摘要
针对传统反向传播(BP)神经网络和支持向量机(SVM)存在的过拟合、维数灾难、参数选择困难等问题,提出了一种基于深度学习算法的航空发动机传感器故障检测方法。对发动机参数记录仪采集的多维数据进行预处理,建立基于深度置信网络(DBN)的故障检测模型,利用预处理后的数据对检测模型进行训练,经过DBN故障检测模型逐层特征学习实现了传感器故障检测。仿真结果表明:在无人工特征提取和人工特征提取的情况下,基于DBN故障检测的准确率均高于BP神经网络和SVM模型。
Aiming at the problems of traditional back propogation（ BP） neural network and support vector machine（ SVM） learning algorithm,such as over fitting,dimension disaster and difficulty of parameter selection,put forward an aircraft engine sensor fault detection method learning algorithm based on deep learning algorithm.Preprocess the multi dimensional data acquired by aero-engine parameter recorder; fault detection model based on the deep belief network（ DBN） is set up; fault detection model is trained using proprocessed data,after DBN fault detection model characteristics learning layer by layer,sensor fault detection is realized. It is shown from the simulation results,in the absence of artificial feature extraction and feature extraction,accuracy based on DBN fault detection is higher than that of BP neural network and SVM model.
作者
刘云龙
谢寿生
郑晓飞
边涛
LIU Yun-long XIE Shou-sheng ZHENG Xiao-fei BIAN Tao(College of Aeronautics and Astronautics Engineering, Air Force Engineering University, Xi＇ an 710038, China)
出处
《传感器与微系统》
CSCD
2017年第9期147-150,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(51476187
51506221)
