基于人工神经网络技术的结构布局优化设计

使用PCL（Patran Command Language）实现了Patran环境下的机翼参数化模型。其优化模型包含两类设计变量：几何位置变量和几何尺寸变量。在采用Nastran软件实现几何尺寸优化的基础上，结舍均匀试验设计方法，利用神经网络的高度非线性映射功能，建立了目标函数与位置设计变量的映射关系。在Matlab环境下，编写了使用改进的可行方向法的优化程序，并对翼梁位置完成优化，最终完成了整个机翼的布局优化设计。可以看出。将参数化建模与神经网络功能结舍进行结构优化，能更好地发挥神经网络的映射功能，使优化结果更加精确、高效。所提方法可以解决在Patran环境下的复杂结构位置变量优化问题，弥补了该软件的不足之处，具有很好的应用推广价值。

Optimization design of structural layout based on techniques of artificial neural network

The parameterized model of aircraft wing was achieved under Patran environment by the use of PCL(Patran Command Language).The optimization model contains two kinds of designing variables: geometric position variable and geometric dimension variable.On the bases of adopting Nastran software to realize optimization of geometric dimension and combining with the designing method of even experiment and utilizing the high degreed nonlinear mapping function of neural network this paper established the mapping relationship between objective function and positional designing variables.The optimization program was compiled by using the modified feasible directional method under Matlab environment,and completed the optimization on the position of wing beam,and finally achieved the layout optimization design of the entire aircraft wing.It can be seen through this paper that let the parameterized modeling be combined with the function of neural network to carry out structural optimization could give the reins to the mapping function of neural network nicely and let the result of optimization be more precise and effective.The method provided in this paper could be able to solve the optimization problem of complicated structural positional variables under the Patran environment,thus counteracted the deficiency point of this software and let it possesses pretty good applicable popularization value.