摘要
结合混沌序列的相空间重构理论和BP神经网络预测理论,构建了一个基于时间序列预测的混沌神经网络模型;考虑基本BP神经网络采用的梯度学习算法收敛速度较慢的缺点,文章利用改进的Levenberg-Marquart(L-M)优化学习算法对网络进行训练;最后对一组飞机舵面卡死故障数据进行仿真实验,结果表明该模型不仅提高了预测精度,而且网络收敛速度也得到明显的改善,有效避免神经网络局部极小问题,可以较好地对飞机舵面卡死故障进行预测。
This paper combines the phase space reconstruction theory of chaos series and BP neural network theory on prediction , establish a chaos neural network model based on time series prediction. Considering shortage of slow convergence rate based on the basic BP neural network, an improved learning algorithm Levenberg--Marquart (L--M) is presented to train the network. Finally , the paper do a experiment on simulating a fault data of airline steer stuck, the result show that the new model not only will improve the prediction accuracy, but also the convergence rate obviously, and could avoid the local minimum and predict the fault of airline steer stuck effectively.
出处
《计算机测量与控制》
CSCD
北大核心
2010年第5期1011-1013,共3页
Computer Measurement &Control
基金
陕西省自然科学基金(SJ08F20)
航空科学基金(2008ZD53035)