摘要
从BP网络的工作原理出发,利用多水平正交表选取BP神经网络训练样本,通过正向传播和误差反向传播建立BP网络的拓扑结构。对二级减速器BP网络模型进行了结构修正与优化计算。实验表明在机械优化设计领域针对实体结构的动态分析计算,采用BP神经网络可以提高了优化收敛速度和精度,这表明神经网络理论与传统的数值方法相结合的方法具有重要的现实意义。
This paper established the topology structure of network by means of error forward and backward spread, and selected BP network train samples by choosing multi-level orthogonal algorithm forms based on the BP network working principle. The structure revise and optimization calculation has been made to secondary decelerator BP network. The experiment results proved adopting BP network in dynamic analysis calculation of solid structure in the domain of mechanical optimization design can improve the convergence speed and precision of the optimization, and it fully demonstrated the method of combination of network theory with traditional arithmetic method has important reality sense.
出处
《机械设计与制造》
2004年第5期53-55,共3页
Machinery Design & Manufacture
关键词
BP神经网络
优化设计
有限元
样本集
BP network
Optimization design
Finite element
Sample assembly
