A THREE-DIMENSIONAL, ASYMMETRIC, AND TRANSIENT MODEL TO PREDICT GRAIN TEMPERATURES IN GRAIN STORAGE BINS

Abstract

A three-dimensional, transient, combined model (headspace model + soil model + conduction model in bulk grain) was developed to predict grain temperatures in a granary. Different meshes (mesh refinement in the whole domain or at the boundary) including linear and hybrid (linear and quadratic) elements were used to simulate grain temperatures. Prediction accuracies of temperatures produced by the different meshes were compared, and the model was validated using measured temperatures in two flat bottom bins (3.76 m diameter and 5.5 m high filled with wheat up to 3 m) located side by side in the north-south orientation near Winnipeg, Manitoba. Grain temperatures predicted by the model were in close agreement with the measured temperatures throughout a 21-month test in the two bins. By using a hybrid element mesh (mesh refinement at the boundary), the mean, standard error, and maximum of the absolute difference between the measured and predicted temperatures in the south bin were 2.2°C, 0.4°C, and 7.0°C, respectively. The mean, standard error, and maximum of the absolute difference predicted by a linear element model (88 linear elements each layer) in the south bin were 2.1°C, 0.3°C, and 6.3°C, respectively. Including a headspace model improved the prediction accuracy of the conduction model at the top of the grain bulk. Mesh refinement only at the boundary produced a homogeneous distribution of errors in the whole domain; however, mesh refinement in the whole domain gave higher errors at the walls than at the center of the bins. Considering the increased computer time and slightly improved accuracy by mesh refinement at the boundary, a uniform mesh with mesh refinement in the whole domain was preferable for predicting grain temperatures in an entire grain bin.