摘要
针对传统航空发动机状态预测需要大量经验知识,预测过程单一的缺陷,提出一种基于深度稀疏受限玻尔兹曼机网络(DSRBM)的航空发动机状态预测方法。利用深层神经网络中的DSRBM网络对高维复杂类型的发动机数据进行特征提取和降维。实验结果证明,DSRBM相较于传统的主成分分析法(PCA),降维数据辨识度高,挖掘了数据之间的隐藏联系。将提出的算法与支持向量机(SVM)算法结合对发动机状态进行预测,与经典算法进行比较,预测准确率达到87%以上,比经典算法提高3%,是对传统航空发动机状态预测方法的一个很好的补充。
Targeting the traditional aero-engine state prediction needs a lot of experience knowledge and the single process of state prediction,a method of aero-engine state prediction based on deep sparse restricted Boltzmann network(DSRBM)is proposed.The DSRBM network of deep neural network is used to extract and reduce the dimension of high-dimensional complex engine data.Experimental results show that DSRBM has a higher identification of dimensionality reduction data than the traditional principal component analysis(PCA),and it can mine hidden links between data.Combining the proposed algorithm with support vector ma⁃chine(SVM)algorithm to predict the engine state,compared with the classical algorithm,the prediction accuracy is more than 87%,it is 3%higher than the classical algorithm,which is a good supplement to the traditional engine state prediction method.
作者
鲍洋
芮国胜
张嵩
董道广
BAO Yang;RUI Guosheng;ZHANG Song;DONG Daoguang(Naval Aviation University,Yantai 264001)
出处
《舰船电子工程》
2021年第5期113-118,共6页
Ship Electronic Engineering
基金
国家自然科学基金项目(编号:41606117,41476089,6161016)资助。
关键词
状态预测
数据降维
深度学习
DSRBM网络
特征提取
航空发动机
state prediction
data dimensionality reduction
deep learning
DSRBM network
feature extraction
aero-engine
