论文部分内容阅读
FeO含量是球团质量的重要指标之一。为了更加准确地计算球团化学成分指标,提出了FeO氧化系数的概念。FeO氧化系数按照定义需要球团成分的离线计算,无法进行直接检测,通过相关因素的分析实现了FeO氧化系数的软测量。在分析球团生产过程中FeO影响因素的基础上,利用神经网络建立了FeO氧化系数的软测量模型,通过灰色关联分析方法确定了神经网络软测量模型的输入。使用实际的生产数据对模型的参数进行训练和验证,结果表明,提出的球团FeO氧化系数软测量模型能够获得满意的精度。
FeO content is an important indicator of pellet quality. In order to calculate the chemical composition index of pellets more accurately, the concept of FeO oxidation coefficient was proposed. The oxidation coefficient of FeO needs to be calculated off-line according to the definition of pellet composition and can not be detected directly. The soft sensing of FeO oxidation coefficient is realized through the analysis of related factors. Based on the analysis of the influencing factors of FeO during pellet production, a soft sensing model of FeO oxidation coefficient was established by using neural network, and the input of soft sensing model of neural network was determined by gray relational analysis. The actual production data were used to train and validate the model parameters. The results show that the proposed soft sensing model of FeO oxidation can achieve satisfactory accuracy.