基于RBF-PLS的水稻钵盘精量播种装置充种性能预测

Prediction of Filling Performance of Rice Bowl Precision Seeding Device Based on RBF-PLS Model

钵盘精量播种是一种精准的播种技术,实现钵盘精量播种的前提条件是芽种的精准充种。芽种充种性能好坏直接影响精量播种合格率高低,评判指标为芽种充种合格率与芽种充种损伤率。传统充种合格率与损伤率数值多通过人工查数计算获得,劳动投入量大、过失误差发生概率高,结果准确性得不到保证,且由于各试验条件不同,充种结果不能推广使用。因此,建立适用性强、准确率高的充种预测模型,以期为不同试验条件提供预测方法。以钵盘型孔直径、钵盘型孔厚度、种箱速度、芽种含水率及刷种高度为试验因素,以充种合格率与芽种损伤率为性能指标。在播种试验台上对芽种进行充种性能的二次正交旋转组合试验,以期为建模提供数据来源;结合径向基函数神经网络模型的学习方式与偏最小二乘法的统计原理,建立的RBF-PLS网络结构模型,获得充种性能模型曲线。选取曲线中有代表性的试验数据,借助Matlab预测模块,获得水稻钵盘精量播种充种性能合格率、损伤率的预测值,得到第5组充种性能最好,合格率为97.14%,损伤率为0.288%。并将预测试验结果与台架试验试验结果进行对比,判定模型可靠性。研究结果表明:基于RBF-PLS网络的充种模型可以较好的预测排种装置充种性能,合格率与损伤率误差平均值分别为0.864%和0.06%,符合评价标准,RBF-PLS网络模型预测结果准确,可进行推广使用。

Bowl-seeding is an accurate seeding technique.The quality of bowl filling directly affects the qualified rate of bowl filling and the damage rate of seed filling.Because of the limited test conditions,different factors affecting the same crop and the requirements of different crop fillings must be different.The application of the filling result was limited and can not be popularized.Therefore,the key of this study is to establish a prediction model with strong applicability and high accuracy in order to provide prediction methods for different test conditions.The specific research methods were as follows:Taking the diameter of bowl-plate hole,thickness of bowl-plate hole,seed box speed,moisture content of seeds and height of brushing seeds as the experimental factors,and taking the qualified rate of seed filling and the damage rate of bud seed as the performance indices.Based on the study method of radial basis function neural network model and statistical principle of partial least square method,the RBF-PLS network structure model was established and the filling performance model curve was obtained.By selecting the representative test data in the curve and using the Matlab prediction module,the prediction values of the qualified rate and damage rate of seed filling performance of rice bowl plate were obtained.The study showed that the RBF-PLS network-based charging model could better predict the charging performance of the seeding device,the average values of qualified rate and damage rate error were 0.864%and 0.06%,which accords with the evaluation standard.RBF-PLS the network model prediction results can be popularized.