Estimating the SEDIMENT being transported by river flow is one of the important aspects in water resources engineering. Erosion and SEDIMENT transport phenomena in watersheds and rivers are complex hydrodynamic problems. Due to large number of obscure parameters involved in these phenomena, the theoretical governing equations may not be of much advantage in gaining knowledge of the overall process. Researchers have developed practical techniques that do not require much theory, algorithm, or rule development, and thus, reduce the complexities of the problem. One such technique is known as ARTIFICIAL Neural Networks (ANN). In this paper, Auto-Regressive ANN was utilized to estimate suspended SEDIMENT lood in rivers. Various network topology, data partitioning and parameters were examined to find the best network with the best results. For increasing the efficiency of the models, Early Stopping technique has been used. Results of these networks were compared to the conventional SEDIMENT rating curves method and it was shown that ANN presented better results especially in peak flow discharges. Trained networks were able to model the SEDIMENT transport phenomena in rivers successfully, presumably because of the superior capability of ANN in nonlinear mapping, without any extra information from governing equation.

Estimating total SEDIMENT load in rivers is an important and practical issue for water resources planning and management. The SEDIMENT concentration can be calculated by both direct and indirect measurements, but direct methods are usually costly and time-consuming. Further, total SEDIMENT load can be determined by several SEDIMENT load transport models. These equations, however, are applicable in certain circumstances, and in most cases the outcomes do not agree with each other and with measured data. The objective of this study was to propose a method based on ARTIFICIAL neural networks (ANN) to predict total SEDIMENT load concentration. Consequently, two ANNs including multilayer perceptrone (MLP) and radial basis function (RBF) with 200 data were used for the modeling purposes. For training and testing the ANN models, 75 and 25 percent of data were used, respectively. The input variables were designated to be average flow velocity, average depth, water surface slope, canal width and median particle diameter of SEDIMENT, while the models output was total SEDIMENT load concentration. The input variables were included to the models step wisely and the results were evaluated to find out the most suitable ANN models. The predicted values were then compared with five known SEDIMENT load transport equations. The conducted statistical analyses indicated that ANNs models in particular MLP can provide better prediction for total SEDIMENT load with correlation coefficient of 0.96. It was further concluded that to enhance the accuracy of ANN model, training of the network should be accomplished using both hydrological and SEDIMENT data. The Ackers and White equation was very overestimating the total SEDIMENT load, while all other equations were underestimating. Based on the results obtained in this study, the ANN-based models provide better concurrence with the observed data, particularly MLP network which can reasonably well predict the peak point of total SEDIMENT.

Accurate estimation of suspended SEDIMENT in rivers is very important for design and operation of water resources projects. SEDIMENT estimation by conventional methods like rating curves are not able to provide correct results. In this paper, gene expression programming (GEP) model which is an extension to genetic programming (GP) technique, for suspended SEDIMENT estimation is applied. The GEP results are compared with those of the adaptive neuro-fuzzy, neural networks and rating curve models. In this regard, the streamflow and suspended SEDIMENT data from Vanyar station that located on Aji-chay river in East- Azarbaijan are used. The root mean square errors (RMSE) and determination coefficient (R2) statistics are used for evaluating the accuracy of the models. The results show that the GEP model is the best among other models in estimating suspended SEDIMENT. The relative RMSE difference for test period between GEP and ANFIS-Grid Partitioning, ANFIS-Sub Clustering, ANN and rating curves methods was 8, 10, 13 and 21 percent respectively. The R2 values for GEP and ANFIS-Grid Partitioning, ANFIS-Sub Clustering, ANN and rating curves methods was 0.93, 0.84, 0.88, 0.86 and 0.81.

In ARTIFICIAL neural networks (ANN), the available methods of neural learning and calibration is according to multi-layer structure of perceptron, but these methods have some problems results from lack of convergence in learning methods, lack of stability in network weights in conditions that there is great criteria deviation in input data spectrum and finally need for much data and information for network learning. A new compound method of ARTIFICIAL neural network-non linear mathematical optimum was introduced in this research to overcome this problem and ARTIFICIAL neural network designed by using error back propagation method was introduced as a strong device for estimating the rate of SEDIMENT in the reservoir of ecbatan dam. According to that, the designed model by various knots in input and hidden layer was performed by using the equation of SEDIMENT discharge and water current and statistics of Yalfan station at Abshineh River. Calibration results show that 6 knots at input layer and 8 knot at hidden layer should be used to distribute SEDIMENT in reservoir of Ecbatan dam.

The estimation of alongshore SEDIMENT transport rate (LSTR) is the most important factor in analyzing the amount of erosion or accretion along a coast. In the present research, an LSTR measurement was done at daily intervals using SEDIMENT traps in Noor coastal area, north of Iran, from March 21 to June 22, 2012. The existing empirical relations are linear or exponential regressions based on the observations and measurements data. Based on calculations, the yearly average of SEDIMENT transport rate is 928.73 (m3/day) for Noor coastal area. One of the most widely used methods for estimating LSTR, which has advantages compared with others, is setting up and application of an ARTIFICIAL neural network (ANNs) and the present study attempts to develop such a model. Different ANNs with different input configurations and transfer functions were examined. The results reveal that usage of the hyperbolic tangent is better than application of the sigmoid as the transfer functioning. Moreover, the ANN with wave breaking height (), surf zone width (W), and alongshore current velocity (V), as inputs and SEDIMENT transport rate (Q) as output configures the best model and predicts more reliably, with higher correlation coefficient, R2, of 0.96, the L.S.T.R among others. Using the ANNs model presented in this research, therefore, the SEDIMENT transport rate can be estimated with sufficient accuracy.

Over the last years, ARTIFICIAL intelligence models have been widely and successfully applied in many fields. In the present study, the efficiency of adaptive neuro-fuzzy inference system (ANFIS), SVM (Support Vector Machine) and Least Squares Support Vector Machines (LS-SVM) have been investigated to estimate the SEDIMENT concentration in four gauging stations, namely Jangaldeh, Nodeh, Arazkoosh, and Gazaghly along the Gorganrood River in Golestan province, Iran. The models were defined based on the five different combinations of the river flow and precipitation using time lags from 0 to 2 previous days. The results showed that the LS-SVM model with simplex search procedure had a better performance than the grid search method. Meanwhile, the results obtained from ANFIS model which estimated SEDIMENT concentration in Jangaldeh, Nodeh, Arazkoose and Ghazaghli stations with MEF Error of 5. 3, 13. 4, 4. 8 and 2. 8 percent, respectively, suggested a higher performance than other models. Overall, at all stations except Gazaghly, considering the antecedent flow with two-day time lag as the input data of the model increased the error magnitudes. Furthermore, the rainfall of the same day and one-day time lag could only enhance the efficiency of the model at Arazkooseh station.

Accurate estimation of SEDIMENT concentrations in hydraulic SEDIMENT transport from different viewpoint such as SEDIMENT discharge estimation of river, selection of hydraulic structures and etc. are important. With respect to importance of this issue in this study for prediction of SEDIMENT concentration of Karun river multi-layer perceptron ARTIFICIAL neural network (ANN / MLP) was used. For this purpose 125 field data including bottom concentration, flow velocity, nearest distance from the beach, and the total depth of flow and flow depth was used. Three statistical metrics namely mean absolute error (MAE), root mean square error (RMSE) and coefficient of determination (R2) were used to evaluate the performance of ANN model. The result shows that MLP model with one hidden layer, Sigmoid transfer function and 5 neurons have best structure in the modeling of SEDIMENT concentration of Kroon River. The R2 and RMSE value is equal to 0. 953 and 63. 37 mg/l in training stage and 0. 752 and 203. 02 mg/l in testing stage, respectively. Finally, the sensitive analysis also showed that the nearest distance from the beach and flow depth had the most and the least effect on the SEDIMENT concentration, respectively.

This research was conducted to compare the efficiency of some simulation models including SEDIMENT rating curves and ARTIFICIAL neural networks for simulating the suspended SEDIMENT load amount. Optimized model basis of flow discharge in Shahrood watershed upon the hydrometric stations including Glinak, Baghkalaye, Loshan and Rajayi dasht was represented. In order to simulate the suspended SEDIMENT load we compared one linear rating curve and ARTIFICIAL neural network with multi-layer perceptron and radial base function models. Then performance evaluation these models was carried out by NASH and RMSE criteria. The results showed that ARTIFICIAL neural network with multi-layer perceptron method in comparison on SEDIMENT rating curve model in all of these stations simulated better models. So that ARTIFICIAL neural network with sigmoid triggering function in Glinak and Rajayi dasht stations with RMSE as 1.033 and 0.825 ton/day and NASH as 0.84 and 0.839 and this model with tansigmoid triggering function in Baghkalaye and Loshan stations with RMSE as 0.799 and 0.883 ton/day and NASH as 0.772 and 0.895, respectively, have the better efficiency for simulating of suspended SEDIMENT load amount. Also comparison of two neural network models showed that MLP model is better than RBF model for simulating of suspended SEDIMENT load amount. The only benefit of RBF networks is less time needed for training.

In water construction projects, river engineering, and irrigation and drainage engineering, it is vital to estimate the accurate volume of the SEDIMENT transported by rivers. As the SEDIMENT transport phenomenon is an immensely complex problem, therefore presenting an appropriate solution for precise evaluation of the suspended load in rivers is tedious and the mathematical models are not also accurate enough to be applied. Nowadays application of ARTIFICIAL intelligence systems has been developed as a novel solution in analysis of water resources problems. In this research, the Adaptive Neuro-Fuzzy Inference System (ANFIS) and the ARTIFICIAL Neural Networks (ANNs) models were utilized to determine suspended SEDIMENT rate of Ajichay River. Discharge, SEDIMENT load and water level data were used to prepare the models and obtain SEDIMENT rating curves. The statistical period is also divided into three seasons, namely dry, wet and snow melting. The accuracy of the models for these periods has been tested. The results showed that ANFIS neuro-fuzzy had better accuracy for determination of suspend SEDIMENT loads in comparison with both the ANNS and the rating curve.

There is a lack of information about erosion, SEDIMENT transport and SEDIMENTation in our country and usually there is a significant difference between computed and measured data. Due to this fact that the rivers always is under erosion, the study of SEDIMENT transport is very important in river hydraulic and geomorphology. SEDIMENT transport phenomenon is one of the important processes which influence many of the hydraulic and river structures, and one of the biggest problems for using water recourses in the world. In this study, ARTIFICIAL neural network was used as an effective way in order to estimate suspended SEDIMENT load in Doogh river in Golestan province. The flow discharge in present day, past day and hydrograph situation were used as input parameters, while the suspended SEDIMENT load was used as output parameter. The MLP neural network with tangent sigmoid activation function was used for training the network. The results show that the ARTIFICIAL neural networks estimate the suspended SEDIMENT load more accurately (R2=0.98, RMSE=0.015, NASH=0.97) than available method such as rating curve method with and without data classification.