ESR10 Project – Drive-by Detection of Railway Track Settlement and Bridge Damage

PARAIC QUIRKE – ESR10Profile Photo

Hello, my name is Paraic Quirke and I am a researcher with the SUP&R ITN project. I qualified with a degree in Civil Engineering in 2005 from University College Dublin, Ireland. I worked in the transportation infrastructure engineering industry for 7 years before taking on my current role as a researcher. After my studies I spent 4 years in Ireland working with RPS in road design and then 3 years in Australia working with Aurecon in airport design. I am now based in Dublin, dividing my time between UCD and Irish Rail as part of the SUPR&R ITN project.


Title: Drive-by detection of railway track settlement and bridge damage

Brief Summary: The project is concerned with the maintenance and sustainability of railway networks. Railway tracks settle over time which may result in safety issues. Bridges can also deteriorate due to corrosion or to impact from vehicles passing underneath. The purpose of this project is to develop and test methods whereby sensors on a train can be used to detect track settlement and bridge damage. In this project, the concept will be tested of using the existing trains as a means to monitor the track and bridges on an ongoing basis.

The project combines elements of Civil, Electrical and Mechanical Engineering sciences to develop ways to use trains in regular service to provide track and bridge condition monitoring. This has the potential to complement the asset condition information currently gathered periodically from track measuring vehicles and scheduled subjective bridge inspections thereby aiding the maintenance decision making process. The provision of daily feedback on asset condition can aid timely intervention of infrastructure issues with the benefit of reduced cost of rectification.

Tubber Bog - Settled Track - for SUPER ITN Website
Example of track settlement in Ireland

Another advantage of using in-service trains to measure track condition is that the dependence on specialised measurement vehicles, which are expensive to purchase, operate and maintain, is reduced. These specialised vehicles occupy the track, often with maximum operational speeds significantly below the line speed which can delay scheduled traffic.


Paper 1: OBrien, E.J., Bowe, C., Quirke, P., Cantero, D., 2016. ‘Determination of longitudinal profile of railway track using vehicle-based inertial readings’. Journal of Rail and Rapid Transit, DOI: 10.1177/0954409716664936.

A numerical study which tests

2D Vehicle model

the hypothesis that the dynamic response in a train induced by crossing a longitudinal track profile can potentially be used to determine the profile. Comparison of longitudinal profile over time can identify maintenance issues such as degrading track alignment, track settlement and hanging sleepers.

The inverse problem of finding the track profile from vehicle inertial response is solved through an optimisation technique using a least-squares fitting approach: i.e. find the profile that generates a vehicle response that best fits the measured response by minimising the square of the difference between the signals. The algorithm is successfully tested on three classes of track profiles with increasing levels of track irregularity and added sensor noise. A certain level of sensor noise can be tolerated before the inferred profile becomes unacceptably inaccurate. It is also found that the method is better at finding rougher track profiles as a result of the higher amplitude of vehicle acceleration produced by crossing these profiles.

Paper 2: Quirke, P., Cantero, D., OBrien, E.J., Bowe, C., 2016. ‘Drive-by detection of railway track stiffness variation using in-service vehicles’. Journal of Rail and Rapid Transit, DOI: 10.1177/0954409716634752.

A numerical study which investigates the hypothesis that variation in railway track stiffness induces a dynamic response in railway vehicles that can be used to find the variation. Railway track stiffness is an important track property which can also help with the identification of maintenance related problems such as track settlement and hanging sleepers.

The track is modelled as a beam on elastic foundation and variation in track stiffness is introduced by varying the stiffness of the individual supporting springs. The novel approach applied in this paper is to optimise the shape of a stiffness variation template in the form of a ‘normal’ distribution function to find patterns of stiffness variation along the track. The shape of the template can be changed by varying 3 parameters, significantly reducing the dimensionality of the optimisation problem. The algorithm is successfully tested with an input signal generated using measured Swedish track stiffness data.

Example of result from the track stiffness algorithm

Paper 3: Quirke, P., OBrien, E.J., Bowe, C., Cantero, D., Antolin, P., Goicolea, J.M. Article under Review. ‘Railway bridge damage detection using vehicle-based inertial readings and apparent profile’.

A bridge strike in Ireland

Irish Rail considers the striking of bridges by large vehicles to be ‘the single most likely cause of a serious rail safety incident on the network’. There were 85 bridge strikes on the Irish Rail network in 2015 alone. Bridge strikes, if unreported, may go undetected until the next periodic visual inspection which creates a potentially dangerous situation in the intervening months. Using in-service trains to regularly monitor bridges allows asset managers to detect damage caused by bridge strikes soon after occurrence. It also allows identification of trends indicating gradual bridge deterioration and can provide justification for prolonging the service life of a ‘healthy’ asset thereby reducing its life-cycle cost.

The presence of damage changes the modal properties of a bridge and hence alters its response to loading. The crossing vehicle acts as both exciter and receiver: it causes excitation in the bridge which then influences the response of the crossing vehicle. It is hypothesised that damage in a bridge will produce a change in the vehicle response which can be used to detect that damage.

3D Finite Element Bridge Model

An optimisation algorithm is used to find apparent longitudinal profiles i.e. the virtual profile producing a vehicle response that matches the measured response to the bridge crossing event. A 3D bridge vehicle model is used to generate the ‘measured’ vehicle response. It is found that impact events such as bridge strikes can be detected using the method.


Paper 4: Quirke, P., OBrien, E.J., Bowe, C., Cantero, D. Article being prepared. ‘Calibration of railway vehicle model using measured inertial response of in-service train for use in drive-by track monitoring system’

In December 2015 I organised theinstrumentation of the trailing bogie on a 22000 intercity rail car subsequently gathering 2 months of field data. The test equipment consisted of four accelerometers, two displacement sensors and an inertial measurement unit which were time-stamped and geo-tagged using GPS equipment. The instrumented train operated on the Belfast line for 6 weeks providing excellent test repeatability. The analysis of this field data is used to test the hypotheses proven using numerical models in the rest of the project.

A simplified 2-dimensional vehicle model is required to reduce the computational effort in the numerical analysis required to find longitudinal track profiles. This calibration exercise was required to determine the vehicle model parameters. A frequency domain decomposition technique is used to identify the dominant frequencies in the captured data. The data is chosen randomly from different sections of track thereby filtering out the effect of track profile and other noise. The remaining dominant peaks are taken to be vehicle frequencies.

Using prior knowledge as initial estimates for the vehicle mass and suspension stiffnesses, an optimisation technique is used to find vehicle mass and stiffness properties that best match the vehicle eigenfrequencies identified in the frequency analysis.

Identification of vehicle frequencies from measured data


Paper 5: OBrien, E.J., Quirke, P., Bowe, C., Cantero, D. Article being prepared. ‘Determination of equivalent longitudinal rail profile using measured inertial response of in-service railway vehicle’

Sample of Raw Data

Using the train response data measured as part of the field testing, the optimisation technique developed as part of Paper 1 is used to find a track longitudinal profile that generates a numerical inertial response best fitting the measured response. The vehicle model properties used are those found using the calibration exercise developed as part of Paper 4.

A section of track with a known settlement is selected as a case study. The elevation profile through the track section was measured using a level survey at the time of testing. Inferred profiles are compared to the surveyed profile to test the accuracy of the method.

Good agreement is found between the two profiles. There is also good repeatability, with the technique consistently finding the settled profile from all the datasets measured through the area. However, further work is required to improve the accuracy and repeatability of the method to conform to current European Standards.

From the numerical development and field verification of the methods presented in this project it is hoped that advances in sensor technology, wireless data transfer and improvements in computer processing power may be employed to apply the methods to a daily rail infrastructure monitoring system. Such a system would provide regular condition information to infrastructure managers enabling improvement in maintenance forecasting, improvements in safety and reduction in the life-cycle cost of assets.

Inferred track longitudinal profile from optimisation technique
Energy heat map – Using the energy in the measured signals used to find track faults

Conference Papers:

‘Determination of vertical alignment of track from accelerometer readings.’ OBrien, E.J., Bowe, C., Quirke, P. IMechE Stephenson Conference for Railways, London, 2015.

‘Drive-by structural health monitoring of railway bridges using train-mounted accelerometers.’ Bowe, C., Quirke, P., Cantero, D., OBrien, E.J. COMPDYN, Crete, 2015.

‘Drive-by inference of railway track longitudinal profile using accelerometer readings taken by in-service vehicles.’ OBrien, E.J., Bowe, C., Quirke, P., Cantero, D. CERI, Galway, 2016.

Follow me on twitter: @PQwitt

Find my profile on LinkedIn:


t: +353 87 251 4725