Those of us who now need afternoon naps can find the modern world baffling, so it’s good to know that some things endure despite apparently passing their sell-by dates. And I include myself in that. Waving chequered flags to warn those in the four-foot that their lives are endangered by an approaching train spans three centuries and, even on the East Coast main line, you can still find surveying teams using lengths of string to measure curvature. Simple and effective was a global preference before ones and noughts came along.
The pace of technological change over the past 20 years has been breathless but hugely enriching. That said, change can bring consequences nobody envisioned – fires that need to be fought. So, there’s an understandable tendency to take a risk-averse approach when it comes to innovation in safety-critical environments. But, if we’re to benefit from the opportunities it presents, we have to recognise and embrace the associated challenges.
A generation of railway surveyors will never have held string, electronics mostly superseding it over the past couple of decades. As with everything, the capabilities of today’s instruments continue to improve over time. Most share common technologies – inertial measurement units, satellite positioning, laser scanning, photogrammetry – and capture essentially the same things: Easting, Northing, Elevation (x, y, z), perhaps Intensity (distinguishing between different materials) and Red, Green, Blue values. But, to maximise commercial interests, developers often employ proprietary software to process data and translate it into something useful for the engineer.
The railway, of course, comprises many disciplines – signalling, drainage, OLE, renewals – all seeking data for different design or construction requirements, using different methodologies at different levels of detail. So, there’s clear potential for inefficiencies and additional workforce risk through repeat visits to the same place.
Those realities are amongst the drivers of a project being led by Network Rail’s Signalling Innovations Group (SIG). During CP6, it intends to grow the survey and design tools it’s helped to develop, alongside which it will undertake research into the most effective ways of integrating survey data from multiple sources, as well as the complex influencing factors. The hope is to reduce the requirement to venture on or near the line.
SIG’s ambition is for a variety of surveys to be made available via a central open-source repository which everyone can tap into, although there is a recognition that it’s just not practical to have every possible survey type for every location, all gathered under consistent conditions. At the same time, an obvious fundamental requirement is that the extremely high levels of quality and accuracy needed for future automated design cannot be compromised. Dealing with this latter issue demands a much deeper understanding about the various methods of data capture.
Stitch and blend
To establish comparable data sets from a variety of techniques, SIG needed to set up a baseline survey test site away from Network Rail’s main line infrastructure. Initially, the company’s test tracks at either Melton or Tuxford seemed the ideal location, but constraints arising from their successful full-time role for train testing meant that it was instead decided to approach one of the country’s heritage railways.
An agreement was reached with the Midland Railway Centre at Butterley, Derbyshire, delivering a welcome financial boost to an attraction celebrating its 50th anniversary in 2019. At Swanwick Junction station, Bridgeway installed 15 control points through a 200-metre-long double-track section, amidst S&C, signal posts, platform furniture and the like.
“The facility gives us the opportunity to carry out multiple surveys within a narrow time window,” says Natasha Purewal, Network Rail’s scheme project manager, “minimising differences in factors such as weather, lighting and the presence of transient objects.”
The first event took place in mid-March, attended by four suppliers using manual trolley-based equipment, a rail vehicle-mounted system and an Unmanned Aerial Vehicle, otherwise known as a drone. Let’s make the point now that it is not Network Rail’s intention to identify a preference for one technique or product over another, and what follows here is not a sales pitch.
Devil in the detail
SCCS and Korec both demonstrated trolley-based kit, respectively using the Amberg IMS 5000 GPS system and the Trimble GEDO CE.
In their basic forms, these use an inertial measurement unit (IMU) – a device comprising gyroscopes and accelerometers – to measure cant and gauge as they are pushed along the track. But that only creates a relative model hanging in space; by recording GPS/GNSS coordinates alongside it, the IMU trace can be accurately transposed onto a real-world SnakeGrid.
The addition of a 360º laser scanner enables a 3D point cloud to be mapped of the trackside environment, extending out to the boundary fence. The scanners spin up to 200 times per second, collecting around 1,000,000 points.
Although control points are captured as a matter of routine, Amberg and Trimble’s trolley-based systems don’t need control points and can therefore provide a continuous measurement of track conditions, such as twist faults. Typically covering two miles in an hour, the process is quicker than conventional fixed surveying techniques and, by reducing the number of people involved, represents a means of gathering large amounts of data whilst minimising workforce exposure to on-track risks.
Critically, what this approach offers is high resolution. “At walking pace, we’ll have a laser profile every 5mm,” says Dave Dampier from SCCS, “so you’ll easily find the centre of a target; and the same goes for a feature such as a catchpit or a signal post.”
“What’s important is to have an understanding of the accuracy of all the different coordinate systems,” asserts Korec’s Matt Lock, “because that determines whether you can use the data for an engineering application or whether it’s more of an asset registry.”
Mounted to a specially-equipped Class 37 locomotive, provided by Colas, was a portable mobile mapping unit from Balfour Beatty Technical Services, known as OmniCapture3D.
The key advantage of this kit is the ability to quickly survey an entire route – comfortably covering more than 100 miles in a day – with little or no impact on traffic. There is no need for possessions and no boots on the ground, except at the depot when the equipment is being commissioned or decommissioned.
On the Class 37, the system can gather data at 70mph, although speed does have the effect of reducing point density and it therefore offers less detail than a trolley-based system. To help infill gaps or ‘shadows’, two offset Z+F Profilers are used, angled towards each running rail. Again, these can capture one million points in 200 profiles every second – effectively one profile every 80mm or so – and are fully synchronised to IMU and GPS units, as well as a tachograph.
There’s also an imaging device to provide additional context, delivering 4K, high definition, infrared or spherical views.
“It’s certainly appropriate for feasibility studies and the accuracy we’re going to now gets us to a higher level of detail, so you can go on to do your design with it,” says John Garlick, Balfour Beatty’s measurement services manager.
Eye in the sky
In principle, the approaches described so far gather three-dimensional data from a radial viewpoint above the four-foot, moving only in one dimension; as a result, nothing is seen beyond the first obstruction encountered by the laser. However, the advent of unmanned aerial vehicles (UAVs) has revealed alternative perspectives from which both plan and elevation views can be derived.
Bridgeway Aerial showed off a DJI Matrice 210, one of several aircraft operated by the company for inspections, photography, 3D modelling and LIDAR applications. They are capable of acquiring geospatial data that would previously have been impossible to secure.
In the case of a structure, a series of high-resolution photos and associated GPS coordinates are taken from strategically-defined positions – including upward views into ‘hidden’ elements such as arch soffits – and stitched together using photogrammetry software. An orthomosaic representation is then extracted (that’s the task of correcting geometrical errors caused by topographic relief, lens distortion and camera tilt) and, from this, a geo-referenced three-dimensional point cloud is processed. It’s even possible to produce a physical model of the structure using a 3D printer.
The data points have a position relative to each other derived from the UAV’s GPS system which, at any one time, achieves its accuracy by typically connecting to around 18 satellites. But, to obtain an absolute position, the cloud is tied to trackside control points by manually identifying them at pixel level on the photographs.
The resolution is determined by the requirements of the brief and defines the speed at which the survey is undertaken. Paul MacMahon, senior UAV pilot, explained: “For an accuracy of ±20mm, we fly quite slowly – perhaps three metres per second – and gather several hundred pictures. It will also depend on weather conditions – how the UAV is performing on the day.”
Horses for courses
All these devices currently play unique and valuable roles in their own right, having proven themselves to fit within the parameters and constraints of rail. Whilst the technology will doubtless evolve, the consistent challenge will be to fulfil the client’s requirements and ensure data quality compliant with appropriate standards.
“At the moment we have bandings that each system is signed-off against,” says Wayne Cherry, Network Rail’s senior innovations engineer. “We’re now trying to get everything to a level that provides seamless integration, to give us confidence within our own disciplines and throughout the industry – supporting the principle of ‘survey once, use many times’.”
The presence of four key suppliers at the Butterley test site indicates a willingness to engage and collaborate in realising the Signalling Innovations Group’s stated ambitions. There are commercial and safety imperatives that they are clearly buying into. A workshop is being held where the companies will present the information captured on the day – initially demonstrating the capabilities of their own software – before delivering the data in an open format for manipulation by SIG into a single model.
Going into CP6, there is a recognition that, at times, the railway is capturing too much duplicate data across the various disciplines and is not always being clever enough about what to do with it. This project shows an emerging commitment to adopt a much more coordinated approach.