Bayesian Process Optimization for Additively Manufactured Nitinol

Additively manufactured nitinol enables the design and rapid prototyping of the shape memory alloy with great flexibility and cost-effectiveness in various applications. To achieve high-density fabrication of nitinol, we utilize a Gaussian process-based Bayesian optimization method to efficiently optimize process parameters of the laser beam-powder bed fusion (LB-PBF) process in this work. Specifically, Gaussian process regression is applied to formulate a surrogate model between the critical process parameters (i.e., laser power, scanning speed) and the residual porosity of the nitinol samples. Then Bayesian optimization is integrated to successively explore the design space to search for the optimal process parameters. These two methods are integrated to find the global optimum iteratively. Compared with the traditional trial-and-error methods, the proposed method can quickly find the optimal process parameter for the high-quality nitinol samples, especially with many process parameters, and accelerate the innovations with nitinol in additive manufacturing.