基于BP神经网络的涡扇发动机机载模型

Turbofan Engine Onboard Model Based on BP Neural Network

针对发动机机载的工程需求,利用BP神经网络建立一种涡扇发动机机载模型。本文利用部件级模型建立起两个部件“输入参数-输出参数”数据集,并采用BP神经网络算法对数据集进行训练,建立足以替换部分旋转部件和全部尾喷管的BP神经网络。并在设计点状态对建立的机载模型进行仿真验证,结果表明,风扇部件离线神经网络模型较热力学模型节省约40%的耗时,压气机部件离线神经网络模型较热力学模型节省约50%的耗时,尾喷管部件节省耗时达70%以上。在精度比较上,离线建立的神经网络输出与基于气动热力学建立的部件级模型基本精度一致,具有一定的实际使用价值。

A turbofan engine airborne model is established using BP neural network to meet the engineering requirements of the engine.This article uses a component level model to establish two component"input parameter output parameter"datasets,and uses BP neural network algorithm to train the datasets to establish a BP neural network that can replace some rotating components and all tail nozzles.And the established airborne model was simulated and verified at the design point state.The results showed that the offline neural network model of the fan component saved about 40%of the time compared to the thermodynamic model,the offline neural network model of the compressor component saved about 50%of the time compared to the thermodynamic model,and the tail nozzle component saved over 70%of the time.In terms of accuracy comparison,the output of the offline neural network is consistent with the basic accuracy of the component level model based on aerodynamics,which has certain practical value.