors, we have conducted a series of simulations using GEODYN - a parallel Eulerian solid and fluid dynamics code. In this paper, we describe a two-dimensional model used to simulate the grain-scale processes and present preliminary results from this modeling effort.