计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (12): 241-248.DOI: 10.3778/j.issn.1002-8331.1601-0268

• 工程与应用 • 上一篇    下一篇

基于NARX神经网络航空发动机参数动态辨识模型

耿  宏,任道先,杜  鹏   

  1. 中国民航大学 电子信息与自动化学院,天津 300300
  • 出版日期:2017-06-15 发布日期:2017-07-04

Dynamic parameter identification model of aircraft engine based on NARX neural network

GENG Hong, REN Daoxian, DU Peng   

  1. College of Electronic Information and Automation, Civil Aviation University of China, Tianjin 300300, China
  • Online:2017-06-15 Published:2017-07-04

摘要: 针对航空发动机参数非线性动态特性,提出一种基于外部输入非线性自回归(NARX)神经网络的发动机参数动态辨识模型。主要思路是根据NARX网络的非线性时序预测特性,结合发动机参数的稳态和动态参数,提出一种基于偏稳态差值预测的NARX参数动态模型结构。设计了SP-P辨识结构,整定了模型内部结构参数并建立N1(低压转子转速)、N2(高压转子转速)、EGT(涡轮后排气温度)参数非线性差分预测模型。最后依据某发动机试车样本,对推杆加减速时N1、N2、EGT动态辨模型进行仿真。仿真结果表明,N2相对误差小于0.2%,N1相对误差小于0.3%,EGT相对误差小于[1℃],满足发动机试车仿真需要。最后,将所建模型应用于某A320机务维修训练器的发动机仿真系统。

关键词: 航空发动机, 动态模型, 非线性系统辨识, NARX网络

Abstract: According to the nonlinear dynamic characteristics of aero engine, a dynamic identification model of engine parameters based on NARX neural network is presented. The main idea is based on the nonlinear time series prediction characteristics of NARX network, combined with the steady state and dynamic parameters of the engine parameters, a dynamic model structure of NARX parameters is proposed, which is based on the difference between the partial and steady state. The structure of SP-P identification is designed, and the internal structure parameters of the model are established, and the N1(low rotor speed), N2(high rotor speed), and EGT(the temperature at the exit of high-pressure turbine) parameters are established. On the basis of an engine test samples, simulating the accelerating and decelerating N1, N2, EGT dynamic identification models. The simulation results show that the relative error of N2 is less than 0.2%, the relative error of N1 is less than 0.3%, and the relative error of EGT is less than [1℃], which meets the requirements of engine test simulation. Finally, the engine simulation system of the model is applied to the A320 maintenance training device.

Key words: aircraft engine, dynamic model, nonlinear system identification, NARX network