增量式遗传RBF神经网络在铁水脱硫预处理中的应用

Application of Incremental Genetic RBF NN to Hot Metal Desulfurization Pretreatment

铁水脱硫预处理过程是一个非常复杂的多元非线性反应过程 ,针对它提出了基于增量式遗传RBF神经网络的模式识别方法 ,预测脱硫剂加入量 .该算法克服了RBF中心个数选择的随机性 ,较好地解决了样本聚类 .为了保证网络结构能适应不断扩大的数据集 ,提出了增量数据处理方法 ,对原有网络参数进行修正 ,这样就有利于连续生产操作 .现场测试结果表明 ,采用该算法后结果的误差较小 ,满足了终点命中率在90 %以上的指标 ,提高了经济效益 ,这说明该算法具有工程实用性 .

Hot metal desulfurization pretreatment is a very sophisticated reaction which is not only diverse but also non-linear. A model identification method based on genetic RBF neural network to predict desulfurizer quantity is put forward, which can perfectly resolve the problem of random selection of RBF cluster center number and sample data clustering. In order to ensure structure of neural network to fit with continuous incremental data set, a method for dealing with incremental data is presented, which is applied to amend the parameters of neural network . Then the request of continuous production is satisfied . Finally the testing results are given, showing that after adopting the algorithm, the error of result is less than before and end-point hitting ratio satisfies to ninety percent, indicating that the algorithm has the engineering practicability.