Rapid optimization of laser powder bed fusion process: a high-throughput integrated multi-task robust modeling approach

Abstract Transpiration cooling has important applications in the hot-end components of aerospace engines, and its service performance depends on its processing quality and microchannel accuracy. Laser powder bed fusion (LPBF) can synchronously achieve integrated manufacturing of complex components and high-precision processing of microchannels, which has great application prospects in transpiration cooling of engine components. However, LPBF involves a series of process parameters. How to effectively optimize process parameters to achieve precision control of processing quality and microchannel accuracy is the primary issue. Especially, the time-consuming and labor-intensive problems brought about by the combination of a large number of process parameters and subsequent characterization caused by multivariate cross experiments, as well as the effective analysis of a large amount of data results and the revelation of internal correlations, pose significant challenges to the process control of LPBF. This study proposes a methodology that combines high-throughput experiments with Gaussian process algorithms to optimize the processing quality and accuracy of the nickel-based high-temperature alloy engine casing divergent cooling microchannel structures. 250 parameter combinations, including laser power, laser scanning speed, model channel diameter, and spot compensation were designed and integrated into 10 high-throughput specimens to quickly and efficiently characterize the processing quality and microchannel accuracy. A Gaussian process model was further established using the Bayesian Optimization, and effective prediction of processing density and accuracy within a wide range of process parameters was achieved. The relationship between various process parameters and processing quality and accuracy was revealed, and various optimized process combinations were summarized. Finally, CT testing was conducted on high-throughput specimens to verify the effectiveness and accuracy of this method. The method proposed in this study provides a way for quickly and efficiently optimizing the process parameters and establishing process-property relationships for LPBF, which has broad application value.