轴流压气机的正问题分析过程中,传统经验模型对新叶型预测效果差,且其建立过程依赖风洞实验,更新难度大。为解决这个问题,根据以往积累的叶型仿真计算结果和叶栅、部件级实验中测得的流场参数,采用BP神经网络模型替代了轴流压气机中预测落后角和总压损失系数常用的经验模型,在原有的计算程序的基础上改进得到了新的压气机正问题分析程序,并对轴流压气机的正问题进行了求解分析。为提高神经网络模型的预测精度,利用一定范围内的流场参数作为数据集,训练出了从基本参数到落后角和总压损失系数的BP神经网络,并将预测相对误差控制在6%和2%之内。针对两台轴流压气机算例进行了计算校验,将该模型替代传统经验模型之后,用模型预测结果对比实验和替代之前的计算结果,发现该方法有效提高了正问题分析精度。

In the process of positive problem analysis of axial compressor, the traditional empirical model has poor prediction effect on the new airfoil. Besides, to update the model is much difficult for its establishment process depends on huge number of wind tunnel experiments. To solve this problem, according to the results of the previous accumulation of blade shape simulation calculations and the flow field parameters measured in the cascades grid and component-level experiments, the BP neural network model is adopted to replace the empirical model commonly used to predict the deviation angle and total pressure loss coefficient in axial compressor, and a new compressor positive problem analysis program is obtained based on the original calculation program which is used to solve the positive problem of axial compressor. In order to improve the prediction accuracy of the neural network model, the BP neural network was trained from the basic parameters to the deviation angle and the total pressure loss coefficient by using the flow field parameters in a certain range as data sets, which controls the relative error of predictions within 6% and 2%. Computational verification was performed for two axial compressor examples. After the model was substituted for the traditional empirical model, the model prediction results were compared with the experimental results and those before the substitution, and the accuracy of the positive problem analysis was improved effectively by the method.

V231.3

航空发动机及燃气轮机基础科学中心项目（P2022-B-V-004-001)，国家科技重大专项(J2019-II-0004-0024).