A Regression-Based Method for Reduced Order Modelling of Laser Welding Process

With a growing demand for customized consumer products tailored to customer demands, innovative and industrially sustainable solutions need to be developed to overcome the challenges attributed when constant reconfigurations occur in a manufacturing environment. Mainly, there is a need to minimize the reconfiguration time and costs, by avoiding the usual physical tests that are being carried on, now, to sign off a production. Laser welding is a vital manufacturing process throughout the world, due to the high accuracy, and high production rates that can be achieved, by it. In this paper, an innovative method for reduced order modelling of the laser welding process is presented, which is based on decomposing and parameterizing accordingly, first, the geometrical features that are involved during welding (e.g., part thickness), and second the heat source that the laser beam creates (e.g., laser power, laser welding speed). The method developed, can be used to develop models that easily represent different laser welding manufacturing scenarios, and rapidly predict the impact that the different process parameters, have on the end product’s vital characteristics, which usually are accompanied by strict tolerances that need to be respected.