Abstract:
Additive manufacturing (AM) is a rapidly growing process for printing materials in three-dimensional objects layer-by-layer in a predefined digital path, including complex design features with engineered properties. The large number of process parameters, including materials thermophysical properties, feedstock rate, complex geometry, filling pattern, and preheat build temperature, affects the structural and mechanical properties of printed specimens. Hence, the rejection of printed specimens is a severe issue due to the distortion of parts due to residual stresses, which increases the overhead costs. The performance and properties of final parts depend on numerous metallurgical variables like thermal cycles, molten pool shape and size, thermal gradient, cooling rates, and solidification rates due to the selection and optimization of process parameters. Here, mechanistic modeling is important to understand the physics of metallurgical variables, however, there is no guarantee to reduce the number of trial experiments to obtain high-performance parts. Therefore the process, structure, properties, and performance of build parts need to be understood for a minimal number of trial experiments using the machine learning and mechanistic models. The understanding of build parts physics will improve the performance and properties, whereas machine learning will reduce the process parameters in the experiment. The synergistic development of machine learning algorithms with mechanistic models will also open opportunities for printing new future AM materials and alloys. The machine learning technique will improve the design, production, and performance of build parts to reduce cost and time.