| 摘要: |
| 针对高炉冶炼过程复杂多变，影响高炉炉况的因素众多且运行过程复杂，为保证高炉炉况稳定顺行，开发了一种基于主元分析（PCA）和统计过程控制（SPC）的高炉炉况异常检测模型。由于高炉运行参数具有高耦合和非高斯的特点，该模型首先采用主元分析算法对高炉实际生产的历史离线数据进行聚类分析，然后应用基于T2统计量的多元控制图和单值控制图对聚类后的新变量和相关参数进行监测，从而达到监测高炉出现异常炉况的目的。该模型可以实时监测高炉炉况的异常波动，其中PCA算法将高炉本身的高维数据降到低维，大幅去除数据中的噪声和不重要特征，在实际生产和应用中，节省了大量的成本和时间。选取马钢某高炉炼铁过程为应用场景，结合数据特点调整和改进算法，通过历史数据模拟和实时在线运行验证模型的可靠性和算法的有效性，同时也对指导高炉实际操作技术做出了一定的贡献。 |
| 关键词: 高炉 统计过程控制 主元分析 过程监测 |
| DOI： |
| 分类号: |
| 基金项目: |
|
| Research on Abnormal Detection of Blast Furnace Condition Based on SPC |
|
XIAO Weimin, ZANG Jun, YUAN Zhixiang, REN Shiliu
|
|
School of Computer Science and Technology, Anhui University of Technology, Anhui Ma'anshan 243032,China
|
| Abstract: |
| The smelting process of blast furnace is complex and changeable, many factors affect the conditions of blast furnace, and the operation process is complex. In order to ensure the stable and forward flow of the blast furnace condition, an abnormal detection model of blast furnace conditions based on principal component analysis (PCA) and statistical process control (SPC) has been developed. Due to the high coupling and non-Gaussian characteristics of blast furnace operating parameters, the model first used a principal component analysis algorithm to perform cluster analysis on the historical offline data of the actual production of the blast furnace. Then, the multivariate control chart based on T2 statistics and single value control chart were used to monitor the new variables and related parameters after clustering, so as to monitor abnormal furnace conditions in the blast furnace. The model can monitor the abnormal fluctuations of blast furnace conditions in real time. The PCA algorithm reduced the high-dimensional data of the blast furnace to low-dimensional, greatly removing noise and unimportant features in the data, saving a lot of cost and time in actual production and application. The ironmaking process of a certain blast furnace in Ma'anshan Iron and Steel was selected as the application scenario, and the algorithm was adjusted and improved in combination with the data characteristics,and the reliability of the model and the effectiveness of the algorithm were verified through historical data simulation and real-time online operation. The achievements in this paper make certain contributions to guiding the actual operation technology of the blast furnace. |
| Key words: blast furnace statistical process control principal component analysis process monitoring |
