OLErt

BACKGROUND TO THE PROJECT: An electrified railway requires a pantograph on top of the train to connect with the overhead line via carbon strips attached to the pantograph. An uplift force is applied by the pantograph to maintain the contact between the carbon strip and the overhead line. These carbon strips wear due to the motion of the train and so to maximise their life the overhead line is set up with a horizontal “zig-zag” (stagger) to distribute the wear across the full width of the carbon strip. Additionally, variations in the uplift force between the wire and the carbon strip lead to local notches. Excessive stagger or notches can cause the contact point between the wire and the pantograph to come off the end of the carbon strip resulting in pantograph flips and in extreme cases a dewirement. To monitor and maintain the equipment, measurement of the interface between the overhead line and the pantograph is required and set out in in British Standard 50317:2012. The key challenge for OLErt is to provide a system that CONTINUOUSLY monitors the condition of the OLE and pantograph and IMMEDIATELY alerts the relevant rail team when either a catastrophic failure occurs, or an emerging problem has exceeded parameter limits. A further challenge is to deliver the system using existing on-train sensors and cameras so that costs are kept low for both installation and maintenance. KEY TECHNICAL CHALLENGES: The milestones described in question 4 outline the key activities required to deliver OLErt. A number of technical challenges and brand-new innovations will be brought to bear to deliver the OLErt system. These are 1. To re-engineer the existing algorithm so that real-time image processing is achieved. The current algorithm was designed and developed through an R&D project using Matlab academic coding techniques. This has resulted in a processing time ratio of approximately 10 x real-time to produce the output. In order to improve the processing speed of the algorithm, we will build on the previously completed R&D to further evaluate the technical design of the algorithm and re-write the Matlab code using industrial coding techniques and standards. Further research into hardware upgrades will also be investigated if required. 2. To enhance the existing algorithm to detect dewirements, pantograph flips and close-to-failure events IMMEDIATELY so that operational teams can take emergency preventative and damage-limitation actions. To achieve this the algorithm will be enhanced to instantly analyse each frame of video in real-time, by comparing it against tolerance limits and provide data to a centralised system when a tolerance limit has been exceeded so that a failure or emerging alert can be generated. The existing algorithm has partly achieved this, but the volume of false positives must be reduced to make the algorithm fit for purpose. OLErt could also be integrated to auto pan-lower or train stop systems in a next phase. 3. To augment the processed video data-stream with accurate video positional data to allow the trending engine to use accurate positional references for run-on-run pattern matching analysis. The video positional technology has been developed using forward-facing rail camera algorithms. These algorithms will require to be adapted to provide positional referencing output using pantograph cameras and then merged with the vid data. 4. To integrate the algorithm on to existing train hardware. Icomera currently provides onboard communications systems, passenger wifi and sensor aggregation hardware across two thirds of the UK rail fleet. The image processing algorithm will be integrated on to this existing hardware and testing will be conducted to ensure the processing power is sufficient for real-time image processing. Following successful testing, the existing Icomera data transmission process established during the PoC project will be used to transmit the data from the train to a centralised cloud-hosted database provided by Incremental 5. To develop a run-on-run trending algorithm which uses machine learning and data collected during previous R&D studies, to detect and alert on emerging problems so that preventative maintenance can be undertaken by the relevant operational rail teams before a dewirement or failure occurs. 6. To develop joint TOC and NR procedures to respond to alerts and responses provided by the OLErt system. The procedures for response to static intelligent infrastructure are already well developed. These will be expanded to include the results from OLErt, which will be an early use of train borne intelligence. The routing will allow operational critical decisions to be made – diverting or stopping services as necessary. Most notifications will be of indicative change and therefore align with risk-based maintenance requirements. These will route via control to the maintenance scheduling process – allowing a timely adjustment or replacement of OLE or train equipment to mitigate risk of system outage from more serious failures if the fault is left unattended or unobserved.