Additive Manufacturing Process-Informed Topology Optimization
Published: 2023
Publication Name: University of Maine ELECTRONIC THESES AND DISSERTATIONS
Publication URL: https://digitalcommons.library.umaine.edu/etd/3908/
Abstract:
Topology optimization (TO) often results in complex part geometries that are much better suited to additive manufacturing (AM) than traditional manufacturing processes. However, large thermal-gradient AM processes, such as powder bed fusion (PBF), exhibit process-dependent microstructures and properties that are ignored in process-agnostic TO solutions, thus, excluding valuable process-structure-property relationships from the purview of optimum design search and raising doubts about the validity of the TO results. This work seeks to address this shortcoming by developing and implementing a process-informed topology optimization (PITO) algorithm for PBF AM. The iterative PITO algorithm augments a density-based TO search technique with a process simulation model to calculate as-solidified porosity. The design density field is polarized to form a discrete geometry in a manner that reflects the common industry practice of isodensity surface extraction, and the discrete geometry is then translated into a set of manufacturing instructions as inputs to a process simulation model. Spatially-dependent porosities in manufactured regions of the PBF build volume are calculated as a function of scan strategy, beam diameter, beam power, and thermomechanical properties of powdered 316L Stainless Steel. The position- and process-dependent porosity data are utilized to calculate isotropic elastic moduli via a Mori-Tanaka homogenization scheme. Direct update of design variables, where porosity data are used to initialize the next position in the design domain, and indirect update of design variables, where element-wise mechanical properties are updated but position in the design domain is preserved between iterations, are implemented and their effects contrasted. A set of design envelopes, boundary conditions, manufacturing and response constraints, objective functions, and manufacturing orientations are chosen and presented to convey the impact of including processing outcomes into the PITO algorithm. Where process-agnostic TO results exhibit symmetry, PITO design solutions are parameter-dependent and exhibit asymmetry in the build direction. Importantly, a numerical comparison of process-agnostic TO and PITO solutions shows improved accuracy between as-designed and as-manufactured performance. This work contains two significant contributions. First, this work demonstrates that PITO solutions more accurately predict performance than traditional TO solutions. Second, this work demonstrates an approach to incorporate the manufacturing processes into part design, providing a framework for integrated process-product optimization toward improved performance.