Research On Forecasting Model Of Open Channel Roughness Coefficient Based On Typical Support Vector Machine Algorithm

Roughness coefficient is an important and sensitive comprehensive parameter that reflects the boundary of the river bed and hinders the flow of water.In the hydraulic calculation of open channels,the research on the roughness coefficient is one of the key problems that must be tackled urgently in the development of hydraulic calculations in the open channel.At present,the commonly used method of selecting the roughness coefficient is often limited by the engineering conditions and practical experience,so that it can not meet the needs of the actual project.In order to select a more reasonable coefficient of roughness,based on the experimental data of artificial rough open canal,multivariate statistical analysis and support vector machine technology were applied to establish four models of principal component analysis-support vector machine(PCA-SVM),principal component analysis-least square support vector machine(PCA-LSSVM),partial least squares-support vector machine(PLS-SVM)and partial least squares-least square support vector machine(PLS-LSSVM),the prediction of roughness coefficient is studied,and mainly obtained the following conclusions:(1)Principal component analysis PCA and partial least-squares PLS are used to preprocess the artificial roughness coefficient data of rough open channel,and the comprehensive components of the feature information are extracted.(2)By combining PCA and PLS preprocessed data with Support Vector Machine SVM and Least Square Support Vector Machine LSSVM,the four kinds of combined models are trained separately,and the prediction of the roughness coefficient is realized.(3)Comprehensively examine the prediction accuracy and operational efficiency of the four kinds of combination models and use them as criteria for the evaluation model,the results show that the four kinds of combined models have achieved a better prediction effect,and the PLS-LSSVM model is more prominent in the prediction of roughness coefficient,indicating that the PLS-LSSVM model is suitable for prediction of roughness coefficient in artificially roughened channels.(4)Based on the dimension analysis method,the empirical relationship of the roughness coefficient is obtained by numerical fitting,the hydraulic factors included in the formula are Froude number Fr,absolute roughness ?,channel average water depth h and bottom slope i,which are consistent with the selection factors of the prediction model.(5)Using PLS-LSSVM model and fitting formula to predict the roughness coefficient respectively,the results show that the prediction accuracy of PLS-LSSVM model is obviously superior to the prediction result of formula method,and further illustrate the advantages of the combination model in the prediction of open channel roughness coefficient.