不规则零件优化排样的神经网络混合优化算法

 提出一种利用人工神经网络求解不规则件排样问题的混合优化方法.该方法首先把排样和制造工艺联系起来,将多边形各边向外扩充,为零件预留加工余量；然后采用自组织特征映射模型（SOM）和Hopfield人工神经网络相结合的方法,运用SOM神经网络对初始在板材内随机排布的不规则零件进行平移,逐步减小不规则零件之间的重叠面积,求得各零件的最优位置,再运用Hopfield神经网络对平移后的零件旋转,进行迭代运算,当能量函数达到稳定状态时,得到各排样零件的最优旋转角度组合,实现自动排样.算法可以解决不规则件和矩形件在规则板材以及不规则板材上的排样问题,实例证明了该算法的有效性和实用性.

Neural networks hybrid algorithm for irregular parts optimal layout

An irregular parts optimal layout method based on artificial neural networks （ANN） was proposed. Firstly the layout problem was connected with the manufacture process, and each side of the polygon was expanded to preserve machining allowance. Then self organizing mapping （SOM） and Hopfield ANN were integrated. SOM ANN was adopted to translate the irregular parts which originally randomly distributed on the sheet to gradually decrease the overlapping area between irregular parts. As a result, the best location of each part was obtained. Hopfield ANN was utilized to rotate the translated parts. During iteration calculation, when the power function reached stable state, the optimal rotating angle of each part was obtained and automatic layout was realized. The algorithm can solve irregular parts layout problem and rectangular parts layout problem in the given layout region. Illustrations indicate that this algorithm is effective and practical．