查看更多>>摘要：? 2022Throughput is a key performance indicator of combine harvesters and an important basis for the control of its operating speed and loss rate. Current throughput monitoring methods are limited by low accuracy, weak stability, and poor applicability. Aimed at the real-time, accurate, and reliable detection of combine harvester throughput, we propose a throughput monitoring method based on multi-sensor decision level fusion and build a grain combine harvester throughput monitoring system (TMS). The experimental analysis method investigates the correlations of operation speed, crop density, feeding auger torque, conveyor torque, and cylinder torque with throughput, with single variable prediction models of throughput being established. Based on the findings, fusion calculations are conducted with throughput predicted using the single variable prediction models as the inputs and the correlation degrees of different variables with the throughput as the decision weight. Additionally, variations in the prediction results using different variables and results calculated by the decision-level fused model are employed for dynamic correction of the model. In this way, accurate detection of throughput is achieved. Field tests demonstrate that the maximum absolute error, average absolute error, maximum relative error, average relative error, and maximum root mean square error (RMSE) of the throughput monitoring of three combine harvesters by the proposed TMS are 0.49 kg/s, 0.2 kg/s, 4.9 %, 3.3 %, and 0.31 kg/s, respectively. The average absolute error, average relative error, and RMSE of the throughput monitoring of the three combine harvesters by the proposed TMS are 0.19 kg/s, 3.1 %, and 0.22 kg/s, respectively, thus suggesting high monitoring accuracy and stability, as well as good compatibility with various combine harvesters. This study provides important technical support for detection in the study of intelligent control technologies for combine harvesters.
NSTL
Elsevier