Research Of Predictive System For Feed Quantity Of Combine Based On Fuzzy Neural Network

This paper analyzed the power consumption of the threshing cylinder and decided the relationship between the rotating speed and the threshing cylinder and the feed quantity . On the base, in order to raise the efficiency of the combine harvester, this paper suggested to stabilize the rotating speed by adjusting the feed quantity to a stable amount.In the part of measuring feed quantity, this paper used the torque of driving axle of the feed system to reflect the feed quantity and got the equation of the torque and feed quantity by experiment which was carried out by the "simulation and control test equipment DF-1.5 for grain threshing and separating".When measuring feed quantity, this paper calculated the density of grain. But the combine harvester was a delayed, non-linear and time varying system. Particularly when the automatic control system of the combine adjusted the working parameters according to the signal got from the sensors, the problem of delay became more obvious. So this paper designed a predictive system to forecast the density of grain nearby the headers on the ground of density got from the torque of driving axle of the feeding system and adjust the forward speed of combine harvester according to the result forecasted by the predictive control system.Appling the technology of Fuzzy Neural Network, this paper construct the predictive system according to the actual destination and then decided the input and output of the network and its fuzzy dividing. At the same time, this paper investigated the constructing of fuzzy rule base and designed the topological structure of networks.This paper carried out the experiment of paddy threshing and separating by using "simulation and control test equipment DF-1.5 for grain threshing and separating", and got the data about the density and its variation of paddy. From these data, the paper selected the example data for training , test and abstracting initial fuzzy rules. At the same time, after constructing the fuzzy neural network of predictive system using the software of MATLAB, this paper trained the network using the example data and checked the generalizing ability of the network. The result proved thenetwork of predictive system to be practicable.At the end, this paper carried out the simulation experiment using the predictive system of combine. At this experiment, this paper contrasted the effect of adjustment of the automatic control system in which the predictive system was or was not added, the result of simulation experiment proved the predictive control system to be effective.