摘要
为解决分段线性化机载模型精度不足的问题,提出并设计了基于稀疏自动编码器的大包线、具有10输入11输出的发动机机载自适应模型,该模型由稳态、动态两部分组合而成。首先基于一种新的相似准则进行建模所需样本数据的压缩,在保留主要信息的同时,大大降低了数据量及采样时间。用BP算法对简化后的样本数据进行了机载模型稳态部分的建模。针对机载模型动态部分所需样本数据量巨大、BP算法难以训练的问题,建立了基于稀疏自动编码器的动态机载模型。引入准稳态判断逻辑,在动态过程使用稀疏自动编码器的动态机载模型,在稳态过程使用基于BP算法的稳态机载模型。仿真结果表明,所建立的发动机机载模型具有优良的动稳态精度,且实时性好、存储量小,其中动态精度小于1%,稳态精度小于0.6%,一次模型计算时间不大于1ms,模型存储量不大于100kB。
In order to solve the problem of the low accuracy of the piecewise linear model in the develop-ment of the on-board engine model,based on the sparse auto-encoder,an adaptive on-board engine modelwith 10 inputs 11 outputs for the large envelope is proposed and designed,the model consists of steady and dy-namic two parts.In the first place,a new similarity criterion is needed to compress the sample data,which canreduce the amount of data and the sampling time while retaining the main information.Steady on-board enginemodeling work is completed by the BP algorithm with the simplified training data.In view of the huge amount ofdata needed in dynamic modeling,the BP algorithm is difficult to train.Dynamic on-board model is establishedbased on the sparse auto-encoder.By the introduction of quasi steady state judgment logic,in the dynamic pro-cess the dynamic on-board model based on the sparse auto-encoder is used,while in the steady state processthe steady on-board model based on the BP algorithm is used.Simulation results show that the on-board modelobtained has excellent dynamic and steady state accuracy,good real-time performance and a small amount ofstorage.The dynamic accuracy is within 1%,the steady accuracy is within 0.6%,model computation time iswithin 1ms once with the storage capacity no more than 100 k B.
出处
《推进技术》
EI
CAS
CSCD
北大核心
2017年第6期1209-1217,共9页
Journal of Propulsion Technology
基金
国家自然科学基金(51576096)
江苏省“青蓝工程”
“333”人才工程
