根据合成氨实际生产对转化温度的要求 ,提出了定限值控制问题。初步讨论了采用神经网络控制器 ,来实现定值控制向定限值控制逼近的方法。为提高系统响应的快速性 ,在神经网络内部采用反偏差阈值函数 ,在神经网络外部采用双模式神经网络控制结构 According to the actual production of synthetic ammonia conversion temperature requirements, the proposed limit value control problem. The method of using neural network controller to approach the constant value control to definite limit control is discussed preliminarily. In order to improve the rapid response of the system, the inverse deviation threshold function is adopted inside the neural network, and a dual mode neural network control structure is adopted outside the neural network