一种基于CNN-LSTM深度神经网络的智能发动机模型设计方法

打开文本图片集

中图分类号：TP183 文献标识码：A文章编号：1006-8228（2026）02-57-04

Abstract：ThispaperproposesaninteligentmodeldesignmethodforaireraftenginesbasedonaCNN-LSTMhybridmodel.The methodefectivelyintegratesthepowerfulspatialfeatureextractioncapabilityofConvolutionalNeuralNetworks（CNN）andthe superiortemporalmodelingcapabilityofLongShort-TermMemorynetworks（LSTM），providinganeficientsolutionforhandling multidimensionaltimeseriespredictionproblemsinherentinaireraftengineoperation.Byreasonablyconfiguringmultipleinput featuresadmultipleoutputtargets，andconductingexperimentsonapreprocesseddataset，theresultsofthisstudydeostrate thatcomparedtotradionalaero-enginemodeldesignmethods，thismethodnotonlysignificantlyshortensthemodeldesigncycle butalsoachievessubstantiallyhighersimulationacuracyoftheconstructedmodelinthecontrolofdynamicandsteady-state errors，exhibiting remarkable advantages.

Keywords：InteligentModelDesignMethod；ConvolutionalNeuralNetwork（CNN）；LongShort-TermMemory（LSTM）；Time Series Prediction

0引言

航空发动机模型在工程实践中扮演着双重角色，既可用于地面仿真测试，也可服务于发动机的辅助控制策略。（剩余5473字）