Next Article in Journal
Kinematic Analysis of Patients with Charcot–Marie–Tooth Disease Using OpenSim
Previous Article in Journal
Design of Experiments-Based Adaptive Scheduling in Kubernetes for Performance and Cost Optimization
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 15
Issue 18
10.3390/app151810102
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

Role of Generative AI in AI-Based Digital Twins in Industry 5.0 and Evolution to Industry 6.0

by
Izabela Rojek
*,
Dariusz Mikołajewski
,
Adrianna Piszcz
,
Olga Małolepsza
and
Mirosław Kozielski
Faculty of Computer Science, Kazimierz Wielki University, Chodkiewicza 30, 85-064 Bydgoszcz, Poland
*
Author to whom correspondence should be addressed.
Appl. Sci. 2025, 15(18), 10102; https://doi.org/10.3390/app151810102
Submission received: 31 August 2025 / Revised: 7 September 2025 / Accepted: 8 September 2025 / Published: 16 September 2025
Browse Figures
Versions Notes

Abstract

Featured Application

The specific application of the work lies in the new use of more effective digital twins based on generative artificial intelligence in industry, including the Industrial Internet of Things, better adapted to the specifics of cooperation with a human operator.

Abstract

Generative artificial intelligence (genAI) plays a crucial role in improving AI-based digital twins (DTs), enabling more dynamic, adaptive, and accurate industrial simulations, essential as Industry 5.0/6.0 paradigms evolve and are implemented. In industry, genAI can simulate complex manufacturing processes or entire production lines, enabling companies to optimize operations, predict maintenance needs, reduce downtime, and develop more scenarios for correct operation (e.g., for faster transitions to new products or new materials) and address potential failures. GenAI also helps DTs continuously learn and evolve by generating new data and scenarios based on historical and current inputs. This capability ensures that DTs remain current and reflective of the real systems they represent, for both operational and training purposes (e.g., training operators for situations that rarely occur on a real production line).Furthermore, it facilitates the creation of synthetic data, which is important for training AI models when real-world data is scarce or expensive. This accelerates the development and improvement of DTs and increases the predictive accuracy, personalization, and operational efficiency of AI-based digital twins, making them more reliable and versatile tools in medicine and industry. However, in addition to strengths, it is also worth considering threats to prepare for risk mitigation. This article helps capture and maintain a balance between opportunities and threats in this area.

1. Introduction

Generative AI refers to a class of AI models designed to create new content, such as text, images, audio, and even entire virtual environments, by learning patterns from existing data. Unlike traditional AI, which focuses on recognizing patterns or making decisions based on input, generative AI is capable of producing original and creative outputs. These models, such as Generative Pre-trained Transformer (GPT used to generate text) or GAN (Generative Adversarial Networks for creating images and videos), learn from large datasets to understand the structure and relationships in the data. Once trained, they can generate new, unique content that mimics the style or structure of the data they were trained on. Generative AI has a wide range of applications, from art and design to science and industry. It can be used to create realistic images, simulate complex environments, and even compose music and write stories. In industries like healthcare, generative AI can synthesize data to train models where real-world data is sparse or sensitive. It also plays a key role in improving AI-based digital twins by generating synthetic scenarios for predictive modeling. Despite its potential, generative AI raises ethical concerns, such as the risk of creating deepfakes or producing biased content if trained on biased data. As a result, there is an ongoing debate about how to responsibly develop and apply generative AI technologies [1].
The origins of generative AI can be traced back to early AI research, which focused on creating systems that could mimic human creativity. The groundwork was laid in the 1950s and 1960s with the development of basic machine learning algorithms, such as neural networks, which attempted to simulate brain-like processes. However, these early models lacked the computing power and data needed to perform complex tasks. In the 1980s, AI researchers began experimenting with generative models, such as hidden Markov models (HMMs), for tasks such as speech and handwriting recognition, an early step toward content generation. The breakthrough came in the mid-2000s with the development of deep learning, which allowed neural networks to scale and process massive amounts of data. The introduction of variational autoencoders (VAEs) and generative adversarial networks (GANs) in 2014 made a significant leap forward in generative AI. In particular, GANs revolutionized the field by pitting two networks (a generator and a discriminator) against each other to create highly realistic synthetic images. At the same time, advances in natural language processing (NLP), particularly with the Transformer architecture introduced in 2017, led to the development of models like GPT that could generate coherent text at an unprecedented level [1]. These innovations set the stage for the modern era of generative AI, where models are trained on massive datasets and used in applications ranging from creative content generation to scientific simulations. As computing power and access to large datasets increased, generative AI began to flourish, evolving from experimental tools to mainstream technologies. Today, generative AI continues to evolve, driven by innovations in model architecture and training techniques (Figure 1).
Generative AI offers several advantages that make it a transformative technology in various fields. One of the key benefits is the ability to create high-quality, lifelike content such as images, text, and audio with minimal human intervention. This can save time and resources in industries such as entertainment, marketing, and design. Additionally, it can generate synthetic data, helping train AI models in fields such as healthcare and autonomous vehicles, where obtaining real-world data may be limited or expensive. Generative AI also enhances creativity by providing new design possibilities, assisting artists, and aiding research and development through simulations and innovation. It can also personalize experiences in applications such as chat bots, virtual assistants, and recommendation systems, improving user engagement. On the other hand, generative AI has raised concerns about the potential misuse of its capabilities, such as creating deepfakes or fake news that can spread misinformation. There are also ethical issues related to bias in AI-generated content, as models trained on biased data can reproduce or reinforce societal prejudices. Another challenge is the high computational and energy costs associated with training large generative models, which raises sustainability concerns. Generative AI models are prone to generating inaccurate or nonsensical results in some contexts, which can be problematic for high-stakes applications. There are privacy risks, especially when models generate content based on sensitive or personal data, underscoring the need for strong data governance and regulation [1].
GenAI plays crucial role in improving AI-based digital twins (DTs), enabling more dynamic, adaptive, and accurate simulations in industry. A DT is a dynamic, virtual replica of a physical asset, process, or system that is continuously updated with real-time data and AI-powered analysis. It acts as a bidirectional bridge between the physical and digital worlds, enabling monitoring, simulation, forecasting, and optimization. In Industry 5.0, digital twins are enhanced with genAI, allowing them not only to reflect reality but also to create new designs and adaptive strategies in collaboration with humans. As Industry 6.0 evolves, DTs become self-organizing, autonomous, and globally connected ecosystems, capable of learning and evolving with minimal human intervention (Figure 2).
In industry, generative AI can simulate complex manufacturing processes or entire production lines, enabling companies to optimize operations, predict maintenance needs, and reduce downtime [2,3,4]. By analyzing data in real time, generative AI allows DTs to identify patterns, helping industries adapt to changing conditions and proactively resolve issues before they become more severe. Generative AI also helps digital twins continuously learn and evolve, generating new data and scenarios based on historical and current inputs. This capability ensures that DTs remain relevant and reflective of the real-world systems or people they represent [5,6]. It also facilitates the creation of synthetic data, which is important for training AI models when real-world data is scarce or expensive to obtain. This accelerates the development and improvement of DTs, increases the predictive accuracy, personalization, and operational efficiency of AI-based digital twins, making them more reliable and versatile tools in medicine and industry. For example, in healthcare, DTs powered by generative AI can model a patient’s biological systems, offering highly personalized treatment plans. In engineering, these enhanced DTs can simulate the life cycle of infrastructure projects, enabling predictive maintenance and cost-effective asset management. Furthermore, the continuous evolution of digital twins, powered by generative AI, enhances decision-making capabilities across sectors, providing simulations that adapt to changing circumstances. This adaptability not only increases operational efficiency but also helps manage risks by helping to predict potential failures or inefficiencies. However, as generative AI and DT become more integrated, so do the associated risks, including data security, privacy concerns, and potential AI-generated errors. In addition to the strengths, it is also worth considering the threats to prepare to mitigate the risks associated with them. These risks underscore the importance of establishing a solid regulatory framework and ethical guidelines to ensure the responsible use of AI in digital twin technology. In recent years, technological advances and digitalization have revolutionized also healthcare, transforming both care delivery and patient management. The concept of DTs has the potential to significantly impact various clinical environments. DTs are virtual models of physical entities such as patients or organs that are continuously updated with real-time data to reflect their real-world counterparts [7,8].
Industry 5.0 is defined as the next stage of industrial evolution, emphasizing human–machine collaboration, where advanced AI, robotics, and DTs work together to enable personalization, resilience, and sustainability. Unlike Industry 4.0, which focused primarily on automation and cyber-physical systems, Industry 5.0 prioritizes human-centered design, circular economy principles, and energy-efficient production. Its core principles encompass sustainability, resilience, and personalization, ensuring that technology serves societal and environmental needs, not just economic growth. A key feature is the integration of DTs enhanced with genAI, enabling real-time simulation, adaptive decision-making, and collaborative problem-solving between humans and intelligent machines. Industry 6.0, while still in development, represents an evolution toward a fully intelligent and connected industrial ecosystem, where AI, quantum computers, and advanced generative systems drive autonomous value creation. Its definition focuses on self-adaptive, hyperconnected, and regenerative systems that extend beyond factories to encompass entire socio-technical ecosystems. Its core principles include autonomy, self-organization, global sustainability, and ethical AI governance. Industry 6.0 features include self-developing DTs, distributed intelligence, and global optimization across industries, societies, and energy systems. GenAI plays a key role in generating innovative designs, optimizing resource flows, and enabling predictive analytics in evolving DT ecosystems. Industries 5.0 and 6.0 emphasize the need to integrate genAI with AI-based DTs, as these tools combine human creativity, industrial productivity, and sustainable transformation.
The aim of this article is to investigate the current role of generative AI in AI-based digital twins in industry (in the light of the Industry 4.0 and Industry 5.0 paradigms) and to what extent the current capabilities in this area are being used. We note that AI/ML-based data analytics is already being used in the Industry 4.0/5.0 paradigm, but it has not yet reached all business areas and not the full range of AI methods has been fully exploited. This article is an attempt to approach this topic comprehensively, reviewing and systematizing existing research, its limitations, promising future research directions, and related risks.

2. Materials and Methods

2.1. Data Set

This bibliometric analysis aims to explore the research landscape around the current role of generative AI in AI-based digital twins in industry (in the light of the Industry 4.0 and Industry 5.0 paradigms).Bibliometric methods are used to analyze scientific publications. To guide our study, we developed several research questions targeting key areas, including the evolution of research issues over time, geographical distribution of research and publications, most influential authors and articles, collaboration networks between researchers and institutions, and emerging topics that may impact future research. Given the novelty of the above topics, it seems essential to develop a comprehensive picture and a thorough understanding of current research trends, industry practices, and future directions of proposed solutions. The bibliometric analysis procedure formulated in this way (Figure 3) allows for a more accurate analysis and interpretation of bibliometric data and will provide a solid basis for future research.

2.2. Methods

Our bibliometric analysis included scientific articles from two major databases, Web of Science (WoS) and Scopus, known for their broad range of research and rich citation data, which is crucial for reliable bibliometric analysis. Our analysis utilized tools built into both databases, which enable keyword categorization and insight into authors, their affiliations, countries, and funding sources, presenting the results in graphs and detailed tables. To refine the search and align it with the review objectives, we applied filters. This allowed us to focus on relevant literature, limiting the results to original and review articles in English. Each article was then manually reviewed by three senior experts to ensure it met our criteria, resulting in the creation of a final sample. We then proceeded to analyze descriptive statistics to examine key features of the dataset, such as lead authors, research networks, thematic clusters, and emerging trends. This allowed us to better track terminology evolution and important research milestones, and to examine key trends and themes or research areas that received particular attention in previous publications. By analyzing temporal trends, we observed changes in research interests over time and classified publications into thematic clusters, examining their interconnectedness. This process also highlighted key topics and subfields. Indicators such as authorship trends (including author teams), affiliations, and countries of origin enabled a quantitative summary of the selected literature and helped assess the impact and influence of publications, shedding light on their scientific significance. PRISMA 2020 (partial only: 10 items) was used during this research (Figure 4 and PRISMA 2020 Checklist (partial only) in Supplementary Materials).
When reviewing publications on genAI for AI-based DTs in Industry 5.0 and the evolution toward Industry 6.0, databases such as Web of Science, and Scopus provide carefully selected, high-quality, peer-reviewed sources, guaranteeing scientific reliability and credibility. Their built-in search and visualization tools enable structured mapping of trends, co-authorship, and citation networks without relying on external, potentially inconsistent platforms. Unlike Google Scholar, which has limited indexing transparency and includes non-peer-reviewed materials, these databases apply strict inclusion criteria consistent with the level of expertise required for Industry 5.0/6.0 research. Furthermore, domain-specific indexing (e.g., engineering and computer science in Scopus) enables precise tracking of interdisciplinary advances crucial to DT research. Using only these internal tools ensures methodological consistency, reproducibility, and scientific credibility in the review process.

3. Results

In the WoS database, we searched by entering the following keywords in the “All fields” field: “generative AI”“digital twin” industry and related and their combinations. When searching in the Scopus database, we entered the keywords “generative AI”“digital twin” industry and related and their combinations in the field “Article title, abstract, keywords”. The richest result showed 45 publications published in 2023–2025 and we included them for further discussion (Figure 5).
The above publications covered the following areas:
  • Computer science: 27.4%;
  • Engineering: 21.7%;
  • Mathematics: 11.3% (Figure 6).
In terms of document type, the following dominated:
  • Conference paper: 45.7%;
  • Article: 23.9%;
  • Conference review: 15.2% (Figure 7).
A lack of leading affiliation and names of authors or sources of research funding was observed. However, the region of origin of the publications is noteworthy:
The three most common sustainable development goals (SDGs) were:
  • “Innovation and industry infrastructure”;
  • “Responsible production and consumption”;
  • “Good health and well-being”.
  • The above results indicate the selection of specialization and location of centers, but no dominant method of research financing was observed.

3.1. Quantitative Analysis

Gkontzis et al. [9] showed that smart cities use advanced data analytics, predictive models, and digital twin techniques to drive sustainable urban development. Predictive analytics helps cities proactively plan and adapt to future challenges. Digital twin technology creates a virtual model of the city, enabling real-time monitoring and analysis of city systems. This study highlights how real-time simulations can identify problems and improve the performance of smart cities. It examines how citizen report analysis, prediction, and digital twin technologies can improve neighborhoods in particular. The research combines ETL processes, AI techniques, and digital twin methods to analyze city data from citizen reports. The interactive GeoDataFrame in the digital twin approach enables simulations of different scenarios, helping users predict how neighborhoods will respond. These simulations uncover patterns and trends, improving the resilience and functionality of cities. This approach ultimately leads to better decision-making in city planning and improving the quality of life for residents. By leveraging these technologies, cities can create more efficient, adaptive and sustainable environments [9]. Avanzato et al. [10] demonstrated that modern healthcare facilities require new digital tools to better meet the growing demand for technologies that support healthcare professionals.With rapid technological advancements, the use of IoT devices, which enable the collection of biomedical and biometric data, has become widespread.This opens up new opportunities for the healthcare sector in analyzing data from these tools [10].
Software solutions for AI-based DTs in Industry 5.0 and the evolution to Industry 6.0 include both commercial and open-source platforms, increasingly integrating genAI into simulation, optimization, and predictive analytics. Leading enterprise solutions include Siemens NX with Teamcenter, Dassault Systèmes’ 3DEXPERIENCE, and PTC ThingWorx, which provide advanced capabilities for AI-based industrial modeling, lifecycle management, and predictive maintenance. Ansys Twin Builder is widely used for physics-based modeling and AI integration, enabling high-fidelity simulations that can be enriched with generative AI for design innovation. Microsoft Azure Digital Twins and IBM Digital Twin Exchange provide cloud environments that integrate machine learning, IoT, and edge AI for large-scale industrial ecosystems. Among open source solutions and free platforms, OpenModelica and FIWARE stand out as platforms supporting DT modeling with AI extensions, suitable for academic and small-scale industrial applications. Eclipse Ditto is another open source framework designed to manage DTs in IoT ecosystems, offering the flexibility to integrate AI algorithms, including genAI, for data-driven adaptation. New, free tools such as Digital Twin Web (DTW) and Python-based simulation libraries allow researchers to experiment with AI-based twin prototypes, making them valuable for human-centric experiments in Industry 5.0. These software solutions (ranging from enterprise-grade solutions to free, open-source tools) provide the foundation for integrating generative AI with DTs, enabling enterprises to transition to human–machine collaboration in Industry 5.0 and self-organizing, intelligent ecosystems in Industry 6.0.
Industry 6.0 has begun to develop gradually, in part through experimental applications of genAI, quantum computing, and hyperconnected DT ecosystems in select advanced industries. Unlike the rapid shift from Industry 3.0 to Industry 4.0, the shift toward Industry 6.0 is evolutionary, not revolutionary, and is developing alongside the consolidation of the human-centric and sustainable framework of Industry 5.0. Early signs of Industry 6.0 development include autonomous factories, self-adaptive supply chains, and AI-powered global optimization systems that extend beyond individual companies. GenAI plays a key role in enabling DTs that can self-evolve, co-create designs, and optimize energy and resource consumption on a global scale. However, Industry 6.0 currently coexists with Industry 5.0, as industries continue to prioritize human–machine collaboration, resilience, and sustainability before fully transitioning to self-organizing systems. Over time, Industry 6.0 will gradually replace Industry 5.0 as autonomous, intelligent, and ethically managed ecosystems become mature enough to redefine global industrial operations.

3.2. Specific/Case Analysis

Digital twins (DTs) are becoming increasingly important in healthcare as they offer innovative solutions. Many research studies are currently combining artificial intelligence (AI) with DT. This paper presents a case study and proof of concept using microservices, focusing on cardiac DT for the analysis of electrocardiogram (ECG) signals with AI support. The system aims to improve the relationship between patients and doctors, making treatment more efficient. It also aims to shorten the treatment time and reduce healthcare costs by streamlining processes [10]. Research by Orlova et al. addresses the challenge of digital twin engineering in organizational and technical systems. It builds on foundational scientific work in the field of digital twin engineering and applied modeling. The study employs a system approach, statistical analysis, operational research, and artificial intelligence methods. A comprehensive methodology for designing digital twins is proposed to accelerate the engineering process. This approach includes design steps, models, and methods that allow for the synthesis of digital twins for complex systems operating under uncertainty. These systems can reconfigure in response to internal faults or environmental changes and perform preventive maintenance. The technology includes a simulation model based on “what-if” situational analysis and fuzzy logic methods. It is applied to develop a digital twin prototype for a device at the early stages of its life cycle, aiming to reduce the impact of unexpected faults. The study examines potential problems and failures that may arise during the device’s future operation. The model identifies failure scenarios based on internal and external factors, recommending appropriate actions for the device. The research’s practical contribution is the decision support model, which helps monitor technical devices and provides solutions to eliminate dysfunctions [11]. Generative AI is transforming the field of imaging, significantly improving detection and diagnosis, including cancer.AI is impacting the field in particular in terms of expanding data sets, improving image quality, and developing predictive medicine. A key focus is the ethical implications of integrating AI into practice, as well as the new perspective of using AI-generated digital twins for personalized screening. There is no doubt that AI-powered digital twins have the potential to improve screening protocols, facilitate earlier detection, and tailor strategies to personalized DT profile [12]. While efforts have focused on optimizing recruitment, less progress has been made in developing advanced monitoring systems to identify workers at risk of dropout and intervene in a timely manner. Previous research on retention has largely relied on deterministic models that have difficulty capturing the unpredictable nature of the patient’s journey. Existing generative models such as TimeGAN and CRBM do not address key needs such as personalized generation and multivariate time-series modeling required for digital twins of patients. To address these gaps, the current study introduces ClinicalGAN, a model that enables patient-level generation through the creation of digital twins. ClinicalGAN improves trial monitoring by offering personalized generation based on metadata, predicting dropout risk in real time, and using multivariate time-series data to capture complex relationships. ClinicalGAN, validated on Alzheimer’s disease clinical trial data, outperforms state-of-the-art methods by 3x to 4x in generation quality metrics and demonstrates significant improvements of 5% to 10% in predicting patient abandonment [13]. DTs can simulate a worker’s health by incorporating data from wearables, medical devices, diagnostic tests, and electronic health records, enabling personalized healthcare monitoring and treatment planning. Applications of DT in laboratory medicine, especially when combined with AI and AI-generated synthetic data, can personalize applications of biological variability (BV) by taking into account circadian rhythms and population-based BV data. DTs can also improve the interpretation of tumor markers in advanced cancer therapies and help refine reference intervals based on individual variability. Although widespread adoption of DT in healthcare is not expected in the near future, the technology is well on its way to offering innovative solutions for more accurate diagnostics and dynamic treatment assessments in personalized medicine [14]. Digital Twins (DTs) in drug development and clinical trials refer to digital replicas of systems ranging from cells to humans, allowing for simulations and experiments in virtual environments. These DTs enhance the drug discovery process by digitizing activities that are often economically, ethically, or socially challenging. Their impact is broad, improving disease understanding, aiding biomarker discovery, and speeding up drug development, all of which advance precision medicine. GenAI plays a key role in developing DTs by generating realistic and complex data. The authors provide an overview of generative AI and its application in DTs, comparing current uses in drug discovery and clinical trials. They also highlight the technical and regulatory hurdles that need to be addressed before DTs can fully revolutionize these fields. While current DT applications are limited to simulating a few characteristics, generative AI holds significant potential to reshape drug development by utilizing advancements in deep learning [15]. Generative AI models such as OpenAI’s Large Language Models (LLMs) have the potential to revolutionize personalized medicine, especially in fields such as nephrology [16,17].
Digital twins are essential for connecting the physical and virtual worlds, requiring large amounts of data to keep the virtual counterpart up to date. A semantic communication framework based on the YOLO model was used to create a virtual apple orchard, minimizing data transmission costs. The YOLOv7-X object detector extracted key semantic data (from images captured by edge devices), thereby reducing the amount of transmitted data. Importantly, the importance of semantic information is quantified each time using the detector’s confidence metrics. This leads to the selection of one of two main resource allocation strategies: a trust-based scheme and an AI-generated scheme. The diffusion model is then used to generate an optimal allocation strategy that outperforms both traditional and trust-based methods. Furthermore, the performance of the YOLOv7-X detector is improved by incorporating the ELAN-H network and SimAM attention modules, which increases detection while reducing computational requirements. The results show that the proposed framework and resource allocation schemes effectively reduce the transmission cost and improve the transmission quality of important information [18,19]. Industrial Cyber-Physical Systems (ICPS) are essential for modern manufacturing, enabling smarter processes by digitizing data across the product lifecycle. Digital twins (DTs) in ICPS help transform traditional infrastructures into intelligent, adaptive ones. Generative Artificial Intelligence (GAI) enhances DT by improving predictive accuracy and supporting diverse needs of smart manufacturing. However, using Industrial Internet of Things (IIoT) devices to share data in DT construction is challenging due to adverse screening issues. The GAI-based DT architecture for ICPS solved the negative outcome selection problem using a contract theory model. Optimization required the development of a balanced, diffusion-based, soft-actor-critic algorithm. This algorithm utilizes dynamic structural pruning to improve performance. Numerical results confirm the effectiveness of the proposed scheme [18]. To solve the problem of high data transmission cost, this paper presents a semantic communication framework using the YOLO model to build a virtual apple orchard. The YOLOv7-X object detector is used to extract key semantic information from images captured by edge devices, which reduces the transmission demand and cost. Then, the diffusion model optimizes the allocation process, outperforming traditional approaches. Moreover, the YOLOv7-X detector is enhanced with ELAN-H and SimAM attention modules to enhance performance and reduce computational complexity, making it suitable for low-power edge devices. The results show that the framework effectively reduces the transmission cost while improving the data quality of communication services [20].
Several successful DTs case studies are already demonstrating the transformative role of AI, including genAI, in industrial innovation. In the manufacturing sector, Siemens is using AI-enhanced DTs to optimize production lines, reducing energy consumption and increasing efficiency, reflecting the goals of Industry 5.0, which are focused on human-centricity and sustainability. In the energy sector, General Electric (GE) is using DTs to monitor gas turbines and wind farms, where AI-based predictions enable proactive maintenance and improved energy efficiency. In the healthcare sector, Philips has implemented patient-specific DTs, which, when combined with genAI, enable personalized treatment simulations and optimize the performance of medical devices. Smart cities are also leveraging DTs, for example, in Singapore’s Virtual Singapore project, where real-time data and AI enable urban planning, traffic optimization, and improved sustainability. In the automotive sector, Tesla and BMW are using DTs to simulate vehicle performance and accelerate design innovation, while generative AI is creating new prototypes and safety solutions. These cases illustrate how digital twins, already active in Industry 5.0, are gradually evolving towards Industry 6.0, moving from predictive models to self-adaptive, generative ecosystems.

4. Discussion

GenAI is playing a transformative role in Industry 4.0, 5.0 and 6.0 by increasing the ability to simulate complex systems and predict future scenarios with greater accuracy. It can create detailed models and generate realistic data for DT training, leading to better decision-making and optimization of industrial processes. In Industry 4.0, generative AI helps automate processes and adapt to changes in real time, making production systems more flexible and resilient. As we move towards Industry 5.0 and beyond, GenAI integration supports human-centric technologies, enabling more personalized and collaborative interactions between humans and machine [21,22]. This not only increases productivity but also innovation, enabling more dynamic and intelligent DT that can evolve based on human input. Overall, generative AI accelerates the development of intelligent, adaptive DT, which is key to the evolving industrial landscapes of both Industry 5.0 and novel Industry 6.0 [23,24,25].

4.1. Limitations of Current Studies

The role of genAI in AI-based digital transformation processes in Industry 4.0, 5.0, and 6.0 has so far faced a number of limitations that researchers and industry practitioners must address. A fundamental challenge is the high computational cost associated with data processing and real-time simulation. This stems from the fact that genAI requires significant resources to create and maintain accurate DTs. Furthermore, the increasing complexity of industrial technologies and environments makes it difficult for genAI to accurately capture all variables, also due to the limited use of even the most complex models. This inevitably leads to potential inaccuracies in simulations and forecasts. Data quality and availability also limit the effectiveness of genAI in DTs, as these models rely heavily on vast amounts of high-quality, real-time data. When this data is scattered, noisy, or incomplete, the DT may not accurately reflect the actual system. Another limitation is integrating genAI with legacy systems, where older machines and software may not be compatible with the advanced AI algorithms required by DTs, complicating implementation. The black-box nature of many generative AI models creates transparency and explainability issues, making it difficult for users to trust the decisions and predictions generated by DTs. This lack of explainability can slow adoption, especially in safety-critical industries such as aviation and healthcare [26,27,28]. Security concerns also arise, as digital twins connected to industrial networks are vulnerable to cyberattacks, and the use of generative AI can introduce new attack vectors. Ethical concerns, such as bias in AI models, further complicate the implementation of generative AI in digital twins. If the training data is biased, the digital twin can produce skewed results, leading to unfair or unsafe decisions. Another limitation is the difficulty of scaling generative AI-powered digital twins across large, interconnected systems. As these systems grow in complexity, managing and coordinating multiple digital twins becomes increasingly difficult. Furthermore, maintaining real-time synchronization between the physical and digital worlds is difficult, especially when industrial processes change rapidly. Generative AI can struggle to keep up with these changes, leading to outdated or inaccurate digital twins. The costs of developing and deploying generative AI models can also be prohibitive for small and medium-sized businesses, limiting their availability. Additionally, regulatory and compliance issues can limit the use of generative AI in heavily regulated industries where strict rules govern data use and decision-making processes. Finally, the lack of a standardized framework for developing and evaluating generative AI in digital twins can lead to fragmented solutions, hindering interoperability and collaboration across industries. Addressing these constraints will be critical to unlocking the full potential of generative AI in digital twins within Industry 4.0 and 5.0 [29,30].

4.2. Directions of Further Research

Further research on the role of genAI in AI-based DTs within Industry 5.0 and 6.0 addresses the problem: how to improve production processes through greater than before capabilities of simulation, prediction and optimization [31,32].One key area of research is how genAI can automate the creation and continuous updating of DTs, which digitally represent real-world systems to reflect the current state of machines and processes [33]. This research also explores the integration of generative AI for real-time anomaly detection and predictive maintenance, increasing efficiency and reducing downtime. Researchers are investigating how generative AI can optimize design processes by simulating multiple scenarios and generating optimal solutions for complex industrial tasks such as supply chain management or smart manufacturing [34,35]. Another key area is improving decision-making in Industry 4.0 environments, where AI-based digital twins can predict system failures and suggest proactive adjustments, increasing the resilience of industries. As Industry 5.0/6.0 emphasizes a human-centered and sustainable approach, future research will focus on how genAI in DTs can facilitate the personalization and adaptation of production systems, enabling large-scale customization while improving energy efficiency (Figure 9) [36].
Additionally, research is underway to use genAI to improve human–machine collaboration in smart factories by developing dynamic transformation technology. This technology dynamically adapts to human input and preferences (Figure 10) [37].
This could enable a shift toward more intuitive and user-friendly interfaces, bridging the gap between human operators and complex AI systems. Another area of interest is the role of generative AI in improving the cybersecurity of DTs. As DTs become more common, they are becoming attractive targets for cyberattacks, prompting research into how generative AI can improve anomaly detection and threat mitigation [38]. In addition, there is interest in exploring how generative AI can support DT ecosystems in modeling environmental impacts and sustainability goals, key issues in Industry 5.0 (Figure 11).
Researchers are exploring the potential of generative AI to model energy consumption, waste reduction, and circular economy principles using DTs [39]. There is also significant emerging research on the use of generative AI for supply chain optimization within DTs, simulating various disruptions (e.g., pandemics or natural disasters) to make industrial systems more adaptive and resilient [40]. Generative AI can also play a key role in collaborative DTs across industries, enabling collaborative learning and innovation through connected virtual environments (e.g., based on edge-cloud architecture (Figure 12)) [41].
Research into the ethical implications of using generative AI in digital twins is another important area, particularly with regard to data privacy, accountability, and bias in decision-making processes [29,30,31]. There is research focusing on the scalability of digital twins powered by generative AI, looking for ways to effectively deploy them in various sectors, from manufacturing to healthcare, while maintaining high accuracy and low computational cost [42,43,44,45,46,47].
Among the three paradigms (experience-based methods, operations research, and simulation-based engineering), facility layout design (FLD) has evolved to utilize genAI. It integrates genAI, semantic models, and data-driven optimization. It utilizes asset administration shell (AAS) metamodels that digitally represent physical assets and their relationships. Large language models (LLMs) enrich domain-specific knowledge graphs by analyzing AAS metadata in various relevant contexts (spatial proximity, process dependencies, security, data uniqueness) and guide the generation of an asset layout based on them [48,49]. GenAI enables improved facility layout with 3D visualization, AI-based optimization, and DTs within a common architecture [50,51]. This reduces the complexity of modeling, integration (no island structures), and optimization, improving human-system interaction even for complex engineering objects [52,53]. When using genAI, there is a need to separate knowledge (about objects, spatial relationships, constraints, etc.) from the generative process itself (the possible practical applications of that knowledge) [54]. This creates a knowledge base about various aspects and processes, codified in formal ontologies [55]. This is especially useful for complex layout planning problems with a large number of decision-makers, competing goals, and criteria [56,57]. The visualization utilizes multidimensional visual interactive tools for planner/designer collaboration—the knowledge base [58]. The development of increasingly intelligent and (semi-)autonomous human-cyber-physical systems (HCPS) is driven by the exponential increase in complexity of new generations of IIoT systems. GenAI is already required to efficiently and consistently operate machines, and these requirements will continue to grow in the future [59,60,61]. Besides assisting in human decision-making (including through process visualization), DT can actively train, and human interaction serves as oversight, with high-level decisions made via the human–machine interface and, if necessary, even compromising autonomy when necessary [62,63,64,65].

4.3. Economic Implications

GenAI-based digital twins in Industry 5.0 have significant economic implications, enabling mass customization while reducing prototyping and production costs. By simulating complex scenarios before physical deployment, they reduce the risk of failure and minimize resource waste, contributing to streamlined processes [66]. These DTs improve predictive maintenance, extending equipment lifecycles and reducing downtime, translating into direct cost savings for the industry [67].In supply chains, GenAI-based digital twins better optimize logistics, inventory, and energy consumption, improving network efficiency and profitability. The collaborative, human-centric approach of Industry 5.0 allows employees to interact more effectively with these digital twins, fostering innovation and increasing productivity [68]. However, their implementation requires significant upfront investments in infrastructure, AI training, and cybersecurity, which can pose economic challenges for smaller companies [69,70]. The sustainability aspect of Industry 5.0 creates long-term economic value by reducing environmental impact and aligning with environmental policies and market incentives. As Industry 6.0 develops, genAI-based DTs are expected to power autonomous ecosystems in which machines, humans, and AI co-create economic value in real time (Figure 13) [71,72].
This evolution marks a shift from an economy based on efficiency to models based on resilience and innovation, emphasizing adaptability and decentralized value creation (Figure 14) [73,74].
The impact of these technologies on the global economy will transform competitiveness, favoring companies that can rapidly integrate genAI-based DTs while maintaining a balance between sustainability, resilience, and inclusiveness [5,75].

4.4. Societal Implications

GenAI-based digital twins in Industry 5.0 promote human-centric innovation, enabling greater personalization of products and services that better meet societal needs. They improve healthcare through preventative care models and personalized treatment simulations, improving quality of life and extending it. In education, digital twins can create adaptive learning environments that tailor content to individual employees, better preparing them for specialized tasks and reducing inequalities in access to knowledge [76,77].Integrating these systems with smart cities supports sustainable urban planning, reducing congestion, pollution, and resource inefficiency (e.g., through supply chain monitoring) [78,79]. However, societal risks stem from concerns about data privacy, as DTs rely on extensive personal and behavioral data to operate effectively. Implementing genAI-based DTs could deepen the digital divide, creating inequalities between regions and populations with varying access to such advanced and expensive technologies (Figure 15) [80].
In Industry 5.0, collaboration between humans and AI promotes more fulfilling work, but it can also displace certain categories of workers, requiring retraining initiatives [81]. As Industry 6.0 develops, digital twins could lead to the emergence of hyperconnected societies, where decisions are co-created by humans, AI, and autonomous systems [82]. This raises ethical and governance challenges regarding accountability, transparency, and fairness in AI-based social systems [83]. GenAI-based digital transformation technologies have the potential to build more inclusive, resilient, and sustainable societies, but only if governance and equity issues are integrated into technological advancements [84].
Society 5.0 is a human-centered vision of the future, first proposed in Japan, where advanced technologies such as AI, IoT, robotics, and DTs are deeply integrated into everyday life to solve social problems and increase prosperity. It differs from previous industrial societies by focusing on the balance between technological progress and social value, inclusiveness, and sustainability. The development of Industry 5.0 supports Society 5.0 by promoting human–machine collaboration, personalized solutions, and energy-efficient production, aligning industrial innovation with societal needs. As Industry 6.0 develops, Society 5.0 will continue to be shaped by self-organizing intelligent systems, generative AI-based prediction, and global optimization, extending beyond factories to encompass healthcare, cities, and environmental management. These future industrial revolutions will help Society 5.0 transition toward a more adaptive, resilient, and ethically managed socio-technical ecosystem. Industry 5.0 and 6.0 act as the technological engines that enable Society 5.0 to realize its vision of a sustainable, inclusive, and digitally enriched civilization.

4.5. Ethical, Legal and Environmental Implications

From an ethical perspective, genAI-based DTs in Industry 5.0 raise concerns about autonomy, as decisions can be delegated to AI systems without sufficient oversight from a human operator, technologist, manager, or even programmer. Issues of fairness and bias arise because DTs rely on large datasets, which can incorporate social inequalities into their simulations and predictions [85]. From a legal perspective, liability issues arise when genAI-based DTs make recommendations or autonomous decisions that cause harm or economic loss [86].Intellectual property rights become complex, as genAI can create designs, processes, or products that blur the line between human and machine ownership. Regulatory frameworks continue to evolve, creating legal uncertainty for companies implementing DTs in various jurisdictions around the world [87]. From an environmental perspective, genAI-based DTs can promote sustainability by optimizing resource use, reducing waste, and supporting circular economy models [88]. However, the high computing power required by large-scale generative AI models contributes to energy consumption and carbon emissions, presenting a paradox for green innovation [88]. In Industry 6.0, where hyperconnected, autonomous ecosystems dominate, ethical dilemmas regarding governance, privacy, and human dignity are intensifying [89]. Legal systems will need to evolve toward adaptive, technology-aware governance models that account for cross-border data flows and AI accountability [30]. From an environmental perspective, Industry 6.0 could enable the creation of net-zero emissions industrial ecosystems, but only if DTs are developed using energy-efficient AI (so-called Green AI), renewable energy sources, and sustainable lifecycle management [90], including during workshops [91,92].

5. Conclusions

GenAI plays crucial role in advancing AI-based DTs in Industry 4.0/5.0/6.0.It has made it possible to create incredibly accurate, real-time simulations of physical systems by generating massive datasets and models that mimic these real-world environments. This helps with predictive maintenance, optimizing operations, and reducing downtime in industrial processes. Generative AI can autonomously design solutions, suggest improvements, and adapt to new scenarios, making digital twins more flexible and dynamic. It also helps with decision-making by generating insights from complex data, allowing for better resource management and innovation. Furthermore, generative models can predict the future behavior of machines, systems, and workflows, leading to better forecasting and planning.
GenAI significantly enhances AI-based DTs by enabling adaptive modeling, creative optimization, and real-time problem-solving, which is crucial for Industry 5.0. Its integration strengthens human–machine collaboration, ensuring that industrial processes remain productive and human-centric while achieving sustainability goals. By supporting personalized solutions and resilient systems, genAI enables enterprises to transition from mass automation to more flexible and energy-efficient production. In the perspective of Industry 6.0, genAI will drive the evolution of self-organizing, intelligent ecosystems in which DTs continuously learn, adapt, and co-create with humans and machines. This advancement underscores genAI’s crucial role in achieving global optimization, ethical governance, and circular economy practices. Therefore, genAI with DTs represents a transformative contribution to the future of industrial innovation, sustainable development, and socio-technical progress.

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/app151810102/s1: PRISMA 2020 checklist (partial only). Reference [93] is cited in the Supplementary Materials.

Author Contributions

Conceptualization, I.R., D.M., A.P., O.M. and M.K.; methodology, I.R., D.M., A.P., O.M. and M.K.; software, I.R. and D.M.; validation, I.R. and D.M.; formal analysis, I.R. and D.M.; investigation, I.R., D.M., A.P., O.M. and M.K.; resources, I.R., D.M., A.P., O.M. and M.K.; data curation, I.R., D.M., A.P., O.M. and M.K.; writing—original draft preparation, I.R., D.M., A.P., O.M. and M.K.; writing—review and editing, I.R., D.M., A.P., O.M. and M.K.; visualization, I.R. and D.M.; supervision, I.R.; project administration, I.R.; funding acquisition, I.R. All authors have read and agreed to the published version of the manuscript.

Funding

The work presented in this paper has been financed under grant to maintain the research potential of Kazimierz Wielki University.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Data set not generated.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
AIArtificial Intelligence
DTDigital twin
GANGenerative Adversarial Networks
genAIGenerative AI
GPTGenerative Pre-trained Transformer
HMMHidden Markov model
IoTInternet of Things
VAEVariational autoencoders
NLPNatural language processing

References

  1. Lee, C.I.; Chen, J.H.; Kohli, M.D.; Smith, A.D.; Liao, J.M. Generative Artificial Intelligence. J. Am. Coll. Radiol. 2024, 21, 1318–1320. [Google Scholar] [CrossRef]
  2. Rojek, I.; Mikołajewski, D.; Dostatni, E. Digital Twins in Product Lifecycle for Sustainability in Manufacturing and Maintenance. Appl. Sci. 2021, 11, 31. [Google Scholar] [CrossRef]
  3. Menaguale, O. Digital twin and cultural heritage—The future of society built on history and art. In The Digital Twin; Springer International Publishing: Cham, Switzerland, 2023; pp. 1081–1111. [Google Scholar]
  4. Rojek, I.; Mikołajewski, D.; Dostatni, E.; Kopowski, J. Specificity of 3D Printing and AI-Based Optimization of Medical Devices Using the Example of a Group of Exoskeletons. Appl. Sci. 2023, 13, 1060. [Google Scholar] [CrossRef]
  5. Rojek, I.; Dostatni, E.; Mikołajewski, D.; Pawłowski, L.; Wegrzyn-Wolska, K. Modern approach to sustainable production in the context of Industry 4.0. Bull. Pol. Acad. Sci. Tech. Sci. 2022, 70, e143828. [Google Scholar] [CrossRef]
  6. Hawkinson, E. Automationin Education with Digital Twins: Trends and Issues. Int. J. Open Distance E-Learn. 2022, 8. [Google Scholar] [CrossRef]
  7. Dihan, M.S.; Akash, A.I.; Tasneem, Z.; Das, P.; Das, S.K.; Islam, M.R.; Islam, M.M.; Badal, F.R.; Ali, M.F.; Ahamed, M.H.; et al. Digital twin: Data exploration, architecture, implementation and future. Heliyon 2024, 10, e26503. [Google Scholar] [CrossRef]
  8. Meijer, C.; Uh, H.W.; El Bouhaddani, S. Digital Twins in Healthcare: Methodological Challenges and Opportunities. J. Pers. Med. 2023, 13, 1522. [Google Scholar] [CrossRef] [PubMed]
  9. Gkontzis, A.F.; Kotsiantis, S.; Feretzakis, G.; Verykios, V.S. Enhancing Urban Resilience: Smart City Data Analyses, Forecasts, and Digital Twin Techniques at the Neighborhood Level. Future Internet 2024, 16, 47. [Google Scholar] [CrossRef]
  10. Avanzato, R.; Beritelli, F.; Lombardo, A.; Ricci, C. Heart DT: Monitoring and Preventing Cardiac Pathologies Using AI and IoT Sensors. Future Internet 2023, 15, 223. [Google Scholar] [CrossRef]
  11. Orlova, E.V. Design Technology and AI-Based Decision Making Model for Digital Twin Engineering. Future Internet 2022, 14, 248. [Google Scholar] [CrossRef]
  12. Singh, Y.; Hathaway, Q.A.; Erickson, B.J. Generative AI in oncological imaging: Revolutionizing cancer detection and diagnosis. Oncotarget 2024, 15, 607–608. [Google Scholar] [CrossRef] [PubMed]
  13. Chandra, S.; Prakash, P.K.S.; Samanta, S.; Chilukuri, S. ClinicalGAN: Powering patient monitoring in clinical trials with patient digital twins. Sci. Rep. 2024, 14, 12236. [Google Scholar] [CrossRef]
  14. Padoan, A.; Plebani, M. Dynamic mirroring: Unveiling the role of digital twins, artificial intelligence and synthetic data for personalized medicine in laboratory medicine. Clin. Chem. Lab. Med. 2024, 62, 2156–2161. [Google Scholar] [CrossRef]
  15. Bordukova, M.; Makarov, N.; Rodriguez-Esteban, R.; Schmich, F.; Menden, M.P. Generative artificial intelligence empowers digital twins in drug discovery and clinical trials. Expert Opin. Drug Discov. 2024, 19, 33–42. [Google Scholar] [CrossRef]
  16. Nyholm, S. Is Academic Enhancement Possible by Means of Generative AI-Based Digital Twins? Am. J. Bioeth. 2023, 23, 44–47. [Google Scholar] [CrossRef]
  17. Hueso, M.; Álvarez, R.; Marí, D.; Ribas-Ripoll, V.; Lekadir, K.; Vellido, A. Is generative artificial intelligence the next step toward a personalized hemodialysis? Rev. Investig. Clin. Organo Hosp. Enfermedades Nutr. 2023, 75, 309–317. [Google Scholar] [CrossRef]
  18. Du, B.; Du, H.; Liu, H.; Niyato, D.; Xin, P.; Yu, J.; Qi, M.; Tang, Y. YOLO-Based Semantic Communication With Generative AI-Aided Resource Allocation for Digital Twins Construction. IEEE Internet Things J. 2023, 11, 7664–7678. [Google Scholar] [CrossRef]
  19. Wen, J.; Kang, J.; Niyato, D.; Zhang, Y.; Mao, S. Sustainable Diffusion-based Incentive Mechanism for Generative AI-driven Digital Twins in Industrial Cyber-Physical Systems. IEEE Trans. Ind. Cyber-Phys. Syst. 2024, 3, 139–149. [Google Scholar] [CrossRef]
  20. Paul, G.; Abele, N.D.; Kluth, K. A Review and Qualitative Meta-Analyss of Digital Human Modeling and Cyber-Physical-Systems in Ergonomics 4.0. IISE Trans. Occup. Ergon. Hum. Factors 2021, 9, 111–123. [Google Scholar] [CrossRef]
  21. Segovia, M.; Garcia-Alfaro, J. Design, Modeling and Implementation of Digital Twins. Sensors 2022, 22, 5396. [Google Scholar] [CrossRef] [PubMed]
  22. Qiu, C.; Zhou, S.; Liu, Z.; Gao, Q.; Tan, J. Digital assembly technology based on augmented reality and digital twins: A review. Virtual Real. Intell. Hardw. 2019, 1, 597–610. [Google Scholar] [CrossRef]
  23. Murgod, T.R.; Sundaram, S.M.; Mahanthesha, U.; Murugesan, P. A Survey of Digital Twin for Industry 4.0: Benefits, Challenges and Opportunities. SN Comput. Sci. 2024, 5, 76. [Google Scholar] [CrossRef]
  24. Upadhyay, D.; Sharma, T.; Fatima, A. Impact of AI-Driven Digital Twins in Industry 4.0: An Exploratory Analysis. Int. Res. J. Adv. Eng. Manag. 2024, 2, 1548–1557. [Google Scholar]
  25. Espina-Romero, L.; Gutiérrez Hurtado, H.; Ríos Parra, D.; Vilchez Pirela, R.A.; Talavera-Aguirre, R.; Ochoa-Díaz, A. Challenges and Opportunities in the Implementation of AI in Manufacturing: A Bibliometric Analysis. Sci 2024, 6, 60. [Google Scholar] [CrossRef]
  26. Arruda, H.M.; Bavaresco, R.S.; Kunst, R.; Bugs, E.F.; Pesenti, G.C.; Barbosa, J.L.V. Data Science Methods and Tools for Industry 4.0: A Systematic Literature Review and Taxonomy. Sensors 2023, 23, 5010. [Google Scholar] [CrossRef] [PubMed]
  27. Gürses, A.; Reddy, G.; Masrur, S.; Özdemir, Ö.; Güvenç, I.; Sichitiu, M.L.; Sahin, A.; Alkhateeb, A.; Dutta, R. Digital Twins for Supporting AI Research with Autonomous Vehicle Networks. arXiv 2024, arXiv:2404.00954. [Google Scholar] [CrossRef]
  28. Zhu, Z.; Liu, C.; Xu, X. Visualisation of the digital twin data in manufacturing by using augmented reality. Procedia CIRP 2019, 81, 898–903. [Google Scholar] [CrossRef]
  29. Boje, C.; Guerriero, A.; Kubicki, S.; Rezgui, Y. Towards a semantic Construction Digital Twin: Directions for future research. Autom. Constr. 2020, 114, 103179. [Google Scholar] [CrossRef]
  30. Afzal, M.; Li, R.Y.M.; Shoaib, M.; Ayyub, M.F.; Tagliabue, L.C.; Bilal, M.; Ghafoor, H.; Manta, O. Delving into the Digital Twin Developments and Applications in the Construction Industry: A PRISMA Approach. Sustainability 2023, 15, 16436. [Google Scholar] [CrossRef]
  31. Li, S.; Lin, X.; Li, G.; Chen, L.; Liao, S.; Wang, J.; Li, J. DPG-DT: Differentially Private Generative Digital Twin for Imbalanced Learning in Industrial IoT. In Proceedings of the 2023 19th International Conference on Mobility, Sensing and Networking (MSN), Nanjing, China, 14–16 December 2023; pp. 270–276. [Google Scholar] [CrossRef]
  32. Rojek, I.; Macko, M.; Mikołajewski, D.; Saga, M.; Burczynski, T. Modern methods in the field of machine modeling and simulation as a research and practical issue related to Industry 4.0. Bull. Pol. Acad. Sci. Tech. Sci. 2021, 69, e136719. [Google Scholar] [CrossRef]
  33. Yang, X.; Liu, X.; Zhang, H.; Fu, L.; Yu, Y. An ontology-based shop-floor digital twin configuration approach. Procedia CIRP 2023, 120, 326–331. [Google Scholar] [CrossRef]
  34. Dai, Y.; Wang, S.; Xiong, N.N.; Guo, W. A survey on knowledge graph embedding: Approaches, applications and benchmarks. Electronics 2020, 9, 750. [Google Scholar] [CrossRef]
  35. Xia, Y.; Xiao, Z.; Jazdi, N.; Weyrich, M. Generation of Asset Administration Shell with Large Language Model Agents: Towards Semantic Interoperability in Digital Twins in the Context of Industry 4.0. IEEE Access 2024, 12, 84863–84877. [Google Scholar] [CrossRef]
  36. Amangeldy, B.; Tasmurzayev, N.; Imankulov, T.; Baigarayeva, Z.; Izmailov, N.; Riza, T.; Abdukarimov, A.; Mukazhan, M.; Zhumagulov, B. AI-Powered Building Ecosystems: A Narrative Mapping Review on the Integration of Digital Twins and LLMs for Proactive Comfort, IEQ, andEnergy Management. Sensors 2025, 25, 5265. [Google Scholar] [CrossRef]
  37. Zamini, M.; Reza, H.; Rabiei, M. A review of knowledge graph completion. Information 2022, 13, 396. [Google Scholar] [CrossRef]
  38. Cao, Y.; Li, S.; Liu, Y.; Yan, Z.; Dai, Y.; Yu, P.; Sun, L. A survey of ai-generated content (aigc). ACM Comput. Surv. 2025, 57, 1–38. [Google Scholar] [CrossRef]
  39. Gao, R.X.; Krüger, J.; Merklein, M.; Möhring, H.-C.; Váncza, J. Artificial Intelligence in manufacturing: State of the art, perspectives, and future directions. CIRP Ann. 2024, 73, 723–749. [Google Scholar] [CrossRef]
  40. Rojek, I.; Mikołajewski, D.; Mroziński, A.; Macko, M. Machine Learning- and Artificial Intelligence-Derived Prediction for Home Smart Energy Systems with PV Installation and Battery Energy Storage. Energies 2023, 16, 6613. [Google Scholar] [CrossRef]
  41. Rojek, I.; Mroziński, A.; Kotlarz, P.; Macko, M.; Mikołajewski, D. AI-Based Computational Model in Sustainable Transformation of Energy Markets. Energies 2023, 16, 8059. [Google Scholar] [CrossRef]
  42. Xu, M.; Niyato, D.; Chen, J.; Zhang, H.; Kang, J.; Xiong, Z.; Mao, S.; Han, Z. Generative AI-empowered Simulation for Autonomous Driving in Vehicular Mixed Reality Metaverses. arXiv 2023, arXiv:2302.08418. [Google Scholar] [CrossRef]
  43. Liu, Y.Q.; Du, H.Y.; Niyato, D.; Kang, J.; Xiong, Z.; Kim, D.I.; Jamalipour, A. Deep Generative Model and Its Applications for Effective Wireless Network Management: Tutorial and Case Study. IEEE Wirel. Commun. 2024, 31, 199–207. [Google Scholar] [CrossRef]
  44. Martini, B.; Bellisario, D.; Coletti, P. Human-Centered and Sustainable Artificial Intelligence in Industry 5.0: Challenges and Perspectives. Sustainability 2024, 16, 5448. [Google Scholar] [CrossRef]
  45. Hang, C.-N.; Yu, P.-D.; Morabito, R.; Tan, C.-W. Large Language Models Meet Next-Generation Networking Technologies: A Review. Future Internet 2024, 16, 365. [Google Scholar] [CrossRef]
  46. Scalise, P.; Boeding, M.; Hempel, M.; Sharif, H.; Delloiacovo, J.; Reed, J. A Systematic Survey on 5G and 6G Security Considerations, Challenges, Trends, and Research Areas. Future Internet 2024, 16, 67. [Google Scholar] [CrossRef]
  47. Musa, A.A.; Hussaini, A.; Qian, C.; Guo, Y.; Yu, W. Open Radio Access Networks for Smart IoT Systems: State of Art and Future Directions. Future Internet 2023, 15, 380. [Google Scholar] [CrossRef]
  48. Hu, F.; Wang, C.; Wu, X. Generative Artificial Intelligence-Enabled Facility Layout Design Paradigm. Appl. Sci. 2025, 15, 5697. [Google Scholar] [CrossRef]
  49. Hosseini-Nasab, H.; Fereidouni, S.; Fatemi Ghomi, S.M.T.; Fakhrzad, M.B. Classification of facility layout problems: A review study. Int. J. Adv. Manuf. Technol. 2018, 94, 957–977. [Google Scholar] [CrossRef]
  50. Heinbach, B.; Burggräf, P.; Wagner, J. Gym-flp: A Python Package for Training Reinforcement Learning Algorithms on Facility Layout Problems. Oper. Res. Forum 2024, 5, 20. [Google Scholar] [CrossRef]
  51. Sulaiman, S.S.; Jancy, P.L.; Muthiah, A.; Janakiraman, V.; Gnanaraj, S.J.P. An evolutionary optimal green layout design for a production facility by simulated annealing algorithm. Mater. Today Proc. 2021, 47, 4423–4430. [Google Scholar] [CrossRef]
  52. Klar, M.; Langlotz, P.; Aurich, J.C. A framework for automated multiobjective factory layout planning using reinforcement learning. Procedia CIRP 2022, 112, 555–560. [Google Scholar] [CrossRef]
  53. Pérez-Gosende, P.; Mula, J.; Díaz-Madroñero, M. A bottom-up multi-objective optimisation approach to dynamic facility layout planning. Int. J. Prod. Res. 2024, 62, 626–643. [Google Scholar] [CrossRef]
  54. Caneparo, L. Semantic knowledge in generation of 3D layouts for decision-making. Autom. Constr. 2022, 134, 104012. [Google Scholar] [CrossRef]
  55. Yao, Z.; Chen, Y.; Cui, J.; Zhang, S.; Li, S.; Hao, A. Conditional room layout generation based on graph neural networks. Comput. Graph. 2024, 122, 103971. [Google Scholar] [CrossRef]
  56. Shi, Y.; Shang, M.; Qi, Z. Intelligent layout generation based on deep generative models: A comprehensive survey. Inf. Fusion 2023, 100, 101940. [Google Scholar] [CrossRef]
  57. Cohen, Y.; Aperstein, Y. Generative Shop floor Layout Design: Challenges and Proposed Modelling Approach. IFAC-Papers OnLine 2024, 58, 748–753. [Google Scholar] [CrossRef]
  58. Tian, L.; Zhou, X.; Wu, Y.-P.; Zhou, W.-T.; Zhang, J.-H.; Zhang, T.-S. Knowledge graph and knowledge reasoning: A systematic review. J. Electron. Sci. Technol. 2022, 20, 100159. [Google Scholar] [CrossRef]
  59. Różanowski, K.; Piotrowski, Z.; Ciolek, M. Mobile Application for Driver’s Health Status Remote Monitoring. In Proceedings of the 9th International Wireless Communications and Mobile Computing Conference (IWCMC), Sardinia, Italy, 1–5 July 2013; pp. 1738–1743. [Google Scholar]
  60. Sondej, T.; Zawadzka, S. Influence of cuff pressures of automatic sphygmomanometers on pulse oximetry measurements. Measurement 2022, 187, 110329. [Google Scholar] [CrossRef]
  61. Akter, N.; Molnar, A.; Georgakopoulos, D. Toward Improving Human Training by Combining Wearable Full-Body IoT Sensors and Machine Learning. Sensors 2024, 24, 7351. [Google Scholar] [CrossRef]
  62. Chen, X.; Eder, M.A.; Shihavuddin, A.; Zheng, D. A Human-Cyber-Physical System toward Intelligent Wind Turbine Operation and Maintenance. Sustainability 2021, 13, 561. [Google Scholar] [CrossRef]
  63. Mikołajczyk, T.; Kłodowski, A.; Mikołajewska, E.; Walkowiak, P.; Berjano, P.; Villafañe, J.H.; Aggogeri, F.; Borboni, A.; Fausti, D.; Petrogalli, G. Design and control of system for elbow rehabilitation: Preliminary findings. Adv. Clin. Exp. Med. 2018, 27, 1661–1669. [Google Scholar] [CrossRef]
  64. Kawala-Janik, A.; Podpora, M.; Baranowski, J.; Bauer, W.; Pelc, M. Innovative Approach in Analysis of EEG and EMG Signals—Comparision of the Two Novel Methods. In Proceedings of the 19th International Conference on Methods and Models in Automation and Robotics (MMAR), Miedzyzdroje, Poland, 2–5 September 2014; pp. 804–807. [Google Scholar]
  65. Bickel, S.; Goetz, S.; Wartzack, S. Symbol Detection in Mechanical Engineering Sketches: Experimental Study on Principle Sketches with Synthetic Data Generation and Deep Learning. Appl. Sci. 2024, 14, 6106. [Google Scholar] [CrossRef]
  66. Miny, T.; Thies, M.; Lukic, L.; Käbisch, S.; Oladipupo, K.; Diedrich, C.; Kleinert, T. Overview and Comparison of Asset Information Model Standards. IEEE Access 2023, 11, 99189–99221. [Google Scholar] [CrossRef]
  67. Salins, S.S.; Zaidi, S.A.R.; Deepak, D.; Sachidananda, H.K. Design of an improved layout for a steel processing facility using SLP and lean Manufacturing techniques. Int. J. Interact. Des. Manuf. (IJIDeM) 2024, 18, 3827–3848. [Google Scholar] [CrossRef]
  68. Burggräf, P.; Adlon, T.; Lehde, N.; Lindholm, N. Uncovering the behaviour of facility layout problem solutions in relation to factory design applications. Procedia CIRP 2024, 126, 93–98. [Google Scholar] [CrossRef]
  69. Rojek, I.; Mikołajewski, D.; Kotlarz, P.; Tyburek, K.; Kopowski, J.; Dostatni, E. Traditional Artificial Neural Networks Versus Deep Learning in Optimization of Material Aspects of 3D Printing. Materials 2021, 14, 7625. [Google Scholar] [CrossRef]
  70. Wang, W.Y.; Zhang, S.; Li, G.; Lu, J.; Ren, Y.; Wang, X.; Gao, X.; Su, Y.; Song, H.; Li, J. Artificial intelligence enabled smart design and manufacturing of advanced materials: The endless Frontier in AI+era. Mater. Genome Eng. Adv. 2024, 2, e56. [Google Scholar] [CrossRef]
  71. Rao, S.; Neethirajan, S. Computational Architectures for Precision Dairy Nutrition Digital Twins: A Technical Review and Implementation Framework. Sensors 2025, 25, 4899. [Google Scholar] [CrossRef]
  72. Rojek, I.; Kotlarz, P.; Dorożyński, J.; Mikołajewski, D. Sixth-Generation (6G) Networks for Improved Machine-to-Machine (M2M) Communication in Industry 4.0. Electronics 2024, 13, 1832. [Google Scholar] [CrossRef]
  73. Zachariades, C.; Xavier, V. A Review of Artificial Intelligence Techniques in Fault Diagnosis of Electric Machines. Sensors 2025, 25, 5128. [Google Scholar] [CrossRef] [PubMed]
  74. Coelho, P.; Bessa, C.; Landeck, J.; Silva, C. Industry 5.0: The arising of a concept. Procedia Comput. Sci. 2023, 217, 1137–1144. [Google Scholar] [CrossRef]
  75. Jin, J.; Xu, H.; Leng, B. Adaptive Points Sampling for Implicit Field Reconstruction of Industrial Digital Twin. Sensors 2022, 22, 6630. [Google Scholar] [CrossRef]
  76. Akar, N.; Turgay, S. Optimizing Cellular Manufacturing Facility Layout Design through Digital Twin Simulation: A Case Study. Ind. Eng. Innov. Manag. 2023, 6, 1–12. [Google Scholar] [CrossRef]
  77. Martinez, E.M.; Ponce, P.; Macias, I.; Molina, A. Automation Pyramidas Constructor for a Complete Digital Twin, Case Study: A Didactic Manufacturing System. Sensors 2021, 21, 4656. [Google Scholar] [CrossRef]
  78. Rojek, I.; Kowal, M.; Stoic, A. Predictive compensation of thermal deformations of ball screws in cnc machines using neural networks. Teh. Vjesn. Tech. Gaz. 2017, 24, 1697–1703. [Google Scholar]
  79. Gao, T.; Wang, L.; Song, W.; Cheng, Y.; Zuo, Y.; Xiang, F.; Zhang, H.; Tao, F. Ten industrial software towards smart manufacturing. J. Manuf. Syst. 2025, 79, 255–285. [Google Scholar] [CrossRef]
  80. Singh, R.; Akram, S.V.; Gehlot, A.; Buddhi, D.; Priyadarshi, N.; Twala, B. Energy System 4.0: Digitalization of the Energy Sector with Inclination towards Sustainability. Sensors 2022, 22, 6619. [Google Scholar] [CrossRef] [PubMed]
  81. Damisa, U.; Nwulu, N.I.; Siano, P. Towards Blockchain-Based Energy Trading: A Smart Contract Implementation of Energy Double Auction and Spinning Reserve Trading. Energies 2022, 15, 4084. [Google Scholar] [CrossRef]
  82. Huang, Z.; Shen, Y.; Li, J.; Fey, M.; Brecher, C. A Survey on AI-Driven Digital Twins in Industry 4.0: Smart Manufacturing and Advanced Robotics. Sensors 2021, 21, 6340. [Google Scholar] [CrossRef]
  83. Liu, X.; Qiu, C.; Shi, J.; Huang, J.; Zhu, C.; Ni, Z.; Zhu, M.; Liu, T. A digital twin modeling method for production resources of shop floor. Int. J. Adv. Manuf. Technol. 2023, 128, 743–761. [Google Scholar] [CrossRef]
  84. Rojek, I. Hybrid Neural Networks as Prediction Models. In Artificial Intelligence and Soft Computing, Lecture Notes in Artificial Intelligence; Rutkowski, L., Scherer, R., Tadeusiewicz, R., Zadeh, L.A., Zurada, J.M., Eds.; Springer: Berlin/Heidelberg, Germany, 2010; pp. 88–95. [Google Scholar]
  85. Han, X.; Lin, Z.; Clark, C.; Vucetic, B.; Lomax, S. AI Based Digital Twin Model for Cattle Caring. Sensors 2022, 22, 7118. [Google Scholar] [CrossRef]
  86. Shaposhnyk, O.; Lai, K.; Wolbring, G.; Shmerko, V.; Yanushkevich, S. Next Generation Computing and Communication Hub for First Responders in Smart Cities. Sensors 2024, 24, 2366. [Google Scholar] [CrossRef] [PubMed]
  87. Rojek, I. Models for Better Environmental Intelligent Management within Water Supply Systems. Water Resour. Manag. 2014, 28, 3875–3890. [Google Scholar] [CrossRef]
  88. Choi, H.; Yu, S.; Lee, D.; Noh, S.D.; Ji, S.; Kim, H.; Yoon, H.; Kwon, M.; Han, J. Optimization of the Factory Layout and Production Flow Using Production-Simulation-Based Reinforcement Learning. Machines 2024, 12, 390. [Google Scholar] [CrossRef]
  89. Tasmurzayev, N.; Amangeldy, B.; Imanbek, B.; Baigarayeva, Z.; Imankulov, T.; Dikhanbayeva, G.; Amangeldi, I.; Sharipova, S. Digital Cardiovascular Twins, AI Agents, and Sensor Data: A Narrative Review from System Architecture to Proactive Heart Health. Sensors 2025, 25, 5272. [Google Scholar] [CrossRef]
  90. Xie, R.; Chen, M.; Liu, W.; Jian, H.; Shi, Y. Digital Twin Technologies for Turbo machinery in a Life Cycle Perspective: A Review. Sustainability 2021, 13, 2495. [Google Scholar] [CrossRef]
  91. Negri, V.; Zanella, S.; Mingotti, A.; Tinarelli, R.; Peretto, L.; Barchi, F.; Acquaviva, A. Towars the DT of educational building: An AI-based distributed measurement system for the power forecasting. In Proceedings of the 2024 IEEE 14th International Workshop on Applied Measurements for Power Systems (AMPS), Caserta, Italy, 18–20 September 2024; pp. 1–6. [Google Scholar]
  92. Hevesli, M.; Mohammed Seid, A.; Erbad, A.; Abdallah, M.M. Energy efficient delay-aware design for MEC-enabled DT-assisted air-ground network. In Proceedings of the 2024 IEEE 35th International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Valencia, Spain, 2–5 September 2024; p. 106. [Google Scholar]
  93. Page, M.J.; McKenzie, J.E.; Bossuyt, P.M.; Boutron, I.; Hoffmann, T.C.; Mulrow, C.D.; Shamseer, L.; Tetzlaff, J.M.; Akl, E.A.; Brennan, S.E.; et al. The PRISMA 2020 statement: An updated guideline for reporting systematic reviews. BMJ 2021, 372, n71. [Google Scholar] [CrossRef] [PubMed]
Applsci 15 10102 g001
Figure 1. DTs inIndustry 4.0/5.0 emerging against the background of the development of AI and ML (own version).
Figure 1. DTs inIndustry 4.0/5.0 emerging against the background of the development of AI and ML (own version).
Applsci 15 10102 g001
Applsci 15 10102 g002
Figure 2. Simple schematic definition of genAI-based DT.
Figure 2. Simple schematic definition of genAI-based DT.
Applsci 15 10102 g002
Applsci 15 10102 g003
Figure 3. Bibliometric analysis procedure.
Figure 3. Bibliometric analysis procedure.
Applsci 15 10102 g003
Applsci 15 10102 g004
Figure 4. PRISMA 2020 flow diagram.
Figure 4. PRISMA 2020 flow diagram.
Applsci 15 10102 g004
Applsci 15 10102 g005
Figure 5. Documents by year.
Figure 5. Documents by year.
Applsci 15 10102 g005
Applsci 15 10102 g006
Figure 6. Documents by area.
Figure 6. Documents by area.
Applsci 15 10102 g006
Applsci 15 10102 g007
Figure 7. Documents by type.
Figure 7. Documents by type.
Applsci 15 10102 g007
Applsci 15 10102 g008
Figure 8. Documents by country.
Figure 8. Documents by country.
Applsci 15 10102 g008
Applsci 15 10102 g009
Figure 9. Idea of the hybrid genAI-based DT of new generation [36].
Figure 9. Idea of the hybrid genAI-based DT of new generation [36].
Applsci 15 10102 g009
Applsci 15 10102 g010
Figure 10. Operation, training and optimization within genAI-based DT of new generation [37].
Figure 10. Operation, training and optimization within genAI-based DT of new generation [37].
Applsci 15 10102 g010
Applsci 15 10102 g011
Figure 11. GenAI-driven environmental adjustment of DTs [38].
Figure 11. GenAI-driven environmental adjustment of DTs [38].
Applsci 15 10102 g011
Applsci 15 10102 g012
Figure 12. GenAI-based DTs within edge-cloud architecture [41].
Figure 12. GenAI-based DTs within edge-cloud architecture [41].
Applsci 15 10102 g012
Applsci 15 10102 g013
Figure 13. Layered architecture of genAI-based DTs [71].
Figure 13. Layered architecture of genAI-based DTs [71].
Applsci 15 10102 g013
Applsci 15 10102 g014
Figure 14. Multi-sensor signal fusion and image feature extraction for genAI-based DTs purposes [73].
Figure 14. Multi-sensor signal fusion and image feature extraction for genAI-based DTs purposes [73].
Applsci 15 10102 g014
Applsci 15 10102 g015
Figure 15. Resultant place of genAI-based DTs within smart factory [80].
Figure 15. Resultant place of genAI-based DTs within smart factory [80].
Applsci 15 10102 g015
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Rojek, I.; Mikołajewski, D.; Piszcz, A.; Małolepsza, O.; Kozielski, M. Role of Generative AI in AI-Based Digital Twins in Industry 5.0 and Evolution to Industry 6.0. Appl. Sci. 2025, 15, 10102. https://doi.org/10.3390/app151810102

AMA Style

Rojek I, Mikołajewski D, Piszcz A, Małolepsza O, Kozielski M. Role of Generative AI in AI-Based Digital Twins in Industry 5.0 and Evolution to Industry 6.0. Applied Sciences. 2025; 15(18):10102. https://doi.org/10.3390/app151810102

Chicago/Turabian Style

Rojek, Izabela, Dariusz Mikołajewski, Adrianna Piszcz, Olga Małolepsza, and Mirosław Kozielski. 2025. "Role of Generative AI in AI-Based Digital Twins in Industry 5.0 and Evolution to Industry 6.0" Applied Sciences 15, no. 18: 10102. https://doi.org/10.3390/app151810102

APA Style

Rojek, I., Mikołajewski, D., Piszcz, A., Małolepsza, O., & Kozielski, M. (2025). Role of Generative AI in AI-Based Digital Twins in Industry 5.0 and Evolution to Industry 6.0. Applied Sciences, 15(18), 10102. https://doi.org/10.3390/app151810102

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop