Array Ultrasonic Inspection of Welds in Lined Pipes

TWI Ltd, Granta Park Great Abington, Cambridge, CB21 6AL, UK
Universidade Federal do Rio Grande do Sul, Lamef, Porto Alegre, Rio Grande do Sul, Brazil

Channa Nageswaran
TWI Ltd, Granta Park Great Abington, Cambridge, CB21 6AL, UK

Thomas Clarke
Universidade Federal do Rio Grande do Sul, Lamef, Porto Alegre, Rio Grande do Sul, Brazil

Paper presented at NDT 2016. The 55th annual conference of the British Institute of Non-Destructive Testing. 12-14 Sept. 2015, Nottingham, UK.

Abstract

In recent years the need for new technologies in oil and gas industry has increased substantially due to the discovery of new oil and gas fields which are hard to access. One specific concern in this industry is the need to increase the reliability of mechanically lined pipes employed offshore. This kind of pipeline contains a complex girth weld joint which experiences fatigue during its operation and plastic deformation is inflicted during its assembly and deployment, i.e. reeling and unreeling. This paper presents the early stage development of a non-destructive inspection technique aiming to assess flaws in lined pipes joints. The proposed technique is based on ultrasonic array methods, including the total focusing method and the time reversal method. This preliminary stage will be conducted by computational modelling and experimental testing in an Alloy 182 dissimilar weld block, which mimic some aspects of lined pipes girth welds. Previous works in the same block investigated its metallurgical structure and grain texture, allowing to account for the inherent anisotropic and inhomogeneity effects of such coarse grained material. The present block analysis is expected to provide useful insight about the liner-weld overlay interface inspection problem.

1. Introduction

The intrinsic anisotropic and inhomogeneous characteristics of austenitic welds are known to create serious issues in the ultrasonic inspections of components which present this kind of joint ^(1-3). This technical difficulty is of concern in some industry sectors such as nuclear and oil & gas which commonly employ austenitic welds in critical joints ⁽¹⁾. The safety operation requirements for components in those industry sectors impose the need for stringent defect assessment of welded parts. Despite the limitations imposed by the microstructure of the weld, ultrasonic inspection is still among the techniques capable of relatively effective inspection of such kind of joints. However, the anisotropy and inhomogeneity issues must be addressed in order to obtain meaningful and reliable results.

Among the components in which austenitic welds are present there are the mechanically lined pipes. Such pipes are the result of the union of two pipes; the first one is made of a corrosion resistant alloy and is assembled inside the second pipe, which is made of carbon steel. The pipes are bonded together through hydroforming ⁽⁴⁾. Lined pipes are employed when it is needed to transport corrosive fluids and is preferred due to lower cost in comparison with a pipe made entirely of a corrosion resistant alloy or metallurgically cladding ^{(5, 6)}.

The union process between two pipe segments is usually done by a girth weld. However, the girth weld is complex due to the presence of the internal liner pipe. In order to produce a weld joint of good quality, initially the liner is removed from the extremity of the pipe. After this, a seal weld is applied in order to avoid the entrance of corrosive material between both pipes. The next step is to apply a clad overlay from the seal weld until the extremity of the pipe. Finally the preparation for the girth weld is made ⁽⁷⁾. A schematic image of the weld region of lined pipes is presented by Figure 1.

It is common to weld the pipe segments together and store it in reels, then they are transported to the installation site and unreeled ⁽⁷⁾. This procedure plastically deforms the pipe and can harm the weld joint. Then during their service life the pipes are subjected to cyclic loads which could damage the piping through fatigue in the girth weld joint region ^{(8, 9)}.

Hence it is essential to use non-destructive techniques in order to evaluate the structural integrity of the girth weld joint and overlay cladding during the service life. Such techniques are intended to avoid the loss of human lives, environmental damage and financial losses. Ultrasonic inspection techniques are frequently employed in the detection of cracks and defects in welded joints ^{(10, 11)}. However, the employment of standard ultrasonic techniques for the lined pipe girth welds and overlay cladding are troublesome and usually cannot provide an accurate estimation of the presence and size of defects ⁽¹²⁾.

2. Literature review

This literature review provides an introduction to anisotropy and inhomogeneity, as well a description of their effects on ultrasonic inspections. Subsequently, a brief description about some of the methods employed in order to take into account the anisotropy and inhomogeneity effects on ultrasonic inspection performance.

2.1 Anisotropy and inhomogeneity

Connolly ⁽¹³⁾ explains that the mechanical anisotropy comes from the direction dependence of the inter-atomic forces, which leads to differing elastic properties in different directions. Since the sound wave velocities depend on the elastic properties, the wave velocities will vary with angle of propagation in an anisotropic material.

In essence, every polycrystalline metal can be classified as anisotropic and inhomogeneous. However, this only becomes a problem for ultrasonic inspection when the grain size is close to the wavelength of the sound wave employed ⁽¹⁾. Generally, as the grains approach one tenth of the wavelength ⁽¹⁴⁾ significant scattering loss will begin to take place.

The main adverse effects of an anisotropic and inhomogeneous material on inspection using ultrasonic methods are listed below ⁽¹⁾:

1. Increased backscattered energy;
2. Increased attenuation;
3. Increased distortion.

When a sound beam reaches a grain boundary some of the energy is transmitted across the boundary, while some of the energy is reflected back. This backscattered energy is responsible for an increase in the background noise of the signals ^{(1, 15, 16)}. Due to the backscattered energy, less energy is available to propagate further into the material, which means the attenuation is increased ^{(1, 16)}. The attenuation is also composed of the absorption of the wave energy by the material; however, the increased backscattered energy is considered the main reason for the increased attenuation ⁽¹⁾. The increased distortion is related to the anisotropy. In an isotropic material the sound beam is expected to travel in a uniform way; however, for anisotropic materials every time the sound beam reaches a grain boundary it changes its direction ⁽¹⁾ and mode conversion may take place. This is known as beam skewing and it is explained in more depth in the book from the German Society for Non-Destructive Testing ⁽¹⁷⁾.

2.2 Time reversal methodology

The first approach to avoid or, at least, reduce the problems caused by the anisotropy and inhomogeneity of coarse grained materials is to reduce the sound frequency. This will increase the wavelength and consequently moves the inspection system towards the Rayleigh regime ⁽¹⁾. Sometimes, however, this approach is not sufficient to bypass the coarse grain issue. Therefore, several methods or procedures have been developed in order to address the inhomogeneous and anisotropic nature of austenitic welds.

Some approaches do not require any information about the weld microstructure, for example in the methods presented by Shahjahan et al ⁽¹⁶⁾ and Aubry and Derode ⁽¹⁸⁾, only the average material velocity are used. Some authors, such as Ye et al ⁽³⁾, Connolly ⁽¹⁴⁾, Ye et al ⁽¹⁵⁾ and Fan et al ⁽¹⁹⁾ employ mathematical models of the weld in order to establish the grain orientations throughout the joint. Another approach makes use of actual measurements of the grain orientations of the weld joint. The orientations can be obtained with, for example, the construction of an orientation map from electron backscatter diffraction (EBSD) data. This is the method chosen for the work presented in this paper and can be found as well in the works of Sorawit ⁽²⁾, Chen et al ⁽¹³⁾ and Nageswaran et al ⁽²⁰⁾. The work follows on from investigations undertaken in the DISSIMILAR project ⁽¹⁾.

The approach followed in this work and elsewhere ^{(1, 2, 13, 20)} uses the EBSD orientation map and the stiffness matrix of the material to create adapted delay laws (ADLs). Through ADLs it is possible to overcome the problems imposed by the austenitic microstructure, since it compensates the distorted wave path and the changes in wave velocity. The ADLs are the equivalent of ordinary delay laws (DL) which are the fundamental mechanism when using array probes, where time delays are applied to the elements of the array to fire each element in a probe at different times ⁽¹⁾ in order to control the sound field in the component. Due to the complexity of the austenitic materials microstructure, the ADL is not as easily calculated as DL and requires more complex mathematical methods to be obtained, such as finite element modelling (FEM). The ADL calculation involves the modelling of the path and time taken for a wave beam leaving an emitting element, reaching a designated position inside the material and returning to the receiving element. Figure 2 illustrates the difference between ADL and DL.

As found in the DISSIMILAR project ⁽¹⁾ it is possible to obtain the ADL experimentally under certain circumstances. However, only the modelling approach is presented in this paper.

3. Methodology

This section presents the methodology employed in this study. Firstly, the test sample will be described. Secondly, the FEM method will be introduced and then, the imaging method. Finally, the experimental methodology will be explained.

3.1 Test sample

In this a dissimilar weld block presented in the Figure 3 was used. This test sample was originally prepared for the DISSIMILAR project ^{(1, 2, 20, 21)}. EBSD maps have been created for this block and similar maps are being created for the lined pipe application.

The EBSD map for this sample is presented in Figure 4. The microstructure was considered to be composed of 6 different grain orientations that are presented in Table 1. The block contains 5 side drilled holes with diameter of 3 mm to be used as targets for the ultrasonic inspections. In addition, the block is made mainly of Alloy 182 ^{(1, 2, 20)}. The stiffness constants, for cubic symmetry, as used in previous works ^{(1, 2, 10)}, namely C11=203.6 GPa, C12=133.5 GPa and C44=129.8 GPa, were used in the present work in order to calculate the mechanical properties for each grain orientation.

Table 1: Grain orientations from the orientation map.

3.2 FEM method

Between the several FEM packages available, PZFlex was chosen to be used in this work. Its main advantages includes relatively low computation times and the ability to introduce the orientation map into the model with high precision. However, the modelling results were not highly accurate due to the 2D expression ⁽²⁾. Nonetheless, it was found that, for this work, 2D FEM was sufficiently accurate for the purpose of generating ADLs.

The model constructed for this study was capable of imaging the central area of the block with 40 mm in the dimension parallel to the front wall and the whole depth of the block which was 35 mm. A 15 mm water gap was used between the transducer and the block. The transducer used in the model was created with the geometrical characteristics of the one used in the experimental evaluation in order to be able to compare both results. A structured mesh composed of squares was used in this model. The mesh size was calculated as one thirtieth of the wavelength of the sound wave in water for the water layer. While, for the block a mesh size equal to one thirtieth of the longitudinal sound wave wavelength for isotropic Alloy 625 material was used.

When simulating or acquiring experimental data with a linear phased array probe it is possible to obtain the Full Matrix Capture (FMC) data, which is a N,N,t matrix where N denotes the number of transducers in the array and t the number of time samples acquired. By manipulating this data matrix it was then possible to calculate the ADL within it. This adapted FMC data can then be employed in image reconstruction algorithms, such as the total focusing method or TFM, and also for conventional point focusing. The approach used to obtain the time of flight (TOF) from the simulated data was similar to the one employed by Pamel et al ⁽²²⁾. In their work, instead of insonifying the model with the transducer elements, the authors employed the defect to insonify the model. In the present work, in order to image an area of the block a grid (with step size of 1 mm) of points was created over the region to be imaged (16 mm wide and 25 mm deep). In every node of the grid, one at a time, a circular vacuum region was assigned to simulate a reflector. The boundary of this vacuum region was excited with a tone burst whose central frequency was set to be equal to the transducer central frequency. The simulation was run for a time long enough for the sound waves to reach the transducers on the boundary of the model. With this approach it was possible to calculate the TOF for every point in the region to be imaged.

The approach described above was employed in this work with a view to reducing the time taken to run a complete model over the grid of points to obtain the TOF. While this method needs to be run only once for every grid point, the traditional method of insonifying the transducer elements needs the model to run once for every element in the array and for every point in the grid. Since, only half of the wave path is modelled, i.e. from the focusing point to the transducer, for time domain models there is a further reduction by a factor of two in the total modelling time. Roughly, the total time required to run the model with this approach was reduced by a factor equal to twice the number of transducers in the array.

Once in possession of the TOF for the grid of points, the FMC data was required to be able to calculate the results. To obtain the FMC data from the FEM simulation each element was excited one by one while a circular vacuum region was inserted with the dimensions and in the position of the side drilled hole in the actual block. The model was run for a time long enough for the sound wave reach the furthest grid point from the transducer and return to all the array elements.

3.3 Imaging method

Once the FMC data is available, regardless of its FEM or experimental origin, it is possible to manipulate it in several ways to create meaningful images and graphs. Here, two of the possible methods, developed in Matlab, will be described, namely the ADL point focusing linear electronic scan and a modification of the TFM method. Both methods depend on the TOF data previously calculated with FEM. The point focusing employs only the diagonal of the FMC matrix, where the transducer which insonifies the material is the same one that receives the signal reflections back after its interaction with the piece being analysed. However, the TFM needs the whole FMC matrix to produce the images.

The algorithms for both the methods are similar and they rely upon equations 1a and 1b, below:

In equation 1a, “LPF” is the summation of the signal strength received by the transducers employed in the step “s” for the coordinate “xs”; A is the scan aperture; “hm,m” is the A-scan extracted from the FMC matrix for the case where the sender and receiver is the transducer “m”; “TOFxf,yf,m,m” is the time taken by a sound beam leaving transducer “m” reaches the focusing point (xf, yf) and returns to the same transducer “m”. In equation 1b, “ITFM” is the summation of the signal strength received by all sender/receiver transducer combinations in the array; “hm,n” is the A-scan extracted from the FMC matrix for sender “m” and receiver “n”; “TOFx,y,m,n” is the time of flight taken by a sound beam leaving transducer “m” reaches the focusing point (x, y) and returns to the transducer “n”. Figure 5 illustrates both analysis methods.

As a way to increase the quality of the results, a cross correlation test was used to discard the A-scans which do not correlate well with the input tone burst. Usually, the discarded A-scans are those with a high incidence angle with respect to the front wall of the block. Since there are reflections from inside the block that are not related to any defect, a FEM model was run in order to collect the baseline FMC data from the block without the defect. The data obtained in this way was subtracted from those with defects in order to obtain the simulated images.

3.4 Experimental methodology

The experimental data for this work was collected in an immersion tank using a linear phased array probe with 64 elements, 0.75 mm pitch, elevation of 12 mm and 2.25 MHz central frequency. From the 64 available elements, only 32 were employed. The probe was controlled by a M2M MultiX++ phased array system and data collected using the M2M Multi 2000 software. A gain of 62 dB was used in order to compensate for the high attenuation in the material. As a consequence, the A-scans saturated in the vicinity of the front wall. However, since the side drilled hole to be imaged is positioned 15 mm below the front wall, this saturation was not a problem. A Matlab algorithm, similar to those applied to the analysis of simulation results, was employed on the experimental FMC data.

4. Results and discussion

The results obtained in this work are presented in this section. First, a reference image built with a standard TFM algorithm considering the block as isotropic and homogeneous will be presented. Following this, the simulated and experimental results from both single point focusing scan and proposed TFM methods are shown.

An experimental reference image of the side drilled role created with TFM considering the block as homogeneous and isotropic is presented in Figure 6, below. The longitudinal wave velocity was assumed to be 5800 m/s.

Figure 6 shows that the indication from the side drilled hole is positioned on its right side with its maximum located approximately 3.5 mm away from the expected position which is the upper surface of the hole. Also, there is a strong signal level in the upper part of the figure that is due to the high gain which was required to compensate for the attenuation. The high signal intensity in the vicinity of the front wall, which is located at depth equals to 15 mm, makes it difficult to image any reflector in this region. One can also find a spurious indication in the bottom of the image. The highest amplitude from the spurious indications is 70% of the maximum amplitude from the side drilled hole.

The TFM image obtained with the algorithm which applies the Equation 1b over the simulated data was capable of detecting the side drilled hole with higher accuracy and with much weaker spurious indications, as can be seen in Figure 7a. Note that due to the nature of the model the saturation observed in the top of the experimental image shown in Figure 6 is not present in the simulated data. In Figure 7b it shows, for comparison, the same image as in Figure 7a without applying the cross-correlation test or the subtraction of the baseline data. The increase in the image quality obtained with the cross-correlation test and baseline data subtraction allowed for a reduction of the spurious indication amplitude levels below 6 dB of the maximum amplitude of the side drilled hole indication.

Figure 8 presents the TFM image obtained with the simulated TOF data and the experimental FMC. Similarly to Figure 6, the saturation on the top of the image is present. However, note that the quality of the image resembles that of Figure 7a. The side drilled hole indication is, as in Figure 7a, located in the expected place which validates the model for this case. Except for the saturation at the top, the spurious indications are below 6 dB of the maximum amplitude from side drilled hole indication – ie good signal-to-noise performance was achieved. The data from which Figure 8 was produced were subjected to the cross-correlation test but the subtraction of the baseline simulated image was not applied as it worsens the image quality.

The single point electronic scan results from the algorithm which uses Equation 1a are presented in Figure 9. The beam was focused using an ADL for the point located in coordinates (41.75 mm, 28.50 mm), which is the top of the side drilled hole. The aperture of the scan was 15 elements which meant that there were 18 steps to sweep throughout the 32 elements available on the simulated array probe. To match the simulation, only 32 elements of the experimental phased array probe were used. The amplitude of both scans is in arbitrary units and normalized for the highest value. It should be noted that, away from the defect region, there is a strong disagreement between experiment and simulation. However, in both scans the peak value is 6 dB above the background noise. These results suggest that the single point focusing method can be used as an alternative to the TFM method. Nevertheless, further investigation about the discrepancies between experimental and simulated results is required.

5.  Conclusions

This work presents an imaging procedure capable of detecting side drilled holes in an anisotropic and inhomogeneous weld block. The imaging grid resolution significantly affects the image quality. However, using very fine grid would require an excessively long simulation times, hence an appropriate balance must be found. The evidence shows that it is possible to predict and image with good signal-to-noise performance and sufficient accuracy the side drilled hole both using TFM and single point focusing electronic scan using ADL. However, further work must be conducted in order to fully optimize the image resolution. The results presented in this work suggest that it is worth to investigate the applicability of the proposed method for lined pipes girth weld and overlay cladding integrity assessment.

6. Acknowledgements

The authors would like to thanks the Brazilian governamental agency Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and BG Group for sponsoring this research throught the Science Without Borders Program.

7. References

1. Dissimilar Consortium, 'Phased Array Ultrasonic Inspection of Dissimilar Metal Joints', TSB Project No: P11/MFE/6/I/AA058J, January 2011.
2. S. Limthongkul, 'Validation of wave propagation in austenitic welds using vibrometer. MSc Dissertation, Brunel University London, London, 2015.
3. J. Ye, J. Moysan, S.J. Song, H. J. Kim, B. Chassignole, C. Gueudre, O. Dupond, 'Influence of welding passes on grain orientation - the example of a multi-pass V-weld'. International Journal fo PressureVessels and Piping, Vol 93-94, pp.  17-21, 2012.
4. S. Denniel, T. Tkaczyk, A. Pepin, 'Reeled Mechanically Lined Pipe - Cost efficient Solution for Static and Dynamic applications in corrosive Environment', Deep Offshore Technology Conference, Perth, 27-29 November 2012.
5. B. E. Urband, G. Prideco, 'CRA Clad Downhole Tubing - An Economical Enabling Technology', National Drilling Conference, Houston, 2001.
6. D, Wilmot, P. Montague, 'The Suitability of CRA Lined Pipes for Flowlines Susceptible to Lateral Buckling', SUT Global Pipeline Buckling Symposium, Perth, 2011.
7. L. Yuan, S. Kyriakides, 'Liner wrinkling and collapse of bi-material pipe under bending',International Journal of Solids and Structures, Vol 51, pp 3-4,  2014.
8. T, Sriskandarajah, G. Roberts, V. Rao, 'Fatigue Aspects of CRA Lined Pipe for HP/HT Flowlines', Offshore Technology Conference, Rio de Janeiro, 2013.
9. T, Tkaczyk, A. Pépin, S. Denniel, 'Fatigue and fracture of mechanically lined pipes installed by reeling'. Proceedings of the ASME 2012 31st International Conference on Ocean: Offshore and Arctic Engineering, Rio de Janeiro, 2012.
10. P. A. Mix, 'Introduction to non-destructive testing: A training guide', 2 ed, Hoboken: Wiley, 2005.
11. L. Schmerr, S. J. Song, 'Ultrasonic Nondestructive Evaluation Systems Models and Measurements', New York: Springer, 2007.
12. D, Hu, Q. Wang, K. Xiao, Y. Ma, 'Ultrasonic phased array for the circumferential welds safety inspection of urea reactor', Procedia Engineering, Vol 43, pp 459–463, 2012.
13. G. D. Connolly, Modelling of the propagation of ultrasound through austenitic steel welds, PhD Thesis, Imperial College, London, 2009.
14. Y. Chen, Z. Luo, Q. Zhou, L. Zou, L. Lin, 'Modeling of ultrasonic propagation in heavy-walled centrifugally cast austenitic Stainless steel based on EBSD analysis', Ultrasonics, Vol. 59, pp. 31–39, 2015.
15. J. Ye, H. J. Kim, S. J. Song, 'Development of an Ultrasonic Ray Model for Phased Array Ultrasonic Testing in Austenitic Weldments'. 17th World Conference on Nondestructive Testing, Shanghai, 25-28 October 2008.
16. S. Shahjahan, A. Aubry, F. Rupin, B. Chassignole, A. Derode, 'Improvement of Flaw Detection with Ultrasonic Array Probes in Multiple Scattering Polycrystalline Materials by Means of a Random Matrix Approach'. 13th International Symposium on Nondestructive Characterization of Materials, Le Mans, 20-24 May 2013.
17. Deutsche Gesellschaft für Zerstörungsfreie Prüfung (German Society for Non-Destructive Testing, Handbook on the ultrasonic examination of austenitic and dissimilar welds. DGZfp/DVS Media: Düsseldorf, 2008.
18. A. Aubry, A. Derode, 'Random Matrix Theory Applied to Acoustic Backscattering and Imaging In Complex Media', Physical Review Letters, Vol. 102, 2009.
19. Z. Fan, A. F. Mark, M. J. S. Lowe, P. J. Withers, 'Nonintrusive Estimation of Anisotropic  Stiffness Maps of Heterogeneous Steel Welds for the Improvement of Ultrasonic Array Inspection', IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control, Vol. 62, pp. 1530-1543, 2015.
20. C. Nageswaran, C. Carpentier, Y. Y. Tse, 'Microstructural quantification, modelling and array ultrasonics to improve the inspection of austenitic welds' Insight, Vol 51, No 12, December 2009.
21. L. J. Cunningham, A. J. Mulholland, K. M. M. Tant, A. Gachagan, G. Harvey, C. Bird, 'The detection of flaws in austenitic welds using the decomposition of the time-reversal operator'. Proc. R. Soc. A, 2016.
22. A. Van Pamel, P. Huthwaite, C. R. Brett, M. J. S. Lowe, Numerical simulations of ultrasonic array imaging of highly scattering materials, NDT and E International, Vol 81, pp 9-19, 2016.