Page 63 - IJAMD-2-2

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

Section 2 reviews related work; Section 3 redefines the and industrial application areas. On the other hand, Zhang
role of industrial AI; Section 4 introduces the proposed et al. developed a comprehensive reference framework
15
foundation framework; Section 5 provides a case study on structured around seven critical dimensions (object,
rotating machinery diagnosis; Section 6 discusses future domain, stage, requirement, technology, function, and
directions; and Section 7 concludes the paper. solutions). In addition, Yang et al. proposed an intelligent
16
manufacturing framework combining human-machine
2. Related work cooperation with autonomous intelligent control systems
17
This section reviews existing studies and frameworks specifically tailored to the process industry. Ahmed et al.
focusing on the integration of AI within various industrial highlighted AI and explainable AI (XAI) methodologies,
applications. Several research works have proposed categorizing various XAI methods and their applicability
1
frameworks aimed at enhancing industrial productivity, in Industry 4.0 contexts, while Jan et al. structured a four-
decision-making, and operations by leveraging AI stage AI data pipeline, outlining data acquisition/validation,
capabilities. Table 1 provides a clear summary of these data processing/fusion, model training/testing, and model
3
previous industrial AI frameworks, highlighting their interpretation. Furthermore, Leng et al. introduced a four-
primary contributions. For instance, Lee et al. introduced layer technical reference framework encompassing layers
4
an industrial AI ecosystem framework that integrates from hardware to industrial applications. Most recently,
7
enabling technologies, such as data technology, analytic Lee and Su proposed a unified industrial large knowledge
technology, platform technology, and operation technology, model (ILKM) framework emphasizing domain-specific
within the established CPS (5C) architecture, guided by knowledge integration through LLMs and machine
key ABCDE elements. Building on this foundation, Peres learning (ML) approaches.
et al. expanded the framework into a more comprehensive Although these frameworks have successfully identified
2
conceptual framework, emphasizing challenges, design important aspects of AI integration across different
principles, essential technologies, capabilities, attributes, industrial applications, several limitations remain.

Table 1. Summary of existing frameworks for industrial artificial intelligence applications
Authors Year Proposed industrial artificial intelligence (AI) frameworks
Lee et al. 4 2018 Proposed an industrial AI ecosystem framework to integrate enabling technologies – data technology, analytic technology,
platform technology, and operation technology – within the Cyber-Physical System (5C) architecture. The framework
systematically guides AI implementation in smart manufacturing under the guidance of five key ABCDE elements: analytics
technology, big data technology, cloud or cyber technology, domain know-how, and evidence.
Zhang et al. 15 2019 Presented a comprehensive industrial AI reference framework consisting of seven key dimensions: Object (who), domain
(where), application stage (when), application requirement (why), intelligent technology (which), intelligent function
(what), and solutions (how); and offers a detailed overall planning for systematically integrating AI across diverse industrial
scenarios.
Peres et al. 2 2020 Proposed a conceptual industrial AI framework highlighting essential enabling technologies (data, analytics, platforms,
operations, and human-machine interaction) while systematically identifying critical challenges, attributes, capabilities,
design principles, and common application domains in Industry 4.0.
Yang et al. 16 2021 Presented a two-tier intelligent manufacturing framework designed for the process industry, integrating human-machine
cooperation, and intelligent autonomous control systems to achieve intelligent optimal decision-making.
Ahmed et al. 17 2022 Conducted a comprehensive survey on AI and explainable AI (XAI) methodologies within Industry 4.0, categorizing
various XAI approaches (model-specific, model-agnostic, local/global, visualization-based methods, and surrogate models),
highlighting their applicability, benefits, and challenges in industrial contexts.
Jan et al. 1 2023 Proposed a structured, four-stage industrial AI data pipeline (data acquisition/validation, data processing/fusion, model
training/testing, and model interpretation). They identified common themes, issues, and industry-specific solutions related to
AI integration in various sectors, providing insights into practical challenges and opportunities within Industry 4.0.
Leng et al. 3 2024 Presented a technical reference framework structured in four layers (hardware infrastructure, computing engine, AI
algorithms, and industrial application layer together with related empowering technologies across layers), identifying three
core opportunities (collaborative intelligence, self-learning intelligence, and crowd intelligence) crucial for realizing Industry
5.0’s vision of human-centric, resilient, and sustainable manufacturing.
Lee and Su 7 2024 Proposed a unified industrial large knowledge model framework consisting of four systematic steps: (i) construction of a
large knowledge library; (ii) preparation of domain-specific instruction data; (iii) development of domain-specific knowledge
large language models; and (iv) establishment of intelligent domain expert machine learning systems; guided by the “6S
Principle” (specialized knowledge, scrutability, safety, scalability, sustainability, and systematization).

Volume 2 Issue 2 (2025) 57 doi: 10.36922/IJAMD025080006