Page 75 - ESAM-1-1

Engineering Science in
Additive Manufacturing ML in additive manufacturing

Figure 7. First application of an ML model or algorithm in AM. The vertical axis contains a combination of algorithm names and types extracted as unique
string terms from research articles at the intersection of ML and AM. The parent algorithm categories are highlighted in red to mark the onset of enhanced
learning capacity through model complexity while also introducing data and computation challenges.
Abbreviations: AM: Additive manufacturing; ML: Machine learning.

Over the past decade, the majority of AM concerns Small-scale DL architectures (FFNNs, CNNs) remain
modeled through ML were related to the AM process popular for modeling AM tasks. Similarly, linear and tree-
chain (per-process, in-process, and post-process phases). based models for classification and regression continue to
Processing technologies such as LPBF, DED, and MEX have be effective for simple learning tasks.
become more mature in the application of ML techniques as
compared to other AM process technologies (e.g., material 5. Advanced approaches
jetting and vat photopolymerization). Vision and sequence
data modalities from AM processes have been widely This section reviews advanced approaches in AI-driven
utilized in applications concerning in situ predictions AM that address both data and model aspects. These
of process defects and product qualities. This progress methods aim to improve data quality for enhancing model
provides a strong basis to enhance and adapt the learning performance or to design ML architectures guided by
algorithms in the industrial and production environments. data engineering and domain expertise. With growing
Figure 8 represents the overall share of major model and awareness of data-centric modeling, researchers have been
algorithm categories in relative and absolute comparison. implementing and developing data integration, knowledge

Volume 1 Issue 1 (2025) 8 doi: 10.36922/ESAM025040004