基于BP神经网络的多参数气膜冷却效率研究

Prediction of the Adiabatic Film Cooling Effectiveness Influnenced by Multi Parameters Based on BP Neural Network

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。

Film cooling is necessary for modern gas turbine.Its cooling effectiveness is sophisticated influenced by multi parameters.The BP neural network is applied to predict the adiabatic film cooling effectiveness of the cooling system with multi geometry and flow parameters.The input parameters of neural network are chosen as blowing ratio,density ratio,free stream turbulence intensity,area ratio and length ratio.A database covering the real operation range is build up.Prediction from the neural network trained by Bayesian Regulation backpropagation is compared to an existing correlation.The result shows a good accuracy and wide application range of the neural network model.It implicates that the developed model is promising to be applied on the film cooling system.