通过 NC系统用软件方法补偿热误差是抵消热变形的有效方法。多输入多输出模型可以用于建立机床热行为数学模型。文章介绍了神经网络和多元线性回归方法 ,可用来建立大型数控铣床的热变形数学模型。试验和计算表明 ,这两种多变量模型能够预报大部分的热变形误差 Using software to compensate for thermal errors through the NC system is an effective way to counteract thermal deformation. Multiple input and multiple output models can be used to establish a mathematical model of machine tool thermal behavior. This paper introduces neural network and multiple linear regression method, which can be used to establish the mathematical model of thermal deformation of large CNC milling machine. Experiments and calculations show that these two multivariable models can predict most of the thermal deformation errors