谷物联合收获机喂入量神经网络模型建立

The Research Situation of Monitoring and Controlling System of Combine Harvester

考虑物料的物理特性,对喂入量理论方程进行分析,利用正交试验获得喂入量的主要影响因素,建立了喂入量BP神经网络模型。该模型输入参数为物料湿度、油压力、谷草比,输出参数为喂入量。仿真试验结果表明:训练样本、确证样本和测试样本网络输出与网络目标的误差较小,相关系数可达0.999 5,模型可以准确地反映喂入量。用未参加训练的10组样本进行喂入量测定,网络模型的误差平方和为0.769 2,远小于回归方程的误差平方和2.656 2,可见网络模型优于回归方程。

Considering the physical characteristics of crops, the feed rate theory equation was analyzed in this paper. The main influence factors of feed rate had been get through orthogonal test, and then a feed rate BP networks model was modeled, whose input parameters are crop moisture content, oil pressure and grain-straw ratio, and output parameter is feed rate. The simulation test results show, the output error of training samples, validating samples and testing samples were small. The R-square of regression function between output and target of networks model for testing samples is 0. 999 5, so this model can accurately reflect the feed rate. The trained networks model was applied to measure the feed rate of ten samples, which had not been used to establish the networks model, the sum squared error of the networks model is 0. 769 2, better than the sum squared error of regression function, which is 2. 656 2.