Prediction Of Flow Boiling Heat Transfer Coefficient By Artificial Neural Networks And Genetic Algorithms

The flow boiling heat transfer studies the fluid inside the pipe which contains nuclear boiling and convection boiling.Artificial Neural Network(ANN)is a simplified model which abstracts the human brain neural network mechanism of information processing using Method of Mathematical Physics.Genetic Algorithms(GA)is a search algorithm based on biological evolution model which has the advantage of global searching.The Genetic Neural Network(GNN)takes the advantage of the parallelism which belongs to GA and can help us maximize the interest with less calculation.Because of the global searching,its convergence speed is high and the search time is shortened.Due to the complexity of the heat transfer mechanism,artificial neural networks and genetic algorithms have been widely used to predict CHF and other issues.The heat transfer problem of nuclear reactors has been the focus of attention.The flow boiling heat transfer process in the tube is complicated.The neural network has the advantages of less calculation and faster calculation speed than other algorithms.In this study,the experiment data from literature is used to train the ANN and the GA to fit the expression of flow boiling heat transfer.Comparatively,the root mean square error of the GNN is smaller,which means that its precision is better.The error of correlation generated by GA is larger than those of the two Neural Networks mentioned above.Additionally,we use the trained ANN to study the effect of the pressure,the vapor quality,the mass flux,the heat flux,the inside diameter and the ratio of area on the heat transfer coefficient from nuclear boiling and convection boiling.Research results show that the heat transfer coefficient increases with the increase of pressure,the mass flux,the heat flux,the vapor quality and the ratio of area,and decreases with the increase of the inside diameter of the tube.