A Comprehensive Review of AI-Based Digital Twin Applications in Manufacturing: Integration Across Operator, Product, and Process Dimensions
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors summarised the digital twin's applications for manufacturing in operator, product, and process aspects. However, there are some questions the authors need to answer before publication. Here are some comments:
-
The authors mentioned this is a digital twin application, but readers cannot find the relationship between physical and digital assets in this manuscript. The authors might think this manuscript just discussed intelligent applications in manufacturing, as it does not provide details about digital twins.
-
What are the new findings from this manuscript? There have been many similar literature reviews for digital twins in manufacturing for operation and process. What are the differences between authors' work and other existing work?
-
“Only five studies addressed the product dimension.” This might not be true. Life cycle analysis, inspection, maintenance, and other topics might be related. The authors need to do more research on this aspect.
-
The authors need to do more research on the machine-learning algorithms for digital twins in manufacturing. Not just from the aspects of algorithm classification but also their impacts on digital twin applications
-
The authors might add section 3.3 for digital twins for products.
-
The authors need to rewrite the discussions. New research questions, goals, and fields need to be discovered.
-
The conclusion is weak. The authors need to provide clear take-home messages for readers.
- This special issue is "Generative Artificial Intelligence", this manuscript might not fit.
Author Response
Response to Reviewer's Comments
Title of the Paper: A Comprehensive Review of AI-Based Digital Twins Applications in Manufacturing: Integration Across Operator, Product, and Process Dimensions
Dear reviewer, we are deeply grateful for the comments made to the document. Below, we present your comments along with our responses, highlighted in blue. The corresponding changes have been integrated into the corrected paper. In addition, we have attached a change control where you can compare the original paper with the modifications made.
- The authors mentioned this is a digital twin application, but readers cannot find the relationship between physical and digital assets in this manuscript. The authors might think this manuscript just discussed intelligent applications in manufacturing, as it does not provide details about digital twins.
A new paragraph has been added to the document to explain the connection between physical and digital assets, as well as their bidirectional communication. It explains in a better way how what happens in the real asset is reflected in the virtual one and vice versa.
- What are the new findings from this manuscript? There have been many similar literature reviews for digital twins in manufacturing for operation and process. What are the differences between authors' work and other existing work?
A paragraph mentioning the contribution of the research has been included in the introduction of the paper:
“The main contribution of this work lies in providing a comprehensive classification of digital twin applications along three fundamental dimensions: operator, product, and process. This classification facilitates the identification of potential gaps in each dimension and the purpose of the application. In addition, it enables recognition of the technologies and tools used in their design and implementation. The proposed framework guides the selection of appropriate technologies aligned with the specific objectives of each dimension. It also offers a clear perspective on the areas within the manufacturing industry where digital twins have been widely adopted, as well as those that remain relatively unexplored.”
- “Only five studies addressed the product dimension.” This might not be true. Life cycle analysis, inspection, maintenance, and other topics might be related. The authors need to do more research on this aspect.
There are numerous studies and applications on digital twins; however, not all of them are framed in the manufacturing field. Many of these applications are oriented to sectors such as construction, agriculture, nuclear plants, among others. In this context, the classification proposed in this research, which organizes the applications according to the established dimensions, is not completely applicable to all of them. Therefore, we have specifically selected those digital twin applications that integrate the use of artificial intelligence for data processing.
- The authors need to do more research on the machine-learning algorithms for digital twins in manufacturing. Not just from the aspects of algorithm classification but also their impacts on digital twin applications
The research not only covers the use of Machine Learning algorithms, but also includes Deep Learning models. In each section corresponding to the dimensions of the proposed classification (operator, product and process), it is analyzed how artificial intelligence models are applied to data analysis, highlighting examples such as prediction and optimization.
- The authors might add section 3.3 for digital twins for products.
In the paper, Section 3.3 discusses and describes in detail the applications of digital twins (DT) in the product dimension. In this section, DT applications are presented whose main purpose is to use the information obtained and processed to update and optimize future iterations of the product.
- The authors need to rewrite the discussions. New research questions, goals, and fields need to be discovered.
The “Discussions” section has been reformulated and expanded to include new challenges faced by DTs, such as cybersecurity for information protection, and the analysis of interoperability challenges has been expanded to include new bibliographic sources to support and give depth to the analysis. New scientific articles have been included to further analyze the various challenges that affect the development and implementation of DT.
- The conclusion is weak. The authors need to provide clear take-home messages for readers.
The conclusions of the document have been strengthened to provide clear and specific messages. They now include a more focused summary of key findings, critical challenges identified and key opportunities. In addition, specific future directions for research, such as the integration of cybersecurity and human-centric technologies, are highlighted, ensuring that readers gain a clear understanding of the study's contribution and its potential impact on manufacturing and beyond.
- This special issue is "Generative Artificial Intelligence", this manuscript might not fit.
We consider generative artificial intelligence as a fundamental tool with enormous potential for the development and evolution of digital twins (DT). In this manuscript, we highlight how the integration of generative AI can facilitate significant advances in several key areas of DTs, such as synthetic data generation to train models, simulation of complex scenarios, and real-time process optimization. These capabilities not only complement traditional DT applications, but also open up new possibilities for implementation in sectors such as manufacturing, construction and agriculture.
Furthermore, in prior consultation with the publisher, it was confirmed to us that the focus and contents of this review fit within the scope of the special issue dedicated to “Generative Artificial Intelligence”. The inclusion of this paper in the special issue brings a comprehensive and substantiated analysis on the emerging role of generative AI in the context of DTs, strengthening the connection between these two disruptive technologies and their impact on the digital transformation of the industry.
Reviewer 2 Report
Comments and Suggestions for AuthorsThis research paper reviews the applications of AI-based digital twins in manufacturing, categorising them across three dimensions: operator, product, and process. The operator dimension focuses on improving safety and ergonomics through intelligent assistance. The product dimension uses digital twins to enhance product design and quality. The process dimension employs digital twins to optimise production flows and dynamically reconfigure systems. The paper identifies key challenges, such as interoperability and high implementation costs, whilst highlighting the potential for AI-driven digital twins to advance sustainable, human-centred manufacturing.
The paper addresses the lack of a structured framework for analysing the specific capabilities and trends of digital twins (DTs) in manufacturing while identifying knowledge gaps in current research. The authors note that previous studies have explored specific areas such as human-robot collaboration, plant reconfiguration, production scheduling, and monitoring. This addresses this gap by providing a comprehensive classification of digital twin applications along three fundamental dimensions: operator, product, and process. This classification facilitates the identification of potential gaps in each dimension, the purpose of the application, and the technologies and tools used in their design and implementation.
The main contribution of this review is the way that it classifies works in the literature. While previous studies have often focused on specific applications of digital twins, this research provides a comprehensive classification system along three key dimensions: operator, product, and process. This classification system offers advantages over previous approaches to understanding digital twins in manufacturing, as follows. (1). Provides a holistic view: By considering the operator, product, and process dimensions, the classification system provides a more holistic view of the potential applications of digital twins in manufacturing, (2) Facilitates identification of knowledge gaps, and (3) aids in technology selection and implementation, serving as a guide for practitioners when selecting and implementing digital twin technologies.
The conclusions presented in the research are consistent with the evidence and arguments presented, and they effectively address the main question posed. The research addresses the lack of a structured framework for analysing and understanding the diverse applications of DTs in manufacturing. This gap was identified due to the tendency of previous studies to focus on specific applications rather than providing a holistic perspective. The authors propose a classification system along three dimensions: operator, product, and process. The research presents a comprehensive analysis of DT applications within each dimension, examining various implementations and considering factors such as AI algorithms, visualization tools, data types, and communication protocols. The conclusions reinforce the value of the proposed classification system, emphasising its role as a tool for guiding future research and promoting a more balanced development of DT applications across all three dimensions. The research also acknowledged the ongoing challenges faced by the field, including technological interoperability, data integration, and high implementation costs, advocating for efforts to address these issues to maximise the potential of digital twins in manufacturing.
Most of the cited works are relatively recent, with a significant portion published within the last few years. This suggests that the research is based on up-to-date knowledge and considers the latest developments in the field.
The paper is well-written, with clear and concise English that makes the content accessible to readers. The authors have articulated their ideas with clarity, ensuring the study's main contributions are easy to understand.
In general, the paper is in very good shape, however there is room for some improvements.
Figure 1, is not needed. Its contents are fully described in lines 72-78.
The number of citations is rather small for a comprehensive literature review (87 papers).
There is no discussion about industrial processes performed by humans using handheld tools or hands. You can consider looking at works that do such modelling, under the topic of traditional craft digitisation and modelling.
Author Response
Response to Reviewer's Comments
Title of the Paper: A Comprehensive Review of AI-Based Digital Twins Applications in Manufacturing: Integration Across Operator, Product, and Process Dimensions
Dear reviewer, we are deeply grateful for the comments made to the document. Below, we present your comments along with our responses, highlighted in blue. The corresponding changes have been integrated into the corrected paper. In addition, we have attached a change control where you can compare the original paper with the modifications made.
Figure 1 is not needed. Its contents are fully described in lines 72-78.
We believe that, although the text provides a detailed description of the articles analyzed, the figure provides a visual summary that facilitates the reader's understanding and analysis.
The number of citations is rather small for a comprehensive literature review (87 papers).
Although there are hundreds of studies on digital twins (DT), we have chosen to focus this review on representative works that integrate DT, artificial intelligence (AI) and the dimensions addressed in this research. However, we have expanded the analysis specifically on the product dimension, incorporating a larger number of DT applications to provide a more complete picture.
There is no discussion about industrial processes performed by humans using handheld tools or hands. You can consider looking at works that do such modelling, under the topic of traditional craft digitisation and modelling.
These processes were not included because they did not meet the requirements established for AI-Based Digital Twins (AI-Based DTs). However, work related to human-robot collaboration was considered, where DTs are used to ensure the safety of operators and assist them in the performance of their duties.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe article undertakes a literature review on the use of digital twins in production processes. The authors proposed an analysis of the literature across three domains: processes, products, and users (operators). The article offers an intriguing perspective on digital twin theory from various angles however, this raises several questions:
-
In the article, the authors address the implementation of DT in specific applications, overlooking the frameworks that support their implementation. The literature contains articles on process modelling oriented towards DT theory, AI, and IoT. These themselves form the foundation for building DT solutions and can be considered in terms of processes, operators, and products. For example, they allow the modelling of processes using BPMN notations and their representation within multi-agent systems that simulate device operations. I believe it is worthwhile to at least mention such solutions.
-
Why did the authors limit the scope of their research to the last five years?
-
Regarding the proposed categorisation, it is logical to indicate the processes in which DT participates. However, in the opinion of the reviewer, the authors should better describe the differences related to the categories of product and user (operator). For instance, in section 3.2.1, instead of focusing on discussing the significance of DT for the relationship with the user, they point to process optimisation, improved planning, reduced cost and time, which are features related to the process in which the user and DT are involved.
-
The section “3.2.3. Production Planning” is part of section “3.2 Operator”. PP is the process of managing resources, manpower, schedules, and other aspects of producing goods and services. Why was PP not included in section “3.1. Process”?
-
In section 3.1, the authors categorised process optimisation under strategic optimisation. This is only partially accurate. Within an organisation, various business processes can be executed, for example, sales processes, which are considered at the operational level of the organisation. These can also be optimised.
Author Response
Response to Reviewer's Comments
Title of the Paper: A Comprehensive Review of AI-Based Digital Twins Applications in Manufacturing: Integration Across Operator, Product, and Process Dimensions
Dear reviewer, we are deeply grateful for the comments made to the document. Below, we present your comments along with our responses, highlighted in blue. The corresponding changes have been integrated into the corrected paper. In addition, we have attached a change control where you can compare the original paper with the modifications made.
Why did the authors limit the scope of their research to the last five years?
There are several reviews on digital twins that generally address their impact in various areas, such as manufacturing, healthcare, agriculture and aerospace, among others. However, the focus of our research is specifically focused on the manufacturing domain, with the objective of classifying DT applications within the proposed dimensions. In addition, the initial approach consisted of performing a review of the current state of the art, considering the full integration of artificial intelligence in DT applications.
Regarding the proposed categorisation, it is logical to indicate the processes in which DT participates. However, in the opinion of the reviewer, the authors should better describe the differences related to the categories of product and user (operator). For instance, in section 3.2.1, instead of focusing on discussing the significance of DT for the relationship with the user, they point to process optimisation, improved planning, reduced cost and time, which are features related to the process in which the user and DT are involved.
The focus of this section 3.2.1 has been adjusted to emphasize more explicitly the relationship between operators and digital twins (DT). In particular, DT applications designed to improve the safety of operators during the execution of their tasks, either in human-robot collaboration (HRC) scenarios or in the use of industrial machinery, are presented. Several technologies integrated in DTs are described, such as biometric sensors, artificial intelligence (AI) algorithms for action recognition, and ergonomic monitoring platforms, which contribute to ensure a safer work environment. These modifications in approach seek to clearly highlight how DTs interact directly with operators to prevent risks and optimize safety in their tasks, providing the reader with a more accurate understanding of this key relationship.
The section “3.2.3. Production Planning” is part of section “3.2 Operator”. PP is the process of managing resources, manpower, schedules, and other aspects of producing goods and services. Why was PP not included in section “3.1. Process”?
Indeed, PP (Production Planning) refers to the process of managing resources, labor, schedules and other aspects related to the production of goods and services. In particular, section 3.2.3 describes a Human Digital Twin (HDT) model that includes digital profiles of workers' dynamic skills and behaviors. This model is specifically designed for operators, including those with some kind of disability, and its main objective is to provide accurate information for decision making in the allocation of tasks within production planning. In this way, HDT allows for a person-centered approach, optimizing the assignment of workers according to their specific needs and capabilities.
In section 3.1, the authors categorised process optimisation under strategic optimisation. This is only partially accurate. Within an organisation, various business processes can be executed, for example, sales processes, which are considered at the operational level of the organisation. These can also be optimised.
The aforementioned processes can be effectively optimized; however, in the categorization proposed in this research, we focus specifically on manufacturing activities that involve the transformation of raw materials into final products. This approach allows us to analyze more precisely how digital twins (DT) contribute to optimize key processes within the production cycle, from planning and execution to real-time control and adjustment.