Calibration and Validation of a Discrete Element Model of Corn Using Grain Flow Simulation in a Commercial Screw Grain Auger

Abstract

Screw augers are primary grain conveying equipment in the agriculture industry. Quantitative prediction of grain conveyance using screw augers requires better understanding and measurement of bulk particle-particle and particlerigid-body interactions. Discrete element modeling (DEM) has potential to simulate particle dynamics and flow within a screw auger and thus provide simulation-based guidance for auger design and operating parameters. The objective of this study was to develop a DEM corn model calibration methodology and validation for combine-harvested corn flow in a commercial screw auger. The methodology used a virtual design of experiment (DOE) varying DEM corn parameters and calibration to match grain pile formation expressed in a normalized angle of repose (AOR). DEM corn particle shape was approximated using 1-sphere and clumped spheres (2-sphere, 5-sphere, and 13-sphere) matching the measured physical parameters of equivalent geometrical diameter, 2D axial dimensions, 3D axial dimensions, and detailed CAD-approximated corn dimensions, respectively. For each DEM corn shape approximation, a virtual DOE using Latin square hypercube design with four independent DEM Hertz-Mindlin contact model interaction coefficients was developed. The DEM assembly of particles matching the initial conditions of the AOR test was created in EDEM 2.7. From the quasi-static AOR of corn flow in the AOR tests and EDEM simulations, the mean square error (MSE), a sum of square difference in grain heights in the AOR tests and EDEM simulations, was used as a bulk material dependent response for the calibration process. The DEM 2-sphere corn shape model and the material interaction coefficients showed the minimum MSE (5.31 mm) compared to the 1-sphere, 5-sphere, and 13-sphere models. With the best DEM corn shape model (2-sphere) and DEM model parameters with the minimum MSE, validation of the DEM in predicting corn flow in a commercial screw auger in laboratory tests at two rotational speeds (250 and 450 rpm) was performed and showed good prediction (within 5% relative error) in matching the change in mass flow rate with the change in auger rotational speed.

Discrete Element Model Calibration Using Multi-Responses and Simulation of Corn Flow in a Commercial Grain Auger

“…Descriptive statistics (means and standard deviations) for the corn kernel dimensions, aspect ratios, and material properties are presented in table 1 (Mousaviraad et al, 2017). The mean corn particle density was within the ranges reported by Boac et al (2010) and González-Montellano et al (2012).…”

“…2, left). The bucket dimensions, material type, and testing procedure were similar to those reported by Mousaviraad et al (2017). A box (height = 200 mm, length = 250 mm, and width = Gate 120 mm) was fabricated from acrylic sheet.…”

“…For measurement of the corn moisture content, particle density, and particle mass, the methods reported by Mousaviraad et al (2017) were used. The moisture content was measured by the oven-drying method at 105°C for 24 h (ASABE, 2006).…”

“…Similarly, Fernandez et al (2011) used DEM simulation of a grain auger application to obtain DEM properties that agreed with the experimental results before using the DEM for virtual screw blade designs. Using a calibration approach with a single bulk response from bucket AOR tests to obtain the minimum MSE (Mousaviraad et al, 2017), the DEM predicted the steady-state mass flow rate with prediction errors of 27% and 29% for 250 and 450 rpm, respectively. The underprediction of steady-state mass flow rate might have been due to limitations in capturing particle-to-particle friction behaviors (sliding and rolling).…”

“…Developing a robust numerical methodology along with experimental testing for controlled, reproducible field-harvested grain conditions are important research challenges in using DEM for the design and performance analysis of grain handling equipment. In our previous work (Mousaviraad et al, 2017), a single AOR calibration and optimization approach was introduced to generate corn DEM friction parameter values and resulted in DEM simulations of grain auger mass flow rates with prediction errors of 27% and 29% for 250 and 450 rpm grain auger speeds, respectively. The prediction accuracy could be improved by using calibration for multiple-response AOR to investigate the effects of grain-to-grain, and grain-to-rigid-body interactions in quasi-static grain flow and by including such multi-response effects during the surface response optimization.…”