EMAT Generation of Polarized Shear Waves for Pipe InspectionJulie Gauthier, Vasile Mustafa, Ahmad Chahbaz
Corresponding Author Contact:
|TABLE OF CONTENTS|
An ultrasonic guided wave inspection technique was evaluated to detect and locate defects in pipes using SH (Horizontally Polarized Shear) plate waves. Electromagnetic Acoustic Transducers (EMATs) were designed and constructed for bi-mode SH waves applications. Advantages of SH waves, guided by the wall surfaces for circumferential propagation and full volume inspection, are demonstrated in the pulse-echo setup. Mode selection criteria were investigated to tune the SH waves to the geometry of the inspection specimen. Examples of mode selection based on the interaction characteristics of SH wave modes with a defect are described. An application of multi-mode inspection for geometrical parameter evaluation (length, depth and orientation) of individual cracks was also demonstrated. The experimental work was performed on cylindrical steel pipe samples having several types of defects ranging from through wall cracks to shallow cracks (10% of wall thickness) both in the circumferential and longitudinal directions. Results from laboratory investigation of the influence of defect depth and orientation on reflectivity of various modes of SH waves are reported. Multi-mode SH-wave inspection is shown to be capable of detecting shallow discontinuities (microcracks) and of locating defects accurately. Inspection results on a pipe section are represented as 3-D images with B-Scan projections and their interpretation is discussed.
Defects found in pipeline structures range from large discontinuities to small and subtle deterioration process defects such as stress corrosion cracking (SCC). Standard ultrasonic techniques applied for the non-destructive testing (NDT) of pipes include the straight beam method using longitudinal waves and the angle beam method using vertical shear (SV) waves. Although these methods have proven effective, they do not have the advantages of SH waves, which could provide a sensitive, fast, and cost effective NDT method to detect, locate and validate discontinuities due to a variety of materials deterioration mechanisms. To investigate the potential of this approach, the authors developed an EMAT guided wave capability to address specifically the challenge of reliable inspection of pipeline for small defects with SH-waves. The tasks included in this work are inspection procedure development and design of transducers for inspection on strenuous surface conditions. This paper presents initial results on the use of SH-plate waves for pipe integrity monitoring to detect crack-like defects.
SH-plate waves are a family of Lamb waves. These waves can propagate in plate-like structures of a few wavelengths thick or even of the order of one wavelength. They are two dimensional stress waves in infinite plate structures whose surfaces are free of stresses. Their propagation characteristics are tailored to the geometry of the structure inspected (the elastic wave-guide). Their elastic motion covers the whole thickness of the structure (wave-guide) due to the guiding effect of the inner and outer surfaces of the pipe. SH-plate waves have small divergence losses and are attenuated less rapidly than bulk waves, resulting in longer propagation ranges than those for bulk waves with the same frequency and higher sensitivity for defect detection. Therefore, SH-plate waves can propagate for distances of several meters in pipes having a thickness of several millimeters. Furthermore, SH-plate waves can follow curvature thus enabling inspection along bends and other irregular geometry. Principles and techniques for guided waves in piping application waves have been developed (1-5) in the last 15 years by many researchers.
Unlike longitudinal and shear bulk waves, SH-plate waves offer improved inspection potential due to the availability of a host of different propagating modes, each with its own unique characteristics. Modes can therefore be selected to enhance detection of different types of defects such as cracks, holes, and corrosion. SH-plate waves were used (6-8) in various applications.
Another characteristic of the shear horizontal wave family of modes is the propagation displacement, characterized by a null normal displacement on the surface of the material and a particle displacement perpendicular to the propagation direction and always parallel to the surface (9). This feature provides a number of significant advantages over conventional ultrasonic waves. These waves are insensitive to loading, therefore there are no leakage losses in acoustic media such as tar, coatings and water. They do not suffer mode conversion at defect interfaces, a fact which enhances received signal quality since the reflected signals are not encumbered with the noise of parasitic conversion signals and there is no conversion loss of energy. Another advantage provided by these waves is their sensitivity to adhesion defects (6) such as disbond between coating and pipe wall, a site favorable to corrosion problems.
Prior to investigating detection and interaction of SH wave modes with defects in materials with cylindrical geometry, it was necessary to represent their theoretical dispersion curves. This was required to verify the group velocity of the pre-Frequency selected modes and to select the frequency on the dispersion curves at which these modes are excited. The large diameter of the pipe structure allows us to use a flat plate approximation. Therefore, it was not necessary to use the more complex formulation of cylindrical shell modes for a hollow circular cylinder. Dispersion curves depend upon material properties and specimen geometry; in this case the curves are for a steel plate. The curves in Fig. 1 describe the relation between the plate thickness to wavelength ratio versus the frequency-thickness product. Another type of dispersion curve is a plot of group velocity versus mode frequency. Group velocity graphs are used for defect location from the measured arrival time (time-of-flight) of the mode waveform. Figure 2 shows the group velocity for the three first SH modes in a 6.8 mm steel plate.
|Fig 1: Dispersion curves for SH modes in plates|
SH waves were generated in the structure using an EMAT (10, 11). Due to the difficulty of excitation and generation of SH waves, an EMAT is an excellent device, providing the capability to match the excitation force with a specific particle motion. When SH-waves are introduced using these transducers, a number of experimental variables that could not be controlled with specially cut piezoelectric transducers can be selected through frequency adjustment. Frequency control permits wave modes to be generated individually in a given structure. Another advantage of using EMATs is that they do not require a liquid couplant, making it simple to scan the inspection structure. SH wave inspection scans cannot be performed with piezoelectric transducers since they require a viscous couplant to generate horizontally polarized shear oscillations in the material. EMATs can also generate SH-waves over layers of rust or dirt of thickness less than 1-2 mm.
|Fig 2: Group velocity dispersion curves for the three first SH modes in a 6.8 mm steel plate.|
The excitation frequency of individual modes can be found from a simple relation between the EMAT periodicity, the shear wave velocity and the mode number (7). The excitation points are shown in Fig.1 along the horizontal line.
To optimize the guided wave inspection, it is possible to analyze the wave structures to determine the most suitable mode for a geometry/defect combination. These can be calculated for each point on the dispersion curve of each mode. For this preliminary work the analysis was based on the power distribution of each mode across the plate thickness. The time average longitudinal power flux per unit width, Pz, can be derived using the shear stress and the displacement (9) in the plate structure. For optimal defect detection, in-plane displacement and power need to be concentrated at the defect location through the wave guide thickness (12).
All the defects introduced in the cylindrical sample for this project were either located on the inside or outside surface of the pipe wall which implies that the energy and in-plane displacement should be maximized at the pipe wall surface. It is also preferable to have an energy with a reasonable depth distribution to avoid excessive sensitivity to surface defects. After analysis of these two parameters (energy and in-plane displacement) for various modes, SH1 and SH0 were found to be the most appropriate for this preliminary work. The graphs in Fig. 3 show energy and displacement distribution along the normalized wall cross section for the points on the dispersion curves where the two modes are to be excited. These graphs illustrate that the energy and the in-plane displacement for SH1 are located mostly at the pipe inner and outer surfaces with a good penetration and that SH0 has an even "detectability" along the entire wall. This wave distribution allows mode selection and optimization for different types of defects.
|Fig 3: In-Place displacement (a) and power density (b) distribution for SH1 and SH0 modes in 6.8 mm steel plate (thickness normalized).|
The design of the EMAT for this work was based on probes specifically constructed for in-field applications. These probes are sufficiently rugged to slide over rough conditions at high velocities. The EMAT consisted of two coils, one acting has a receiver and the other has a transmitter, having a periodicity (wavelength) of 10 mm. The probe was used to generate the fundamental mode SH0 and the first antisymmetric mode SH1. Propagating modes were selected by changing the excitation frequency according to the mode number and the periodicity of the EMAT. The coil was excited with a burst length of 15 to 20 µs. These values are a compromise between a sufficiently long RF burst to have a stationary resonant excitation of the structure while providing a reasonable resolution in the circumferential direction. A gated RF pulse amplifier excited the emitter coil and the signal from the receiver coil was sent to a broadband receiver. Impedance matching between the probe and the electronic units (amplifier and receiver), was performed for each of the SH modes. The following measurements were made in the pulse-echo set-up with the ultrasonic pulse propagating in the circumferential direction around the pipe, as shown in Fig. 4.
|Fig 4: Experimental setup for SH wave inspection|
The EMAT is by nature bidirectional; therefore waves emanating from the probe travel in the clockwise and counterclockwise directions. Reception being also bidirectional, indications along the acoustic path can be "seen" from two directions. The transducer was excited at 0.30 MHz for SH0 and at 0.39 MHz for SH1; their respective group velocities were 3.22 mm/µs and 2.45 mm/µs. The work was performed on a 1200 mm length of steel pipe having an internal diameter of 490 mm and a wall thickness of 6.8 mm. Results are presented in the form of either A-Scans or B-Scans. Scanning was performed using a small portable scanner unit having a lateral resolution of 0.08 mm. The received signal was pre-amplified and analog filtered using a narrow frequency pass band. Up to 100 successive ultrasonic signals were averaged to improve the signal to noise ratio. A summary of the defects machined on the sample is found in Table 1.
|Machined slot, Axial direction, 0.7 mm, 10% depth ratio, 25.4 mm length||D1|
|Machined slot, Axial direction, 2 mm, 30% depth ratio, 25.4 mm length||D2|
|Machined slot, Axial direction, 3.3 mm, 50% depth ratio, 25.4 mm length||D3|
|Machined slot, 45 o direction, 25% depth ratio, 25.4 length||D4|
|Machined slot, Circumferential direction , 25% depth ratio, 25.4 mm length||D5|
|2 similar machined slots, Axial direction, 20% depth ratio, 1 on the external pipe surface (a) and 1 on the internal surface (b) , 25.4 mm length each||D6a|
|EDM slit, 2.5 mm, 35% depth ratio, 25.4 mm length||D7|
|Machined erosion patch 3 cm 2||D8|
|Three 30% depth ratio defects axially aligned, 38.1 mm length - 10 mm separation||P1|
|Three 30% depth ratio defects circumferentialy aligned, 12.7 mm length - 10 mm separation||P2|
Reference signals for the SH0 and the SH1 modes propagating around the entire circumference of the steel pipe are shown in Fig. 5. The full circumferential travel path is referred to as C. There is a 'dead zone' at the start of the A-Scan of around 70 to 100 mm in the steel due to the main bang, depending on the gain adjustment and the number of cycles in the tone burst. The SH guided waves can travel several circumferences because of their low attenuation properties. The time axis results must be divided by two because of the pulse-propagation echo configuration, except for the full circumference path where the wave pulse travelled the entire perimeter. Both the SH0 and the SH1 modes were sensitive to the presence of the longitudinal weld. The ultrasonic energy reflected from the weld is higher than normal because of the presence of a severe undercut found along the entire weld. The weld reflections are indicated as W1 and W2 (for the two propagation path possibilities). SH1 being a dispersive mode, the received signal is spread much more. The pipe inner surface was very rough in some sections; this condition did not seem to cause amplitude loss of the circumferential signal.
|Fig 5: A-Scan display of propagation of (a) SH0 mode and (b) SH1 mode around pipe.|
The second set of results, shown in Fig. 6, are A-mm Scans of the signal in the pulse echo mode from three defects located in the same circumferential path for the SH0 and the SH1 modes. These defects had different depth ratios: 10%, 30% and 50%. For the first graph, 6a, the EMAT was placed at 224 mm from the 30% defect (D2). The defect caused a good reflection, however the energy loss was sufficient to reduce considerably the reflection from the second defect (D3 - 50% depth ratio). The SH0 was barely sensitive to the 10% (D1) defect. For the SH1 mode the probe was placed at 140 mm from the 10% defect and the two consecutive defects (30% at 540 mm and 50% at 760 mm) were also detected, although the SNR was reduced appreciably by the presence of D2. This is shown in Fig. 6b. This representation demonstrates that SH guided waves are sensitive to the axial defect and that it is possible to detect a series of defects located in line with each other if the depth of the defect is smaller than the effective sound beam. However, power and in-plane displacement of the SH0 mode were not sufficiently high at the surface to have a good detectability for shallow defects located after in line defects.
|Fig 6: A-Scan display of propagation of (a) SH0 mode and (b) SH1 mode around pipe with three defects.|
Two similar defects were machined, one on the inside surface of the pipe in the axial direction and the other on the outside surface in the same direction. Amplitude as a function of time of flight was plotted, Fig. 7, for both the SH0 and SH1 modes. The EMAT probe was placed between the defects at 200 mm from the outer surface defect and 400 mm from the inner surface defect. The location, inner or outer surface, of the defect with respect to the probe did not influence the amplitude response of the ultrasonic echo. The response from the inner surface defect was smaller, but this was attributed to a small misorientation with respect to the pipe axis. These results confirm that the ultrasonic energy was propagating in the entire volume of the pipe wall and that sensitivity for inner and outer surface defects is comparable. The probe could be placed inside the pipe and still detect the defect located at the outer surface.
Two circumferential defects of 25% depth ratio were also machined on the surface, one oriented at 90° and the other at 45° with respect to the pipe axis. The circumferential defect could not be detected by either of the SH modes propagating in the circumferential directions, however it was detected by the same EMAT, with the adaptable curvature, in the axial direction at a distance of 1.1 meter. The 45° notch was detected with a reduction of SNR of more than 20 dB than for a similar notch more favorably oriented. The ground patch, D8 was not detected.The reason for this was the minimal removal of material and the extremely smooth surfaces of the ground area. The ultrasonic energy reflection response to the EDM slit was comparable to the other larger slits. It must be mentioned that the very smooth walls of all these machined defects are excellent reflection surfaces, which would not be the case for a naturally occurring shallow defect such as SCC; this is currently being investigated.
|Fig 7: A-Scan display (a) SH0 mode and (b) SH1 mode around pipe having a defect on the inner andouter surfaces.|
The severity of a colony of microcracks, such as SCC, is related to the length and depth characteristics of an individual crack. Another important consideration for defect colony characterization and assessment is the determination of axial spacings and circumferential spacing between individual cracks since, as cracks grow, they can join to form a larger single crack leading to rupture. It is important to both detect and characterize stress corrosion to accurately assess the pipeline integrity. Examples of guided wave imaging capability are now shown.. The following scans were performed to assess the ability of the SH waves to discriminate between closely aligned defects. These results are a preliminary attempt to analyze geometric parameters of individual crack length, and, axial and circumferential spacing between cracks.
|Fig 8: B-Scan image of (a) single defect, (b) four defects separated circumferentialy and (c) three defects separated axially.|
The scans were performed with the EMAT located at 300 - 400 mm from the defect region. The first serie of B-Scans are shown in a 3-D visualization view for which the amplitude is both color coded in a grey scale format and plotted in the z-axis. The scan is in the x direction, and the time of flight (propagation direction) is in the y direction. The first B scan, Figure 8a, shows the image from a single defect scanned with the SH0 mode, the second image, 8b, shows the image resulting from scanning three defects separated by 10 mm in the axial direction. The last one, 8c, shows the image resulting from three defects separated by 12.7 mm in the circumferential direction. These B-Scan images demonstrate that defects can be isolated in the axial direction but not as easily in the circumferential direction. This poor circumferential resolution is attributed to the low frequency and multiple cycles in the tone burst that are required to efficiently generate the SH modes. However the reasonable divergence of SH0 allows an excellent axial resolution.
The second set of images in Fig. 9 is a result of scanning the same defects with the SH1 mode. The axial resolution obtained with this mode is comparable to the SH0 mode. The ultrasonic amplitude is higher because energy is distributed at the materials surface. It was expected that the circumferential resolution would not be improved by the use of SH1 since this mode is dispersive. This is demonstrated in Fig. 9 where two consecutive images are shown; the left one shows three circumferentialy separated defects and the right one shows a single defect. The circumferential spacing information would require either numerical analysis of the signal or artificial intelligence methods to extract further information.
|Fig 9: B-Scan image using SH1of (a) a single defect followed by four defects separated circumferentialy and (b)three defects separated axially.|
To quantify the parameters of length and spacing, for the series of three 30% defects separated axially, the standard B-Scan display was used, as shown in Fig. 10. The images were analyzed using ultrasonic data analysis software. The parameters measured are the axial spacing between two consecutive defects, S, and the length of the individual defect, L. Cursors were aligned with what seemed to be the interface between a defect and the pipe material and values were calculated using the scanning tool resolution. The results obtained with this method using SH1 correlated well with defects. The separation S1 was established at 10 mm and the individual defect length, L1, at 40 mm. The actual values were of 12 mm and individual defect length of 38 mm. The results obtained with SH0 were not as good. At the current time no analysis has been performed to quantify the individual crack depth.
|Fig 10: Standard B-Scan image of three defects separated axially using SH0 (a) and SH1 (b) modes.|
The results found in this preliminarily work demonstrates the potential of using the SH-plate waves for rapid inspection of piping structures. SH modes can be complimentary to one another in establishing geometrical parameters of individual defects. This technique can be used to detect defects located either on the inner surface or the outer surface of a structure and with the appropriate transducer design the inspection technique can be automated on variable curvature pipe having rough surface conditions. The technique can be especially useful in assessing values of spacing between individual defects and produced repeatable results. Theoretical analysis of the ultrasonic characteristics of the power density and the in-plane displacement still requires further investigation but demonstrates to assist in the selection of propagation mode for a particular application. The next step in this research program is to demonstrate the ability of this method to detect real defects such as SCC.
This work was jointly funded by Tektrend International Inc. and by Precarn Associates as part of the Smart Sensory Structures project.