For existing problems, such as the complex interactions between a crop and a machine, the measuring difficulty and the limited measurement precision of the feeding quantity within the corn silage harvester, a method of feeding rate measurement based on key conditions data, working data cleaning, and multiple variate regression is proposed. Non-destructive rotation speed, rotation torque, and power consumption sensors are designed for the key mechanical components. The data conditions, such as rotating speed, rotating torque, power consumption, hydraulic pressure, and hydraulic flow for the key operation of parts including cutting, feeding, shredding, and throwing are monitored and collected in real-time during field harvesting. The working data are screened and preprocessed, and the Mann-Kendall boundary extraction algorithm is applied, as is multiple component time lag correction analysis, and the Grubbs exception detection method. Based on a Pearson correlation analysis results, one-factor and multiple-factor regression models are respectively developed to achieve an accurate measurement of the corn feeding rate. The field validation tests show that the working data boundary extraction results among the load-stabilizing components such as shredding roller and throwing blower are highly reliable, with a correct rate of 100%. The power monitoring data of the shredding roller and throwing blowers are significantly correlated with the crop feeding rate, with a max correlation coefficient of 0.97. The determination coefficient of the single-factor feeding rate model based on the shredding roller reaches 0.94, and the maximum absolute error of the multi-factor feeding rate model is 0.58 kg/s. The maximum relative error is ±5.84%, providing technical and data support for the automatic measuring and intelligent tuning of the feeding quantity in a silage harvester.
Full Text Finder