PREDICTION OF FRICTION FACTOR IN THE TURBULENT FLOW REGIME Artificial Neural Network (ANN) is an intelligent data-driven modeling tool that can capture and represent complex and non- linear input/output relationships. They can continuously improve their performance through dynamic learning o training. ANN can be employed as a function of approximation, pattern recognition and classification, prediction, memory recall, optimization, and noise filtering The artificial neural network (ANN) provides a viable solution to the problem of predicting the friction factor in pipe flow at very high speed because it is based on training not on the analytical and statistical assumptions. STATEMENT OF THE PROBLEM
Most previous works did not provide any
applicable equation or function that can be applied in the prediction of the friction factor even without access to Artificial Neural Network software. Therefore the need to investigate the inner workings of the ANN and also not only predicting the friction factor of the turbulent flow regime but also extract a functional model from the network which will be carried out in this project AIMS AND OBJECTIVES
The objective of the work to be carried
out in this project is to develop a functional model which will be extracted from the Artificial Neural Network for the prediction of friction factor in the turbulent flow regime. SIGNIFICANCE OF THIS STUDY It is difficult to determine a good network topology only from several inputs and outputs. The fact is that a suitable number of hidden layers can reduce training time with very high accuracy. This project will draw attention to a simple ANN-based functional model which can be used for the prediction of the friction factor in the turbulent flow regime. SCOPE OF THE STUDY
The scope of this project will be
a simulation-scale study using MATLAB as the ANN toolbox. LITERATURE REVIEW Over the years, several configurations of ANN have been developed and tested to determine the most suitable network to be used in predicting friction factor. The tested ANN formats included single and double hidden layers. Program file with code written and implemented in MATLAB to create, train, validate and test the networks. Most of the previous works simply demonstrated the ability of ANN to match experimental output data and did not provide any applicable model or function that can be applied by interested researchers for friction factor prediction. This project is unique by not only predicting the experimental output data, but will also provide detailed stepwise procedure for developing a functional model for friction factor prediction in the turbulent flow regime MATERIALS AND METHODS SOURCE OF DATA
The data to be used for this project will be
generated using MATLAB by using random datasets. The data will be collected from previous indexes and will be normalized and divided into two groups, where one will be used for training and the other for testing and validating the model independently. PROCESSING OF DATA
The inputs of the neural network are
the data that will be adjusted to be able to correctly predict the friction factor of the turbulent flow regime. MATLAB will be used to create, train, validate and test the networks as well as to generate extract a functional model for the network. TRAINING THE NETWORK The Artificial Neural network will be trained by feeding in training dataset and letting it change its weights according to the learning rule such as using back- propagation training algorithms which will help to determine the structure of the model to be extracted TESTING OF THE FUNCTIONAL MODEL
The test data will be provided to the
trained network to obtain the model to be used for prediction The results will be compared with the desired output to get the test accuracy of the previous network prediction.