RAILWAY embankments and cuttings are critical but vulnerable parts of the rail network around the globe. In some countries, like Britain, many are over a century old and under increasing pressure from heavier rainfall, changing land use and ageing construction.

“Traditionally, slopes have been monitored by sending engineers out to site to read boreholes or other instruments,” says Jayson Jones, senior engineer at global engineering supplier Dywidag. “That works, but it is slow, labour‑intensive and leaves a blind spot: what happens between visits?”

To fill the gap, Dywidag has partnered with specialist sensor supplier World Sensing to develop a system comprising highly sensitive tilt sensors, cameras and bespoke software to deliver near real‑time alerts when the ground moves.

World Sensing had already developed a tilt sensor with the ability to notice and report movement. On its own, that sensor was a useful component, but Jones says Dywidag’s experience of failing embankments and slopes made it clear that the railway needed something more integrated and automated. The case against manual reading wasn’t hard to understand, based on expensive site visits, with limited recording points, that could easily miss sudden or night-time movements and no dedicated link between movement, imaging and operational decisions.

The system developed by Dywidag, branded Smart Guard, continuously watches critical slopes, automatically triggers cameras when movement occurs, and gives asset owners and operators usable, near real‑time information. The system integrates tilt sensors on embankments and cuttings, cameras covering the same assets from key angles, communications hardware connecting field devices to the cloud and a software platform and user interface designed for rail users.

“Traditionally, slopes have been monitored by sending engineers out to site to read boreholes or other instruments, that works, but it is slow, labour‑intensive and leaves a blind spot: what happens between visits?”

Jayson Jones, senior engineer at global engineering supplier Dywidag

“A key technical challenge was that sensors and cameras use different hardware and communication paths,” Jones says. “To turn them into one coherent system, we needed to receive sensor alerts as soon as significant movement is detected, pass those alerts to a central platform, automatically trigger the appropriate camera or cameras, capture and return images taken at, or just after, the time of movement, tie each event and image to a specific location and device and present everything in a clear dashboard for engineers and decision‑makers.”

To achieve this, the development team wrote a substantial amount of bespoke software, including scripts to interpret sensor messages, decision logic to select and trigger cameras, and API integrations to push and pull data between field equipment and the cloud. In addition, they built an interface that offers a live site overview, a list of alerts, time‑series charts and image comparison tools.

Without this software, connecting up the raw hardware would typically mean configuring cameras to take images every 15 or 30 minutes. Dywidag’s system goes much further by tying camera triggers directly to detected events and packaging everything into a system built around the realities of rail earthworks management.

Three technical characteristics underpin the system:

  1. robust communications at 868MHz: the sensors operate at this long‑range radio frequency because it is well suited to rail infrastructure environments.
    It penetrates soil and water better than many higher frequency bands. In field tests, sensors buried about 1m underground or exposed to water have maintained their communications link, an important factor on rail sites where devices may be surrounded by vegetation, saturated ground or snow. Some other systems use 2.4GHz, the same band as Wi‑Fi and microwave ovens, which is more easily absorbed by water and can be less reliable in poor conditions.
  2. very fast response times: under good conditions, tilt sensors typically detect and report movement within around three to five seconds using standard mobile networks. Because images are relatively large files, they take slightly longer to transmit than the initial alert, but in practice decision‑makers typically receive both a warning and supporting pictures within about a minute of the event. “That is effectively real-time, and fast enough to support decisions on stopping trains, closing lines or sending people to site,” Jones says.
  3. extremely high sensor sensitivity: the tilt sensors used by the system can detect very small changes in angle - on the order of 0.001°. In practice, even a light tap can register. However, not every small movement matters, so the system is set up to start with a baseline phase. After installation, sensors run for around a week with alerts turned off. During this time, the system observes how the slope moves naturally as temperature and conditions fluctuate.

Engineers then set alert thresholds based on that behaviour, usually in several bands, for example 5, 10 and 15° of tilt, to separate early warning signs from more serious movements. The result is high sensitivity, tuned to real‑world conditions, helping avoid unnecessary alarms without missing important changes.

Field operation

Last winter, Dywidag deployed the system as a trial at the preserved Keighley & Worth Valley Railway in northern England. A slope was chosen that had known issues with movement. Eight tilt sensors were installed at the top and bottom of an embankment, along with two cameras, one focused on the slope and one looking into the valley. Three positional GNSS sensors were fitted to provide location data - two on the embankment and one on a bridge as a reference point, as well as an automatic total station (ATS) to measure prisms and targets for detailed surveying. This allowed the system to be tested alongside more traditional monitoring methods.

On screen, the interface presents a map‑style view: cameras as icons, sensors as points along the slope, the track clearly marked, and a status panel showing live device and alert information. Time‑series charts show how tilt measurements change over time, making trend analysis straightforward. Image tools support side‑by‑side comparisons, overlays and slider views between dates, making subtle changes in the landscape easier to spot. For asset engineers and project managers, this turns thousands of raw data points into an accessible picture of slope behaviour. Jones says the results of the trial were extremely encouraging and have already resulted in real-world deployment on the British national network (see panel below).

Looking ahead, Jones says that he expects the system to evolve in two main ways: simpler hardware in the field, and smarter use of data. On the hardware side, the aim is to consolidate logger boxes, reduce power consumption and cut the amount of equipment needed at each site. That should lower installation and maintenance costs and make large‑scale deployment more feasible.

On the data side, Dywidag plans to integrate external sources such as national meteorological and rainfall records, correlating them with on‑site sensor readings. The goal is to shift from simple alerts, such as “the slope has moved”, towards more predictive, risk‑based insight, such as “given recent and forecast conditions, this site is at elevated risk”.

Discussions are currently underway for the deployment of Smart Guard at around 15 potential sites in British infrastructure manager Network Rail’s Anglia region, with more expected in the North West and Central areas. “Ultimately, our goal is to give power to the people that can make decisions while there is still time to keep trains and passengers safe,” Jones says.

Real-world use

RECENTLY the Smart Guard system has been deployed at two sites in Warwickshire, in central England, at Harbury and Chilton. Both are full installations, funded for five years. The site at Harbury involves a steep cutting on a busy route, originally envisaged as a cut‑and‑cover tunnel but left open. It has a history of major earthworks movement on one side, and nearby houses have seen gardens affected as the ground has shifted. A total of 85 tilt sensors have been installed along the embankment, with six cameras covering both the slope and the track, providing an approximately 180m-long monitored section. Solar panels and batteries provide power, with panels set in carefully cleared gaps in the tree line.

“The system hasn’t recorded a failure on the slope in the short time it has been operational so far, so we haven’t seen its full capabilities yet, but we’ve found the interface to be responsive and customisable,” says Matthew Wood, principal route engineer, geotechnical, Central route, at Network Rail.

He says the real-time data from Harbury is currently reviewed by the infrastructure manager on a weekly basis and a scheduled programme of manual checks means that the site will likely be inspected later this year to ensure that the system is functioning correctly. “These types of installations will become increasingly common across the British network, with systems provided by a number of suppliers, so I imagine competition will be intense,” Wood says.

“The system hasn’t recorded a failure on the slope in the short time it has been operational.”

Matthew Wood, principal route engineer, geotechnical, Central route, at Network Rail.

Chilton is another substantial embankment, where around 80 sensors have been installed, supported by three cameras. Fewer are required than at Harbury, thanks to clearer sightlines and fewer trees. Sensors are concentrated at the bottom of the embankment to give a monitored length of around 180m. In addition, boreholes have been installed at mid‑slope and upper‑slope levels to deepen the dataset.

In both cases data is fed through to both the Tier 1 contractors who maintain the sites on behalf of Network Rail, as well as directly to Network Rail asset engineers. Jones says that the longer‑term ambition for these and all other sites, is for the system to be embedded directly in the operational environment, ideally in signalling control centres, something he describes as the “gold standard.”

This would allow signallers to be immediately notified of alerts on a 24/7 basis giving the fastest possible response times. Jones notes that achieving this will require organisational as well as technical change. At present different Network Rail business units own earthworks, and are responsible for monitoring and operations, but the transition to the planned Great British Railways organisation, which is designed to bring train and infrastructure operations closer together, could be the ideal opportunity for systems like Smart Guard to be integrated into the operational railway.