Comparative Evaluation of Optimization Algorithms for Automatic Build Orientation for Powder Bed Fusion of Metals Using a Laser Beam

For additive manufacturing (AM) using Powder Bed Fusion of Metals using a Laser Beam (PBF-LB/M), digital preparation steps (part design, part orientation in the chamber, supporting, etc.) are required. In the current state of the art digital process chains, these steps are done manually by an engineer, utilizing specialized AM preparation software, e.g., Autodesk Netfabb or Materialise Magics. Especially the part orientation has a significant impact on part quality and costs. However, setting up a suitable orientation algorithm is a complex and multi-dimensional engineering problem, so computationally intensive evaluation procedures are necessary. The orientation selection is usually done by comparing the pre-evaluated part orientations relative to each other. The selection strategy thus significantly influences the quality of the orientation schemes and the computational effort. The selection strategy often follows brute force logic to investigate the entire orientation space. This approach divides the solution space into equidistant steps that limit the search resolution. Therefore, this paper conducts a comparative evaluation of different new strategies to replace the brute force algorithm with a suitable optimization algorithm that selects the calculated partial orientations based on a priori decision logic. Three optimization algorithms Covariance Matrix Adaption Evolution Strategy (stochastic method), Multilevel Coordinate Search (direct search method), and Efficient Global Optimization (surrogate model optimization) are evaluated regarding their orientation evaluation results. The number of objective function evaluations required is compared to brute force. Each orientation optimization algorithm is applied to 42 parts. All algorithms reduce the number of orientation evaluations compared to the discrete method of brute force, but differ in the evaluated orientation quality. The Covariance Matrix Adaption Evolution Strategy produced the best-evaluated orientation but also required the highest number of orientation evaluations. In the future, the work should help to find a suitable optimization algorithm for the automatic part orientation finding for PBF-LB/M.