基于相关性分析和神经网络的直接推力控制

传统的发动机控制是将一些可测量参数作为被控制量，如转速和压比，因为可由它们推算出推力。但由于推算的过程不准确，为保证发动机安全运行，在发动机的设计过程中常常保留了较大的裕度。而直接推力控制可以降低这些裕度，充分发挥发动机的性能。针对在飞行中推力无法直接测量的情况，本文使用了相关性分析和神经网络技术设计了非线性推力估计器，使用若干测量参数作为输入，实时计算出发动机的推力值。并针对发动机使用的实际情况，使设计出的估计器不但可以估计额定发动机的推力，而且可以估计非额定发动机的推力。在仿真结果中，估计出的推力跟踪上了正常和非额定发动机的推力。

Direct Control of Aeroengine Thrust Based on Correlation Analysis and Neural Networks

In conventional engine control system some sensed parameters, such as rotor speed and pressure ratio, are used as controlled signals. The signals can be used to calculate the real interested para-meter-the thrust. Due to this unaccuracy of the calculated thrust, large margins are built to protect aeroengines. Direct control of the thrust can reduce these margins and exploit potential performance of the engine. But the engine thrust cannot be measured when aircrafts are flying. In this paper,correlation analysis and neural networks techniques are used to design nonlinear thrust estimators, which can calculate aeroengine thrust in real time with several sensed parameters as the inputs. The estimator designed in this paper is used for estimating the thrusts of normal engine and abnormal engine. Simulation results show that the estimated thrust closely tracks normal and abnormal engine thrusts. It provides a way to reduce engine safety margin and increase the thrust-to-weight ratio.