Research on Defect Prediction of Laser-arc Hybrid Additive Manufacturing Based on Bayesian Optimization XGBoost

Laser-arc hybrid additive manufacturing technology has broad application prospects in various important technical fields. However, there are still technical problems in the control of defects inside the formed parts. Aiming at the binary classification problem of whether there are defects in the cladding layer of TC4 alloy, the process test was designed by trial and error method combined with orthogonal rotation experiment to obtain whether there are defects in the cladding layer under different process parameters. A binary classification prediction model based on Bayesian optimization extreme gradient boosting (BOXGBoost) is established. The experimental results show that the BO-XGBoost prediction model has an accuracy rate of 92.9%, a recall rate of 100%, an accuracy rate of 85.7%, and a F1 score of 92.3%. It can provide guidance value for the identification of internal defects of composite heat source additive forming parts.