Estimation of mass flow of seeds using fibre sensor and multiple linear regression modelling

摘要

A new methodology to estimate the mass of grain seeds, which flow in the shape of clumps, was suggested in this paper. The methodology used an off-the-shelf digital fibre sensor to detect the behaviour of the clumps and multiple linear regression modelling to estimate the mass by the parameters detected by the sensor which were the length and the density of the clumps. An indoor apparatus was used for modelling which resembled the sowing process using the grain drill. A fluted roller was installed in the apparatus to regulate the flow of seeds. It was rotated by a motor at 2,4 and 6 rpm to cover all the possible mass flow rates which might occur during actual sowing. Modelling was based on several assumptions, such as the linearity between the sensor parameters and the mass. Although errors were observed while estimating each clump independently, the methodology was able to estimate the mass of a continuous flow of seed clumps. The average difference in estimation at 4 rpm was 0.9%, while there was overestimation at 2 rpm and underestimation at 6 rpm resulting in an overall estimation error of 5.3%. These results showed that the digital fibre sensor could be used for estimating mass flow of seeds at variable sowing rates within the speed limits of the grain drill.