摘要 为建立适用于畸变容限控制技术的航空发动机进气总压畸变估算方法,采用基于神经网络方法,以飞机试飞测量数据为基础,分别开展了发动机进口总压流场和稳态周向畸变指数重构方法研究。以壁面静压、发动机进口面中心点总压以及攻角、侧滑角数据分别构建了三种不同输入参数组合的神经网络模型,并对比分析了各模型的重构精度和适用性。结果表明:采用神经网络方法可以较好地建立壁面静压、飞行参数与发动机进口总压分布和稳态周向畸变指数的映射模型;以壁面静压、发动机进口面中心总压、攻角和侧滑角为输入的神经网络模型,对于总压流场重构和稳态周向畸变指数重构都具有最佳的重构精度;采用先重构流场总压分布再计算稳态周向畸变指数的间接方法相对于直接重构稳态周向畸变指数的方法具有更高的重构准确性;对于与建模飞机具有相同进气道结构和布局的飞机,训练完成的神经网络模型对总压流场重构具有一定的适应性,对稳态周向畸变指数重构适用性差。 In order to establish an aero-engine inlet total pressure distortion estimating method for the ap⁃plication of engine distortion tolerance control,total pressure field and steady state circumferential total pressure distortion index reconstruction were studied respectively based on the flight test data and neural network method.By using different combination of the aero-engine inlet wall static pressure and center to⁃tal pressure and angle of attack,sideslip angle as input parameters,three neural networks were established.And the reconstruction accuracy and feasibility of three neural networks were analyzed and compared.The results show that neural network method could satisfactorily establish relating models between wall static pressure,flight parameters and total pressure field,circumferential total pressure distortion index on engine inlet face.The neural network with static pressure,center total pressure on the engine inlet face,angle of at⁃tack and sideslip angle as inputs had higher accuracy on both reconstruction of total pressure field and dis⁃tortion index than other models.The distortion index reconstruction accuracy of the indirect method which first constructing total pressure field and then calculating distortion index was better than that of the direct method which constructing distortion index directly by neural network.The trained neural network had de⁃graded reconstruction accuracy for the airplane having the same inlet type and layout with the modeling air⁃plane,but was not suitable for distortion index reconstruction.
机构地区 中国飞行试验研究院
出处 《燃气涡轮试验与研究》 北大核心 2021年第1期21-26,33,共7页 Gas Turbine Experiment and Research
关键词 航空发动机 神经网络方法 总压畸变 壁面静压 稳态周向畸变指数 畸变重构 aero-engine neural network method total pressure distortion wall static pressure steady state circumferential total pressure distortion index distortion reconstruction
