混沌对角递归神经网络的船舶横摇预报方法

船舶运动在一定条件下会出现混沌特性,因此可以利用混沌神经网络对其进行预报.对传统的混沌对角递归神经网络模型各权值的训练进行优化,给出了基于Lyapunov函数的各层权所通用的学习速率调整算法的收敛定理并加以证明.仿真结果表明,采用优化采样时刻可提高各权值的精确度,使收敛性得到改善,能有效提高预报精度和延长预报时间.与前向神经网络BP预测相对比,优化后的模型具有很好的预测效果.

Approach of prediction of ship rolling based on chaotic diagonal recurrentneural networks

There is chaotic characteristics in ship motion under certain conditions, so the chaotic diagonal recurrent neural
network(CDRNN) is proposed to predict ship swaying motion. A convergence theorem of each weight learn algorithm based
on Lyapunov function is given and proofed. Simulation results show that, the value of the optimized sampling time ?? is
applied to increase the accuracy of all of the weight, which improves algorithm convergence, and the predicted precision
and the forecast time are advanced efficiently. The optimized modeling has better predictive effect using CDRNN than using
feed-forward BP neural network to predict.