摘要
利用人工神经网络的理论 ,尝试将神经网络理论应用于机械优化设计。利用时变回归神经网络的可行性、渐近稳定性和最优性 ,通过神经网络系统的演化 ,使之演化到系统的平衡态 ,计算网络的 L yapunov函数 ,并将该函数与机械优化设计问题的最优解相对应 ,将人工神经网络在此时的网络参数与机械优化设计问题的设计变量相对应 ,从而实现机械优化问题的求解。描述了神经网络用于机械优化设计的仿真算法 ,并以弹簧为例进行了优化计算 。
We apply artificial neural network theory to mechanical optimal design. We first construct an artificial neural network, calculate the network's Lyapunov function, evolve it and lead it to its stabilization status. In order to get an approach to the mechanical optimal design, let the network's evolving process corresponds to the searching process. Then, this stabilization status corresponds to the optimization status of mechanical design problem, the Lyapunov function of this network at this stabilization status corresponds to the optimization solution of mechanical design problem and the network's parameters correspond to the design variations of mechanical optimal design. The example of spring's optimal design by using this strategy proves that this method is efficiency and validity.
出处
《机械科学与技术》
CSCD
北大核心
2000年第6期898-900,903,共4页
Mechanical Science and Technology for Aerospace Engineering
关键词
弹簧
优化设计
时变回归神经网络
Mechanical optimal design
Artificial neural network
Spring
