August 16, 2017

Optical fibre networks facilitate shift to predictive maintenance

Written by  Shun-Yee Liu, Hwa-Yaw Tam, and Kang-Kuen Lee
  • Print
  • Email

Railways are using optical fibre sensing networks to switch from scheduled to condition-based and predictive maintenance, explains Shun-Yee Liu, Hwa-Yaw Tam, and Kang-Kuen Lee from Hong Kong Polytechnic University’s electrical engineering department, and Kei-Chun Cheng from Volant Railway Technologies, Hong Kong.

RISING expectations from both passengers and freight customers have been imposing tremendous pressure on railway operators to enhance service reliability by improving both rolling stock and infrastructure maintenance in order to reduce disruption arising from train and track failures. There is also a trend in the railway industry to move away from inefficient and costly traditional scheduled maintenance regimes to predictive and condition-based maintenance.

NS IC emu Keith FenderA prerequisite for adopting condition-based and predictive maintenance for railway assets is the availability of condition monitoring systems that can effectively and continuously monitor mission-critical components. These systems produce big data in respect of railway asset maintenance, which aids the development of advanced fault identification and prediction techniques. Thus, the development of railway condition monitoring systems has recently become a hot topic in the industry.

Fibre-optics have experienced enormous advances over the last 20 years in different applications including telecommunications and fibre-optic sensing. Fundamentally, fibre-optic sensing systems incorporate optical-fibre-based sensors that are embedded in the optical fibres which also transmit sensor signals. An all-optical sensing network has many advantages as it is small and lightweight, non-conductive, immune from electromagnetic interference, corrosion free, has a long transmission distance, produces little noise, and has high signal fidelity. This makes it an excellent platform for the development of railway condition monitoring systems.

Optical fibre sensors can be transformed into different types of transducers for measuring many different parameters including strain, temperature, pressure, displacement and acceleration. The various transducers can also be multiplexed and interrogated by one single interrogation system using optical fibres. Furthermore, electric power is not required to operate the optical fibre sensors. This simplifies the sensing network substantially and eliminates the need to integrate large numbers of different conventional monitoring systems, thus improving system maintainability and data manipulation.

In addition, fibre-optic sensors can be installed in locations which are hostile or difficult to access in order to acquire condition data not readily available using conventional electrical sensors. The availability of new types of condition data enriches the railway condition database, thereby aiding the development of pristine fault detection capabilities and prognostic functions through advanced data analysis and deep learning.

Figure 1 shows a comprehensive railway condition monitoring system based on optical fibre sensing technology for both rolling stock and infrastructure condition monitoring. Such monitoring systems can oversee the condition of mission critical components on both rolling stock and infrastructure including train running gear, track, overhead line and third rail. Moreover, as the two sub-systems countercheck each other, this safeguards the proper functioning of the two sub-systems and ensures the data’s reliability.


An effective train condition monitoring system can improve train availability and service quality by providing operators with real-time train condition information so that they can act promptly in case of a fault. A track-based train condition monitoring system based on fibre-optic sensing technology is shown in Figure 2 (pXX). The monitoring system consists of track-mounted fibre-optic sensors that are connected to a sensor interrogation unit, a data analysis unit and a client workstation that interfaces with the operator. By interrogating wheel-rail interactions followed by advanced data processing and analysis, the monitoring system can address issues with mission critical train components and running gear such as underframe headstock, wheels, axle bearings and suspension systems.

In addition to functional aspects, the design of the train condition monitoring system has also taken account of system reliability and user perceptions. The front-end sensing system consists of fibre-optic sensors and optical fibre cables which are not only extremely reliable, but also maintenance free. In addition, there is no track side equipment which would be vulnerable to interference and maintenance issues.

AUG38fig1The track-mounted sensors are easy to install and multiplex with redundant sensors to improve system reliability at minimum cost. In the back office, the design of the man-machine interface at the client workstation is train operation oriented. The interface home page only provides key condition information of trains in service and uses a simple alert, alarm, three-level warning system to draw the operator’s attention to faults. This design avoids distracting the operator with an abundance of information that is not crucial to train operation.

Apart from real-time monitoring functions that can support the daily operation of the railway and condition-based maintenance, a continuous monitoring system can also acquire massive amounts of train condition data to provide indispensable information for prognostic maintenance. For example, by making use of wheel condition deterioration rates and past maintenance activities, the system can project the best time for wheel turning in order to maintain wheel condition. By contrast, traditional inspection conducted in the field or in a workshop can only provide sparse and discrete condition data which is insufficient to realise health projection and prognostic maintenance.

The train condition monitoring system can also detect train structure problems that could cause a serious accident. For instance the train condition monitoring system is able to pick up structural defects such as broken headstock screws or a crack on the underframe. These incidents produce abnormal vibrations, which were identified by the monitoring system.

Infrastructure inspection and maintenance are becoming more challenging for railways as their work load has increased significantly with the global industry boom and introduction of many new lines within a short time span.

A major challenge is that as the number of lines and track length increases, conventional methods which rely on the very limited non-traffic hours for infrastructure inspection and maintenance are becoming inadequate. Besides, increasing traffic can cause the infrastructure to deteriorate more quickly, resulting in more unattended breakages and service disruptions.

Consequently, the trend is to use in-service trains as inspection vehicles to carry out infrastructure inspection during traffic hours leaving the non-traffic hours for maintenance. Moreover, in-service inspection trains can interrogate an entire line for many parameters and can provide up-to-date infrastructure condition information to the train operator to avoid unexpected breakdowns and service outages. By contrast, conventional inspection methods using engineering trains are less efficient as they only focus on one aspect at a time.


A train-borne infrastructure condition monitoring system based on optical fibre sensing technology for installation on in service trains can incorporate many different types of sensors to interrogate the condition of track, overhead line and third rail. For example, track issues such as rail corrugations can be revealed by the system. By applying different data processing and filtering techniques, other track issues such as rail cracks and dip welds can also be identified.

AUG38fig2The intrinsically non-conductive and electromagnetic interference-immune properties of optical fibre sensors make it possible to install sensors on pantographs and conductor shoes. By monitoring key parameters such as contact position, contact force and displacement, it is possible to identify issues with the catenary and third rail such as stagger registration shift, wire sag, hard spots, section insulator misalignment and loss of contact.

The performance of the train-borne overhead line condition monitoring system has been tested in-depth in the Netherlands and Hong Kong, and shown good performance in detecting OCS abnormalities such as sagging and excessive wearing of contact wire.

Hong Kong Polytechnic University’s department of electrical engineering has been developing optical fibre sensing networks for condition-based maintenance of railway systems for over a decade. The department has been applying the sensing technology in many field trials, including on high-speed trains in China, Taiwan, and India. In 2014, five track-based optical-fibre sensing networks were successfully installed on the MTR network in Hong Kong. Two track-based and two train-borne systems were designed in early 2016 and will be installed in Singapore by end of 2017, while the optical fibre sensing systems installed so far have proven to be highly reliable and require minimal maintenance.

Get the latest rail news

IRJ Rail Brief newsletter covers global railway news