论文部分内容阅读
气膜冷却作为当代燃机高温透平中必需的冷却手段,其冷却性能在多种参数的影响下表现复杂。采用BP神经网络模型对多种几何、流动参数变化下的气膜冷却系统的绝热气膜冷却效率进行预测。选择气膜冷却系统的吹风比、密度比、主流湍流度、面积比和长径比作为神经网络的输入参数,以燃气轮机透平叶片气膜冷却的实际运行工况为范围建立数据库。计算结果表明,采用贝叶斯归一化法训练后建立的气膜冷却神经网络模型在预测精度上要优于经验公式法,而且参数适用范围更广,具有良好的发展应用前景。
As the necessary cooling method in the high temperature turbine of modern gas turbine, the cooling performance of gas film is complicated under the influence of many parameters. BP neural network model was used to predict the adiabatic film cooling efficiency of gas film cooling system under various geometrical and flow parameters. The blowing ratio, density ratio, mainstream turbulence ratio, area ratio and aspect ratio of the film cooling system are selected as the input parameters of the neural network, and the database is established based on the actual operating conditions of gas turbine turbine blade cooling. The calculation results show that the model of the film cooling neural network established by Bayesian normalization method is superior to the empirical formula method in prediction accuracy, and has a wider range of parameters and a good development and application prospect.