Application Of Neural Networks And Genetic Algorithms In Inverse Heat Conduction Problem

The research to the IHCP that extensively exist in the most areas of the engineering has very important worth. In this paper, the history of IHCP is reviewed, and a classification to the IHCP is proposed. A introduction to the IHCP theory and some methods of solving the IHCP are presentedIn this paper, because of the shortcoming of the solution methods in the past, the ANN and the GA are applied to solving the IHCP. The comprehensive research of solving IHCP with ANN and GA is developed, and the solution models are presented, and some design methods and solution rules are proposed, which are the innovations of this paper.The one-dimension IHCP was solved with BP networks, RBF networks and GA, and the two-dimensions IHCP was solved with BP networks and GA, and the solution methods presented in this paper were tested with the IHCP to that the direct problem has analytical solution opposite. The results of simulations show that :1) Both of the ANN and the GA can get the good results to the one-dimension IHCP ,2)the two methods also can get the good results to the two-dimensions design IHCP, 3) the two methods have poorer anti-noise performance to the two-dimensions multi-varies IHCP. 4) the GA has good anti-noise performance to the two-dimensions unique-varies IHCP, nor has the ANN. The results got in this paper have higher use value.