Rail Engineer has reported on the use of data analytics before (issues 136 and 148, February 2016 and February 2017) following attendance at Rolling Stock Maintenance conferences organised by London Business Conferences.
The latest conference on the same topic was held recently in a new venue in west London that enjoyed a lot more space. So what has changed over the last few years?
Perhaps it is best to start with a recap.
Despite the general title, this series of conferences has tended to emphasise the use of data to improve rolling stock performance and maintenance efficiency. It focuses on a number of approaches to maintenance that rely on data: predictive maintenance, where sensors send data to help analysts predict when maintenance is required; Reliability Centred Maintenance – a process of analysis to determine the optimum maintenance approach for each component; and Remote Condition Monitoring, leading to Condition Based Maintenance (CBM) where maintenance is carried out with knowledge of condition.
The recent conference included 25 presentations, a panel session (in which your writer unexpectedly participated) and a small exhibition including Perpetuum, Trimble, Humaware, Lucchini, NEM Solutions, Nomad Digital, RSSB, Siemens, SKF, UBIMET, and Unipart Rail.
What follows is a selection of the highlights.
Inevitably, in a conference with so many papers, there was a certain repetition, albeit with differences of emphasis. The overall objectives were probably summed up best by Lee Braybrooke of Trimble who highlighted that fleet managers should be seeking higher utilisation of their fleets, seeking to optimise the use of their resources – both people and facilities, monitoring leading indicators to identify failure points, reduce risk, minimise unscheduled maintenance, and deliver cost savings. He added that fleet managers should also be moving to condition based and predictive maintenance.
Mark Johnson, South East Trains’ engineering director and chair of the RDG Fleet Reliability Focus Forum, opened the conference. He highlighted the large number of new trains being introduced onto the railway and which are generally equipped with data systems that bring the opportunity to implement CBM. That was the good news.
However, he explained that many of the new trains are replacing comparatively modern trains that perform well, compared with the more usual situation where new trains replace those that are life expired. Fleet Focus has encouraged a general improvement in fleet performance and, over the last 10 years, the distance between technical incidents (MTIN) has increased from 5,000 miles to roundly 10,000 miles. The bad news was that, with the influx of new trains, Mark was forecasting a dip in MTIN results whilst the new trains “bed in”.
With that sobering thought, the conference moved to a case study from Paolo Masini from Trenitalia’s rolling stock technology and maintenance engineering department. His organisation’s aim is to get to the most appropriate maintenance for the component concerned. He illustrated this with the examples of doors.
There is no point maintaining doors on a distance-run basis as doors usually only work when trains are stationary, resulting in doors on non-stopping services being over-maintained compared with those on stopping services. If the maintenance planner knows the frequency of door operation, then maintenance can be organised based on that frequency, but if more information is available, such as speed of door operation, then maintenance might be delayed until performance starts to degrade, provided no prior safety checks are required.
Moving on to a practical example, Pedro Conceição from Nomad Digital highlighted work on a customer’s fleet over a period of nearly 15 years, which has seen over 70 per cent of maintenance tasks redesigned following comprehensive Reliability Centred Maintenance studies.
Some 50 per cent of tasks are still preventive, with three quarters of these based on condition or to find hidden failures (those failures that are only evident if something else fails). Around 45 per cent of components are allowed to run to failure and about five per cent of components needed re-engineering following the failure modes and effects analysis carried out as part of the Reliability Centred Maintenance process.
As a result, availability has grown from about 70 per cent to over 90 per cent and the failure rate has reduced by two thirds from about nine failures/million km operated to less than three. He did make the point that no condition-based or predictive maintenance programme can overcome fundamentally unreliable systems or components.
Justin Southcombe of Perpetuum has featured in the previous articles. The company makes vibration energy harvesting devices that power an accelerometer that transmits its data wirelessly to the train, enabling simple retrofitting. It is also maintenance free. It was originally sold to monitor the condition of bearings, but since then has been used to monitor wheel-tread condition and track condition.
Perpetuum has embarked on a £1 million project with the University of Southampton and TWI (The Welding Institute) to use tri-axial accelerometers which will exploit the natural frequencies of axles, excited by wheel/track impacts, to measure load and detect cracks with the prospect that, one day, routine non-destructive axle testing might be eliminated.
Justin highlighted a commercial innovation introduced through a partnership between Perpetuum and Shaeffler, the parent company of FAG bearings. Shaeffler is offering the opportunity for railways to pay a flat rate per month for the use of both Shaeffler’s bearings and Perpetuum monitoring devices.
The rationale for this offer is as follows. Rolling stock bearings do not, in general, wear out, indeed they are generally replaced at a defined point in order to ensure they do not wear out. The overwhelming majority of the population of these conservatively rated bearings might last a great deal longer than the nominal life assigned to them, whilst a minority might fail prematurely (often causing service disruption). The proposition is that monitored bearings that give ample warning of failure could be allowed to run for an extended service life. This is the basis of the Shaeffler offer.
In return for the monthly fee, Shaeffler will guarantee that a large percentage of the bearings, for example 90 per cent, will reach an extended target life, and will supply replacement bearings for fitment at the extended interval. The supplier provides fewer bearings over the life of the train and the train operator saves money by replacing them less frequently.
Bearings that fail prematurely will warn of impending failure and replacement will be carried out “funded” by the “savings” on routine replacement.
Shaeffler said that the savings for a legacy five-car train over 20 years amounted to €0.09 per train/km as, effectively, at least one bogie overhaul is eliminated.
In a different solution to the same bearing monitoring task, Maurizio Giomavelli presented the SKF Insight “totally wireless” sensor to monitor bearing condition. It is battery powered and uses a 2G/3G Internet of Things data service to transmit data “directly from sensor to cloud” and into SKF’s Bearing Application Knowledge Back End.
Maurizio made the point that all bearings solutions are designed, calculated and tested for a specific bogie application: bearing manufacturer engineers are involved in the whole OEM process. The notable difference between SKF’s sensor and Perpetuum’s is that the SKF product is somewhat smaller.
David McGorman, digital director of Unipart Rail and managing director of Instrumental, spoke about a world where the supply base is linked into the Remote Condition Monitoring system and is able to dispatch a spare part to a depot almost before the train has got there to have the defective component changed. Several speakers echoed the benefit of the entire supply chain having access to all the data for their respective components. This would allow much more useful data/information exchange than, for example, blue card labels with fault symptoms and paper reports of the repairs.
Philippe de Leharpe from SNCF spoke about the challenge of a large, but dispersed, fleet where any one depot only sees a few vehicles. Data systems add value by allowing managers of small fleets to measure the performance of their vehicles against the overall fleet performance. He also reported that they were gaining value from adding RCM technology to older vehicles. The Corail trailer coaches from the mid 1970s have had sensing and data transmission fitted to warn both train staff, via their mobile phones, and control room staff of issues with the train doors – open/close status, pressure of door closing cylinder and status of the doors closed lamp.
Gerald Schinagl of ÖBB, the Austrian state operator, introduced their process. He summed up his presentation in a few words, that “this is a journey, not a pre-determined destination; although it’s increasingly evident that it’s a worthwhile journey”.
He said that ÖBB had set up a team of 12 diverse people under the brand DIGI@ttack, reporting directly to the managing director. They were referred to as agile and holistic in activity with an evolutionary organisation and a challenging mindset. They had to deal with enormous expectations from people who expected instant results and who thought it was just a new tool.
Emphasising the importance of looking at the rolling stock and infrastructure holistically, Gerald warned his audience not to assume that, “just because one rolls on the other”, models or algorithms for rails will work for wheels.
He added that this is not an IT topic, but it is very technological; that supplier solutions might be limiting. He said it is a mindset and culture programme and, although a project is a good way to start, it needs to move into business as usual. Technology and Data Science competences are located in DIGI@ttack but are organised as a joint team with ÖBB’s Engineering Technical Services team.
The aim is to keep know-how inside the company and respond to day-to-day challenges, including the deployment of “minimum viable products” to address those challenges. Gerald particularly emphasised the importance of providing management with information about the impact on job roles three to five years ahead, for example so that people are not recruited for roles no longer needed!
He also spoke about the use of “sexy” devices such as virtual or assisted reality as a good ice breaker when introducing new ways of working.
Something completely different
Stefan Eisenbach of UBIMET, a meteorology company specialising in high-resolution and very local weather models and severe weather alerts, introduced how local weather forecasts, specific to an area or route, could be used to optimise operation and maintenance.
Creating specialist weather forecasts for railways started in 2005 when the Austrian Federal Railways asked UBIMET to implement a nationwide meteorological monitoring and warning system in order to face the challenges of climate change and more frequent extreme weather events. He said that general weather forecasts, which provide information about areas, are unsuitable for railways as they require information about routes and, particularly, local features such as cool valleys or cuttings or exposed embankments or bridges.
UBIMET has developed algorithms to provide precise weather forecasts along all railway lines, severe weather warnings for the network and critical assets and special risk predictions for a safe operation (icing risk, trees on tracks).
The objective of this is to reduce safety risk whilst improving network availability together with improved efficiency and reduced cost.
Nguyen Manh Phuc from Leo Express described some of the challenges affecting a small open access operator. The company operates three million train kilometres per year using a fleet of five new five-car Stadler Flirt EMUs in the Czech Republic and Poland and second-hand locomotives and carriages in Germany.
The five Flirt units, which are operated intensively, were the basis of his case study. He said that, during their first two years, the company relied on the manufacturer for maintenance, but thereafter recruited its own team, reducing costs by over 40 per cent.
A further cost saving is made by buying spares directly from the original producer rather than from the train manufacturer, although Stadler still overhauls the bogies. This needs careful planning – the maintenance frequency is every one million kilometres, which for Leo Express is every two years. The company has thirty bogies (six on each train due to its articulated design) plus three spares. Three bogies are dismounted and swapped for the spares, then sent away for overhaul – half train by half train. The overhaul takes two weeks and the whole programme takes six months.
Equally, Leo Express were proud of the work carried out in house to make the Flirt units suitable for operation in Poland, which included modifications to the safety systems, energy consumption meter and to lights. The first unit took one month to modify and the rest were completed over three months. It was reported that the supplier estimated three months to modify one train at much higher cost.
The final challenge reported was that Leo Express does not have its own depot and has to “fit in” others’ depots. Plans are in hand to build a dedicated depot in the Czech Republic near the Ostrava-Prague corridor.
Simon Jarrett, head of technical services for Chiltern Railways, talked about robotics in train maintenance, specifically cab-front cleaning and replenishment of fluids (Issue 159, January 2018). Since then, a one-eighth scale model has been completed by Cranfield University and partners Garrandale Rail and this was being demonstrated. Simon said that the motivation for developing robotic solutions is to reduce the number of people that have to work trackside around moving trains and that it is increasingly difficult to recruit staff to carry out these comparatively low-skilled jobs.
Reflecting after the conference, it was clear that speakers had generally recognised that Remote Condition Monitoring is largely a people and process issue, enabled by the IT. This is a significant change in the four years that Rail Engineer has covered this series of conferences, when one might have concluded that all one needed were sensors, analytics and “sexy front ends”.