Prediction of Temperature Fields induced by Natural Convention in a Cylindrical Enclosure using Fuzzy LS-SVM

Abstract

Abstract-Support Vector Machines (SVM) is a machine learning algorithm basing on the statistical learning theory. In this study, SVM is used for the prediction of temperature field induced by natural convection in a cylindrical enclosure. Because of the large amount of computing and poor real-time characteristics of conventional SVM, and also to solve the non optimal problem in the whole situation and the over-fitting phenomenon, Fuzzy LS-SVM is adopted. The heat transfer in the enclosure is an unsteady process. The heat transfer process is firstly simulated with CFD software, then part of simulated data is picked for training of LS-SVM model and the rest of data is used for validation of the model. The prediction results are successfully validated from the mean relative error (MRE), max relative error (MAE), mean square error (MSE) and absolute fraction of variance (R 2 ). Besides, by comparison with artificial neural network based on back propagation (BP-ANN), the fuzzy LS-SVM shows more superior performance in both prediction accuracy and computation efficiency.