航空发动机推力衰退缓解的神经网络控制

针对航空发动机气路部件性能退化导致的推力下降问题,提出一种基于变增量线性规划(LP)优化 神经网络控制方法用于航空发动机推力衰退缓解控制。该方法通过内环控制转速和发动机压比,外环修正发动机指令信号以缓解发动机推力衰退。其中内环非线性自回归滑动平均(NARMA-L2)转速控制器由神经网络训练得到;外环指令修正回路利用变增量LP优化方法调整发动机指令信号。以某型小涵道比涡扇发动机为对象进行仿真验证,结果表明,在4组仿真条件下,设计的控制方法在保证性能退化的发动机不超限的条件下使推力衰退至少缓解了46.5%,验证了该方法的有效性。

Neural network control of aircraft engine thrust ,degradation mitigation

A thrust degradation mitigation neural network control method of aircraft engine was proposed based on variable incremental linear programming (LP) optimization, due to performance deterioration of the gas path components. This method alleviated the engine thrust degeneration through control of high-pressure rotor speed and the engine pressure ratio by the inner loop, and correction of the engine command signal by the outer loop. The inner loop nonlinear autoregressive moving average (NARMA-L2) speed controller was obtained by neural network, and the outer command correction loop used the variable incremental LP optimization method to adjust the engine command signal. Simulations on a low-bypass-ratio-turbofan engine were performed. Results showed that under the 4 sets of simulation conditions, the designed control method can mitigate the thrust at least 46.5%, ensuring that the engine with performance deterioration was not overrun. The effectiveness of the method has been verified.