Advances in Magnetic Resonance Imaging
Climate change is a hot topic these days, with great attention focused on the need to reduce atmospheric carbon dioxide (CO2) concentrations to slow the process of global warming. Such reduction will rely, in part, on developing better quality—that is, lower emission—automotive fuels, and among the most promising approaches to improving fuel quality is the use of gas-to-liquid (GTL) refining technologies.
But making GTL competitive with conventional oil refining technology will require improved design trickle-bed reactors (TBRs). TBRs are essentially a column filled with catalyst particles, through which gas and liquid flow downward. Improved design and operation of TBRs could reduce CO2 emissions by 400 million metric tons a year, or about 2 percent of global CO2 emissions.
TBRs are currently designed by using imprecise empirical correlations, which leads to significantly oversized reactors and higher operating and capital costs, thereby putting TBR technology at a competitive disadvantage. The best way to improve TBR design is to enhance our understanding and modeling of the reactors through new phenomenological models and better closure laws for computational fluid dynamics simulations. Both of these require measurements of the velocity of the liquid and gas inside the reactor, as this will determine the overall performance of the reactor.
The only technique that can provide these measurements is magnetic resonance imaging (MRI), but the signal-to-noise ratio of conventional gas phase imaging is too low to achieve the required resolution. Collaboration between the Magnetic Resonance Research Centre at the University of Cambridge and Microsoft Research has produced a compressed sensing algorithm that enables gas phase velocity mapping at a resolution an order of magnitude greater than in the past. These measurements are of sufficiently high resolution to enable characterization of the interfacial velocity between the liquid and gas, which is critical to improved understanding of the reactor’s behavior and prospective design improvements.
In addition to its potential for helping to reduce atmospheric CO2 emissions, the compressed sensing algorithm could lead to greater use of MRI as a means for testing hypotheses about the process under study. A particular target is to reduce data acquisition times by an order of magnitude; this opens up new opportunities for studying chemical engineering processes, as well as enabling the implementation of magnetic resonance measurements with low magnetic field hardware, which would enable the use of on-site MRI as a process analytics tool.
Most recently, we have been looking at radical techniques that avoid altogether the need to produce an image in order to perform a particular analysis. Producing an intermediate image is very costly in terms of the amount of data required, and may be unnecessary when what is needed is a simple decision, or an estimate of the value of a few parameters. An example is estimating the density and shape distribution of bubbles in a reactor. We have shown that this can be done directly, without any intermediate image, resulting in much shorter acquisition times and tolerance to noise. Ultimately this could allow the use of more compact MRI machines, avoiding the need for strong magnetic fields, and thus avoiding the need for supercooling. This would simplify the machinery both in terms of weight and cost, and allow certain kinds of measurements which were previously impossible. The resulting machines could have an impact both in chemical engineering and, potentially, medical diagnosis.
Learn more about this research:
Primary Researchers
Lynn Gladden, CBE, FRS, FREng is pro-vice-chancellor for Research for the University of Cambridge. She is the Shell Professor of Chemical Engineering, and head of the Department of Chemical Engineering and Biotechnology.
Andrew Blake is a Microsoft Distinguished Scientist and managing director of Microsoft Research Cambridge. A Fellow of the Royal Academy of Engineering and The Royal Society, his research interests include probabilistic principles of computer vision software, with applications to motion capture, user interface, image editing, remote collaboration, and medical imaging.
