基于线性遗传规划的航空发动机NOx排放开环控制

An Open-Loop Control Method of NOx Emission of Aeroengine Based on Linear Genetic Programming

关键词:  航空发动机  NO_x排放  线性遗传规划  开环控制  机器学习

Key words:  Aeroengine  NO_x emission  Linear genetic programming  Open loop control  Machine learning

摘要:

传统航空发动机减排方法主要优化发动机空间结构，这使得其结构变得很复杂，且面临回火、燃烧不稳定、宽范围适应性差等诸多问题。本文使用基于线性遗传规划在时域上调制燃料流量的方法来实现NO_x排放控制。实验基于双旋流的单头部燃烧室，搭建了开环控制实验系统；使用线性遗传规划算法对中心燃料控制律进行迭代优化；最后结合机器学习，分析了控制律的优化过程。结果表明：线性遗传规划算法通过对重要搜索路径的选取来获得最优控制律，在最优控制律下，NO_x排放量相较于无控制状态有效下降了43.1%。

Abstract:

Optimising the spacial structure is a traditional method to reduce aeroengine emission， which makes the structure complicate and leads to many problems like tempering， instable combustion and poor wide-range adaptability. Based on linear genetic programming， this paper uses a method that adjusts the fuel flow in time domain to control the emission of NO_x. The experiment is conducted in a dual-swirl cylindrical burner， and an open-loop controlled experimental system is established. Linear genetic programming is used to iteratively optimize the control law of the inner fuel. Combined with machine learning， the optimization process is analysed. The results show that linear genetic programming algorithm obtains the optimal control law by selecting major search paths， and using the optimal control law， the emission of NO_x effectively decreases by 43.1% compared with the uncontrolled status.