基于PSO-DE-XGBoost的小方坯缺陷预测模型  

Prediction model for defects in small billets based on PSO-DE-XGBoost

作　　者：田炎霖 闫文青[1] 高峰 顾进广[2] Tian Yanlin;Yan Wenqing;Gao Feng;Gu Jinguang(Wuhan University of Science and Technology;MCC South Casting Technology Engineering Co.,Ltd.)

机构地区：[1]武汉科技大学材料学部 [2]武汉科技大学计算机科学与技术学院 [3]中冶南方连铸技术工程有限责任公司

出　　处：《冶金能源》2025年第2期75-80,共6页Energy For Metallurgical Industry

基　　金：国家自然科学基金(U1836118);科技创新“2030”新一代人工智能技术(2020AAA0108500);富媒体数字出版内容组织和知识服务实验室开发基金(ZD2022-10/05)。

摘　　要：针对某厂的小方坯缺陷分类数据,文章提出了一种基于粒子群优化(PSO)、差分进化(DE)和极限梯度提升(XGBoost)算法的小方坯缺陷预测模型。首先通过过采样、随机采样和特征工程三种数据增强的方法对不平衡数据集进行预处理,随后采用PSO和DE方法优化XGBoost算法的超参数,并验证模型性能。实验结果表明,经过特征工程增强的数据集训练的模型性能最优,对于中心缩孔等级、内部裂纹等级、表面缺陷等级和形状缺陷等级的预测准确率分别达到79.07%、80.6%、97.5%和72.7%。该研究方法不仅为小方坯缺陷预测提供了一种新的思路,同时也为铸坯的制造过程质量控制提供了参考。The article proposes a small billet defect prediction model based on Particle Swarm Optimization(PSO),Differential Evolution(DE),and the XGBoost algorithm.Firstly,it preprocesses the imbalanced dataset of defect classifications from a certain factory using three data augmentation methods:oversampling,random sampling,and feature engineering.Subsequently,PSO and DE are employed to optimize the hyperparameters of the XGBoost algorithm,and model performance is validated through five-fold cross-validation.Experimental results show that the model trained on the feature-enhanced dataset achieves optimal performance.Specifically,the prediction accuracy for central shrinkage level,internal crack level,surface defect level and shape defect level reaches 79.07%,80.6%,97.5%,and 72.7%,respectively.This study not only provides a new approach for predicting small billet defects but also serves as a reference for quality control in the manufacturing process of billet.

关 键 词：小方坯缺陷 粒子群优化 差分进化 XGBoost 数据增强 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TF777.3[自动化与计算机技术—控制科学与工程]