Abstract：Aiming at complexity and difficulty of field control of FeO content in sinter，a prediction system of neural network was developed. A 4layer modified feedforward neural network based on multifactor input and network output by adding main factor linear relativity with hypofactor nonlinear relativity, is the newest feasible method of field control of FeO content in sinter. The network structural design is an advanced one, has high accuracy and strong generalization ability. The network training sum of squared error is 0.01508846. The output of FeO of training sample set was tested that the absolute average error is 0.135665, with a hitting accuracy of 97.78%. The prediction after network training has an absolute average error of 0.189226, with a hitting accuracy of 91.14%.
蒋大军. 烧结矿FeO含量预报系统开发与应用[J]. 钢铁, 2006, 41(9): 0-13.
JIANG Dajun. Development and Application of Prediction System of FeO Content in Sinter. Iron and Steel, 2006, 41(9): 0-13.
2006, Vol. 41 