Parametric energy conserving sampling and weighting for the thermal analysis of Selective Laser Melting

This work investigates the application of Model Order Reduction (MOR) to the case of a non-linear heat transfer simulation, which pertains to Selective Laser Melting (SLM). A hyper-reduction strategy, relying on the Energy Conserving Sampling and Weighting (ECSW) method, is presented before being implemented within an adaptative POD-Greedy algorithm to build a more general reduced order base composed of versatile thermal modes. To this end, an a-posteriori error estimator is used alongside Ordinary Kriging to evaluate the degree of error committed during the parametrized-MOR process. The approach introduces a novel scheme called parametric Energy Conserving Sampling and Weighting (pECSW), which extends the applicability of the sparse selection of finite elements for larger parametric ranges than what is possible with the regular ECSW approach. Hence, the generic property of the final layout avoids launching an optimization strategy for new model analysis.