Final Report Summary - TRAINWHEELS (A novel technology for detecting train wheels surface cracks by non destructive testing based on thermography)
Executive Summary:
TRAINWHEELS is a novel technology for detecting train wheels surface cracks by a Non-Destructive testing. This technology aims to respond to a necessity that is being more predominant in the European rail network regarding the inspection and maintenance requirements. In this regards one of the biggest concerns about transportation safety is broken wheels due to cracks and in fact in the latest years, frequent accidents and regular breakdowns have been registered due to this failure.
The main novelty of Trainwheels is based on the use of the Induction Thermography phenomena, which automatically provides the captured images and identifies the cracks through the employed artificial vision tools and without dismantling the wheel-set. Trainwheels system can be used by maintenance companies, installed in their underfloor wheel lathe, or by train wheels manufacturers in order to detect cracks caused by faults in the production process. In fact the aim of Trainwheels is to improve train circulation safety by enhancing the effectiveness of non-destructive tests on wheels and to provide the early detection of surface cracks. For this reason, Trainwheels is envisaged as a solution to the current visual inspection methods, or as a complementary one to the more spread internal inspection method (Ultrasound testing, which lacks surface inspection capabilities). This hybrid system should be needed to suit the purpose of delivering a complete and self-consistent detection method: Ultrasound to provide a crack detection for the innermost part of the wheel and Induction Thermography to detect cracks on the surface and in the outermost part of the wheel.
A first prototype for the validation of the technology has been constructed with the required integration and communication between the involved components and software tools. The adjustment of the hardware, software and electrical parameters was performed in order to maximize the detection of the cracks. Obtained results from the demonstrator gave very good results and thus a successful validation of the technology. A further demonstration event was carried out in Asturias, Spain (at the end user facilities, TAM), to show the suitability of the system to detect surface defects through an automated and faster solution. The goal was also to show the potentiality of the system to the train maintenance companies and related train transportation sector companies. They could evidence the usefulness of the technology as a surface crack detection system, that may replace the current visual inspection and other surface inspection technologies, reducing inspection times and thus increasing productivity.
The trainwheels solution as a system for the early prediction of cracks of wheel manufacturers will make possible to repair them or improve altogether the new ones, with expected savings of up to 25%. The controlled heating and feedback by the infrared camera and the coupled software, makes the system to run under efficient parameters and operating costs. Sales forecasts for the period 2017-2021 have been estimated in terms of units and in terms of revenue for the three types of target customers. A sales strategy has been also developed which includes a growing segment of sales via licenses to Wheel Maintenance Equipment Manufacturers. Target accumulated sales for the five-year period are of the order of 660 units and 52 M€.
Trainwheels has been depicted to comply with all European Safety Standards. The future commercialization stage of the project involves three consecutive phases which will target at the local (Italy, Germany and Spain), European (EU28) and international (out of EU28) markets.
Project Context and Objectives:
Context and Introduction
The European rail network is getting more and more predominant in the transportation sector: it hosts high speed-trains, higher loads of more passengers or goods, with increased frequency. This determines a higher mechanical load for the moving parts of the carrier involved: therefore, increasing inspection and maintenance are required in order to guarantee safety and well-being of trains. One of the biggest concerns about transportation safety is broken wheels due to internal cracks. In fact frequent accidents and regular breakdowns have been registered in the latest years due to this failure.
Three testing methods have been investigated to identify cracks or problems on the surface or near the surface of the wheel: Magnetic Particle Inspection, Dye Penetrant Inspection, and eddy current testing. These methods show the following drawbacks: MPI and DPI generates high amount of hazardous waste and none of them is automatable, while the eddy current system is sensitive to lift-off variations and probes need to be positioned at a constant distance. In conclusion a new, more efficient, automated and faster method is needed to detect surface cracks.
In addition the most of the internal cracks usually begin on the surface of the wheel, it is therefore urgent to deliver a technology for superficial inspection and this vital task can be accomplished by a system based on Electromagnetic Induction phenomenon, as Trainwheels. Nowadays, maintenance technicians use Induction thermography to locate overheating joints and sections; but we require a more sophisticated system.
In fact thermography crack detection method would mean a big improvement in surface and near-surface train wheel crack control, and therefore minimization of the related accidents. Induction thermography is going to be part of a system which aims to establish a more competitive and safer railway network. Through the application of this technology as a detection method, it will be possible to gain an increased security level and a lower wheel and maintenance cost. This system is mainly oriented to be used by railway maintenance companies or by train wheels manufacturers.
Railway technology is a specialized market with specific requirements, which result from high safety standards linked to the transportation of passengers and goods. Technical experts with the required certifications are the basic capital of the companies operating in the rail industry. In a relatively restricted market with a limited total volume, very high entry barriers exist. In Europe, however, with the liberalization of rail transport, companies can now access to previously closed markets.
By using thermography, it is possible to create a new and more efficient way of train wheel surface inspection and maintenance, which is necessary if we take a look at the accidents statistics resulted in the latest years. It is therefore necessary to improve the efficiency of train wheel maintenance procedures, because they have to be inspected at regular intervals for internal and surface defects. Ultrasound system is the state-of-art technology to detect internal cracks, but thermography would mean an automated method to detect surface and near surface cracks, with a considerable lower global cost compared to other surface crack inspection methods.
External costs are a side effect of transportation, which is not internalised into the price paid by the user and is therefore not taken into account by them when they make a transport decision. Electromagnetic induction Thermography makes the defect detection process more efficient, due to the fact that is an automated control process (time reduction compared to manually operated processes) and does not generate any waste in comparison with other surface inspection methods. Thermography as non-destructive method is a fast way to detect surface and near surface cracks and does not require great operator specialization because easy software is used to interpret the results.It is not only a more advanced and automated method to detect cracks, but is even faster and easier to use. By mean of an infrared camera and powerful software, the operator is able to detect surface cracks on screen, without needing new maintenance training. Environmental benefits can be achieved because of the fact that thermography does not generate any waste, in comparison with MPI and LPI methods which generate contaminated magnetic particles and liquid; that means savings in waste treatment and time inspection. In conclusion, it is a greener and cleaner way to detect surface cracks.
Thus Trainwheels is presented as a solution based on automated Induction thermography crack detection, and thus on the physical phenomenon of the electromagnetic induction, a far know - yet very powerful - technique that was first developed as a thermal treatment for the hardening of metallic components. An alternate current generator provides an oscillating current at a specified frequency, flowing into a conductive coil which is shaped such mirroring the train wheel. The coil is positioned as close to the wheel as possible and the varying magnetic field in its cross-section produces an electric field in the metallic material of the wheel. This electric field translates itself into a current, flowing only in the outermost shield of the wheel (for the so called “skin effect”) which, due to its electric resistivity, is then heated. Cracks, if present, will disturb the current flow and so generate changes in the temperature profile in the crack area. These changes of temperature are visualized using an infrared camera. The image acquired by the infrared camera is evaluated through an image processing system.
The potential benefits for SMEs involved in the Consortium are:
o A decrease in accident occurrence of up to a 25% from current yearly rates
o Savings up to 25% for wheel manufacturers who, through an early prediction of cracks in their wheels, will be able to repair them or improve altogether the new ones.
o Savings of up to 20% in the induction process, because the heating will be controlled and feed backed by the infrared camera and the coupling software, in order to run thermography under efficient parameters and operating costs.
Objectives
Trainwheels, based on the Induction thermography phenomenon is presented as an automated train wheel surface crack detection method that will meet all European Safety Standards. Induction thermography as a non-destructive test method is already used in other sectors of industry. Using thermography for train inspection would result in a big improvement compared to previously quoted methods: as an automated and faster approach as well as cleaner and cheaper.
Thus the challenge of Trainwheels is to optimize Induction thermography in order to use it as an automatic train wheels surface inspection system.
For this purpose firstly, an induction generator will be improved to heat up the wheel by passing electric current through it. By means of a thermographic camera, a scan of the wheel will be provided in chromatic scale. The temperature profile is shown on screen and the presence of a crack will be highlighted by a hotter spot on the image. For this reason, it is also necessary to provide modern and easy-to-use software that shows the profile of the emitted radiation on the screen. The development of a new sensitive infrared camera will open new fields for thermography applications.
Following the technical as well as the scientific objectives of the project are described. The specific goals as defined by DOW are also included.
In general terms Trainwheels aims to achieve the following goals:
- To develop a method to detect train wheel surface cracks based on thermography. Faster and cheaper for at least 20% than previous methods.
- Up to 20% of reduction in inspection times. Energy savings up to 20% from current induction systems
- Develop and optimize an induction device. To obtain a novel application for thermography applied to train wheels inspection
- Develop and optimize an infrared system
The scientific objectives of Trainwheels, according to the needed research and development task are:
- To gain understanding of train wheels operations, railway environments and behaviour of materials under those circumstances. To be able to develop a thermography system as quick and effective as an ultrasound system
- To increase the understanding and physics involved in the induction system by simulation (Coil dimensions, configuration, relative positioning, electric power, voltage and current, inspection time and heating process
- Increase understanding of thermography and thermal artificial vision. Up to 20% increase in defect detections.
From the project development point of view, the objectives are specified and distributed in the following work packages:
WP1. System Characterisation.
Define and characterize the type of wheel set systems available on the market and their properties (material magnetic and thermal properties). Define, characterize and classify the types of wheel surface cracks. The following deliverables must be available:
WP2. Electromagnetic induction heating system design.
Define and evaluate through simulation, the electromagnetic induction physic and thermal transfer pattern, for different configurations according to the data obtained in WP1 and WP3 (type of cracks and type of thermographic camera). In a second phase the goal was develop an optimized induction system for an optimal heating of the train wheel surface.
WP3. Thermographic methods.
Evaluate available thermography methods to find the best suitable parameters and configuration. Test and design a suitable inspection system consisting on an infrared camera and suitable algorithms for the software. Measurements and analyses of characterized failures was also done to proof the capability and suitability of the developed system.
WP4. Artificial Vision.
The objective of this work package was to develop and design pre-processing algorithms to increase signal quality. Implement state of the art data and design new computer vision algorithms to recognize and classify defects in thermal images. This will led to the development of an automatic result analysis system, to assist personnel to interpret and categorize the cracks.
WP5. Engineering
Definition of the boundary conditions for an industrial environment in the railway sector. Simulation and construction of the testing system mechanical assembly. Manufacturing support for the prototype production and iterative redesign measures.
WP6. Technology Integration and Validation.
The first part of this Demo work package was dedicated to the assembly and pre-testing of the prototype at TERMO’S facility. Once shipped to TAM’s premise, the second part will aim at the integration and validation of the components, the monitoring system and the whole TRAINWHEELS system by the Consortium partners.
WP7. Dissemination of knowledge
The purpose of this work package is threefold. Firstly to develop an Exploitation plan, made the SME consortium and led by the Exploitation Manager. Secondly, after signing a Consortium Agreement between the partners to define and protect the future Intellectual Rights that will derive from the R&D activities, the most important results obtained will be protected from the competition. Finally a Dissemination strategy will was elaborated to communicate the results within the Consortium and to exploit and popularize the TRAINWHEELS results to the scientific communities and to potential end-users.
WP8. Consortium and Project Management
The main objectives planned for WP8 are the following:
• Coordinate knowledge management.
• Coordinate the production of deliverables, milestones reports & cost statements.
• Coordinate legal, contractual, ethical, financial & administrative issues of the consortium.
• Organise and coordinate the Project Management and Exploitation Board Members meetings.
• Coordinating payments, audits and distribution of money.
• Coordinate communications between the consortium and the EC.
• Coordinate the preparation and signature of the Consortium Agreement.
Project Results:
The overview of the progress of the work to gain the aforementioned objectives is presented here, in line with the structure defined by DOW (WP and corresponding task). The most significant results are highlighted.
System Characterisation
A in-depth study of the wheels geometry and material properties was performed comprising the typical defects and cracks.
Different information sources were used to obtain the most complete current situation:
- Directives from the UIC “Union Internationale des Chemins de Fer”
- Railway Group Standards
- British Railway Standards
- Several European Norms and Standards (UNE)
- Relevant scientific and technical papers
- Visits to key actors in the sector: wheels and trains manufacturers, maintenance companies, railway companies and potential end-users (Talgo, CAF, Nertus, Danobat railway division, etc.)
The work performed comprised:
- The description of the train wheels and axles from a geometrical and physical point of view: applications, figures, thermal properties, electromagnetic properties, etc. It will be very relevant for the following work packages concerning simulation.
- The geometrical characteristics and classification of surface cracks was done: type, dimensions, inspections, etc. This data will be very required for the “Artificial Vision” Work Package, where the artificial vision software to detect failures will be developed.
- Relevant information from important Railway actors during several visits made by Inspiralia and TAM is reported.
- A market study was performed in parallel to the technical work, in order to help us defining the final TRAINWHEELS solution and at which market will target.
The main information and data obtained regards the railway network statistics and the contemporary situation analysis. One of the most important conclusions stands on the revolution of the infrastructure and its consequences. The European rail network is getting busier every day with high speed-trains, more passengers, higher loads, and an increased frequency. These factors produce heavier axle and wheel loads with the subsequent infrastructure inspection and frequencies reformulation requirement. In fact according to European Directives on Safety, one of the most important indicators relating to precursors of accidents are broken wheels; even though severe rail accidents are relatively rare within Europe, their frequency is still at an intolerable level, and at a second level network disruptions and breakdowns may cause several social problems.
Thus increasing and improving inspection and maintenance are clearly required in order to guarantee the maximum efficiency and safety as possible. No clear definition of the required inspection technologies, beyond the required ultrasonic method for internal defects detection, is defined on the standards. The companies used to employ different type of inspection technologies. The most common surface inspection methods found are: Magnetic Particle Inspection (MPI), Dye Penetrant Inspection (DPI), and Eddy current testing. These methods show the following drawbacks: MPI and DPI generates high amount of hazardous waste and none of them is automatable, while the Eddy current system is sensitive to lift-off variations and probes need to be positioned at a constant distance. In conclusion a new, more efficient, automated and faster method is needed to detect surface cracks.
The most important points comprise:
Wheels classification by: configuration, brakes, dimensions, and material:
- Regarding the configuration two types exist: solid or monobloc wheels (wheels manufactured in one piece) and tyred wheels which are manufactured by several pieces. Tyred wheels allow to replace only the tyre instead of the whole wheel, but are less reliable and in fact are actually forbidden for high speed trains. Both types are representative for Trainwheels
- Two types of brake systems are identified. The first one is the brake shoes which applies the load directly to the wheel tread the flange. These brakes are mainly used on freight car wagons and they produce thermal fatigue loadas on the tread which may induce cracks on the surface. Thus this type is the most relevant for Trainwheels. The second type are disk brakes which press the rim of the wheel and thus surface defect will not be as frequent.
- Dimensions (Diameter and Profile) of wheels are defined on standards and must be respected for safety reasons. The profile is a key factor as it defines the geometry to be heated and tested.
- Materials are clearly defined and specified on standards. For Trainwheels we must consider ER7&ER8. Since these specifications do not take into account electromagnetic properties, these must be defined early on following tasks.
Cracks classification based on location and size:
- Regarding the location the most frequent cracks are located (Figure 2) at end face & tread (the most important), and at clamping places and at marking places too.
- The maximum size allowed for surface cracks is defined by any of the dimensions being 2mm.
Market study. The study comprised two main points: the type of inspection required to understand where to place the system and the type of train (freight or passengers) to determine our client target.
- By type of Inspections the following were found:
_Routine inspections: mounted wheels, train in movement, automatic, often every 15 or 30 days.
_Complete inspections: disassembled wheels, in special shops or pits, with operator presence, and every 10^6 Km
- Regarding the inspection stage, different strategic options apply: at manufacturing and/or reprofiling stage (pit lathe), where the wheel will be clean and after which no cracks should be found, during maintenance inspections which depend on company’s strategies and for manufacturing companies as a final inspection stage.
- By type of market we have found different scenarios and constrain:
_Passenger trains: Inspections in passenger trains are more restrictive due to, traditional mind, much more complex design and low accessibility, and more common interior cracks (fatigue).
_Freight trains: inspections in freight trains show more opportunities in the market thanks to: new regulations in the way, Wagons and wheel-set design allowing better accessibility, low inspections levels (frequency and control), and importance of surface cracks detected due to thermal loads.
This data was the starting point for the following work on the project and an important basis to perform the design of the electromagnetic induction heating system, together with the artificial vision component and the thermography methods.
Electromagnetic induction heating
Computational simulation was employed for the design of the inductor geometry and electrical parameters, required to improve the crack detection as well as the efficiency of the induction system. A deep study was carried out with a number of different models performed for the purpose:
- Initial model: simple metallic plate and a simple coil to study the different possible configurations and to better understand the physics.
- Validation with experimental test to validate the simulation tool
- Heat diffusion study, of the wheel was carried out when induction is stopped.
- Rotating wheel study. To simulate the real Trainwheels case
Initial simple model
Considering the parameters obtained by the Initial studies (crack dimensions, location and type) the first numerical models were developed. These simulations are based on a coupled electromagnetic-thermal problem that were initially carried out in a simple metallic plate, to understand the physic and its influence under different conditions: electrical parameters, geometric positions, material properties, crack orientations and dimensions, etc. (Deliverable 2.1). Results of the temperature patterns were obtained which in general terms show the maximum heat on the tips and colder flanks as shown in the following (Figure 3).
The most important conclusions obtained from the simulation is that the crack will show up more clearly on thermal gradient solutions than on thermal maps images. Moreover, the border effect is eliminated on the thermal gradient results, showing its maximum on the crack tips. The thermal gradient plot will highlight the crack onto the surface at every time step. However small time step will be preferable, before the heat diffusion starts, which increases very rapidly (Figure 4).
From these results it was concluded that the optimum electrical parameters combination is that with the highest frequency (400kHz for the given FHG inductor) due to the more evident crack detection and the most efficient use of power. This goes related to the fact that the frequency increase will generate a smaller skin depth and thus the current will be forced to flow through a smaller depth, which will be translated into a higher current concentration and thus higher temperatures. From this initial study it was also concluded that any kind of crack dimension (with a minimum of 1mm in any direction), location and orientation can be detected at high frequency electrical inputs.
The final electrical inputs and final coil geometry for the inductor generator were derived from these simulations, for which the impedance value was required to design the capacitor. This value will allow defining the whole system electrical behaviour as an integrated generator-capacitor-inductor system. In fact simulation was a mandatory tool to predict the cracks detection behaviour.
Validation
A validation to prove the suitability of the FEM model to predict the behaviour of the system was carried out, thanks to the experimental test performed by FHG. This will be essential to prove the consistency of the tool for future analysis.
Very accurate and similar results (temperatures in the crack around 34°C and around 27-30°C in the surroundings between the coils), were obtained from this comparison Figure 5, which led to a qualitative validation of the electromagnetic model. This approximation is due to the fact that the results from the thermographic camera cannot be used to read the exact temperature, in a high extent due to the image noise, but an overall thermal map can be obtained as shown.
Hence, further details around the crack and tips of the crack are difficult to validate also due to the resolution or the camera and the fact that the sequence image composition, will not provide accurate temperatures from the IR maps. Simulation provides also the temperatures inside de crack, on the plate deeper area and thus not only on the surface as it does the camera. The IR signals from this deeper area are not received from the camera, and thus only the higher temperature of the edges will be detected.
Heat diffusion study
Once computational time was reduced (through the wheel geometry simplification study, and using just a portion of the wheel), an initial study of the crack heat diffusion was done for a static wheel. A simple coil and a crack were modelled and the temperature on the tip of the cracks along time was obtained.
If the induction is stopped the heat diffusion will occur rapidly. It will depend on the electrical parameters, etc. but in around 0.5s the temperatures on the target will be again near the ambient temperature.
The temperature along the crack was also plotted (Figure 7). The highest temperature jump is observed between the tip and the inner of the crack and the thermal gradient between the tip and its exterior surroundings will be much lower (blue arrow). In a very short period of time the spatial thermal gradient will decrease in a high extent. However this highest temperature jump won’t be noticed by the camera since only top surface IR emission and temperatures will be detectable (proven in the experimental test). Thus the thermal gradient between the tip and its exterior surroundings will be the data that the camera will be able to read, in which the spatial thermal gradient will be lower.
Rotating wheel model
The simulation of a rotating wheel model was performed to replicate the final model. The same model was used (Figure 6) and the electrical parameters were those to achieve the maximum values allowed by the generator (400kHz and 5kW). The thermal gradient (Figure 8) for the different time steps showed the influence of having the coil above the crack (thermal gradient maximum value).
Thus the optimum place to detect the crack would be, right above the coil, but this will not be possible since it would be covering the wheel. Higher wheel angular velocities (0.5rpm) were also tested in which the heat influence along time was higher and thus the crack detection improved.
Coil geometry study
To obtain the optimum coil geometry different analysis were performed. Various coil distances and orientations were studied in which the magnetic cancellation in a two parallel wires configuration was observed (Figure 10). The U shape and the following shapes were studied to check the different influence on the piece of the coils with different direction currents.
The induced current densities were plotted and higher values for lower distances were observed. With the multiturn configuration a 5% of higher values were obtained. For the final coil geometry design the way in which the crack will be more obvious had to be also considered and thus a pattern that assured the detection of every kind of crack (dimension and orientation) was studied. Since the induced current distribution must be high in order to be able to detect it at the distance at which the camera will be positioned, the multiturn case was preferred. The final coil design was a pancake shape inductor to detect 360º craks (Figure 11). With this design the thermal shift was increased and thus a longer detection capability is provided (Figure 12).
The final coil design was carried out following the pancake shape as stipulated the optimum one but with squared shape, in order to provide an homogenous patter along the whole tread.
Main conclusions
The main conclusions obtained from these models are:
- Initially the optimum electrical parameters will be based on the maximum frequency (in this case and for the TERMO induction system 400kHz). The intensity to be supplied was defined together with the final coil geometry (2kW). These parameters were then optimized based on the results obtained from the first induction test and obtained images. There it was proven that due to the complex shape of the inductor and electrical conditions not was not possible to achieve such a high frequencies. Possible frequencies were 50kHz an 150kHz. As it was then noticed the lower frequencies generate higher signal to noise ratio and thus that was preferred to obtain a clearer detection.
- The temperature jump due to the current density concentration is observed right under the coil. Very fast heat diffusion is observed for very small time steps.
- The most important result that will highlight the crack is the spatial thermal gradient pattern, and to be applied in the artificial vision software.
- Higher thermal gradient was found within the crack, but the deeper values will not be detected by the camera and thus only surface values and thermal gradients between the crack edges and surroundings should be considered.
- Magnetic field cancelation is lower for lower wire distances. A pattern that assures the detection of every kind of crack dimensions and orientation was designed, with a pancake geometry.
- Simulation provided the information required to predict how the cracks will be observed trough the camera, saving time and energy. Thus the most appropriate electrical parameters, the optimum inductor geometry and other very interesting issues were obtained from the analysis, to finally design a system which assures a safe and secure crack detection
Thermography methods
The central part of the Trainwheels thermography system is the thermal camera which will be used for data acquisition. The properties of the thermal camera determine the quality of the data that will be available for the automated crack detection and categorization with the software algorithms.
After some tests with different cameras which fulfilled these specifications for the Trainwheels system the Xenics Gobi 640 CL was the chosen thermal camera, because this camera fulfilled all specifications and seems to deliver the best additional features for the Trainwheels system.
Technical data of the selected thermal camera Xenics Gobi 640 CL can be found in D3.3 and the second review report. First the chosen camera was compared with the camera used for the basic tests. Therefore the signal quality was evaluated with the calibration body (Figure 15). It can be seen, that the Trainwheels camera delivers a sharper image compared to the camera used for the initial test. This should be self-evident because of the better resolution.
Analyses of the amplitudes of the reference cracks show that the noise of the Trainwheels camera is higher than the noise of the camera used for the initial tests, but also the amplitudes are higher. For the images acquired under the conditions as described above, a signal noise ratio for the Trainwheels camera of 2.4 dB can be calculated and 2.6 dB for the camera used before.
Induction Frequency
As the theory of the electromagnetic induction tells, the excitation frequency determines the skin depth of the inducted eddy currents. A lower frequency leads to a deeper skin depth but with a lower eddy current density. For the crack detection it is important to penetrate as much as possible with the skin effect inside the material but also it is important to generate an eddy current density, which is high enough to generate thermal effects on the cracks that can detected by the camera.
For testing the influence of the induction frequency the Thermomacchine AURORA induction generator was modified in the way that it can be operated at 50 kHz and at 150 kHz. The skin depth of steel is about 0.06 mm at 150 kHz and about 0.1 mm at 50 kHz. A comparison of the influence of the inductor frequency is demonstrated (Figure 17).
The lower frequency delivers an image with more contrast also with lower induction power. With the high frequency at low power heat is only generated at the edges of the reference cracks. Most of the energy is concentrated at the surface and heats up the surface. The contrast increases with the power and above a power of 40 % the surface effects are smaller in comparison to the Joule effects at the cracks. It seems that then also crack flanks are heated by the Joule effects and the thermal contrast rises. At the lower frequency the thermal contrast between the reference cracks and the surface is quite better also on low power. The inducted eddy currents penetrate deeper inside the material because of the higher skin depth and the crack is heated at is flanks also on low power.
The signal analysis along the crack lines illustrates this very clear. At high frequency most of the induction power is used to heat up the surface and there is a bad signal noise ratio for a high frequency (Figure 18) at low power. At the low frequency the surface isn't heated up as much but the flanks also generate heat so we get a good signal noise ratio already on low power. When the induction power increases the surface isn't heated up more than on low power but now also the crack flanks generate heat and the signal noise ratio is getting even better.
Table 1 shows the calculated signal noise ratio at 50 kHz and 150 kHz induction frequency, being higher for low frequencies and high power.
Induction Power
Beside the induction frequency also the needed induction power is of interest for the project, because another aim of the Trainwheels project is to keep the costs low what means that also the used energy for testing has to be kept low.
After finding out that the lower frequencies of the induction system are leading to better results, the influence of the induction power was studied. For this the signals of the reference crack of the calibration body were analysed at induction frequencies of 50 kHz and 150 kHz while the induction power was varied.
The images (Figure 19) show that there is a good thermal contrast already at 10 % induction power at the frequency of 50 kHz. But there are also indications of scratches in the calibration body’s surface which are caused from manufacturing. These scratches disappear above an induction power of 30 %, so that only the reference cracks and some artefacts caused by differences in the emissivity of the calibration body’s surface can be seen. The images on the right, for which an induction frequency of 150 kHz was used deliver useful data not before an induction power of 30 %.
The analysis of the signal along the lines of the calibration body which are containing cracks are given (Figure 20). At an induction power of 10 % there is already a good signal noise ratio when using 50 kHz excitation frequency, and the signal noise ratio is increasing steadily until 50 % of induction power. For 150 kHz a signal noise ratio above 2, which allows to distinguish a signal properly from the noise, cannot be achieved below an induction power of 30 %.
The signal noise ratio is given for different induction powers at the frequencies of 50 kHz and 150 kHz which are available by the Trainwheels induction generator (Figure 21). It can be seen there that the signal noise ratio at 50 kHz is always higher than on 150 kHz independent of the power.
Wheel Velocity
The velocity of the wheel influences the measurement results. The relation between the cameras frame rate and the wheel velocity is given by:
1 1
The aim of the Project is to test a wheel in one minute or faster. This means a circumference velocity of 50 mm/sec is at least needed to check a wheel of about 900 mm diameter in one minute.
To study the influence the calibration body was moved with different velocities during the measurements. The measurement conditions were the same as in the tests with the induction parameters. For data acquisition the Trainwheels thermal camera and the basic Trainwheels software was used. The results of these tests are shown as line scans (Figure 22).
In Figure 22 can be seen, that until a velocity of 100 mm/sec there is no big influence to the images. Above a velocity of 100 mm/sec the cracks are starting to smear and also the edges are blurring. Also a loss of signal in the angular reference cracks of the calibration body can be registered.
In summary the analysis of the influence of the circumference velocity proofed the consideration done in deliverable 3.1 where a velocity of 50 mm/sec was calculated when using a camera with a resolution of 384 x 288 pixels. Now a camera with 640 x 480 pixels was used which leads to a faster circumference velocity of 100 mm/sec which means that theoretically the testing of the wheel can be done in about 30 sec when one round for testing is needed (2rpm).
Artificial Vision
Graphical Software User Interface
The software user interface can be seen below (Figure 23). The software was designed to be simple to use for its main task: executing measurements. On the left, the user enters certain data related to the wheel under test or to the test itself: examiner name; client name; train, wagon and wheel number; wheel diameter. The diameter is not only stored in the result file for information purposes, it is also used in translating line numbers into distances in millimeters. Finally, the user has to enter the line number that is used to construct the unwinded image (see section Image Processing).
On the top, the user can activate and deactivate a live image of the camera and manually execute a non-uniformity correction. Most importantly, when the live image is activated, it is possible to initiate an automatic measurement.
Data Acquisition and communication
The Trainwheels system consists of five main components which had to be controlled by the PC (except the mechanical system) that they work together proper. Those components are:
• Induction system
• Thermal camera with frame grabber
• Wheel set drive
• Positioning monitoring
• Mechanical system
Figure 24 gives an overview how the subsystems are connected and controlled by the PC. The communication respective the signal flow between the subsystems was realized by an IO card which delivers the signals and information to the software module which controls the workflow during the measurement. The signal flow is shown in Figure 25. The control parameters as well as the measurement and image processing parameters are stored in a configuration file which can be adapted to each specific inspection task
Image Processing
Once the measurement is started and the trigger sensor passes for the first time, the camera captures frames with a rate of 50 frames per second for one whole turn of the wheel, until the magnetic sensor passes for the second time. In every captured frame, only one line – whose vertical position in the image is entered by the user – is actually used in further processing. Hence, this line is cut out of each frame and stored in the main memory; the rest of the frame is discarded.
When the magnetic sensor passes for the second time, the data acquisition finishes. The stored lines are assembled into one single image which presents an projected 360 degree view of the wheel (Figure 26). All image processing is done based on this projection image.
The carried out image processing algorithm is as follows: First, a line segment detection is executed. The used algorithm is the “line-segment detector” (LSD) as described in von Gioi, Jakubowicz, Morel, Randall: “LSD: A Fast Line Segment Detector with a False Detection Control”, IEEE Transactions on Pattern Analysis and Machine Intelligence (2010), and implemented with tiny variations in OpenCV in the class LineSegmentDetector. The result of the detection is a set of lines, each with a starting and an ending point.
Second, each of these lines is treated as a candidate that has to fulfil further conditions to be finally treated as a defect. The line length of a defect line must be equal to or larger than a certain minimum length (first condition). Then, a pairwise comparison of all candidates is executed and for each line that is to be considered a defect, another line must exist that is parallel (second condition) and not too far away (third condition). Lines that meet all three of these conditions are considered defects and drawn in red, lines that remain in the candidate stage are drawn in yellow (Figure 33).
Technology integration and validation
The Trainwheels prototype integration and validation was carried out. It was mounted at TAM facilities (end-user) where the integration of the different components and final test was carried out. The components and assembly were validated in a lab environment and the successful results are highlighted here.
Prototype integration
The mechanical prototype design with the detailed drawings was described in deliverable 5.2. The main characteristic and purpose of this design is to provide a flexible system that can be mounted both on the top or the bottom of the wheel. The goal to be installed on the top of the wheel is to provide an easier handling, for the prototype testing that will be carried out at TAM.
However the real configuration will be that installed on the bottom of the train wheel, which responds to the necessity of the companies to inspect the wheel in an automated manner. The system starts to inspect the wheels automatically once the train enters in the inspection station, and thus it can be mounted directly in the rails of the inspection line.
The mechanical parts were manufactured. The components of the system are very briefly described here (more details are given in Deliverable 6.1):
- Mechanical frame
- Thermographic camera
- Generator and Induction system
- Wheel turning motor
- Electrical box
- System control and Computer hardware and software
- Encoder
- Chiller
- Electrical connections
An initial pre-prototype was integrated at FhG (Figure 29). The assembly of the system initially performed in Fraunhofer served as a first validation, from which some initial results were obtained.
The final integration for demonstration purposes was carried out at TAM (Figure 30), the end user of the technology. There the final wheel was mounted (provided by Arcelor), which corresponds to a Freight train. All the components were integrated and connected (Figure 31).
Results and validation
The most important criteria to proof the functionality of the Trainwheels demonstrator and to show that the principle of inductive thermography can be used for the detection of surface cracks on railway wheel is stated here.
Data acquisition and evaluation
The data acquisition concept based on multiple line scanning and use those lines for creating a projection around the circumference of the wheels. This concept was chosen to get spatial information as well as time dependent information about the crack. The tests showed that this concept principally works, but therefore the rotational speed of the wheel has to be constant, otherwise the speed deviations lead to images which can’t be matched and the time dependent information will be lost. Unfortunately with the wheel set at TAM facilities it wasn’t possible to get the needed constant rotational velocity, because of the damages and impacts of the wheel set. So for data evaluation at TAM no time dependent information was available and the data evaluation could be done only with the spatial information. Nevertheless also on TAM’s wheel set the system could detect the cracks only by optimizing the evaluation parameters of the spatial image evaluation.
The tests also showed that the aimed inspection period of 1 minute is possible. The tests were done with a circumference speed of 60 mm/s this means an inspection period of 50 s for a wheel set with a diameter of 930 mm.
Software
The software interface was designed in way that a minimum of user interaction is needed. The GUI allows the operator only to setup metadata for the inspection report, to start and stop a measurement or to load a former measurement. The operator has no direct access to the measurement or evaluation parameters. These parameters are stored in a configuration file which can be created for each specific inspection and reused for repeating inspections. This ensures that the parameters can’t be changed during the inspection by the operator which is conforming with NDT standards where only administrators can change such parameters.
To give the operator information about the subsystems a status bar is available which indicates warnings and errors from the subsystem. During the measurement the main window of the software shows the live image of the thermal camera so the operator can follow the measurement what might be helpful for the result interpretation. After the measurement has finished the software report is generated automatically as a pdf file and the evaluated thermal image where the evaluated cracks are indicated is displayed. The software allows the operator to rotate the wheel that a crack will be moved under the camera by right clicking on that crack. This is also helpful for the result interpretation.
What is missing on the software is that the operator cannot change an obvious false indication in n the results or comment them.
Induction system
An aim of the project was to decrease the power consumption for in relation to the crack detection with magnetic particle testing. A modern standard magnetic particle testing device for parts up to 100 kg and a maximum diameter of 400 mm has a power consumption of about 25 kW to 30 kW. But such a device cannot be used for the wheel set testing (wheel weight 300 kg, diameter > 900 mm). The Trainwheels demonstrator has a maximum power consumption of 20 kW and was used at 30 % to 40 %. So the effective consumption is about 5 kW for wheel testing.
The for the Trainwheels system developed coil allows to test in all three inspection positions. There is no need to change the coil between testing on the tread or on the side faces. This makes the demonstrator very time efficient. Also the coil and the induction system are very robust against deviations in the distance between coil and wheel surface. During the tests no influence on the results could be asserted when the distance differs of ± 1 mm.
Results
The first results show a promising detection potentiality. There the differences between a clean wheel (brushed wheel from FhG) and a very used wheel with natural cracks (form TAM) were observed (Figure 34).
The smaller and sharp-edged defects of the Fraunhofer wheel set generate a much better thermal indication than the defects of TAM’s wheel set. The reason therefore is that the sharp-edged defects are leading to a higher eddy-current density and thus to more heat generation at the defect. Nevertheless, the Trainwheels thermography system is able to detect the cracks on both wheel sets. Figure 35 shows a comparison of the result images of the two wheel sets, evaluated with identical evaluation parameters. The red markers are indications classified as a crack, the yellow markers are indications which were detected by the software but not classified as crack. The reference cracks are marked blue. In Figure 35 can be also seen, that the pitting around the surface of TAM’s wheel set produces a lot of false indications, because the edges of the small pitting craters act like a crack. The false alarms in the flange area of the Fraunhofer wheel set are caused by 0.5 mm to 1 mm deep rills from the lathe form feed.
The evaluation software allows to decrease the rate of the false alarms by setting up these parameters more coarse. Figure 36 and Figure 37 are showing the difference of a very sensitive and a coarse setup of the Trainwheels system evaluation parameters. In Figure 36 can be seen, that it is possible to eliminate most of the false alarms in the flange but to keep the defects. The same can be seen in Figure 37 for the TAM wheel set.
In the context of the detection possibilities on the TAM wheel set should be mentioned that railroad experts told, a wheel set like the one that was used for the tests at TAM’s facilities shows the worst case and in common a wheel set like this would never be inspected without t re-profiling before.
About the reproducibility of the Trainwheels thermography system can be stated that the reproducibility of the crack detection is given. In about 30 tests during the testing period the all cracks were detected when the system was set up correct and the cracks were representative for natural cracks. There is only a slight unsteadiness in detecting radial cracks on the tread when the cracks are 100 % parallel to the side faces over their complete length.
Conclusions
The validation showed, that the realized demonstrator fulfils the aims of the Trainwheels project. The main purpose for the project, to use inductive thermography as an alternative method to the classic NDT methods for crack detection on railway wheel sets was reached. With the demonstrator can be proofed, that the inductive thermography generates comparable or better results as the classic methods. Also the advantages of the inductive thermography like no need for cleaning, automated measurement evaluation etc. can be shown.
In summary can be stated, that with a few changes the demonstrator can be used under industrial conditions for crack detection on railway wheel sets. Some of these changes are:
• Automatization of the manual axes.
• Modification of the wheel drive to reach a constant rotational speed also on heavy used wheel sets.
• Some improvements in the mechanical structure like:
o Modifying the vertical stoppers of the self-centring.
o Decreasing the weight of the system to make the moving easier (i.e. possibility of undocking the induction unit during docking the system to the wheel set).
• Some software changes like:
o Add the possibility for the operator to comment and correct false alarms in the result.
o Adding additional features for additive manual evaluation.
The most important improvement will be the modification of the wheel drive, because this will offer the possibility to use time dependent evaluation algorithms and thus a gain in information which can be used for increasing the crack detection performance.
Potential Impact:
Dissemination activities
TrainWheels has shown an important commitment to disseminate the project results through different channels: Internet presence has been consolidated by keeping updated regularly the project website and by including information in the different websites of the consortium members. Several press releases have been published in different media with the participation of all the consortium members, as an activity that continued during the second project reporting period. We also approached relevant actors of the railway industry in the target markets such as Talgo, Nertus, Danobat, Arcelor and CAF, to which the product has been introduced.
The dissemination activities are summed up here (more details in deliverable 7.6):
- PROJECT WEBSITE
- INTERNET PRESENCE
- CONFERENCES, SEMINARS AND TRADE-FAIRS
- PROJECT BROCHURE
- PRESS RELEASE
- PROJECT VIDEO
- DISSEMINATION AMONG POTENTIAL CUSTOMERS
The seminars, fairs and conferences attended are listed here. There the Trainwheels technology was shown and the information was made available for the public
- AWZ Seminar in Radsatz. April 2014
- 8th Symposium of Non Destructive Testing in Railways in Wittenberg. March 2014
- London Rail Infrastructure Exhibition. May 2014
- Innotrans 2014. September 2014
- Rail Vehicle Conference Rad-Schiene in Dresden etc. September 2015
A scientific article was also published at the Rad-Schiene Conference proceedings and a presentation was delivered to explain the most relevant aspects of the technology and the main project achievements.
A project brochure and summary presentation was prepared and used in dissemination and communication events to attract the audience and keep their interest in its highest level. For this purpose, a video presenting a solution in a very intuitive way and easy to understand was created and included among our dissemination material.
A technology demonstration event was held on October, 9th 2015 with the participation of a wide audience among members of the railway industry, potential customers and other project stakeholders. This half-day TRAINWHEELS technology demonstration event which was held at TAM facilities in Avilés, Spain allowed Trainwheels technology and prototype to be shown to potential customers and interested parties by members of the technical and commercial teams of the consortium partners. The participation in the event was numerous and the results of the project were discussed with great interest by the audience (Figure 34).
The consortium is satisfied with the level of communication and dissemination achieved. The activities performed allowed for relevant interaction and feedback from potential customers and other market agents that are considered a key factor to the future development of our commercial and partnership strategies.
Exploitation plan
The exploitation activities performed during the TRAINWHEELS project, include: the results of the market analysis performed by the project partners, the projected sales forecasts, an estimation of the investments ahead, an implementation plan for product exploitation and a business plan with conclusions on the project financial projections, etc (more details in deliverable 7.4).
TrainWheels project have allowed the consortium partners to develop, and successfully test TrainWheels technology by the construction of a system prototype that have proven that the induction thermography technology is valid for train wheels surfaces crack detection. Potential customers have been identified and the target market for the product has been defined. Needs and requirements in terms of quality, costs and efficiency have been investigated by means of interviews with a selected group of potential customers. The system performance up to date has given the consortium partners sufficient confidence on the system capability to satisfy market needs. A complete analysis of alternative options in the market has shown the product will be competitive and the TrainWheels target price has been estimated.
Goals for TrainWheels market penetration have been revised from previous versions of this report, based on the latest outcomes of the market and competitive analysis performed. Team competences have been revised and the necessary partnership strategy for future TrainWheels development and exploitation has been identified. Moreover, two potential partners have already been identified to collaborate with the consortium in the future construction of TrainWheels demonstration units to test the product in industrial environment with real data and real trains.
The consortium strategy for product exploitation has been unveiled. It consists of a 4-phase project, starting with a demonstration phase and three commercialization stages.
The demo phase allowed the partners to achieve the necessary references for product introduction to the market. As a very conservative market, particularly when safety is involved, the railway industry, and the partners themselves, need to achieve sufficient confident in the product reliability and its cost related issues before launching the product to the international markets. CAF and TALGO have shown their interest in participating in the construction and testing of these TrainWheels demonstrations. This has been considered by the project partners as a great opportunity and a direct result of the exploitation work that has been performed during the FP7 project.
After finishing the demonstration phase it is expected that the future project will enter its commercialization stage in three consecutive phases which will target at the local (Italy, Germany and Spain), European (EU28) and international (out of EU28) markets. Partners will create a joint company for the commercialization of TrainWheels. Potential customers among Entities in Charge of Maintenance, Wheel Manufacturers and Wheel Maintenance Equipment Manufacturers will be informed about the TrainWheels capacities and added values.
Sales forecasts for the period 2017-2021 have been estimated in terms of units and in terms of revenue for the three types of target customers. A sales strategy has been developed which includes a growing segment of sales via licenses to Wheel Maintenance Equipment Manufacturers. Target accumulated sales for the five-year period are of the order of 660 units and 52 M€.
Costs and investment for the exploitation plan have been estimated. They include additional R&D investments, pre-commercialization and industrialization costs for a total estimated 1.480.000 €. Commercialization and manufacturing costs has been estimated at an additional 4,3 M€ and 39,5 M€ respectively, which, added to the nearly 300.000 € already invested by the partners in the project, yields a total investment of 45,6 M€. With these numbers, the estimated project profit at the end of 2021 is of 6,45 M€. This means a 60% return on investment and the achievement of project break even by year 2019 with a maximum negative cash-flow that will be reached in 2016 just below 1,0 M€. The creation of new jobs by the project partners as a consequence TrainWheels development and exploitation has been estimated at 64 new staff in 2021.
The combination of the prototype technical results obtained, the feedback received from the consulted potential customers and the project financial figures described above, give the consortium partners sufficient confidence on the feasibility of continuing the efforts to further develop and launch TrainWheels to the market.
List of Websites:
WESITE: www.trainwheels.eu
PROJECT COORDINATOR CONTACT DETAILS:
Andrea Gili
Termomacchine SRL
Tel. +39 011 9008811
Fax +39 011 9034066
andreagili@termomacchine.com
Skype: gili.andrea74
TRAINWHEELS is a novel technology for detecting train wheels surface cracks by a Non-Destructive testing. This technology aims to respond to a necessity that is being more predominant in the European rail network regarding the inspection and maintenance requirements. In this regards one of the biggest concerns about transportation safety is broken wheels due to cracks and in fact in the latest years, frequent accidents and regular breakdowns have been registered due to this failure.
The main novelty of Trainwheels is based on the use of the Induction Thermography phenomena, which automatically provides the captured images and identifies the cracks through the employed artificial vision tools and without dismantling the wheel-set. Trainwheels system can be used by maintenance companies, installed in their underfloor wheel lathe, or by train wheels manufacturers in order to detect cracks caused by faults in the production process. In fact the aim of Trainwheels is to improve train circulation safety by enhancing the effectiveness of non-destructive tests on wheels and to provide the early detection of surface cracks. For this reason, Trainwheels is envisaged as a solution to the current visual inspection methods, or as a complementary one to the more spread internal inspection method (Ultrasound testing, which lacks surface inspection capabilities). This hybrid system should be needed to suit the purpose of delivering a complete and self-consistent detection method: Ultrasound to provide a crack detection for the innermost part of the wheel and Induction Thermography to detect cracks on the surface and in the outermost part of the wheel.
A first prototype for the validation of the technology has been constructed with the required integration and communication between the involved components and software tools. The adjustment of the hardware, software and electrical parameters was performed in order to maximize the detection of the cracks. Obtained results from the demonstrator gave very good results and thus a successful validation of the technology. A further demonstration event was carried out in Asturias, Spain (at the end user facilities, TAM), to show the suitability of the system to detect surface defects through an automated and faster solution. The goal was also to show the potentiality of the system to the train maintenance companies and related train transportation sector companies. They could evidence the usefulness of the technology as a surface crack detection system, that may replace the current visual inspection and other surface inspection technologies, reducing inspection times and thus increasing productivity.
The trainwheels solution as a system for the early prediction of cracks of wheel manufacturers will make possible to repair them or improve altogether the new ones, with expected savings of up to 25%. The controlled heating and feedback by the infrared camera and the coupled software, makes the system to run under efficient parameters and operating costs. Sales forecasts for the period 2017-2021 have been estimated in terms of units and in terms of revenue for the three types of target customers. A sales strategy has been also developed which includes a growing segment of sales via licenses to Wheel Maintenance Equipment Manufacturers. Target accumulated sales for the five-year period are of the order of 660 units and 52 M€.
Trainwheels has been depicted to comply with all European Safety Standards. The future commercialization stage of the project involves three consecutive phases which will target at the local (Italy, Germany and Spain), European (EU28) and international (out of EU28) markets.
Project Context and Objectives:
Context and Introduction
The European rail network is getting more and more predominant in the transportation sector: it hosts high speed-trains, higher loads of more passengers or goods, with increased frequency. This determines a higher mechanical load for the moving parts of the carrier involved: therefore, increasing inspection and maintenance are required in order to guarantee safety and well-being of trains. One of the biggest concerns about transportation safety is broken wheels due to internal cracks. In fact frequent accidents and regular breakdowns have been registered in the latest years due to this failure.
Three testing methods have been investigated to identify cracks or problems on the surface or near the surface of the wheel: Magnetic Particle Inspection, Dye Penetrant Inspection, and eddy current testing. These methods show the following drawbacks: MPI and DPI generates high amount of hazardous waste and none of them is automatable, while the eddy current system is sensitive to lift-off variations and probes need to be positioned at a constant distance. In conclusion a new, more efficient, automated and faster method is needed to detect surface cracks.
In addition the most of the internal cracks usually begin on the surface of the wheel, it is therefore urgent to deliver a technology for superficial inspection and this vital task can be accomplished by a system based on Electromagnetic Induction phenomenon, as Trainwheels. Nowadays, maintenance technicians use Induction thermography to locate overheating joints and sections; but we require a more sophisticated system.
In fact thermography crack detection method would mean a big improvement in surface and near-surface train wheel crack control, and therefore minimization of the related accidents. Induction thermography is going to be part of a system which aims to establish a more competitive and safer railway network. Through the application of this technology as a detection method, it will be possible to gain an increased security level and a lower wheel and maintenance cost. This system is mainly oriented to be used by railway maintenance companies or by train wheels manufacturers.
Railway technology is a specialized market with specific requirements, which result from high safety standards linked to the transportation of passengers and goods. Technical experts with the required certifications are the basic capital of the companies operating in the rail industry. In a relatively restricted market with a limited total volume, very high entry barriers exist. In Europe, however, with the liberalization of rail transport, companies can now access to previously closed markets.
By using thermography, it is possible to create a new and more efficient way of train wheel surface inspection and maintenance, which is necessary if we take a look at the accidents statistics resulted in the latest years. It is therefore necessary to improve the efficiency of train wheel maintenance procedures, because they have to be inspected at regular intervals for internal and surface defects. Ultrasound system is the state-of-art technology to detect internal cracks, but thermography would mean an automated method to detect surface and near surface cracks, with a considerable lower global cost compared to other surface crack inspection methods.
External costs are a side effect of transportation, which is not internalised into the price paid by the user and is therefore not taken into account by them when they make a transport decision. Electromagnetic induction Thermography makes the defect detection process more efficient, due to the fact that is an automated control process (time reduction compared to manually operated processes) and does not generate any waste in comparison with other surface inspection methods. Thermography as non-destructive method is a fast way to detect surface and near surface cracks and does not require great operator specialization because easy software is used to interpret the results.It is not only a more advanced and automated method to detect cracks, but is even faster and easier to use. By mean of an infrared camera and powerful software, the operator is able to detect surface cracks on screen, without needing new maintenance training. Environmental benefits can be achieved because of the fact that thermography does not generate any waste, in comparison with MPI and LPI methods which generate contaminated magnetic particles and liquid; that means savings in waste treatment and time inspection. In conclusion, it is a greener and cleaner way to detect surface cracks.
Thus Trainwheels is presented as a solution based on automated Induction thermography crack detection, and thus on the physical phenomenon of the electromagnetic induction, a far know - yet very powerful - technique that was first developed as a thermal treatment for the hardening of metallic components. An alternate current generator provides an oscillating current at a specified frequency, flowing into a conductive coil which is shaped such mirroring the train wheel. The coil is positioned as close to the wheel as possible and the varying magnetic field in its cross-section produces an electric field in the metallic material of the wheel. This electric field translates itself into a current, flowing only in the outermost shield of the wheel (for the so called “skin effect”) which, due to its electric resistivity, is then heated. Cracks, if present, will disturb the current flow and so generate changes in the temperature profile in the crack area. These changes of temperature are visualized using an infrared camera. The image acquired by the infrared camera is evaluated through an image processing system.
The potential benefits for SMEs involved in the Consortium are:
o A decrease in accident occurrence of up to a 25% from current yearly rates
o Savings up to 25% for wheel manufacturers who, through an early prediction of cracks in their wheels, will be able to repair them or improve altogether the new ones.
o Savings of up to 20% in the induction process, because the heating will be controlled and feed backed by the infrared camera and the coupling software, in order to run thermography under efficient parameters and operating costs.
Objectives
Trainwheels, based on the Induction thermography phenomenon is presented as an automated train wheel surface crack detection method that will meet all European Safety Standards. Induction thermography as a non-destructive test method is already used in other sectors of industry. Using thermography for train inspection would result in a big improvement compared to previously quoted methods: as an automated and faster approach as well as cleaner and cheaper.
Thus the challenge of Trainwheels is to optimize Induction thermography in order to use it as an automatic train wheels surface inspection system.
For this purpose firstly, an induction generator will be improved to heat up the wheel by passing electric current through it. By means of a thermographic camera, a scan of the wheel will be provided in chromatic scale. The temperature profile is shown on screen and the presence of a crack will be highlighted by a hotter spot on the image. For this reason, it is also necessary to provide modern and easy-to-use software that shows the profile of the emitted radiation on the screen. The development of a new sensitive infrared camera will open new fields for thermography applications.
Following the technical as well as the scientific objectives of the project are described. The specific goals as defined by DOW are also included.
In general terms Trainwheels aims to achieve the following goals:
- To develop a method to detect train wheel surface cracks based on thermography. Faster and cheaper for at least 20% than previous methods.
- Up to 20% of reduction in inspection times. Energy savings up to 20% from current induction systems
- Develop and optimize an induction device. To obtain a novel application for thermography applied to train wheels inspection
- Develop and optimize an infrared system
The scientific objectives of Trainwheels, according to the needed research and development task are:
- To gain understanding of train wheels operations, railway environments and behaviour of materials under those circumstances. To be able to develop a thermography system as quick and effective as an ultrasound system
- To increase the understanding and physics involved in the induction system by simulation (Coil dimensions, configuration, relative positioning, electric power, voltage and current, inspection time and heating process
- Increase understanding of thermography and thermal artificial vision. Up to 20% increase in defect detections.
From the project development point of view, the objectives are specified and distributed in the following work packages:
WP1. System Characterisation.
Define and characterize the type of wheel set systems available on the market and their properties (material magnetic and thermal properties). Define, characterize and classify the types of wheel surface cracks. The following deliverables must be available:
WP2. Electromagnetic induction heating system design.
Define and evaluate through simulation, the electromagnetic induction physic and thermal transfer pattern, for different configurations according to the data obtained in WP1 and WP3 (type of cracks and type of thermographic camera). In a second phase the goal was develop an optimized induction system for an optimal heating of the train wheel surface.
WP3. Thermographic methods.
Evaluate available thermography methods to find the best suitable parameters and configuration. Test and design a suitable inspection system consisting on an infrared camera and suitable algorithms for the software. Measurements and analyses of characterized failures was also done to proof the capability and suitability of the developed system.
WP4. Artificial Vision.
The objective of this work package was to develop and design pre-processing algorithms to increase signal quality. Implement state of the art data and design new computer vision algorithms to recognize and classify defects in thermal images. This will led to the development of an automatic result analysis system, to assist personnel to interpret and categorize the cracks.
WP5. Engineering
Definition of the boundary conditions for an industrial environment in the railway sector. Simulation and construction of the testing system mechanical assembly. Manufacturing support for the prototype production and iterative redesign measures.
WP6. Technology Integration and Validation.
The first part of this Demo work package was dedicated to the assembly and pre-testing of the prototype at TERMO’S facility. Once shipped to TAM’s premise, the second part will aim at the integration and validation of the components, the monitoring system and the whole TRAINWHEELS system by the Consortium partners.
WP7. Dissemination of knowledge
The purpose of this work package is threefold. Firstly to develop an Exploitation plan, made the SME consortium and led by the Exploitation Manager. Secondly, after signing a Consortium Agreement between the partners to define and protect the future Intellectual Rights that will derive from the R&D activities, the most important results obtained will be protected from the competition. Finally a Dissemination strategy will was elaborated to communicate the results within the Consortium and to exploit and popularize the TRAINWHEELS results to the scientific communities and to potential end-users.
WP8. Consortium and Project Management
The main objectives planned for WP8 are the following:
• Coordinate knowledge management.
• Coordinate the production of deliverables, milestones reports & cost statements.
• Coordinate legal, contractual, ethical, financial & administrative issues of the consortium.
• Organise and coordinate the Project Management and Exploitation Board Members meetings.
• Coordinating payments, audits and distribution of money.
• Coordinate communications between the consortium and the EC.
• Coordinate the preparation and signature of the Consortium Agreement.
Project Results:
The overview of the progress of the work to gain the aforementioned objectives is presented here, in line with the structure defined by DOW (WP and corresponding task). The most significant results are highlighted.
System Characterisation
A in-depth study of the wheels geometry and material properties was performed comprising the typical defects and cracks.
Different information sources were used to obtain the most complete current situation:
- Directives from the UIC “Union Internationale des Chemins de Fer”
- Railway Group Standards
- British Railway Standards
- Several European Norms and Standards (UNE)
- Relevant scientific and technical papers
- Visits to key actors in the sector: wheels and trains manufacturers, maintenance companies, railway companies and potential end-users (Talgo, CAF, Nertus, Danobat railway division, etc.)
The work performed comprised:
- The description of the train wheels and axles from a geometrical and physical point of view: applications, figures, thermal properties, electromagnetic properties, etc. It will be very relevant for the following work packages concerning simulation.
- The geometrical characteristics and classification of surface cracks was done: type, dimensions, inspections, etc. This data will be very required for the “Artificial Vision” Work Package, where the artificial vision software to detect failures will be developed.
- Relevant information from important Railway actors during several visits made by Inspiralia and TAM is reported.
- A market study was performed in parallel to the technical work, in order to help us defining the final TRAINWHEELS solution and at which market will target.
The main information and data obtained regards the railway network statistics and the contemporary situation analysis. One of the most important conclusions stands on the revolution of the infrastructure and its consequences. The European rail network is getting busier every day with high speed-trains, more passengers, higher loads, and an increased frequency. These factors produce heavier axle and wheel loads with the subsequent infrastructure inspection and frequencies reformulation requirement. In fact according to European Directives on Safety, one of the most important indicators relating to precursors of accidents are broken wheels; even though severe rail accidents are relatively rare within Europe, their frequency is still at an intolerable level, and at a second level network disruptions and breakdowns may cause several social problems.
Thus increasing and improving inspection and maintenance are clearly required in order to guarantee the maximum efficiency and safety as possible. No clear definition of the required inspection technologies, beyond the required ultrasonic method for internal defects detection, is defined on the standards. The companies used to employ different type of inspection technologies. The most common surface inspection methods found are: Magnetic Particle Inspection (MPI), Dye Penetrant Inspection (DPI), and Eddy current testing. These methods show the following drawbacks: MPI and DPI generates high amount of hazardous waste and none of them is automatable, while the Eddy current system is sensitive to lift-off variations and probes need to be positioned at a constant distance. In conclusion a new, more efficient, automated and faster method is needed to detect surface cracks.
The most important points comprise:
Wheels classification by: configuration, brakes, dimensions, and material:
- Regarding the configuration two types exist: solid or monobloc wheels (wheels manufactured in one piece) and tyred wheels which are manufactured by several pieces. Tyred wheels allow to replace only the tyre instead of the whole wheel, but are less reliable and in fact are actually forbidden for high speed trains. Both types are representative for Trainwheels
- Two types of brake systems are identified. The first one is the brake shoes which applies the load directly to the wheel tread the flange. These brakes are mainly used on freight car wagons and they produce thermal fatigue loadas on the tread which may induce cracks on the surface. Thus this type is the most relevant for Trainwheels. The second type are disk brakes which press the rim of the wheel and thus surface defect will not be as frequent.
- Dimensions (Diameter and Profile) of wheels are defined on standards and must be respected for safety reasons. The profile is a key factor as it defines the geometry to be heated and tested.
- Materials are clearly defined and specified on standards. For Trainwheels we must consider ER7&ER8. Since these specifications do not take into account electromagnetic properties, these must be defined early on following tasks.
Cracks classification based on location and size:
- Regarding the location the most frequent cracks are located (Figure 2) at end face & tread (the most important), and at clamping places and at marking places too.
- The maximum size allowed for surface cracks is defined by any of the dimensions being 2mm.
Market study. The study comprised two main points: the type of inspection required to understand where to place the system and the type of train (freight or passengers) to determine our client target.
- By type of Inspections the following were found:
_Routine inspections: mounted wheels, train in movement, automatic, often every 15 or 30 days.
_Complete inspections: disassembled wheels, in special shops or pits, with operator presence, and every 10^6 Km
- Regarding the inspection stage, different strategic options apply: at manufacturing and/or reprofiling stage (pit lathe), where the wheel will be clean and after which no cracks should be found, during maintenance inspections which depend on company’s strategies and for manufacturing companies as a final inspection stage.
- By type of market we have found different scenarios and constrain:
_Passenger trains: Inspections in passenger trains are more restrictive due to, traditional mind, much more complex design and low accessibility, and more common interior cracks (fatigue).
_Freight trains: inspections in freight trains show more opportunities in the market thanks to: new regulations in the way, Wagons and wheel-set design allowing better accessibility, low inspections levels (frequency and control), and importance of surface cracks detected due to thermal loads.
This data was the starting point for the following work on the project and an important basis to perform the design of the electromagnetic induction heating system, together with the artificial vision component and the thermography methods.
Electromagnetic induction heating
Computational simulation was employed for the design of the inductor geometry and electrical parameters, required to improve the crack detection as well as the efficiency of the induction system. A deep study was carried out with a number of different models performed for the purpose:
- Initial model: simple metallic plate and a simple coil to study the different possible configurations and to better understand the physics.
- Validation with experimental test to validate the simulation tool
- Heat diffusion study, of the wheel was carried out when induction is stopped.
- Rotating wheel study. To simulate the real Trainwheels case
Initial simple model
Considering the parameters obtained by the Initial studies (crack dimensions, location and type) the first numerical models were developed. These simulations are based on a coupled electromagnetic-thermal problem that were initially carried out in a simple metallic plate, to understand the physic and its influence under different conditions: electrical parameters, geometric positions, material properties, crack orientations and dimensions, etc. (Deliverable 2.1). Results of the temperature patterns were obtained which in general terms show the maximum heat on the tips and colder flanks as shown in the following (Figure 3).
The most important conclusions obtained from the simulation is that the crack will show up more clearly on thermal gradient solutions than on thermal maps images. Moreover, the border effect is eliminated on the thermal gradient results, showing its maximum on the crack tips. The thermal gradient plot will highlight the crack onto the surface at every time step. However small time step will be preferable, before the heat diffusion starts, which increases very rapidly (Figure 4).
From these results it was concluded that the optimum electrical parameters combination is that with the highest frequency (400kHz for the given FHG inductor) due to the more evident crack detection and the most efficient use of power. This goes related to the fact that the frequency increase will generate a smaller skin depth and thus the current will be forced to flow through a smaller depth, which will be translated into a higher current concentration and thus higher temperatures. From this initial study it was also concluded that any kind of crack dimension (with a minimum of 1mm in any direction), location and orientation can be detected at high frequency electrical inputs.
The final electrical inputs and final coil geometry for the inductor generator were derived from these simulations, for which the impedance value was required to design the capacitor. This value will allow defining the whole system electrical behaviour as an integrated generator-capacitor-inductor system. In fact simulation was a mandatory tool to predict the cracks detection behaviour.
Validation
A validation to prove the suitability of the FEM model to predict the behaviour of the system was carried out, thanks to the experimental test performed by FHG. This will be essential to prove the consistency of the tool for future analysis.
Very accurate and similar results (temperatures in the crack around 34°C and around 27-30°C in the surroundings between the coils), were obtained from this comparison Figure 5, which led to a qualitative validation of the electromagnetic model. This approximation is due to the fact that the results from the thermographic camera cannot be used to read the exact temperature, in a high extent due to the image noise, but an overall thermal map can be obtained as shown.
Hence, further details around the crack and tips of the crack are difficult to validate also due to the resolution or the camera and the fact that the sequence image composition, will not provide accurate temperatures from the IR maps. Simulation provides also the temperatures inside de crack, on the plate deeper area and thus not only on the surface as it does the camera. The IR signals from this deeper area are not received from the camera, and thus only the higher temperature of the edges will be detected.
Heat diffusion study
Once computational time was reduced (through the wheel geometry simplification study, and using just a portion of the wheel), an initial study of the crack heat diffusion was done for a static wheel. A simple coil and a crack were modelled and the temperature on the tip of the cracks along time was obtained.
If the induction is stopped the heat diffusion will occur rapidly. It will depend on the electrical parameters, etc. but in around 0.5s the temperatures on the target will be again near the ambient temperature.
The temperature along the crack was also plotted (Figure 7). The highest temperature jump is observed between the tip and the inner of the crack and the thermal gradient between the tip and its exterior surroundings will be much lower (blue arrow). In a very short period of time the spatial thermal gradient will decrease in a high extent. However this highest temperature jump won’t be noticed by the camera since only top surface IR emission and temperatures will be detectable (proven in the experimental test). Thus the thermal gradient between the tip and its exterior surroundings will be the data that the camera will be able to read, in which the spatial thermal gradient will be lower.
Rotating wheel model
The simulation of a rotating wheel model was performed to replicate the final model. The same model was used (Figure 6) and the electrical parameters were those to achieve the maximum values allowed by the generator (400kHz and 5kW). The thermal gradient (Figure 8) for the different time steps showed the influence of having the coil above the crack (thermal gradient maximum value).
Thus the optimum place to detect the crack would be, right above the coil, but this will not be possible since it would be covering the wheel. Higher wheel angular velocities (0.5rpm) were also tested in which the heat influence along time was higher and thus the crack detection improved.
Coil geometry study
To obtain the optimum coil geometry different analysis were performed. Various coil distances and orientations were studied in which the magnetic cancellation in a two parallel wires configuration was observed (Figure 10). The U shape and the following shapes were studied to check the different influence on the piece of the coils with different direction currents.
The induced current densities were plotted and higher values for lower distances were observed. With the multiturn configuration a 5% of higher values were obtained. For the final coil geometry design the way in which the crack will be more obvious had to be also considered and thus a pattern that assured the detection of every kind of crack (dimension and orientation) was studied. Since the induced current distribution must be high in order to be able to detect it at the distance at which the camera will be positioned, the multiturn case was preferred. The final coil design was a pancake shape inductor to detect 360º craks (Figure 11). With this design the thermal shift was increased and thus a longer detection capability is provided (Figure 12).
The final coil design was carried out following the pancake shape as stipulated the optimum one but with squared shape, in order to provide an homogenous patter along the whole tread.
Main conclusions
The main conclusions obtained from these models are:
- Initially the optimum electrical parameters will be based on the maximum frequency (in this case and for the TERMO induction system 400kHz). The intensity to be supplied was defined together with the final coil geometry (2kW). These parameters were then optimized based on the results obtained from the first induction test and obtained images. There it was proven that due to the complex shape of the inductor and electrical conditions not was not possible to achieve such a high frequencies. Possible frequencies were 50kHz an 150kHz. As it was then noticed the lower frequencies generate higher signal to noise ratio and thus that was preferred to obtain a clearer detection.
- The temperature jump due to the current density concentration is observed right under the coil. Very fast heat diffusion is observed for very small time steps.
- The most important result that will highlight the crack is the spatial thermal gradient pattern, and to be applied in the artificial vision software.
- Higher thermal gradient was found within the crack, but the deeper values will not be detected by the camera and thus only surface values and thermal gradients between the crack edges and surroundings should be considered.
- Magnetic field cancelation is lower for lower wire distances. A pattern that assures the detection of every kind of crack dimensions and orientation was designed, with a pancake geometry.
- Simulation provided the information required to predict how the cracks will be observed trough the camera, saving time and energy. Thus the most appropriate electrical parameters, the optimum inductor geometry and other very interesting issues were obtained from the analysis, to finally design a system which assures a safe and secure crack detection
Thermography methods
The central part of the Trainwheels thermography system is the thermal camera which will be used for data acquisition. The properties of the thermal camera determine the quality of the data that will be available for the automated crack detection and categorization with the software algorithms.
After some tests with different cameras which fulfilled these specifications for the Trainwheels system the Xenics Gobi 640 CL was the chosen thermal camera, because this camera fulfilled all specifications and seems to deliver the best additional features for the Trainwheels system.
Technical data of the selected thermal camera Xenics Gobi 640 CL can be found in D3.3 and the second review report. First the chosen camera was compared with the camera used for the basic tests. Therefore the signal quality was evaluated with the calibration body (Figure 15). It can be seen, that the Trainwheels camera delivers a sharper image compared to the camera used for the initial test. This should be self-evident because of the better resolution.
Analyses of the amplitudes of the reference cracks show that the noise of the Trainwheels camera is higher than the noise of the camera used for the initial tests, but also the amplitudes are higher. For the images acquired under the conditions as described above, a signal noise ratio for the Trainwheels camera of 2.4 dB can be calculated and 2.6 dB for the camera used before.
Induction Frequency
As the theory of the electromagnetic induction tells, the excitation frequency determines the skin depth of the inducted eddy currents. A lower frequency leads to a deeper skin depth but with a lower eddy current density. For the crack detection it is important to penetrate as much as possible with the skin effect inside the material but also it is important to generate an eddy current density, which is high enough to generate thermal effects on the cracks that can detected by the camera.
For testing the influence of the induction frequency the Thermomacchine AURORA induction generator was modified in the way that it can be operated at 50 kHz and at 150 kHz. The skin depth of steel is about 0.06 mm at 150 kHz and about 0.1 mm at 50 kHz. A comparison of the influence of the inductor frequency is demonstrated (Figure 17).
The lower frequency delivers an image with more contrast also with lower induction power. With the high frequency at low power heat is only generated at the edges of the reference cracks. Most of the energy is concentrated at the surface and heats up the surface. The contrast increases with the power and above a power of 40 % the surface effects are smaller in comparison to the Joule effects at the cracks. It seems that then also crack flanks are heated by the Joule effects and the thermal contrast rises. At the lower frequency the thermal contrast between the reference cracks and the surface is quite better also on low power. The inducted eddy currents penetrate deeper inside the material because of the higher skin depth and the crack is heated at is flanks also on low power.
The signal analysis along the crack lines illustrates this very clear. At high frequency most of the induction power is used to heat up the surface and there is a bad signal noise ratio for a high frequency (Figure 18) at low power. At the low frequency the surface isn't heated up as much but the flanks also generate heat so we get a good signal noise ratio already on low power. When the induction power increases the surface isn't heated up more than on low power but now also the crack flanks generate heat and the signal noise ratio is getting even better.
Table 1 shows the calculated signal noise ratio at 50 kHz and 150 kHz induction frequency, being higher for low frequencies and high power.
Induction Power
Beside the induction frequency also the needed induction power is of interest for the project, because another aim of the Trainwheels project is to keep the costs low what means that also the used energy for testing has to be kept low.
After finding out that the lower frequencies of the induction system are leading to better results, the influence of the induction power was studied. For this the signals of the reference crack of the calibration body were analysed at induction frequencies of 50 kHz and 150 kHz while the induction power was varied.
The images (Figure 19) show that there is a good thermal contrast already at 10 % induction power at the frequency of 50 kHz. But there are also indications of scratches in the calibration body’s surface which are caused from manufacturing. These scratches disappear above an induction power of 30 %, so that only the reference cracks and some artefacts caused by differences in the emissivity of the calibration body’s surface can be seen. The images on the right, for which an induction frequency of 150 kHz was used deliver useful data not before an induction power of 30 %.
The analysis of the signal along the lines of the calibration body which are containing cracks are given (Figure 20). At an induction power of 10 % there is already a good signal noise ratio when using 50 kHz excitation frequency, and the signal noise ratio is increasing steadily until 50 % of induction power. For 150 kHz a signal noise ratio above 2, which allows to distinguish a signal properly from the noise, cannot be achieved below an induction power of 30 %.
The signal noise ratio is given for different induction powers at the frequencies of 50 kHz and 150 kHz which are available by the Trainwheels induction generator (Figure 21). It can be seen there that the signal noise ratio at 50 kHz is always higher than on 150 kHz independent of the power.
Wheel Velocity
The velocity of the wheel influences the measurement results. The relation between the cameras frame rate and the wheel velocity is given by:
1 1
The aim of the Project is to test a wheel in one minute or faster. This means a circumference velocity of 50 mm/sec is at least needed to check a wheel of about 900 mm diameter in one minute.
To study the influence the calibration body was moved with different velocities during the measurements. The measurement conditions were the same as in the tests with the induction parameters. For data acquisition the Trainwheels thermal camera and the basic Trainwheels software was used. The results of these tests are shown as line scans (Figure 22).
In Figure 22 can be seen, that until a velocity of 100 mm/sec there is no big influence to the images. Above a velocity of 100 mm/sec the cracks are starting to smear and also the edges are blurring. Also a loss of signal in the angular reference cracks of the calibration body can be registered.
In summary the analysis of the influence of the circumference velocity proofed the consideration done in deliverable 3.1 where a velocity of 50 mm/sec was calculated when using a camera with a resolution of 384 x 288 pixels. Now a camera with 640 x 480 pixels was used which leads to a faster circumference velocity of 100 mm/sec which means that theoretically the testing of the wheel can be done in about 30 sec when one round for testing is needed (2rpm).
Artificial Vision
Graphical Software User Interface
The software user interface can be seen below (Figure 23). The software was designed to be simple to use for its main task: executing measurements. On the left, the user enters certain data related to the wheel under test or to the test itself: examiner name; client name; train, wagon and wheel number; wheel diameter. The diameter is not only stored in the result file for information purposes, it is also used in translating line numbers into distances in millimeters. Finally, the user has to enter the line number that is used to construct the unwinded image (see section Image Processing).
On the top, the user can activate and deactivate a live image of the camera and manually execute a non-uniformity correction. Most importantly, when the live image is activated, it is possible to initiate an automatic measurement.
Data Acquisition and communication
The Trainwheels system consists of five main components which had to be controlled by the PC (except the mechanical system) that they work together proper. Those components are:
• Induction system
• Thermal camera with frame grabber
• Wheel set drive
• Positioning monitoring
• Mechanical system
Figure 24 gives an overview how the subsystems are connected and controlled by the PC. The communication respective the signal flow between the subsystems was realized by an IO card which delivers the signals and information to the software module which controls the workflow during the measurement. The signal flow is shown in Figure 25. The control parameters as well as the measurement and image processing parameters are stored in a configuration file which can be adapted to each specific inspection task
Image Processing
Once the measurement is started and the trigger sensor passes for the first time, the camera captures frames with a rate of 50 frames per second for one whole turn of the wheel, until the magnetic sensor passes for the second time. In every captured frame, only one line – whose vertical position in the image is entered by the user – is actually used in further processing. Hence, this line is cut out of each frame and stored in the main memory; the rest of the frame is discarded.
When the magnetic sensor passes for the second time, the data acquisition finishes. The stored lines are assembled into one single image which presents an projected 360 degree view of the wheel (Figure 26). All image processing is done based on this projection image.
The carried out image processing algorithm is as follows: First, a line segment detection is executed. The used algorithm is the “line-segment detector” (LSD) as described in von Gioi, Jakubowicz, Morel, Randall: “LSD: A Fast Line Segment Detector with a False Detection Control”, IEEE Transactions on Pattern Analysis and Machine Intelligence (2010), and implemented with tiny variations in OpenCV in the class LineSegmentDetector. The result of the detection is a set of lines, each with a starting and an ending point.
Second, each of these lines is treated as a candidate that has to fulfil further conditions to be finally treated as a defect. The line length of a defect line must be equal to or larger than a certain minimum length (first condition). Then, a pairwise comparison of all candidates is executed and for each line that is to be considered a defect, another line must exist that is parallel (second condition) and not too far away (third condition). Lines that meet all three of these conditions are considered defects and drawn in red, lines that remain in the candidate stage are drawn in yellow (Figure 33).
Technology integration and validation
The Trainwheels prototype integration and validation was carried out. It was mounted at TAM facilities (end-user) where the integration of the different components and final test was carried out. The components and assembly were validated in a lab environment and the successful results are highlighted here.
Prototype integration
The mechanical prototype design with the detailed drawings was described in deliverable 5.2. The main characteristic and purpose of this design is to provide a flexible system that can be mounted both on the top or the bottom of the wheel. The goal to be installed on the top of the wheel is to provide an easier handling, for the prototype testing that will be carried out at TAM.
However the real configuration will be that installed on the bottom of the train wheel, which responds to the necessity of the companies to inspect the wheel in an automated manner. The system starts to inspect the wheels automatically once the train enters in the inspection station, and thus it can be mounted directly in the rails of the inspection line.
The mechanical parts were manufactured. The components of the system are very briefly described here (more details are given in Deliverable 6.1):
- Mechanical frame
- Thermographic camera
- Generator and Induction system
- Wheel turning motor
- Electrical box
- System control and Computer hardware and software
- Encoder
- Chiller
- Electrical connections
An initial pre-prototype was integrated at FhG (Figure 29). The assembly of the system initially performed in Fraunhofer served as a first validation, from which some initial results were obtained.
The final integration for demonstration purposes was carried out at TAM (Figure 30), the end user of the technology. There the final wheel was mounted (provided by Arcelor), which corresponds to a Freight train. All the components were integrated and connected (Figure 31).
Results and validation
The most important criteria to proof the functionality of the Trainwheels demonstrator and to show that the principle of inductive thermography can be used for the detection of surface cracks on railway wheel is stated here.
Data acquisition and evaluation
The data acquisition concept based on multiple line scanning and use those lines for creating a projection around the circumference of the wheels. This concept was chosen to get spatial information as well as time dependent information about the crack. The tests showed that this concept principally works, but therefore the rotational speed of the wheel has to be constant, otherwise the speed deviations lead to images which can’t be matched and the time dependent information will be lost. Unfortunately with the wheel set at TAM facilities it wasn’t possible to get the needed constant rotational velocity, because of the damages and impacts of the wheel set. So for data evaluation at TAM no time dependent information was available and the data evaluation could be done only with the spatial information. Nevertheless also on TAM’s wheel set the system could detect the cracks only by optimizing the evaluation parameters of the spatial image evaluation.
The tests also showed that the aimed inspection period of 1 minute is possible. The tests were done with a circumference speed of 60 mm/s this means an inspection period of 50 s for a wheel set with a diameter of 930 mm.
Software
The software interface was designed in way that a minimum of user interaction is needed. The GUI allows the operator only to setup metadata for the inspection report, to start and stop a measurement or to load a former measurement. The operator has no direct access to the measurement or evaluation parameters. These parameters are stored in a configuration file which can be created for each specific inspection and reused for repeating inspections. This ensures that the parameters can’t be changed during the inspection by the operator which is conforming with NDT standards where only administrators can change such parameters.
To give the operator information about the subsystems a status bar is available which indicates warnings and errors from the subsystem. During the measurement the main window of the software shows the live image of the thermal camera so the operator can follow the measurement what might be helpful for the result interpretation. After the measurement has finished the software report is generated automatically as a pdf file and the evaluated thermal image where the evaluated cracks are indicated is displayed. The software allows the operator to rotate the wheel that a crack will be moved under the camera by right clicking on that crack. This is also helpful for the result interpretation.
What is missing on the software is that the operator cannot change an obvious false indication in n the results or comment them.
Induction system
An aim of the project was to decrease the power consumption for in relation to the crack detection with magnetic particle testing. A modern standard magnetic particle testing device for parts up to 100 kg and a maximum diameter of 400 mm has a power consumption of about 25 kW to 30 kW. But such a device cannot be used for the wheel set testing (wheel weight 300 kg, diameter > 900 mm). The Trainwheels demonstrator has a maximum power consumption of 20 kW and was used at 30 % to 40 %. So the effective consumption is about 5 kW for wheel testing.
The for the Trainwheels system developed coil allows to test in all three inspection positions. There is no need to change the coil between testing on the tread or on the side faces. This makes the demonstrator very time efficient. Also the coil and the induction system are very robust against deviations in the distance between coil and wheel surface. During the tests no influence on the results could be asserted when the distance differs of ± 1 mm.
Results
The first results show a promising detection potentiality. There the differences between a clean wheel (brushed wheel from FhG) and a very used wheel with natural cracks (form TAM) were observed (Figure 34).
The smaller and sharp-edged defects of the Fraunhofer wheel set generate a much better thermal indication than the defects of TAM’s wheel set. The reason therefore is that the sharp-edged defects are leading to a higher eddy-current density and thus to more heat generation at the defect. Nevertheless, the Trainwheels thermography system is able to detect the cracks on both wheel sets. Figure 35 shows a comparison of the result images of the two wheel sets, evaluated with identical evaluation parameters. The red markers are indications classified as a crack, the yellow markers are indications which were detected by the software but not classified as crack. The reference cracks are marked blue. In Figure 35 can be also seen, that the pitting around the surface of TAM’s wheel set produces a lot of false indications, because the edges of the small pitting craters act like a crack. The false alarms in the flange area of the Fraunhofer wheel set are caused by 0.5 mm to 1 mm deep rills from the lathe form feed.
The evaluation software allows to decrease the rate of the false alarms by setting up these parameters more coarse. Figure 36 and Figure 37 are showing the difference of a very sensitive and a coarse setup of the Trainwheels system evaluation parameters. In Figure 36 can be seen, that it is possible to eliminate most of the false alarms in the flange but to keep the defects. The same can be seen in Figure 37 for the TAM wheel set.
In the context of the detection possibilities on the TAM wheel set should be mentioned that railroad experts told, a wheel set like the one that was used for the tests at TAM’s facilities shows the worst case and in common a wheel set like this would never be inspected without t re-profiling before.
About the reproducibility of the Trainwheels thermography system can be stated that the reproducibility of the crack detection is given. In about 30 tests during the testing period the all cracks were detected when the system was set up correct and the cracks were representative for natural cracks. There is only a slight unsteadiness in detecting radial cracks on the tread when the cracks are 100 % parallel to the side faces over their complete length.
Conclusions
The validation showed, that the realized demonstrator fulfils the aims of the Trainwheels project. The main purpose for the project, to use inductive thermography as an alternative method to the classic NDT methods for crack detection on railway wheel sets was reached. With the demonstrator can be proofed, that the inductive thermography generates comparable or better results as the classic methods. Also the advantages of the inductive thermography like no need for cleaning, automated measurement evaluation etc. can be shown.
In summary can be stated, that with a few changes the demonstrator can be used under industrial conditions for crack detection on railway wheel sets. Some of these changes are:
• Automatization of the manual axes.
• Modification of the wheel drive to reach a constant rotational speed also on heavy used wheel sets.
• Some improvements in the mechanical structure like:
o Modifying the vertical stoppers of the self-centring.
o Decreasing the weight of the system to make the moving easier (i.e. possibility of undocking the induction unit during docking the system to the wheel set).
• Some software changes like:
o Add the possibility for the operator to comment and correct false alarms in the result.
o Adding additional features for additive manual evaluation.
The most important improvement will be the modification of the wheel drive, because this will offer the possibility to use time dependent evaluation algorithms and thus a gain in information which can be used for increasing the crack detection performance.
Potential Impact:
Dissemination activities
TrainWheels has shown an important commitment to disseminate the project results through different channels: Internet presence has been consolidated by keeping updated regularly the project website and by including information in the different websites of the consortium members. Several press releases have been published in different media with the participation of all the consortium members, as an activity that continued during the second project reporting period. We also approached relevant actors of the railway industry in the target markets such as Talgo, Nertus, Danobat, Arcelor and CAF, to which the product has been introduced.
The dissemination activities are summed up here (more details in deliverable 7.6):
- PROJECT WEBSITE
- INTERNET PRESENCE
- CONFERENCES, SEMINARS AND TRADE-FAIRS
- PROJECT BROCHURE
- PRESS RELEASE
- PROJECT VIDEO
- DISSEMINATION AMONG POTENTIAL CUSTOMERS
The seminars, fairs and conferences attended are listed here. There the Trainwheels technology was shown and the information was made available for the public
- AWZ Seminar in Radsatz. April 2014
- 8th Symposium of Non Destructive Testing in Railways in Wittenberg. March 2014
- London Rail Infrastructure Exhibition. May 2014
- Innotrans 2014. September 2014
- Rail Vehicle Conference Rad-Schiene in Dresden etc. September 2015
A scientific article was also published at the Rad-Schiene Conference proceedings and a presentation was delivered to explain the most relevant aspects of the technology and the main project achievements.
A project brochure and summary presentation was prepared and used in dissemination and communication events to attract the audience and keep their interest in its highest level. For this purpose, a video presenting a solution in a very intuitive way and easy to understand was created and included among our dissemination material.
A technology demonstration event was held on October, 9th 2015 with the participation of a wide audience among members of the railway industry, potential customers and other project stakeholders. This half-day TRAINWHEELS technology demonstration event which was held at TAM facilities in Avilés, Spain allowed Trainwheels technology and prototype to be shown to potential customers and interested parties by members of the technical and commercial teams of the consortium partners. The participation in the event was numerous and the results of the project were discussed with great interest by the audience (Figure 34).
The consortium is satisfied with the level of communication and dissemination achieved. The activities performed allowed for relevant interaction and feedback from potential customers and other market agents that are considered a key factor to the future development of our commercial and partnership strategies.
Exploitation plan
The exploitation activities performed during the TRAINWHEELS project, include: the results of the market analysis performed by the project partners, the projected sales forecasts, an estimation of the investments ahead, an implementation plan for product exploitation and a business plan with conclusions on the project financial projections, etc (more details in deliverable 7.4).
TrainWheels project have allowed the consortium partners to develop, and successfully test TrainWheels technology by the construction of a system prototype that have proven that the induction thermography technology is valid for train wheels surfaces crack detection. Potential customers have been identified and the target market for the product has been defined. Needs and requirements in terms of quality, costs and efficiency have been investigated by means of interviews with a selected group of potential customers. The system performance up to date has given the consortium partners sufficient confidence on the system capability to satisfy market needs. A complete analysis of alternative options in the market has shown the product will be competitive and the TrainWheels target price has been estimated.
Goals for TrainWheels market penetration have been revised from previous versions of this report, based on the latest outcomes of the market and competitive analysis performed. Team competences have been revised and the necessary partnership strategy for future TrainWheels development and exploitation has been identified. Moreover, two potential partners have already been identified to collaborate with the consortium in the future construction of TrainWheels demonstration units to test the product in industrial environment with real data and real trains.
The consortium strategy for product exploitation has been unveiled. It consists of a 4-phase project, starting with a demonstration phase and three commercialization stages.
The demo phase allowed the partners to achieve the necessary references for product introduction to the market. As a very conservative market, particularly when safety is involved, the railway industry, and the partners themselves, need to achieve sufficient confident in the product reliability and its cost related issues before launching the product to the international markets. CAF and TALGO have shown their interest in participating in the construction and testing of these TrainWheels demonstrations. This has been considered by the project partners as a great opportunity and a direct result of the exploitation work that has been performed during the FP7 project.
After finishing the demonstration phase it is expected that the future project will enter its commercialization stage in three consecutive phases which will target at the local (Italy, Germany and Spain), European (EU28) and international (out of EU28) markets. Partners will create a joint company for the commercialization of TrainWheels. Potential customers among Entities in Charge of Maintenance, Wheel Manufacturers and Wheel Maintenance Equipment Manufacturers will be informed about the TrainWheels capacities and added values.
Sales forecasts for the period 2017-2021 have been estimated in terms of units and in terms of revenue for the three types of target customers. A sales strategy has been developed which includes a growing segment of sales via licenses to Wheel Maintenance Equipment Manufacturers. Target accumulated sales for the five-year period are of the order of 660 units and 52 M€.
Costs and investment for the exploitation plan have been estimated. They include additional R&D investments, pre-commercialization and industrialization costs for a total estimated 1.480.000 €. Commercialization and manufacturing costs has been estimated at an additional 4,3 M€ and 39,5 M€ respectively, which, added to the nearly 300.000 € already invested by the partners in the project, yields a total investment of 45,6 M€. With these numbers, the estimated project profit at the end of 2021 is of 6,45 M€. This means a 60% return on investment and the achievement of project break even by year 2019 with a maximum negative cash-flow that will be reached in 2016 just below 1,0 M€. The creation of new jobs by the project partners as a consequence TrainWheels development and exploitation has been estimated at 64 new staff in 2021.
The combination of the prototype technical results obtained, the feedback received from the consulted potential customers and the project financial figures described above, give the consortium partners sufficient confidence on the feasibility of continuing the efforts to further develop and launch TrainWheels to the market.
List of Websites:
WESITE: www.trainwheels.eu
PROJECT COORDINATOR CONTACT DETAILS:
Andrea Gili
Termomacchine SRL
Tel. +39 011 9008811
Fax +39 011 9034066
andreagili@termomacchine.com
Skype: gili.andrea74
