基于NARX网络的航空发动机性能参数预测优化

Prediction method of aeroengine performance parameters based on optimized NARX network

航空发动机的构造复杂性使得发动机工况监测成为一项困难的任务。为了准确预测航空发动机运行时传感器数据的变化趋势,提出一种基于粒子群算法优化的NARX神经网络预测模型。通过准确预测航空发动机输出的传感器数据,达到有效监测航空发动机工况状态的目的。实验结果显示,预测模型所得性能参数——发动机工作站温度、高压转子转速、发动机工作站压力值和低压转子转速的均方误差分别为0.006 52,0.005 25,0.009 3,0.009 12。结果表明,所提出的基于粒子群算法优化NARX网络能够有效预测发动机性能参数,相较于传统NARX网络、传统BP神经网络和粒子群算法优化的BP神经网络,在预测准确度上有较大优势,为基于飞参数据进行发动机健康管理与监测提供了良好的数据支持。

The structural complexity of the aeroengine makes the engine condition monitoring a difficult task.This paper presents a NARX neural network prediction model based on particle swarm optimization to accurately predict the change trend of sensor data during aeroengine operation.By accurately predicting the sensor data of the aero-engine output,the working condition of the aero-engine can be effectively monitored.The results show that the mean square errors of the performance parameters obtained from the prediction model are 0.00652,0.00525,0.0093 and 0.00912,respectively.The results show that the optimized NARX network based on particle swarm algorithm can effectively predict engine performance parameters and has greater prediction accuracy than the traditional NARX network,the traditional BP network and the optimized BP network based on particle swarm algorithm,which provides good data support for engine health management and monitoring based on flight parameter data.