P J Mudge, S Chan
Granta Park, Cambridge, UK
Plant Integrity Ltd
Granta Park, Cambridge, UK
E Andersen, J K F Andersen
Paper presented at BINDT annual conference 2008, 15-18 September 2008, Macclesfield, Cheshire, UK.
The potential for low frequency ultrasonic guided waves to monitor large areas of structures offers considerable scope for long term monitoring of a structure's condition. This paper reports on the work carried out within the European Commission funded OPCOM project which has examined the application of this technology for monitoring both oil and gas production installations and offshore wind turbine towers. Developments have included design of sensor arrays capable of withstanding marine environments, long term environmental influences on the test data and modelling of the influence of component geometry and condition on the test.
This work is concerned with the development of tools and techniques for the 100% examination and monitoring of very large diameter tubular steel structural components using ultrasonic guided waves. Such tubulars exist in large numbers in offshore installations and their structural integrity may be affected by both corrosion and fatigue cracking over the service life of the installation, which is normally in excess of 20 years. This work has been carried out under the European Union funded OPCOM project.
The low frequency (i.e. approximately 20 to 100kHz) ultrasonic guided wave technique was developed in the 1990s for the rapid survey of pipes, for the detection of both internal and external corrosion. The propagation of the so-called guided waves is affected by changes in thickness of the component, so that they are sensitive to metal loss defects, notably corrosion. The principal advantage is that long lengths, 30m (~100ft) or more in each direction, may be examined from a single test point. The main area of application has been to pipes and pipelines and, in the decade since this technique was introduced commercially, it has been enhanced and refined and has become widely accepted as a method for rapid screening of pipes for corrosion.
The properties of these guided waves in cylinders, i.e. pipes, are well understood and this knowledge has governed the development of commercial guided wave testing systems. For these an encircling bracelet tool is placed around the outside of the pipe to be tested, with the aim of generating an axially-symmetric wave, to send a circular wave along the pipe. An example of such a tool is shown in Figure 1.
Fig.1. Bracelet tool for long range guided wave testing on a 10" (273mm) diameter pipe
The guided waves propagating in pipes are a special case of the Lamb waves which can exist in plates. One of the parameters which affects the propagation characteristics is the pipe diameter. It may be appreciated that, as the pipe diameter increases, the behaviour becomes more like that of the Lamb or plate waves. Equally, as pipe diameters become very large the practicality of mounting an external bracelet tool becomes difficult owing to the size and weight of the tool itself. The aim of this work was to develop tools and techniques applicable to large diameter steel structural tubulars, with particular reference to the 4m diameter foundation piles for 'Monopile' offshore wind turbines.
2. Structure to be examined
The Monopile design for offshore wind turbines, developed by DONG Energy, relies on a large diameter (4.7m) tubular pile driven vertically into the sea bed to support the tower (which may be up to 80m high) and a wind-turbine generator of up to 3.6 MW. Figure 2 shows the pile after driving into the sea bed and the installation of the superstructure.
Fig.2. Top, the monopile after driving and, Below, installation of the superstructure
The 4.7m diameter piles are substantial structures, being up to 50m long and with a wall thickness varying between 45 and 90mm. The weight of each is in excess of 150 tonnes. They are fabricated from ten 3m long tubular cans welded together. Examples of these piles at the fabrication yard are shown in Figure 3.
Fig.3. 4.7m diameter x 30m long piles. The circumferential welds may be seen
Of principal concern is the fatigue loading on the pile caused by the bending stresses resulting from wind loading on the turbine blades, tower and, to some extent, wave loading. Fatigue cracks, should they occur, are likely to be at the circumferential welds, but the initiation of cracking from localised corrosion cannot be ruled out. Consequently, the requirements of an inspection/monitoring system are for detection of both corrosion and cracking. As may be seen from Figure 2
, most, if not all, of the pile is below sea level, so that it is hard to access for inspection. Further, the section below the sea bed is also required to be examined as the bending stresses continue affect the region in the sea floor, at least for part of its length.
The scale of the inspection task is enormous. Whilst the welds are at known locations, so that inspections may be concentrated at those points, the total length of circumferential weld in the pile is 118m. Further, corrosion may potentially occur at any position on the outer or inner surface of the pile. This total surface (inside and out) is 886m2. Thus to make the examination of this component effective it is necessary to employ a method which is capable of covering such a large test area rapidly and one that does not require a sensor to be scanned over the whole surface.
3. Inspection development
3.1 Approach adopted
The ultrasonic guided wave technique is highly suited for the rapid inspection and monitoring of such a large volume of material. However, it is impractical to consider an encircling tool, such as is shown in Figure 1, for the 14.8m long circumference. Consequently it was decided that a linear array of finite length (curved to fit the shape of the pile) would be used to test a strip running down the length of the pile. This tool would then be moved around the circumference in steps to cover the whole of the circumference. The top edge of the pile, clearly visible in Figure 2, is accessible from the inside of the column, as the superstructure is fitted over the pile. It was decided to use an array of compression wave transducers mounted on this top edge to generate the waves of interest.
3.2 Initial experiments
The properties of linear arrays for ultrasonic testing are well known from both phased array applications and SONAR. The following principles were used to develop an array design. Initial studies had shown that the optimum test conditions were achieved when exciting the S0 plate wave mode at a frequency of around 30kHz.
When a number of transducers are assembled together in an array and driven electrically, the velocity of motion of the individual elements is not constant, but varies from element to element in a complex manner due to acoustic interactions between them. These interactive effects may be reduced by:
- Separating the elements of the array,
- Changing the number of element and so the length of the array,
- Making the individual elements large so their self radiation overwhelms the mutual radiation between elements,
- Using individual amplifiers to drive each element at the correct amplitude and phase to obtain a uniform velocity of motion across the array.
3.3 Distance between elements
The distance between each element will affect the number and the dimension of side-lobes from the ultrasound beam. These are not wanted and could produce noise. Ideally, the transducer elements should be infinitely close, in which case there will be few side-lobes with low amplitude. Increasing the distance between elements up to λ/2 will increase the number of side-lobes but not their amplitude. For a separation distance beyond λ/2, the amplitude of the side-lobes perpendicular to the main beam direction will increase.
3.4 Length of the array
The length of the array is directly linked to the beam width of the main beam of the emitted signal. Indeed, a good approximation of the beam divergence is given by:
- λ is the wavelength of the signal,
- L the length of the array,
- θ20db the divergence angle of the -20dB profile of the sound beam.
Using this equation it occurs that, using a distance between elements of λ/2, a reasonable length would be 4λ, which mean using an array of 8 elements.
3.5 Size of the transducer
As explained previously, the bigger the diameter of the transducers, the lower the acoustic interactions, but this reduces the steering ability of the array. It was decided therefore to use a 40mm diameter transducer as a good compromise.
These design principles were tested first by modelling the beam profile of the array using Civa and Quicksonic software. These both indicated that the array had good directionality and acceptable side lobe properties. Second, a series of experiments was carried out on heavy (50mm thick) plates in the TWI laboratory. The set up is shown in Figure 4.
Fig.4. Initial laboratory test on the array. The well defined responses from the far end of the plate may be clearly seen on the screen
The array was driven by a multi-channel Plant Integrity Teletest®
unit which allowed the signal supplied to each element to be controlled independently. This enabled beam focusing and steering to be implemented. Following the successful laboratory tests, the array was tested on the full size piles at the fabrication yard. An example of the test set up is shown in Figure 5
. The aims of this test were:
- To demonstrate that the whole length of the pile could be tested from the top end,
- To establish the sensitivity by examining two 200mm diameter holes present in the pile (these are access holes for electrical cables - one of these holes may be seen in the top left corner of Figure 5),
- To test the effect of focusing the beam,
- To test the effect of beam steering.
Fig.5. Linear array mounted on one of the piles shown in Figure 3
The results were very encouraging. The far end of the pile was clearly seen and the holes were readily detected. Clear signals were also observed from each of the nine circumferential welds. A typical result is shown in Figure 6.
Fig.6. Result from a test on a full scale pile. The far end signal may be seen on the right hand side
The pile end signal may be seen on the right hand side of the trace. However, the most significant feature of these tests was the presence of high amplitude Rayleigh waves which were propagating around the circumference of the pile at the top end. These waves appeared to be generated from each end of the array and were virtually impossible to dampen. Rayleigh waves at megahertz frequencies are readily damped by placing a layer of a substance such as grease on the surface. However, this was ineffective in this case. There were also issues of coupling of the individual transducer elements. Differences in efficiency between the elements produced non-optimum conditions for array performance and affected the focusing and steering behaviour. Consequently, it was decided to conduct a thorough investigation of the array characteristics and coupling effects.
It was clearly impractical to do a long-term study on a full size pile, so a small-scale mock up was required. To do this, advantage was taken of a useful property of guided waves, that their behaviour scales with size. An important parameter which describes guided wave behaviour is the frequency x thickness product, expressed in MHz.mm. For the 50mm thick pile, testing at 30 kHz gives an f x t value of 1.5 MHz.mm. A length of 22" (559mm) diameter schedule 20 line pipe, with wall thickness 9.53mm was used for the studies. By using smaller ultrasonic transducers, operating at 157 kHz the equivalent f x t value was obtained to that on the full scale pile. Diameter to wall thickness ratio is also an important factor. For the full size pile this is 94:1. For the 22" pipe this ratio is 62:1, not as large as for the full scale pile, but big enough that a realistic approximation to the geometry of the full size pile is achieved.
4. Small scale array optimisation
4.1 Set up and hardware (properties of the array)
The array consisted of 8 independent compression transducers (see Figure 7) designed to provide waves at frequencies identified as optimum for the inspection for this pipe size (in the range 80 to 200 kHz). Using the array on the edge of the structure, as shown, the probes will generate longitudinal waves propagating along the pipe. For the experimental set up, the data was collected via a Teletest Focus system.
Unlike the initial design, the transducers were mounted on a rigid frame, which allowed the spacing between the elements to be varied. The frame also ensured that a uniform load was applied to each probe. Even using the mounting frame ensure a uniform load, the coupling efficiency was not the same for each transducer. The coupling therefore needed to be optimised.
Fig.7. Experimental set up
4.2 Improving coupling consistency
In order to improve the consistency of transmission of the wave into the component, some coupling medium has to be used. The most suitable for this application is to use dry coupling by adding a soft material such as neoprene rubber to the face of the transducers (see Figure 8). This reduces the difference of acoustic impedance across the interface and thus allows a better transmission of the wave between the probes and the material under test. A study was carried out into the transmission of the ultrasonic energy for different surface roughnesses. It was shown that for a roughness above an Ra value of 3µm, the transmission coefficient is low, and that the transmission is better at lower frequency.
Fig.8. Rubber coupling layer
In order to see the effect of the rubber on the output signal, some experiments using different thickness of rubber were carried out (see Figure 9):
Fig.9. A-scans, different thickness of rubber
For each A-scan, the Signal-to-Noise ratio (SNR) was calculated, see Table 1:
Table 1. Signal to noise ratio for different thicknesses of rubber
|Thickness of the rubber (mm)||Frequency (kHz)||Ringing zone duration (µs)||Amplitude of the signal (mV)||Amplitude of the noise (mV)||Signal to noise ratio|
It may be seen from Figure 9 and Table 1 that the presence of rubber decreases the transducer ringing and thus improves the quality of the signal. Also, the amplitude of the signal increases considerably with the thickness of rubber. However, the amplitude of the noise behaves in the same way. It may be seen that the signal to noise ratio is maximum when 3mm rubber was used. Therefore, this thickness of rubber was adopted for the experiments.
4.3 Rayleigh wave suppression (spacing of the array)
As found in the field tests on the full scale piles, the linear array also generates Rayleigh waves propagating around the end of the pipe. The Rayleigh wave is a non dispersive surface wave propagating with a velocity slightly less than the shear wave (about 2970 m.s-1) in a steel material. This wave is of high amplitude and can thus hide the other peaks which may appear. One way to cancel this Rayleigh wave is to space the transducers in the array at half the wavelength of the Rayleigh wave, see Figure 10.
Fig.10. Cancellation of the Rayleigh wave
The Rayleigh wave component generated by the first transducer is cancelled by the out of phase wave from the second transducer when the spacing is λ/2. Thus, for a given frequency, the spacing between the transducers (in order to cancel the Rayleigh wave) may be calculated, Table 2 presents the optimum spacing between the transducers, depending on the frequency:
Table 2. Optimum transducer spacing with frequency
The minimum spacing between two transducers is 25 mm, as at this point adjacent elements touch each other. It may be seen from Table 2, this distance is optimum at 60 kHz. Above this frequency, it is impossible to reach the optimum spacing, so it is not possible to suppress the Rayleigh wave by this method at the frequency of interest for the small scale mock-up, which is at around 150kHz.
However, the observed results were not as bad as the predictions in Table 2 would suggest. This may be explained by the fact that most of the energy of the Rayleigh wave is concentrated within one wavelength of the surface. Using a high excitation frequency increases the concentration of the energy in the wave near the surface, so that the Rayleigh wave component is more highly attenuated than at lower frequencies. The effect of this is shown in Figure 11. At 100 kHz the amplitude of the Rayleigh wave is high, with repeat echoes being observed which correspond to the wave travelling around the pipe circumference at the end. The first reflection of the pipe end is hidden behind the peak of the Rayleigh wave at approximately 1300 µs. At 170 kHz, the Rayleigh wave is not present and the reflections from the pipe ends may be clearly seen.
Fig.11. A-scans, influence of the frequency on the Rayleigh wave
4.4. Optimising transducer array design
One of the characteristics of a linear array is that there will be a spreading of the beam from the origin with increasing range. This is illustrated in Figure 12. One of the aims of this study was to reduce the natural divergence of the beam to enable the inspection to be carried out in a series of strips around the pile circumference with reasonable resolution. The circumferential resolution is important as this enables the position of any reflectors in the pile to be located.
Fig.12. Effect of beam divergence
A study was made of the key parameters of the array which affect divergence of the beam. These are: the number of transducers, the spacing between them (see Figure 13) between their centre (Pitch) and their edge (Aperture), the frequency, the velocity of the wave, the material under testing, the couplant and the focus distance.
Fig.13. Spacing between transducers
A number of different arrangements of transducer elements were considered. Table 3 presents the results for different spacing between the probes and for a different numbers of elements. The width of the beam was calculated at a range of 2 metres from the transducer.
Table 3. Influence of the spacing and the number of probes on beam width at 2m.
These results show that the principal influence on the divergence is the width of the array. The number of transducers does not affect the beam geometry, provided that they are close enough to act as a single array.
Nevertheless the beam width is large for all cases studied and over the whole length of the pile there would be insufficient lateral resolution. It was therefore decided to use a delay profile in the signals applied to each of the transducers in order to focus the beam. In fact a number of measures were used to achieve enhanced beam characteristics. These were:
- Increase in frequency to 200 kHz (f x t = 2 MHz.mm) to reduce the Rayleigh wave component, and to reduce the divergent angle (which is proportional to frequency)
- Application of the delay profile to concentrate the energy,
- Use of the smallest array length (200mm) to start with the narrowest beam.
The results are shown in Figure 14. The dark trace is the unfocused signal. The reflection from the far end of the 8.18m long 22" pipe may be seen at approximately 6300µs. The pale trace is the focused result (the focus was at the pipe end). It may be seen that there is an increase in amplitude of 14.5dB (a factor of 5) when the beam is focused.
Fig.14. A-scan, focusing test
These design principles are now being applied to the array geometry for the full scale tests and the procedures are being based on the focusing regime demonstrated on the small scale specimen.
This work was carried out under the European Union funded OPCOM project, which is partly funded by the EC under the Nanotechnologies and Nanosciences, Knowledge Based Multifunctional Materials, New Production Processes and Devices (NMP) programme, Contract Number: NMP2-CT-2005-516993.
- Cawley P & Alleyne DN, 'The use of Lamb Waves for the Long-Range Inspection of Large Structures', Proceedings of Ultrasonics International 95, published in Ultrasonics, Vol. 34, pp287-290, 1996.
- Mudge P J, 'Practical Enhancements Achievable in Long Range Ultrasonic Testing by Exploiting the Properties of Guided Waves', proceedings of the 16th World Conference on NDT, Montreal, Canada, September 2004.