Feed-forward Process-parameter Optimization for Control of Laser Powder Bed Fusion Melt-pool Size
- today
- access_time -
- location_on404 AB
- blur_circularConference
Conference Abstract: Laser Powder-Bed Fusion typically applies a set of nominal machine parameters to different regions of a part (e.g. downskin, and core). The parameters are pre-certified experimentally using simple test coupons, and then applied to actual parts based on geometric heuristics (e.g. overhang-angle). For complex geometries this approach proves problematic and can result in non-uniform melting (and thus defects). For example, a geometry with thin features on the lower layers that thicken as build height increases, may achieve good quality on the thin features but then experience overheating as the build transitions into the thicker regions. This non-uniform heating creates a challenge for certifying part quality as the typical parameter certification process relies on the assumption that actual parts will have the same thermal history as the test coupons. Attempts to address these issues have focused primarily on feedback-control which alters process parameters on the fly. Feedback-control poses an issue for many applications (e.g., aerospace and medical) that require inputs to be known prior to manufacturing. In this work a feed-forward approach to parameter optimization is proposed such that near-uniform melt pool size is achieved over the entire build. The approach uses a highly parallelized and vectorized part-scale finite element simulation with periodic adaptivity that can predict inter-layer temperatures for even extremely complex AM parts. A look-up table is input into the part-scale simulation to determine the laser power that should be applied on a point-by-point basis based on the predicted inter-layer temperatures. The look-up table can be determined empirically or numerically, and an example of each approach is presented. A real-time controller is used to implement the adaptive control scheme. A baseline experimental case using nominal machine parameters is compared to an experimental case using optimized laser power by sectioning the parts and characterizing the melt-pools.