A Method to Design Hybrid Lattice Support Structures for LPBF Additive Manufacturing

 Support structure design is imperative in the design of additively manufactured parts with overhang features in the build direction, especially those fabricated using laser powder bed fusion (LPBF). When designed effectively, support structures quickly dissipate heat and mitigate part distortion without driving up excessive costs. Lattices, composed of individual unit cells strategically arranged to achieve a desired function, are a promising solution as a support structure (e.g., tunable properties, reduction in manufacturing costs). Despite their potential, two main drawbacks are (1) the computational cost to find an optimally directed solution and (2) the adaptability to complex support structures. Prior research has designed lattice support structures utilizing both gradient- and non-gradient-based optimizers; however, there still exist limitations within current work. Gradient-based optimizers pose challenges regarding limited design exploration and non-differentiable objective function. Non-gradient-based optimizers, proven to be an effective alternative solution, are known to be too slow in comparison to gradient-based optimizers and have not yet been applied to consider the multi-physics functionality of support structure. Furthermore, the box-like structure of the unit cells makes it difficult to apply to complex structures. Current methods employing non-solid and solid pin supports are not reliable for curved or inclined surfaces. Thus, to improve the current state of the design of support structures, a robust method to facilitate the design of lattice support structures and a multi-sized unit cell approach using LPBF is proposed.

The proposed method addresses the two defined roadblocks of computational cost and adaptability and provide validation for each through three works. The first two works address the high computational cost associated with significant design exploration and using simulation-informed evaluations are addressed by modifying non-gradient based optimizer, simulated annealing (SA). The modified SA-based method is utilized to quickly optimize the distribution of commonly employed, pre-defined unit cells while adhering to user-defined manufacturing constraints. By incorporating a stage-dependent annealing swapping strategy, a decrease in iteration count to explore the domain is achieved for multiple scales. Combined approximation technique of homogenization and equivalent static loading reduce the computational cost of simulation-informed evaluations. Furthermore, a multi-sized unit cell approach is proposed to enable lattice support structure design for complex geometries, addressed in the last work. The problem is formulated to find the optimal configuration to dissipate heat while considering structural integrity and manufacturing costs (e.g., material and post-processing costs). The method is validated through several case studies, including a cantilever beam, aerospace bracket, and heat exchanger adapter. Within each case study, consistently obtained increased heat dissipation is accomplished and compared to the uniformly distributed benchmark design while still satisfying manufacturing constraints. For the case study of a cantilever beam and aerospace bracket, up to 61% reduced material cost and 62% reduced post-processing costs are achieved, while satisfying constraints. For the intricate design of the heat exchanger adapter, at least 50% of material cost savings are achieved which is about 19% of the total cost of the entire build. Overall, this dissertation seeks to help users design complex LBPF geometries with customizable support structure properties. It will also advance the current research within a range of disciplines, including design, optimization, and AM of lattices.