ABSTRACT

The iron ore sinter is still the main raw material for the blast furnace ironmaking process, its properties, such as strength and reducibility, are of vital importance to the productivity and the smooth operation of the blast furnace. In the present study, one model based on an artificial neural network (ANN) was established to predict the sinter strength. The ANN model was trained with the sample set, which was generated from the credible data from the published papers. The comparison between the direct prediction model and the indirect prediction model with the amount of liquid phase and spinal phase calculated with thermodynamic theory as the middle layer was carried out. The results show that the indirect ANN model gave much higher accurate prediction results than that of the direct one without the middle layer.The parametric study with the validated model shows that the sinter strength increased first with increasing the SiO₂ to 5.4% and then decreased with further increasing the SiO₂.

摘要

铁矿石烧结矿仍然是高炉炼铁过程的主要原料，其强度和还原性等性能对高炉的生产率和平稳运行至关重要。在本研究中，建立了一种基于人工神经网络 (ANN) 的模型来预测烧结强度。ANN 模型使用样本集进行训练，该样本集是根据已发表论文的可靠数据生成的。对直接预测模型与以热力学理论计算的液相量和脊髓相量为中间层的间接预测模型进行了比较。结果表明，间接ANN模型比没有中间层的直接ANN模型给出了更高的准确预测结果。₂至 5.4%，然后随着 SiO _{2 的}进一步增加而降低。