Using a digital twin to drive operational decisions when it comes to maintenance is about turning what could be a cost into an asset.
By 2025, the world will be creating 175 zettabytes annually, according to market research firm IDC’s Data Age 2025 report. To put that in context, one zettabyte is equivalent to one trillion gigabytes. How rapidly this data is growing can be demonstrated by the fact that in 2012, only one zettabyte of data existed.
But, with all this data being produced, how much of it is actually useful? While a rail organisation is only a small proportion of the global data total, according to Andrew Smith, solutions executive responsible for Bentley’s Rail and Transit solution, they are still producing a significant amount of data.
“Rail organisations typically are very data rich,” said Smith. “They’ve got a large number of asset disciplines because it’s a huge complex system and each of those asset disciplines has a number of inspection and measurement mechanisms that can produce data.”
This data on its own, however, is not yet a useful resource.
“Data is a discrete fact about something,” said Smith. “For example, the distance between the left and right rail at this location is X, but data is no use to you when you’re actually trying to either work out short term what you’re going to do or longer term what may happen in the future. What you need to do is start a transformation process, so the first step of that is to go from data to information, which is data in context with meaning attached.”
Giving data its context turns what can be seen as a cost, the accumulation and storage of data, into a resource, information that can be used to make a decision.
“In order to be able to do that, you need to have a framework in place that allows you to pull all the different classes of data together, such that you can see all of that data in context,” said Smith. “And to me, that’s at the heart of the digital twin.”
Digital twins are a replica or model of a system or asset that can be used to take the information that a rail organisation has, in the form of data, to create insights, that are conclusions drawn from data and information.
“When you bring all this information together, the digital twin can tell you how as well and why things are happening, and it can give you contextual history,” said Smith. “The digital twin can give you design intent information that you wouldn’t necessarily have otherwise, as well as the as-constructed record. Critically, a railway is a system, it’s not just a set of isolated components, and what a digital twin allows us to do is understand specifically the relationships between those components and how they can be affecting each other.”
While digital twins are widely used in many fields, including construction and manufacturing, they have a distinct role to play when it comes to the maintenance and management of rail assets. As the complexity of operating a railway requires various departments covering different skills and mandates, applying a digital twin can overcome the data and organisational silos. Smith, who has been working in the rail industry for over 20 years, highlights one way in which this can be applied.
“For anywhere that’s got overhead electrification for example, if you’re on ballasted track you can move the track from side to side through maintenance, but you need to maintain the relationship with the overhead wires, but these are often managed by two different teams. The digital twin will manage by design the relationship between the two. The maintenance records, where you’re going to go, and the type of maintenance you’re doing means that there is a chance that you will actually introduce a change to the overhead wire relationship. Therefore, you need to tag that work order as needing somebody to go out and actually measure the overhead wire relationship as well, whereas historically that relationship wouldn’t be as tightly coupled.”
DESIGNING A RAIL-BASED DIGITAL TWIN
Getting to this level of maturity with a digital twin takes a deep understanding of how a rail network operates and how best to design a digital twin that fits the reality of a rail organisation. Bentley, as part of its portfolio of solutions in the rail and transit space, has experience working with rail operators around the globe to design and deploy digital twins. From this experience, Smith highlights, the usual understanding of what a digital twin is can be re-evaluated.
“Normally if you think about a digital twin you actually start with a four-dimensional model, however railways often don’t think in terms of XYZ axes. They tend to think in terms of linear distances with lateral and vertical offsets and that drives the way that measurements are made, the way that inspections are made, but also the way that maintenance is actually managed. If you’re sending someone to go out and do some tamping along a piece of track, you don’t send them to an XYZ coordinate or a latitude- longitude coordinate, you’ll send them this many metres past kilometre post seven on such and such a track.”
With this in mind, Smith suggests that digital twins in the rail space can be more useful if they are designed to fit the way that railways are understood. Then, the data that makes up the digital twin can be overlaid on the representation of the network. When needed, for example at a station or in yards, this data can be visualised as a three-dimensional model, but linear visualisations may be more appropriate for a section of track.
To get to the point of having a representation of a rail network, a large amount of data will have to be collected and interpreted. As managers of an array of legacy assets, rail organisations can turn to the use of artificial intelligence (AI) to sort and organise the vast streams of data, said Smith.
“One of the challenges that we see with a digital twin for a lot of brownfield sites in particular is that there are a large number of assets in place that are not being represented digitally. Being able to use image recognition or identifying features from reality meshes and then being able to put an attribution against them is a great use of AI to be able to identify where the assets are.”
With this data in place, the twin must be maintained and kept up to date. With networks spanning across hundreds of kilometres, rail organisations can use automated surveys of a network to provide the constant data upkeep needed.
With the digital twin now operating as a living representation of a rail network, defect detection can be done in a way that gets to a root cause, rather than just addressing individual issues. One example, that Smith describes is if measurement scans identify vertical deterioration. A digital twin would then allow for a cross referencing against other assets that are in place, to see if there is a culvert on that section of track.
“Then I’m not going to send a tamper out,” said Smith. “The first thing that I’m going to do is send a crew out to inspect a culvert to see if it’s collapsing over time. The next thing I might want to do there is ask, if I’ve got twin track, am I seeing the same deterioration on both tracks? Normally they’d be considered in isolation, separate from each other. Then I would ask, has any maintenance taken place at this region? That’s not just maintenance of this asset, but all maintenance records, so I could say, ‘Hang on, someone actually went in there and did some maintenance work on the drainage in-between, but it happens to be in an area that’s close enough that it could’ve had an unexpected knock on onto the condition of the track.’”
These kinds of insights can only be gained through the kinds of insights a digital twin is able to offer, by bringing together disparate data and putting that data into context.
DRIVING THE SOLUTION
While a digital twin may seem like a laudable goal on its own, according to Smith, the implementation of such a tool only makes sense when a rail organisation has identified what are the issues that it needs to solve.
“The driver here is not a technology change. The driver is to change the way of working, so an organisation has to first understand its current working practices, where the efficiencies and inefficiencies are, where the limitations and constraints may be, and then we can understand the aspirational state, where they actually want to be at some stage in the future.”
Implementing a digital twin begins with understanding the process of going from a current state to an aspirational state in the future. Rather than jumping in straight to a predictive maintenance solution, the first step may be to identify where the current most significant issue is, with a plan or vision to have a predictive system at a point in the future. Understanding where the technology is going to be implemented comes down to working with the people who are going to be using the software.
“It is absolutely critical that those people are engaged right from the outset, not just the management but the end users,” said Smith.
To get people on board, Bentley has used model offices where representative users are invited to be involved in the design process and give their insights into the particular challenges they face.
“Then there’s buy in,” said Smith. “There’s engagement at that side, which means that the final product is a tool that the engineers have designed and set up to help them do their job better that means they’re positive about the tool and they’re positive about the process change that’s in place to be able to do it.”
Rather than success looking like a piece of software that is installed to contract specifications, Smith outlines how in developing a success plan for the implementation of the software, the outcome is about delivering value.
“Owner operators of railways aren’t installing these systems because they like technology. Technology is an overhead to them – it’s a cost, an expense, and it’s a risk, so the only time that it’s worth doing is when they can show that the value is greater than the cost associated with it, so what we’re moving to is making sure that the focus is now on the value to the users instead,” said Smith.
“You can look into the future and run ‘what if’ scenarios. So, I’m going to increase the tonnage over a particular length of rail and I’m going to run a simulation of what that’s going to do to my rail replacement strategy that I have in place. We can use AI on top of this to look both tactically how do I optimise right now, where do I best spend money, but also starting to look further out by running simulations and trying to predict what the impact the change is going to have.”
This value can be defined in any number of ways, but as Smith highlights, it is the process of creating insights out of data.