Page 48 - IJAMD-1-2

International Journal of AI for
Materials and Design
A unified ILKM in smart manufacturing

Recognizing this gap, an industrial large knowledge process of assessment and feedback is integral to ensuring
model (ILKM) framework is proposed for domain-driven, the relevance and effectiveness of ILKM solutions. Overall,
data-centric industrial systems in Industry 4.0 and smart ILKMs underscore the transformative potential of data-
manufacturing. In addition, the “6S Principle” is proposed driven approaches, offering detailed and comprehensive
as a guideline for the development of ILKMs. The role of optimization and enhancement directions for industrial
ILKMs and their comparison with LLMs are discussed in products in Industry 4.0 and smart manufacturing.
detail. Through this exploration, this paper aims to provide
a comprehensive understanding of the transformative 3. ILKM framework
power of ILKMs in the modern manufacturing landscape, The proposed ILKM framework, shown in Figure 2,
highlighting their significance and opportunities as a provides a step-by-step guideline for developing and
cornerstone of the ongoing industrial revolution. deploying ILKMs using industrial data to enhance

2. The role of ILKMs in Industry 4.0 and manufacturing capabilities in areas such as predictive
maintenance, process optimization, quality control,
smart manufacturing engineering design, question-answering(QA) platforms,
In Industry 4.0 and smart manufacturing, the deployment and data analytics. The ILKM framework consists of four
of ILKMs emerges as a pivotal element. Figure 1 shows pivotal steps: (i) the construction of an LKL categorized by
the general process of how ILKM works within Industry human-interpretable and structured machine-generated
4.0, where ILKM functions at the core of this advanced data; (ii) the preparation of domain-specific instruction
manufacturing paradigm. The process begins with the data; (iii) the development of a domain-specific knowledge
acquisition and management of a vast array of industrial LLM based on the domain-specific data and domain
data, derived from diverse industrial products. 10,11 This instruction data; and (iv) the establishment of an intelligent
data are categorized into two primary forms: human- domain expert ML system. As illustrated in Figure 2,
interpretable data and structured machine-generated data. the details of the ILKM framework are outlined in the
Leveraging technologies such as LLMs, a comprehensive subsequent Sections 3.1–3.4.
large knowledge library (LKL), along with various ML 3.1. Large knowledge library construction
and AI techniques, ILKMs serve as artificial general
12
intelligence, which plays a vital role in enabling advanced The initial step in constructing an ILKM involves the
13
and sophisticated data analytics and problem-solving. These creation of an LKL. This library is pivotal for accommodating
advanced analytical capabilities, therefore, pave the way for the breadth and diversity of industrial data, thus serving
more insightful and informed decision-making processes. as a foundational resource for subsequent analytical
Beyond this, ILKMs can also interface with and improve tasks. During this phase, it is essential to categorize the
supply chain management, leading to more efficient, data into domain-specific categories systematically. Such
14
resilient, and customer-focused operations. In addition, an organization enables researchers and data scientists
the solutions generated by ILKMs undergo evaluation by to streamline their efforts, allowing for efficient retrieval
subject matter experts, who play a crucial role in validating of domain-specific data to inform the development of
and refining the relevant solutions, thereby aiding in the ML models tailored to address distinct industry-related
continual optimization of ILKM outputs. This iterative challenges. Within these categories, based on the usage

Figure 1. General process in Industry 4.0 and smart manufacturing using industrial large knowledge model
Abbreviations: AI: Artificial intelligence; LKL: Large knowledge library; LLM: Large language model; ML: Machine learning.

Volume 1 Issue 2 (2024) 42 doi: 10.36922/ijamd.3681