Experimental study of flow resistance in the presence of rigid vegetation and its prediction with intelligent models

Document Type : Research Paper


1 Depratment of Irrigation and Reclamation Engineering, University College of Agriculture and Natural Resources, University of Tehran, Karaj, Iran

2 Department of Water Science and Engineering, Arak University, Arak, Iran


Determining the resistance coefficients and reducing the uncertainty in selecting this parameter is one of the most essential factors in achieving the flow characteristics in rivers and open channels. Therefore, the appropriate selection of roughness coefficient in different conditions, such as vegetation, has been one of the important research topics. This research first determined the Manning’s roughness coefficient in a laboratory flume with different vegetation arrangements. Then, the ability of five intelligent models, including GMDH, ANN-RBF, RT, ANFIS, and ANFIS-PSO, to predict the Manning’s roughness coefficient was evaluated. The models were coded in the MATLAB software. Due to the creation of a gradually varied flow in the laboratory flume, the water level profile obtained through Euler's method was compared with the experimental values. The results showed an acceptable agreement between the experimental water level profiles and the estimates made by Euler's method. The evaluation of the results based on the statistics showed that the ANFIS-PSO model performs better than other models in predicting the Manning’s coefficient. Hence, the results of this model are RMSE=0.0096, R2=0.9984 and KGE=0.9922 in the training phase and RMSE=0.0099, R2= 0.9982 and KGE=0.9873 in the test phase. The ANN-RBF, GMDH, ANFIS, and RT models are in the next ranks. By evaluating the results of different combinations in modeling, it was found that three parameters of vegetation density (D), vegetation arrangement (N) and Reynolds number (Re) had, respectively, significant effect in estimating the correct results.


Main Subjects