Monitoring and predicting corn grain quality on the transport and post-harvest operations in storage units using sensors and machine learning models

• Published in: Scientific reports Vol. 14; no. 1; p. 6232
• Main Authors: Rodrigues, Dágila Melo, Coradi, Paulo Carteri, Teodoro, Larissa Pereira Ribeiro, Teodoro, Paulo Eduardo, Moraes, Rosana dos Santos, Leal, Marisa Menezes
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 14.03.2024; Nature Publishing Group; Nature Portfolio
• Subjects: 639/705/1046; 639/705/258; 704/172; Air temperature; Artificial intelligence; Carbon dioxide; Corn; Dry matter; Drying; Electrical conductivity; Environmental conditions; Food preservation; Germination; Grain; Grain quality; Humanities and Social Sciences; Humidity; Loss reduction and grain conservation; Moisture content; multidisciplinary; Neural networks; Post-harvest technologies; Predictive models; Relative humidity; Respiration; Science; Science (multidisciplinary); Sensors; Water content; Loss reduction and grain conservation; Predictive models; Artificial intelligence; Grain quality; Post-harvest technologies
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-024-56879-5

Monitoring the intergranular variables of corn grain mass during the transportation, drying, and storage stages it possible to predict and avoid potential grain quality losses. For monitoring the grain mass along the transport, a probe system with temperature, relative humidity, and carbon dioxide sensors was developed to determine the equilibrium moisture content and the respiration of the grain mass. These same variables were monitored during storage. At drying process, the drying air and grain mass temperatures, as well as the relative humidity, were monitored. For the prediction of the physical and physical–chemical quality of the grains, the results obtained from the monitoring were used as input data for the multiple linear regression, artificial neural networks, decision tree, and random forest models. A Pearson correlation was applied to verify the relationship between the monitored and predicted variables. From the results obtained, we verified that the intergranular relative humidity altered the equilibrium moisture content of the grains, contributing to the increased respiration and hence dry matter losses along the transport. At this stage, the artificial neural network model was the most indicated to predict the electrical conductivity, apparent specific mass, and germination. The random forest model satisfactorily estimated the dry matter loss. During drying, the air temperature caused volumetric contraction and thermal damage to the grains, increasing the electric conductivity index. Artificial neural network and random forest models were the most suitable for predicting the quality of dry grains. During storage, the environmental conditions altered the moisture contents causing a reduction in the apparent specific mass, germination, and crude protein, crude fiber, and fat contents. Artificial neural network and random forest were the best predictors of moisture content and germination. However, the random forest model was the best predictor of apparent specific mass, electrical conductivity, and starch content of stored grains.