05.03.19
Predicting system failures before they occur
Source: RTM Feb/March 2019
Professor Lukumon Oyedele is assistant vice-chancellor and chair professor of enterprise and project management at the University of the West of England (UWE Bristol). He also runs its Big Data Enterprise and Artificial Intelligence Laboratory (Big-DEAL). Here, he explains i-RAMP, a potentially revolutionary approach that could help detect faults on a railway line – before they happen.
One of Network Rail’s key objectives is to improve service – and safety – for users and stakeholders. But it knows only too well that to achieve this, it must reduce the occurrence of system failures.
A lot can go wrong: faulty signal boxes, broken tracks, even a defective escalator can add unwanted minutes to travel times and, in some cases, affect passenger safety. The knock-on effect is hundreds of work hours lost due to train delays, which ultimately hampers UK productivity.
Now, with HS2 on the horizon, it seems more important than ever to ensure these system failures occur as little as possible or not at all. One way of doing this could be to predict them before they happen – and this is where we come in.
Within our Big Data and AI lab, we are working on a research project called i-RAMP (IoT-enabled Platform for Rail Assets Monitoring and Predictive Maintenance), which harnesses the power of big data and artificial intelligence to help detect and predict faults in advance. It also uses augmented reality (AR) to assist engineers with repairs.
The project is a collaboration with smart engineering solutions company Costain and engineering technology start-up Enable My Team, which is the project lead.
i-RAMP will work as follows: a network of Internet of Things (IoT) sensors will first be installed on tracks and on station facilities like ventilation systems, barriers, or lighting.
IoT sensors can transmit a whole variety of data including vibration, strain, pressure on a structure, humidity, or temperature. Using several such components will enable train companies and station managers to monitor many parts of a train network at the same time.
Once it has gathered these parameters, this method will use AI techniques to analyse the data and predict when a fault is likely to occur, highlighting any stress points or component failures on a 3D virtual model of the station and tracks.
This approach will also allow engineers to use AR-based tools that offer them information about the location of faulty components and provides guidance on how to fix them. As well as orienting them to the exact place where the problem lies, it will supply them with real-time instructions and warn of potential dangers when they carry out the repairs.
For example, by wearing a headset or using their mobile phones, engineers will be able to view these instructions superimposed on the joint or electrical circuit that they are repairing or replacing.
It therefore might give information or warnings about the presence of high voltage in a section of a control panel, or instructions on how to disassemble an electric circuit in a signal box safely.
The research project finishes in the second quarter of 2020, with large-scale testing planned for 2021. We are currently in discussion with London Bridge station and others for the use of their rail assets for trialling the solution.
Currently, rail companies fix a problem once it has already occurred. The repair is a reactive approach and while they are fixing the problem, everything stops.
This system hopes to change that by enabling companies to fix an issue before it even becomes one, and do this in a timeframe when commuting is not disrupted.
Enjoying RTM? Subscribe here to receive our weekly news updates or click here to receive a copy of the magazine!