基于NARX神经网络的试车台压缩机模型辨识方法研究  

Identification method of compressor model for test facility based on NARX neural network

作　　者：彭冲 李楠 王信 王同庆 宋野 白长青[1] PENG Chong;LI Nan;WANG Xin;WANG Tong-qing;SONG Ye;BAI Chang-qing(State Key Laboratory for Strength and Vibration of Mechanical Structures,Shaanxi Key Laboratory of Environment and Control for Flight Vehicle,School of Aerospace Engineering,Xi’an Jiaotong University,Xi’an 710049,China;AECC Sichuan Gas Turbine Establishment,Mianyang 621703,China)

机构地区：[1]西安交通大学航天航空学院机械结构强度与振动国家重点实验室陕西省先进飞行器服役环境与控制重点实验室,西安710049 [2]中国航发四川燃气涡轮研究院,四川绵阳621703

出　　处：《燃气涡轮试验与研究》2022年第5期57-62,共6页Gas Turbine Experiment and Research

基　　金：中国航发四川燃气涡轮研究院稳定支持项目(GJCZ-0013-19)。

摘　　要：针对试车台供气抽气系统压缩机工作范围广、动态特性预测困难的问题,提出了基于NARX神经网络的试车台压缩机模型辨识方法。在对供气抽气系统结构和运行原理分析的基础上,将压缩机及调节阀作为整体进行建模,研究系统各参数变化对压缩机流量的影响机制。为改善NARX神经网络的辨识精度,引入Gamma Test方法确定网络的最佳时延阶次。基于所建立的神经网络模型,预测了试车台供气、抽气压缩机多工况切换过程的实时流量,最大相对误差分别为0.53%和0.59%。研究所提出的方法可准确反映试车台压缩机流量的动态特性,为实现试车台高精度流量控制提供了技术基础。In order to solve the problems of wide working range and difficult dynamic characteristics prediction of compressor for test facility,an identification method of compressor model for test facility based on NARX neural network was proposed.Based on the analysis of the structure and operation principle of the air supply and extraction system,the whole modeling of the compressor and the regulating valve was carried out to study the influence mechanism of system parameters on the flow rate of the compressor.To improve the identification accuracy of NARX neural network,the Gamma Test method was introduced to determine the optimal time delay order of the network.Based on the established neural network model,the real-time flow rate of the air supply compressor and the air extraction compressor was predicted during the multi-condition switching process,and the maximum relative error was 0.53%and 0.59%,respectively.The method proposed can accurately reflect the dynamic characteristics of the flow rate of the air compressor,and provide a technical basis for the realization of high precision flow control of the air compressor.

关 键 词：试车台 压缩机 NARX神经网络 流量特性 动态模型 GammaTest方法 

分 类 号：V263.47[航空宇航科学与技术—航空宇航制造工程] TP271[自动化与计算机技术—检测技术与自动化装置]