What are digital twins?
IBM technologists provide a definition of digital twins that is supported by a reference architecture
What are digital twins? After a great deal of effort, workshops, conference calls, client validation, and industry research, the the IBM UK Technical Consultancy Group (TCG), an affiliate of IBM’s Academy of Technology (AoT), agreed upon the following definition of digital twins:
“A digital twin is a dynamic virtual representation of a physical object or system, usually across multiple stages of its lifecycle. It uses real-world data, simulation or machine learning models, combined with data analysis, to enable understanding, learning, and reasoning. Digital twins can be used to answer what-if questions and should be able to present the insights in an intuitive way.”
Digital twins have one fundamental purpose: To model the behavior of real world systems to enable people to make better business decisions that impact the real world. This can be directly, through decision support; for example, a digital twin can simulate a large number of scenarios that are possible in a Formula 1 race to determine whether a driver should be called in for a pit stop when the pace car comes onto the track. Or, this can be done more indirectly by using a digital twin to hone the skills of operators in an Electricity Systems Operator control room who are dealing with a sudden decrease in electricity generation.
Using digital twins to solve real world problems
Digital twins can be used to address various problems that organizations face, including but not limited to:
Capturing requirements. Digital twins can ensure that the requirements that are captured during the earliest stages of a product lifecycle are maintained, verified, and validated as the product evolves, is built, enters service, and is ultimately retired, decommissioned, and recycled.
Designing products. One of the benefits of digital design over the last 30+ years is the ability to evaluate alternatives during the ideation phases and then quickly discard concepts that don’t meet the original intent. In addition, a design digital twin can be used to simulate and test the design before any manufacturing work takes place. Also, 3D models can be visualized in context to produce a configurable digital mock-up of the final product for early user acceptance testing, which has been a commonplace practice in automotive and aerospace industries for several decades.
Project planning: A project planning digital twin can be used to compare different lifecycle plans based on impact from other digital twins as they evolve to assist with contingency and resilience management to ensure the plan is achievable.
Reliability engineering. The ability to digitally reflect sensor information from a real-world instance of an asset has become increasingly feasible over the last few years with the evolution of Industrial IoT solutions; again, this is not a new concept but one that has improved in scalability, security, cost, and resilience in recent years. Being able to monitor asset performance in as near to real time as is needed means that reliability engineers are able to make better decisions about asset maintenance and replacement thus improving overall asset performance, increasing system efficiency, and optimizing asset behavior, all of which allows the reliability engineer to predict and manage risk based on high quality data rather than assumptions based solely on experience.
Training. As assets become ever more complex and the experienced knowledge workers get closer to retirement, the usefulness of a digital twin as an aid to training is gaining significant momentum. No longer are long apprenticeships or mentoring needed when all the information is available to a new user with a digital twin; certainly assistance will often be needed by new team members but the digital twin has been successfully proven in many instances to allow teams to fix things right the first time.
Real-time decision making. Digital twins allow decision makers to rapidly understand the implications of any changes that are made to an asset at any point in its lifecycle. For example, if a material change is made, what will the impact be on the project plan model, the design model for mass and centre of gravity, the cost model for overall financial impact, and so on. The digital twin thus enables an organization to execute simulations to answer “what if” questions, sometimes repeatedly, with adjustments to the parameters, rather than going through the process of creating physical prototypes.
Decommissioning resources. With only finite levels of certain resources available globally, there has been a significant focus in recent years on how assets are recycled, decommissioned, or scrapped to encourage a circular economy. For example, with steel being a finite resource, there is great focus by major steel producers to understand where their products are being used, how long they will be used for, how they will be maintained, and what condition they will be in at the end of their first life, to ensure that they can be reused (possibly at a lower grade) for future products. In addition, there are global initiatives to monitor plastics and other hazardous materials to ensure safe usage and disposal, which means that the digital twin can be used to improve reporting and regulatory compliance.
The digital twin trade-off
As mentioned previously, digital twins allow us to carry out simulations at multiple points in the product lifecycle to improve decision making, but there is a trade-off. Simulations are, by necessity, bounded and an approximation, so we need to understand the business value and impact of creating digital twins before we invest a great deal of time and money. We must first answer the simple question of “why are we doing this?” Sometimes the answer is obvious: to reduce project costs, get a product to market faster to achieve regulatory compliance. But, in other instances, the value is not so clear. In all cases, the three considerations before undertaking a digital twin project should be:
- Complexity: How expensive (time and cost) will it be to create a digital twin?
- Breadth: How generic or how specific will the digital twin be?
- Depth: How detailed and accurate will the results from the digital twin be?
In reality, due to the wide variety of problems that digital twins can address and due to the trade-offs that are inherent in each problem, it is likely that you will end up with multiple, federated digital twins, addressing different needs such as representing various phases in the product lifecycle or answering different “what-if” questions. And, areas of your business, and possibly other organizations in your ecosystem, will need to share data, and that data will potentially need to be integrated in real time to ensure changes to any one digital twin are correctly represented in another.
Digital twin reference architecture
Digital twins can be highly complex but the following figure provides a simplified representation of how information from the real world is consumed and analyzed securely within the digital twin. Data can be visualized using the right tool, at the right time, by the right person to provide accurate and timely information to drive effective decision making.
High-level component view of a digital twin
As the previous diagram shows, data underpins digital twins; the representation of the real world (model data) is key to modeling the real-world. The sensor data provides the current state of the real world, and the modeling output is generated by the simulations.
Consider these two key challenges about the data and simulations:
Managing this data but also ensuring that a set of output data can be traced back to both the system configuration at the time and back to the initial conditions are both important and quite difficult.
Simulations can generate huge amounts of data, potentially much more data than the real world can generate (via IoT devices). Dealing with such volumes of data is crucial. Aston Martin Red Bull is using IBM Spectrum and Flash Storage solutions to deal with such volumes of data.
This component view is further supported by a high-level and a detailed reference architecture, which contains:
- Seven layers of information management and manipulation
- Three columns which ensure the digital twin is secured, appropriately coupled, and governed to ensure accuracy and quality of data.
It is vital to note that the digital twin does not stand alone; it must be integrated with the overall enterprise architecture. As a matter of fact, some elements that are used in the digital twin are likely to already exist within the organization and they can be extended or re-purposed to support the digital twin models.
Simplified digital twin reference architecture
As can be seen in the following detailed reference architecture, digital twins are not standalone applications. Digital twins integrate into the organization’s existing enterprise application suite to support the intended business outcomes.
Detailed digital twin reference architecture
The digital twins that we have encountered are not simple products that can be bought over the counter, but instead they are the result of a significant systems integration (SI) effort. It’s not likely that this will change in the near future.
IBM and digital Twins
IBM has been involved with Digital Twins since the Apollo space program. IBM’s Real-Time Computer Complex (RTCC) was an IBM computing and data processing system at NASA’s Manned Spacecraft Center in Houston. It collected, processed, and sent to Mission Control information that directed every phase of an Apollo mission. The RTCC was so fast, there was virtually no time between receiving and solving a computing problem.
Today, an example of a very complex digital twin is The Weather Company. It is fed by a large number of sensors, both commercially operated weather stations and the Weather Underground, which is a network of amateur meteorology stations connected to the Internet and statistically filtered to weed out inaccurate readings. The Weather Company models a hugely complex physical system, the Earth’s Atmosphere, and then analyzes the output and presents it in an intuitive way in the forms of weather maps and forecasts that enable users to make better decisions.
IBM is uniquely positioned to bring to market all the critical parts of the IT landscape to deliver a successful digital twin project because:
- IBM has a large portfolio of software, services, and hardware solutions. For example, IBM Maximo Asset Health Insights can transform managed assets into digital twins by attaching sensor data to them.
- IBM has deep industry and technology knowledge across all industries.
- IBM can support all stages in the product lifecycle from inception to recycling or disposal.
IBM delivers projects from three perspectives, which are controlled by an overarching layer of change management:
People: IBM ensures that all users in the value chain are educated and trained on the use of digital twins to improve cross-organizational collaboration and worker efficiency.
Process: IBM delivers leaner processes with automation where needed to reduce time to market and improve quality.
Technology: IBM authors, delivers, and supports a vast portfolio of class-leading products that are used in the creation of digital twins.