Experimental-Numerical Investigation of a Steel Pipe Repaired with a Composite Sleeve

: Pressure vessels are subjected to deterioration and damage, which can signiﬁcantly reduce their strength and loading capabilities. Among several procedures nowadays available to repair damaged steel pipelines, composite-repairing systems have become popular over the past few years to restore the loading capacity of damaged pipelines. This study reports a numerical-experimental investigation performed for a composite-repaired pipeline made of API 5L X60 steel. An experimental burst test was carried out on a 4 m long pipe section, closed by two lateral caps, and tested up to failure by means of high-pressure water. In parallel, the test was numerically replicated through a FEM model of the composite-repaired steel tank, allowing for a cross-comparison of results. It was found that the composite repairing system has almost eliminated both the noteworthy thickness reduction of 80% and the related stress concentrations in the pipe body. These outcomes allow for a better understanding of these repairing procedures in order to drive their subsequent optimization.


Introduction
The main goal of high-pressure vessel maintenance is to avoid leaks and ruptures that may result in substantial environmental consequences, as well as potential risks of fire of flammable gas clouds, not to mention the enormous risks of injuries and human fatalities. To reduce the failure probability and at the same time keep the system design capacity, the technical conditions of pipelines are periodically controlled by in-line inspections, which are scheduled in order to detect potential damages or defects that can endanger the system loading capacity [1][2][3][4][5]. Part-wall metal loss and defects of buried pipelines are mostly a consequence of electrochemical corrosion caused by oxidation. In these cases, repairing of damages is performed when exceeding threshold damage sizes according to the codes [6]. According to [7], more than 60% of the world's oil and gas transmission pipelines are over 40 years old. Most of these pipelines are in urgent need of rehabilitation in order to work at the desired, and continuously increasing, operating capacity. Within this scenario, the failures due to corrosion and their associated repairing techniques have recently attracted intense research interest [8][9][10][11][12][13].
Replacements of the pipe segments are especially complicated when the pipeline maintenance operations require a pause in the fluid transport and distribution, which can generate substantial expenditures or income losses. When areas of corrosion or other damages to operating pipelines are identified, there are often significant economic and environmental incentives for repairing without removing the pipeline from service. From

Experimental Test
An experimental test was carried out on a steel tank obtained by cutting a 4 m long pipeline section and by using two semi-elliptical end caps welded to the pipe ending sections (Figure 1). The material of the pipe was the API 5L X60 steel. Mean diameter and average thickness were equal to 510 mm and 8.8 mm respectively. The damage was simulated in the middle of the pipe by machining a localized area so as to achieve a considerable local wall loss (nearly 80%), in such a way as to simulate the effects of corrosion. The so-obtained defect presented sizes of 254 mm × 128 mm × 7.1 mm, see Figure 1. The machining process was carried out in a controlled manner using a milling machine in order to obtain a defect uniform in size. This is a common practice in simulating external corrosion effects on a pipeline [16].  The overall procedure to obtain the repaired tank is shown in Figure 2. The repairing procedure with the composite sleeve consisted of the following steps: the whole test. Readings of the manometer were recorded in a video format to enable the analysis of pressure over time. A monotonically increasing internal fluid pressure was applied up to the burst of the tank.

Numerical Simulations
A numerical replication of the experimental burst test was performed for validation and to improve the understanding of the failure mechanisms. ABAQUS [41] was used as FEM package software to create the CAD model for the global and local models, generate the finite element meshes, perform the global-local FEM calculations, and post-process the results.
The CAD model of the repaired pipe according to the burst test was reported in Figure 3 and in Figure 1 with some highlights of the defect position and size. The considered material was the API 5L X60 steel used for the seamless pipe shown in Figure 2. The main mechanical properties considered here are listed in Table 1, whereas the true stress true strain diagram is reported in Figure 4. With reference to the repairing materials, Tables 2 and 3 comprise the main mechanical properties of the composite sleeve and the adhesive filler (Methyl Methacrylate Monomer), respectively. The sleeve was made up of multiple E-glass fibre composite layers with the direction of fibres in the tangential direction to the pipe (direction of the hoop stress). Since the composite sleeve presented orthotropic mechanical properties (Table 2), a local cylindrical coordinate system was adopted, see Figure 5. considered material was the API 5L X60 steel used for the seamless pipe shown in Figure  2. The main mechanical properties considered here are listed in Table 1, whereas the true stress true strain diagram is reported in Figure 4. With reference to the repairing materials, Tables 2 and 3 comprise the main mechanical properties of the composite sleeve and the adhesive filler (Methyl Methacrylate Monomer), respectively. The sleeve was made up of multiple E-glass fibre composite layers with the direction of fibres in the tangential direction to the pipe (direction of the hoop stress). Since the composite sleeve presented orthotropic mechanical properties (Table 2), a local cylindrical coordinate system was adopted, see Figure 5.    considered material was the API 5L X60 steel used for the seamless pipe shown in Figure  2. The main mechanical properties considered here are listed in Table 1, whereas the true stress true strain diagram is reported in Figure 4. With reference to the repairing materials, Tables 2 and 3 comprise the main mechanical properties of the composite sleeve and the adhesive filler (Methyl Methacrylate Monomer), respectively. The sleeve was made up of multiple E-glass fibre composite layers with the direction of fibres in the tangential direction to the pipe (direction of the hoop stress). Since the composite sleeve presented orthotropic mechanical properties (Table 2), a local cylindrical coordinate system was adopted, see Figure 5.     3D CAD and FEM models were set up. The final assembled CAD model is shown in Figure 6, whereas the corresponding FEM model is shown in Figure 7. The machined volume was considered as completely filled with the adhesive material. Due to the insignificance of the stress distribution in the filler material, only its elastic behaviour was considered (material data listed in Table 3). A total of 8 composite reinforcing layers were 3D modelled around the damaged section. The orientation of the fibres was assigned by defining local coordinates for each layer, as shown in Figure 5. The direction of axis 1 represents the (tangential) direction of fibres that were directed in such a way to reinforce the sleeve against the highest stresses (hoop stress). Axes 2 and 3 represent the directions perpendicular to the fibres inside and outside the layer planes, respectively.  3D CAD and FEM models were set up. The final assembled CAD model is shown in Figure 6, whereas the corresponding FEM model is shown in Figure 7. The machined volume was considered as completely filled with the adhesive material. Due to the insignificance of the stress distribution in the filler material, only its elastic behaviour was considered (material data listed in Table 3). A total of 8 composite reinforcing layers were 3D modelled around the damaged section. The orientation of the fibres was assigned by defining local coordinates for each layer, as shown in Figure 5. The direction of axis 1 represents the (tangential) direction of fibres that were directed in such a way to reinforce the sleeve against the highest stresses (hoop stress). Axes 2 and 3 represent the directions perpendicular to the fibres inside and outside the layer planes, respectively.
A coarse mesh was produced for the global FEM model, see Figure 7a. Consequently, the sub-modelling technique was used to improve the accuracy of the FEM results. This sub-modelling consisted of the modelling of smaller parts of the overall CAD model, thus allowing for much finer meshes and more detailed modelling in these localized regions. Displacement boundary conditions, calculated by means of the global FEM model, were applied to the cut sections of submodels in order to replicate the displacement fields within these localized regions of interest. Once the displacement fields were replicated locally, highly accurate local stresses and strains were calculated accordingly. For these reasons, sub-modelling represents a widely applied technique when there is the need to obtain the local fields of interest with high accuracy, e.g., for fracture [42][43][44][45][46] or fatigue [47] assessments. Two submodels were created so as to calculate the data for the repaired and the healthy pipe sections, see Figures 6 and 7. With submodel 1 (SM1), the stress distribution in the defective section, filled with the adhesive and reinforced with the composite layers was examined. With submodel 2 (SM2), the stress distribution was examined in an undamaged section. The number and type of elements used in each model are listed in Table 4.
A non-linear quasi-static analysis was performed with a constantly increasing internal pressure applied to the entire internal surface reaching up to 204 bar. Such pressure value was set up according to the maximum pressure value recorded during the burst test.     A coarse mesh was produced for the global FEM model, see Figure 7a. Consequently, the sub-modelling technique was used to improve the accuracy of the FEM results. This sub-modelling consisted of the modelling of smaller parts of the overall CAD model, thus allowing for much finer meshes and more detailed modelling in these localized regions. Displacement boundary conditions, calculated by means of the global FEM model, were applied to the cut sections of submodels in order to replicate the displacement fields within these localized regions of interest. Once the displacement fields were replicated locally, highly accurate local stresses and strains were calculated accordingly. For these reasons, submodelling represents a widely applied technique when there is the need to obtain the local fields of interest with high accuracy, e.g., for fracture [42][43][44][45][46] or fatigue [47] assessments.
Two submodels were created so as to calculate the data for the repaired and the healthy pipe sections, see Figures 6 and 7. With submodel 1 (SM1), the stress distribution in the defective section, filled with the adhesive and reinforced with the composite layers was examined. With submodel 2 (SM2), the stress distribution was examined in an undamaged section. The number and type of elements used in each model are listed in Table 4. A non-linear quasi-static analysis was performed with a constantly increasing internal pressure applied to the entire internal surface reaching up to 204 bar. Such pressure value was set up according to the maximum pressure value recorded during the burst test.

Results and Discussion
The failure occurred for an unrepaired section, demonstrating that the repairing technique was able to carry a pressure value even higher than the undamaged part. This can be observed in Figure 8, in which the deformations obtained during the experimental test and from the FEM global model are compared. It is worth noticing that the repaired part presents a deformation much smaller than the undamaged part.
Two pictures of the pipe after failure are shown in Figure 9. It can be noticed from Figure 9a that the failure occurred for an integer part of the pipe with a fracture direction perpendicular to the hoop stress, i.e., the maximum stress. This is rather typical for pipelines, pressure vessels as well as rotating components [42,43,46,47]. Moreover, Figure 9b shows that the machined defect did not present any further damage after the test, thus highlighting again the load-carrying capacity of this composite-repairing technique. Figure 10 shows stress and plastic strains calculated numerically for the overall pipe (i.e., for the global FEM model). It is worth noting that the composite sleeve allowed to significantly reduce the deformation of the steel in the repaired section, i.e., a null plastic strain was calculated, in turn, corresponding to a behaviour of the material within the linear-elastic range. On the contrary, the integer parts present stresses having much higher values, corresponding to significant plastic deformations. This added structural loading capacity was allowed by the filler material that transferred the pressure from the steel to the outer composite layers.
The distribution of results in the global FEM model was calculated in order to have a general understanding of the overall stress-strain distribution. More accurate results were calculated by means of the two sub-models shown in Figure 6, for which further geometric details and much finer meshes were adopted. Stresses and strains were extracted from the submodels' results along the two paths shown in Figure 11. Namely, Path1 (Figure 11a) included all the nodes for the inner surface along one-quarter of the repaired pipe section starting from the middle of the damage. Path2 (Figure 11b) was similar to Path1 but considered the inner surface of the composite sleeve. Finally, Path3 was simply defined along the radial direction (i.e., through-the-thickness) of an unrepaired part of SM2. geometric details and much finer meshes were adopted. Stresses and strains were extracted from the submodels' results along the two paths shown in Figure 11. Namely, Path1 (Figure 11a) included all the nodes for the inner surface along one-quarter of the repaired pipe section starting from the middle of the damage. Path2 (Figure 11b) was similar to Path1 but considered the inner surface of the composite sleeve. Finally, Path3 was simply defined along the radial direction (i.e., through-the-thickness) of an unrepaired part of SM2.  geometric details and much finer meshes were adopted. Stresses and strains were extracted from the submodels' results along the two paths shown in Figure 11. Namely, Path1 (Figure 11a) included all the nodes for the inner surface along one-quarter of the repaired pipe section starting from the middle of the damage. Path2 (Figure 11b) was similar to Path1 but considered the inner surface of the composite sleeve. Finally, Path3 was simply defined along the radial direction (i.e., through-the-thickness) of an unrepaired part of SM2.  The variations of von Mises stresses along Paths1-2 are plotted against the related normalized distances in Figure 12. Radial displacement on the deformed shape of the steel part in the damaged section is shown in Figure 13. To better understand the related stress variations, some markers were added in Figure 12 corresponding to the positions shown in Figure 13. Vertical and horizontal axes in Figure 13 were selected as the starting and ending point of the normalized distance, respectively, whereas the latter increased in a clockwise direction. The size of the defect along the normalized distance was equal to 0.16 (Point 1 in Figures 12 and 13).  The variations of von Mises stresses along Paths1-2 are plotted against the related normalized distances in Figure 12. Radial displacement on the deformed shape of the steel part in the damaged section is shown in Figure 13. To better understand the related stress variations, some markers were added in Figure 12 corresponding to the positions shown in Figure 13. Vertical and horizontal axes in Figure 13 were selected as the starting and ending point of the normalized distance, respectively, whereas the latter increased in a clockwise direction. The size of the defect along the normalized distance was equal to 0.16 (Point 1 in Figures 12 and 13).
Considering the results of Figures 12 and 13, the maximum value of von Mises stresses occurred between the range 0 (T direction) to 0.16 (Point 1), thus along the damaged area. Such a result is consistent with the fact that a reduced thickness induces a steep increment of stresses. At Point 1, further stress concentrations were expected due to the sharp changes in the thickness of the pipe, whereas between Points 1 and 2, the increment of the pipe wall resulted in a considerable reduction of stresses. According to these explanations, also the radial displacements presented a similar trend (Figure 13), and the highest displacement was observed between Points 1 and 2. Even though the lowest thickness was in the damaged area, the presence of the filler material allowed to transfer pressure stresses from the thin layer of steel to the outer sleeve, in turn resulting in high but not excessive magnitudes for displacement and stresses. As a matter of fact, maximum displacement occurred just outside the damaged part between Points 1 and 2.  The variations of von Mises stresses along Paths1-2 are plotted against the related normalized distances in Figure 12. Radial displacement on the deformed shape of the steel part in the damaged section is shown in Figure 13. To better understand the related stress variations, some markers were added in Figure 12 corresponding to the positions shown in Figure 13. Vertical and horizontal axes in Figure 13 were selected as the starting and ending point of the normalized distance, respectively, whereas the latter increased in a clockwise direction. The size of the defect along the normalized distance was equal to 0.16 (Point 1 in Figures 12 and 13).
Considering the results of Figures 12 and 13, the maximum value of von Mises stresses occurred between the range 0 (T direction) to 0.16 (Point 1), thus along the damaged area. Such a result is consistent with the fact that a reduced thickness induces a steep increment of stresses. At Point 1, further stress concentrations were expected due to the sharp changes in the thickness of the pipe, whereas between Points 1 and 2, the increment of the pipe wall resulted in a considerable reduction of stresses. According to these explanations, also the radial displacements presented a similar trend (Figure 13), and the highest displacement was observed between Points 1 and 2. Even though the lowest thickness was in the damaged area, the presence of the filler material allowed to transfer pressure stresses from the thin layer of steel to the outer sleeve, in turn resulting in high but not excessive magnitudes for displacement and stresses. As a matter of fact, maximum displacement occurred just outside the damaged part between Points 1 and 2. Further, stress and displacement variations between Points 2 to 4 can be explained by looking at the plastic hinge generated between positions 3 and 4, see also Figure 14. Such plastic hinge occurred in the position where the minimum displacement was calculated. This represents also the reason why the increase in stress in the composite layers in this geometric position was observed ( Figure 12). After Point 4, the plastic zone vanished and small changes in stress and displacements were noticed, with the stress distribution becoming uniform moving away from the damaged area. The trend of von Mises stresses for Path2 were similar to those for Path1. Considering that the radius and thickness of the composite layers were larger than those of the pipe, the magnitude of stresses were slightly lower but presented a comparable behaviour. Further contour plots of von Mises stresses and plastic strains for SM1 are reported in Figure 15. It can be observed that the maximum values of stress and plastic strains were reduced if compared to the global model ones. This is due to the addition of the fillet radii to the defect that was not explicitly modelled for the global model. In addition, an appropriate density of elements in this zone was also considered for the SMs, thus improving the accuracy of the analyses.     Considering the results of Figures 12 and 13, the maximum value of von Mises stresses occurred between the range 0 (T direction) to 0.16 (Point 1), thus along the damaged area. Such a result is consistent with the fact that a reduced thickness induces a steep increment of stresses. At Point 1, further stress concentrations were expected due to the sharp changes in the thickness of the pipe, whereas between Points 1 and 2, the increment of the pipe wall resulted in a considerable reduction of stresses. According to these explanations, also the radial displacements presented a similar trend (Figure 13), and the highest displacement was observed between Points 1 and 2. Even though the lowest thickness was in the damaged area, the presence of the filler material allowed to transfer pressure stresses from the thin layer of steel to the outer sleeve, in turn resulting in high but not excessive magnitudes for displacement and stresses. As a matter of fact, maximum displacement occurred just outside the damaged part between Points 1 and 2. Further, stress and displacement variations between Points 2 to 4 can be explained by looking at the plastic hinge generated between positions 3 and 4, see also Figure 14. Such plastic hinge occurred in the position where the minimum displacement was calculated. This represents also the reason why the increase in stress in the composite layers in this geometric position was observed ( Figure 12). After Point 4, the plastic zone vanished and small changes in stress and displacements were noticed, with the stress distribution becoming uniform moving away from the damaged area.
The trend of von Mises stresses for Path2 were similar to those for Path1. Considering that the radius and thickness of the composite layers were larger than those of the pipe, the magnitude of stresses were slightly lower but presented a comparable behaviour. Further contour plots of von Mises stresses and plastic strains for SM1 are reported in Figure 15. It can be observed that the maximum values of stress and plastic strains were reduced if compared to the global model ones. This is due to the addition of the fillet radii to the defect that was not explicitly modelled for the global model. In addition, an appropriate density of elements in this zone was also considered for the SMs, thus improving the accuracy of the analyses.
Considering SM1, the von Mises stress distribution through the thickness (Path3) was also extracted, see Figure 16. The maximum von Mises stress values measured along the paths in the repaired section, in the composite layer, and for an undamaged section were 506.5, 390, and 554.5 MPa respectively. For the undamaged section (SM1), the maximum stress value was measured as 9% higher than the maximum value for the repaired part (SM1), thus demonstrating that the repairing technique was able to increase the maximum loading capacity of the repaired structure. Considering SM1, the von Mises stress distribution through the thickness (Path3) was also extracted, see Figure 16. The maximum von Mises stress values measured along the paths in the repaired section, in the composite layer, and for an undamaged section were 506.5, 390, and 554.5 MPa respectively. For the undamaged section (SM1), the maximum stress value was measured as 9% higher than the maximum value for the repaired part (SM1), thus demonstrating that the repairing technique was able to increase the maximum loading capacity of the repaired structure.
Additionally, if looking at the overall maximum stress values (Figures 15a and 16), the highest stresses in the damaged part and in the healthy part are nearly 3% different from each other. This clearly indicates that repairing has almost eliminated both the noteworthy thickness reduction of 80% and the related stress concentration in the pipe body.   Considering SM1, the von Mises stress distribution through the thickness (Path3) was also extracted, see Figure 16. The maximum von Mises stress values measured along the paths in the repaired section, in the composite layer, and for an undamaged section were 506.5, 390, and 554.5 MPa respectively. For the undamaged section (SM1), the maximum stress value was measured as 9% higher than the maximum value for the repaired part (SM1), thus demonstrating that the repairing technique was able to increase the maximum loading capacity of the repaired structure.
Additionally, if looking at the overall maximum stress values (Figures 15a and 16), the highest stresses in the damaged part and in the healthy part are nearly 3% different from each other. This clearly indicates that repairing has almost eliminated both the noteworthy thickness reduction of 80% and the related stress concentration in the pipe body.

Conclusions
This study reported an experimental and numerical investigation of a compositerepaired steel pipeline.
A burst test was performed on an API 5L X60 steel pipeline section, for which Additionally, if looking at the overall maximum stress values (Figures 15a and 16), the highest stresses in the damaged part and in the healthy part are nearly 3% different from each other. This clearly indicates that repairing has almost eliminated both the noteworthy thickness reduction of 80% and the related stress concentration in the pipe body.

Conclusions
This study reported an experimental and numerical investigation of a compositerepaired steel pipeline.
A burst test was performed on an API 5L X60 steel pipeline section, for which corrosion damage was replicated by machining 80% of the local wall thickness. Such damage was repaired by means of a Methyl Methacrylate Monomer adhesive filler enclosed by a sleeve of E-glass composite layers. Numerical FEM simulations were then aimed at replicating the experimental burst test, providing a good agreement with the experimental observations.
The conclusions of this investigation can be summarized as follows: • inspections of pipelines are scheduled to detect potential damages or defects, but their complete understanding must be provided upfront in order to be effective during quick maintenance actions. The understanding of the stress-strain distributions in the surroundings of the damaged and repaired area proposed here provide useful data for engineers on such advanced repairing techniques. • a plastic hinge is generated by the internal pressure in the composite repaired section of the pipe. Few explanations of such phenomenon are currently available in similar studies, whereas an accurate description of the mechanical behaviour of the plastic hinge on the repaired pipe was here provided. • the values of the highest stresses in the damaged part and in the healthy part are nearly 3% different from each other. This result indicates that repairing has almost eliminated both the noteworthy thickness reduction of 80% and the related stress concentration in the pipe body.