Entangled
When I hear the term Digital Twin my mind draws a parallel with quantum entanglement rather than 3D simulation models. Quantum entanglement is a perplexing phenomenon, that describes particles that are connected (“entangled”) in such a way that they are not independent of one another. According to the Stanford Encyclopedia of Philosophy, entangled particles are strongly correlated and unified. Measurements of one of the entangled particles automatically gives you the quantum state of the other particle. The core idea of creating a Digital Twin is to be able to create a digital replica of a physical object that can accurately reflect the current state of the physical object. Measurements taken of either the physical object or its digital counterpart produce the same accurately readings. In the world of Digital Twins, the physical asset and its digital counterpart are in a way “entangled”.
History
Houston, we’ve had a problem here
It was John Vickers from NASA who coined the term — Digital Twin way back in 2010. Gartner named it as one of the Top 10 Strategic Technology Trends for 2017. However, the core idea of remotely studying physical objects has been around for at least since the 1970s. When the oxygen tanks exploded early into the mission of Apollo 13 in 1970, technical issues needed to be resolved from 200,000 miles away. NASA had a physical-digital twin model of Apollo 13 on earth which allowed engineers to test possible solutions from ground level. That approach turned out to be the key to the rescue mission and one which brought back Apollo 13 safely home.
Why do we need Digital Twins?
Collecting, maintaining, aggregating, processing, and providing information about assets is the core idea behind a Digital Twin. The physical asset being studied is fitted with various sensors to measure vital areas of its functionality. The sensors collect data relating to aspects of the physical object’s working, such as temperature, vibrations, pressure, noise levels etc, this data is then relayed to a central system and applied to a digital copy of the physical object. The data is analysed to generate valuable insights, such as to detect performance issues or apply potential improvements or predict failure conditions. The outcome from the analysis is then applied back to the physical object, thereby improving the operation of the physical asset.
Anatomy of a Digital Twin
Another way of look at Digital Twins is to consider them as avatars for physical assets. The anatomy of a Digital Twin’s avatar is composed of the following parts.
Physical Asset
The physical asset that is being modelled. This can be anything from a microwave oven to the international space station.
Sensors
Devices that collect data of interest from the physical asset.
Actuators
Electronic devices that control the physical asset, such as a pressure release valve, a pitch controller on a wind turbine or a thermostat.
Edge Processing
Computing devices deployed to the edge that are typically used for data pre-processing or edge-based data inference using pre trained AI models. Such as turning down the temperature of an engine using an actuator when an AI model deployed to the edge detects an anomaly. Often telemetry data from sensors are mapped or used to calculate values derived from a mathematical combination of values from one or more telemetry data points or reference data. For example, calculating the speed of a car can be derived by combining the telemetry data coming from the rotating drive shaft sensor, magnetic sensors and wind speed at the car’s hood using a simple calculation to work how fast the car is travelling. At times this calculation can be done at the edge to remove even the smallest of latency.
Gateways (Edge/Cloud)
A central messaging hub that sits between the physical asset and the Digital Twin platform which is usually hosted on the cloud. The gateways can also translate protocol formats, and when deployed to the edge (Edge Gateways), can also act as multiplexers for multiple sensors. Digital Twin solutions do not always have edge gateways but will almost always have a cloud gateway.
Telemetry Data
In a Digital Twin solution data is generated by sensors associated with the physical asset. The data from the sensors is mapped to the defined information model and sent to an events processing system. The data points are usually mapped to specific properties defined in the information model and take the form of key value pairs. Data from sensors is the lifeblood of a Digital Twin.
Security
A Digital Twin enforces a security domain to protect and manage access to data associated with it. Digital Twins are guarded by several security controls associated with authentication, authorization, and communications. Usually, digital certificates, are used to identify and safely connect physical assets. Security is also applied to the bidirectional communication channel between the physical asset and the central Digital Twin platform.
Information model
At the heart of a Digital Twin is an information model. It is like a class in an object-oriented programming language. The model details the physical asset and includes semantic links/relations associated with the asset together with a define taxonomy and attribute set. The information model is a crucial aspect of a Digital Twin and is composed of labels, names, types, properties, and unite of measure etc.
Stream Processing
A mechanism to ingest telemetry data generated by a Digital Twin is a core part of the anatomy of a Digital Twin. The data can be ingested via a batch process or in real-time. In a batch processing model, a collection of data points from incoming telemetry data is grouped for a specific time interval and processed. In the real-time model, data is processed in near real-time. Digital Twins use both models. Data destined for batch processing flows down the cold path and data destined for real-time processing flows down the hot path. See this post for more details on architectural patterns for processing telemetry data via the hot and cold paths. Telemetry data flowing through the cold or hot path is also ultimately persisted in a physical data store, usually a big data data store.
Digital Twin Instance
When you entangle a physical asset with its matching digital model you get a Digital Twin instance. The instance is the binding of the physical asset with the Digital Twin using the information model. The Digital Twin instance stays linked (entangled) to the physical asset through its lifecycle and typically, contains data relating to in-use conditions as captured through the sensors, historical state, and predicted state etc. A Digital Twin instance also consumed data from other data sources within the enterprise systems via an enterprise gateway, using REST APIs or connectors.
Often incoming telemetry data from sensors are mapped or calculated into values that do not match 1:1 with data values from the sensors. Data can be derived from a mathematical combination of values from one or more telemetry properties and even reference data. For example, calculating the speed of a car can be derived by combining the telemetry data coming from the rotating drive shaft sensor, magnetic sensors and wind speed at the car’s hood using a simple calculation to work how fast the car is travelling. These mappings and transformations take place within the Digital Twin instance.
Insights
Rendering data associated with a Digital Twin is substantially different to simulation or static models. Simulations usually don’t interact with real-time data. Digital twins manage the bidirectional flow of information that first occurs at the sensors and then again when insights created by the Digital Twin are applied back to the physical asset. Insights derived via the Digital Twin can be highly complex and virtual models that are the exact counterpart (or twin) of a physical asset that needs to be rendered and presented to Stakeholders who then make actionable data-driven decisions. The rendering of the information can take several forms, such as 3D models to rich dash-boards. Stakeholders looking at the digital twin should be able to easily see crucial information about how the physical asset is doing in the real world.
Digital Twins also interact with AI models to infer and generate valuable insights. AI models can be trained and deployed o address various aspects of physical assets. The AI models will be trained on historic data composed of incidents of failure and telemetry data reading leading up to the failure. The model when fed with real-time data from the Digital Twin is then able to detect abnormalities. For example, an AI model can be built to monitor the temperature levels of the oil in transformers and recommend a change when abnormal temperature patterns are detected.
Day in the life of a Digital Twin
The interconnect physical asset and its virtual representations (the Digital Twin) constructs a digital thread as data flows between them. The digital thread typically encompasses the following interactions:
- Data from the physical object collected by sensors is sent to the centralized platform (usually the cloud) via the gateways. In some implementations this data may be used by the edge devices for processing, translation or even inference.
- Once incoming telemetry data passes the security domain it is ingested and processed into a defined information model and send it to the Digital Twin.
- The digital twin uses this data to mirror the physical asset in real-time.
- The Digital Twin applies analytics to generate valuable insights, such as to detect performance issues or apply potential improvements or predict failure conditions.
- Insights from analytics are visualized and presented via the dashboards, and 3D models.
- Stakeholders make actionable, data-driven decisions.
- The insights are applied back to the physical object, thereby improving the performance and operation of the physical asset.
Types
Component Twins
Component twins are Digital Twins of the smallest functioning component of a physical asset. Such as the bearings of an electromagnetic motor. Creating a Digital Twin for the bearings component to monitor its temperature, acoustic levels and vibrations can help decide which machines need attention. Maintenance teams can then be directed to fix issues before they occur and prevent machine breakdown. For example, bearings are a vital component of a wind turbine and prevention maintenance can reduce costly repairs that usually involve cranes, barges, and specialised teams to fix offshore wind turbines in the middle of the rough seas. Whilst the whole of the wind Turbine can be modelled into a Digital Twin, simply modelling one component using component twins can also be of significant value.
Asset Twins
An asset twin is the modelling of one components of a physical asset. Asset twins aid in the modelling and study of interaction between components.
System Twins
System twins are a combination of more than one asset twin.
Process twins
Process twins model how systems composed of several physical assets and components work together to create an entire production facility.
Industry Examples
GE Engines
GE created Digital Twins to forecast the degradation of the GE90 engines over time. The Digital Twin represents just one aspect of the GE90 engines – its composite fan blades. The blades are prone to spallation. The Digital Twins helps pinpoint the right time for maintenance before any issues arise.
Tesla Cars
Sensors in Tesla cars continually stream data from the car to a digital copy that lives in the cloud. AI algorithms analyse the data to identify abnormalities and any problems are fixed by sending over-the-air software updates back to the physical car. Using Digital Twins, Tesla is able to adapt its car’s configurations to different climate conditions, improves its performance, and provides remote diagnostics.
Bridgestone Tyres
Bridgestone uses Digital Twins to gain insights on how speed, road conditions, and driving style affect the performance and lifespan of their tyres. Bridgestone also uses Digital Twins to design and test new types of tires. Bridgestone note that this approach has cut development time by half.
The future of Digital Twins
During the IEC and ISO Joint Technical Committee (JTC 1) plenary in Lillehammer, Norway in 2016, a decision was made to establish a Joint Advisory Group (JAG) Group on Emerging Technology and Innovation (JETI). JETI assessed several emerging technologies factoring recent trend analysis from research firms such as Gartner, IDC and others and identified 15 top priority technologies. Digital Twins was listed as one of the top four in that list. That recognition is a good gauge of the importance the industry is giving to Digital Twin technology. Digital Twins are impacting the full spectrum of industries as discussed in the few examples above. According to a Gartner survey, 75% of organizations implementing IoT already use digital twins or plan to within a year. Gartner predicts that by 2022, over two-thirds of companies that have implemented IoT will have deployed at least one digital twin solution.
