Exploring the Fusion of Knowledge Graphs into Cognitive Modular Production

Abstract

Modular production has been recognized as a pivotal approach for enhancing productivity and cost reduction within the industrialized building industry. In the pursuit of further optimization of production processes, the concept of cognitive modular production (CMP) has been proposed, aiming to integrate digital twins (DTs), artificial intelligence (AI), and Internet of Things (IoT) technologies into modular production systems. This fusion would imbue these systems with perception and decision-making capabilities, enabling autonomous operations. However, the efficacy of this approach critically hinges upon the ability to comprehend the production process and its variations, as well as the utilization of IoT and cognitive functionalities. Knowledge graphs (KGs) represent a type of graph database that organizes data into interconnected nodes (entities) and edges (relationships), thereby providing a visual and intuitive representation of intricate systems. This study seeks to investigate the potential fusion of KGs into CMP to bolster decision-making processes on the production line. Empirical data were collected through a computerized self-administered questionnaire (CSAQ) survey, with a specific emphasis on exploring the potential benefits of incorporating KGs into CMP. The quantitative analysis findings underscore the effectiveness of integrating KGs into CMP, particularly through the utilization of visual representations that depict the relationships between diverse components and subprocesses within a virtual environment. This fusion facilitates the real-time monitoring and control of the physical production process. By harnessing the power of KGs, CMP can attain a comprehensive understanding of the manufacturing process, thereby supporting interoperability and decision-making capabilities within modular production systems in the industrialized building industry.

1. Introduction

In recent years, the fusion of Cyber–Physical Production Systems (CPPSs) with cognitive technologies has enabled a new era of intelligent manufacturing [1,2]. The CPPS is a concept that refers to the fusion of physical systems (such as machines and equipment) with digital systems (such as sensors, data analytics, and control systems) in the production process [3,4]. The CPPS allows for the real-time monitoring, control, and optimization of the production process. The incorporation of cognitive technologies, like artificial intelligence (AI) and machine learning (ML), within the CPPS has allowed for the creation of intelligent agents that can adapt and learn from their environment, optimizing production processes and reducing costs [5]. However, to fully realize the potential of cognitive technologies in the CPPS, the seamless fusion of various modules is crucial. This is where knowledge graphs (KGs) come into play.

KGs have become an important tool for organizing and managing data in a way that allows for the fusion of different modules in a seamless manner [6,7]. KGs are structured representations of knowledge that capture relationships between entities, enabling the creation of intelligent agents that can reason and learn from data [8]. Cognitive digital twins (CDTs), introduced by Eirinakis et al. [9] and Abburu et al. [10], are virtual replicas of physical systems or processes that are created by integrating data from various sources, such as sensors, machine logs, and historical data. CDTs are designed to simulate and predict the behavior of the physical system and can be used to identify inefficiencies or problems in the production process [11,12]. KGs have the potential to be incorporated into CDTs as a way of organizing and representing the data that are used to create the virtual model. By structuring the data in a way that captures the relationships between different entities, a KG can improve the accuracy and reliability of predictions made by the CDT [13]. This can help manufacturers make better decisions and optimize their production processes based on the insights provided by the CDT [14]. In essence, CDTs with KGs can act as a decision support tool for manufacturers, allowing them to anticipate and mitigate potential problems, reduce downtime, and optimize their production processes [11,15,16].

Additionally, cognitive modular production (CMP) is an emerging approach to manufacturing that combines the benefits of CPPSs and CDTs to create a more intelligent and automated manufacturing system [6]. CMP involves breaking down the manufacturing process into smaller modular components, each of which can be optimized individually to improve the overall performance. This approach allows for greater flexibility and agility in the manufacturing process, as well as improved efficiency and reduced downtime. By using CDTs, CMP allows for virtual testing and the simulation of the modular components before they are deployed, while the CPPS enables the monitoring and control of the physical production process in real time [9,11].

The fusion of a KG into CMP has the potential to transform the way manufacturing processes are conducted. The ability to seamlessly integrate different modules and create intelligent agents that can reason and learn from a large amount of data can lead to a more efficient and effective production process. However, despite the potential benefits of integrating KGs into CMP, there is a lack of knowledge in terms of the understanding of the methods and techniques that can be used to effectively integrate KGs into CMP. Existing research has primarily focused on the individual components of CMP, such as the use of AI and the Internet of Things (IoT) for autonomous operations, but the potential benefits of incorporating KGs into CMP have not been thoroughly investigated. KGs offer a unique approach to organizing and representing complex data, which could significantly enhance the decision-making processes in CMP systems [6,17,18,19]. However, there is a lack of empirical evidence on the effectiveness of integrating KGs into CMP and the potential benefits that this fusion could bring to the industrialized building industry. Hence, the purpose of this study is to investigate the potential fusion of KGs into CMP to support decision making on the production line. The research questions include the following:

-

How can CDT applications be utilized to support the production line in the CMP process?

-

How can KGs be integrated into CDTs and used to structure and represent data in the real-time view of the CMP process?

-

What are the opportunities that KGs can provide for production optimization, allowing proactive decision making in the CMP process?

The research questions for this study were initially formulated through a comprehensive scoping review of the relevant literature in the field to identify key themes, trends, and gaps in the current knowledge. Thereafter, the questions were refined by incorporating insights and experts’ perspectives gained from the survey, which served as the primary research method for this study. This integration not only enhanced the clarity of the research questions but also sharpened their focus. A more detailed account of the survey’s execution is provided in the Section 3 of this study.

In the Section 2, this research paper introduces the theories that underpin this study. Following this, Section 3 outlines this study’s design, the materials incorporated, and the procedures implemented. Section 4 critically examines the accumulated data, while the succeeding Section 5 delves into the implications of the findings. Lastly, this research paper concludes by presenting the outcomes drawn from this study. The research methodology is presented in Figure 1.

2. Theoretical Background

2.1. Cognition in Modular Production

Modular production is a manufacturing approach that involves creating a product by assembling standardized, pre-made components or modules. These modules are designed to fit together seamlessly and can be easily interchanged, allowing for flexibility and customization in the production process [19,20]. The use of modular production can lead to increased efficiency, reduced costs, and an improved product quality. Additionally, the ability to quickly modify and adapt production systems to changing market demands makes modular production an attractive option for manufacturers in a variety of industries [16].

Cognitive modular production (CMP) refers to the fusion of cognitive abilities, such as perception, attention, memory, and decision making, into the production process to enhance its efficiency and quality. This approach involves the fusion of AI, ML, and the IoT into modular production systems to create intelligent production modules and as a result improve the speed, accuracy, and adaptability of the production process [21]. Furthermore, modular production systems often lack standardization, which can make it difficult to integrate cognitive processes in a standardized and consistent manner. Another obstacle to integrating cognition into modular production systems is the need for specialized knowledge and expertise in both cognitive science and modular production systems [22,23].

2.2. CDT Application on the Production Line of Modular Production Systems

CDTs are a relatively new technology that combines the concepts of digital twins (DTs) [24] and cognitive computing [25]. CDTs have the potential to revolutionize the way decisions are made on the production line of modular production systems by enabling real-time decision making based on the analysis of large amounts of data [9]. CDTs can provide a real-time view of the production process and its performance, enabling operators to make informed decisions based on the current state of the system [26]. They can also provide insights into potential issues or opportunities for optimization, allowing for proactive decision making rather than reactive responses to problems. Additionally, CDTs can be used to simulate different scenarios, enabling operators to evaluate the impact of potential decisions before implementing them on the production line [23].

One application of CDTs in modular production systems is in predictive maintenance. By analyzing data from sensors and other sources, a CDT can predict when a machine or component is likely to fail, allowing operators to schedule maintenance proactively and minimize downtime [15,27]. This can lead to significant cost savings and increased productivity. Another application of CDTs is in quality control. By analyzing data from sensors and cameras, a CDT can detect defects in real time and provide feedback to operators. This can enable operators to quickly identify and resolve issues, reducing scrap and reworkings [23].

CDTs can also be used to optimize the production process. By analyzing data on factors such as machine utilization, production rates, and energy consumption, a CDT can identify opportunities for optimization and provide recommendations to operators. This can lead to improved efficiency and reduced costs [21,28]. However, CDT technology is still in early stages and has limited application in modular production systems. There are some challenges to implementing CDTs in modular production systems. One challenge is the availability and quality of data. CDTs rely on high-quality data to provide accurate predictions and recommendations. If data are incomplete, inaccurate, or unavailable, the performance of the CDT can be compromised [23]. Another challenge is the complexity of the production system.

Modular production systems often involve a large number of components and processes that must be coordinated. This can make it difficult to develop a CDT that accurately models the system and provides meaningful insights. To address these shortcomings, researchers and practitioners in the field of CDTs are working on developing standardized data formats and interfaces, as well as addressing issues around data privacy and security. As a comprehensive summary,

Table 1. Literature on cognitive digital twins.

n	Author(s)	References	Year	Industry	Applications
1	Kalaboukas et al.	[29]	2023	Engineering and management	Emphasized the concept of supply chain CDTs and proposed a holistic governance approach.
2	Yitmen et al.	[6]	2023	AEC industry	Presented an IIoT-enabled KG-based data representation approach for cognitive modular production and CDT.
3	Yitmen and Alizadehsalehi	[7]	2023	AEC industry	Studied Construction 4.0-based DTs and examined the CDT’s vision and characteristics.
4	Yitmen and Alizadehsalehi	[13]	2023	AEC industry	Reviewed previous studies pertinent to the CDT concept, its definitions, its characteristics, and its required technologies.
5	Unal et al.	[18]	2022	Engineering	Presented a Digital Twin Pipeline Framework of the COGNITWIN project that supports hybrid and cognitive digital twins, through four Big Data and AI pipeline steps adapted for digital twins.
6	D’Amico et al.	[30]	2022	Manufacturing	Reviewed the semantic digital twins in the maintenance context.
7	Zheng et al.	[27]	2022	Manufacturing	Discussed the emergence of the cognitive digital twin.
8	Rožanec et al.	[31]	2021	Manufacturing	Aimed to capture specific knowledge related to demand forecasting and production planning.
9	Berlanga et al.	[32]	2021	Computer science	Proposed a platform for social networks.
10	Yitmen et al.	[26]	2021	AEC industry	Investigated the applicability, interoperability, and integrability of an adapted model of CDTs for BLM (Building Lifecycle Management).
11	Kalaboukas et al.	[17]	2021	Manufacturing	Implementation of CDT in connected and agile supply networks.
12	Abburu et al.	[33]	2020	Engineering	Proposed a framework for the implementation of hybrid and cognitive twins as part of the COGNITWIN toolbox.
13	Zhang et al.	[34]	2020	Computer science and engineering	Discussed how the different levels of self-awareness can be harnessed for the design of CDTs.
14	Du et al.	[35]	2020	AEC industry	Established methods and tools for the intelligent information systems of smart cities.
15	Eirinakis et al.	[9]	2020	Management	Proposed enhanced cognitive capabilities for the DT artifact that facilitate decision making.
16	Albayrak and Ünal	[36]	2020	Engineering	Smart steel pipe production plant via CDT-based systems.
17	Abburu et al.	[10]	2020	Engineering	Proposed a CT control system for automation in the process control system.
18	Essa et al.	[37]	2020	Computer science	Introduced the automation of defect detection.
19	Saracco	[38]	2019	Computer science	Proposed to bridge physical space and cyberspace.
20	Fernández et al.	[39]	2019	Engineering	Introduced the concept of an associative CDT, which explicitly includes the associated external relationships of the considered entity for the considered purpose.

presents the diverse applications of CDTs in different industries based on the latest research (2019–2023).

2.3. Establishment of Production Line and CDT and Their Implementation

In order to integrate CDTs into the physical system, it is important to have a clear understanding of the system’s architecture and components. This can involve creating a detailed schematic of the system and mapping out the flow of materials, information, and energy through the system [15]. Once the digital model has been created, data must be collected to create the DT. These data may come from a variety of sources, including historical performance data, real-time sensor data, and information on the performance characteristics of different components [28]. The DT can then be used to simulate different scenarios and test potential changes or improvements to the system. For example, the DT could be used to simulate the impact of introducing a new machine or process into the production environment or to test the impact of changing the layout of the production line.

In a modular production environment, implementing a production line system that incorporates CDTs involves integrating the DT with the physical system. This may involve installing sensors and other monitoring devices to collect data on the performance of different components and processes and integrating these data into the digital twin in real time [40].

One of the key benefits of using CDTs in a modular production environment is the ability to quickly reconfigure the system to accommodate changing production needs. For example, if a particular product is in high demand, the system can be reconfigured to prioritize the production of that product while minimizing the impact on other products being produced on the same line. To support this level of flexibility, it is important to have a modular system architecture that allows for easy reconfiguration and adaptation. This may involve the use of standardized components and interfaces, as well as the use of flexible manufacturing cells that can be easily reconfigured to support different production requirements [28].

2.4. KG Representation for CMP

In recent years, the use of KGs has become increasingly popular, particularly in the field of artificial intelligence and data science. KGs provide a way to represent and organize complex information and knowledge in a way that is both intuitive and computationally tractable [8]. In the context of CMP systems, KGs can be used to capture information about the system’s components, processes, and interactions, as well as the relationships between them. To create a KG for a CMP system, one would first need to identify the relevant entities and relationships within the system. This might include information about the physical components of the system (e.g., machines, sensors, and actuators), the processes that take place within the system (e.g., assembly, machining, and inspection), and the various data streams that are generated by the system (e.g., sensor readings, control signals, and production data) [11,27,41].

Once the relevant entities and relationships have been identified, they can be represented as nodes and edges in the KG. Each node would represent a specific entity within the system and each edge would represent a relationship between two entities (e.g., a machine is connected to a production line or a sensor provides data to a control system).

By organizing information in this way, a KG has the potential to provide a powerful tool for decision making and interoperability within a CMP system. For example, one could use a KG to identify potential inefficiencies in the production process or to optimize the allocation of resources within the system. KGs can also be used to support interoperability between different components or systems, by providing a common framework for communication and data exchange [8].

Another usage of KGs is to support predictive maintenance, where the system can use the data collected from sensors and other sources to predict when maintenance may be required for a particular component. This can help to reduce downtime and increase the overall efficiency of the system. Additionally, KGs can be used to support traceability and quality control by tracking the movement of components and materials throughout the production process and recording any quality issues that may arise [27].

Table 2. Literature on knowledge graphs.

n	Author(s)	References	Year	Industry	Applications
1	Peng et al.	[42]	2023	Science and sustainability	Presented a systematic overview of knowledge graphs.
2	Yitmen and Alizadehsalehi	[6]	2023	AEC industry	Presented an IIoT-enabled KG-based data representation approach for cognitive modular production and CDTs.
3	Milošević et al.	[43]	2023	Biomedical	Discussed the comparison of biomedical relationship extraction methods and models for knowledge graph creation.
4	Zhou et al.	[44]	2023	Engineering	Proposed a semantic-aware event link reasoning over an industrial knowledge graph embedding time series data.
5	Tiddi et al.	[45]	2022	Computer science	Provided an extensive overview of the use of knowledge graphs in the context of explainable machine learning.
6	Tiwari et al.	[46]	2021	Computer science	Presented a characterization of different types of KGs along with their construction approaches.
7	Sun et al.	[47]	2020	Computer science	Presented a model that incorporates multi-modal knowledge graphs into recommender systems.
8	Rasmussen et al.	[48]	2019	Construction industry	Discussed the AEC knowledge graphs.

shows the recent literature on KGs.

3. Materials and Methods

A scoping review was carried out to identify and categorize the fundamental principles, main sources, and various forms of evidence that are pertinent to this study [49]. The purpose of this review was to establish the theoretical background for the current study. This was followed by a computerized self-administered questionnaire (CSAQ) survey in order to collect data regarding the analysis of the factors involved for integrating KGs into CMP.

3.1. Scoping Review

The scoping review was conducted in Scopus and Google Scholar, searching in the keywords, abstract, and title. The following search strings were used:

-

(“Digital Twin” OR “DT”) AND (“Modular Production” OR “Cognitive Modular Production” OR “CMP”) AND (“Knowledge Graph” OR “KG”).

A filter was applied to the initial searches to further narrow the search. The following was included in the filter: year of publication: 2016–2023, language: English, and document type: journal articles and conference papers. The information flow of the scoping review process is presented in Figure 2 based on the PRISMA-ScR (Preferred Reporting Items for Systematic re-views and Meta-Analyses extension for Scoping Reviews) [50].

3.2. Sampling and Data Collection

As primary data to collect and analyze for the research, a computerized self-administered questionnaire (CSAQ) survey based on the scoping review is offered to bolster the findings of this study. The survey focusing on a methodological approach for the analysis of the factors involved in the potential fusion of KGs into CMP was distributed to private experts and organizations operating in Europe, Scandinavia, the Far East, the Middle East, and North America. The questionnaire targeted key stakeholders involved in building projects, including modular production/manufacturing firms, building construction firms, and IT consultancy firms. The participants represented a diverse range of roles and positions within these organizations, ensuring a holistic perspective on the topic. To reach out to the desired participants, a professional network platform (LinkedIn) was used to contact a total of 250 industry specialists. Prior to their involvement, the contributors were duly informed about the objectives of this study, and strict measures were taken to ensure the privacy and anonymity of their responses. The survey managed to collect a total of 87 completed responses, accounting for 35 percent of the total attempted contacts. These respondents were experts in their respective fields and were specifically asked to provide insights and evaluations concerning their work experiences, observations, and organizational contexts. To gauge their level of agreement, the participants utilized a five-point Likert scale, where a rating of one denoted strong disagreement and a rating of five indicated strong agreement, in relation to statements derived from the literature as an outcome of the scoping review.

Table 3. Questionnaire respondents: Overview.

Company Type	Modular Production/Manufacturing		Industrialized Building Construction		Consultancy in Construction		IT in Construction
	Role	%	Role	%	Role	%	Role	%
	Production manager	9%	Project manager	9%	Design manager	8%	Digitalization specialist	11%
	Design engineer	5%	Site manager	8%	BIM manager	9%	Systems analyst	4%
	Production supervisor	7%	Construction manager	9%	BIM coordinator	12%	Data analyst	4%
	-	-	-	-	Business development manager	5%	-	-
Company Size
Small (<50)	6%		6%		6%		6%
Medium (50–250)	9%		9%		9%		9%
Large (>250)	12%		12%		5%		5%
Operating Region
Scandinavia	6%		3%		8%		6%
West Europe	5%		5%		9%		7%
N. America	4%		9%		2%		4%
Far East	3%		2%		3%		3%
Middle East	7%		6%		3%		5%

presents the characteristics of the participants who responded. In addition, the statistical analysis program SPSS has been utilized to analyze the questionnaire outcomes. The table illuminates distinct technology-driven operating regions spanning diverse geographical areas. Scandinavia, where the countries Sweden, Denmark, Norway, and Finland flourish with Nordic charm and innovation, constitutes a dynamic hub for technological advancements. West Europe, home to technological leaders, such as Germany, Italy, France, and the Netherlands, showcases a rich tapestry of digital cultures and innovative economies. North America, powered by the technical prowess of Canada and the USA, stands as a global technology powerhouse offering boundless opportunities. The Far East, encompassing tech-driven giants, like Hong Kong, South Korea, and China, propels groundbreaking technological innovation and economic growth, shaping a pivotal aspect of the table. Lastly, the Middle East, epitomized by Dubai’s technological modernity amid rich traditions, forms an intriguing operating region at the crossroads of tech and culture. Each of these technology-laden regions presents unique challenges and prospects, contributing to a comprehensive global tech landscape for strategic considerations.

3.3. Data Analysis

The acquired data were summarized, and their initial trends were identified through descriptive statistics. Principal component analysis was then utilized for factor analysis. Measures of sampling adequacy were computed to assess the data’s suitability for factor analysis. The internal consistency of each factor was also assessed using Cronbach’s alpha. Subsequently, relationships among key study components were validated through Spearman’s rank-order correlation. An in-depth discussion was carried out to interpret these statistical findings, facilitating comparisons with the existing literature and suggesting implications for future research and industry applications.

3.4. Bias Mitigation Measures

To address the potential sources of bias within this study, we ensured that our CSAQ survey was designed to be representative of the target population, mitigating selection bias. In terms of performance bias, we ensured that all groups received the same level of care or exposure. To counter publication bias, we conducted a comprehensive literature search in different databases and also tried to carefully design our study using randomization to minimize the study design bias.

4. Results

4.1. Descriptive Statistics

Table 4. Descriptive statistics, factor analysis, and reliability test. Five-point Likert scale (1 = strongly disagree, 5 = strongly agree).

	Questionnaire Statement	Mean	SD	Median	Loadings	Cronbach α	Rank
Cognitive Digital Twins (CDTs)	CDTs can analyze enormous volumes of data in real time to transform modular production system production line for decision-making.	3.13	1.09	3	0.713		15
	CDTs can provide a real-time view of the production process and its performance, enabling operators to make informed decisions based on the current state of the system.	3.46	1.17	3	0.842		6
	CDTs can be used to simulate different scenarios, enabling operators to evaluate the impact of potential decisions before implementing them on the production line	3.58	1.16	4	0.722	0.766	17
	By analyzing data from sensors and other sources, a CDT can predict when a machine or component is likely to fail, allowing operators to schedule maintenance proactively and minimize downtime.	3.64	0.97	4	0.843		5
	By analyzing data on factors such as machine utilization, production rates, and energy consumption, CDTs can identify opportunities for optimization and provide recommendations to operators	3.92	1.14	4	0.712		14
Knowledge Graphs (KGs)	KGs capture information about the system’s components, processes, interactions, and relationships enabling the creation of intelligent agents that can reason and learn from data.	4.21	1.12	4	0.709		13
	KG is a tool to organize and manage data in a way that allows for the integration of different modules in a seamless manner.	4.34	0.99	4	0.814		10
	KG can be incorporated into Cognitive Digital Twins (CDTs) as a way of organizing and representing the data that is used to create the virtual model.	3.47	1.17	3	0.859	0.782	2
	KG can improve the accuracy and reliability of predictions made by the CDTs by structuring the data in a way that captures the relationships between different entities	3.56	0.93	4	0.707		11
	KG can help manufacturers make better decisions and optimize their production processes based on the insights provided by the CDT.	4.38	1.13	4	0.824		8
Cognitive Modular Production	Cognitive Modular Production (CMP) combines the benefits of Cyber Physical Production System and CDTs to create a more intelligent and automated production system	3.17	1.22	3	0.723		18
	CMP involves breaking down the production process into smaller modular components, each of which can be optimized individually to improve overall performance	3.39	0.97	3	0.847		4
	CMP allows for greater flexibility and agility in the production process, as well as improved efficiency and reduced downtime	3.68	1.13	4	0.708	0.778	12
	By using CDTs, CMP allows for virtual testing and simulation of the modular components before they are deployed	3.65	1.18	4	0.726		19
	Cyber Physical Production Systems (CPPS) enable the monitoring and control of the physical production process in real-time	3.54	1.11	4	0.885		1
Fusion of KG into CMP	Extraction of information of standards enable to gain a thorough understanding of the interoperability existing across the data sources describing standardization frameworks Industry 4.0.	4.18	0.98	4	0.823		9
	KG population using the Standard Ontology (STO) (concept of standards and standardization frameworks I4.0) enable to represent the data in a structured and easily understandable format, making it more accessible for analysis and integration with other KGs.	4.34	0.94	4	0.747		20
	KG integration defining connections between instances in KGs to resolve semantic interoperability conflicts enable to create a unified KG that can be used to support decision-making and optimize the production process.	4.37	1.14	4	0.852	0.793	3
	KG reasoning identifying patterns and relationships within the data and using this information to make inferences and predictions support to extract valuable insights from the data and enhance the capabilities of the fusion of of KG in CMP.	4.23	0.95	4	0.714		16
	KG interlinking (Linked Open Data Cloud) using unique identifiers and standard vocabularies to connect the different KGs and make them accessible to a wider audience support to enhance the interoperability and accessibility of the fusion of KG in CMP.	3.38	1.09	3	0.832		7

presents the descriptive statistics summarizing the responses obtained from the questionnaire survey. These statistics provide valuable insights into the perceptions of the architecture, engineering, and construction (AEC) industry regarding the analysis of the factors involved in the fusion of KGs into CMP. The mean values and standard deviations are reported, offering a comprehensive overview of the industry’s understanding and opinions on the concepts discussed.

According to the analysis of the questionnaire survey results, it was found that the mean scores for 14 out of the 20 questions exceeded 3.65 on a scale of 5.00. This indicates that industry professionals have shown strong support for the fusion of KG into CMP. The proposed model received an overall mean rating of 4.03, suggesting a high level of endorsement and indicating the industry’s inclination towards adopting the fusion of KG into CMP. One notable finding from the survey was the high mean rating of 4.38 for the statement that focused on the impact of KGs in providing manufacturers with better decision making and optimizing production processes. This statement aimed to assess the relative importance of variables such as decision-making support, learning, optimization, and reasoning. This finding highlights the industry’s desire to leverage advanced technologies to enhance decision-making processes, facilitate learning and adaptation, optimize operations, and improve reasoning abilities.

4.2. Factor Analysis and Reliability

Factor analysis was employed to identify the main dimensions within the variables of cognitive digital twins, knowledge graphs, cognitive modular production, and the fusion of KGs into CMP. Principal component analysis was conducted to empirically test and validate the variables. A summary of the outcomes can be found in Table 3. To assess the appropriateness of the data for factor analysis, overall and individual measures of sampling adequacy were calculated. Values above 0.5 are considered acceptable. The reliability of each extracted factor was evaluated using Cronbach’s alphas, which assess the internal consistency of the factors based on the average correlation between variables within each factor. A minimum acceptable value for Cronbach’s alpha is 0.7. The examination of Cronbach’s alpha values indicated that all reliability coefficients α for the constructs listed in Table 3 demonstrated acceptable levels of reliability. Some constructs exhibited higher reliability than others. Specifically, the constructs “Fusion of KG in CMP” and “Knowledge Graphs (KG)” exhibited the highest reliability coefficients α, with values of 0.793 and 0.782, respectively. The variables “KG can be incorporated into Cognitive Digital Twins (CDTs) as a way of organizing and representing the data used to create the virtual model” and “KG fusion defining connections between instances in KGs to resolve semantic interoperability conflicts enable to create a unified KG that can be used to support decision making and optimize the production process” demonstrated the highest factor loadings, with values of 0.859 and 0.852, respectively.

4.3. Correlation Analysis

Spearman’s rank-order correlation was employed to validate the relationships, and the matrix evaluation demonstrates a significant correlation. Notably, a positive linear relationship was observed between several key components: CDTs for real-time data analysis and optimization; KGs for data organization and decision optimization; CMP for intelligent, optimized, and flexible production; and the fusion of KGs into CMP for enhanced interoperability and production optimization. The highest correlation was found between CMP for intelligent, optimized, and flexible production and the fusion of KGs into CMP for enhanced interoperability and production optimization (ρ < 0.01, r = 0.852). A secondary significant positive correlation was observed between the CDT for real-time data analysis and optimization and CMP for intelligent, optimized, and flexible production (ρ < 0.01, r = 0.815). The correlation calculations pertaining to the respondents’ perception of the CDT’s decision support capabilities can be seen in

Table 5. Correlational analysis of cognitive digital twins’ perception of decision support capabilities.

	CDT for Real-Time Data Analysis and Optimization	KG for Data Organization and Decision Optimization	CMP for Intelligent, Optimized, and Flexible Production	Fusion of KG into CMP for Enhanced Interoperability and Production Optimization
CDT for real-time data analysis and optimization	1.000
KG for data organization decision optimization	0.675	1.000
CMP for intelligent, optimized, and flexible production	0.815	0.786	1.000
Fusion of KG into CMP for enhanced interoperability and production optimization	0.719	0.792	0.852	1.000

Notes: N = 87. Correlations have a (2-tailed) level of significance “Sig. < 0.000”. Correlation is significant at the 0.01 level.

.

5. Discussion

The motivation for this research stemmed from the recognition of the potential of KG technology and its future applications in the field of CMP systems. The lack of attention given to KGs in the literature related to CMP prompted the authors to investigate this area further. The aim of this study is to analyze the factors that contribute to the fusion of KGs into CMP, enabling stakeholders to leverage the benefits of knowledge-based decision making and optimization. By combining the strengths of KGs and CMP, this fusion could provide a promising solution to address the identified challenges.

5.1. Theoretical Contributions

5.1.1. Cognitive Digital Twins (CDTs)

The survey results show that CDTs in production lines could improve decision making. CDTs’ real-time data analysis allows operators to see the production line’s performance in real time. CDTs provide a real-time view of the production line and its performance, enabling operators to make informed decisions. CDTs allow operators to simulate production line scenarios before making decisions. This capability improves operational efficiency, risk mitigation, and decision making [11,21].

CDTs use sensor data and other data to predict machine or component failures. Operators can reduce unplanned downtime and optimize production line performance by using predictive maintenance. CDTs analyze critical factors, like machine utilization, production rates, and energy consumption. CDTs can optimize production lines using these insights. This includes optimizing machine usage, production rates, and energy savings [40]. Optimizations improve productivity, waste, and costs.

There are many advantages for CDTs in production lines. Real-time data analysis, scenario simulation, predictive maintenance, and optimization help operators optimize performance and stay competitive. CDTs change decision making, encouraging a proactive and dynamic approach to production line productivity and efficiency.

CDTs bring a multitude of benefits to manufacturing and production systems. As semantically enhanced versions of digital twins, CDTs support autonomous quality by developing data models that facilitate ontology development, leading to quality improvements. They also aid in process optimization and decision making across the entire building life cycle through knowledge graph modeling and reasoning. CDTs are especially valuable for complex production systems with multiple subsystems and stakeholders from diverse domains or life cycle phases, as they offer a unified framework for orchestrating interactions among subsystems and processes. By storing information about their physical counterparts throughout their life cycle, CDTs enable predictive analytics that can substantially enhance decision making and collaboration and reduce business risks. CDTs also support the shift from time-based to condition-based maintenance, allowing early fault detection and minimizing the waste of working components or time, thus reducing through-life maintenance costs and preventing unexpected breakdowns. Additionally, CDTs facilitate feedback loops for product life cycle management (PLM), enabling information from later life cycle stages to inform and enhance the design and creation phase of other assets. By integrating dynamic knowledge bases with digital twin models, CDTs further enable knowledge-based intelligent services for autonomous manufacturing, optimizing production processes and outcomes [14,27,30].

5.1.2. Knowledge Graphs (KGs)

This study found significant implications for using KGs with the discussed data. A system’s components, processes, interactions, and relationships are stored in KGs. KGs enable intelligent agents to reason and learn from data by structuring data. KGs help data management by integrating modules. KGs organize and manage data through their interconnected nodes and edges, ensuring data fusion across components and modules. KGs organize and represent data for CDTs to create virtual models. KGs can capture entity relationships, improving CDT predictions [8]. KGs improve CDTs’ data analysis, insights, and decision making based on graph knowledge.

KGs in CDTs let manufacturers use data insights. Manufacturers can optimize production by using KGs’ structured information. KGs help manufacturers identify patterns, trends, and optimization opportunities by representing the system comprehensively [8,27]. KGs aid CDT development and implementation. They capture, organize, integrate, improve predictions, and aid decision making. Manufacturers can optimize operations, improve performance, and gain deeper insights into their production processes using KGs.

Furthermore, KGs play a pivotal role in facilitating collaboration and knowledge sharing among different departments within manufacturing organizations. With KGs, cross-functional teams can access a unified and comprehensive view of the production processes, enabling them to make informed decisions collaboratively. This promotes efficient communication and enhances the overall workflow, as teams can leverage the insights and recommendations derived from KG-based analyses to address challenges and seize opportunities in real-time.

In addition to aiding manufacturers in optimizing existing production processes, KGs also support the development of new and innovative manufacturing techniques. As manufacturing technologies continue to evolve, KGs provide a robust foundation for experimentation and simulation. Engineers and researchers can create virtual models of proposed production systems within KGs, allowing them to simulate various scenarios and assess the potential outcomes before implementing changes on the factory floor. This not only reduces the risks associated with process modifications but also accelerates the innovation cycle by enabling rapid prototyping and testing in a virtual environment. In essence, KGs serve as a dynamic playground for refining manufacturing strategies, enabling companies to stay at the forefront of technological advancements while minimizing disruptions.

5.1.3. Cognitive Modular Production (CMP)

This study found that CMP in manufacturing has the potential to adapt for production system optimization and automation. CMP uses CPPSs and CDTs to create an intelligent and automated production environment. CMP breaks down production into modular parts. Optimizing each component improves the performance. CMP improves production efficiency and productivity by optimizing individual modules [28].

CMP also improves production system agility. CMP’s modularity allows it to adapt to changing production needs. Manufacturers can quickly adapt to market shifts and customer needs with this flexibility. CMP relies on CDTs to simulate modular components before deployment. Manufacturers can simulate the module performance, functionality, and behavior using CDTs [41]. Virtual testing identifies issues, optimizes designs, and streamlines module integration into the production system, reducing errors and improving efficiency.

CMP also uses CPPSs to monitor and control physical production in real time. The CPPS integrates sensors, actuators, and control systems. This fusion allows real-time decision making and predictive maintenance to minimize downtime and maximize productivity by monitoring, collecting, and analyzing production system data.

In conclusion, CMP integrates CPPSs and CDTs to transform manufacturing. CMP optimizes modular components, improving performance, flexibility, and efficiency. CDTs enable virtual testing and simulation, reducing errors and easing fusion. The CPPS also allows real-time production monitoring and control. CMP could transform manufacturing and production systems.

In addition to its impacts on production optimization and automation, the concept of Cyber–Physical Production Systems (CPPSs) plays a crucial role in enhancing the safety and security of manufacturing environments. With the integration of sensors, communication networks, and data analytics, the CPPS enables the real-time monitoring of production processes. This monitoring capability extends beyond production performance, encompassing factors such as worker safety, equipment health, and environmental conditions. By collecting and analyzing data from various sources, the CPPS can proactively identify potential hazards, malfunctions, or deviations from safety protocols. This ability to predict and mitigate risks contributes to a safer working environment for personnel and safeguards against the occurrence of accidents that could disrupt production continuity. Furthermore, the secure communication protocols inherent in the CPPS help prevent unauthorized access and cyberattacks, ensuring the integrity of production data and the overall operational infrastructure.

While the adoption of Cyber–Physical Production Systems (CPPSs) brings about substantial benefits, it also presents several challenges that necessitate careful consideration. One of the key challenges involves the integration of legacy systems and technologies with the new CPPS framework. Many manufacturing facilities have established processes and equipment that were not originally designed to be interconnected or digitally monitored. Retrofitting these systems to accommodate CPPS functionalities may require substantial investments in terms of time, resources, and capital. Moreover, the transition to a CPPS demands a skilled workforce proficient in both traditional manufacturing practices and the intricacies of digital technologies. Bridging this skills gap through training and education is essential to fully harness the potential of the CPPS. Additionally, the increased reliance on data-driven decision making and real-time analytics exposes manufacturers to potential cybersecurity vulnerabilities. Safeguarding sensitive production data from cyber threats becomes imperative, necessitating the implementation of robust cybersecurity measures and continuous monitoring. Addressing these challenges will be pivotal in realizing the transformative potential of the CPPS while ensuring its seamless integration into existing manufacturing ecosystems.

5.1.4. Fusion of KGs into CMP

The fusion of KGs into CMP offers many benefits. First, extracting information from standards helps understand the interoperability landscape within Industry 4.0 standardization framework data sources. Effective interoperability strategies are built on this knowledge. Second, the standard ontology (STO) population of the KG organizes and simplifies data. The KG can organize and access information by incorporating Industry 4.0 standards and standardization frameworks. This structured format allows seamless analysis and fusion with other KGs, enabling comprehensive insights and informed decision making [15].

KG fusion also helps resolve semantic interoperability issues. KG instances are linked to identify and resolve inconsistencies. This process creates a unified KG that aids decision making and production optimization. The integrated KG streamlines project management and boosts productivity. KG reasoning also reveals data patterns and relationships. This fusion can infer and predict, revealing valuable insights and actionable intelligence. Organizations can improve performance and efficiency by using the cognitive capabilities of the KG integrated into CMP to understand complex production dynamics and make informed decisions. Finally, KG interlinking in the Linked Open Data Cloud improves interoperability and accessibility. Unique identifiers and standardized vocabularies can connect KGs from different domains and sources to form a knowledge network. Interconnectedness promotes collaboration, knowledge sharing, and information access. The fusion of KGs into CMP allows stakeholders to share knowledge and advance CMP [8].

The fusion of KGs into CMP provides many benefits, including interoperability understanding, structured data representation, seamless integration, intelligent reasoning, and enhanced interconnectivity. Data-driven decision making, optimized production processes, and collaboration and innovation could transform construction project management [21].

The results of a quantitative analysis conducted among key stakeholders in the CMP field, including design managers, design coordinators, BIM managers, BIM coordinators, digitalization specialists, project managers, and construction managers, demonstrate a potential inclination towards adopting the fusion of KGs into CMP. This analysis provides empirical evidence supporting the fusion of KGs into CMP having the ability to facilitate real-time analysis through data-driven models enhanced by cognitive resources. The fusion of KGs into CMP aims to enable effective decision making and enhance comprehension, optimization, and critical thinking. Additionally, it proves to be a valuable monitoring and control mechanism, contributing to overall device optimization. It empowers project management teams to self-organize, respond to unforeseen events, and make informed decisions in the context of complex systems involving physical actors [11,21].

The fusion of KGs into CMP work has the potential to revolutionize the AEC industry by redefining the design, construction, and operation processes within complex project environments. The industry is poised to embrace cognitive technologies, such as DTs, cognitive computing, AI, ML, and cloud-based systems, as part of its inevitable evolution. Figure 3 depicts the model built on a multi-source real-time data flow that monitors the physical twin, applies ontology to the IoT, and enables the standardized representation and semantic interoperability of data. KGs as structured representation of knowledge that captures relationships, entities, and attributes enable the fusion and analysis of diverse data sources, providing a comprehensive and contextual understanding of the incorporation of KGs into CDTs for the CMP process. This fusion of interconnected KGs into CMP provides learning, event identification, and prediction and contextual reasoning skills through a developed ontology and algorithms. The integration of these two areas in the schema was inspired by the survey results, which indicated a strong consensus among experts for the need for such a fusion to advance the field in line with the existing literature.

This research emphasizes that the fusion of KGs into CMP serves as an intelligent system seamlessly connecting engineering operational data, information, and models throughout the entire life cycle of modular production. It leverages self-learning capabilities and predictive analysis to provide real-time results in the appropriate context, empowering production managers to proactively prevent or resolve potential issues. By facilitating knowledge-driven decision making, the fusion of KG into CMP enhances the industry’s ability to optimize performance and achieve efficient outcomes. Throughout monitoring the operations of the production plant (physical twin) that will provide specific behavior and capabilities to CDTs, AI models will provide cognitive capabilities for the following: real-time optimization algorithms regarding an aligned predictive production schedule at the production-plant level; energy-aware machines (self-identification of optimal model of operation); self-configurable production lines and machines; and proactive behavior for risk management (hazard analysis and critical material).

5.2. Practical Implications

The integration of KGs and CPPSs offers significant practical implications for manufacturing and industrial processes. This integration entails adopting a KG-based approach to model CPPSs while incorporating contextual information and graph embeddings to enhance the system’s insights. One notable application of this approach is evident in the realm of self-organized reconfiguration management within CPPSs. This approach effectively identifies and integrates contextual influences from the environment that impact product quality. Moreover, it facilitates quality assessment by revealing intricate dependencies within the architecture of the intelligent digital twin, thereby enabling manufacturers to achieve a more holistic understanding of their production processes and optimize them accordingly [51].

The synergy between KGs and CMP not only leads to enriched insights but also opens avenues for transformative decision making and optimization. The incorporation of KGs into CMP offers a comprehensive framework that holds the potential to bolster Industry 4.0 principles. By leveraging KGs’ capabilities, CMP gains the capacity to enhance decision making processes, streamline optimization efforts, and fully realize the potential of Industry 4.0’s interconnected and data-driven paradigm. The fusion of KGs into CMP equips production managers with the ability to effectively analyze vast datasets, simulate diverse scenarios, predict potential failures, and identify optimization opportunities. As this fusion takes root in industries like AEC, it carries the promise of revolutionizing production systems, elevating overall efficiency, and propelling innovation in modular manufacturing processes. Detailed insights into the operational processes, opportunities, and challenges associated with the fusion of KGs into CMP are presented in

Table 6. Sample of the fusion of KGs into CMP.

	Process	Opportunities	Challenges
Unveiling Industry 4.0	Understanding the principles, concepts, and standardization frameworks of Industry 4.0.	Harnessing the power of advanced technologies, like the IoT, AI, and robotics to revolutionize manufacturing processes and enable smart factories. Creating new business models and opportunities for growth through interconnected systems, data-driven decision making, and enhanced supply chain management.	Overcoming resistance to change and ensuring widespread adoption of Industry 4.0 technologies and practices across different sectors and organizations. Addressing cybersecurity risks and data privacy concerns associated with the integration of advanced technologies and interconnected systems in Industry 4.0 implementations.
Building a Knowledge Graph	Applying the standard ontology (STO) to ensure a consistent and easily understandable format, facilitating analysis and integration with other KGs.	Enabling comprehensive data integration and analysis, leading to enhanced insights, informed decision making, and improved operational efficiency. Facilitating the development of intelligent systems and applications by providing a structured and interconnected knowledge representation, fostering innovation and collaboration.	Acquiring and integrating diverse data sources and ensuring data quality and accuracy for constructing a comprehensive and reliable knowledge graph. Overcoming the complexities of ontology design and semantic modeling to accurately represent and capture the relationships and concepts within the domain of interest.
Seamless Integration	Semantically defining connections between instances in knowledge graphs to resolve interoperability conflicts. Utilizing techniques like mapping and alignment to ensure consistency and compatibility across different KGs, resulting in a unified and interconnected system.	Enabling seamless data exchange and interoperability between different knowledge graphs, fostering a unified and comprehensive view of information for improved decision making and system optimization. Enhancing the scalability and extensibility of knowledge graphs, allowing for the integration of diverse data sources and facilitating the development of advanced applications and services in CMP.	Resolving semantic interoperability conflicts and ensuring consistent data representation across multiple KGs. Managing the scalability and complexity of integrating large-scale KGs while maintaining data integrity and preserving the integrity of relationships between entities.
KG Reasoning	Applying reasoning algorithms to identify patterns and relationships within the KG data. Utilizing this information to make inferences and predictions, uncovering valuable insights and enhancing the capabilities of the KG-based framework.	Unveiling valuable insights and patterns within the knowledge graph data, empowering informed decision making, predictive analysis, and the optimization of cognitive modular production processes. Leveraging reasoning algorithms to extract hidden knowledge and make inferences, enabling proactive problem solving, enhanced system performance, and continuous improvement in the KG-based framework.	Dealing with the computational complexity and scalability of reasoning algorithms when applied to large-scale KGs. Addressing the uncertainty and ambiguity inherent in the data and making accurate and reliable inferences and predictions based on incomplete or noisy information.
KG interlinking	Establishing connections between different KGs using unique identifiers and standard vocabularies. Enabling accessibility and enhancing the interoperability of the KG-based framework by linking the graphs in the Linked Open Data cloud.	Enhanced interoperability and accessibility of the KG-based framework through interconnected KGs, enabling seamless data exchange, collaboration, and knowledge sharing in cognitive modular production. Facilitating a broader audience access to linked KGs, promoting cross-domain insights, and fostering innovation and development in the field of CMP systems.	Ensuring the compatibility and alignment of different KGs with varying data models, formats, and ontologies. Resolving the issue of data heterogeneity and inconsistency when integrating KGs from diverse sources, domains, and languages.

, underlining the transformative potential of this integration.

Furthermore, the integration of KGs and CMP holds broader implications for various sectors beyond manufacturing. The intelligent utilization of KGs enables a shift towards more data-driven, context-aware decision making, which can be applicable in fields such as healthcare, supply chain management, and smart cities. For instance, in healthcare, KG-CMP integration could facilitate personalized treatment plans by considering a patient’s medical history, genetic information, and current health status, leading to more effective and precise medical interventions. In the context of smart cities, KG-CMP integration could aid in optimizing resource allocation, urban planning, and infrastructure maintenance by analyzing data from various sources, such as sensors, social media, and public records. As a result, the symbiotic relationship between KGs and CMP has far-reaching implications that extend beyond manufacturing, reshaping how various industries operate and innovate in an increasingly data-centric world.

5.3. Limitations and Future Study

This study is limited by its relatively small sample size, limited exploration of parameters, and a focus on a specific subset of regions. Due to the research topic focusing on novel technologies that are not widely known or implemented in the AEC industry, respondents were purposefully sampled. While this approach may have reduced the sample size, it likely increased the quality of the data. However, being overly selective may have compromised the congruence between the sample and the entire population, in this case, modular production in the industrialized building industry. Given the novelty and scope of the research topic, it was challenging to identify individuals with expertise across all areas. Alternatively, a study design involving focus groups or workshops with subject matter experts could have been employed to allow for responses and discussions to build upon each other, resulting in more comprehensive insights.

This study highlights the requirement for further research into the development of the methodologies, infrastructure, and modeling to support data-driven analytics for the adaptation of CDTs in CMP. Analytics system requirements and design specifications could be described, the analytical platform for the manufacturing plant for modular production could be developed, a holistic model of uncertainty and causal relations could be created, and an anomaly detection system could be established.

6. Conclusions

The literature, supported by the quantitative analysis conducted with industry professionals, shows that the fusion of KGs into CMP offers immense potential for revolutionizing the construction industry and enabling efficient and intelligent decision-making processes. Through the extraction of information from standards, the population of a KG, seamless integration, reasoning algorithms, and the interlinking of KGs, this fusion enhances cognition, analytics, and optimization in CMP. The process of unveiling Industry 4.0 through this fusion involves extracting valuable information from standards to analyze interoperability, identify key challenges, and gain a comprehensive understanding of the current state of the industry. This provides a potential foundation for the development of the fusion of KGs into CMP. Building a KG using the standard ontology for Industry 4.0 frameworks facilitates the structured representation of entities and their relationships within the data sources. This enables easy accessibility, analysis, and fusion with other knowledge graphs, enhancing the overall understanding of the CMP domain.

The seamless fusion of multiple KGs resolves semantic interoperability conflicts by semantically defining connections between instances. Techniques such as mapping and alignment ensure consistency and compatibility, resulting in a unified knowledge graph that supports decision making and optimizes the production process in CMP. KG reasoning utilizes algorithms to unveil new knowledge, identify patterns, and make inferences and predictions. This process extracts valuable insights from the data, empowering stakeholders with actionable information for enhanced decision making and the optimization of cognitive modular production. KG interlinking connects knowledge graphs using unique identifiers and standard vocabularies, facilitating interoperability and accessibility. This expands the reach of the fusion of KGs into CMP, enabling a wider audience to benefit from its insights and contribute to the advancement of CMP. While the fusion of KGs into CMP presents numerous opportunities, it also faces several challenges. These challenges include data heterogeneity and inconsistency when integrating knowledge graphs from diverse sources, as well as the need to ensure compatibility and alignment across different data models and ontologies. Overcoming these challenges requires robust data management strategies, advanced data fusion techniques, and continuous efforts to maintain data quality and integrity.

In conclusion, the fusion of KGs into CMP represents a transformative approach to the construction industry, enabling intelligent decision making, optimization, and cognitive capabilities. By leveraging the power of KGs, this fusion has the potential to revolutionize how construction projects are designed, executed, and managed. As Industry 4.0 continues to evolve, the fusion of KGs into CMP has the capability to drive innovation, enhance efficiency, and maximize the potential of CMP in the AEC industry.