摘要
为了提高产品质量预测模型精度和降低过拟合能力,提出了一种基于Stacking集成的研究构思,将LightGBM、GBDT和Lasso三种预测算法进行Stacking集成,以此构建多算法集成的产品质量预测模型。首先,采用XGBoost对制造过程数据进行关键特征提取,同时引入SHAP模型挖掘关键特征对产品质量的影响差异。其次,将关键特征导入基学习器LightGBM、GBDT和Lasso中,并将三种基学习器的预测结果作为元学习器LightGBM的输入,以此完成多算法集成的质量预测。最后以注塑成型加工过程为例,验证所提模型和方法的可行性。
In order to improve the accuracy of product quality prediction and reduce over ftting,this paper proposes a research idea based on stacking integration,which integrates lightGBM,GBDT and Lasso prediction algorithms to build a multi-algorithm integrated product quality prediction model.Firstly,XGBoost is used to extract the key features from the manufacturing process data,and the SHAP model is introduced to mine the differences in the influence of key features on product quality.Secondly,the key features are imported into the basic learners lightGBM,GBDT and Lasso,and the prediction results of the three basic learners are used as the input of the meta learner lightCBM,so as to complete the quality prediction of multi-algorithm integration.Finally,the injection molding process is taken as an example to verify the feasibility of the proposed model and method.
作者
钟武昌
战洪飞
林颖俊
余军合
王瑞
ZHONC Wuchang;ZHAN Hongfei;LIN Yingjun;YU Junhe;WANG Rui(College of Mechanical Engineering and Mechanics,Ningbo University,Ningbo Zhejiang 315211,China;Ningbo beiteri Energy Technology company,Ningbo Zhejiang 315040,China)
出处
《机械设计与研究》
CSCD
北大核心
2023年第5期100-107,共8页
Machine Design And Research
基金
国家重点研发计划项目(2019YFB17071d)
国家自然科学基金资助项目(71671097)。
