基于EasyEnsemble和XGBoost算法的焊缝超声波检测结果预测模型  

Prediction Model of Weld Ultrasonic Inspection Results Based on EasyEnsemble and XGBoost Algorithm

作　　者：陈毓 陈亮[1] 汪琰 郑宇[1] CHEN Yu;CHEN Liang;WANG Yan;ZHENG Yu(School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学机械与动力工程学院,上海200240

出　　处：《船舶工程》2022年第4期134-139,共6页Ship Engineering

摘　　要：针对某船厂提出的降低不合格焊缝漏检率的需求,基于焊缝超声波检测历史数据,在筛选焊接质量关联参数并进行归一化、编码等特征工程处理的基础上,通过主成分分析提取有效特征作为模型输入,提出一种基于Easy Ensemble和XGBoost算法的焊缝超声波检测结果预测模型。该模型针对样本数据分布极度不均衡而导致负样本召回率低的问题,选取Easy Ensemble算法进行数据增强获得均衡训练样本集。试验证明,所提方法极大地提高了负样本召回率,降低了不合格焊缝的漏检率。According to the demand of a shipyard to reduce the missed inspection rate of unqualified welds,based on the historical data of weld ultrasonic testing,on the basis of screening the welding quality related parameters and carrying out normalization,coding and other feature engineering processing,effective features are extracted through principal component analysis as model input,and a prediction model of weld ultrasonic testing results based on Easy Ensemble and XGBoost algorithm is proposed.Considering the low recall of negative samples caused by extremely unbalanced sample data distribution,the Easy Ensemble algorithm was adopted to obtain a balanced training sample set and XGBoost algorithm was used as the base classification model of EasyEnsemble algorithm.Experiments show that the proposed method greatly improves the recall rate of negative samples and reduces the missed inspection rate of unqualified welds.

关 键 词：焊缝质量预测 分段制造 XGBoost Easy Ensemble 

分 类 号：U673.2[交通运输工程—船舶及航道工程]