基于数据挖掘与清洗的高炉操作参数优化

为了提高企业生产力，实现“智慧钢厂”，对企业的海量生产数据信息进行有效挖掘，收集了某钢厂最近4年的高炉生产数据，利用箱型图进行数据清洗，提高数据质量.采取工艺理论和专家经验结合随机森林算法筛选出23个影响铁水质量和产量的特征参数.以铁水产量和铁水［Si+Ti］质量分数为目标参数，通过k-means聚类分析法将其分为3类.将分类结果与特征参数整合后进行分析，得到造成铁水产量和质量大范围波动的13个参数，同时提供了相应参数的合理控制范围.研究可对高炉稳定顺行以及数据挖掘在钢铁行业的应用提供指导.

Operation Parameters Optimization of Blast Furnaces Based on Data Mining and Cleaning

Effective mining of mass production data helps to improve the productivity and the level of management to realize “smart steel works”. The production data for the last 4 years of one steel works were collected， and the box plot was used to clean the data to improve their quality. Twenty-three characteristic parameters affecting hot metal quality and yield were selected by technological theory and expertise combined with the random forest algorithm. Hot metal yield and Si+Ti content were taken as the objective parameters， which could be divided into three categories by k-means cluster analysis. Thirteen parameters contributing to the wide range fluctuation of hot metal yield and quality were obtained after the comprehensive analysis of classification results and characteristic parameters， and the reasonable variable ranges of corresponding parameters were provided. The research can guide the stable operation of blast furnaces and the application of data mining in the iron and steel industry.