Real-time monitoring of sugarcane harvester feed rate is great significance for guiding harvesting operation and improving efficiency. In this study, a feed rate monitoring system of sugarcane harvester is designed and developed. The system adopts the proposed iterative wavelet threshold denoising technology to enhance data quality. Compared with Fourier transform and traditional wavelet threshold method, the signal-to-noise ratio of the collected signal is increased by 41.6% and 10.5% respectively, and the root mean square error is reduced by 32.5% and 12% respectively. A nonlinear adjustment particle swarm optimization back propagation neural network (NAPSO-BPNN) is introduced and established with the hydraulic motor outlet pressure of the base cutter, the hydraulic motor outlet pressure of the lower conveyor roller, the displacement of the upper conveyor roller, and the flow rate of the hydraulic motor of the chopper as inputs, and the feed rate as the output. The NAPSO-BPNN demonstrated lower uncertainty in initial weight and threshold settings, with determination coefficients increasing by 0.12 and 0.06, and average relative errors decreasing by 8% and 3.8% compared to traditional BPNN and PSO-BPNN. Finally, the accuracy and reliability of NAPSO-BPNN feed monitoring model were verified in three plots with sugarcane growing well, growing poorly, and seriously lodging. The determination coefficients of NAPSO-BPNN feed monitoring model on three plots are 0.954, 0.93 and 0.911 respectively. The average relative errors are 7.43%, 8.16% and 9.26% respectively, and the root mean square errors are 0.157, 0.223 and 0.247 respectively. Therefore, the monitoring system of feed rate developed in this study is accuracy and reliability in different plots. The outcomes of this study are expected to provide robust technical support for adjusting the operational status of harvesters and optimizing real-time parameters.
Read full abstract