Guest writer: Tim Flower – professional head of maintenance at Network Rail.
Intelligent Infrastructure is Network Rail’s digital asset performance management programme, using technology to turn data into intelligent information so the frontline and supporting teams can work smarter and more safely to deliver improved services for passengers and freight customers.
Ultimately the goal is to reduce expenditure whilst improving infrastructure availability by:
- Understanding the probability of individual asset failure;
- Predicting when failure will occur;
- Forecasting the impact on the operational railway;
- Planning intervention prior to disruption to train services.
The programme isn’t just about introducing huge amounts of new technology, it has been carefully designed to look at how we can maximise the value from the data we have whilst working closely with our research and development programme to make sure we continue to be at the forefront of technology introduction.
People and culture transformation
Throughout all stages of the programme, we have recognised that there will be significant changes to how our teams interact with data and technology, and how work will be specified, planned and delivered. Successful delivery of our plans will rely on our ability to enthuse and inspire our teams to work with us through this change, giving them tools that they will want to use because it will make their working lives easier.
The cornerstone of the programme is the engineering assessment we undertake against each asset or asset system. This is performed utilising reliability-centred maintenance techniques that originated in the aviation industry and which were restructured to be applied across other industries by John Moubray in the 1980s, in a process he named RCM2.
The process applies failure modes effects and criticality analysis (FMECA), enabling us to assess and revise our maintenance standards and inform our infrastructure monitoring and asset management plans. By summer 2020, we are aiming to have completed the safety case utilising RCM2 techniques to remove some of our cyclical track circuit maintenance through reliance on embedded monitoring, which, if successful, will be rolled out across other asset types such as level crossings and busbars.
A further benefit of this process is that it allows us greater understanding of the legacy assets on our network, allowing us to pinpoint which asset types are unable to deliver the availability requirements for a route section. This information is used to focus our research and development programme on creation of next generation assets, which are developed in accordance with our ‘design for reliability’ processes.
The outputs from the FMECA are also used to specify the deployment of asset monitoring. There are three main methods of monitoring assets depending on the requirements identified.
Assets are fitted with sensors that monitor their condition, reporting back to a central system known as RADAR. This is monitored 24/7 by Intelligent Infrastructure technicians who use the information provided by the system to predict an asset failure, providing guidance to front line teams to help us to diagnose the likely failure mode and prevent the failure occurring. Use of this type of monitoring is widespread in Network Rail, as shown in the diagram below.
The introduction of Internet of Things devices will increase this footprint massively in the coming years, so research and development is focusing on how we can generate the best return on investment from the myriad of sensors that are now available.
Currently, there are two separate approaches to this, either having a dedicated set of infrastructure measurement trains or by fitting monitoring equipment to in-service vehicles.
The use of a dedicated measurement fleet is much more mature, and we now have 13 dedicated vehicles measuring the profile, depth and internal and external cracks in rails, the condition of the formation and geometry of the track, contact force and position of the overhead line, the loading gauge and quality of the radio signals.
The most recent innovations in this area are:
- Plain-line pattern recognition (PLPR): With this system, pixel-width images of the track are recorded at up to 70,000 times per second across seven cameras. These images are integrated with laser profile and track geometry data and processed using machine vision to deliver defect reports to section managers, replacing the need to manually inspect plain line track. So far, this has been rolled out across approximately 9,000 miles of track, with a further 5,000 miles planned for the future.
- Eddy-current inspection: Another replacement for manual inspection, this helps us understand the extent of cracking in the surface of a rail as the system is able to measure the depth of cracks rather than just their length. This system is now live across much of the network.
- Switches and Crossing (S&C) Dynamic Measurement: Here, we run a specialist train through the S&C in both the normal and reverse position to give us a greater understanding of how it behaves under load.
Monitoring infrastructure using in-service trains is currently much less developed but has huge opportunities in the future to provide a near real-time indication of how the railway system is performing. Currently, several service trains have track geometry monitoring systems or accelerometers fitted and work is in development to utilise this data more completely in order to support maintenance activities. In 2020, a dynamic overhead line force measurement system will be fitted to an Avanti train on West Coast main line and work is progressing to fit similar systems onto East Coast and Great Western main lines.
One challenge that remains is the ability to predict driver-reported rough rides (DRRR). Drivers and traincrew are trained to report any suspicious vehicle movements to the control centre, in case there has been a sudden failure such as a rail break, track buckle or embankment failure which may pose a risk of derailment. This is a last line of defence because it is very unlikely that a sudden infrastructure failure will be picked up during routine asset measurement recordings, which only take place every four weeks.
The default response to a DRRR is to send a track maintenance team out to find what is wrong with the track and fix it. However, in most cases, and after many trains have been delayed, the maintenance intervention results in “No Fault Found”. We are launching a design contest to solve this issue, whilst also exploring the ability of smaller, simpler systems to give us an understanding of the track asset movement – without the need for full track geometry systems. It is expected that we will have prototypes in place and under trial within the next twelve months.
The ultimate aspiration from a train-borne measurement and monitoring perspective is to blend the right mix of dedicated measurement capability with in-service monitoring to ensure both safety and performance risks are mitigated.
Network Rail has several fixed monitoring systems that are installed on the infrastructure and monitor trains as they pass. These measure axle bearing condition, using hot axle-box detectors (HABD), wheel condition using wheel impact load detectors (WILD) and pantograph condition using pantograph monitoring devices. The information generated from these systems is shared with train operators to help them improve their management of the condition of their vehicles. In the event of a serious defect being detected, operational rules are in place to mitigate safety risks, such as vehicle derailment due to rail breaks.
Network Rail is now working with TOCs and FOCs (train and freight operating companies) to understand how these systems can be rolled out further, to provide greater coverage, or enhanced, for example to included acoustic bearing monitoring.
This review is also considering whether there is a requirement for additional wheel lathes across the network to enable proactive wheel turning. It is expected that an improved average condition will not only extend wheel life but will also reduce wear and tear on the rail and ballast, improving the whole life cost of track – truly a whole industry benefit!
A huge amount of value will be derived from utilising advanced analytics and machine learning techniques to drive a greater understanding of asset condition and rates of degradation. These techniques, which will be applied across all asset systems, will use our existing data sources and also help us to understand what new data is required to drive efficiency and performance.
Initial delivery has concentrated on track and signalling, building on the decision support tools that we delivered as part of our ORBIS programme (issue 174, May 2019). For track, the initial challenge was to make sure the geometry measurement systems aligned across multiple runs to enable the degradation models to be as accurate as possible. The following diagram shows around 100 yards of twist measurements across three runs. Here the peaks and troughs of the measurements are offset between runs by approximately ten yards, making analysis difficult.
The teams have developed an algorithm that sets a baseline run and shifts all other runs to align to this baseline, as demonstrated in the above diagram, which shows seven runs over one mile, now in complete alignment.
The aligned data can then be used to predict when the track will degrade past alert and intervention limits, allowing intervention to be planned proactively.
Next to be delivered is cyclic top capability, which is being delivered in instalments. The first instalment will focus on a visualisation of the data to help maintenance teams get a better understanding of the nature and potential cause of the fault and how it might progress.
Future capability will begin to add predictive elements to the tool, showing the rate of degradation of cyclic top faults and highlighting any sites which are expected to become faults shortly. This will allow maintenance teams to plan their interventions earlier and treat a defect before it affects the service.
For signalling the teams are initially focusing on delivering diagnostic and predictive capability from the points and track circuit condition monitoring data. An example of how we are utilising the points data is below:
These are just some of the examples of analytics we plan to deploy as part of the programme. We have now mobilised the delivery teams and expect to be delivering capability across all disciplines from late 2020 onwards.
Visualising the railway
When Network Rail rolled out imagery and data from the first national aerial survey in 2016, it marked a major milestone in railway analysis and early project work that could be carried out from the safety of the office.
Using the Geo-RINM Viewer, planning and maintenance teams could carry out inspections, measurements and analysis of the railway without the need for manual inspections.
Desktop access to high-resolution images and 3D digital terrain and surface model data, the survey – carried out by the ORBIS programme – proved a resounding success. Routes were soon requesting updated data to keep pace with the changes taking place across the infrastructure.
To meet this demand the Intelligent Infrastructure programme was asked to develop and improve on the first survey. Working with Network Rail’s Air Operations team, new surveys were carried out over the past two winters – benefiting from reduced leaf and foliage cover to improve clarity of the network. This refreshed data is now being rolled out to the routes. Date labels and time sliders have been added to allow comparisons between old and new imagery so that accurate earthwork changes can be measured and changes to the infrastructure can be clearly seen.
Currently we use lots of different planning tools that have been developed by different routes, which means planning is inconsistent across Network Rail. The tools have been developed by local experts, which means that, while they are fit for purpose, they aren’t supported by our IT department. As a result, they often require a lot of ‘handle turning’ to update their core information and they can’t talk to other systems. This means that allocating people, materials and equipment to deliver the work in the access available is difficult and time consuming.
The planning workstream has therefore been scoped to create a new common set of planning tools holding information on the whole network. These will be supported by Route Services Information Technology (RSIT) and Asset Information Services (AIS), which can be adapted to suit different routes and regions as they choose.
There are three main outputs to this work:
- Asset Lifecycle Planning System: This will help to manage the long and short-term workbank, helping to optimise maintenance and renewals work efficiently and effectively.
- Work Planning and Scheduling System: This new scheduling system will take information from across the whole network and create schedules, taking into account unplanned work, asset conditions, criticality, best access times, resource available, materials and equipment. Work schedules will become more stabilised. This will help make plans more stable in the long term, allowing for work to be planned in a safer manner.
- Time Recording System: This will accurately record the time Network Rail staff have spent at work against various activities, reducing administration and helping managers to understand the true unit cost of a job.
We have already deployed handheld devices to our frontline teams and supported management and engineering staff. To date, Network Rail has launched over 60 ‘apps’ to help our teams deliver effectively and efficiently, and all our analytics capability will be mobile device compatible.
In late 2020, we will make our condition monitoring trace information available to our faulting and maintenance teams, enabling them to check in real time that their activity has delivered the planned output.
The next steps will be to bring all our disparate data sources into an application that gives all the applicable information about a particular asset to our teams to enable them to pinpoint faulting activity, enabling a faster fix. The teams will be able to update this information on site with measurements or asset information, which will then be used to drive analytics improvements.
The goal with all our mobile capability is to provide the end user with a quality interface that is simple to use, intuitive, provides them with the information they require and only asks for input where it is absolutely necessary. In this respect, there is room for improvement with a lot of our apps, and we will be working closely with end users to give them what they need.
Delivering at pace
We have, wherever possible, introduced an agile methodology to deliver incremental capability, obtaining regular feedback from end users to drive future development.
By delivering something that users can both use and provide feedback on immediately, we are able to iterate through the development cycle every two weeks to quickly deliver an end product that meets all of their needs. Previously, it has taken up to two years from requirements gathering to an end user seeing a product, now it can be as little as two months for a first release!
We are starting to see the first benefits of the new way of working. More consistent and connected data will enable us to make earlier and better decisions that will result in a safer and more reliable infrastructure, both reducing our expenditure on emergency repairs and giving our customers and passengers a more reliable service – all through Intelligent Infrastructure.
Digital Surface Model (Hillshade) created from LiDAR data acquired in Winter 2018.