摘要
为实现航空发动机飞行试验实时监控,分析整理了涡扇发动机实际飞行试验数据,并以三层前向人工神经网络为基础,通过引入输出层反馈至输入层,形成该涡扇发动机的NNARX模型。对包括高压转子转速在内的11个发动机关键参数变化模型进行研究,并在额外架次全程飞行试验数据上验证和讨论辨识模型的推广能力。结果表明,辨识模型样本点上最大相对误差在5%以内,辨识模型可以应用到该型发动机的试飞实时监控中,同时也可为后续建立涡扇发动机的全包线自适应实时监控模型提供参考。
To realize the real-time monitoring of flight test of aero-engine, the actual turbofan engine flight test data has been analyzed, and based on three-layer feedforward artificial neural network, which has been revised so that there is a backforward connection between output layer and input layer, finally the dynamic NNARX model of two-spool turbofan engine has been identified. During the identification of NNARX mod- el, about 11 key parameters of the engine, which includes high pressure spool speed, have been studied. The identified model was implemented on totally new flight data, which has been used as non-samples in order to test the model, and the precision of identified model in new data has been discussed. It has been shown that good consistency has been achieved in both real flight samples and non-samples, and identified method and results can be used in such turbofan engine flight test to monitor the engine state. It could be a good reference for building real-time and adaptive model of turbofan engine in global flight envelope.
出处
《燃气涡轮试验与研究》
北大核心
2016年第6期21-25,共5页
Gas Turbine Experiment and Research
关键词
航空发动机
飞行试验
人工神经网络
NARX模型辨识
全飞行包线
趋势监控
健康管理
aero-engine
flight test
artificial neural network （ANN）
NARX model identification
global flight envelope
trend monitoring
health management
