Agricultural DEM study harvests information on grain damage
Published on: March 10, 2021
The U.S. Grain Council’s Corn Harvest Quality Report 2017-18 noted that about 16% of the country’s crop has some type of damage resulting from harvesting and processing including broken kernels, kernels with chipped edges, external cracks, and internal stress cracks. This damage results in crop waste, reduced market price, nutrient variability, and inconsistent product quality.
During agriculture production operations, grain kernels are subject to complex loading conditions. In an effort to reduce these problems, a leading agricultural equipment developer has teamed with Purdue University to develop tools to predict and ultimately reduce grain damage during harvesting. Zhengpu Chen, a Ph.D. student from Purdue’s School of Agricultural and Biological Engineering, is working with his research advisors, Prof. Kingsly Ambrose and myself, to develop discrete element method (DEM) models for predicting kernel damage. Specifically, Zhengpu is developing models for mechanical damage, which can occur during material handling operations such as threshing, separation, and conveying. For this study, DEM is an ideal tool, since grain is composed of many individual particles with complex shapes and dynamic behavior.
Although Zhengpu is focused on rigid corn and wheat kernels, DEM modeling can be applied to study a wide variety of materials. The complex nature of agricultural material coupled with the complexity of harvesting equipment requires sophisticated computational software, such as Rocky DEM, to simulate the loads on and movement of the grain. For example, the models describing the forces and movement of rigid corn kernels are different from those used to model high-moisture-content grapes or long, flexible grasses. Moreover, in order to help companies with their simulation work, the software must be flexible, fast, user-friendly, and integrate well with other computational solutions. Our work specifically has been focused on implementing experimentally validated grain damage models within the Rocky DEM framework so that grain damage can be predicted, offering insight to industry engineers as they modify equipment designs.
DEM parameters, testing and validation
A critical part of creating a good DEM simulation is to have models that accurately reflect the material’s behavior. For example, in our studies of corn and wheat kernels, we used x-ray micro-computed tomography to obtain 3D kernel sizes and shapes. Our initial simulations attempted to capture these shapes using a glued-spheres approach, which consists of using bonded and overlapped spheres to approximate the real shape of kernels. These “lumpy” grains worked well in capturing the bulk behavior of the grain in some cases, but performed poorly in others. Our current studies now use polyhedral representations of the grains in order to more accurately capture the true shapes. We believe these new shapes will result in improved predictions of bulk behavior.
In addition to grain size and shape, we have developed methods for measuring grain-grain and grain-boundary friction coefficients, grain elastic behavior, and grain−boundary coefficients of restitution. These parameters are used directly within the DEM model.
To verify that our models are capturing the packing behavior of the grain accurately, we have simulated two agricultural industry-standard tests: one for bulk density and one for the poured angle of repose. Comparing the bench test for bulk density with the simulation using glued-sphere particles produced a 4% error for corn and an 8% error for wheat. Although these values are acceptable, we expect they can be reduced with better kernel-shape representation. The angle of repose tests showed an error of less than 2%.
Grain damage model development
Different mechanisms can damage grains during handling, and each mechanism requires a different model. Kernels that are compressed and squeezed can exhibit stress fractures. A moving kernel pressed against a surface may have external wear damage and might even cause machine erosion. A kernel that hits a machine surface (or another grain) can bounce, with damage resulting from the impact.
To calibrate our damage models, we developed a series of bench experiments for measuring kernel damage over a range of compression forces and impact energies. These experiments included grain compression testing, pin-on-disk wear testing, and rotary impact testing. We quantified damage for individual grains via manual inspection for a large number of grains in order to produce statistically reliable data.
We found that wheat and corn are generally resistant to wear damage. Most of the kernel damage comes from single and repeated impacts of kernels with each other and especially with boundary surfaces. An impact-damage model described in the literature does an excellent job at describing our grain impact-damage data. We’re in the process of implementing this model in Rocky DEM and will compare it to experimental data for validation.
Why use Rocky DEM to improve grain damage prediction
Zhengpu’s studies to date have provided insight that will guide the next step in his research. For example, improvements in particle shape and contact force modeling have made the packing and damage simulation predictions more accurate. Although not as critical for our current model validation work, improvements in computational speed are essential for the practical application of DEM. One item of particular importance for us is the ability to adapt the DEM sub-models, such as the force models describing particle interactions, and also to extract internal DEM variable information, such as impact energies, impact locations, and the total work done on particles. It’s also critical to have an API in order to track the damage variables. Without Rocky’s API, we’d need to perform the tedious and time-consuming tasks involved in maintaining our own DEM codes. This commercially available and widely-used software, plus Rocky’s team of support staff who help us with model implementation while we focus on model development and validation, have greatly improved the efficiency of our work. One last item that shouldn’t be overlooked is the ease of technology transfer. Ultimately our goal is to develop models that can be used by industry to improve product performance. Working with the same software tools that industry uses makes technology transfer easy.
Polyhedral representations of the grains in order to more accurately capture the true shapes.
Customization using API
Ease of technology transfer to industry
Researchers at Purdue and external collaborators have been using Rocky DEM for projects other than grain damage. For example, we have used Rocky DEM to simulate the feeding of highly compressible, milled corn stover through a compression feed screw into a high-pressure chemical reactor. We’ve also been investigating the compaction behavior of systems consisting of spherical particles and flexible fibers. Each of these systems has unique modeling challenges, and Rocky DEM has been adaptable enough to meet our needs.
The following publications are related to this work:
Carl Wassgren is a professor in the School of Mechanical Engineering with a courtesy appointment in the Department of Industrial and Physical Pharmacy at Purdue University. His research has focused on developing models and experiments for predicting the dynamics of particulate systems. He has worked on a wide range of projects such as coating, blending, hopper flow, segregation, wet granulation, powder compaction, and attrition. He also teaches several powder-related courses at Purdue, including particulate systems and powder characterization. Wassgren has received a number of awards honoring his research and teaching activities. He also serves as the director for Purdue’s Center for Particulate Processes and Products.