摘要
提出一种基于极限学习算法的离散过程神经网络模型,用于解决液体火箭发动机状态预测这一难题。首先,在历史数据的基础上建立离散过程神经网络(DPNN)预测模型;然后,根据在线更新的数据样本,采用递推极限学习(EL)算法对双并联前馈离散过程神经网络(DPFDPNN)隐层到输出层的权值进行更新,并应用权值更新后的过程神经网络对发动机状态进行预测;最后,以液体火箭发动机状态预测中氢涡轮泵扬程预测为例,分别采用有权值更新和无权值更新两种预测模型进行了试验。结果表明,通过更新过程神经网络权值可以使模型具有更高的预测精度和更好的适应能力,该方法能够为液体火箭发动机状态预测提供一种有效的解决途径。
Aimed at the problem of liquid propellant rocket engine condition prediction, a double parallel feedforward discrete process neural network (DPFDPNN) model based on extreme learning (EL) algorithm is proposed. The discrete process neural network (DPNN) , which was trained via off-line data, is firstly adopted to make prediction of liquid propellant rocket engine condition. In order to improve the accuracy and efficiency of the DPNN for condition prediction, the weights connecting the hidden layer and output layer are then direct- ly updated by the EL algorithm based on recursive algorithm with the real data stream. The corresponding com- putational steps are given and the DPNN with weights update is compared with the DPNN without weights update by predicting the lift of oxygen turbo pump. The result shows good accuracy and adaptibility of the DPNN with weights update and this work provides an effective way to solve the problem of liquid propellant rocket engine condition prediction.
出处
《北京航空航天大学学报》
EI
CAS
CSCD
北大核心
2016年第8期1675-1681,共7页
Journal of Beijing University of Aeronautics and Astronautics
基金
国家自然科学基金(51206181
51506219)~~
关键词
液体火箭发动机
状态预测
离散过程神经网络
极限学习算法
递推算法
liquid propellant rocket engine
condition prediction
discrete process neural network
extreme learning algorithm
recursive algorithm