Conference Agenda

A particularly wicked problem faced by modern farmers is the safe storage and preservation of cereal grains. A common approach to find the bulk moisture content of grain uses the electrical permittivity. Permittivity is a complex quantity, consisting of both real and imaginary parts. Most common capacitive moisture sensors operate by measuring the real part only and use the permittivity of water to make inference to volumetric moisture content of grain. However, valuable information is stored in both real and imaginary parts of the permittivity. Due to the close relationship with density, temperature and moisture content, a true understanding of the electrical permittivity of grain is vital for high-accuracy physical measurements.

We have obtained the full complex electrical permittivity of cereal grains using a coaxial transmission-line device (the ``permittivity cell'', or PC) in the frequency range 400 kHz to 1.3 GHz. Analysis is performed using a combination of numerical and analytical methods, including HFSS. We calibrate the device using fluids described by the Debye model. With non-parametric numerical inversion we recover the both components of the complex permittivity from the scattering matrix elements. We perform simultaneous inversion of both open and short cap measurements, ensuring the physicality of recovered solutions. Using the Kramers-Kronig approach, we find that non-parametric cereal grain permittivity results can be represented by an analogous parallel RC circuit. The implications of these findings and the possibility of utilizing this approach for advanced sensing applications will be discussed.

Grain storage is the process of holding back entropy; outside of removing waste, there is no known way to improve the quality of grain post-harvested. As a result, the farmer is faced with the difficult issue of storing grain, risking spoilage, or selling it at a lower price. Current grain management technologies are reactive and can only identify spoilage after it has occurred. In contrast, the challenge of a more predictive approach is that it requires a detailed understanding of the grain. One essential feature that needs to be understood to represent a grain bin is the pressure that the grain experiences, which affects grain compaction, aeration efficiency, and overall storage management.

An innovative approach to radial pressure distribution analysis in grain bins, utilizing a novel triaxial pressure sensor, is introduced in this study. Such sensors can be placed anywhere in a bin while filling to measure the x, y, and z pressures experienced by the grain in that specific region. Our objective involves using these sensors to refine and validate new mathematical models, specifically addressing the unknown variability of the vertical-to-lateral pressure ratio (k) throughout a bin. These sensors are deployed at strategic locations within a grain bin to generate a detailed pressure distribution map. This research aims to develop a comprehensive framework that improves our understanding of grain pressure fields and serves as a cornerstone for future predictive grain management approaches, including the application of digital twins to grain storage.

Ensuring a seamless and consistent flow of bulk grain is vital for efficient handling, processing, and storage, particularly with wheat, where impurities can significantly affect handling dynamics and storage properties. This study examined the impact of three dockage sizes (smaller than 1.1 mm, 1.1 to 2 mm, and larger than 3.3 mm) at four dockage percentages (from 0 to 10%) on the repose angle of dry wheat with 7.5% moisture content (wet basis). Employing an exponential growth model with two parameters as a regression framework, the study revealed complex relationships between dockage attributes and wheat's repose angle, assessed using both filling and emptying methods. The findings indicated variations between predicted and measured repose angles, with differences ranging from 0.01 to 1.62o for filling angles and from 0.56 to 4.56o for emptying angles. By illuminating the intricate interplay between dockage attributes and wheat behavior, especially in controlled silo environments, this research could enhance our understanding of storage condition optimization. These insights might offer valuable guidance for improving grain flow efficiency and ensuring optimal storage practices in agriculture.

Deoxynivalenol (DON) contamination from Fusarium graminearum colonization poses a significant challenge in Canadian wheat production. Both high atmospheric and interstitial CO2 concentration in field and storage, respectively, have been associated with fungal growth and mycotoxin production in grains. This study aimed to investigate DON production under different airtight conditions during the storage of wheat. Wheat samples with different moisture contents (17, 20 and 23%) were treated by: 1) natural infestation, 2) supplemental Fusarium inoculation, and 3) disinfestation followed by inoculation with Fusarium. The prepared samples were stored in sealed flasks at room temperature. The flasks were opened at intervals of 1, 3, 7, or 12 days. Prior to each opening, the concentrations of oxygen and carbon dioxide were measured using gas chromatography. Differences in CO2 and DON production under different storage condition were observed. The preliminary results will be presented.

Harvested grain is stored until the harvested grain is processed and delivered to customers for consumption. Insect infestation in stored grain not only causes nutrient degradation and dry matter loss, but also initiates mold growth inside the stored grain bulks. Detecting insect activities and monitoring insect populations is critical during storage as it helps in taking immediate corrective action to prevent further damage. Traditional sampling methods, although cheap, require skilled labor and are incapable of continuously monitoring insect activities and their population dynamics. Timely and in situ insect monitoring is needed. This study reviewed the application of in situ insect detection techniques combined with machine learning. Machine learning has been emerging for insect detection because it can classify various types of stored product insects. Technologies including image processing, acoustic measurements, mass spectroscopy, polymeric chain reaction (PCR) and electronic nose have been studied in literatures to produce real time detection of different species and different life stages of an insect in stored grain bulks. While suitable models including artificial neural networks, YOLO models, multivariate curve fittings are trained to classify or quantify the infestation level. These artificial intelligent-support techniques providing the next level of sensing and handling huge data at any point of time during the insect monitoring. The role of these new techniques in Integrated Pest Management of the stored grain products is also reviewed.

Plant breeding and genetic engineering have been used to establish high-yielding, drought-tolerant, disease-resistant, short-maturity corn varieties. However, there is no certainty that these varieties conform to the established thermal property values in literature, and the existing moisture content-based and proximate composition-based predictive models. This study evaluated the thermal properties of ten (10) corn varieties at five different moisture content levels, ranging from 13 to 21%. The aim was to examine the potential application of moisture content- and proximate composition models in predicting the thermal properties of corn varieties. Thermal properties were evaluated using existing reference scientific methods. The Moisture content – variety interaction significantly affected most of the thermal properties. The thermal conductivity, diffusivity, and specific heat ranged between 0.11 – 0.55 W m-1 K-1, 0.09-0.21 mm2 s-1, and 1.44 – 4.24 kJ kg-1 K-1, respectively. The predictive strength of both moisture content and proximate composition-based models varied greatly among corn varieties for thermal properties. The corn thermal properties were largely described by the Gaussian Process Regression (GPR) models compared to Linear Regression machine learning models. The existing moisture content and proximate composition-based predictive models are more suited for descriptive than predictive application with corn varieties in U.S.