by Mike London, University of Manitoba and Alberta Research Council
The Alberta Research Council is Canada’s oldest provincial research organization. It was founded in 1921 to study Alberta’s resources, including the large reserves of heavy oil in the Athabasca tar sands. The extraction and recovery of the oil from the sand presents several technical challenges, largely owing to the high viscosity of the oil. We are exploring a wide range of recovery methods and conduct extensive laboratory experiments, CT and NMR imaging, theoretical studies and numerical simulations to support this work.
We have made use of IRIS Explorer as a visualization tool and application builder in the development and analysis of pore-scale numerical simulations of multi-phase flow, oil reservoir simulations, CT imaging and tar sand extraction experiments, some examples of which are shown in the image on this page. It should perhaps be noted that the routine use of interactive visualization in reservoir simulation is a fairly recent phenomenon; traditionally, graphics were confined to the offline generation of material for presentation. Commercial reservoir simulation products now incorporate a standard complement of visualization functionality that is suitable for production use. Research, however, cannot present such a finite list of demands. For instance, heavy oil recovery often requires a more detailed look at mechanisms through the display of new variables or new views of old variables. In this context, we have found that IRIS Explorer is most useful, not as a replacement for canned visualization packages (which themselves may, of course, have been implemented using IRIS Explorer) but as a complement to them. The great advantage of IRIS Explorer lies in the leverage gained by incorporating standard modules into specialized applications. Indeed, a great deal can be accomplished just by writing a suitable data reader module.
To illustrate our use of IRIS Explorer, we describe one of the applications in more detail. Our pore-scale simulations are directed at exploring the complicated behaviour of the flow of fluid through the irregular geometries representative of the pores and channels found in the unconsolidated sands in most heavy oil deposits. We model the pore space as a flow network, paying particular attention to the pore-scale geometries and topologies. Because of their irregular structure, visualization is an essential part of this work, not only for viewing the final results, but also for debugging the development and monitoring the various stages of modeling. We have used IRIS Explorer as the framework for all of these stages.
We begin by constructing a model of the pore space by creating an artificial packing of spherical particles of various radii. A simple algorithm, which incorporates the idea of particle rearrangement in the presence of gravity, is encoded in an IRIS Explorer module. The final packing is provided as an IRIS Explorer lattice (a generalized array with data and coordinates) which can easily be rendered as geometry using the Ball module. In addition, the intermediate configurations can be usefully visualized in order to check on the packing algorithm, and these results can be assembled into an animation of the complete process.
Following the creation of the packing, the pore space is divided into nodes, where phases mix and flow without hindrance, and channels, where Stokes flow dominates and phases are separated by distinct interfaces. The resulting network is encoded into an IRIS Explorer pyramid. This data structure, used frequently for finite element grids, consists of layers of lattices. Links between layers associate related elements together. Hence, the base layer contains the original particle lattice, while the next layer contains nodes with links to enclosing particles, and the next layer contains channels, each with links to two nodes. A custom module converts the pyramid to a geometry for viewing.
Once the visual check of the network is satisfactory, another module converts the pyramid to ASCII data required by the simulator. Thus, except for some fluid property descriptions and boundary conditions, the entire setup is carried out graphically. The actual flow simulation is run offline, since it frequently takes days or weeks to complete.
The simulation output is encoded in a custom binary format, for which a special reader is constructed. The simulator models single-phase slugs of fluid moving through the channels, with interfaces between them. This is accommodated in IRIS Explorer by adding a slug layer to the pyramid, with links to the appropriate channels. This means that the same module can be used to display both the input and output data. A variety of views are available on output, including colormapped pressures and flow rates. The final stage of the analysis is the calculation of macroscopic properties such as the permeability tensor. Again, this part of the simulation is carried out entirely within the IRIS Explorer map.
We have found that the flexibility of IRIS Explorer cannot be matched by any application-specific visualization package. It has been easy to use it to produce VRML and other common graphics files. Moreover, the leverage gained by incorporating standard modules allows a single user to harness state-of-the-art visualization techniques without the services of a development team.
We thank the Alberta Research Council and the University of Manitoba for the support of this work. For more information, please see http://flow.arc.ab.ca/london/vde2000/