A Study on Automating Rolling-stock Maintenance in the Rail Industry

using Robotics

Randika K. W. Vithanage, Colin S. Harrison and Anjali K. M. DeSilva

Glasgow Caledonian University, Glasgow, U.K.

Keywords:

Industrial Robots, Automated Maintenance, Sensor Fusion.

Abstract:

Maintenance cost of United Kingdom’s rail rolling stock is a substantial portion of its whole life costs. There-

fore, it is vital to conduct these maintenance tasks in an efﬁcient and cost-effective manner to minimize op-

erational costs while maximizing safety, quality, and consistency of service. The introduction of robotics and

other intelligent mechanisms to maintenance processes would be an ideal solution to these challenges. Hence,

this research suggests introducing autonomous maintenance systems equipped with industrial robots to tasks

within the railway system, speciﬁcally for rolling-stock maintenance. The paper summarizes on-going and

future work of a case-study conduct in conjunction with a UK railway operator.

1 INTRODUCTION

1.1 Maintenance in General

Maintenance can be deﬁned as a task or series of tasks

which protect or reinstate the anticipated condition

of a system and these tasks include all technical, ad-

ministrative and managerial actions taken (M

arquez,

2007). Further, maintenance is categorized into three

major groups: preventive maintenance, scheduled

maintenance and unforeseen maintenance (Parker and

Draper, 1998). Proper maintenance of infrastructure,

machines, systems and other resources are essential

for any industry to provide a safer, reliable and re-

silient output (M

arquez, 2007). In general, mainte-

nance tasks are costly and on the other hand, mainte-

nance could be hazardous; being accountable for 25-

30% of fatalities in the United Kingdoms manufactur-

ing industry (Fraser, 2014)(HSE, 1999).

1.2 Involvement of Robots in

Maintenance

Robots that are designed for maintenance tasks can

be found in many different applications; especially

in the nuclear industry there are robots deployed for

inspection, scheduled maintenance, disaster manage-

ment and rescue operations (Pegman et al., 2006)(Na-

gatani et al., 2013)(Lee et al., 2013). Some of the

other applications are semi-automated highway main-

tenance tasks such as crack sealing, automatic warn-

ing corn dispensing, data collection and lamp post

maintenance (Lasky and Ravani, 2000)(Chan et al.,

2015)(Armada et al., 2005). Further, robots des-

ignated for maintenance can be seen in the elec-

tric power distribution sector for maintenance of live

wires (Kochan, 2001)(Maruyama, 2000), in facil-

ity management ﬁeld to clean glass-wall of high-

rise buildings (Tokhi et al., 2007)(Onori and Kochan,

2005), and in the railway industry to lay railway

tracks, rail grinding and ongoing research in robotic

train front-end cleaning (Villedieu and Francois,

1995)(Farnsworth and Tomiyama, 2014)(Moura and

Erden, 2017)(Tomiyama et al., 2017). Due to the

high levels of dexterity required, and a limited abil-

ity to cope with non-rigidly organized environments,

the majority of robots currently deployed in the main-

tenance sector are teleoperated, remotely controlled

or require close human supervision (Farnsworth and

Tomiyama, 2014). Further, due to the complex nature

of the maintenance tasks these robots need to be heav-

ily customized and specially designed for a particu-

lar task and therefore, lack the ﬂexibility and recon-

ﬁgurability common to industrial robots. Also, these

ﬁxed robotic and hard automation systems have large

initial investments compared to systems based on

standard industrial robots (Gupta and Arora, 2009).

278

Vithanage, R., Harrison, C. and DeSilva, A.

A Study on Automating Rolling-stock Maintenance in the Rail Industry using Robotics.

DOI: 10.5220/0006410702780283

In Proceedings of the 14th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2017) - Volume 2, pages 278-283

ISBN: Not Available

1.3 Application of Industrial Robots

Nationally and Internationally

Automated work-cells ﬁtted with industrial robots are

ubiquitous in the manufacturing sector, due largely

to the nature of task repetition. The total percentage

of robots installed in direct production-related tasks

amounts 78% of the total robots installed in the UK

see ﬁgure 1 (BARA, 2012). Though industrial robots

are commonly found in the manufacturing sector, op-

portunities available for introducing such systems into

maintenance are more challenging and this is mainly

caused due to lack of available technologies to cope

with complexities linked to maintenance tasks.

Figure 1: UK robot application by process.

According to the World Robotics 2014 data, an-

nual shipments of industrial robots have increased

over time despite the drastic drop in 2009 which was

due to world’s economic recession see ﬁgure 2 (IFR,

2015). Further, as shown in ﬁgure 2 it can be seen

that the same trend exists within UK (BARA, 2012).

Therefore, it is evident that industrial robots are be-

coming increasingly popular nationally and interna-

tionally. Over the period of 15 years from 1990, the

mean quality-adjusted unit price of industrial robots

in the UK, US and four other EU countries has

dropped down by nearly 80% - refer ﬁgure 3. There-

fore, robots are becoming more popular among indus-

trial practitioners perhaps due to their reduced prices

and improved quality, hence allowing shorter payback

periods.

1.4 Railway Industry in the United

Kingdom

According to the Ofﬁce of Rail and Road, in 2014 UK

shows growth of passenger rail journeys by 143% and

rail has carried 10% more freight compared to 2004.

Furthermore, in 2014, over 1.6bn passenger journeys

were made see ﬁgure 4 (ORR, 2014). Also, the UK

railway industry can be identiﬁed as one of the lead-

ing factors in the transport sector and the economy be-

ing responsible for 12bn spends a year (UKTI, 2014).

Therefore, UK rail is among the top most rapidly

growing industries. Due to rapid expansion and in-

creased competition, companies associated with the

Figure 2: Industrial robots over time (World Robotics and

BARA Robot Facts).

Figure 3: Price of industrial robots over time (IFR 2006).

railway industry progressively drive towards continu-

ous improvement of existing facilities, resources and

processes. That said, optimizing the usage, inspec-

tion, and maintenance of train ﬂeets in a cost-effective

manner remains a key challenge for the sector. Reduc-

ing inspection and maintenance costs, whilst main-

taining/improving safety is a priority for not just train

operating companies, but the rail industry both na-

tionally and internationally. According to the Value

for Money Study published in 2011 May, which was a

joint assignment by the Department for Transport and

the Ofﬁce of Rail Regulation Network, rolling stock

maintenance and ﬁnancing has been estimated to be

1.78bn a year which accounts for 15% of total UK

railway costs (McNulty, 2011).

Rolling stock maintenance could be clustered, dis-

tributed over local contractors and encourage them

to be specialists in the ﬁeld, and apply techniques

to achieve increased productivity through automation

(McNulty, 2011). Also, it has been identiﬁed that in-

creased automation is a key objective in the Rail Tech-

nical Strategy 2012 and is expected to obtain opera-

tional cost beneﬁts by intelligent maintenance tech-

A Study on Automating Rolling-stock Maintenance in the Rail Industry using Robotics

279

Figure 4: UK’S overall passenger railway journeys (ORR -

2014).

niques (TSGL, 2012). Further, ”Robotics and Au-

tonomous Systems (RAS)-2020” has identiﬁed that

UK could save 1tn over the following 20 years by

applying RAS to the transport sector (RAS, 2014).

Therefore, the introduction of autonomous systems in

order to eliminate or minimize human intervention in

maintenance processes would be an ideal technique

and it can be viewed that UK railway industry is in the

right stage to invest in automation (RRUKA, 2015).

1.5 Motivation and Scope of the Study

Based on above evidence it would be highly beneﬁcial

to introduce off the shelf industrial robots for rolling

stock maintenance. Therefore, this research seeks a

set of advanced skills and techniques to introduce in-

dustrial robots into autonomous maintenance appli-

cations in the railway industry through identiﬁcation

of viable maintenance tasks for automation. Further,

this research will focus on adapting automation tech-

niques in existing manufacturing to maintenance pro-

cesses and address the key challenges in-cooperated

with such automation exercises. The scope of this

project is to examine the feasibility of introducing

off the shelf industrial robots to automate ”Siemens

Class 380 Desiro” power bogie gear ﬂuid changing

task and develop advance sensing modules/algorithms

required for robot manipulation in maintenance en-

vironments. Shown in ﬁgure 5 is the main gearbox

of Siemens Class 380 illustrating wire lock, ﬁll/drain

plugs and inspection window, and the curved tooth

coupling cover of Class 380 power bogies. Techni-

cians have to drain the oil out from both main gearbox

and curved coupler and reﬁll them with new oil after

inspecting for the presence of water in oil or any metal

debris in the magnetic ﬁller caps. All the guidelines

for this process are provided in detailed maintenance

procedures provided by the manufacturer.

Figure 5: (a) Class 380 main gearbox where (1) Inspec-

tion window, (2) Drain plug, (3) Wirelocks and (4) Filler

plug, and (b) Curved coupling cover Class 380 where (5)

Drain/ﬁller plugs.

Figure 6: (a) and (b) Technician performing Class 380

curved coupler oil change.

Both these maintenance processes involve com-

plicated practices such as removal and application of

wire-locks in the main gearbox and locating the po-

sition of drain/ﬁller plugs of curved coupler since it

rotates as the train drives. These processes should be

relatively straightforward to an experienced technical

person but are quite challenging for an automated sys-

tem due to lack of positional repeatability. Further,

these items are located in difﬁcult to reach positions

(refer ﬁgure 6) where technicians need to execute the

bulk of the work related to these maintenance tasks

in unergonomic environments which may pose poten-

tial health and safety hazards, for example, oil spill

slip/trip risks. Therefore, the team has identiﬁed this

as one of the tasks to be explored further to introduce

automation.

2 WORKFLOW STRATEGY AND

EXPERIMENTAL SETUP

This study consists of several stages. At the initial

phase, as a part of the preliminary feasibility study,

data collection and computer-based simulations will

be conducted. The approach includes the develop-

ment of physically modelled off line tasks in a lab-

oratory setting to replicate the automation context.

A fully functional 6 axis articulated industrial robot

- Fanuc LR Mate 200iD will be used for the mock

tests to gain detailed knowledge of the task prior to

implementation and how difﬁculties can be overcome

ICINCO 2017 - 14th International Conference on Informatics in Control, Automation and Robotics

280

Figure 7: Fundamental workﬂow of the project.

with reduced ﬁnancial and technical risk. The funda-

mental workﬂow of this project is shown below and

visualized in green are the actions to be taken dur-

ing the phase 1 which are currently ongoing (ﬁg-

ure 7). As it can be seen project steps will take a

cyclic pattern to optimize the output to the desired

level. Work study data collection is currently under-

way and these data will be collected by observations

and video analysis of the actual maintenance tasks,

referring to maintenance instructions and communi-

cating with expertise in the ﬁeld such as technicians,

engineers, and managers at Abelio Scotrail, Shields

Depot, Glasgow. Data will be processed using stan-

dard work study methods: Method study and Work

measurement and compared with that by Farnsworth

2014 (Farnsworth and Tomiyama, 2014). Also, key

geometric details and critical dimensions of the parts

to model the experimental setups will be recorded in

parallel to work study data collection.

2.1 Prototype Design and

Computer-based Robot Simulations

Computer aided designs of scaled-down Class 380

curved gear coupler and end of arm tool were mod-

elled as shown in ﬁgure 8. It is been assured that the

Figure 8: (a) Proposed CAD of end of arm tooling where

(1) Floating tip to detect precise location of guide pin, (2)

4 X Micro load cells, (3) 4 X Wheatstone bridges, (4) 2 X

Arduino Nano, (5) ToF and RGB sensors and (6) Tip to in-

spect electric connectors, and (b) Fanuc Roboguide virtual

work cells designed for Class 380 curved coupler.

mock coupler contains key physical features required

to conduct the pilot test. The ﬁnal design is to be

conﬁrmed based on the input from Abelio-ScotRail

expertise, the output of computer-based simulations

and manufacturability. Rapid prototyping and CNC

milling will be utilized for the fabrication. Vision

and force sensing modules are incorporated to the

robots end of arm tooling (EOAT) and equipped with

a micro-controller to process machine learning algo-

rithms expected to develop in future. Further, it is

aimed to enhance tools reconﬁgurability with auto-

matic tool changing techniques. The software pack-

age ROBOGUIDE provided by Fanuc has been used

to conduct computer-based simulation of the process

and experimental setup. Roboguide is an advanced

simulation which allows users to model robotic work-

cells, process veriﬁcation, ofﬂine code generation and

robot motion conﬁrmation such as collision detec-

tions, program veriﬁcation, cycle time and payload

validation. Further, this software permits users to

program robots using a ﬂexible programming lan-

guage similar to Pascal (KAREL) other than tradi-

tional teach pendant programming.

Illustrated in ﬁgure 9 is a conceptual work-cell of

the train gear ﬂuid changing platform designed us-

ing Roboguide software for a preliminary feasibility

study. The virtual work-cell would assist the team

members to visualize robot positioning inside the ex-

isting service bay, design and allocation of multiple

end of arm tooling (EOAT), set up of robotic vision

systems, placement of force-torque sensing equip-

ment, oil discharging and dispensing methods and

other factors such as health and safety requirements.

2.2 Proposed Sensing Model

In order to cope with complexities and moderate lev-

els of disorder commonly found in maintenance envi-

ronments, the robot should be able to sense its envi-

ronment. Therefore, this research suggests a combi-

A Study on Automating Rolling-stock Maintenance in the Rail Industry using Robotics

281

Figure 9: Preliminary work-cell designed by Roboguide

Class 380 gear oil change platform.

nation of both vision and haptic data processing tech-

niques to bring up a smart sensing methodology. A

time of ﬂight and a RGB sensor is proposed to cap-

ture depth and intensity data of the environment that

will be used for the initial positioning of the robot.

The precise manipulation of the robot is supposed to

achieve through force sensing module equipped with

four micro load-cells coupled to Wheatstone bridges

- see ﬁgure 8(a). Moreover, both the main gearbox

and curved coupler are located in dirty and difﬁcult to

access positions, and it is quite difﬁcult to precisely

control parking position of the train itself. Therefore,

proposed system should be robust to noise and accom-

modate the high level of anticipated tolerances.

3 RESULTS

Object identiﬁcation and localization methodologies

are developed at this stage of the research by fusing

ToF and RGB sensors. Conventional edge detection

techniques didnt prove to be successful in detecting

objects due to the high level of environmental noises

presented, geometric ambiguities and surface charac-

teristics of the targets. Therefore, a template matching

algorithm which encapsulates geometric and intensity

data is developed to detect electrical pin candidates of

the automatic train coupler. The developed method-

ology is robust to rotation, ambient light and surface

ambiguities, and able to detect all electrical connec-

tors effectively within the anticipated work envelope

- see ﬁgure 10. Further, detection and localization of

the drain/ﬁller plugs of the curved coupler and main

gearbox is equally challenging and advanced machine

learning algorithms are used to detect and localize the

said targets refer ﬁgure 11. Moreover, the images

used in validation phase are captured in normal work-

ing conditions without the aid of any artiﬁcial lights

or ﬁltered backgrounds.

Figure 10: Successfully identiﬁed and localized electrical

connector of the Siemens Class 380 Scharfenberg train cou-

pler.

Figure 11: Identiﬁed drain/ﬁller plugs of curved coupler (1

row), identiﬁed ﬁller plug of the main gearbox (2

row)

and identiﬁed drain plug of the main gearbox (3

row).

4 CONCLUSION

Not to overlook the rapid progress made by recent in-

dustrial robots and sensory equipment, the majority

of present day robots in the manufacturing industry

manipulate in rigidly organized worlds. In most in-

stances these highly organized environments are cre-

ated by the precise position of the parts, implemen-

tation of jigs and ﬁxtures to guide tools, creating

clear, clean and bright environments. The introduc-

tion of such techniques entirely to the maintenance

sector is not straightforward and this could be viewed

as one of the primary hindrances to implement fully

autonomous systems equipped with industrial robots

in maintenance tasks. Therefore, in a wider scope,

this research pursues a set of advanced sensor (RGB,

depth, and force) fusion techniques to create a hy-

brid sensing model that could be coupled with indus-

ICINCO 2017 - 14th International Conference on Informatics in Control, Automation and Robotics

282

trial robots enabling them to cope with low-moderate

levels of disorders commonly found in maintenance

sectors. More importantly these techniques are lim-

ited to maintenance environments but certainly can be

adapted to manufacturing industry as well.

REFERENCES

Armada, M. A., Tavakoli, M., Zakerzadeh, M., Vossoughi,

G., and Bagheri, S. (2005). A hybrid pole climbing

and manipulating robot with minimum dofs for con-

struction and service applications. Industrial Robot:

An International Journal, 32(2):171–178.

BARA (2012). Robot facts 2012. Technical report, British

automation and Robotic Association.

Chan, B., Guan, H., Jo, J., and Blumenstein, M. (2015). To-

wards uav-based bridge inspection systems: A review

and an application perspective. Structural Monitoring

and Maintenance, 2(3):283–300.

Farnsworth, M. and Tomiyama, T. (2014). Capturing, clas-

siﬁcation and concept generation for automated main-

tenance tasks. CIRP Annals-Manufacturing Technol-

ogy, 63(1):149–152.

Fraser, K. (2014). Facilities management: the strategic se-

lection of a maintenance system. Journal of Facilities

Management, 12(1):18–37.

Gupta, A. and Arora, S. (2009). Industrial automation and

robotics. Laxmi Publications.

HSE (1999). Hazards during maintenance. http://

www.hse.gov.uk/safemaintenance/. (Accessed on

05/28/2017).

IFR (2015). World robotics 2014. https://ifr.org/img/

uploads/Executive-Summary

WR-Industrial-Robots-

0161.pdf. (Accessed on 05/28/2017).

Kochan, A. (2001). Tele-operated maintenance of live

power lines. Industrial Robot: An International Jour-

nal, 28(4):316–319.

Lasky, T. A. and Ravani, B. (2000). Sensor-based path plan-

ning and motion control for a robotic system for road-

way crack sealing. IEEE Transactions on Control Sys-

tems Technology, 8(4):609–622.

Lee, W., Hirai, M., and Hirose, S. (2013). Gunryu iii: recon-

ﬁgurable magnetic wall-climbing robot for decom-

missioning of nuclear reactor. Advanced Robotics,

27(14):1099–1111.

arquez, A. C. (2007). The maintenance management

framework: models and methods for complex systems

maintenance. Springer Science & Business Media.

Maruyama, Y. (2000). Robotic applications for hot-line

maintenance. Industrial Robot: An International

Journal, 27(5):357–365.

McNulty, R. (2011). Report of the rail value for money

study. Technical report, Department for Transport and

the Ofﬁce of Rail Regulation.

Moura, J. and Erden, M. S. (2017). Formulation of a control

and path planning approach for a cab front cleaning

robot. Procedia CIRP, 59:67–71.

Nagatani, K., Kiribayashi, S., Okada, Y., Otake, K.,

Yoshida, K., Tadokoro, S., Nishimura, T., Yoshida,

T., Koyanagi, E., Fukushima, M., et al. (2013).

Emergency response to the nuclear accident at the

fukushima daiichi nuclear power plants using mobile

rescue robots. Journal of Field Robotics, 30(1):44–63.

Onori, M. and Kochan, A. (2005). Robot cleans glass roof

of louvre pyramid. Industrial Robot: An International

Journal, 32(5):380–382.

ORR (2014). Rail value for money study — ofﬁce of

rail and road. http://www.orr.gov.uk/rail/publications/

reports/rail-value-for-money-study. (Accessed on

05/28/2017).

Parker, L. E. and Draper, J. V. (1998). Robotics applica-

tions in maintenance and repair. Handbook of indus-

trial robotics, pages 1023–1036.

Pegman, G., Luk, B., Liu, K., Collie, A., Cooke, D., and

Chen, S. (2006). Tele-operated climbing and mobile

service robots for remote inspection and maintenance

in nuclear industry. Industrial Robot: An Interna-

tional Journal, 33(3):194–204.

RAS (2014). Robotics and autonomous systems - ras 2020.

Technical report, The Knowledge Transfer Network.

RRUKA (2015). Robotics and autonomous systems

for rolling stock maintenance dissemination event

— rail research uk association (rruka). http://

www.rruka.org.uk/events/robotics-2/. (Accessed on

05/28/2017).

Tokhi, O., Zhang, H., Zhang, J., Wang, W., Liu, R., and

Zong, G. (2007). A series of pneumatic glass-wall

cleaning robots for high-rise buildings. Industrial

Robot: An International Journal, 34(2):150–160.

Tomiyama, T., Garc

ıa, L. R., Kr

slin, A., and Taykaldira-

nian, G. (2017). Systems and conceptual design of a

train cab front cleaning robot. Procedia CIRP, 59:61–

66.

TSGL (2012). Rail technical strategy 2012. https://

www.rssb.co.uk/library/future%20railway/innovation-

in- rail- rail- technical- strategy-2012.pdf. (Accessed

on 05/28/2017).

UKTI (2014). The uk rail industry: A showcase of excel-

lence. Technical report, UK Trade & Investment.

Villedieu, E. and Francois, D. (1995). Vehicle for the auto-

matic laying of a track by a vehicle travelling on said

track and track designed for installation by such a ve-

hicle. US Patent 5,385,102.

A Study on Automating Rolling-stock Maintenance in the Rail Industry using Robotics

283