基于改进ConvNeXt模型的压气机变几何系统T-step预测方法  

T-step Prediction Method for Compressor Variable Geometry System Based on Enhanced ConvNeXt Model

作　　者：旷典 詹于游 谭燕[1] KUANG Dian;ZHAN Yu-you;TAN Yan(Engineering Techniques Training Center,Civil Aviation Flight University of China,Guanghan Sichuan 618300,China)

机构地区：[1]中国民用航空飞行学院工程技术训练中心,四川广汉618300

出　　处：《航空发动机》2023年第6期19-26,共8页Aeroengine

基　　金：中央高校基本科研业务费专项基金(J2022-014);中国民航飞行学院科研基金(青年基金)(Q2019-056)资助。

摘　　要：为了实时监控航空发动机压气机变几何系统的状态并获取警告信号,提出一种基于改进ConvNeXt模型的T步(T-step)预测方法。与仿真数据和特定试验条件下生成的数据集相比,T-step预测方法采用了飞机数据采集系统记录的实际飞行数据。证实了采用改进ConvNeXt模型预测压气机变几何系统参数的可行性,并在发动机过渡状态和稳态下分别进行了试验验证。结果表明:采用改进ConvNeXt模型的T步(T-step)预测方法能精准地预测压气机VSV角度和VBV开度的变化,最低可达2.132°和7.077°,预测误差在可接受范围内。该方法能识别和预测各类型航空发动机不同运行状态的变几何系统参数的角度,获得相对准确的结果。In order to monitor the status of the compressor variable geometry system and acquire warning signs in real-time,a T-step prediction method based on an enhanced ConvNeXt model was proposed.Compared to simulation data and the datasets generated in specif-ic lab conditions,the T-step prediction method adopts the actual flight data recorded by aircraft data acquisition systems.The feasibility of using the enhanced ConvNeXt model to predict the compressor variable geometry system parameters was demonstrated,and experimental verification was conducted under the transient state and the steady state respectively.The results show that the T-step prediction method based on the enhanced ConvNeXt model can accurately predict the changes in compressor VSV angle and VBV opening,with the lowest prediction errors of 2.132ºand 7.077º,respectively,which are within acceptable ranges.The method is applicable to different types of aero-engines and can identify and predict the angles of variable geometry system parameters under different operating states,and obtain relative-ly accurate results.

关 键 词：压气机变几何系统 改进卷积神经网络模型 预测方法 航空发动机 

分 类 号：V239[航空宇航科学与技术—航空宇航推进理论与工程]