基于多级动态主元分析的电熔镁炉异常工况诊断

电熔镁熔炼过程中的异常工况(如半熔化工况)直接影响产品质量、威胁人员和生产安全, 有必要及时诊断. 但与异常直接相关的超高温熔池温度(>2850 ℃)难以利用温度传感器检测, 目前现场主要依靠工人在定期巡检时人眼观察炉壁来诊断, 工作强度大、安全度低、诊断不及时. 针对上述问题, 本文提出一种炉体动态图像驱动的电熔镁炉异常工况实时诊断方法. 结合电熔镁炉熔炼各区域温度分布的空间特征、正常工况下熔炼温度变化和水雾扰动引入的图像时序特征、以及异常工况下温度异常区域持续发亮扩大的特征, 在对炉体动态图像进行空间多级划分的基础上, 提出了一种多级动态主元分析(Multi-level dynamic principal component analysis, MLDPCA) 动态图像分块建模方法. 在此基础上, 提出基于MLDPCA的逐级诊断方法与基于贡献图的异常定位方法. 最后, 采用某电熔镁生产现场的实际图像进行方法验证, 结果表明了所提方法的有效性.

Multi-level Dynamic Principal Component Analysis for Abnormality Diagnosis of Fused Magnesia Furnaces

The abnormalities during the melting process of fused magnesia furnace (FMF) such as semimolten situation may significantly affect the product quality, the safety of personnel and manufacturing process. The abnormal condition diagnosis deserves more attentions. However, the ultra-high temperature within the melting zone of the FMF is not measurable, that makes the diagnosis of FMF abnormality be difficult. The practitioners can only perform occasional visual inspections which often fail to detect the abnormalities in time. In order to resolve this challenge, this paper proposes a novel dynamic image analysis based real-time abnormality diagnosis method for the FMF. The proposed method exploits the spatial and temporal characteristics of temperature fluctuation in FMF in normal condition as well as the partial glowing of the furnace wall and continuous expanding of the glowing area in abnormal conditions. In order to extract these spatial and temporal features from the dynamic images, a new multi-level dynamic principal component analysis (MLDPCA) algorithm is developed. A hierarchical monitoring method is then proposed to perform the abnormality diagnosis and locate the abnormality by using the MLDPCA based contribution plot. The application result on a practical FMF using the collected field images has demonstrated the effectiveness of the proposed method.