Detecção precoce e predição da perda de matéria seca e qualidade de grãos de milho em tempo real durante o transporte

• Autor(a) principal: Nunes, Camila Fontoura
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Manancial - Repositório Digital da UFSM
• dARK ID: ark:/26339/00130000059bb
• Texto Completo: http://repositorio.ufsm.br/handle/1/24646
• topic: Biosensores; Fatores bióticos; Inteligência artificial; Monitoramento de grãos; Pós-colheita; Transferência de grãos; Biosensors; Biotic factors; Artificial intelligence; Grain monitoring; Post-harvest; Grain transfer; CNPQ::CIENCIAS AGRARIAS::ENGENHARIA AGRICOLA

Taking into account that the transport of grains can be carried out over long distances and that the mass of grains during transport, many times, has high contents of water, there may be risks of heat and moisture transfer and heating of the mass of grains, proving quanti-qualitative losses. Thus, this study aimed to validate a probe system for monitoring temperature, relative humidity and diffusion of the concentration of carbon dioxide in the mass of corn grains, in real time, during transport, as a function of different initial water contents (12, 16 and 25% w.b.), to detect early dry matter losses and predict possible changes in the physical quality of the grains. Portable equipment was developed to monitor the grain mass during road transport. The equipment consists of an Arduino Mega 2560 microcontroller as a control core. The system's hardware included three digital sensors to detect air temperature and relative humidity, a non-destructive infrared sensor to detect CO2 concentration, real-time clock modules and a micro-SD card. The output data from the digital sensor, infrared sensor and modules were connected to the I/O communication terminals of the microcontroller which were responsible for physical communication, component integration and data calculation. The temperature and relative humidity sensors were fixed at three ends of a threaded bar and the CO2 sensor was fixed in the central part. The real-time clock module and the micro-SD card were packaged in a plastic back box. The software used on the Arduino board was programmed based on the C++ programming language, with most of the libraries provided by the platform. The Arduino IDE (Integrated Development Environment) was used to develop the embedded firmware for the Atmega 2560 microcontrollers. To monitor the mass of corn grains, simulating a transport condition, a box was experimentally made with plywood material. in the dimensions of (0.2 x 0.2 x 1.8 m). The probe was inserted into the grain mass profile to assess the detection level of temperature, relative air humidity and carbon dioxide (CO2) in the grain mass. To define the hole diameter, the temperature, relative humidity and carbon dioxide sensors were placed in probes with different hole diameters (7.5, 7.0 and 6.5 mm), drilling height (470, 235 and 117.5 mm) and grain moistures contents (12.16 and 25%). The holes were made to allow the entry of air and facilitate the response of the sensors. In order to choose the diameter and drilling height of the probe that best fits, one of the requirements was that they meet the two analyzed moistures contents. The readings on the sensors were carried out until the values of temperature, relative humidity and carbon dioxide concentration stabilized. With the validation of the equipment, the definitive probe was made with a polyvinyl chloride tube with a diameter of 50 mm and a height of 1500 mm, with three perforated regions (upper, central and lower). A metallic grain sampler tube was developed to attach the probe. To evaluate the quality of the grains, the hygroscopic equilibrium moisture was obtained from the monitored intergranular air temperature and humidity; the concentration of carbon dioxide (CO2) to indirectly determine the dry matter consumption of the grains, the electrical conductivity test and the germination of the grains were carried out. To early predict physical changes in grain mass, Machine Learning and linear regression algorithms were used. The models tested were: artificial neural networks (ANN), linear regression (LR), M5P algorithm, reduced error pruning tree (REPTree), random forest (RF), and support vector machine (SVM). Among the results it was observed that the elevation of the parameters temperature, relative humidity of the air accelerated the metabolic activity of the grains and intensified the respiration of the grain mass, causing consumption of dry matter and alterations in the physical quality of the corn grains. The monitoring system with sensors for real-time measurement of temperature, relative humidity and concentration of carbon dioxide (CO2) in the mass of corn grains obtained satisfactory results, in which the probe was validated with a hole diameter of 6.5 mm and drilling height of 225 mm. The real-time monitoring of the variables indirectly and precociously determined the changes in the physical quality of the grains during transport, confirmed by the physical analyzes of electrical conductivity and germination. In the 2-hour period, the monitored variables indirectly indicated the physical changes that occurred in the grain mass. The condition of 16% of water content of the corn and the position of superior of the profile of the mass of grains suffered the biggest physical alterations of quality, mainly in the loss of dry matter, in function of the high equilibrium moisture content and respiration of the mass of grains. Real-time monitoring of corn grain mass and the application of Machine Learning algorithms predicted the quantitative and qualitative losses of corn grains in transport. All Machine Learning models, with the exception of the support vector machine algorithm, obtained good results, however, the multiple linear regression reached the best fits, being indicated for the prediction of grain losses in corn transport.