本文应用Ｋｏｈｏｎｅｎ 自组织神经网络求解时延、功耗和连线三重驱动的门阵列布局问题．算法用自组织学习算法和分配算法确定关键单元的位置，用迭代改善的方法确定非关键单元的位置，从而获得关键线网最短、散热大的单元离得尽可能远并且单元连线总长尽可能短的布局．本文还介绍了面向线网和功耗的样本矢量的概念，与面向单元的样本矢量相比，面向线网和功耗的样本矢量不仅可以直接处理多端线网，而且能够描述时延信息和热信息．实验表明这是一种有效的方法 In this paper, Kohonen self-organizing neural network is used to solve the gate array layout problem of triple-drive delay, power consumption and connection. The algorithm uses self-organizing learning algorithm and allocation algorithm to determine the location of the key elements, iteratively improve the method to determine the location of non-critical elements to obtain the shortest key network, the heat dissipation of large cells as far as possible and the total length of the unit connection as much as possible Short layout. This paper also introduces the concept of sample vector oriented to line network and power consumption. Compared with cell-oriented sample vector, the sample vector oriented to line network and power consumption can not only directly process multi-end wire network but also can describe delay information and heat information. Experiments show that this is an effective method