Subscribe to our newsletter to receive the latest news and events from TWI:

Subscribe >
Skip to content

Advances in rail inspection technology (June 2005)

Colin R Bird, Principal Consultant NDT


Jonathan Bray, Rail Economist
Tube Lines

Presented at Rail Engineering 2005, University of Westminster, 29 June 2005.


The rail industry, the public and regulatory authorities all seek to see improvements in the delivery of a safe, reliable rail infrastructure at affordable cost. This paper describes inspection techniques and methods for the combination of data, which have been developed by TWI for the inspection of rail. The technology used in these inspection techniques has benefited from TWI's experience developed with Tube Lines and insight gained in other industrial sectors. This paper describes progress in two inspection technologies:-

  1. The application of ultrasonic phased array technology for train mounted wheel probe inspection to provide improved signal to noise and continuous compensation for the effects of rail wear on the performance of the inspection.
  2. The application of magnetic flux leakage for high speed inspection of rail heads.

This paper further discusses complementary inspection technologies and how, by careful combination of these inspection technologies, a more powerful and reliable inspection may be achieved when compared with simple analysis of the individual inspection results.


Industrial Need

Tube Lines are one of the Public Private Partners (PPP) of the London Underground system. Under the 30 year PPP Contract Tube Lines is responsible for the safe and efficient maintenance, renewal and upgrade of the Underground infrastructure, including track, trains, signals, civil works and stations. Tube Lines has direct responsibility for about 200 miles of track but is also responsible for providing some services to Metronet, the other PPP Contractor. These additional services include the ultrasonic testing of all rail on the Underground.

The Underground has a very good safety record, especially with respect to train derailments. Tube Lines is acutely aware however that a single derailment can cause very severe consequences. There are many inspection and quality control systems in place to minimise these risks and Tube Lines is constantly working to improve the effectiveness and efficiency of these processes. To minimise the risk of unplanned maintenance and failures Tube Lines is investing in improved rail inspection procedures and systems. This includes a thorough review of existing technology and procedures used together with the application of new technology and procedures to support this.

The current inspection processes are completely manual. There are a number of tiers of surveillance which range from frequent (can be daily) manual walks for visual examination of the rails to manual ultrasonic rail inspection every few months using hand held rail testing equipment. Tube Lines has an aspiration to supplement these inspections with mechanised testing which would be far more efficient in terms of the ground covered. Note that current work is undertaken during engineering hours when there are no trains running. Mechanisation could mean that works are undertaken during normal traffic hours. Such a mechanised approach could also be more comprehensive in the section of the rail inspected and more consistent/repeatable. Mechanisation could involve technology other than ultrasonics such as magnetic flux or guided waves.

Development work

This paper gives an overview of work carried out within the EC funded CRAFT Project Rail-Inspect, project number CRAF-1999-70907. The work was carried out by six SMEs supported by two RTDs. Partners are listed in the final section.

The Rail-Inspect project sought to explore and develop techniques with the objectives of:

  • Improving the probability of detection of significant defects while reducing the incidence of false positive indications ('false calls').
  • Increasing the extent of the rail head and web which can be tested, even if the rail head is severely worn.
  • Carrying out inspection at up to 80km/hr.

Recognising that no single NDT technique provides all the required information from the cross-section of a rail, the project investigated a number of techniques. The strategy was to develop the strengths of individual techniques, and then combine them into a system that provided the operator with a user friendly means of gaining maximum useful information on the state of the track. The combination of techniques was a key novel feature of the project. Three techniques were investigated, for the following reasons:

  • Ultrasonic, for volumetric inspection: rail head, web and foot.
  • Eddy current (EC), for sensitive rail head surface inspection.
  • Magnetic Flux leakage (MFL), which is less sensitive than eddy current but applicable to large surface breaking defects.

As with all railway equipment, it is important to ensure that there is no interference with signalling systems. This aspect requires further investigation.

The following sections summarise the results obtained from the different techniques and how the data was fused to provide a single report for the operator.

Development of the ultrasonic inspection system

One issue with ultrasonics is how to achieve a reliable way of coupling the ultrasound into the rail. Water coupled sliding probes are not mechanically reliable and so a wheel probe was developed by the consortium. The wheel consists of a flexible membrane which contains the transducers. The membrane supplied by Sonatest contains a liquid which couples the ultrasound to the track as the wheel is rolled over the track.

As part of the RAIL-INSPECT project, the use of phased array probes was investigated in order to assess its potential for simplifying and improving the ultrasonic inspection of rails. The intention was to exploit the ability of phased array systems to vary beam angle and beam focus by applying various focal laws under electronic control. The geometry of rail testing is complex, particularly when the effect of wear on the rail head is taken into account. ACAD-based computer model was used to assist with technique development. The modelling enabling the specification for the phased array probe and the wheel to be established along with the ultrasonic procedure.

Inspection modelling

Unlike many industrial components the railhead wears during its relatively short life and any inspection system must allow for this wear. Figure 1 below provides the basic section of the rail and a ray diagram showing a high beam spread probe.

Fig. 1. Wheel probe with 10mm conventional probe
Fig. 1. Wheel probe with 10mm conventional probe
 Fig. 2. Phased array probe showing skewed focused beam for gauge corner cracking detection
Fig. 2. Phased array probe showing skewed focused beam for gauge corner cracking detection

There are three basic methods of accommodating wear within an inspection procedure: an array of probes providing a spread of sound, broad beam spread probes to capture the defect even after beam bending and steered focused probes. As described in the introduction a wheel probe was chosen to deploy the ultrasonic probes. The tyres conform to the rail surface but this does not prevent variable deflection of the beam with rail wear. Figure 2 above shows the beam model being used to design a phased array focused beam for gauge corner cracking detection.

A model of the phased array wheel probe generating a 40° beam directed towards the bottom of the rail is shown in Figure 3. This beam was used for the detection of fatigue cracks from the bottom of the rail.

This model is an add-on to AutoCAD 2000 ® and provides ray tracing for phased array probes and the complex time delay calculations required to generate the focal laws.  

Fig. 3. 3D Model of ultrasound beam to the bottom of the rail
Fig. 3. 3D Model of ultrasound beam to the bottom of the rail
Fig. 4. Photograph of the wheel probe showing the three phased array probes
Fig. 4. Photograph of the wheel probe showing the three phased array probes

Figure 1 shows that without a focused probe signals are generated from the underside of the rail head as well as the bottom of the rail. Further without a focused probe it is difficult to generate a high beam angle which will detect gauge corner cracking without interference from surface waves. Surface wave generation is a significant problem with soft wheels where a small amount of water can build up between the tyre and the track. Figure 2 shows that with modelling it was possible to design and apply ultrasonic beams free from interfering geometric echoes or surface waves.

The model was used to provide focal laws, which compensated for wear on the rail head. The modelling which was confirmed by trials with the wheel probe showed that the ultrasonic beam was deflected by up to 5° due to normal rail wear.

Wheel probe design

A close up photograph of the wheel probe containing the three phased array probes is given in Figure 4.

The wheel consists of a flexible tyre containing three phased array transducers. The three transducers are orientated to produce three non steered beam angles: 0°, 40° and nominally 58°. These basic non steered beam angles are then modified by the electronic steering available in the phased array system to produce focused beams in a variety of locations within the rail section.

The tyre supplied by Sonatest contains a liquid which couples the ultrasonic transducers to the tyre material which in turn couples the ultrasound to the track. The tyre material was designed to be sufficiently rugged to withstand the impact and wear caused by the rail whilst maintaining relatively low acoustic attenuation. The chosen inspection frequency was 3.5MHz, which provided good beam focusing and defect resolution without suffering attenuation from the tyre material or coupling loss into the rail. The liquid was designed to match the acoustic impedance and velocity of the tyre material minimising wheel internal reflections and noise generation.

The phased array transducers were designed and manufactured by Imasonic. Use of phased array ultrasonic probes allowed a reduction in the required number of probes and, most importantly, allowed the system to focus and steer to the desired inspection locations. The 0° probe was used for detection of rail foot corrosion, horizontal splits and bolt hole cracks. The 40° probe for bolt hole and rail foot fatigue cracks and the 68° probe for rail head cracking. Of particular interest was the detection of gauge corner defects that are both tilted and skewed with respect to the rail axis. For ideal defect detection and sizing conditions a skewed, focused 70° ultrasonic beam was required. Further the rail head can also contain squat defects in the centre portion of the rail head which require high beam angles for ideal defect detection. By steering and focusing the nominal 58° probe both non-skewed beams aligned with the centre of the rail and skewed 70° ultrasonic beams for gauge corner crack detection were generated. These focused beams avoided surface waves and ultrasonic noise generated by the relatively rough rail surface.

Using the R/D Tech focus system a 3D record of defect position was obtained. The phased array system can record full A-scan data. To minimise the data volume and achieve the required rail inspection speeds multi-zone peak amplitude data was recorded with respect to volumetric position. Defect reporting software was developed to merge the data from all the ultrasonic channels into a single strip chart, which identified the location and amplitude of defects with respect to rail axial position. The TWI system was based on the R/D Tech Focus. With this instrument the system operates at a maximum speed of 16km/hr without loosing defect resolution. There are faster instruments that would enable the system to run at least 10 times faster. The limiting factor for pulse echo ultrasonic inspection is the velocity of sound in the rail. As an example, for the rail bottom inspection at 80Km/hr the wheel will have moved between 2mmand 4mm on return of the pulse to the probe depending on the upon the ultrasonic beam angle being used, thus limiting the minimum defect size that can be detected.

As a final system demonstration the wheel probe was tested on a track at TWI made up from five ex-service rails containing a variety of service induced defects and an implanted calibration block. Figure 5 provides photographs of the wheel probe being tested on the track at TWI. The wheel was mounted in a small trolley together with eddy-current and flux leakage sensors. To prove the integrity of the wheel, the wheel containing the phased array probes was driven up to 80Km/hr without loss of coupling or distortion of the ultrasonic data.

Fig. 5. Final trials of the inspection system containing the Phased array UT, EC and MFL techniques
Fig. 5. Final trials of the inspection system containing the Phased array UT, EC and MFL techniques

Development of eddy-current system

The eddy current technique was seen as having two applications within the Rail Inspect project. The first is the detection of shallow surface breaking cracks, where the eddy current results can be combined with ultrasonic results to offer a high capability throughout the rail section. The second application lies in the measurement of lift-off between the inspection device and the surface of the rail. Variations in lift-off can arise both from small vertical movements in the system which positions the inspection device and from uneven wear on the rail head. Measurements of the variation in lift-off could be used to control the gain of the eddy current crack detection system and to control the direction of the beam from an ultrasonic phased array transducer to compensate for uneven wear on the rail head. Both of these applications of the eddy current technique were investigated in this study.

The eddy current sensor assembly consists of a total of four probes, one pair of which are designed to detect surface breaking fatigue cracks in the centre of the rail head and the other pair of probes are designed to detect gauge corner cracking. Each pair of probes consists of an absolute probe, sensitive to changes in lift-off and a double differential probe sensitive to defects. All four probes are positioned in a single housing which is contoured to suit the rail head. Figure 6.

Fig. 6. Eddy current probe assembly
Fig. 6. Eddy current probe assembly

The system proved to work efficiently for finding real surface breaking fatigue defects at all orientations with a gap between the rail and the probes of 3mm.

Fig. 7. Results of eddy-current trial on track containing gauge corner cracking
Fig. 7. Results of eddy-current trial on track containing gauge corner cracking

The eddy-current results from the system when tested on a rail, containing many minor gauge corner cracks is shown in Figure 7. This section of data provides a new rail towards the left where the signal amplitudes are low followed by a rail join and then an ex-service rail containing gauge corner cracks. The maximum crack depth in this rail are 2mm from the gauge corner.

Graph lines 1 and 2 are the lift off channels for measuring the rail profile and lines 3 and 4 are from the double differential crack detection channels. Line 4 is from the centre of the rail where there was no surface cracking and line 3 was from the gauge corner where there were multiple cracks over the 22m length of ex-service rail.

Successful trials were carried out at up to 80km/hr.

Development of magnetic flux system

Flux leakage inspection of rail is not a new idea and was used many years ago before electrification and electronic signalling. Previous systems applied current to the rail, the current potentially interfering with rail signalling. Using basic flux leakage wall thickness measurement equipment the system was developed with six pairs of flux leakage probes across the width of the rail. This system uses coils for signal detection rather than Hall effect sensors. Hall effect sensors respond relatively slowly to flux changes but provide accurate signal measurement when static. This system is non-contact, with permanent magnets generating the magnetic flux. Using coils provides the benefit that the faster the coil is passed over the rail the larger the signal developed by any cracks. Trials were carried out at up to 80km/hr. The high speed test equipment is shown in Figure 8. This was a test disc created with new and fatigued rail to prove the technique at differing speeds with artificial and real defects. Some of the inspection results are shown in Figure 9, these show the rail ends, a large fatigue crack (broken rail) and three artificial defects with sizes of 5mm to 2mm.

Fig. 8. High speed test set-up
Fig. 8. High speed test set-up

The trials clearly demonstrated that the system reliably reported major defects of depth 4-5mm or deeper, while not reporting surface blemishes and minor defects of depth 1-2mm.

The developed probe pan shown in Figure 9 was applied to the ex-service track shown in Figure 5 containing 2mm deep gauge corner cracking. These cracks provided no reportable defect indications with the MFL system but an increased noise level. This inspection system is both rugged and rapid, relatively insensitive to small defects <2mm and sensitive to significant defects >3mm. For this reason it is believed that it would provide a rapid inspection tool for rolling surface defects.

Fig. 9. Magnetic Flux leakage probe pan
Fig. 9. Magnetic Flux leakage probe pan
Fig. 10. Results of inspection on a mixture of artificial and real defects
Fig. 10. Results of inspection on a mixture of artificial and real defects


The project was not intended to provide a total industrial solution but to provide additional ideas to the complex problem of rail inspection. This paper has shown the development of three inspection techniques for rail all of which can be train mounted. Phased array ultrasonic with the ability to focus and steer the ultrasonic beam; eddy-current for the detection of shallow rolling surface defects and magnetic flux leakage for the detection of relatively large rolling surface defects.

By combining the results of the techniques, their complementary features can be highlighted to useful effect. For example, if a defect is detected in the rail head by the eddy current system, but is not detected by the magnetic flux system, it will be known that the defect is likely to be less than 2mm deep. Such a defect may not require immediate attention, but its presence can be fed into a regime of planned maintenance. If the MFL system detects a defect in the rail head but UT does not then the defects are likely to be less than 5mm deep. Conversely if the UT system detects a defect in the rail head but the EC does not it is likely to be a sub surface defect. An example of the combined data on a single presentation is given in Figure 11 below which was from an ex-service rail containing heavy rail wear and minor gauge corner cracking.

Fig. 11.
Fig. 11.

In particular the use of the MLF technique has provided a very rapid inspection technique with a low equipment cost. The system does not contact the rail. The system cannot find all defects but it can form part of an inspection regime for rapid initial inspection without false calls due to small amounts of rail degradation.


  1. A dry coupled phased array ultrasonic system for the detection of defects in rail was developed in the Rail Inspect project.
  2. Phased array probes can be used to correct for wear on the head and to skew beams to maximise signals from defects.
  3. An eddy current system has been developed for the detection of minor flaws in the running surface of the rail.
  4. A flux leakage system has been developed to detect major surface breaking flaws in the rail running surface which has been tested up to 80kmhr.
  5. Software has been developed to identify significant signals from the large volume of ultrasonic data while ignoring lower amplitude noise.


The authors wish to thank all the partners in the Rail-Inspect consortium for their contributions to the work described and for their permission to publish this paper: Sonatest (UK), Computerised Information Technology (UK),Imasonic (France), Isotest (Italy), Tecnitest (Spain), Zenon (Greece), TWI (UK) and Technical University of Sofia (Bulgaria).

The consortium acknowledges with thanks the funding provided by the EC under Contract Number G3ST-CT-2001-50169.


  1. Rail-Inspect, project number CRAF-1999-70907.

For more information please email: