Digital Twins and the Railway: One Framework Many Implementations
The components of a Digital Twin
Essentially, a digital twin is a dynamic digital profile of an object, system or process. Figure 1 is a diagram of a basic digital twin in context. It shows how sensors located on physical assets in the real-world measure behaviour, and generated data are linked via communications infrastructure to the twin’s software. The software manipulates the data it receives into an up-to-date virtual model, i.e. a real-time representation of the monitored object (or process). A record of historical behaviour is formed by historical status profile and the designed capability, this is sometimes termed a digital through-life thread. Additionally, there may be a feedback link from the digital twin back to its physical counterpart via the control actions that can be implemented, as either a manual or an automated process.
Figure 1 A digital twin system, showing the connections between the twin and the real world that it represents
A digital twin becomes powerful and valuable when it is used to make high-quality decisions resulting from the transformation of data into actionable insight. This is also known as feedback from the twin and it is made possible by using analysis and/or simulation models and techniques, which require human expertise to set up and monitor.
Where have twins been deployed before?
NASA made the first use of a digital twin, in 2002, by remotely monitoring and controlling the status of its spacecraft. However, in the years immediately after, no further use of twins was put to action. More recently there has been an increase in this area with factories using twinning processes to streamline their operations in semiconductor, automotive and aerospace production.
Harley-Davidson’s ‘smart’ factory—a highly digitized and connected factory that allows processes to be monitored in a manner that anticipates problems—reduced operating costs by $200m and production cycles from 21 days to 6 hours. GE Transportation have discussed embedding sensors into locomotives to create a valuable through life thread for this significant asset. Siemens and IBM have applied a digital twin throughout the lifecycle of the Finnish power grid by integrating eight software products to act as a single source of truth. In these examples digital twin is the platform for data analytics and artificial intelligence to access real-time data and translate it into data-led decisions. The decisions might be implemented automatically using advanced processes like additive manufacturing.
What will twins mean for rail?
The use of digital twins in rail assets and systems could make the railways more reliable, competitive and efficient. They can enable the railways to deliver a high-quality service that meets the demand of users. The railway, in its huge size and complexity differs in scope considerably from the closed controllable environment of a factory, and it would be impossible to fully twin the whole system. Most likely, there will be many digital twins for the railway, modelling different systems and processes. The challenge and opportunity is to understand how such developments led by different players and with different motivations, can be compatible and interoperable.
Figure 2 Schematic of proposed digital twin layers
Researchers at Birmingham Centre for Railway Research and Education (BCRRE) propose that fully mature digital twins for the railway can be usefully described in five layers, see Figure 2 and Table 1. The first three layers take care of the representation of the railway system or process of interest. The existence twin describes static and strategic elements of the railway e.g. infrastructure, staff, and processes. It answers the question ‘What is there?’ The status twin describes an up-to-date state for railway system elements changing in days, weeks or a few months. The operational twin adds event-based real-time data about fast-changing elements. Together the status and operational twin answer the question ‘What is the railway doing?’ The distinction between them is the update rate, set to represent ‘real-time’ status by frequently inputting data to capture relevant changes in the value of the characteristics of the real world. The close-to-real-time updates provided by internet of things (IoT) enabled sensors are needed only for dynamically changing factors. Periodically updated data on more slowly changing statuses, for example drone scans of substructure condition, and digitised survey data giving the location of static assets are all relevant data inputs types for digital twins of the railway.
In order to assist decision making, digital twins in the railways need to provide a platform for user friendly, semi- or fully-automated decision support and implementation processes. Simulation twins are invaluable and act as a test bed for what if questions, for example, ‘What might the railway do if X changes?’ Finally, the cognitive twin allows automated data-driven analyses. These can be used to derive insights about processes or make predictions about future performance. It aims to answer the question ‘What is the railway going to do?’ or, better, ‘How do we derive best value from the railway?’
Table 1 Proposed digital twin layers
Beyond the technological and practical aspects—for example fitting sensors to legacy equipment and structures—the required cultural, organisational and human adaptations to move towards effective digital twin implementation pose a formidable challenge for the railway industry. Persuasion may come with the design of valuable and exciting railway digital twin use case studies alongside an associated common framework that clearly supports expansion and integration of early developments.
Rules and policy for the twin must be set out in advance, and include aspects such as data quality, scope and completeness of representation and model compatibility to answer the question ‘What must a railway digital twin do?’ Furthermore, key operational aspects such as response to threats e.g. cybersecurity strategy, use policy etc should also be covered.
Now is the time to ask, does the rail industry already (at least partially) have existing digital twin(s); if so, where? Determining how rail can make greater use of digital twins, will provide the opportunity to take a proactive approach to harnessing this technology, and understand whether accelerating to full digital twins of the railway is desirable and how it can be achieved.
Blog written by Gemma Nicholson, Research Fellow, Electronic, Electrical and Systems Engineering, Birmingham Centre for Railway Research and Education (BCRRE).
