为了直接、准确预测灌水器流量,引入支持向量机预测方法,取灌水器6个工作压力和8个几何参数作为因素,正交设计安排300组灌水器训练样本和30组检测样本,并采用精度较高的SST k-ω模型模拟计算灌水器流量,同时利用遗传算法对支持向量机参数进行优化计算,得到与模拟流量误差最小的流量预测值.结果表明,惩罚参数为100、核函数参数为20时检测样本的流量预测值与模拟值的误差最小,平均相对误差为1.91%,决定系数为0.98,而回归拟合方法计算得到的平均相对误差为6.45%,决定系数为0.93,表明支持向量机预测流量的优越性,且30组试验验证样本的平均相对误差为2.25%,证明支持向量机预测的准确性和可靠性.预测模型建立可有效地提高灌水器研发效率,对水力性能评估和流道结构设计与优化提供依据.%To carry out the prediction and calculation of the flow rate for further study the hydraulic performance and the structure optimization of the flow channel in drip irrigation emitter is of great significance. In order to predict and calculate the flow rate of the emitter accurately, in this study, the prediction and calculation method of Support Vector Machine (SVM) with strong generalization ability was introduced, and the flow rate prediction model of the SVM was built. We chose six working pressures and eight geometric parameters of the flow channel as factors, and arranged 300 sets of emitter schemes as training sample of SVM according to the orthogonal experimental design method, and 30 sets of schemes as test sample. Based on these, the prediction model sample set of flow rate of SVM was established. The flow rate of the emitter was simulated by the SST k-ω model with high precision in the sample set, and compared with the predicted value of flow rate of the SVM. The pressure and geometric parameter of the emitter was taken as the input item, and the flow rate was taken as the output item of SVM. The prediction and simulation of the flow rate were carried out in State Key Laboratory Base of Eco-hydraulic Engineering in Arid Area, Xi'an University of Technology. In order to eliminate the impact of each factor on the predicted results, the input and output item in the emitter sample were normalized before predicting flow rate. At the same time, the Genetic Algorithm was used to optimize the C and δ parameter in the Radial Basis Function (RBF) kernel of the SVM, and then the minimum error between the predicted value and simulated value of flow rate was obtained. The results showed that the relative error between the predicted value of flow rate using SVM and the simulated value was from 0.09% to 6.43%, the average relative error was 1.91%, and the determination coefficient was 0.98 when the optimal values of SVM parameter C and δ were 100 and 20, respectively. The predicted value of flow rate of SVM had a good correlation with the simulated value, which satisfied the predicted demand for the flow rate of the emitter. However, when the regression fitting method was adopted and calculated, the relative error between the predicted value and the simulated value was from 0.15% to 26.69%, the average relative error was 6.45%, and the determination coefficient was 0.93, which indicated excellent superiority based on SVM. To further verify the reliability of SVM, the five experimental verification schemes were chosen, and manufactured by using high-precision engraving technology. The flow rate value of experimental verification sample was tested under different pressure range, and was compared with the predicted value of flow rate. The relative error between the predicted value of flow rate using SVM and the experimental value was from 0.14% to 5.13%, and the average relative error was 2.25%, which were within the error range, verifying the accuracy and reliability of predicting flow rate using SVM. The establishment of the flow rate prediction response surface based on SVM can effectively improve the development efficiency of the emitter, and provide the evidence and guidance for the hydraulic performance evaluation, the flow channel structure design and optimization.
展开▼
