نوع مقاله : مقاله پژوهشی
نویسندگان
1 گروه مهندسی آبیاری و آبادانی، دانشکدگان کشاورزی و منابع طبیعی، دانشگاه تهران، کرج، ایران
2 گروه علوم و مهندسی آب، دانشگاه اراک، اراک، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
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.
کلیدواژهها [English]
Experimental study of flow resistance in the presence of rigid vegetation and its prediction with intelligent models
EXTENDED ABSTRACT
The presence of vegetation in rivers and waterways is one of the main factors that increases the roughness coefficient, therefore, it should be considered in the designs. Research related to vegetation in waterways has often been conducted with the assumption of uniform flow conditions (Zhang et al., 2019). In addition, limited research has been done on different types of vegetation. Therefore, it is necessary to investigate Manning's coefficient in the presence of vegetation under gradually varied flow conditions.
This research aims to provide a high-accuracy model for determining the Manning's coefficient of the bed with different arrangements of rigid vegetation in the gradually varied flow conditions. There are various methods for calculating Manning's coefficient, including laboratory and intelligent models. Although intelligent models have high accuracy and low-cost, which are faster than laboratory works, experimental data is needed to train these models. Considering the need to accurately calculate the Manning’s coefficient in river beds for proper design, using intelligent methods to determine the Manning’s coefficient can be helpful. Based on the results of previous research, five models of GMDH, ANN-RBF, RT, ANFIS, and ANFIS-PSO were used to predict the Manning’s coefficient.
In this research, 86 experiments were conducted in a flume with vegetation with different arrangements under the gradually varied flow conditions, so that these data can be used for calibration and validation of the models. Then, five models of GMDH, ANN-RBF, RT, ANFIS, and ANFIS-PSO were used to evaluate the Manning’s coefficient. All models were coded in the MATLAB software. All parameters affecting Manning's coefficient were extracted and used as input and output parameters in modeling. Also, the experiments of this research were carried out in the central water research laboratory of the Department of Irrigation and Reclamation Engineering, University of Tehran. The examined flume has a rectangular cross-section with a width and height of 0.5 m and a length of 12 m, respectively. The bottom of the flume is made of plexiglass, and its wall is made of glass. The slope of this flume is constant and equal to 0.002. The laboratory's circulating water distribution system was used to supply the stream water used in this channel.
The results showed that there is an acceptable agreement between the laboratory water surface profile and the predicted results by Euler's method. The evaluation of the results based on some important statistics showed that the ANFIS-PSO model has a better performance 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. In the following order, the ANN-RBF model with the accuracy of RMSE=0.0157, R2=0.9962 and KGE=0.9663, the GMDH model with the accuracy of RMSE=0.0246, R2=0.9894 and KGE=0.9595, the ANFIS model with the accuracy of RMSE=0.0328, R2=0.9826 and KGE=0.9302 and the RT model with the accuracy of RMSE=0.0538, R2=0.9558 and KGE=0.9106 are in the test phase. Considering that different combinations were used in the modeling, it was determined by evaluating the results that three parameters, respectively, vegetation density (D), vegetation arrangement (N), and Reynolds number (Re), had a greater effect in providing correct results.
Using the Euler's numerical method, it was shown that the laboratory data matched the water surface profile calculated by this method. By examining the results, it was found that the ANFIS-PSO model has the most accuracy compared to other models. The lowest performance was assigned to RT. The parameters of vegetation density (D), vegetation arrangement (N), and Reynolds number (Re) played the most crucial role in the development of models. Therefore, the ANFIS-PSO model should be used to predict the Manning’s coefficient in the mentioned conditions.
)