在分析神经网络模型模拟非线性函数的学习训练过程及曲线曲面的数控插补原理的基础上，提出一种基于三层ＢＰ网络的任意空间曲线插科方法，并以平面内三次曲线为例，进行仿真计算。结果表明，选取合适的插补周期以及足够的隐层节点数，神经网络可以精确地模拟曲线。 On the basis of analyzing the training process of neural network model for simulating nonlinear function and the principle of NC interpolation of curve and surface, a method of arbitrary space curve interpolation based on three-layer BP network is proposed. Taking the in-plane cubic curve as an example, Simulation calculation. The results show that the neural network can accurately simulate the curve by selecting the appropriate interpolation period and sufficient number of hidden nodes.