以矩形人工加糙渠道为研究对象, 采用主成分分析-支持向量机方法建立糙率系数预测模型。根据前期试验研究成果, 选取佛汝德数 Fr、绝对粗糙度△、渠道平均水深 h、底坡 i 这四个主要影响因素, 采用主成分分析方法提取两个主成分, 获得影响糙率系数大小的综合性指标并用于支持向量机对数据的训练、测试及预测。研究结果显示: 模型的训练集均方根误差 RMSE 为 3. 85 ×10﹣4、预测相关系数 R 为 0.997, 测试集均方根误差 RMSE 为 5.37×10﹣4、预测相关系数 R 为 0.992、预测相对误差小于5% 。研究结果表明, 基于主成分分析-支持向量机所建模型适合 人工渠道糙率系数的预测。

With the rectangular artificially roughened channel as the research object, we established a prediction model of roughness coefficient using the principal component analysis-support vector machine method. According to the preliminary experimental results, we selected four main influence factors: Froude number Fr, absolute roughness△, channel average water depth h, and bottom slope i. We used the principal component analysis method to obtain two main components, and obtained the comprehensive indexes influencing roughness coefficient, and used them for data training , testing, and prediction of the support vector machine. The research results showed that the RMSE and prediction correlation coefficient R of the training setwere 3.85×10﹣4 and 0.997 respectively, while those of the test set were and 0.992 respectively. The relative error was less than 5% . The results showed that the model based on principal component analysis-support vector machine is suitable for predicting the roughness coefficient of artificial channels.

新疆维吾尔自治区自然科学基金项目( 2015211A025)