Using DEM to improve corn stover biofuel production efficiency
Share this
Published on: April 17, 2020
A study of 40 solids processing plants in the U.S. and Canada found that 80% of facilities experienced solids-handling problems, and most of the plants were slow coming online. Once operational, handling problems continued, resulting in performance of only 40% to 50% that of design. Among global research efforts to develop renewable energy sources, converting biomass to fuel holds much promise. Consider corn: a huge world crop, over a billion tons grown annually. After harvest, the leftovers — leaves, stalks, and cobs of plants left in a field — make up about 40 percent of that yield. Certainly sustainable as a biofuel, corn stover is both the largest underutilized crop and the primary source of biomass in the United States.
Corn itself (and other high-starch grains) have been converted into ethanol for thousands of years, yet only in the past century has its use as fuel greatly expanded. But scientists and engineers have hit a roadblock: understanding the flow of this grainy, fibrous raw material in refining facilities. In the particular case of corn stover, it often behaves unpredictably while moving through augers on its way to being processed, forcing costly shutdowns for cleaning and repairs.
Particle simulation is an obvious tool to solve such problems; however, discrete element modeling (DEM) only recently has become sophisticated enough to realistically model random particle behavior on an industrial scale. Purdue University’s Center for Particulate Products and Processes (CP3) has teamed with government (U.S. Department of Energy, Idaho National Laboratory), academic (Penn State) and industry (Rocky DEM, AdvanceBio and Forest Concepts) partners to advance the technology.
Corn stover biorefinery process
Overall, the milled corn stover project is investigating how to make ethanol biorefineries more reliable, specifically related to the work that mechanical engineers perform. One of the biggest causes of refinery downtime is at the front end of the process. A feed screw consistently supplies materials into the process, and a compression screw feeder compresses the feedstock into a compact plug, which forms a barrier preventing reactor gas back-flow. As loose corn stover material is conveyed through the auger (Figure 1), its flow is irregular. The biomass particles sometimes knit together and pack tight, jamming the machine without warning. Some equipment incorporates hatches where workers can access the compression screw feeder to break up the jam. When material is packed as densely as particle board, it can require power tools to clear. In some cases, the whole process must be shut down. Consequently, screw-feeder jams and related problems contribute greatly to inefficient plant function.
Figure 1. At the front end of the biorefinery process, a system of feed screws prepares and conveys milled corn stover through the hopper to a pressurized reactor. Improving biofuel equipment and processes relies on accurately predicting particle behavior parameters in this equipment, including mass flow rate, torque, elasticity, pressure and bulk density.
Corn stover modeling
None of these problems is easy to solve. Studying corn stover material and its flowability, its mechanical properties, its rheology, in some cases the tribology is really an interesting body of work. Due to small particle size, large compressibility and low bulk stiffness, corn stover requires novel modeling approaches related to proper screw design (for example, pitch, channel depth, effects of wear). From literature, we found that R&D teams traditionally run a slew of parametric studies, since there is not much practical experience to be considered as a design springboard. They identify reasonable input parameters, vary them widely, then run many DEM simulations all at the same time — which requires great computational power — and finally perform post-processing. In our research, we uniquely integrated Rocky DEM into the optimization process using an API. This way, an algorithm runs only the simulations that are necessary to calibrate our simulations to match our experiments.
Corn stover modeling challenges relate to high compressibility, significant material densification, and complex screw geometry.
CP3 has a long history of investigating pharmaceutical industry issues, including material compaction that involves large plastic deformation force laws. Existing DEM models mis-predict forces beyond 5 percent strain; in a typical biomass application, volumetric change can be up to 70 percent! Consequently, the team implemented a large plastic stress model —nonlinear elasto-plastic spring — into Rocky (Figure 2).
Figure 2. Rocky DEM large plastic stress model introduced into simulation process. Initially contact is locally plastic, but as strains become large, plastic regions interact, causing a saturation.
Validation
Because input parameters are difficult to measure directly, the team has taken a calibration-style approach that includes the Rocky DEM model in the simulation process (Figure 3). Penn State’s cubic triaxial tester (CTT), which determines fundamental mechanical properties of powder materials, will be used to calibrate material properties related to compressive material behavior under large strains. The iterative calibration algorithm optimized against only CTT data gives material properties that can approximate behavior in other simple systems.
Figure 3. Calibration procedure
Rocky DEM simulation using CTT data
As the final part of our research study, the team will validate the model against compression feed-screw experimental data from academic and corporate sources, then perform parametric studies. Our research is a step toward realizing corn stover biomass refining as a predictable system. Right now, we’re putting a lot of energy into growing the material and harvesting it, but we still have a lot of untapped energy.
Eventually, the numbers that we end up with in our research will go right out into the agriculture and biofuel industries and be used in a real system! That’s what I’ve come to appreciate in our team’s work: the ability to be a true bridge between raw science and pure application.
Nathan Gasteyer
Center for Particulate Products and Processes, Purdue University
Nathan Gasteyer is a graduate research assistant pursuing a master’s degree in mechanical engineering at Purdue University, where he also received his undergraduate degree in mechanical engineering. He has worked at Argonne National Laboratory and Fiat-Chrysler Automobiles. Gasteyer’s technical interests include design thinking, machine learning and machine design, and he looks forward to bringing these to his post-graduation job as a Northrop Grumman engineer.
