Deep-learning approach for predicting laser-beam absorptance in full-penetration laser keyhole welding

Laser-beam absorptance in a keyhole is generally calculated using either ray-tracing method or electrodynamic simulation, both physics-based. As such, the entire computation must be repeated when geometry changes. In this study, data-based deep-learning model for predicting laser-beam full-penetration laser welding proposed. The uses set of top- and bottom-aperture as inputs. From these, an artificial intelligence (AI) trained to predict laser-energy value. For training dataset, various geometries (i.e., shapes) are hypothetically created, upon which employed compute corresponding values. An image classification model, ResNet, learning recognizer features absorptance. regression, several modifications applied structure. Five depths tested, optimal AI architecture used with R 2 accuracy 99.76% within 1.66 s 740 shapes. Using parameters affecting identified.