Adhere & V/TSIC: monitoring and treatment

The collision at Fisherton Tunnel, Salisbury in October 2021 was primarily caused by poor wheel/rail adhesion.

That said, RAIB’s investigation pointed to numerous other factors arising from management and communication within and between organisations in the industry, as well as autumn leaf fall management and treatment. One issue was knowledge of rail head contamination. Part of the management process is for a Mobile Operations Manager (MOM) to go to locations where poor adhesion has been reported to assess the presence of contamination. However, RAIB’s report stated that MOMs had been prohibited from going on to the track to investigate reports of poor adhesion, unless a line block had been put in place in the interests of their safety. As a result, they were expected to observe railhead contamination from lineside fencing of bridges (Report 12/2023: Collision between passenger trains at Salisbury Tunnel Junction, paragraphs 237-240).

Network Rail often issues problem statements seeking innovators to propose solutions. For this challenge Network Rail said: “Every Autumn, Routes dispatch their MOMs to inspect the condition of the railhead in high-risk areas. This inspection requires the MOM to go lineside to determine the level of contamination at each site.

“The inspection of these sites is undertaken up to three times per week and totalled circa 1,400 on Wessex alone in 2024 (additional inspections were required following ROLA (reports of low adhesion) or Wrong Side Track Circuit Failure (WSTCF) incidents).

“This process is placing our staff lineside during some of the most challenging and dangerous times of year and is something the industry must look to reduce. Slips, trips, and falls account for a significant number of lost time injuries, and sending staff out during wet and windy conditions will inevitably lead to incidents.

“The inspection process itself may also only account for a small proportion of the overall high-risk site due to limited access and other restrictions. This may lead to dangerous conditions going undetected if not adequately managed.”

This was reinforced by RAIB’s Salisbury recommendations.

Machine vision

At the ADHERE seminar in March 2024, One Big Circle’s Emily Kent explained how her company’s machine vision system, AIVR (Automated Intelligent Video Review), is assisting Network Rail’s Network Services team with adhesion monitoring and treatment and helping to get ‘boots off the ground’.

Brian Whitney’s presentation to the Vehicle/Track System Interface Committee (V/T SIC) seminar also spoke highly of machine vision for identifying track features and defects.

Emily reported that following the October 2021 accident, Network Rail, SouthWestern Railway, and expert suppliers worked collaboratively to deploy a train-borne monitoring system that could be deployed for Autumn 2024 to monitor contamination on the railhead. The objective was to create an automated system on in-service vehicles that would provide images and location details to enable digital examination of contaminated areas. These data could be examined and responded to in near real time without the need for personnel to attend sites and make visual inspections from hazardous lineside positions.

Equipment was fitted to two SWR Class 158 trains and to one of Network Rail’s Multi Purpose Vehicles (MPV) assigned to railhead treatment duties. The installation included the AIVR Connect module (data acquisition and transmission), AIVR Tachometer, two cameras, and lighting. The cameras are capable of capturing high resolution images of the rails even at 90mph. A forward-facing-camera was also installed on some of the trains to provide context of the line. The data was automatically transmitted to One Big Circle’s AIVR ‘cloud’ and could be rapidly reviewed using the AIVR platform. External data sets, such as known high risk areas, were integrated with AIVR.

The trial has successfully demonstrated and validated areas of low adhesion on the railhead and provided location data for these areas through the AIVR platform. All the data captured could be filtered by location so that several months’ data could be compared quickly to look for signs of degradation and/or confirm the success of railhead treatment. The ability to integrate with other datasets has been demonstrated; the location of known high risk areas were pre-loaded and footage was only captured of these areas rather than the whole route. In addition, data from the trains’ wheel slide protection systems was interfaced with AIVR data. This allowed alerts to be generated to show users where wheel slip or slide has occurred and allow the user to be able to see both the railhead and the general area of the alert.

Further progress

Since the initial trial, Wessex route has equipped two more MPVs (a total now of three) with the AIVR hardware installed and the passenger train installations have been maintained. It was noted that the Class 158/159 currently have no forward-facing video, but these images are captured by other fleets. To date over 80,000 miles of data has been captured on the Wessex route alone. In addition, Transport for Wales has equipped three Class 153 units, including Forward Facing Video (FFV), and these are in service.

East Midlands Railway has one Class 158 unit in action and two Class 170s due to deploy shortly (including FFV). The ‘cloud’ storage means that all required users – operators and Network Rail – can access the same information and this is enabled by Network Rail’s enterprise licence.

Further change is being introduced under the headline ‘Smart Activation’. Firstly, one MPV with AIVR fitted front and rear is being used to monitor the effectiveness of rail head treatment, by assessing the rail head images from the front and rear cameras. This might improve the accuracy of low adhesion response and allow treatment to be varied or applied discontinuously depending on railhead condition.

Rail Engineer wonders whether the output of the University of Sheffield’s research into adhesion estimation using machine vision learning might have a role to play.

There is also potential to integrate the information from the FFV with the railhead view. For example, in high risk of low adhesion areas, other requirements could be configured including specific trees and plants. Over time this would increase knowledge of leaf fall of certain types of trees and increase knowledge of variables and impact.

An important factor in all this development has been lots of engagement and insight from expert end users who want to guide the development to meet their need rather than having something foisted on them. This helps deliver solutions that fit with operational processes, leading to improved safety and effectiveness. Examples shown during the seminar were identification of squats and rail welds.

Advancing remote monitoring

Whilst all this technology was being developed, onsite inspections by MOMs continued as outlined in the problem statement. This has provided a way of validating the information from AIVR and MOMs have also been involved learning how to use the data to make assessments without going trackside.

This work involving Wales, Wessex, and Eastern is being followed closely by other regions advancing their monitoring capability, with other routes and regions following progress closely. By providing these digital tools to those who need to monitor the critical lineside environment safely and remotely, ‘boots on ballast’ can be reduced.

Clearly, this system provides ‘eyes in the control room’ which will help staff managing autumn adhesion issues to provide much more timely advice to those who need to manage operations.

Image credit: RAIB