These Designs Can Drive Digital Twinning
Image Source: Wright
Studio/shutterstock.com
By Paul Pickering for Mouser Electronics
Edited January 28, 2021 (Originally published May 31, 2018)
Spurred by the ubiquitous connectivity of the Internet of Things (IoT) and low-cost sensors, the digital twin (DT)
model is quickly becoming a part of manufacturing and other industries. But implementing a DT program imposes stiff
requirements at all levels of the signal chain, particularly at the edge node close to or on the machines being
twinned. Here, we’ll outline sensors and edge-node architecture, review edge node’s importance, and
discuss edge-node communications, all lending themselves to maximizing a DT’s full potential.
Sensor and Edge-Node Architecture
Digital twinning architecture closely resembles three-level IoT architecture (Figure 1).
- Sensors at the edge node gather real-time information about a functional unit (such as an industrial robot,
aircraft engine, or wind turbine) and transmit this information over a wired or wireless local area network (LAN).
- A gateway node communicates with multiple edge nodes (possibly using various protocols) and combines this
information into a wide area network (WAN).
- An enterprise node receives gateway data, applies it to the digital model, and communicates the results.
Figure 1: Digital twinning architecture resembles IoT architecture with
sensors at the edge node, gateway node, and enterprise node.
With a sufficiently accurate model and high-quality data, a DT can predict failures, boost efficiency, and even
make changes in real-world operations.
The Importance of the Edge Node
DTs require a constant stream of high-quality, real-world data to validate a virtual machine’s performance
against its physical counterpart. Otherwise, the real and virtual worlds will gradually diverge, and the
calculations and predictions of a DT will be of little value.
An edge node is fundamental to this data-gathering process because it contains sensors that gather real-world,
operational, and environmental data and includes communication links that send this information upstream. If a DT
can make changes to the physical process, an edge node has actuators that allow this process to happen.
Sensor measurements fall into two categories:
- Operational measurements (relating to the physical performance of a machine or device), such as
tensile strength, speed, flow, displacement, torque, operating temperature, or vibration
- Environmental or external data (affecting the operations of the physical process), such as
ambient temperature, barometric pressure, and humidity
Edge-node sensors can take many forms. Devices such as temperature sensors, pressure sensors, load cells, and
accelerometers measure real-world characteristics and generate numerical information. Sensor fusion systems combine
the results of multiple sensors to generate insights that are not possible from any single device. Cameras and
microphones create video and audio streams using complex, unstructured information that requires extensive
processing to interpret.
Legacy Machines Pose Challenges
Ideally, DT design begins with a digital design that serves as a model for the real-world installation. The sensors
providing the real-time data can be included in the model and carry over to the final version. This is certainly the
case in many high-technology applications in the oil and gas, nuclear energy, aerospace, and automotive industries.
But problems can arise if a machine design predates a virtual model’s implementation. Upgrading an edge node
to activate digital twinning brings a new set of challenges.
Designers in more traditional industries rarely have the opportunity to design a DT’s real-world counterpart
from the ground up. Instead, they must work with an existing factory infrastructure that might have been operating
quite well for years, if not decades. If so, the DT infrastructure must be grafted onto the existing system.
Although an underlying system can be a good candidate for digital twinning, the integration process becomes
exponentially more complex if an existing machine has few or no sensors monitoring its performance. In this case,
dozens to hundreds of sensors must be grafted onto a machine that was never designed to accommodate such progressive
technology.
Even if an original machine has sensors already in place, the accuracy of the sensors might not be sufficient to
provide useful data to a digital model. An installed temperature sensor, for example, can function only to detect an
over-temperature fault but not to provide the quality of data necessary to identify a pattern of temperature
overstress, which might help predict an early failure.
The capacity of an installed communications network is another potential issue. A traditional IoT installation uses
many different wired and wireless standards to connect edge nodes to their respective gateways. These include such
industry standards as:
- Zigbee®—for low-power mesh applications
- Sub-1GHz—for low power and long range
- Wi-Fi—for a direct internet connection with high speed
- Bluetooth®—for the lowest power
Each standard must undergo a careful evaluation to decipher whether it can handle burden increases by digital
twinning data.
The Common Thread in Existing DT Applications
Although digital twinning is in its infancy in many industries, many technology products undergo design, testing,
and validation in the virtual world long before the first prototype ever sees daylight. These products also tend to
gather large numbers of specialized, data real-time sensors. Aircraft engines and Formula 1® racing
cars are two good examples:
Aircraft Engines
Aircraft engines are already highly instrumented. A traditional turbofan engine (Figure 2)
contains sensors to measure pressure, temperature, flow, vibration, and speed. Multiple specialized sensors exist
within each category: For example, for pressure, there are turbine pressure, oil pressure, oil- or fuel-filter
differential pressure, stall detect pressure, engine control pressure, bearing compartment pressure, and more
sensors types.
Figure 2: An airplane turbofan contains hundreds of sensors, so adding a DT
requires an order-of-magnitude increase. (Source: patruflo/Shutterstock.com)
A DT requires much more data than a traditional monitoring application, so its sensor design must accommodate the
increased requirements. Although most airplane engines in service today contain fewer than 250 sensors,
manufacturers demonstrate next-generation, DT-compatible products containing more than 5,000 sensors. Additional
data comes from sensors monitoring fuel flow, fuel and oil pressure, altitude, airspeed, electrical load, and
outside air temperature. Rolls-Royce, GE, and Pratt & Whitney are already using DTs to increase reliability,
boost efficiency, and reduce maintenance costs.
Formula 1 Racing
DT technology helps improve driver and car performance in the high-pressure world of Formula 1. The McLaren-Honda
team (Figure 3), for example, uses over 200 sensors to transmit real-time data relative to the
engine, gearbox, brakes, tires, suspension, and aerodynamics. During a race, the sensors transmit up to 100GB of
data to the McLaren Technology Centre in Woking, U.K., where analysts study the data and use the DT to relay optimal
race strategies back to the driver. The DT virtually drives in the same race as the physical car, even adjusting to
the same road conditions, weather, and temperature.
Figure 3: The McLaren-Honda F1 Team uses over 200 sensors to transmit
real-time data relative to the engine, gearbox, brakes, tires, suspension, and aerodynamics. (Source:
ZRyzner/Shutterstock.com)
The Future of DT Edge-Node Architectures
If a DT model is to undergo full implementation, it must solve several issues with the existing edge-node
architecture:
Smart Sensors and Edge-Node Processing
As sensors gather ever-increasing amounts of data, it’s important to have a clear understanding of how to use
the data in a digital model and where to process the data, whether at the node, at the gateway, or in the cloud.
Processing at the node reduces network bandwidth but risks losing information to reduce DT performance.
The type of sensor has a bearing on the decision. Although many sensors transmit information in a structured format
that’s easy to use (such as the digital transmission that denotes pressure, for example), those such as
microphones and image sensors produce massive volumes of raw data that’s unstructured and useless without
extensive processing, no matter where it’s performance occurs.
Enhanced Communication Interfaces
Despite increasing edge-node processing, vastly increased data flow will require system designers to perform
network bandwidth increases at all levels. For example, aircraft engines generate up to 5GB per engine per second
and up to 844TB per day for a twin-engine airplane in commercial service.
Traditional industries produce large volumes of data, with an additional complication. Many remotely located edge
nodes in traditional, Industrial IoT (IIoT) applications use battery-power and low-performance wireless protocols
for low-power consumption optimization. The existing design tradeoffs might need a re-evaluation to detect
communication bottlenecks.
Robust Edge-Node Security
Existing IoT installations have created new security issues in edge-node devices, and security measures such as
encryption, secure hardware designs, application keys, and device certificates are becoming more common. Increases
in DT adoption will raise the importance of these technologies, especially in nodes that add Internet Protocol (IP)
connectivity—which opens an entry point for would-be hackers.
Conclusion
Implementing a DT program imposes stiff requirements at all signal-chain levels, particularly at the edge node
located close to or on the machines being twinned. The edge node is fundamental to digital twinning because it
contains the sensors that gather real-world, operational, and environmental data and contains the communication
links that send this information upstream. Digital twinning is currently being used in aerospace and automotive
industries, where many kinds of specialized sensors are essential for digital twinning are available. In
retrofitting equipment for digital twinning, dozens to hundreds of sensors must be grafted onto a machine that was
never designed to accommodate them. For digital twinning to continue gaining traction, several solutions are vital
and must live within sensors as well as in edge-node processing, communication processing, and edge-node security.
Author Bio
As a freelance technical writer, Paul
Pickering has written on a wide range of topics including: semiconductor components & technology, passives,
packaging, power electronic systems, automotive electronics, IoT, embedded software, EMC, and alternative energy.
Paul has over 35 years of engineering and marketing experience in the electronics industry, including time spent in
automotive electronics, precision analog, power semiconductors, embedded systems, logic devices, flight simulation
and robotics. He has hands-on experience in both digital and analog circuit design, embedded software, and Web
technologies. Originally from the North-East of England, he has lived and worked in Europe, the US, and Japan. He
has a B.Sc. (Hons) in Physics & Electronics from Royal Holloway College, University of London, and has done graduate
work at Tulsa University.
