A Cornell University based team investigates the intricate dynamics within a fluidized bed reactor.
by Caitlin Elizabeth Rockett
Petroleum has enhanced human life in a multitude of ways, but it’s an imperfect resource—decades of burning petroleum-based fuels have taken their toll on Earth’s ecosystem, and while the industry may be far from pumping its last drop of oil, it is a non-renewable resource. As such, scientists across the globe are searching for sustainable fuel alternatives that are as portable and efficient as petroleum.
Cornell University assistant professor Olivier Desjardins is part of a team of researchers in the pursuit of next-generation fuels. His team collaborates with experimentalists led by Dr. Mark Nimlos at the National Renewable Energy Laboratory (NREL) in Colorado, and Cornell assistant professor Perrine Pepiot.
Desjardins’ team develops numerical methods and models to investigate turbulent, multiphase and chemically reacting flows, such as fuel atomization and combustion in jet engines. Supercomputers allow Desjardins and his students to simulate the intricate internal processes of devices like engines on a microscopic scale that would otherwise be out of reach. Their work can be applied to a number of technologies, including the thermochemical conversion of biomass (such as wood chips or switchgrass) to produce liquid transportation fuel. Depending on the operating conditions, the first step of the conversion process can result in a gaseous mixture of carbon monoxide and hydrogen (gasification), or a liquid-phase bio-oil (pyrolysis). Each route presents significant challenges that need to be resolved before large-scale commercial production becomes possible.
Using two resources at the National Institute for Computational Sciences (NICS)—a Cray XT5 known as Kraken and an SGI Altix UV 1000 called Nautilus—the Cornell team has delved into the inner workings of fluidized bed reactors (FBR), which can be used in thermochemical conversion of biomass.
“One of the fundamental aspects of our project is the large size of the simulations that we run,” Desjardins said, explaining that graduate assistant Jesse Capecelatro used Kraken to run turbulent simulations up to 50,000 cores, close to half of the behemoth machine’s 112,896 cores. “The ability to run high core counts is essential to our progress, and this makes Kraken a vital aspect of our research.” Nautilus, with 4 terabytes of global memory, allowed the team to produce visualizations of their large datasets describing the complex flow inside a FBR.
An actual lab-scale (four-inch) FBR at NREL. This reactor served as the model for Desjardins’ team’s simulations. (Image courtesy of Mark Nimlos, National Renewable Energy Laboratory)
An FBR is essentially a cylinder filled with a granular material such as sand; super-heated steam is injected from the bottom of the cylinder at high speed, causing the sand to bubble and flow as though it were a high-density (much heavier than water) fluid. At this point the system has reached a uniform temperature of 600 degrees Celcius—that’s 1,112 degrees Fahrenheit—and biomass (wood pellets in this case) is slowly added. When biomass is added, constant uniform temperature within the reactor is necessary to provide a homogenous end product.
As the wood pellets heat up, the molecules that form the biomass are destroyed and the solid matter begins to sublimate, moving from a solid to a gas. This gas reacts within the reactor, and a mixture of hydrogen and carbon monoxide (referred to as syngas) rises to the top of the reactor where it is collected for further transformation into a liquid fuel source. While FBRs have been used in the coal industry for quite some time, there are still uncertainties and drawbacks in their use for biofuel production—exactly what Desjardins’ team is aiming to rectify. According to NREL scientists, there are two main issues.
“One of them is product variability,” Desjardins explains. “Researchers build lab-scale reactors—small versions of the industrial-sized machines—to use for experimental testing. These test reactors will produce a predicted amount of gas, but when you implement the same features in a 10-meter or larger reactor, it becomes far more difficult to predict the end product, how much hydrogen you have compared to carbon monoxide.”
The second issue is tar. “When you go through the conversion process you start creating large molecules called PAHs—polycyclic aromatic hydrocarbons. They get stuck together and form a heavy, gooey substance that deteriorates the reactor.” Desjardins explains that when tar builds up, it can impact the reactor performance and even deteriorate its components. Reactor and product cleanup becomes necessary, generating a higher operating cost. It’s not fully understood what mechanisms lead to tar buildup, but it is speculated that bubbles rising through the reactor from the torrent of injected steam play a key role in this process.
“We see gas bubbles that rise inside the bed of sand,” said Desjardins. “We see other regions of the reactor that remain very densely packed with sand. It’s pretty clear that whether a fluid particle goes through a bubble or stays in a really dense region will result in a very different residence time.” And while it may seem that fluidizing the bed more (injecting steam at a higher velocity) would increase bubbling frequency and decrease residence time, Desjardins says their simulations show that it’s not so simple.
“Faster fluidization of the bed tends to increases the distribution of residence times,” he explained. “We end up with some fluid particles not spending enough time in the bed, and we still see particles remain in the reactor long enough to build up tar.” With this knowledge, the bubbling dynamics of FBRs could be optimized to control residence time. If their work is successful, it might lead to novel reactor designs that experimentalists could implement and test in their lab-scale facilities.
Instantaneous snapshot of the simulation shown within the reactor. Volumetric rendering of the gas velocity on the left, iso-surface of gas volume fraction on the right. Gas velocity is the greatest inside the bubbles. (Image courtesy of Jesse Capecelatro and Olivier Desjardins, Cornell University)
NREL experimentalists operate an actual lab-scale reactor located on-site. Desjardins’ team is currently modeling the reactor isothermally (without temperature factored in) at a two-inch scale, about half the size of the actual reactor—a huge computational feat involving around 16 million sand particles and using 8,000 cores on Kraken. These models track the movements of each individual sand particle using a computational grid of 20 million cells. The team can follow particles as they collide and form bubbles of varying sizes. They can also measure the velocity of the bubbles as they rise as well as the frequency of bubble formation.
Instantaneous shot of the particles within the reactor—15.6 million particles tracked individually. (Image courtesy of Jesse Capecelatro and Olivier Desjardins, Cornell University)
The team has used available literature regarding the behavior of bubbles inside FBRs to validate their observations, and so far their simulations have very strong overall agreement with experimental data.
In addition to their role in residence time and tar formation, bubbles also affect how gas is mixed in the reactor. Observations by the team indicate that bubbles could be adversely influencing mixing, particularly in the denser parts of the FBR. As gas enters a bubble it is pushed through the reactor until it reaches the top, significantly reducing further mixing in the densely packed regions. The team may investigate ways to mitigate bubbles to ensure proper mixing.
The ultimate goal is to simulate a fully reactive FBR. This requires delving into the chemistry of biomass, which is the focus of Pepiot’s team. Desjardins says that Pepiot and her students already have the code in place to do this work, but they are currently doing two-dimensional testing of their current research. While Desjardins is unsure whether full-scale modeling of the 4-inch reactor will yield any more information than their current half-scale simulations, it is still a possible goal—the team has already produced full-scale models of a 2-inch reactor.
About NICS: The National Institute for Computational Sciences (NICS) is a joint effort of the University of Tennessee and Oak Ridge National Laboratory that is funded by the National Science Foundation (NSF). Located on the campus of Oak Ridge National Laboratory, NICS is a major partner in NSF’s Extreme Science and Engineering Discovery Environment (XSEDE).