Skip to main content

Whole Life Rail Axle Assessment and Improvement

Final Report Summary - WOLAXIM (Whole Life Rail Axle Assessment and Improvement)

The European rail network is targeting a considerable expansion of passenger and freight traffic by 2020. In order to achieve this, increased reliability and availability of rolling stock is necessary whilst maintaining the same or a better level of safety. The axle life is a crucial part of the both the safety and economic performance of the vehicles and the axle deteriorates through its lifetime by means of fatigue and corrosion mechanisms. Periodic inspection is used to ensure that these mechanisms have not compromised the axle safety; however, inspection which takes a vehicle out of service impacts on the economic aspects of train operation. It is clear that errors are made in both aspects.

Recently, serious incidents due to axle failure in Cologne (2008) and Viareggio (2009) were due to the failure of the crack detection regime and continues to affect the inspection periodicity. On the other hand, axles of both passenger and freight vehicles with suspected corrosion are being taken out of service after only 3 years when their lifetime should be at least 20 years because of fears that cracks may initiate at corrosion pits.

The idea of the project was to improve the efficiency in the use of axles by extending their life. Improved inspection technologies, with knowledge of the conditions causing crack initiation and growth under corrosion conditions associated with reliability theory can achieve this objective. Hollow axles are advantageous because of their lighter weight and are likely to be increasingly used. Current inspection methods for hollow axles are slow and relatively inefficient. If new technologies can be introduced that have both the capabilities of a faster inspection, corrosion characterisation and fatigue crack detection these should be highly sought after.

There is therefore a clear need to fully understand the inspection regime and to improve it, particularly if this can be done efficiently without dismantling of the wheel set. Industry therefore requires a method of establishing the risk of cracking from specific corroded areas.

WOLAXIM has shown the potential of three new and better methods of crack detection and corrosion assessment for railway axle inspection. One method was to improve the measurement of corrosion and therefore the sentencing of corroded axles. This method, using a portable microscope together with software analysis tools and supported by the completion of significant background work on corrosion fatigue in rail axle steels, has already experienced significant interest from the rail industry. The primary application may be for it to be used with reliability software which has been developed to predict future probability of failure with a view to reducing scrapping of corroded axles. A tool has thus been provided to address the problem of scrapping corroded axles.

A second method is specifically for the hollow axles of high speed trains and aimed to improve the speed of the inspection and improve crack detection reliability. This was to be deployed while the train is in the depot overnight and without dismantling the wheel set. These objectives were achieved and an inspection time of around 5 minutes compared to existing methods taking over 20 minutes should be attractive to the industry. A final method, for the exposed body of the axles (intended primarily for freight wagon or passenger trailing axles) carried out as a vehicle passed an inspection station, was shown to be feasible, although the power available in the prototype needs to be increased for full scale operation.

Project context and objectives:

The idea of the project is to improve the efficiency in the use of train axles by increased safety and extending their life, which is often reduced by withdrawal from service because of observation of minor corrosion. New and improved inspection technologies, with efficiencies advanced in a number of different ways can achieve this objective. Avoiding manually deployed inspection by automating it and deploying it from trackside is one way, detailed knowledge of the conditions causing crack initiation and growth under corrosion conditions with an instrument to analyse this, and associated with reliability theory gives an alternative to solve a slightly different problem. Also, hollow axles are advantageous because of their lighter weight, and are likely to be increasingly used. Current inspection methods for hollow axles are slow and relatively inefficient and a faster method of inspecting these is desirable.

In order to provide the SMEs with a potential for growth after the project, it was planned to have a series of results that would lead to future sales for the SMEs. These are tied to the technical objectives described below.

Technical objectives

In summary the technical objectives were as follows:

- Produce a prototype instrument to detect cracks on the exposed part of an axle of a moving train. The instrument should be capable of being easily installed in an inspection station. It should be detecting cracks of the order of 2 - 3mm deep on an exposed axle of a train passing the inspection station at 5 - 10 km/h.
- Produce an instrument to measure and assess corrosion damage and risk of fatigue crack growth within an area of 0.1 m2 in a 10 second measurement period.
- Produce an instrument to inspect hollow axles using a non-rotating probe. The device should be able to detect cracks of 3 mm deep anywhere within the axle and should complete an inspection within 5 minutes.
- Produce equations for installation into STRUREL software to calculate inspection intervals incorporating crack growth by corrosion fatigue.

Project results:

The European rail network is targeting a considerable expansion of passenger and freight traffic by 2020, this is supported by a significant mandate given to the European rail sector in 2011 by the European Commission to shape the future of transport in Europe. The Transport White Paper presented by Commission Vice-President Siim Kallas outlined a step change for passenger and freight transport in Europe to be operated largely by rail by 2050 [1]. In order to achieve this, increased reliability and availability of rolling stock is necessary whilst maintaining the same or a better level of safety. The axle life is a crucial part of the both the safety and economic performance of the vehicles and the axle deteriorates through its lifetime by means of fatigue and corrosion mechanisms. Periodic inspection is used to ensure that these mechanisms have not compromised the axle safety; however inspection which takes a vehicle out of service impacts on the economic aspects of train operation. It is clear that errors are made in both aspects: Recently, serious incidents due to axle failure in Cologne (2008) and Viareggio (2009) were due to the failure of the crack detection regime, and continues to affect the inspection periodicity. On the other hand, axles of both passenger and freight vehicles with suspected corrosion are being taken out of service after only 3 years when their lifetime should be at least 20 years because of fears that cracks may initiate at corrosion pits.

Axles can be inspected either in the depot (while still on a train) with limited access, or at overhaul when worn wheels are removed and there is good access to the surface. At overhaul, there is no additional disruption of the train service for inspection and generally this time for inspection is preferred by the train operating companies. However, currently it is not usually possible to extend the inspection period to overhaul times and depot inspection is necessary.

Production inspection of axles is carried out by surface inspection methods (dye penetrant and MPI) and ultrasonics for solid axles. Inspection of hollow axles is by automated ultrasonics with a group of rotating probes in a similar way to the service inspection.

Methods of inspection at depot and overhaul and degrees of automation vary from country to country. In Germany highly automated phased array ultrasonic methods are used at overhaul [2], whereas in the UK surface methods such as MPI are used most commonly at this time.

In service, axles often tend to crack either in mid-span or under or close to the wheel seats. Various inspection methods have been tried or developed for this particular inspection. In the UK, surface inspection methods (particularly MPI and electromagnetic) have been introduced since the Rickerscote accident in 1996 [3] for accessible areas. However, where the crack initiates from an inaccessible surface (e.g. fretting cracks under a wheel) the inspection is by ultrasonics. The methods adopted are generically known as the high angle scan (applied from the axle body), the near end scan and the far end scan (applied from the axle end). The WIDEM project [4] carried out a complete survey of current axle inspection methods and their performance in terms of POD. This showed that in general the surface methods, particularly eddy current, were more sensitive when they could be deployed.

Hollow axles are also used, particularly for high speed passenger trains, where the loss of weight has advantages, and the ultrasonics used in this case is an angled beam scan from a rotating probe in the bore. This inspection is mechanised. This inspection requires incrementing and rotating the probe, a very slow process. Portable devices for use in depots have been manufactured but these tend to be unreliable due to the long reach and sensor rotation required. There is also some uncertainty in their capability for crack detection in the radius between the axle body and the wheel seat. The results of an inspection performance trial in the WIDEM Project indicated that a 90 % probability of detection may only be achieved with cracks of 10 mm depth or greater using this method.

Manual methods of inspecting the wheel seats of hollow axles (using high angle scans from the axle body) have been introduced for inspection without removal of the end caps. Surface inspection on site can be carried out by manual eddy current methods, and these can also scan the radius. These are highly sensitive but are also subject to the expertise of operators.

The primary difficulty and skill required for the inspections is discriminating between geometrical echoes and crack signals. This can be assisted greatly by better use of the phased array probe giving a scanned picture of the whole surface. Such an approach was being investigated in the SAFERAIL project [5]. SAFERAIL was also attempting to carry out this on solid axles (as opposed to hollow axles in this project) and was inspecting the surface by means of ACFM, which requires scanning of the surface with an array probe.

Manual methods are also regarded with some suspicion in some circles because they are dependent on the operator, and human factors in ultrasonic inspection are known to be the most significant factor in the performance capability of the techniques. This was studied for axle inspection in WIDEM [4] and by the UK Health and Safety Executive [6]

An inspection tool similar in deployment to the proposed AC current method was patented by TTCi and Tecnogamma. This relies on the laser ultrasound method to detect cracks as the vehicle passes the inspection station [7]. Its limitation is in its sensitivity.

The Rickerscote accident report also concluded that the cracks had initiated as a result of corrosion, although all previous analysis had focussed on fatigue and very little work on the process of corrosion fatigue had been carried out.

The analysis of fatigue properties of railway axles had been investigated earlier (Snell [8], Beretta et al. [9]) and the crack growth in railway axles had also been studied (Zerbst et al. [10]) in order to better define fatigue design (WIDEM [4]) and the scheduling of inspection for these components. However, until recently the diffused corrosion that can appear on some areas of the axles and the possibility that the pits caused by corrosion can promote the nucleation and the subsequent propagation of fatigue cracks had not been considered to date when defining axles' fatigue properties.

Some bibliographic notes report cases of axle failures due to crack propagation from corrosion pits. Hoddinott [11] reports that about five mid-span failures of in-service axles occurred in the UK from 1996 to 2003, four of which have been connected to the presence of diffused axle surface corrosion and corrosion pits. On the other side of Atlantic, the Transportation Safety Board of Canada [12] reported one axle failure to have been caused by corrosion pits under the journal bearing. It also mentions another seven failures with similar features occurring between 1998 and 2000.

The effects of corrosion on fatigue properties can firstly be observed as a number of surface defects or pits which obviously reduce the fatigue strength of the axle body (this seems to be the effect described by Hoddinott in one case). Whereas, in general, 'corrosion fatigue' detrimentally influences both crack initiation and crack growth (Schijve [13]), this phenomenon is characterised by the formation of a consistent number of small cracks, whose nucleation is favoured by the pits due to the aggressive environment. These small cracks are then able to cross the 'microstructural barriers' (Akid & Miller [14], Miller [15]) with ease and a much faster growth rate than in air. These effects result in a large decrease in fatigue properties even in gentle environments with a disappearing of the 'knee' of the S-N diagram. Previous research has provided some clues / indications as to the fatigue properties of carbon steels:

i) rotating bending fatigue tests by Endo and Miyao of a mild carbon (reported by Schjive [13]) steel in tap water showed a reduction of fatigue strength of the order of 50 %;
ii) fatigue tests in tap water for an AISI 1018 steel showed fatigue strengths of 120 MPa at 2106 cycles (approximately 50 % of fatigue limit in air) and 90 MPa at 107 cycles (Ragab et al. [14]).

This experimental evidence shows a detrimental effect of the environment upon the S-N diagram (which can be also found in many engineering manuals) which can be taken into account applying a reduction of the fatigue design limit. This was the method chosen in BASS 503/4 BR [15 - 16] which suggested a reduction of axle design limit from 160 MPa to 110 MPa to allow for corrosion protection degradation. However, current EN13103/4 standards [17 - 18] do not consider this type of reduction of fatigue strength and they rely on adequate axle maintenance.

This topic is of real and practical interest as it plays an important role in the maintenance cost of wheel-sets. It bears particular relevance to freight cars where the 'one million mile' approach would imply avoidance of maintenance operations (Lonsdale & Stone [19]).

Recently, a series of investigations carried out to assess the effect of corrosion upon fatigue properties of A1N, steel widely adopted for manufacturing railway axles [20]. Rotating bending corrosion fatigue tests on both smooth and micro-notched specimens machined from axles were performed. The S-N data for complete failure of the specimen have been obtained: results have shown that corrosion has a significant influence on the fatigue life especially at very high cycles (> 107 cycles) where the disappearance of the fatigue limit seems to be present although this was not tested at a full range of loads. In some countries, in order to prevent these effects, railway axles are protected by means of the applications of suitable coatings. Unfortunately, impacts due to the ballast or lack of adhesion due to low cleanliness can cause corrosion to attack axles' surfaces at the point of disruption and begin its detrimental effects on fatigue strength locally.

Although corrosion is beginning to be recognised as an important part of axle fatigue, the assessment of corrosion, in order to sentence corroded axles has not been well advanced. The current state of the art for this seems to be only visual inspection, at best supported by a pit depth gauge or similar device. Corrosion can be measured in the laboratory by optical methods to a high degree of accuracy [21] but these tend to be slow, requiring precision scanning of small areas and do not give a suitable output for sentencing. For corrosion assessment generally, there are a number of standards (NACE) for the measurement and classification of pits, but these are not related to high cycle fatigue. There was therefore no instrument that can be used directly for the quantitative on-site inspection of axles.

The net result of this lack of knowledge is that axle with some a small evidence of corrosion are withdrawn from service (particularly in the UK), with no real knowledge of the effect the corrosion has on the axle life, and although this is a fail-safe approach, it has resulted in a economic issue when the axles have useful life.

The WOLAXIM project was conceived with a view to solving some of the inspection problems mentioned above. It was an R4SME project in which the foreground IP from research carried out belongs to the SMEs. It was structured in the form of work packages. Within each of these work packages was a task structure leading to the completion of the work package. This, together with a series of deliverable reports was used as tools to manage the project.

The main S+T results from the project were the research into corrosion fatigue, which produced significant new knowledge of the process and characteristics of fatigue crack initiation in a corrosive environment, the development of the three novel axle inspection techniques, two of which are close to commercial application, together with development of the software that linked one of the instruments through models of corrosion fatigue to axle life estimation. These, and how they were achieved, are described within the structure of the project below.

Preliminary work

The project started with the formation of a test plan for the fatigue tests, specification of the electronic instruments and acquisition of test samples. It is important to note that fatigue tests by their nature are time consuming and also have an uncertain duration. The corrosion fatigue test plan indicated 10 small scale tests on A1N steel, 16 on A4T steel and 6 on 30 CrMb steel, together with some full scale tests. The specification of the electronic instruments was produced. The sample collection included some small scale samples and some larger axles. The small scale samples include those produced in the Polimi corrosion fatigue tests and those with fatigue cracks produced by TWI for the other small scale experiments. Full scale solid axles were made available from ATM and some of those used in the WIDEM project. Some hollow axles with cracks were also available.

Research into corrosion fatigue of rail axle steels

Initially, a literature survey was carried out to ensure that no prior worked on high cycle corrosion fatigue had been carried out in other industries. It showed that very little work had been carried out on this subject, the majority of the work having been carried out on large structures exposed to corrosion and fatigue such as offshore oil platforms, which have low cycle counts by comparison. Therefore, as high cycle corrosion fatigue is a relatively newly identified phenomenon in the rail industry, this enabled the consortium to be sure that the planned work had not been carried in another industry earlier. The work carried out by Polimi prior to and during this project was therefore the main source of data in this area.

Tests of small scale samples to failure in corrosion fatigue to produce the SN diagram for A1N, A4T and 30CrMo materials were carried out. An increased number of tests were made than originally planned in order to extend the data to obtain results at lower stress levels.

In order to monitor the corrosion situation, the free corrosion potential was monitored during the tests. Only the first stage of the test is plotted, when the initial value of about -500 mV, corresponding to the absence of any corrosion pit onto the surface of the specimen, this decreases until the value of -730 mV, which corresponded to the formation of the uniform oxide layer onto the surface of the specimen. Afterwards, the value of the free corrosion potential does not change significantly until the final failure of the specimen.

The first series of rotating bending tests on smooth small scale specimens was run on A1N, to add data in the S-N diagram of this material. The total number of test run until the final failure of the specimen is six, others were interrupted before failure to study the surfaces.

Data clearly show that the atmospheric corrosion has a significant influencen on the fatigue life. At stress levels, below the air fatigue limit of the material (?S = 255 MPa), where cracks would not be able to nucleate and propagate in the air, the presence of a corrosive environment allows crack nucleation, from corrosion pits, and environmental assisted crack growth. This results in a disappearance of the air fatigue limit and in a continuously decreasing S-N diagram in the corrosive environment (failures, with Nf > 107 cycles, also occurred for a stress level equal to ?S = 240 MPa which corresponds to approximately 50 % of the fatigue limit in air). It is also of some importance to compare the present data with the design limits outlined in axle standards: the presence of rainwater can cause failures at stress levels well below the EN13103/4 design limit (?S = 362 MPa), while the BASS design limit (?Sdesign =220 MPa), which is valid for steels with UTS of 550 - 650 MPa, is comparable to present experimental results.

The small scale tests were continued on A4T. and 30CrMo. This led to a significant result from the SN data was the comparison, in terms of corrosion-fatigue life, between A1N, A4T. and 30CrMo, which are very similar despite the different strength properties of the steels. The crack growth during the corrosion fatigue process, as it occurs at lower loads than previously known, has an effect on the Paris curve and in the end the equations for corrosion fatigue.

In addition to the planned work obtaining failure data, tests have been interrupted to observe the surface in detail using a scanning electron microscope. These were tests interrupted at a fixed percentage of the fatigue life and were been performed to study the phenomena of corrosion fatigue at different stages and for different stress levels. Replicas were used as a means of permanently recording data from part completed fatigue tests and this was shown to be a useful and accurate procedure to retain the data and allow continuation of the fatigue process.

The number of readings was increased from a plan of 12 tests. 43 were in fact carried out. It was noted that the progress of a corrosion fatigue test on the small samples consisted of a four distinct phases:

(1) pitting (corrosion alone);
(2) formation of microcracks (typically tens of microns in length);
(3) coalescence of microcracks when depth of cracks was around 0.3 mm (length 0.5 - 3 mm) is reached;
(4) further coalescence and growth of cracks to a size detectable by conventional NDT.

The evolution of corrosion pits, the pit-to-crack transition and the coalescence of cracks at different stages and for different stress levels, were established as separate stages of the corrosion fatigue process. Moreover, it has turned out that knowledge about each of these separate stages of the corrosion fatigue damage process was important for development of the equations that would later be used to estimate axle life. The main results from this work were that it was shown that A4T and 30CrMo (high strength axle steels) had a similar pattern for crack initiation under corrosion fatigue to the standard axle steel A1N. The detailed crack growth rate data in corrosive environment was established, by collecting data at different stress levels in A1N and A4T materials.

Additional study of the pit to crack transition showed that the stages of initiation of a microcrack from a pit were studied and this growth appeared to occur with pits at around 30 - 50 µm diameter, a secondary pit appears at the bottom of this and the crack initiates at the bottom and progresses to the original pit

A more intensive study of the crack to pit transition in A4T was carried out. In this material three different stage of the transition between pit and crack can be documented. The first is the formation at the bottom of the pit of a secondary pit, probably due to an electrochemical effect (Type A), the second is the development from the secondary pit of a short crack at the bottom of the primary pit (Type B), the third is the growth of the short crack until the mouth of the primary pit is reached (Type C). The presence of each type of previous defined corrosion pit has been studied in the early stage of the corrosion fatigue life. As for the analysis of the surface damage, the image of the pits has been collected. An interesting feature of this data is the low percentages of life that the pits are seen. This indicated to a certain extent that the absence of cracking indicated a low percentage of useable life.

Finally, the crack patterns from interrupted and final tests were studied. The crack aspect ratio changes dramatically as the cracks coalesce during a fatigue test and the crack densities for different tests were analysed using Weibull probability distributions. Thus, the process was fully understood and it was possible to develop model solutions.

The full scale tests included examination of the surfaces with the optical instrument at intermediate stages of the corrosion fatigue process, which involved partial dismantling and rebuilding of the testing system at each stage. Eventually of the planned six tests, four different tests were agreed with slightly different load patterns to better reach the goal of a comparison, in terms of maximum crack length, between the experimental full scale tests and the prediction of the corrosion fatigue model. One of the full scale tests was a long 'load spectrum test' (15 106 cycles). and one full scale test was a very long 'end life' test (30 106 cycles). The main result of the full scale tests confirmed the nature of the crack initiation process and the equations developed in the small scale tests.

A good estimation of the corrosion fatigue life of the full scale axles was obtained by means of the crack growth model under corrosion fatigue corrosion that has been previously set with the small scale tests results.

The modified Nasgro equations were produced. This was a difficult task because corrosion fatigue crack growth at the low stress levels caused a significant change to the Paris parameters in the region normally associated with the fatigue limit.

Corrosion assessment technique and instrument development

To take advantage of the corrosion fatigue information, a device for analysing the surface of the axle with a view to assessing its life in the presence of corrosion and fatigue was planned. It was originally envisaged that the device would need to measure the depth of pitting and to analyse the surface statistically to enable the risk of cracks developing to be estimated. The primary application was expected to be the sentencing or otherwise of axles that show evidence of corrosion but at present only use visual inspection for this purpose. A number of devices were investigated initially which would have been possibly developed into the required instrument for 3D surface measurement. However, examination of the surfaces of the samples from the corrosion fatigue tests described above were showing that surface analysis requiring the detection of microcracks may give the information required. These tests showed that unless these microcracks are present then the axle is not suffering from corrosion fatigue or is at a very early stage of it. This being the case the emphasis of the design of this instrument shifted to the use of convenient means of detecting the microcracks or getting images from the surface that can be analysed to distinguish between cracks and pits and give a quantitative measure of the degree of cracking.

A number of microscopes were investigated and the one eventually chosen was a USB microscope with x50 - 200 magnification and a polarisation option. A scanning system and a holder for the microscope were added and this made a workable prototype. The system enabled the microscope to be supported in a stable position (either against the axle body or a radius) while a scan around the circumference took place. However, this as it stood required manual analysis of the images. Initial information from end-users suggested that the instrument would be often used to analyse small areas of corrosion initially detected visually, so it was ideally suited for this, without further modification. This could also meet the specification of inspection speed as far as data acquisition was concerned, but the analysis could not be done at this speed.

It was clear that surface preparation was an important part of the procedure for use of the device. In order to facilitate the proposed optical survey a rusty surface must be cleaned to enable the analysis. Laboratory methods required the sample to be immersed in a solution for 20 minutes at 75°C so this would not be practical on site. Therefore the use of three proprietary rust removers and various mechanical treatments were investigated for preparing the surface. A suitable procedure was identified.

(1) The mechanical cleaning procedure must not be too abrasive. Abrasion marks look like cracks (especially to an automated system.
(2) The scanner could usefully have a facility for stepping in along the axle as well as round it.
(3) The scanning of a complete axle is too slow at present. This is a problem both with data collection and interpretation. The data can be collected automatically with a few modifications, but the software for automatic interpretation will require development.
(4) Mechanically it would be advisable for the microscope to be incorporated into a more rugged structure (although no specific problems were experienced it should be noted that the equipment was deployed by research staff), and the focussing system also automated.

To better meet the specification for data analysis, it was necessary to further enhance the instrument by the addition of crack counting software. When using the correct sizing parameters, the features on the images could be counted and boxed. It reduces as well the time required to count manually the defects on the images and may help in distinguishing defects from simple image artefacts. However, the software could be further refined as there are a lot of variables to take into account such as lighting of the environment, the quality of the images (due to the microscope or the operator), corrosion of the sample (if the sample is too corroded, the analysis is more difficult due to the agglomeration of pits in the entire image). The sensitivity of the microscope was checked against a conventional but more powerful optical microscope and the results showed that cracks of around 200 µm long were detected similarly by both instruments. Consideration of the pixel size suggested that features down to about 3 µm would be equally observed.

After familiarisation of staff with the system, it was tested on site as well as on the laboratory. A large number of images were recorded and cracks were detected. A comparison of sensitivity with MPI and eddy current NDT was also made, showing that the microscope method was the most sensitive, although slower in deployment.

A training course for use the equipment was also developed, and a pilot course carried out with one of the SME members of the consortium.

Axle life software and application of the instrument

It was required in this task to provide a software that would integrate into the proprietary Strurel software suite. In this context, it was necessary to provide an algorithm for inputting high cycle load spectra (which would be typically be provided by a vehicle operator), and the inclusion of the crack growth equations described above. The work involved the study of the WIDEM inputs for high cycle variable amplitude loads. However, an effective (sufficiently fast and stable) software solution of the integral equations involved turned out to be quite difficult, although it was eventually achieved.

A stand alone Fortran program for calculating crack growth during corrosion fatigue and including the modifications to the Paris equations was produced and checked, and this was also incorporated into Strurel to complete the task. Some case studies were carried out with sample load spectra and detected crack sizes to see how the software could be used.

There are a number of inputs to STRUREL which are shown below:

- diameter = 160 mm,
- crack aspect ratio (crack depth/ half crack length) = 0.8
- critical crack height = 10 mm,
- R-ratio = -1,
- K1c (critical stress intensity factor) = 100 MPam1/2.

This represents a total cycle count of approximately 800 000 000, roughly corresponding to 2 400 000 km (assuming 1 m diameter for the wheel).

The other input required is the initial crack length. A number of values were chosen for this, the lowest value 35 µm corresponds to the pit to crack transition or where no crcak could have been detected by an instrument. Higher values are for example those which may be detected by the microscope on inspection.

Phased array inspection of hollow axles

Initially, the probe design for the hollow axle inspection and the modelling of beams focal laws was carried out. This enabled the concept of a conical shaped probe to be developed into a detailed design. Design parameters such as inner and outer cone diameters, probe frequency and probe element size and number of active elements to enable beam steering without sidelobes, choice of coupling liquid needed to be incorporated. Each of these was investigated in turn. Initial measurements of the sound velocity in a suitable coupling medium with changes in temperature were investigated. How the number of elements affects the circumferential directivity pattern and the echo height was determined. This, with the axle bore dimensions, gave the cone angle and its inner and outer diameters.

The electronic rotation of the sound field by switching the active element groups gives a natural circumferential scanning resolution of 360° divided by 48 elements, i. e. 7.5°. To get smaller steps than 7.5° we can steer the angle of incidence with delay laws. The patterns of the switched element group without angular steering are drawn with the blue curves (-7.5° and 0°). These curves have the largest amplitude. The amplitude loss of the steered beams is very small. In the superimposed directivity patterns we have a maximum amplitude loss of less than 1 dB in the echo signal.

For the complete rotation of the sound field with 1.5° resolution, we need only two delay laws ±1.5° and ±3° for the angular steering. The directive patterns of these delay laws show only a very small increase of the grating lobes and a very small decrease of the maximum amplitude. So we can apply these angular stepping without loss of probe sensitivity.

A mock up of the probe with 10 elements was thoroughly tested to ensure that the parameters suggested by the modelling worked well in a practical system. The final probe consisted of dual conical 48 element probes working at 3 MHz.

Ultrasonic testing system

The axial scanning will be realised by a commercial motor driven linear axis. The pumping system for the coupling liquid is realised with a commercial hose pump for silicon hoses.

The Compas phased array device, made by Diatek, is applicable without changes. Interesting parameters for the system performance are the data reduction, storage rate and the pulse repetition frequency. The data reduction will be done by variable pixel distance by using the maximum amplitude detection. A reduction down to 1 pixel per A-scan is possible. The maximum storage rate is approximately 2 MByte per second. The maximum pulse repetition frequency is 20 kHz for small number of pixels. The Compas device is optimised for synchronised parallel operation of up to 16 PA probes.

The construction involved considerable challenges in electrical connection of the elements in the presence of an oil filled volume and the oil supply for the couplant. Inside the probe is a real space with only 10 mm diameter for the 96 coaxial cables of the transducers, for 2 copper pipes for the inflow and the outflow of the coupling liquid, for 1 copper pipe for the equalising of the oil pressure in the bellows and for 1 copper pipe for the air pressure-balance in a hermetic sealed bore.

The prototype was completed with only a minor modification needed to rebuild after an oil leakage had been found which would have compromised the coupling. The equipment also included a special user interface for the standard phased array software used.

The equipment was tested and the set up optimised using a test axle with known slots and a procedure produced. This confirmed that the objective inspection time for an axle of five minutes could be achieved. The system detected all the slots in the body of the axle, but not those in the diameter transition radius. This would need another angle of the probe.

Development of AC thermography system

The development using AC thermography to detect cracks in exposed train axles as the pass an inspection station. The principle of this is to generate an alternating current in an axle that heats it slightly. Currents passing around a crack cause hot spots and it should be possible to detect these with a thermal camera. In the induction format, it has been investigated recently by Almond et al. [24] and showed a good probability of detection albeit on much smaller components than were used in WOLAXIM.

Since this was a completely novel technique, there was a stage concentrated on obtaining some preliminary results showing the feasibility of the technique. This was challenging because the equipment is not readily available and some parts had to be built and the circuitry constructed. Obtaining clear images of cracks with the instruments available was not easy initially and better understanding of the phenomenon reported in the literature was required. To this end, models of the technique in both 2D and 3D have been developed.

The 2D models showed that 1 V across the system would give a temperature difference across the crack of about 1 deg K at 20 kHz. It was also calculated that 114 A at 100 KHz in the small scale samples would give a useable temperature difference. The surface current density for this is around 2 A/ (mm of circumference x depth of penetration). To obtain the same current density in an axle would require 750A, assuming the same depth of penetration and 150 mm diameter. It was therefore expected that provided this could be supplied with a system supplying 10 kW or thereabouts (power = I2R = 11.25 kW). On the basis of the resistance calculations above, the voltage required across the axle would be 750/I = 0.02 = 15 V.

The initial feasibility studies were able to show the detection of a crack in a small steel bar at high frequencies. However, there was initially considerable difficulty in sourcing a suitable electrical system (the high powers at high frequencies puts considerable stress on electronic components) for the full scale prototype and it was eventually decided to manufacture a system in house based on circuitry used for arc welding, which should have been able to deliver high powers for short durations.

A two stage approach was adopted, firstly a 3 kW system was produced to test the principle. The first prototype was reasonably successful. With this it was possible to produce indications from a crack in an axle by induction. It was also possible to detect a small (2 mm deep) crack in a small scale axle directly at a lower frequency than had been done earlier.

An experimental arrangement representing the system on track was then set up. Extensive tests were carried out with this system to increase the power delivered to the axle by using impedance matching techniques with capacitor banks and transformers. Although it was possible to increase the current significantly, it did not have enough power to produce detectable heat in an axle on a track. The problem seemed to be the imaginary (inductive) impedance caused by the circuit. With no additional circuit components connected, the voltage waveform (a square wave) is transformed to a triangular wave for the current, indicating that the system was primarily inductive.

However, the results were sufficiently encouraging to suggest that a higher power system would achieve these objectives. The second system, with 25 kW of power, was then produced. Again, the power delivered to the axle was found to be only a small part of the output of the system. There were a large number of attempts to increase the current, by matching the impedance with transformers and using an additional set of experiments with series and parallel resonant circuits to eliminate the imaginary impedance. Unfortunately, obtaining the necessary real currents to flow through the axle to produce an image by direct heating in a full scale axle was not achieved, although it was possible to observe a heating effect.

A comparison of the surface current density of the successful detection of the 2 mm deep crack noted above and that in the axle suggested that the current was around 75 % of that needed to reveal the crack.

One other aspect of the modelling compared to the experiment which was noted was that in all the models, it seemed that the crack indication would continually increase with time, provided that the current was maintained. It was therefore anticipated that even if the current was rather low then the indication would be visible in time. This in general turned out not to be the case experimentally, and the indication was best seen with rapid heating early in the heating cycle. The reason for this difference is not clear, but is possible due to the colour palette relationship with temperature.

In order that the thermal camera would cover one axle completely, it needs to be at a certain distance from the axle. To achieve this, the camera must be placed at some depth below the track. In order to do this without an exceptionally deep trough or excavation, a reflector was designed. This enables a smaller space below the track in exchange for a longer space along the track, which is usually more convenient. The unit is positioned at the centre between the rails and in the centre of the electrodes.

The spatial and thermal resolution of the proposed thermal camera was checked against a real axle. However the spatial resolution will ultimately depend on the size of the hot spot. It may be that two cameras are required to cover the whole area.

The high power system was installed on site in field trials and it was possible to test the system on a real vehicle and to show that the currents generated could be similar to those in the laboratory, and that a heating effect was detected. The presence of sunlight reflections caused some problems and the system needs to be more mechanically robust.

Conclusions

The scientific and technical objectives of the WOLAXIM project were substantially met as follows.

Objective 1: Produce a prototype instrument to detect cracks on the exposed part of an axle of a moving train

The instrument should be capable of being easily installed in an inspection station. It should be detecting cracks of the order of 2 - 3 mm deep on an exposed axle of a train passing the inspection station at 5 - 10 km/h. It will be tested statically by month 18 (milestone 2) and tested in field conditions in month 22 (milestone 3). This objective was partially achieved, a system was built and tested extensively in the laboratory and field trials were also carried out. However detection of a crack in an axle was only achieved on a small scale or by induction, as a greater amount of power was needed for full scale operation than had been expected. However it can be stated that the proof of principle was achieved.

Objective 2: Produce an instrument to measure and assess corrosion damage and risk of fatigue crack growth within an area of 0.1 m2 in a 10 second measurement period

This objective was almost fully achieved (the speed of scanning the surface could be achieved but the analysis required a separate step which took longer). Extensive testing on site, together with training of the lead SME in the use of the instrument was carried out. This instrument together with the background in corrosion fatigue crack growth established give an operator the tools to sentence corroded axles.

Objective 3: Produce an instrument to inspect hollow axles using a non-rotating probe

The device should be able to detect cracks of 3 mm deep anywhere within the axle and should complete an inspection within 5 minutes. This objective was fully achieved and the inspection time was somewhat better than the objective, In effect use of this system makes the time to move the system between axles the main time in the overall inspection operation.

Objective 4 Produce equations for installation into Strurel software to calculate inspection intervals incorporating crack growth by corrosion fatigue

This objective was fully achieved. A full set of equations was delivered and tested. It provides one of the analysis tools for axle life estimation under corrosion conditions.

References

[1] UNIFE Annual Report, 2012.
[2] Kappes, W., Fraunhofer IZFP, Saarbrecken (Germany) Bohr, W., Fraunhofer IZFP, Saarbrucken (Germany) Kruning, M., Fraunhofer IZFP, Saarbrucken (Germany) Rockstroh, B., Fraunhofer IZFP, Saarbrucken (Germany) Rodner, C., Fraunhofer IZFP, Saarbrucken (Germany) Goetz, J., Fraunhofer Technologie-Entwicklungsgruppe, Stuttgart (Germany) Nemec, D., Fraunhofer Technologie-Entwicklungsgruppe, Stuttgart (Germany), Application of new front-end electronics for NDT of railroad wheel sets, paper Th 1.4.1 European Conference of NDT, Berlin 2006.
[3] Health and Safety Executive (HSE), Railway accident at Rickerscote, November 1996. ISBN 071761171X, HMSO.
[4] WIDEM EU Project. Wheelset integrated design and effective maintenance. Website: http://www.widem.org
[5] http://www.saferail.net.
[6] UK Health and Safety Executive, PANI project reports 1-3 (1998 - 2004).
[7] TTCi, Remotely detecting cracks in moving freightcar axles, TTCi report SAFETY, 8 August 2006.
[8] Snell J. R. Key issues in the application of unified railway axle standards. J. Rail and Rapid Transit 2004; 218:279 - 282.
[9] Beretta S., Ghidini A., Lombardo F. Fracture mechanics and scale effects in the fatigue of railway axles. Eng. Fract. Mech. 2005; 72:195 - 208.
[10] Zerbst U., Vormwald M., Andersch C., Madler K., Pfuff M. The development of a damage tolerance concept for railway components and its demonstration for a railway axle. Eng. Fract. Mech. 2005; 72:209 - 239.
[11] Hoddinot D. S. Railway axle failure investigations and fatigue crack growth monitoring of an axle. J. Rail and Rapid Transit 2004; 218:283 - 292.
[12] Transportation Safety Board of Canada. Main track derailment: Canadian national train No.G-894-31-14. Railway Investigation Report R01Q0010, 2001.
[13] Schijve J. Fatigue of structures and materials. Kluwer Academic Publishers, New York, USA, 2004.
[14] Akid R., Miller K. J. Short fatigue crack growth behaviour of a low carbon steel under corrosion fatigue conditions. Fatigue Fract. Engng Mater. Struct. 1991; 14:637 - 649.
[15] Miller K. J. Material science perspective of metal fatigue resistance. Mat. Science Tech. 1993; 9:453 - 462.
[16] Ragab A., Alawi H., Sorein K. Corrosion fatigue of steel in various aqueous environments. Fatigue Fract. Engng Mater. Struct. 1989; 12:469 - 479.
[17] BR BASS 503. Design guide for the calculation of stresses in driving axles. Railway Companies of 1996, Issue C, 1996.
[18] BR BASS 504. Design guide for the calculation of stresses in non-driving axles. Railway Companies of 1996, Issue C, 1996.
[19] EN13103. Railway applications- Wheelsets and bogies - Non powered axles - Design method. CEN, 2001.
[20] EN13104. Railway applications- Wheelsets and bogies - Powered axles - Design method. CEN, 2001.
[21] Lonsdale C. P., Stone D. H. North American axle failure experience. J. Rail and Rapid Transit 2004; 218:293 - 298.
[22] Beretta S., Carboni A., Lo Conte A., Palermo E. An investigation of the effects of corrosion on the fatigue strength of A1N steel railway axles. International Journal of Fatigue, 2007.
[23] http://www.alicona.com/cms/front_content.php?changelang=1&idcat=47&idart=25
[24] D. P Almond, B.Weekes Teng Li, S. G.Pickering E. Kostson, J. Wilson, G. Y.Tian S. Dixon and S. Burrows 'Thermographic techniques for the detection of cracks in metallic components' Insight, Vol 53, November 2011, pp. 614 - 620.

Potential imapct:

In 2011, the European Commission gave the European rail sector a significant mandate to shape the future of transport in Europe. The Transport White Paper presented by Commission Vice-President Siim Kallas outlined a step change for passenger and freight transport in Europe to be operated largely by rail by 2050.The European Commission acknowledged rail as the greenest of transport modes. The impact of this on the development of the rail infrastructure is considerable and expansion of fleets and pressure on rolling stock availability can be expected, while levels of safety are expected to be maintained and any additional environmental impact minimised.

The rolling stock count at present in Europe is 249 110 passenger carriages and 1 245 908 freight wagons (UIC figures for mainland Europe added to known UK figures). Assuming 4 axles per passenger carriage and 2 per freight wagon this gives a total of around 3.5 million axles.

Of course, some axles are painted to avoid corrosion and the usage of these axles varies, but nevertheless their existence and continuous subjection to corrosion fatigue, suggests a large market for the SMEs to approach.

The role of inspection in determining the safe life of an axle is crucial. Inspection of the axles for cracking takes place at intervals set by knowledge of expected loading, known crack growth rates and inspection sensitivities. Usually, the inspection interval is such that an inspection is required between major overhauls, so inspection of axles while the train is in service is still required and this is currently inconvenient and costly for the train operators. Current inspection methods designed for crack detection typically do not attempt to identify or measure the corrosion and lifetime.

If withdrawals due to unspecified levels of corrosion take place while a train is in-service there is a consequent significant cost associated while the train is out of service for additional maintenance and also the financial and energy cost in making and supplying a new axle. This has become a specific problem for the high cost, highly detailed hollow axle designs such as those used in Pendolino trains but also with freight axles.

The major objective of the project was to improve the competitiveness of the SMEs in the project by enabling them to provide solutions to reducing the cost of, and/or improving the safety margins of, rolling stock operation. The opportunity now arises because some of options to solve limitations of current inspection techniques and systems (by increasing speed and convenience of inspection and improving defect detection probability, together with a better knowledge of the fatigue crack growth mechanisms) have been achieved.

The consortium SMEs already have presence in the railway inspection market, and will have improved access the EUR 2 billion per annum market for in-service inspection equipment and services in Europe as a result of the project. Access to the International markets will also be possible through this project.

The ability of inspection to allow a prolonged life of an axle will result in a reduced demand for axles, and for steel. Since the service inspection personnel will be European based, and much of the steel is imported, there should be a net improvement in the employment possibilities for European staff.

The market for services in the rail sector is increasing and this suggests that a growing market exists for the WOLAXIM systems. The exploitation of the systems from the end of the project is likely to be different for each result.

The corrosion assessment system, and the associated performance data and axle life methodology, can be exploited with a small amount of investment, but it is necessary to establish the role that they could take within the existing framework of inspection and axle life estimates currently in use in the industry.

The phased array system has sufficiently well advanced to be tested in the field when the opportunity arises. The system has considerable advantages over some of the existing methods, particularly in the inspection speed and the simplicity of the mechanical arrangement. Ideally a development to include other angles needs to be undertaken.

The thermography system probably requires a more powerful prototype, and therefore needs a fair amount of investment before it can be revisited.

The various aspects of the project were disseminated in various ways. The primary vehicle was the use of the European Structural Integrity Society TC24 meetings in April and October 2012, which specialise in issues concerning rail axles and involve representatives of the major rail companies. In addition, papers were presented at the World Conference of NDT, the European Conference on Fracture and the First International Conference on Railway Technology: Research, Development and Maintenance. The partners also presented at the national NDT conferences.

An initial flyer was available at the beginning of the project and has been updated. A video was created and has been distributed on Youtube (see http://youtu.be/4ZL95SJayEM online).

Project website:
http://www.wolaxim.eu

Contact details:
John Rudlin
E-mail: john.rudlin@twi.co.uk
Telephone: +44-122-3899000

Related documents