神经网络预测改性海绵铁除氧效果的研究  

Study on the prediction of the deoxidization efficiency of modified sponge iron with artificial neural network

作　　者：徐波[1] 贾铭椿[1] 门金凤[1] 

机构地区：[1]海军工程大学,湖北武汉430033

出　　处：《工业水处理》2012年第9期29-31,共3页Industrial Water Treatment

摘　　要：为实现改性海绵铁填充床除氧过程的自动化控制,研究了人工神经网络对改性海绵铁填充床除氧过程模拟仿真的可行性,采用误差反向传播网络(BP网络)建立了进水流量、改性海绵铁填充高度与溶解氧去除率之间关系的动态模型,并对不同的训练样本归一化方法和训练方法进行比较。结果表明,在网络隐含层层数为1、节点数为7时,采用归一化方法 1和有动态修正的训练方法能够较好地预测填充床对溶解氧的去除率,该模型可用于改性海绵铁填充床除氧过程的动态描述。In order to achieve the automatic control of the deoxidization process of modified sponge iron packed bed, the feasibility of simulation of the deoxidization process of modified sponge iron packed bed with artificial neural network （ANN） has been studied. With error back propagation （BP） network, a dynamic simulation model showing the relationship among feed water flow rate, modified sponge iron packed height with the removing rate of DO is established. Furthermore,different training sample normalization methods and training methods of ANN are compared. The results show that using the first normalization method and the training methods having dynamic modification can accurately predict the removing rate of DO by using the packed bed, when the number of layers of the ANN hidden layer is 1, and the number of nodes is 7. The model can be used for describing the dynamic of the deoxidization process of modified sponge iron packed bed.

关 键 词：改性海绵铁填充床 除氧 人工神经网络 隐含层节点 

分 类 号：TG179[金属学及工艺—金属表面处理] TP183[金属学及工艺—金属学]