嵌入式分布式架构下航空发动机模型预测控制方法研究

Research and Implementation of Model Predictive Control for Aircraft Engines with Embedded Distributed Architecture

关键词:  涡喷发动机  发动机控制  分布式控制  模型预测控制  硬件在环仿真

Key words:  Turbojet engine  Engine control  Distributed control  Model predictive control  Hardware-in-the-loop simulation

摘要:

传统集中式架构和离线设计控制器的方法已经很难满足航空发动机复杂多变、全生命周期控制需求。本文提出了一种部分分布式架构的涡喷发动机模型预测控制方法，并进行了鲁棒性分析：采用改进的组合线性模型求解方法，在保证精度的前提下大幅缩短求解时间，解决了发动机多变量线性状态空间模型求解问题；设计了基于线性模型预测控制的涡喷发动机控制器，解决了含有约束条件下发动机动态性能优化问题；考虑到分布式控制系统具有总线网络通信特性，建立了以网络工具箱模拟总线网络的分布式控制系统仿真平台，解决了存在总线网络丢包情况下的分布式控制系统鲁棒性分析问题；采用多节点模块化设计方法，设计并搭建了分布式控制系统硬件在环仿真平台并采用嵌入式矩阵运算优化方法，解决了模型预测控制算法在嵌入式平台上的应用问题。试验表明：模型预测控制的效果优于传统PID控制效果，有效地提高了发动机的动态响应；同时，在存在网络丢包情况下，本文所设计的基于模型预测控制的分布式控制系统依然具有稳定的控制效果和鲁棒性。

Abstract:

It is arduous for the approach to design a control offline with the traditional centralized architecture to satisfy the complex and variable control requirements of an aero-engine during its entire life cycle. Model predictive control （MPC）for turbojet engines with partially distributed architecture is proposed in this paper， and its robustness is analyzed herein： firstly， an improved method is adopted to solve linear mixed models（LMMs）， significantly reducing the time of model solving with guaranteed accuracy， thus addressing issues concerning the solution to the multivariable linear state-space model of an engine； secondly， a turbojet engine controller based on linear MPC is designed to improve the dynamic performance of an engine under constraint conditions；thirdly， a simulation platform of the distributed control system（DCS） is established with a network toolbox simulating the bus network， given the system features communication characteristics of the bus network， thus solving the problem of robustness analysis of the system in the case that packet loss occurs in the bus network；fourthly， a DCS hardware-in-the-loop（HIL） simulation platform is designed and built by using a multi-node modular design. In addition， an optimization method of embedding matrix operations is employed to enable the application of the MPC algorithm on embedded platforms. The experiments unveiled that the effect of MPC is better than that of the traditional PID control. The former effectively improves the dynamic response of an engine， meanwhile， in the presence of network packet loss， the DCS based on MPC contrived in this paper still has a stable control effect and robustness.