Page 65 - IJAMD-2-2

International Journal of AI for
Materials and Design
A unified industrial AI foundation framework

Figure 2. A unified industrial artificial intelligence (AI) foundation framework. The industrial large knowledge model framework image is reprinted
with permission from Lee and Su. The stream of quality image is reprinted with permission from Lee et al. Copyright © 2022 Society of Manufacturing
20
7
Engineers (SME). The 5C-cyber-physical system figure is reprinted with permission from Lee et al. Copyright © 2018 SME. The digital twin image is
4
reprinted with permission from Lee et al. Copyright © 2020 The Institution of Engineering and Technology.
19
Abbreviations: LLM: Large language model; ML: Machine learning.
illustrates the overall structure of the proposed framework, forms, such as reports, manuals, research papers,
highlighting the interconnections and dynamic feedback and maintenance logs. Rather than reflecting an
between knowledge, data, and model modules. Each pair anthropocentric design philosophy in the traditional sense,
of modules is connected by forward and backward arrows, the knowledge module aims to transform these scattered
highlighting that these modules continuously inform and resources into structured, accessible, and reusable formats
refine one another, rather than existing in a linear sequence. that guide data analytics and model development – serving
In addition, forward arrows from all three modules point as a foundation that enables scalable, automated, and
toward the industrial AI methodology platform, indicating context-aware AI system development. It focuses on the
that the platform is strengthened by these three modules. following four aspects.
In Sections 4.1-4.4, the details of the knowledge, data, and
model modules are outlined, followed by an explanation 4.1.1. Knowledge extraction
of how they enhance and extend the capabilities of the Knowledge extraction refers to the systematic process of
industrial AI methodology platform. converting unstructured information into structured,
searchable knowledge representations. This is essential
4.1. Knowledge module for enabling automated reasoning and guiding data
The knowledge module focuses on capturing, organizing, preprocessing and model design. Techniques such as
and utilizing domain knowledge to support the entire knowledge graph construction 21-23 and LLM-assisted
24
industrial AI development cycle. In complex industrial information extraction are commonly used. For instance,
settings, domain expertise often exists in unstructured tools such as OpenKE, TransOMCS, and gBuilder can
26
27
25
Volume 2 Issue 2 (2025) 59 doi: 10.36922/IJAMD025080006