XGBoost-based on-line prediction of seam tensile strength for Al-Li alloy in laser welding: Experiment study and modelling. (April 2021)

• Record Type: Journal Article
• Authors: Zhang, Zhifen; Huang, Yiming; Qin, Rui; Ren, Wenjing; Wen, Guangrui
• Is Part Of: Journal of manufacturing processes. Volume 64(2021)
• Page Start: 30
• Page End: 44
• Publication Date: 2021-04
• Subjects: WAAM wire and arc additive manufacturing -- TIG inert gas tungsten arc welding -- MIG metal-inert gas welding -- CMT cold metal transition welding -- SSC seam strength coefficient -- CC Correlation Coefficient -- PCA principal component analysis -- Fast-ICA fast independent component analysis -- XGBoost extreme gradient boosting decision tree -- RF random forest -- GBDT gradient boosting decision tree -- DT decision tree -- LR linear regression -- MSE mean squared error -- MAE mean absolute error -- R2 determinate coefficient R-Square -- RF-30Dorg original 30 dimension feature selected by RF -- RFPCA-25D 25 dimension of new PCA feature transformed from RF-30Dorg -- RFPCA-F10D the first 10 dimension of feature subset from RFPCA-25D -- RFPCA-L10D the last 10 dimension of feature subset from RFPCA-25D; Laser welding -- Aluminum-lithium alloy -- Optical spectroscopy -- Ensemble learning -- Extreme gradient boosting -- Feature reduction; Production management -- Data processing -- Periodicals; Manufacturing processes -- Periodicals; Procestechnologie; Productietechniek; Production -- Gestion -- Informatique -- Périodiques; Fabrication -- Périodiques; Manufacturing processes; Production management -- Data processing; Periodicals; 670.5
• DOI: 10.1016/j.jmapro.2020.12.004
• Languages: English
• ISSNs: 1526-6125

Abstract

Graphical abstract: Highlights: A new XGBoost prediction model for real-time SSC prediction in laser welding. 68.6 % increase rate of the new model from 0.2947 to 0.9383. Li I at 395.09 nm shows the highest importance followed by Al I at 669.84 nm, Mg I at 518.4 nm and Ar I. A good linear and positive correlation between the spectrum intensity of Mg I (517.27 nm) and seam strength coefficient. Abstract: This paper studies the regression prediction of laser welding seam strength of aluminum-lithium alloy used in the rocket storage tank by means of the optical spectrum and extreme gradient boosting decision tree (XGBoost). First, the relationship between the spectrum intensity and the seam strength coefficient is thoroughly investigated through parameters changing experiments using the developed monitoring system of the optical spectrum. Then, the importance of the metal line spectrum, including Al I, Li I, and Mg I, is quantitatively evaluated, and good complementarity between the Random Fores(RF)t and Principal Component Analysis(PCA) is demonstrated. Finally, a novel regression model, e.g., RFPCA-XGBoost is proposed and is compared with other different feature selection methods, tree-based ensemble learning models and grid search parameters optimization, and the comparison results show that among all the methods, the proposed model has the best performance regarding the R2 value, achieving the R2 value of 0.9383.