An mmWave radar-based mass flow sensor using machine learning towards a peanut yield monitor

Authors: BIDESE-PUHL, RAFAEL - Auburn University; BUTTS, CHRISTOPHER - Retired ARS Employee; REWIS, MATT - Kelley Manufacturing Company; McIntyre, Joseph; MORRIS, JASON - Kelley Manufacturing Company; BRANCH, BENNIE - Kelley Manufacturing Company; BAO, YIN - Auburn University

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/16/2023
Publication Date: 10/29/2023
Citation: Bidese-Puhl, R., Butts, C.L., Rewis, M., Mcintyre, J.S., Morris, J., Branch, B., Bao, Y. 2023. An mmWave radar-based mass flow sensor using machine learning towards a peanut yield monitor. Computers and Electronics in Agriculture. 215: Article 108340. https://doi.org/10.1016/j.compag.2023.108340.
DOI: https://doi.org/10.1016/j.compag.2023.108340

Interpretive Summary: The performance of a radar-based sensor intended to measure the flow of peanuts into harvesting machines was tested on static test setups for low flow rates and for high flow rates. The radar sensor can measure the peanut flow from outside ducts carrying the peanuts and dirt collected during harvesting protecting it from the abrasive conditions inside the peanut flow. The radar measured the velocity and distance from the sensor of the peanuts passing through a plastic duct section. The flow rate was measured for the low flow setup by an industrial electronic scale. The flow through the high flow set up was controlled by an adjustable speed conveyor belt calibrated to produce specific peanut flow rates. The peanut flow rates were compared to the output from the radar sensor using machine learning techniques to develop the conversion program to produce mass flow measurements from the radar results. The developed program successfully measured peanut flow rates in the duct for both test set ups.

Technical Abstract: A millimeter-wave FMCW radar-based mass flow sensor was developed to monitor peanut mass flow rate during harvest. The radar sensor components can be placed outside the abrasive flow in the pneumatic conveyor of a peanut harvester. Two systems to simulate the mass flow conditions of the field were used: one for research scale mass flow rate using a retrofitted 2-row harvester blower and one for commercial scale mass flow rate using a modified 6-row harvester. The ground truth for mass flow rate was obtained and the radar data used to get range-velocity image time-series. The datasets were generated using a sliding window interval, samples were generated to train machine learning models to predict mass flow rate from radar data. After evaluation of 5-fold cross validation, the best performing models achieved an RMSE of 0.14 kg/s (19 lb/min) and a sMAPE of 15 % while having an R² of 0.85 for the research-scale harvester and an RMSE of 0.52 kg/s (69 lb/min) and a sMAPE of 10 % while having an R² of 0.71 for the commercial-scale harvester. Moreover, the sensor can provide the harvester operator with information about the velocity of the peanuts to adjust the air pressure of the pneumatic conveyor to reduce undesired peanut shelling. The mass flow measurement results are promising. Further field investigation is necessary to evaluate the effects of noise caused by combine movement, foreign materials, peanut varieties, moisture content, and soil type.