Distortion minimization of laser beam welded components by the use of finite element simulation and Artificial Intelligence

Abstract Welded components often require costly efforts to compensate for the thermally induced distortions. This is usually done by applying a specific pre-deformation or by reworking, for example by thermal straightening. These approaches are not practicable for joining complex structures. Therefore, a method that is capable of minimizing the distortions of a complex frame structure with multiple welds is presented. A meta-model by means of an Artificial Neural Network is used to predict the local distortions depending on the welding parameters within sub-areas. A genetic algorithm is utilized to efficiently find suitable welding parameters for the global structure. It could be shown that by applying the method, distortion minimized parameters from more than one billion potential parameter combinations are identified both efficiently and reliably.