摘要
针对舰船的水面动力响应提出了一种基于神经网络集成参数辨识模型,通过由系统微分方程导出的完备状态点的概念,给出了个体网络的生成方法和一种特殊的集成神经网络的模型分解法,可使模型的输出动态逼近目标的实际可能状态,进而对舰船的相关参数做出有效估计;并对所提出的辨识模型进行了稳定性和收敛性分析。仿真试验表明模型具有快速准确的逼近能力和很好的泛化能力。
A parameter identification model based on NNE in accordance with the ship responses of water surface state is presented. By the concept of complete state point space conducted by differential equations of the system,a method of the generating and ensemble of individual neural networks is proposed.The method makes the network state to approach the real possible state of the target dynamically, and gives an effective estimation of the related ship parameters. Stability and astringency of the identified model are analyzed. The simulation results show that the model has strong approaching and good generalization ability.
出处
《船舶力学》
EI
北大核心
2009年第4期566-570,共5页
Journal of Ship Mechanics
基金
国防预研基金(51421060505DZ0155)
陕西省自然科学基金(2005A009)资助
关键词
神经网络集成
参数辨识
数字仿真
舰船参数
NNE（neural network ensembles）
parameter identification
computer simulation
ship parameter
