A Gradient Boosting Machine Learning for Predicting Seam Width and Tensile Strength in Pulsed Laser Bonding of Glass to Copper

This study intends to use a gradient boosting machine to estimate the laser welding quality of copper/glass junctions based on the experimental data. Two output characteristics, such as seam width and shear strength, were taken into consideration, along with three other parameters: laser power, scanning speed, and pulse frequency. A total of 64 sets of experimental parameters were run, with each parameter having four levels. It is confirmed that input data has an impact on bonding quality with relation to microstructure characterization. Using the same dataset for comparison, two auxiliary models—an artificial neural network and full quadratic multiple regression analysis—were created in order to confirm the suitability and benefit of the gradient boosting machine. Additionally, the significance of input factors could be illustrated using the random forest model.