Digital Twin Terms and Definitions
This article explains the concept of a Digital Twin, a digital model driven by sensor data to provide future or deeper insight into an existing physical product’s performance. It is characterized by the following traits:
- Driven by Sensor Data: Sensor data is captured from an existing physical product by an Internet of Things (IoT) system.
- Digital Model: Sensor data is fed into a digital model of the product. This model can be a numerical, 1D simulation, or 3D simulation. It may focus on a single engineering domain or span many. This model emulates or simulates the behavior of the existing physical product.
- Predictive or Deeper Insights: The model mimics the performance of the existing physical product, offering insights into its future operation, or deeper insights into its current or past operation.
- Real Time or Offline: Digital Twins can run in real time, meaning they mimic the behavior of existing physical products almost simultaneously. Digital Twins, especially ones that are more computationally intensive and thus cannot keep up in real time, can also be run after the sensor data has been captured.
Capturing Sensor Data for Digital Twins
The data captured from sensors on the product are no different from that of most IoT efforts. Note, however, that this sensor data will act as an input for the digital model. Organizations should plan both the inclusion of the right sensors and correct placement if they are to enable the right type of Digital Twin for their application.
Digital Models for Digital Twins
A number of different kinds of digital models can be used to power Digital Twin efforts, including:
- Numerical Models: These models use algorithms such as machine learning and artificial intelligence. These applications or agents either extrapolate that data and/or correlate data to existing events. Both are an effort to predict future behavior.
- 1D Simulation: These models are a combination of flow diagrams with equations or formulas behind the blocks that simulate the performance of embedded software or multi-disciplinary engineering systems. These models can provide deeper insights into ongoing operation.
- 3D Simulation: These models, often in the form of multi-body dynamics, are commonly used to predict the dynamics and structural performance of products. These models can provide deeper insights into ongoing operation.
To run in real-time, these models must run on the same compute platform as the IoT system, as they will be fed sensor data. In offline scenarios, sensor data can be downloaded in tabulated formats and imported into these models.
Reaping Insights from Digital Twins
The key value of Digital Twins is to (1) predict future performance or (2) gain deeper insight into current or past performance.
Predicting future performance relies on Machine Learning and Artificial Intelligence to correlate events with sensor readings from past operation. Once the right indicators are found, future events can be predicted and avoided.
Gaining deeper insight into current or past performance relies on Virtual Sensors, which digitally track measures in the digital model. Tracking performance through Virtual Sensors delivers a richer set of information than the data gathered from sensors on the physical product. This information can be analyzed independently or combined with sensor data in a numerical model, bringing Machine Learning and Artificial Intelligence into play.
Applications for Digital Twins
- Engineering: During the design phase, engineers can replace the assumptions behind their requirements with real-world data, whether that is directly from sensors on the product or from virtual sensors on a Digital Twin. This information provides more insight into product performance and use, removing ambiguity from design.
- Service and Maintenance: Uptime is a critical characteristic for product-as-a-service strategies. Digital Twins can provide more insight into ongoing and future performance of physical assets, allowing organizations to plan maintenance in order to avoid downtime.