传统回归方法对于围岩变形时程曲线存在反弯点，这种情况的模拟具有难度。提出的基于RBF神经网络的隧道围岩变形预测方法，不仅能很好地描述复杂的围岩变形时程曲线，而且比BP神经网络具有更快的收敛速度和更好的全局搜索能力。实例研究验证了该方法的有效性与可行性。

Owing to the difficulty of traditional multi-variable regression methods to represent the surrounding rock deformation curve with inflexion points, a method for forecasting tunnel surrounding rock deformation using radial basis function neural networks is presented. This method not only can be utilized to approximate the complex deformation curves, but also has higher convergence speed and better globally-searching ability than those using BP neural networks. An example is given to show the effectiveness and practicability of this method.