Phased array inspection of aluminothermic rail welds - the automated solution
TWI Bulletin, January - February 2011
A fast convenient means of inspecting rail joints arrives
Tamara Colombier studied material sciences in Strasbourg, France. She has both an Engineering Degree and a Master Diploma in material sciences and chemistry. Tamara joined TWI in 2008 and she has been Project Leader in the NDT section for 19 months. She is qualified in many non destructive methods of inspection. Tamara was the Project Manager for the second year of the Railect Project.
joined the Non Destructive Testing Technology Group, TWI, as Project Leader in December 2005; her current position is Senior Project Leader. Her expertise lies in the research and development of electromagnetic, thermographic and ultrasonic methods. She has been involved in various projects and site inspections for power and transport industries and has headed responsibility for design, inspection and application of onsite techniques. Before joining TWI, she was awarded a post graduate degree in NDT during which she worked for EDF (Electricité de France),on the development of ultrasonic techniques for the assessment of bond quality between concrete and composite. Angélique was the Project Manager for the first year of the Railect Project.
There are millions of aluminothermic welds made each year to join rail in Europe's rail network and millions are already in-service. Occasionally flaws develop during and shortly after welding that can lead to early failure and this can cause disruption to train services and costs to both the train operating companies and the maintenance operators. Although radiographic inspection is sometimes carried out, this is not very common, and the conventional ultrasonic standard requires many probes and scans because of the complex geometry. As Tamara Colombier and Angélique Raude explain since the welds are made by a casting process, these can be more difficult to inspect than joints made by other welding processes. Hence, the RAILECT project was conceived to provide a convenient means of inspecting the joints using a multiple phased array ultrasonic system.
This article describes the type of flaws concerned, the modelling of the ultrasonics, the practical realisation of a prototype system and the development of acceptance criteria.
The majority of field welding in the railway industry is carried out using aluminothermic procedures. Such welds are primarily associated with replacement of rail or weld defects, installation of insulated rail joint and track construction activities. In the UK, it is estimated that Network Rail produces over 65,000 Thermite welds per year and they have approximately 1.5 million welds installed on track.
Thermite welding is an effective, highly mobile and cost effective casting method of joining heavy steel structures such as rail, however this is also a very skilled welding process which requires experience and expertise. The many process steps can be altered by the welder and/or the environmental conditions and can result in poorer weld performance and in creation of defects within the weld.
Up until this date, only a few techniques of inspection have had some limited success to control the quality of these welds. For instance, manual ultrasonic and radiography techniques are sometimes carried out on the welds but this is not commonly used in the railway industry. Most of the time, only visual inspection is performed. Every new weld is inspected using visual techniques for weld defects, profile and geometry. Although there is no requirement to inspect rail welds using ultrasonics MPI or DPI, there is a conventional ultrasonic standard EN 14730-1:2006 Annex C developed specifically for the inspection of rail welds.
However, EN 14730-1:2006 Annex C describes a relatively long and complex inspection. It involves many ultrasonic probes and scans because of the complex geometry of the weld profile and hence this technique is used very rarely. As the welds use a casting process, these can be more difficult to inspect than joints made by other welding processes and as a result visual inspection is currently the only technique regularly used to inspect aluminothermic rail welds.
Rail flaw detection has an important part to play in ensuring the safety of the railroads. Recent accidents such as the Hatfield disaster, in the UK in 2000 caused by broken rail have focused attention on the technologies that enable the detection of flaws in rails and rail welds. The RAILECT project is a European funded project which aimed to develop a novel automated ultrasonic phased array system of inspection for rail welds. The project ended in December 2010. It included three research centres from two European countries; Kaunas University (Lithuania), University of Newcastle (UK) and TWI (UK) who have been developing an innovative ultrasonic phased array system for the benefit of four SMEs originated from three different European countries: Optel from Poland, Vermon from France, Spree Engineering and Kingston Computer Consultancy from the UK. Each SME has its own specific products and services that the RAILECT project aimed to develop.
Rail profiles and weld dimensions
The project initially focused on three different types of rail profiles CEN 60E1, CEN 60E2 and CEN 56E1, see Fig.1.
Fig.1. Dimensions of the three different rail profiles
These three rail profiles are in accordance with the European standards and are commonly used within Europe. CEN 60E1 and CEN 60E2 are the most recent profiles whereas CEN 56E1 is an older version. New welds follow the most recent profiles but many welds already in service are more likely to be CEN 56E1. The idea was to create a versatile prototype adapted for the quality control of both new and already in service aluminothermic rail welds. Although CEN 60E1 and CEN 60E2 differed slightly in terms of the head geometry, this was relatively negligible when compared to the rail profile CEN 56E1. This latter has a smaller web height ie 158.75 mm instead of 172 mm for the other two rail profiles.
The weld width of concern for the project was 35 mm (Fig.2.), although the finished cast range can vary from 35 mm to 90 mm.
Fig.2. A 35mm wide aluminothermic rail weld in service
The first iteration of the RAILECT prototype fitted the two most recent rail profiles but not the oldest one. The challenge at this stage was to produce a mechanical system that could be used on the three rail weld designs.
The project focused on three types of defects namely porosity, lack of fusion and shrinkage (Fig.3.).
Fig.3. Examples of defective samples
3b) Shrinkage defects
These defects were considered the most common type of defects leading to potential failure of the rail weld when in service and hence critical to rail safety. The rail foot was more predisposed to the lack of fusion defect whereas the shrinkage defect was more likely to be located in the web of the rail. Unlike porosity, lack of fusion and shrinkage were more likely to be found in specific areas of the rail weld. These observations were critical features to take into account for the design of the ultrasonic phased array system.
In total, 23 rail welds were manufactured for the purpose of the project. Five samples were allocated for each type of defect and a set of height reference samples corresponding to non defective welds was produced (Table 1). For each type of weld, three samples were dedicated to the fatigue testing and two samples were reserved for the ultrasonic phased array inspection with the prototype. Additional mechanical testing was performed on the non defective welds for the purpose of the Engineering Critical Assessment.
Table 1 Rail welds manufactured for the purpose of the Railect project
|Samples manufactured ||Weld characteristics |
|8 ||Reference samples - no defect |
|5 ||Porosity |
|5 ||Lack of fusion in the foot |
|5 ||Shrinkage in the lower web |
|Total = 23 samples |
After being manufactured, every sample was inspected using both radiography and manual ultrasonic testing in accordance with EN 14730-1:2006 Annex C so that one could compare the capability and reliability of the different techniques of inspection with the ultrasonic phased array system to be developed.
The device to be developed was expected to be:
- Weatherproof and functional in inclement weather
- Easy to operate
- Give clear indications of defective welds
Furthermore, the system had to inspect rail welds that presented the following characteristics:
- Normal gap width: 25mm wide (± 2mm each rail)
- Gap width range: from 26 to 80mm
- The finished cast width: 35mm (Normal gap)
- Finished cast range: from 35 to 90mm
- Gap between sleepers: from 340mm minimum - welds are not always placed at the centre of the spacing between sleepers
The majority of the work performed within the course of the project was carried out for normal gap corresponding to 35mm cast width.
Development of the UT phased array system
In order to choose the most appropriate design, the Consortium had to consider a design that could provide the most effective coverage of the weld area and a design where data acquisition and analysis could be feasible and relatively fast. The other objective was to minimise the amount of scanning in order to maintain an easy application and analysis of the signal. The coverage of the weld area was investigated using both modelling and experimental tests.
The modelling work investigated primarily the coverage of the weld, and the response from different defect types. This was carried out with both ESBeam tool and CIVA softwares. Figures 4, 5 and 6 show some examples of the modelling output. A 32 element phased array probe was considered for the purpose of the modelling experiments. The probe and focal law parameters were varied during the modelling experiments so that the most appropriate settings for full volumetric weld coverage could be identified.
Fig.4. Modelling experiments using Civa software show the beam coverage of the weld body using two 32 element phased array probes on each side of the weld
Fig.5. Modelling of the beam coverage of the weld body using ESBeam Tool - 32 element phased array probe
Fig.6. Modelling experiments using Civa software for the detection of two 2mm diameter side drilled holes separated by a 2mm gap in the ankle of the rail foot
Thus, it was shown that one phased array probe could be used to cover the whole of the web and the central part of the weld foot from one side, and that another set of probes were needed to cover each side of the foot. In total, eight probes were necessary to inspect one side of the weld. The system would simply be placed on the other side of the weld to complete the inspection and also to perform a volumetric inspection of the weld.
Various probe frequencies were investigated in the modelling but the selection of the most appropriate frequency for the present application was carried out during the experimental testing.
The inspection of the head and the web of the rail weld could be easily covered using 32 element phased array probes however, the inspection of the rail foot and toes was more complex because of the constraints dictated by the geometry and dimensions of the rail in these areas. Only small phased array probes containing a limited number of elements were able to fit onto the rail foot. These constraints lead to a more complicated system design that it was originally thought and as a result various design options were suggested and investigated.
To specify the probes in more detail, it was also necessary to consider the possible instrumentation that could be available. It was eventually decided to use a system commercially available and portable.
For the purpose of data acquisition and analysis, phased array instrument and software were used jointly to carry out the inspection, record and analyse the data. The focal laws were set up and saved initially but their application could be controlled to enable an automatic sequence. The data output were collected for analysis and compared with the acceptance criteria by means of a MATLAB program.
A first prototype was developed based on the modelling and laboratory trials. This system was clamped to the side of the railhead. All phased array probes were then held in place around the rail profile by means of clamp mechanisms.
The majority of the mechanical testing was performed for the purpose of the ECA calculations. The following tests were carried out:
- Hardness testing - BS EN 14730-1:2006
- Tensile testing - BS EN 10002-1: 2001
- Fracture toughness testing - BS 7448-1:1991
- Fatigue testing - BS EN 14730-1:2006
Results for the fatigue testing showed that generally non defective welds failed in the heat affected zone whereas defective welds failed within the weld, stresses being introduced by the defects.
ECA calculations were carried out for different areas of the rail weld in accordance with procedures prescribed in BS 7910:2005.
Both surface breaking flaws and internal flaws were assessed. Sensitivity analyses were carried out to determine some critical conditions such as flaw geometry and relative distance to surface for internal flaws.
A Finite Element model was also created to assess the accuracy of the analytical solutions developed with the ECA calculations. The model was based on the parameters provided by fatigue test activities. The modelling is shown in Figure 7.
Fig.7. Finite element modelling built from the results of the mechanical testing Fig.7. Finite element modelling built from the results of the mechanical testing
Laboratory trials were carried out to verify the correct operation of the focal laws and the detection of defects. Some of the results are shown in the figures Figures 8 and 9.
Fig.8. Sectorial scan performed on the ankle of the rail foot containing two 2mm side drilled holes in the rail foot
Fig.9. Inspection of the weld web using phased array transducer 9a) beam configuration
9c) weld containing porosity
Figure 8 shows the validation of the model described in Figure 6 related to the inspection of the rail foot. The two side drilled holes were detected using the probe and focal law parameters given by the model.
Figure 9 shows the scan obtained for the inspection of the web of the rail. In this case, the scan is given by the probe located on top of the rail at a selected probe offset with regards to the rail weld centreline. Both non defective and defective welds are illustrated on the figure above. One might notice that the bottom of the rail weld can be easily seen on the scan, this indication could be used as a reference position when carrying out the inspection.
When compared with the conventional manual UT technique used for the inspection of rail welds, the phased array system was more successful at detecting defects.
Trials were carried out on site at the Barrow Hill Railway Centre, Chesterfield (Newcastle University Test Track). Figure 10 shows the demonstration of the first prototype.
Fig.10. Demonstration of the RAILECT prototype at the Barrow Hill Railway Centre
Although it was shown that the system was capable of being deployed on site, the mechanical system operation was too fragile for normal use and it was decided to build a second prototype. The second system is more practical and easier to deploy on site as fewer manual operations are required to position the phased array probes around the rail profile. This system was successfully tested on the Network Rail test track (Fig.11) at High Marnham. Positive reactions were received from the project end user, Network Rail.
Fig.11. Field trials on Network Rail test track
The ultrasonic phased array system produced within the course of the RAILECT project proved the feasibility of an automated phased array system for inspection of rail welds. Modelling studies were validated by the experimental trials. Probe locations and parameters such as focal laws were investigated and optimised so that the maximum volumetric coverage of the rail weld could be achieved.
If adopted by the railway industry, the system could be one of the first automatic systems of inspection allowing rapid and simple detection of defects in rail welds and enhancing rail safety significantly it will also help rail operators reduce repair costs considerably. The system developed validated the RAILECT concept and proved to be reliable and technically valuable. Work is currently in progress to improve further the RAILECT system toward future adapability for its entry to market. Ultimately the commercial system will be portable, durable and handy.
The authors would like to acknowledge the European Community's Seventh Framework Programme managed by REA-Research Executive Agency for the funding of the RAILECT project and all the RAILECT partners for their contribution.