Digital Twin in Industry: State-of-the-Art

This paper, titled “Digital Twin in Industry: State-of-the-Art,” authored by Fei Tao, He Zhang, Ang Liu, and A. Y. C. Nee (2018), provides a comprehensive review of the current advancements and applications of Digital Twins (DTs) across various industrial sectors. It positions Digital Twin as a highly promising enabling technology essential for realizing smart manufacturing and Industry 4.0. A defining characteristic of DTs is their seamless integration between cyber and physical spaces.

The authors highlight that despite the concept of DT being initially proposed almost 15 years prior (in 2003) and many successful DT applications in areas like product design, production, and prognostics and health management (PHM), there was a notable absence of a dedicated review focusing on DT applications specifically within industry. This paper aims to bridge that gap by thoroughly examining the state-of-the-art in DT research, covering its key components, current development, major industrial applications, and outlining existing challenges and potential directions for future research.

Methodology and Scope: The review’s methodology involved a detailed examination of various sources to ensure completeness:

• 50 journal and conference articles published from January 2003 to April 2018.
• 8 patents related to DTs.
• Best practices of 6 leading companies. These sources were collected from multiple databases, including ProQuest, ScienceDirect, Scopus, Google Scholar, IEEE Xplore, and Google Patent. To enhance reliability, three searchers independently conducted searches multiple times, comparing and compiling their findings. Papers were evaluated for relevance based on abstracts, introductions, and conclusions, ensuring that keywords like “digital” or “twin” specifically referred to “digital twin”. The paper also explains that an iterative process was used to build a comprehensive list of keywords, leveraging highly cited articles and adding new keywords as needed, with authors (experts in cyber-physical systems, smart manufacturing, and manufacturing service) abstracting keywords to reduce bias.

The paper is structured to answer five key research questions:

1. What is DT?
2. What is the current development of DTs?
3. Which industrial areas are most applicable for DTs?
4. How to implement DTs?
5. What are the main challenges in deploying DTs?

Concept and History of Digital Twins: The concept of the DT was first introduced by Grieves in 2003, initially as a three-part model comprising a physical product, a virtual product, and their connections. The enabling technologies for DTs have since experienced exponential growth. In 2012, NASA formalized the definition of a DT as a multiphysics, multiscale, probabilistic, ultrafidelity simulation that reflects the state of a corresponding twin based on historical data, real-time sensor data, and physical models. The meaning of DTs has become increasingly concrete, leading to specialized notions such as the airframe digital twin (ADT) and experimental digital twin (EDT).

While some researchers emphasize simulation as the core of DTs, others argue for the three-dimension model (physical, virtual, connection). Tao et al. expanded this to a five-dimension model, including physical part, virtual part, connection, data, and service, with data at the center as a precondition for creating new knowledge, and the connection part bridging all elements.

The history of DTs is divided into three stages:

• Formation (2003-2011): The initial concept was proposed, but technological foundations like cloud computing, big data, IoT, and sensor technologies were still maturing, leading to few publications.
• Incubation (2011-2014): The first journal article on DT for aircraft structural life prediction appeared in 2011, and NASA formalized the definition in 2012, leading to increased research efforts.
• Growth (2014-present): The first white paper was published in 2014, and Gartner classified DTs as one of the top ten promising technological trends in 2017 and 2018, indicating its rapid expansion beyond aerospace to many different industries.

Current Development and Theoretical Foundations of DTs in Industry: The theoretical foundations of DTs draw from various disciplines, including information science, production engineering, data science, and computer science. Key areas include:

• DT Modeling, Simulation, and VV&A (Verification, Validation, and Accreditation): This involves physical, virtual, connection, data, and service modeling. Physical modeling extracts features, virtual modeling mirrors them, connection modeling maintains links, data modeling handles definition and storage, and service modeling identifies and upgrades services. Simulation theories are used for operation analysis, while VV&A validates model veracity and provides confidence levels. A current challenge is the lack of a unified, generic DT modeling framework that considers all five dimensions.
• Data Fusion: DTs necessitate handling massive data from physical, virtual, and historical sources. Data fusion involves preprocessing (cleaning, conversion, filtering), data mining (fuzzy sets, rule-based reasoning, intelligent algorithms), and data optimization to discover data evolution laws. Challenges include reducing data dimensionality and integrating heterogeneous data.
• Interaction and Collaboration: This involves physical-physical, virtual-virtual, and virtual-physical interactions. Physical-physical interaction enables multiple physical entities to communicate for complex tasks; virtual-virtual allows virtual models to form networks for information sharing; and virtual-physical enables synchronized optimization of virtual models with physical objects, and dynamic adjustment of physical objects based on virtual orders. Research in this area is currently limited.
• Service: Data-driven DTs enhance services like structure monitoring, lifetime forecasting, and in-time maintenance. Relevant theories include service encapsulation (uniform interface for functions), service matching and searching (choosing suitable services), Quality of Service (QoS) modeling and evaluation, service optimization and integration, and fault-tolerance management. This allows DTs to prescribe services like maintenance.

Industrial Applications of Digital Twins: DTs demonstrate superiority over traditional solutions in several industrial applications.

• DTs in the Product Lifecycle:
  • Product Design: DTs enable more responsive, efficient, and informed product design by allowing data feedback and reinforcing collaboration between design and manufacturing. They can manage 3D product configurations and evaluate product quality early.
  • Production: DTs are crucial for visualizing and updating real-time status, facilitating monitoring, timely adjustments, and process optimization in production. They integrate various data (environment, operational, process) to enable autonomous systems to respond to state changes. DTs also support the digitalization of production facilities, facilitate production optimization (e.g., through virtual engineering tools, real-time geometry assurance, reducing waste, and prolonging machine lifetime), and enable effective real-time production control.
  • Prognostics and Health Management (PHM): This is the most popular application area for DTs, with extensive initial use in aircraft for predicting structural life, monitoring operational states, and predicting tire wear. The application extends beyond aircraft to areas like additive manufacturing processes and cyber-physical systems.
    • Advantages of DT-driven PHM over traditional PHM include:
      1. Enhanced Model Fidelity: Integrates geometry, physics, behavior, and rule modeling for more accurate depiction of practical situations, achieving ultra-high fidelity.
      2. Holistic Data Integration: Merges physical, virtual, real-time, and historical data, aligning with the big data trend in smart manufacturing.
      3. Bidirectional Interaction: Connects physical and virtual spaces, allowing better control of physical entities and progressive optimization of virtual models.
      4. Rational Decision Making: Drives maintenance decisions with high-fidelity virtual models, complementing traditional optimization algorithms.
    • Despite these advantages, current PHM applications often focus on high-value equipment, and there’s potential for broader use in equipment maintenance and repair.
  • Other Areas: DTs are also applied in areas like streamlining development processes and detailed simulations through the Experimentable Digital Twin (EDT), and depicting cloud-based cyber-physical systems for recommendations.
• DT-Related Patents:
  • General Electric (GE): Holds patents concerning wind farms (monitoring, control, performance optimization) and a DT interface for managing multiple digital models. Also, for controlling cooling systems in power systems.
  • Siemens: Focuses on patents related to machine-human interfaces (human programming interface to generate human DTs for autonomous systems), methods for creating a DT of a room (useful for digital factories), energy-efficient asset maintenance, and collision detection in distributed autonomous production systems.
• DT Applications by Industry Leaders:
  • Siemens: Applied DTs for power system planning, operation, and maintenance in Finland, improving automation and decision-making. Also developed a DT for a wastewater treatment plant for real-time pipe monitoring and fault forecasting.
  • GE: Demonstrated how DTs revolutionize wind farm development, operation, and maintenance, increasing operational efficiency by 20%. Also applied DTs in locomotives (tracking lifecycle, optimizing operations based on real-time component conditions) and healthcare (streamlining hospital operations like bed planning and work allocation).
  • British Petroleum (BP): Utilized DTs to monitor and maintain oil/gas facilities in remote areas, improving reliability.
  • Airbus: Aims for factory digitalization using DT-based solutions, developing an assembly line DT to monitor production and optimize efficiency.
  • Systems, Applications & Products in Data Processing (SAP SE): Employs DT-driven PHM services to prevent failures in subsea equipment through cost-effective digital inspections and simulations.
  • International Business Machines Corporation (IBM): Applied DTs in automatic vehicles to analyze critical parameters like engine speed and oil pressure, preventing breakdowns and developing more efficient engines.

Observations and Recommendations: The paper makes several key observations and recommendations for future DT research and application.

• PHM is the Most Popular Application Area: The extensive application of DTs in PHM, particularly for aircraft components, highlights its advantages in model precision, data integration, interaction, and decision-making over traditional methods. However, current applications are often limited to high-value equipment, and broader applicability for maintenance and repair is needed.
• Modeling is the Core of DTs: While the importance of DT modeling is widely recognized, no consensus has been reached on a generic way to build a DT model. A unified DT modeling framework and more modeling tools are urgently needed.
• Cyber-Physical Fusion is a Key Difficulty: Challenges in implementation include achieving effective cyber-physical fusion, which involves data acquisition, transmission, mining, and collaborative control. Issues such as improving robustness and applicability of fusion algorithms, utilizing parallel computing for mass data processing, addressing security threats, and standardizing connection and communication protocols need to be resolved.
• Other Recommendations for Future Areas: The paper suggests exploring DT applications in dispatching optimization and operational control within workshops, as these areas are currently underexplored. DTs can enable more accurate planning, efficient dispatching through real-time monitoring and virtual model analysis, and more adaptable and robust control systems by integrating cyber-physical connections.

Despite rapid growth, DT remains an evolving concept with pressing issues like the need for a unified modeling method that must be addressed to enhance its practical viability.

Reference: Tao, F., Zhang, H., Liu, A., & Nee, A. Y. C. (2018). Digital Twin in Industry: State-of-the-Art. IEEE Transactions on Industrial Informatics, 14(4), 1722–1731. https://doi.org/10.1109/TII.2018.2873186