铸件质量溯源和寻因方法研究  

Study on Traceability and Root Cause Analysis Methods for Quality of Aerospace Castings

作　　者：余朋 邱伟真 王玉 李天佑[1] 毛乐 李文[1] 沈旭[1] 吴晓明 计效园[1] 殷亚军[1] 周建新[1] YU Peng;QIU Weizhen;WANG Yu;LI Tianyou;MAO Le;LI Wen;SHEN Xu;WU Xiaoming;JI Xiaoyuan;YIN Yajun;ZHOU Jianxin(State Key Laboratory of Materials Processing and Die&Mould Technology,Huazhong University of Science and Technology,Wuhan 430074,Hubei China;Xi'an Aerospace Engine Co.,Ltd,Xi'an 710025,Shaanxi China)

机构地区：[1]华中科技大学材料成形与模具技术全国重点试验室,湖北武汉430074 [2]西安航天发动机有限公司,陕西西安710025

出　　处：《铸造工程》2024年第3期63-70,共8页Foundry Engineering

基　　金：国家重点研发计划(2020YFB1710100)。

摘　　要：航天钛合金铸件缺陷对铸件性能影响大,生产铸件废品率高,而铸件因在铸造过程中参数波动等原因产生一些内部或外部缺陷,为了精准快速寻找各个关键工艺参数对铸件质量的影响规律,降低铸件产品缺陷。本文使用华铸ERP提取各个生产工艺参数,使用主成分分析降低参数维度,通过BP神经网络预测与分析模型、随机森林算法和XGBoost算法进行对比,结果显示随机森林的预测准确率和召回率最高。根据预测模型阐述了铸造缺陷与工艺参数之间的关联,找到铸件缺陷产生源头,指导优化生产流程,有效减少缺陷率。研究表明,结合大数据分析的预测模型可以实时监控生产过程,实现航天铸件缺陷的质量管理,为促进航天智能化制造企业的升级提供参考。Defects in titanium alloy castings for aerospace applications can greatly affect casting performance,resulting in high scrap rates.Casting defects arise from fluctuations in parameters during the casting process.To accurately and rapidly identify the effects of key process parameters on casting quality and reduce product defects,this study extracted process parameters from the Huzhou ERP system and used principal component analysis to reduce the parameter dimensions.BP neural network,random forest,and XGBoost models were compared for prediction and analysis.Results showed random forest had the highest prediction accuracy and recall.Associations between casting defects and process parameters were explained based on the predictive model to identify the root causes of defects and optimize the production process for effective defect reduction.The study demonstrates that predictive models incorporating big data analytics can monitor production in real-time to manage quality of aerospace castings,providing insights to facilitate smart manufacturing upgrades in the aerospace industry.

关 键 词：钛合金铸件 主成分分析 神经网络 随机森林 XGBoost 

分 类 号：TG249[金属学及工艺—铸造]