【作者】 智会强；

【导师】 王秀春；

【作者基本信息】 河北工业大学 ， 热能工程， 2004， 硕士

【摘要】 导热反问题(Inverse Heat Conduction Problem，IHCP)广泛存在于工程中的各个领域，对它的研究有着十分重要的意义。本文回顾了导热反问题的发展过程，对其做了一个比较合理的分类。同时，对导热反问题理论做了简要介绍，并提供了一些常用求解方法。 鉴于导热反问题传统求解方法存在的缺陷，本文将近年来得到飞速发展和广泛应用的人工神经网络(Artificial Neural networks，ANN)和遗传算法(Genetic Algorithms，GA)引入了导热反问题的求解中，对用ANN和GA求解导热反问题做了比较深入的研究，建立了相应的求解模型，提出了ANN和GA用于求解导热反问题时的一些设计方法和求解规律，这些正是本课题的创新之处。 本文分别用BP网络、RBF网络和GA求解了一维导热反问题，分别用BP网络和GA求解了二维导热反问题，并用正问题具有解析解的导热反问题对所提出的算法进行了验证。求解结果表明：1)ANN和GA对一维导热反问题均能得到满意的结果，2)对二维设计型导热反问题，两者也能得到高精度的解；对二维多宗量导热反问题，两者的抗噪声性能都比较差，3)对二维单宗量导热反问题，GA的抗噪声性能可以达到要求，ANN则性能较差。对工程实际来说，本文的结果具有较大的应用价值。更多还原

【Abstract】 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.更多还原