基于随机森林和聚类的连铸坯纵裂纹预报方法

Longitudinal crack prediction method of continuous cast slab based on random forest and clustering

纵裂纹是一种常见的铸坯表面缺陷,准确预测铸坯表面纵裂对于提高铸坯质量有着重要意义。针对纵裂纹形成与扩展过程中结晶器热电偶温度在时间、空间上的变化趋势,捕获和提取了热电偶时序温度的典型变化特征,采用随机森林(Random Forest, RF)对捕捉到的特征进行降维,筛选出与纵裂联系密切的相关特征,在此基础上建立了基于K均值(K Means)聚类的纵裂检测模型。结果表明,提出的基于温度时序特征和聚类算法的纵裂预测模型能够正确区分和识别纵裂纹和正常工况样本,将机器学习方法引入连铸过程异常监控提供了新的思路。

Longitudinal crack is a common surface defect of casting slab. Accurate prediction of longitudinal cracking on slab surface is of great significance for improving the quality of casting slab. Aiming at the temporal and spatial variation trends of the temperature of the mold thermocouple during the formation and propagation of longitudinal cracks, this paper captures and extracts the typical variation features of the thermocouple temperature in time series, and the Random Forest(RF) algorithm is used to reduce the dimension of the captured features, and the features closely related to longitudinal cracks are extracted. On this basis, a longitudinal crack detection model based on K-means(K Means) clustering was established. The results show that the proposed longitudinal crack prediction model based on temperature-time series features and clustering algorithm can correctly distinguish and identify samples with longitudinal cracks from samples under normal conditions, which provides feasible way for introducing machine learning methods into abnormal monitoring of continuous casting process.