Page 44 - IJAMD-2-2

International Journal of AI for
Materials and Design ML-driven optimization in additive manufacturing

A B

C

D

Figure 4. Property and design optimizations for polymer 3D printing. (A) Hierarchical machine learning framework integrating experimental knowledge to
reduce data-driven discovery efforts. Reproduced with permission from Bone et al. Copyright © 2020 American Chemical Society. (B) Machine learning-
111
assisted extrusion printing workflow for thermoelectric inks with four input variables and three output properties. Reproduced with permission from Song
et al. Copyright © 2024 Royal Society of Chemistry. (C) Experimental workflow for synthesizing and screening sub-microliter-scale photodegradable
114
hydrogels. Reproduced with permission from Seifermann et al. Copyright © 2023 Wiley. (D) Overview of voxel-level inverse design of 4D-printed active
116
composite plates enabled by machine learning. Reproduced with permission from Rodriguez and Goodman. Copyright © 2024 Springer Nature.
122
In their subsequent study, the same authors extended also present a combinatorial explosion of possible design
116
this approach to active composite plates. While plate variables. To handle this complexity, the authors combined
architectures offer higher design freedom and enable more ML with gradient descent (GD) and EA for the inverse
intricate shape transformations than beam structures, they design of active plates. A ResNet-based ML model was

Volume 2 Issue 2 (2025) 38 doi: 10.36922/IJAMD025130010