This is to certify that the thesis entitled “Prediction of flow and its Resistance in compound open channel” submitted by Mrutyunjaya Sahu, in partial fulfillment of the requirements for the award of Master of Technology in Civil Engineering from National Institute of Technology Rourkela , a bona fide investigation conducted by him under our guidance and supervision. I hereby declare that this entry is my own work and that, to the best of my knowledge and belief, it does not contain any material previously published or written by any other person, nor material accepted in substantial part for the prize from any other person . degree or diploma from university or other institution of higher education, except where due acknowledgment is given in the text.
INTRODUCTION
INTRODUCTION
- ORGANIZATION OF THE THESIS
Turbulent flow structures create complexity in predicting discharge and composite friction factor in composite open channel. Further, this analysis will be continued to predict discharge and composite friction factor in composite open channel flow.
LITERATURE REVIEW
LITERATURE REVIEW
Likewise, local flow conditions determine the rate of erosion and deposition of sediment in the main channel and floodplains. The reason for the inadequacy in the prediction of composite friction factor and discharge in composite channels may be the incorrect calculation of momentum transfer between main channel and floodplain.
NUMERICAL ANALYSIS OF TURBULENT FLOW STRUCTURES
- INTRODUCTION
- GEOMETRY SET UP AND DISCRITIZATION OF DOMAIN
- LARGE EDDY SIMULATION
- BOUNDARY CONDITIONS
- INLET AND OUTLET BOUNDARY CONDITION
- FREE-SURFACE
- SOLVER
- RESULTS
The geometry of the composite channel was created using the ANSYS 13 design modeler and is shown in Figure 3.1. SUBNETWORK SCALE MODEL: From equation (3.2) the nonlinear carrier term uiuj can represent as.
AWb wdA
VELOCITY DISTRIBUTION
It shows that the velocity head peaks in the center of the main channel and slightly in the floodplain. The peak value of the depth-averaged velocity head lies in the main channel just before the junction of the main channel and the floodplain.
BED SHEAR STRESS
Current-wise non-dimensional average secondary velocity contours From Figure 3.10 and Figure 3.11, some secondary currents can be observed on both sides of the junction of the main channel and floodplain. The mean secondary velocity contours show circulation at the main channel corner, main channel-floodplain interface, at the corner of floodplain as shown in Figure 3.11.
LATERAL MOMENTUM TRANSFER
These currents can be considered longitudinal eddies as mentioned by Tominaga and Nezu (1991). The eddy in the floodplain reaches the free surface. To investigate the effect of secondary flow on spanwise momentum transport Tominaga and Nezu (1991) integrated Eq.(3.11) over the full depth separately for the main channel and floodplain, transforming this equation into the depth-averaged momentum equation for fully developed turbulent open channel flow as. From this, it can be inferred that near the junction between the main channel and floodplain, apparent shear stress is negative.
J got an extreme positive and negative peak on both sides of the main channel and floodplain. T has a positive peak at the main channel and a highest negative peak at the floodplain and main channel junction. The transfer of momentum due to the component of secondary circulation and turbulent transport is shown in Figure 3.13.
ANALYSIS AND PREDICTION OF FLOW AND ITS RESISTANCE FACTOR
INTRODUCTION
Simulation by LES is done for a single flow depth of a composite channel as explained in Chapter 3. This helps in predicting the flow variables such as discharge, composite friction factor for a given flow depth with given flow conditions and extracting point-to-point information from the entire domain of computational space. For this situation, a researcher or river engineers working in this field needs a convenient approach, which can easily evaluate the average value of flow variables such as total discharge and total friction drag for different hydraulic conditions.
Therefore, in this present study artificial intelligence methodologies are considered to root out the efficiency involved in predicting discharge and composite friction factor for a wide range of geometric and hydraulic conditions of a composite channel flow. Among these techniques, two advanced adaptive approaches such as Back Propagation Artificial Neural Network (BPNN) and an Artificial-Neuro-Fuzzy Inference System (ANFIS) are currently selected for the present study. These approaches have been applied to predict the composite channel discharge and friction factor for variable hydraulic flow conditions.
Modeling of discharge using Back
- INTRODUCTION
- SINGLE CHANNEL METHOD
- DIVIDED CHANNEL METHOD
- COHERENCE METHOD (COHM)
- EXCHANGE DISCHARGE METHOD (EDM)
- EXPERIMENTAL SETUP AND PROCEDURE
- DEVELOPMENT OF BACK PROPAGATION NEURAL NETWORK (BPNN) 4.2.9.1 BACK PROPAGATION NEURAL NETWORK ARCHITECTURE
- SIGMOID TRANSFER FUNCTION (f)
- LEARNING OR TRAINING IN BACK PROPAGATION NEURAL NETWORK
- SOURCE OF DATA
- SELECTION OF HYDRAULIC PARAMETERS
- RESULTS
- TESTING OF BACK PROPAGATION NEURAL NETWORK
- DISSCUSION
There are several vertical partitioning methods, which are based on changing the wetted perimeter of the subarea to account for the effect of interaction. COHM is more difficult to apply when the roughness of the main channel river bed varies with discharge, as is the case in sandy bottom rivers. The factor χi calculated by equations given in Bousmar and Zech (1999) for each subsection of the flow.
After calculating χi for each subdivision by iterative procedure, it can be used in equation (4.11) to obtain overall discharge of the composite channel. The plan form of the channel, which has the straight composite channel (Type-I) with equal floodplain on both sides of the main channel, as shown in Figure 4.4. The coefficient of determination of the EDM and COHM model of Figure 4.13 and Figure 4.12 respectively shows that the two models are stationary.
MODELING OF COMPOSITE FRICTION FACTOR USING ANFIS
- INTRODUCTION
- MODELING OF ADAPTIVE-NEURO FUZZY INFERENCE SYSTEM (ANFIS)
- SELECTION OF HYDRAULIC PARAMETERS
- FUZZY LOGIC AND FUZZY INFERENCE SYSTEMS
- HYBRID LEARNING ALGORITHM
- TRAINING AND TESTING OF ANFIS NETWORK
- COMPARISION OF DIFFERENT METHODS WITH TRUE VALUE FOR PREDICTION OF COMPOSITE FRICTION FACTOR
- RESULTS OF ARTIFICIAL NEURO-FUZZY INFERENCE SYSTEM
- COEFFICIENT OF DETERMINATION (R 2 )
- DISCUSSION
The formulations proposed by various investigators are expressed in Table 4.5 to calculate composite friction factor. Therefore, in this study, a step has been taken to predict the composite friction factor in a composite open channel flow using ANFIS. The dimensionless parameters used for the present ANFIS model to estimate the composite friction factor in connecting channel are (i) relative width (Br), ratio of floodplain width (B-b) to total width (B), where B = main channel width, b = floodplain width (ii) Ratio of main channel (Pmc) perimeter to floodplain perimeter (Pfp) given as Pr, (iii) Ratio of hydraulic radius of main channel (Rmc) to floodplain (Rfp) given as Rr as usual varies with symmetry, (iv) channel longitudinal slope (S0) and (v) relative depth (Dr) ie.
Application of methods for composite friction factor prediction in Tang and Knight (2001) mobile experimental flow conditions. The distribution of predicted values of composite friction factor over all training data set and test data set are shown in Figure 4.27 and Figure 4.28 respectively. An example set of rule generation for prediction of composite friction factor is shown in Figure 4.30.
Distribution of the composite friction factor over all collected data for the training data set. Residual analysis is also performed by calculating the residuals from the actual and predicted composite friction factor for the training data set and shown in Figure 4.33.
CONCLUSION
CONCLUSION
The discharge and composite friction factor found from the LES simulations are also in good agreement with experimental results. The methods are found to give good results for some composite channels where they do not give good results for composite channels of other geometries and hydraulic conditions. Therefore, two adaptive numerical approaches such as BPNN and ANFIS have been applied to a number of global data systems to successfully predict discharge and composite friction factor to composite channels of different hydraulic conditions.
The ANFIS model is logically adaptive and incorporates the variation within sharp data where BPNN cannot. Furthermore, it is seen that the ANFIS model is computationally cheap and predicts the composite friction factor with less time than the BPNN method. The study can be further extended for predicting boundary shear stress, discharge distribution etc.
REFERNCES
In: Proceedings of the Institution of Civil Engineers, Water, Maritime and Energy, London: Thomas Telford,4.pp.247–57. Atabay S, Knight DW, Seckin G. (2004). Effect of a moving bed on boundary shear in a composite channel. Artificial Neural Network Model for Friction Factor Prediction in Pipe Flow, J Applied Science Research, 5(6), pp.662-70.
Salvetti MV, Zang Y, Street RL and Banerjee S. Large eddy simulation of decaying free surface turbulence with dynamic subgrid scale models. Tang X and Knight DW.(2001).Experimental investigation of phase-discharge relationships and sediment transport rates in a composite channel. Wormleaton PR, Allen J and Hadjipanos P. 1982), Discharge Assessment in Compound Channel Flow, J. Investigation of drag coefficient in compound channels. 2007).Current resistance and its prediction methods in composite channels, Acta Mechanic Sinica; 21, pp.353-61Yen BC.
REBUTTAL AND ERRATA
Can we learn the dominated relations between the controlling factors and discharge from the ANN set-up? Therefore, sensitive analysis of the model results with the chosen
From this, other engineers and researchers can easily find out this information and perform the process to predict the required variables. Answer: The purpose of this study is to predict discharge in composite open channel flow. Firstly, the influential variables as described by Yang et al. 2005) are taken as input of BPNN to predict the discharge.
From this, other engineers and researchers can easily find out this information and perform the process to predict the required variables. The aim of this study is to predict the discharge of compound open channels with accuracy, which was done efficiently. For this reason, an ANFIS has been developed to predict the roughness coefficient (n) of the crew so that the discharge can be predicted using Manning's equation.
ERRATA
1 RMSE and mean analysis of BPNN network and earlier developed models Current data source. To find the robustness of the proposed model, sensitivity analysis is performed. The inputs in the test samples vary one by one systematically by ±10% from its base value while keeping the other items at their original values.
The scaled change in output is calculated with the current input increased by 10% and the current input reduced by 10%. The input parameters of the BPNN network are: (i) influence of flood plain and main channel roughness (fr); (ii) Ratio of area of floodplain to main channel (Ar); (iii) relative depth (Hr), i.e. Sensitivity analysis is inevitable to verify the performance of the proposed ANFIS model, as well as to testify its performance.
Mean
Therefore, the same procedure as discussed above for the BPNN model is also followed for this case. The input parameters for the ANFIS network are: (i) relative width (Br), ratio of floodplain width (B-b) to total width (B), where B = main channel width, b = floodplain width; (ii) The ratio of the perimeter of. Ratio of floodplain depth (H-h) to total depth (H), where H = depth of main channel, h = depth of floodplain.
After this analysis, the output is averaged by changing each input sequentially for the model. It can be seen from Figure 6 and Figure 7 that by increasing and decreasing input 1 by 10%, the error shows an uneven trend, rather the error increases by increasing the input value. Therefore, input 5 (Dr) and input 2 (Pr) are more sensitive to changes and amplify the effect with the changes.
Input changed by 10%
Input decreased by 10%
BRIEF BIODATA OF THE AUTHOR
