Real-time monitoring of combine harvester grain loss in glutinous rice harvesting

The increasing demand for glutinous rice in Malaysia has prompted efforts to improve local production. One of the major challenges in achieving higher yields is significant grain loss during harvesting, largely due to improper management of combine harvester operational parameters. Conventional methods for measuring grain loss are labor-intensive and lack real-time monitoring capabilities. This study developed a real-time sensor-based system to detect and monitor grain loss during harvesting. The specific objectives include evaluating the performance of the combine harvester in glutinous rice fields, optimizing the operational parameters of the combine harvester to minimize grain loss, and developing a grain loss monitoring system. Field experiments, designed using Central Composite Design (CCD), focused on four key factors: forward speed (1.76–2.6 km/h), header height (150–350 mm), feed rate (2–6 kg/s), and cleaning fan speed (1100–1700 rpm). A total of 30 experimental runs were conducted to measure grain loss and evaluate the combine harvester’s performance. Grain loss data were analyzed using Response Surface Methodology (RSM) and Artificial Neural Networks (ANN), with model accuracy assessed using R², mean absolute error (MAE), and root mean square error (RMSE). Statistical analysis of the two models showed strong model performance, with R² values of 0.9909 and 0.9892, MAE values of 1.25 and 1.73, and RMSE values of 1.69 and 3.68, respectively. The optimal operational settings of the combine harvester were determined using the RSM-DF model (2.26 km/h forward speed, 150 mm header height, 1100 rpm cleaning fan speed, and 3 kg/s feed rate) and the ANN-GA model (2.26 km/h forward speed, 252 mm header height, 1100 rpm cleaning fan speed, and 3 kg/s feed rate), resulting in minimum grain loss values of 23.58 kg/ha and 14.70 kg/ha, respectively. A real-time grain loss monitoring system was developed using PVDF piezoelectric sensors and a microcontroller-based circuit, achieving 88.45% accuracy at optimal settings (2.26 km/h forward speed, 350 mm header height, and 1500 rpm fan speed). Additionally, machine learning models were explored for predicting system performance, with the Decision Tree (DT) model outperforming the Random Forest (RF) model, achieving R² values of 0.9992 for training and 0.9987 for testing. The forward speed of the combine harvester was identified as the most critical factor affecting machine performance. Integrating a real-time grain loss monitoring system with machine learning predictions can significantly enhance combine harvester efficiency and reduce grain loss during glutinous rice harvesting.

Text FK 2024 17 - Full Text.pdf Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (68MB) Text FK 2024 17.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (68MB)

Actions (login required)

View Item