一种基于改进的RBF神经网络的铁水脱硫预报模型

A Prediction Model for Molten Iron Desulfuration Based on An Improved RBFNN

脱硫过程是炼钢生产中一个十分重要的环节。脱硫效果的好坏,直接关系到炼钢生产能否保质保量地进行,而这又取决于对脱硫过程的控制,其关键是脱硫剂的加入量。因而必须建立脱硫过程模型,实时和高精度地预报脱硫剂的加入量。而脱硫过程又是一个非常复杂的工艺过程,采用传统的方法建立的模型无法保证稳定和高精度的脱硫效果。笔者提出了一种基于改进的RBF神经网络的铁水脱硫预报模型及其具体设计方法,并在炼钢厂进行了实际投运。结果表明,该模型性能良好,这同时说明了设计方法的有效性和实用性。

Desulfuration process is a very important phase in the steeling production. Whether the result of desulfuration is good or not will primarily decide whether the quality and quantity of the steeling production can be guaranteed. It lies on the control of desulfuration process, most of all, the control of the amount of the desulfuration material. So a model of desulfuration process must be established to predict the amount of the desulfuration material timely and accurately. Since desulfuration process is very complicated, the model estabished by traditional methods is hard to achieve a stable and accurate desulfuration result. A prediction model for molten iron desulfuration based on an improved RBFNN and its specific design methods are presented together. The results from the actual application in a steel plant prove that the model constructed by the specific methods seems to be successful for such applications, and it is possible for the methods to become useful tools for prediction and optimization in iron and steel industry.