Transport Committee
Oral evidence: Future of transport data, HC 84
Wednesday 21 February 2024
Ordered by the House of Commons to be published on 21 February 2024.
Members present: Iain Stewart (Chair); Jack Brereton; Paul Howell; Karl McCartney; Gavin Newlands.
Questions 68–91
Witnesses
I: Milda Manomaityte, Innovation Director, Railway Industry Association; Martin Frobisher, Group Engineering and Safety Director, Network Rail; and Paul Watson, Senior Director, Arcadis.
Written evidence from witnesses:
– Railway Industry Association
Witnesses: Milda Manomaityte, Martin Frobisher and Paul Watson.
Chair: Welcome to this session of the Transport Select Committee, where we are continuing our data in transport inquiry. Today we are focusing on infrastructure potentials and issues. Before we get going, I invite each of you to state your name and organisation, please.
Martin Frobisher: I am Martin Frobisher. I am the group safety and engineering director for Network Rail.
Milda Manomaityte: Good morning. I am Milda Manomaityte. I am the innovation director at the Railway Industry Association.
Paul Watson: I am Paul Watson, a senior director at Arcadis.
Q68 Chair: We are very grateful for your time and evidence today. I will start with a high-level question by inviting each of you to set out how you see the potential of data to improve the monitoring, delivery and maintenance of rail infrastructure assets.
Martin Frobisher: For us, it is absolutely key. What we have done over the course of the last 25 years is to steadily build up a database of all our assets and the scheduled tasks for maintenance against those so that we consistently do maintenance in an organised way. We gather lots of data using telemetry from the network. We monitor that in real time. All of that process and data allows us to prevent failures and provide better service for passengers.
Milda Manomaityte: Over the past years our members have improved the ways we use the data that the industry has been collecting, either manually or in a very technical sort of way. We have improved the way we understand the data that we already have, and to make it more accessible for a wider group of people. We get better insights from it to make better decisions.
There has been a significant technology improvement in the way that we collect data. That involves sensors that can see and feel vibrations and feel the heat. They can see through the cameras. We even use robotic solutions to gather data. There are a lot of interesting opportunities for how we use the data we have gathered, and for how we collect it.
Paul Watson: For me, data is central to driving efficiencies in infrastructure and driving whole-life costs, as well as improving performance in transport infrastructure for the end users. We deal with all-age stages of the asset life cycle. By understanding how assets are performing in the operational phase, by the use of sensors and things like those Milda mentioned, we are able to make better decisions at the design stage, which may lead to reducing embedded carbon or designing assets in ways that are better suited to the end user.
Going forward, it is about bringing multiple datasets together from different elements of the transport network and different parts of the sector. It is about driving the inside, and ultimately the key is making better decisions. It is the same in the maintenance phase: if we have more data around how assets are performing, we can come up with a better maintenance strategy. It comes down to the key of making better decisions in the transport sector.
Q69 Chair: To help to illustrate that, could you give us some examples of a project or an intervention that, if it had been done the old way—if I can classify it like that—would have been much more expensive or taken far longer? How has the use of data transformed how you have delivered it?
Martin Frobisher: There are lots of examples that we could give. One of the best is our plain line pattern recognition technology on our track-recording vehicles. We have trains that monitor the network. They have high-definition cameras that take many photos per second. The image is then processed using artificial intelligence, which identifies any defects. That is so much more efficient than somebody walking the track to identify issues. It is more efficient and more repeatable, and provides a better way of monitoring the network.
Milda Manomaityte: I have a good example. One of our members, Siemens, has developed a solution to take the data that railway interlockings collect. That is the heartbeat of our railways. It collects all sorts of information. As regards the way it was shared or gathered, it was sent to a very specific location—technical spaces where people with specific technical knowledge could understand it. It is complex data.
They have developed a solution where you can take all that information, apply it through algorithms and present it in an application that is very easily understandable. When there is a fault, it explains what the fault is. It has preloaded solutions on how to fix any faults. If you find a different way, you can input that information as well. Not only that, but it connects to Google Maps so that it knows where the interlocking is located and where the fault occurred. Google Maps knows the nearest access to that interlocking, so with this application you can look and say, “Okay, I can plan my route to go and fix it in the best way. I know where it is, what the problem is, and roughly how I can fix the problem.”
Paul Watson: We have a good example where we partnered with a technology company out of Canada called Niricson. We have a solution called Bridge Health, which works with sensor equipment mounted on drones and a technology platform that uses AI and analytics. The benefits are in accessing structures that are difficult to access and where often you need to stop transport networks running. You can access it in more difficult situations. You can access more regularly, so that you can paint a better picture throughout the lifespan of an asset as to how it is performing. The data is collected, brought to a digital twin environment and analysed. Maintenance regimes can then be built around that. It is about better access and getting the data to feed downstream activities.
Q70 Karl McCartney: I have a similar question to what the Chair just asked, but I am going to be more specific. I will come to you, Martin, but maybe Milda and Paul will add something.
You talk about a 25-year timescale for the data that you are looking at collecting. Looking at the east coast main line, part of the problem for passengers and lots of my constituents is obviously when the lines get caught and come down. Is the data you are collecting allowing you to predict where some of the issues might happen because of the maintenance regime? Is the maintenance regime therefore being changed? Are you seeing improvements through that? I am just surmising, but looking to you to give me the answers.
Martin Frobisher: The answer is yes to all of those points. We gather lots of data on the overhead lines. One source of that data is thermal imaging cameras that can identify hotspots and defects. We use cameras to identify, with intelligence to interpret the images. That is one way we prevent failures. We have another optical system that looks at where the balance weights are and the adjustment of the lines for tension. We use technology to observe that.
For our basic maintenance tasks, every asset is logged in our asset database. Every asset has a scheduled task assigned to it. We then make sure that we do the maintenance that is required at the required frequency. Our engineers constantly review the data to make sure that the frequencies we are setting are right.
Q71 Karl McCartney: You talked about hotspots. Are there specific areas or locations where the lines come down and get caught by trains passing, or can it happen anywhere?
Martin Frobisher: The wires could come down for many reasons. If they are out of alignment, they can get hooked on a pantograph. If there is a stranded wire or a defect, it can part or there can be corrosion failures. There are a number of different mechanisms for it to fail, but doing regular, routine maintenance, and observing it with things like thermal imaging cameras, gives us an opportunity to minimise the number of failures.
Q72 Karl McCartney: Is that mounted on rolling stock or drones, as Paul mentioned? Do you use an amalgamation of all those different things?
Martin Frobisher: To cover the whole network, routinely we use our helicopter with a thermal imaging camera because it can fly the whole line of route. We are working to get authority for drones to go beyond line of sight. At the moment, the regulations demand that a drone operator works within line of sight. We have a project that looks at taking it beyond line of sight. At the moment, for efficiency, a helicopter can fly the whole line of route and that is generally the easiest way of observing with a thermal imaging camera.
Karl McCartney: Milda, do you have anything to add?
Milda Manomaityte: We have several members that have developed solutions with train-mounted cameras that look at overhead lines, coupled with cameras recording what is happening and machine learning and artificial intelligence trained to recognise what a healthy overhead line and pantograph interaction looks like. Whenever there is a bit of deviation from a healthy line, you can send an automatic signal to say that maybe the line is sagging a bit or there is a misalignment. During weather events it will detect if debris is flying and hits the line. It automatically sends an alert that something has happened.
The power of that is that, if you can combine it with weather data prediction from the Met Office, you can think about the weather coming in. I had a conversation with one of our members, who was looking at and recording the lines through cameras. They detected where some of the neighbours had trampolines, and sent them weights to weigh down the trampolines, so that they did not fly over the lines. The power of data comes when you combine multiple sources. You then have rich, interesting intelligence that you can get to prevent the faults from happening.
Q73 Karl McCartney: Paul, you mentioned Canada. Are there any other examples? Do we have good practice in this country, or are we taking information from how other people are dealing with data across the world and using that to our advantage?
Paul Watson: A good example is that we work with the Netherlands around performance-based maintenance. That uses a lot of the technologies that Martin mentioned. It is about gathering data around the state of the railway, operations, sensors and all sorts of different technologies, and then bringing that together to devise maintenance strategies and trying to move to more predictive maintenance.
One of the big challenges is around any organisational change that comes with that change. Technology and organisation need to work together to build change. I know that Martin and Network Rail are working with ProRail in the Netherlands to share that knowledge, so that knowledge sharing does happen.
Q74 Paul Howell: On the data collection and data usage angle, you talked about the helicopter and the drawings and that sort of thing, Martin; you didn’t touch as much on better use of cameras on trains. On our recent trip to Tokyo, we met with Hitachi and saw their technology in terms of the way they are looking at the overhead lines in particular, but also just maintenance generally. I don’t know how much you are using their particular thing. Obviously, that runs on certain lines but not on other lines. If you were using their advanced technology on their trains, you would see what is happening on the lines that they run on. Do you have any opportunity to extrapolate that data in any way to see what is happening on other lines, or is that extrapolation not a valid base, if you see what I mean, because of the differences between places? They threw one final thing into the discussion. It was not just about the overhead lines; it was about the vegetation and everything around, because whether it be leaves on the line or whatever, it stops the train. What I am trying to get at is the use of data and the extrapolation opportunities.
Martin Frobisher: It is a really good question. Our main monitoring programme is from trains. Network Rail operates a fleet of 13 dedicated inspection vehicles, which are going round the network all the time monitoring the infrastructure. They have a variety of instruments to measure the track. There are lots of cameras that can provide data processing for observing other issues. We use that data to assess the network and to interpret and generate the maintenance plans that we need. We have 13 vehicles that are scheduled to cover the whole network, measuring everything from cracks in the rail with ultrasonic measurements to optical cameras measuring pantograph uplifts. There are lots of instruments on those trains.
You are quite right when you mention the Hitachi Class 800s and other such modern rolling stock. The future is to gather more and more data from passenger rolling stock. We have some data that is coming from some trains, but what you have just described is the future. Today, we have a dedicated fleet of measurement trains, but the future is about gathering more and more information from passenger and freight trains that are running round the network all the time, processing that data and getting an even better picture.
Q75 Paul Howell: I will come back on one slight point. Initially you said you were collecting it from passenger trains. You mentioned freight in the second part of your answer. I assume that the impact of freight trains on the line is completely different from the impact of passenger trains. I guess it is speed versus weight, or whatever.
Martin Frobisher: A freight train has a much higher load behind it. It performs differently. There are some freight-only lines on the network. At this moment in time our data gathering on freight lines is from our own fleet of 13 dedicated vehicles, but in the future more data from other trains that are running on the network is definitely the direction that we, as an industry, need to move in. We have some good pilots, but as time progresses that is a source of more data than we have today.
Paul Howell: Milda, do you have anything to add on those points?
Milda Manomaityte: Northern is doing a fantastic programme. We are equipping its trains with various sensors. That is such a good addition to the specialised infrastructure monitoring of trains. The passenger and freight trains run on the line every day and we can get so much rich data that complements the very complex data that the specialised trains gather.
Q76 Chair: To follow on from Karl’s question about the international picture, are there other countries that you can point to where they have shown the way and what can potentially be done? Or is the UK up there with the leaders at the moment?
Martin Frobisher: My sense is that the UK is right up there with the leaders. For the last two years I have been European chairman for the International Union of Railways in Paris. I meet with European railways a lot. Different countries are developing different technologies, but on balance the UK is right up there. Italy is doing well on digital twin. The Netherlands, as Paul mentioned, is doing really well on telemetry from assets. You see different technologies in different places, but the UK is right up there.
Milda Manomaityte: I am a bit biased about the Baltic countries, being Lithuanian myself. A good example in the Baltics is the Rail Baltica project. It is run by three Governments, so that is a complication in itself. Right from the beginning the project leaders agreed that data was absolutely important, and that the project would be run on open by default data principles. Any future contractors will have to agree to those principles and submit data to an open portal, where it can be shared. It is a good example. That also means aligning the three countries’ building information model standards. The three countries all developed in slightly different ways, so working in collaboration and making the decision, “This is important for us and we will overcome the challenges, whether it is standards or collaboration,” is really helpful when the Governments are right behind that.
Paul Watson: I echo what Milda and Martin have said. Europe and the world are doing different elements. Part of it is being able to integrate all of that because we automatically all rely on the same supply chain. If work is happening around sensors and things like that, it is about being able to leverage what is happening. Knowledge sharing will be critical for the future.
Q77 Chair: Milda, I want to pick up a comment you made in answer to an earlier question about looking at other potential impacts on the rail infrastructure. It is not just about monitoring the overhead lines, track and whatever. You referenced the Met Office and its weather predictions. Are we capturing related data that could affect the health of rail assets? I have heard stories that the way a farmer ploughs his fields can affect drainage, and therefore the resilience of the track. Looking more widely, are we capturing as much as we could?
Milda Manomaityte: That is absolutely right. There are unexpected outcomes when you have different sources of information coming in and applying minds, whether machine minds or human minds, to analyse the data. There are a lot of interesting projects going on in various sectors, but the beauty and the power of it will come when we actually work in collaboration on all of it. Maybe the earth is a little bit cold in that space and you have train vibration monitoring. You see that the track is vibrating. You have a vision camera and, as I said, the weather prediction. When you bring all of that information together, you get rich insight.
There is an opportunity to do better and to work in collaboration. That is where the guiding mind or the central view comes in to see that this part of the network is doing something very interesting or these people are doing something very interesting. When we bring them together, that is where the benefit is best.
Martin Frobisher: I can give a really good example of that. The UK Rail Safety and Standards Board has developed a brilliant risk model for the risk to the network of extreme rainfall. We are currently trialling that on the north end of the west coast main line. We take the weather forecast—we get a very granular weather forecast—and put that into the model, and it predicts the risk of earthworks failures. That model then tells us the optimum speed to run the trains. Being able to take a really good weather forecast and a really good risk model, and then letting that inform the right speed for trains to operate, is really good technology.
Q78 Chair: What additional support should there be from Government or from GBR, for which we had the draft Bill published yesterday? Who else should take the lead to make sure that there is the co-operation that you referred to?
Martin Frobisher: First of all, I think we get good support. The transport data strategy that has been published by the Department for Transport is a good document. We are getting strong support. Great British Railways will break down barriers between the companies that operate in the railway. Collaboration and the sharing of data is the best way to maximise the benefits. As those barriers come down and more data is shared, I think we will realise more opportunities.
I think Government are doing what is needed. Documents such as the national AI strategy and the geospatial strategy for 2020-25 are good. We get good support from Government. Great British Railways will help even more.
Milda Manomaityte: There is an initiative called the rail data marketplace that is led by the Department for Transport. It came from the rail sector deal. As of yesterday, it has over 600 organisations publishing data, with 80-plus datasets. This is really important. Network Rail, train operating companies and the Office of Rail and Road are also publishing data. This is where that support is really useful to encourage organisations to publish and share their information on the platform. You can come in and see all of the datasets and do something.
The other thing that our members and railway clients are doing with the rail data marketplace is to understand what problems it can solve. What are we trying to fix? If we want to understand how weather impacts infrastructure and how to prevent that, we can think about the datasets we already have, what we are missing and which organisations or other sectors we should work with to ask for data that we may not have within the railway sector.
Q79 Paul Howell: You made reference to digital twin; it would help if you gave us an understanding, in a short story, of what that is. I have read that the national data will not be a single digital data twin. Instead, it will provide the mechanism that will allow people to bring together information and models. Who do you think those people are who are going to bring that together? Who is going to use the data that is provided by the digital twin?
Martin Frobisher: It is a really good question. What is a digital twin? At this point in time there is no specific definition, and there is certainly no definition around data standards and data protocols. Digital twin can mean more than one thing. The way I see it is as a model of the real world, using data to simulate the real world, and then using that data to learn, experiment and draw some sort of conclusions.
The best digital twin model in the UK at the moment is produced by Birmingham University. They have been working on it for many years. It has different elements. Some look at the detail of wheel-rail interface and all the mechanics of how a railway works. It simulates the timetable. Its final interface is a picture of the lineside, and you can actually sit in the cab and drive a train in the model. It is an all-encompassing model of the network. We can use it to test the timetable and analyse the problems with timetables. We can use it to test vulnerabilities with various assets. We can simulate driver operation. It is a really good piece of work that has been done by a leading UK university.
We have smaller models. As Paul mentioned earlier, we use bridge-analysis technology. We fly a drone over a bridge and create a 3D point cloud. That creates a digital twin of the bridge, and then we can use artificial intelligence to analyse that image and identify defects. That saves us scaffolding out the bridge and makes the inspection a whole lot more efficient and more effective. We have many different twins for individual assets such as bridges, but the best all-encompassing twin has been produced by Birmingham University. That is absolutely brilliant and world-class.
Milda Manomaityte: To add to your question on who should be leading on this, the railway system and roads—we are talking about transport data—need clear collaboration and engagement rules. That is where the Government can be of tremendous help, by having a view of what is going on in other sectors and bringing them together. When you start working on something you get wrapped up in it, so it is useful to bring the different sectors together. I am not a specialist or an engineer. I don’t even know whether that could be hosted in one place. Maybe it would be a collection—a neural network of different twins.
Paul Watson: To add to those points, I have another example of the usage and who would be using it. There was a digital twin project in San Francisco where we looked at centres on the traffic network and light rail. There were lots of benefits from optimising, for example, 11 intersections. It gave a roughly 20% benefit in road and rail transit times for passengers. There are lots of good examples of isolated usage of digital twins.
Centrally, to add to Milda’s point on how we make a connected digital twin, it comes back to what Martin mentioned earlier around having standards panels and standards catalogues in place, so that when everyone is publishing data we can embed and link data formats. We can bring that together, and then that builds connectivity. It needs to be cross-sector. It is not just about rail or highways. It all touches weather data, so it needs to be wider than just a rail initiative to make a dual digital twin.
Q80 Paul Howell: I want to move into a slightly different space. We are talking about data collection. The Chair said earlier that even the direction a farmer ploughs his field in could have an impact on different things. There are obviously many different potential impacts of data, so you are collecting so much data. A phrase we’ve all heard many times is, “You can’t see the wood for the trees.” You need to be able to get the right access to that data.
As we know, data is a commercial opportunity for people these days. It is seen as something that has value. There is also the challenge of, for want of a better phrase, malign actors getting into the same data. How do you manage the data pot in such a way that you are getting value out of it without putting risk into it at the same time? Maybe I could start with Milda this time, for a change.
Milda Manomaityte: I refer that to Martin. He is the specialist on safety and data.
Martin Frobisher: There are a couple of questions that you have asked there. First of all, on seeing the wood for the trees, the latest technology helps us to do that. I will give you a good example. We have an experimental technology at London Bridge station on the CCTV system. We have 2,000 CCTV cameras in the station. No operator could scan them all at the same time.
We had a project where we brought lots of actors into the station to assault one another and bring weapons into the station. We trained artificial intelligence to interpret the images, so there is a huge data feed. We have trained the artificial intelligence and now, if somebody does something of concern, it flags it up on the CCTV screens because we trained artificial intelligence to interpret that massive amount of data. I think that is really clever. You can see the wood for the trees using that sort of technology.
Paul Howell: Would that be replicated across other stations?
Martin Frobisher: It is something that we are developing at the moment. It is a technology under development.
The question of cyber-security is absolutely vital. We have invested heavily in a cyber-security centre in Manchester. We have people round the clock in that centre monitoring our databases and our network. We can see who is dialling into our network at any moment in time, and where they are dialling from. When a database gets downloaded, alerts flag up on the screen.
I was in there a couple of weeks ago and somebody was downloading a structures database about bridges. The alarm flags up on the screen. The operators check who is downloading it. It is a structures engineer. It is sensible. It is not something to raise a concern. But if people are accessing data that they shouldn’t, or people are dialling in from places that would be of concern to us, we have all sorts of continuous monitoring on the network. Our cyber-security centre is there to protect the network from those kinds of issues.
Paul Howell: Milda, do you have anything to add?
Milda Manomaityte: On the vast amount of data, and coming back to the point I was making, we need to start to understand what we are trying to solve. Then it becomes a bit clearer as to what data we need. You are right that there is a lot. Sometimes when we say, “Can you share your data?”, we get the answer, “Which one? We have so much.” If we understand the challenges for the infrastructure owners, the train operating companies or local authorities, we can think about what we need. We are starting from a problem and then seeing what we have and don’t have.
Paul Watson: As well as having too much data, we must remember that we cannot always collect data about everything. It is about recognising what is in the data and what is not, and how we collect data to fill the gaps. It comes back to the discussion on standards. It is about making sure that the data that is published and used has gone through data governance, is checked and is right to be used and that we know what it is being used for. Those are just two points to build on.
Paul Howell: To build on your point, Paul, as parliamentarians we get involved in many things. I once went to Fylingdales to look at what they were doing there. The integrity of data was more important than anything else because that is what the decisions were going to be on. The testing of integrity is critical to what you are doing, and I endorse that.
Q81 Chair: I would like to pick up on one of the points that Paul mentioned, which is the commercial value of the data. It is a commodity. The transport data strategy document says that there should be an open-by-default approach, and I would like to tease that out a little bit more. As the railways are currently structured, is there a disincentive financially for operators, Network Rail or anyone else to share the data? As the railways are restructured, what considerations do we need to give to enable that open-by-default approach?
Milda Manomaityte: I mentioned in my pitch that the data only has value when you start using it and getting insight. There are various ways of using it. There are datasets that are, and will be, open by default and accessible for everyone. There will be datasets that might be under commercial licences. Rail data marketplace allows that. If you want to publish it and make it open for everyone, you can do that. If you only want to share it with universities, you can do that as well.
What is happening already is that the leaders—the pioneers—are starting to use data and show the value, and the rest follow. We need champions, innovators and trailblazers to start using and showing. It is a conversation. We see it and we react. It doesn’t need to be a blanket approach.
Chair: Paul, do you want to add anything?
Paul Watson: From a project perspective, when we do transportation projects, new contracting models are coming to the fore. Network Rail has recently embraced the project 13 model, which very much incentivises organisations to work together in an ecosystem approach and share data throughout the value chain of data and understand the ecosystem. Potentially, traditional models become a barrier, but the ecosystem models, collaborative models and working across industry mean that there are ways of unlocking that. That comes down to leadership and getting the right approaches in place.
Martin Frobisher: The principle of open data is sound because you never know how people can make use of it, collaborate and provide good solutions for passengers. There is a really good example. We put data on the web about the availability of our lifts and escalators. If a disabled passenger gets to King’s Cross station and wants to go to Waterloo, knowing that the lifts are working at both ends means that they do not go down to the underground and then get to Waterloo and have to turn back. That sort of data made available to an app developer can offer real benefits to passengers who need the information.
We also need to think very carefully about what we release. We would not want to release data that was of use to terrorists. We would not want to release data that was commercially sensitive. The principle is sound, but we always need to be thoughtful about how the data might be used and to think quite carefully about what we release.
Q82 Chair: Who polices that? Who makes that decision?
Martin Frobisher: For Network Rail data, we take a view on what we release. By and large, if it is not commercial, or security or people-sensitive, we want it to be out there. People developing apps are clever: they can find all sorts of solutions that help passengers. The principle of it being open where we can is sound, but we need to be mindful of sensitive data to make sure that we think about that and apply judgment. That is how we work.
Q83 Chair: I appreciate that is for your data in Network Rail but, building on what we have been discussing, if we get to a point where you are accumulating data from Network Rail, from Hitachi trains or whoever, and you have built up a much bigger model, who polices that?
Martin Frobisher: That is where the legislation that you will be scrutinising comes in. Having one guiding mind for the railway allows you to make better decisions about that sort of thing. Great British Railways will be able to take an overview about railway data. If it is more joined up and there are fewer barriers between parts of the railway, it is possible to make better decisions.
Q84 Chair: If I might be slightly cheeky, given that the legislation in draft was only published yesterday and we will be doing our scrutiny of it, from what you have seen of the draft Bill, is it going to enable GBR to do those kinds of tasks?
Martin Frobisher: I think it will. The first clause is the key: creating a network guiding mind, a governing body. The creation of an organisation that can take an overview, break down barriers and get things working in a more joined-up way is tremendously helpful.
Milda Manomaityte: We are always stressing the importance of involving the suppliers, the contractors and the people who develop these technologies day in, day out. They are specialists in it, and they work in or outside the railways. It is absolutely crucial to engage early. There is a good opportunity now to have those dialogues and build something that works for the entire sector.
Q85 Chair: I would like to turn to the operational decisions that are made on the back of the data you collect. In the storms that hit the country in recent weeks, a number of operators—for example, ScotRail and others—decided to suspend all services. I am not asking you to judge whether that was the right call or not, but to what extent would they have used the predictions from the data available that it would be too dangerous to run trains over a particular line or network, given the conditions? What I am trying to ask is whether we are being over-cautious at the minute. Is there a role for data to make more informed decisions as to whether services should run or not?
Martin Frobisher: If you take Scotland as an example, they have somebody in their control centre who constantly monitors the weather data that is coming in. We have the best weather models that we can get. Following the Carmont accidents we employed an expert, Dame Julia Slingo, who was the chief scientist at the Met Office, to give us advice on how we could get the best weather models and data. That advice has helped us to improve the weather data we get. The Scottish control centre is getting good information on weather forecasts. Our engineers have provided guidance on how a weather forecast will translate into an impact on the network. They have applied that guidance.
What we can do with data is to create ever more sophisticated models. The model that I described and that we are trialling on west coast north takes that to the next level. It is really clever because it takes the whole- system risk. If you impose a speed restriction, trains see more red signals and you increase the risk of signals passed at danger. Imposing a speed restriction creates a secondary risk. It protects the risk of a train hitting a bank slip, but it increases the number of red signals. The RSSB model is all-encompassing. It allows us to find the optimum speed and to balance what speed is reasonably practicable to run the trains. It is a really clever way of answering that question.
We have good data. We do analysis of that data and we have people in our control centres, but the next generation of models will allow us to take that to the next level.
Milda Manomaityte: This is where the UK Rail Research and Innovation Network can really come in handy. It is a partnership of academia and private sector rail clients. The UK Rail Research and Innovation Network has a centre of excellence and digital systems specifically focusing on the railway sector. That is where you can take the data and the modelling and work with researchers and people who work on the railways every day. You can run the models and see how your decision impacts the change. You can do a lot of that with academia, and it is really good.
Paul Watson: My only comment is to agree with all the points raised. Obviously, currently, whether in the scenario you gave or in design and engineering, there are a number of assumptions. The more data we capture and the more sophisticated the models get, the more we can either validate those assumptions or realise that those assumptions might be conservative. Using data and more sophisticated models can help address that.
Q86 Paul Howell: We have talked a lot about the amount of data that you are getting, and where it can be used or secured. An obvious question is: where are the gaps? Are there any data pots that are not getting filled, where the information is difficult to find, that you think would be particularly helpful?
Paul Watson: Part of what we suffer from is legacy assets and new assets, as well as legacy systems and new systems. Some of the older systems across all transportation were not necessarily designed to get information out. They were more designed for information in. With older assets, we might not have full digital twin models. As we build more and more new assets, we will survey more assets and get more information. For me, it is about the whole piece of having legacy infrastructure and legacy systems.
There are lots of examples with organisations now introducing new systems to bring data together. We did work for Govia Thameslink Railway around exactly that and where the gaps are. Sometimes it comes back to data sharing. Sometimes we look to share data, but it is difficult to share data. You also need to understand the data to be able to interpret it, because sometimes it was not built to standards and things like that, as we have touched on. Those are probably the gaps that we generally see at the moment.
Milda Manomaityte: I absolutely agree with those points. Collaboration is important to plug the gaps. We have a lot of pilots. A lot of research has been done. Things are happening in the sector in various areas, so bringing all of that together and working in collaboration is important.
I had an example of somebody saying, “We have a tunnel.” To understand how cracks appear in a tunnel, just one tunnel is not enough. You need to have a lot of data to start predicting patterns. If there is an opportunity to plug into other infrastructure asset owners and say, “Can we share this data and work together to figure out how the infrastructure behaves?”, that is where it is important.
Martin Frobisher: I agree with what has been said. As Paul said, the biggest issue is old assets. When the Victorians built the railway, they put in good drainage. Over the years, some of the records have got lost. Over the last two to three years, we have been surveying the whole network to find drainage assets and to make sure that our database is up to date. We have invested a lot of work in making sure that we gather that sort of information about hidden old assets. That is a good example where we have had to invest extra time and resource in mapping and surveying the network.
There is lots of opportunity in the future to gather more data from passenger trains, as we said earlier. Another area that we need to work on is that good asset data requires very precise positioning. GPS only has a certain accuracy, and to really analyse assets you need to get it down to much greater accuracy. We end up using other sensors in addition to GPS to get positional accuracy. I am hoping that quantum technologies could take that a step further in the future. I don’t think we have major gaps in our asset databases, but there are things that we can improve and there is lots of opportunity for the future.
Q87 Chair: The railways have a finite budget to spend. There is always going to be tension between maintaining existing assets and investing in new assets, be they new railway lines, new stations, or whatever. Do you get a sufficient proportion of the investment budget to look at data, particularly given that it is a spend to save sector? If you are able to intervene on infrastructure maintenance earlier, it prevents a more costly repair later. Is a sufficient amount of the budget given to the data side?
Martin Frobisher: I believe the answer to that is yes. In the last funding control cycle for Network Rail, the settlement for maintaining the network for CP6 was £43.5 billion. For CP7, which starts in April this year, it is £43.1 billion. The funding for the network is sufficient.
The maintenance budget will increase. The renewals will reduce slightly, but investment in data is healthy. Our intelligent infrastructure project for our CP6 control period was £300 million. We have allocated £120 million already for control period 7. We are investing heavily in data and our intelligent infrastructure project. I believe that the funding for the network that has been determined by the Office of Rail and Road is right.
Q88 Chair: Another potential impediment to maximising the use of data is skills gaps and shortages. Milda, in your written evidence you noted that the rail sector “suffers a perennial skills shortage in technical roles”. How do we address that to make sure that we have the skills base that can maximise the use of the potential of data?
Milda Manomaityte: The beauty of working with new technologies and data, and new analytic tools, is that yes, on the one hand you need highly skilled people who understand the technology and the data, and you need those people in the traditional engineering companies. It also makes it more accessible. I shared the example about the interlockings. Previously, the data was only understandable by very specific people with very specific technical knowledge. It becomes more widely understandable and accessible to people with some training.
It is about having the ability to make sure that people who already work in the industry are aware of the technology changes and are being trained and upskilled. If we are optimising one area where we can have predict and prevent maintenance, maybe we can think about those themes and think about what we can do instead of going on to the track and fixing faults all the time. That requires organisational change. We need to think about the skills. On the one hand, we have a big shortage. On the other hand, we are optimising our workforce. It is how we can think about the organisation and where best to use the people we already have.
Q89 Chair: I appreciate that the country has a wider challenge in having people with relevant engineering and technical skills. Martin, from Network Rail’s perspective, what is your strategy to increase the skills base?
Martin Frobisher: When we are developing new technologies, our research programme works with lots of small to medium enterprises that have innovative people with real, deep technical skills. I could quote about 17 research projects that are using artificial intelligence in one way or another as an example. What we are generally doing is turning to organisations that have deep, technical and specialist skills in those areas. Once we have found something that works, our strategy is to train our workforce to use the new tools.
We have a healthy research programme. There was £250 million during the last control period. We have used that with lots of small to medium enterprises that have deep technical skills to develop new products. Once we have got them to the point of roll-out, we are training people to use those tools. I think that combination works effectively.
Q90 Gavin Newlands: Good morning. I’m sorry for being late for the session. I want to come back to something you said earlier, Martin. I think you said you can see who is dialling in and where they are dialling from. Theoretically, would it not be possible to mask that from a cyber-security point of view? Maybe I watch too many films and TV programmes.
Martin Frobisher: It is always going to be a game of cat and mouse between the people who are manning the security centre and the cyber criminals who are developing better tools, using AI and all sorts of other technology. We have a really well-resourced security centre. We get good advice from the experts. We don’t just rely on one support contract. We use Microsoft to a large extent, but we have other tools from other suppliers. We have a diversity of tools from a number of suppliers. We have a well-organised cyber-security centre, and we recruit graduates who have really good skills to that centre. We have a strong team. We are supported by good suppliers, and we are using the latest technology.
I can’t sit here and say “Never” because, of course, the hackers are constantly developing their technology.
Gavin Newlands: I wouldn’t ask you to.
Martin Frobisher: We take this seriously. We resource it properly with good people and good suppliers.
Q91 Gavin Newlands: I have been contacted by a couple of folk over the last few months, not necessarily about this but it has made me think. This is a bit of a left-field question for you. Since the pandemic, the conspiracy theorists all seem to have united in many ways and on many issues. The sharing of data is one of the issues that they seem to coalesce around, with all the privacy issues that that entails. Is that something that has come up in your line of work? I rather suspect it is more likely to come up for us as politicians, as we are challenged by the public at large on what we are doing with their data and what we are allowing to be done with their data. Has that come up in your line of work thus far in potential problems, or do you have any advice for us in our dealings with the public? It is potentially an increasing concern, and something we will have to be aware of as politicians.
Martin Frobisher: It is something that we have to manage carefully. We are subject to freedom of information requests in exactly the same way that you are. I imagine that the way our organisation works is very similar. We are subject to data protection. Every year, we get each department to confirm that it is compliant with the relevant policies. We have an owner for each of our corporate systems, so that we make sure that we are compliant with the various requirements. When the public ask questions, we are totally transparent in what we can provide, unless it is commercially sensitive or security related. We always try to be very open.
Gavin Newlands: Logical answers don’t always satisfy a conspiracy theorist, I’m afraid, but there we are.
Martin Frobisher: But we try to be open. It is in our interests.
Chair: Thank you. I fear the clock is against us, and we have another panel to hear from this morning. For now, I thank the three of you very much indeed for your time and evidence.