In this digital age, railways are harvesting increasing volumes of data (‘Big Data’) on the performance of their assets but converting raw data into meaningful information remains a challenge for many organisations in the industry.
One example of a successful foray into the processing of ‘Big Data’ is in the area of electrification overhead line contact system performance. The technology application is helped by the introduction of cheap and small sensors, declining computing and data storage costs, new abilities to process and analyse data and ubiquitous connectivity. The coming together of these key enablers is driving the adoption of the Internet of Things in rail and, with it, demands for analytic applications to turn data into operational intelligence.
Rail Engineer was invited to the new traction maintenance depot at Reading to meet the team that has developed the OLErt monitoring system and hear of the advantages it can bring to the railway.
Critical interface
Whilst considerable attention is paid to the interface between wheel and rail, there is another mechanical moving interface between a train and the infrastructure – that between the pantograph and the contact wire. Whilst a failure of the wheel/rail interface can result in a very serious safety incident, a failure to maintain the connection between train and power supply can be operationally disastrous. There are also safety connotations in the case of failure.
For some time, there have been various methods of monitoring the electrification overhead contact system relationship, and processes to avoid that relationship being out of correspondence have been developed.
One major risk emerges if the extremity of the pantograph (the ‘horn’) becomes separated from the lateral position of the contact wire while that wire is still close to the pantograph. In that case, the contact wire may then get underneath the pantograph horn and either severely damage the pantograph or the contact wire and its associated overhead line equipment (OLE) may be torn down – or both.
Traditionally, stagger (the deliberate deviation of the contact wire from a central position to spread wear on the pantograph contact surface) of the contact wire has been monitored by observation from a train, such as the test car ‘Mentor’. Instrumented processes have been used and the height and stagger of the wire may also be checked by patrolling and suitable equipment. Cameras have been mounted on train roofs for some time, but with little recovery of data by technology. More recent developments have involved bespoke equipment to monitor the contact system and the interface – this has not been a low cost or generic option.
The recent development of cameras being fitted to the roofs of modern electric trains has generated an opportunity to develop a more real time and wide-ranging monitoring process.
At present, the cameras provide useful output but, in reality, they only show the circumstances of an incident after it has occurred, such as the contact system/pantograph interface becoming non-compliant with unfortunate results. Using the data to improve performance required a system to identify potential problems and highlight them before an incident occurs.
These events have a high impact, with passenger and electric freight line closure as a minimum. They are also expensive, with each one costing Network Rail time and money for infrastructure repair and compensation to train operators. Despite receiving this compensation, the train operators would prefer better reliability and no disruption for passengers.
Understanding the data
So, data was required – Big Data in this modern world – to discover how the pantograph and contact wire interact and to detect any potential faults before they can cause trouble.
Network Rail is already receiving information from in-service trains on the condition of its track, and the wheel/rail interface, but it did not have a system for retrieving contact force data from the pantograph interface at over 100 mph.
Ideally, a pantograph will exert a constant upward force on the contact wire and a good maintenance regime will keep the carbon strip on the top of the pantograph – the point of contact between the train and the wire – in good condition.
With the Great Western electrification programme set to increase markedly both the number of trains running under the wires and the number of pantograph passes, the possibility of increased failures had to be taken seriously.
A consortium of interested parties came together to meet this challenge under the auspices of Network Rail. Software specialist Incremental, communications specialist Icomera, Great Western Railway (GWR) and the University of Oxford began a development project using a modern EMU and capturing data from the monitoring camera. The aim of this ‘OLErt’ project was to make the camera intelligent.
Team contributions
Incremental Solutions, a high-growth and relatively young company, has been exploiting the concept of ‘intelligent mobility’ in the rail industry to improve understanding of vehicle movements and to solve critical issues of performance, reliability and journey optimisation.
Its core aims are to assist in the creation of a connected transport network with a proactive approach to predicting and preventing, resulting in improved incident management, more accurate timetabling and, overall, operationally more efficient.
It does this by utilising best-in-class solutions, working with other innovators and liaising with academia. This creates opportunities for SMEs to scale up their ideas, promotes innovation and challenges the status quo.
Of the other contributors, the University of Oxford brought algorithms from its own research groups and Icomera contributed its expertise in wireless connectivity, while Network Rail and Great Western Railway integrated the projects activities. The group has now successfully completed the first phases of the project and, for that, has been voted ‘highly commended’ at the 2019 Rail Partnership Awards for Driving Efficiencies.
From the group came the inspiration to take the existing camera output from the train and to digitise it, producing a data stream that could continuously monitor the performance of the pantograph and its movement in line with the OLE contact wire. That data stream could then be analysed to extract changes in a normally consistent signature, so identifying potential incident sites.
In order to test the theory, Great Western Railway made available a class 387 electric multiple unit to which was fitted the pantograph monitoring camera. Following a calibration exercise to eliminate any lens distortion and ensure the pantograph was within height range, a network video recorder (NVR) was installed into the train roof-space. This was linked to the Ethernet backbone to enable a secure download of the recorded footage via Wi-Fi, removing risks associated with access to the train roof.
Processing the data from the camera, utilising algorithms from the Oxford University research, supplied the proof of concept for the project. It was noted that the data from the widely fitted cameras on the fleets was encrypted, so special arrangements had to be made for the OLErt-related measurement camera on the chosen class 387.
In the realm of fixed equipment, a Wi-Fi network was installed in the East yard of Reading Train Care Depot. This included three access points onto the noise abatement wall to cover the stabling locations of the unit – plus the necessary network ancillaries.
A number of test sites were chosen as examples of differing environments, including high-speed running (over 100mph), increased contact-wire height over level crossings, neutral sections, tangential wires crossed at speed and a complex layout.
Results
The series of tests, and monitoring and processing of the outputs, revealed interesting data and proved the concept of the proposals. As stated earlier, the potential is for a mass of data which would not be capable of analysis, but the results passed through the Oxford algorithms allowed the ‘signature’ of the OLE to be observed and potential problem spots identified. In fact, during the trials, a location with a serious risk of dewirement was picked up and attended to as a matter of urgency.
A conclusion emerged that pantograph stress is more complex than thought – during the trials 51 different signalled patterns were achieved in the five geographical areas. Pantograph force assessment did show that the average pantograph force was compliant with that expected at speed and in service. Interestingly, concerns over dirt on the camera lens, which might have obscured the image, were addressed by application of a special film ensuring that, after four weeks testing, the image obtained was still good.
In summary, the OLErt system has usefully built on existing data acquisition, itself a recent development, and applied advanced methodology to harvest large amounts of data. It has then used algorithms supplied by academia to turn that ‘Big Data’ into a format that will allow individual sites with potential problems to be identified so they can be investigated before a significant operational or safety incident occurs.
This research will definitely benefit all partners in the operation of the railway and help enhance a continuing good service to customers.
Thanks to the OLErt team for explaining their work and its concepts – Paul Barnes and Dean Shaw (Network Rail), Daniel Lee-Bursnall (Incremental), Stephen Duncan (Oxford University), Rich Fisher and David Eveleigh (Great Western Railway).
This article first appeared in Issue 177 of Rail Engineer, Aug/Sep 2019.
