In the oil and gas transportation industry, CO₂ corrosion is a primary cause of failure of carbon steel pipelines. Carbon steel is still widely utilized due to its economic applicability despite inevitable corrosion [[1], [2], [3], [4]]. The general mechanisms governing the uniform CO₂ corrosion in stable aqueous solutions are pretty much well understood [[5], [6], [7], [8]]. The most predominant product of carbon steel CO₂ corrosion consists of iron carbonate (FeCO₃) [9], which deposits on carbon steel surface as concentrations of Fe²⁺ and CO₃^2- near the wall exceed the solubility constant of FeCO₃ (K_sp) [10]. Some studies confirmed that formation of dense CO₂ corrosion scales might block ion transfer since the corrosion scale possess good corrosion resistances when deposited on carbon steel surfaces [11]. Sometimes, corrosion products could contain Fe₃C characterized by effective conductivity and porous properties, useful for promoting the corrosion processes [12,13]. Therefore, the composition and morphology of the corrosion scales should impact the corrosion processes of carbon steel.

The corrosion processes become more complex when the pipelines are subjected to flow conditions. Numerous studies have shown that the flow conditions accelerated the corrosion processes significantly [[14], [15], [16], [17], [18], [19], [20]]. Under flow conditions, the composition and micro-morphology of the corrosion scales would change, not only due to the obvious increase in mass transfer but will also impact the corrosion scales. In recent decades, numerous reports dealing with the destruction of corrosion scales by wall shear stress (WSS) under flow conditions have been published. The outcomes of these studies look similar whether using rotating disk electrode [[21], [22], [23], [24]], impingement jet [25,26] or regular loop systems [[27], [28], [29], [30]]. However, since large diameter circular or rectangular pipes are often used in loop systems, the fluid could hardly excise strong stress force on the wall, even under high flow velocities.

To accurately evaluate the effects of WSS on corrosion scales, a thin channel test system was previously tested [28]. This kind of test channels is characterized by high aspect ratio and strong WSS in the flow. The WSS in the flow could be measured by a floating element wall probe [31]. However, the high single WSS difficulties to mechanically destroy the protective corrosion scales or corrosion inhibitor films. The presence of uneven defects in the pipeline steel would influence the local flow field, resulting in elevated WSS and turbulence. However, these WSS and turbulence are still insufficient for destruction of the protective corrosion scales or corrosion inhibitor films. Therefore, local corrosion is suspected to occur as a result of cavitation from fluctuating pressure.

Numerous investigations dealing with local corrosion have so far been reported [[32], [33], [34]]. After initiation of local corrosion, the surface defects formed on carbon steel surfaces would certainly impact the local flow field. These surface defects should induce higher turbulence, superior WSS, and elevated flow field fluctuation in the local flow field. Under the influence of local defects-induced flow field, the changes would affect composition and microstructure of the corrosion scale. However, how these changes affect the corrosion process on surface defect and nearby area remain unclear. The study of such problems is closely related to development of local corrosion under flow conditions, and will be of great engineering significance to find solutions for expansion and deepening of local corrosion.

To investigate the influence of local defect on the flow field near the defect, as well as the effect of change in flow field on the corrosion scale morphology and corrosion process, loop system and improved thin test channel were used. × 80 pipeline steel planar specimens (PS) and specimens with arc defects (SAD) were prepared, and then the corrosion tests in CO₂-saturated NACE solution at different flow velocities were tested. Electrochemical measurements during the corrosion test, including open circuit potential (OCP) and electrochemical impedance spectroscopy (EIS) were measured. After the corrosion test, the micro-morphologies and compositions of the corrosion scales deposited on the test surfaces were characterized by scanning electron microscopy (SEM), X-ray diffraction (XRD) and energy dispersive spectrometry (EDS). The localized corrosion pits after the corrosion test were characterized by 3D measuring laser microscopy Computational fluid dynamics (CFD) flow field simulations were also performed to clarify the effects of surface defects on the formed local corrosion scales and corrosion processes. The results suggested that the local defect induced-flow field had a significant effect on both composition and micro-morphology of the corrosion scales in different regions, relevant to gain a better understating of the development mechanisms of local corrosion in pipeline steel.

All the corrosion test specimens were cut from a sheet of X80 pipeline steel, with chemical compositions listed in Table. 1. To compare the effects of surface defects on corrosion processes, PS and SAD were machined into square shapes with the corrosion test surface area of 1 cm². The PS possessed the same shape with SAD but without arc defect, and the SAD details are shown in Fig.1. The arc defects on the specimens were similar with corrosion pits produced by local corrosion on X80 pipeline steel. All specimens were sequentially ground with 400, 600, 800 and 1000 grit silicon carbide sand papers, degreased with acetone, and then rinsed with deionized water.

A CO₂-saturated NACE solution containing 5.0 g sodium chloride and 0.5 g acetic-acid dissolved in 94.5 g pure water was prepared for the corrosion tests. The solution was deoxygenated by purging CO₂ (99.96 %) gas for 12 h prior to the corrosion tests. The pH of the CO₂-saturated NACE solution was measured as 2.7 ± 0.02. The flow of CO₂ was maintained throughout the corrosion tests to ensure CO₂-saturation. All the corrosion tests were conducted at 40 ± 1 ℃ under atmospheric pressure.

A home-made high WSS experimental system was used (Fig. 2). The test channel consisted of plexiglass with 600 × 100 × 4.5 mm in size. The fluid distribution box was arranged at both ends of the channel to ensure uniform distribution of the solution. The test port was located at bottom of the test channel, by adjusting the vertical height followed by connection with a screw thread (Fig. 3). The other containers and passageways in contact with the solution were all made of plastic, and solution pump was constructed from corrosion resistant plastic. The solution temperature was controlled by a cooling pipe set up in the solution tank and made of 316 L steel. The mean volumetric flow was measured by a vertical rotor flow meter. The electrode locations were displayed in Fig. 4. The working electrode (WE), counter electrode (CE) and reference electrode (RE) were all arranged in an equilateral triangle with electrode center spacing of 15 mm [29].

The electrochemical measurements were conducted on a three electrodes system. The X80 pipeline steel specimens were used as WE, high purity zinc as RE, and platinum as CE. The electrochemical measurements were tested by a Solartron 1287 + 1255B electrochemical system. The corrosion time of each group was set to 16 h. The OCP and EIS measurements were carried out continuously during the whole corrosion period. EIS was monitored during the corrosion process, with a sinusoidal potential excitation of 5 mV at frequencies varying from 1 MHz to 10 mHz.

After the corrosion test, removed the specimens from the test channel, rinsed with deionized water, and then dried in vacuum dryer. The micro-morphologies and elemental compositions of the corrosion scales were analyzed by SEM (FEI QUANTA FEG2500), XRD (Bruker D8 Advance), and EDS (Oxford Inca Energy X-Max-50). The surface of the SAD was divided into three parts according to the flow direction and location of the defects: a) area in front of defect, b) defect area, and c) area behind defect. By considering different regions were affected by defects induced turbulence, two points were selected in area (a) and labeled as (a-1) and (a-2). Three points were selected in area (b) and labeled as (b-1) to (b-3). Six points were selected in area (c) and labeled as (c-1) to (c-6). The locations of SEM test areas and test points are illustrated in Fig. 5. The localized corrosion pits of scale-removing corrosion tests specimens were analyzed using 3D measuring laser microscopy.

The flow field near the defects in the test channel was obtained by computational fluid dynamics (CFD) simulations. Two-dimensional model of 600 × 4.5 mm longitudinal section was built with the arc defect located in its center. The fluid was assumed as incompressible, and a standard k-epsilon turbulent model was employed to numerically solve the simulations. The turbulence intensity was set to 5%, and hydraulic diameter (D_h) was calculated according to Eq. (1):

where A is the cross section area of test channel (m²) and P is cross section perimeter (m). The initial conditions and boundary conditions for CFD simulations asfollow:

(1) Inlet: set to velocity-inlet.

(2) Outlet: settooutflow, an environment pressure of 101,325 pa.

(3) The grid: 10 mm length centered on the defect set to 0.025 × 0.025 mm with grid size of both ends of 0.5 × 0.025 mm.

(4) Solid wall: considered as adiabatic with a roughness of 10 μm.

(5) Flow type: turbulent flow as calculated.

During the corrosion tests, carbonic acid is formed in CO₂-saturated solution [35]. The cathodic reactions could be summarized by Eqs. (2) to (5):

The cathodic reaction is mainly affected by pH of the solution. Low pH (2.70 ± 0.02 in this solution) media contain high H⁺ concentrations. The increase in pH would directly affect the formation of HCO₃^- and H₂CO₃ [9].

The multi-step dissolution of carbon steel might be summarized by the following anodic reactions Eqs. (6) to (7) [36]:

During the corrosion of carbon steel, the Fe²⁺ and CO₃^2- (or HCO₃-) concentrations near the interface between carbon steel and solution would increase gradually until reaching super-saturation of FeCO₃ and formation of FeCO₃ corrosion scale on the surface [26]:

3.2.1. Planar specimens

The EIS spectra of the PS at flow velocity of 3 m/s are depicted in Fig. 6(a). During the first 6 h of the corrosion, only one-time constant was identified as a capacitive semicircle, characterizing the active stage of the interface when X80 pipeline steel was exposed to CO₂-saturated NACE solution. An electrochemical equivalent circuit obtained by fitting the experimental impedance data is depicted in Fig. 7(a), where R_s is the solution resistance, R_ct is charge transfer resistance, and Q_dl is constant phase element (CPE) related to the double-charge layer capacitive [28]. By considering the roughness and inhomogeneity of the material surface, constant phase angle element Q_dl was used in place of ideal capacitive C. The CPE impedance $({{\text{Z}}_{\text{CPE}}})$was described as in Eq. (11):

where Q is a proportional factor, j equals to (-1)^1/2, ω represents frequency, n is a factor with values between 0 and 1.

The impedance parameters are listed in Table. 2. During the first 6 h of the corrosion, Q_dl increased with corrosion time but R_ct decreased. This was mainly attributed to existence of Fe₃C in X80 pipeline steel as skeleton of the metal remained on the surface after the corrosion [37]. The existence of Fe₃C was conducive to formation of micro-batteries with ferrite (alpha-Fe) [29]. The residual Fe₃C after the corrosion of steel acted as an interfacial electrical conductor to promote the corrosion reactions during the initial stage of the process [38,39]. After 6 h of the corrosion (Fig. 6), a capacitive semicircle at high frequencies and another capacitive semicircle at low frequencies were identified. An electrochemical equivalent circuit was employed to fit the experimental impedance data (Fig. 7(b)), where Q_pf is a constant phase element of porous scale and R_pf is resistance of the porous scale. The impedance parameters are compiled in Table 2. Q_dl and Q_pf rose with the corrosion time but R_ct and R_pf declined. This indicated that after a certain time of reaction, the residual Fe₃C issued from steel and FeCO₃ formed by the reaction were mixed to form a porous scale. The micro-pores acted as ions transfer micro-channels during the reaction, further promoting the corrosion rate [40].

As flow velocity reached 5 m/s and 7 m/s, the electrochemical impedance spectra looked basically the same for the PS corrosion test. The only difference was the shortening in formation time of the porous scale, confirming the negative impact of high flow velocity on formation of porous scale. However, the single high WSS caused by high velocity was insufficient to destroy the loose porous corrosion scale.

The SEM micro-morphologies of the PS after the corrosion are shown the formed scales loose and porous (Fig. 8). Fig. 9(a) shows the XRD patterns of the PS after the corrosion at flow velocity of 3 m/s. The XRD analysis show that the corrosion scale on the corrosion test surface are Fe₃C and FeCO₃. The surface elemental compositions analyzed by EDS are gathered in Fig. S1, with molecular ratio of Fe₃C: FeCO₃ calculated as 18: 5. This also confirmed that the corrosion scales under single high WSS were composed of Fe₃C skeleton with less FeCO₃ precipitate. Therefore, the loose porous mixed scales of PS is not stripped under single WSS of flow field. The WSS is cause by the viscosity of the fluid. The fluid near the static wall slows down due to the viscosity, and the slows down fluid reduced the velocity of the fluid layer above them, cause the velocity gradient of fluid, and the WSS defined as:

where μ is the viscosity of the fluid, $\frac{du}{dy}$dud_y is the flow velocity gradient with a vertical wall distance y. For a rectangular channel of single phase was determined according to Eq. (13) [31]:

where τ is time-averaged shear stress (Pa), ρ is fluid density (kg/m³), C_f is fanning factor, and V is mean flow velocity (m/s). C_f is a function of the surface roughness and Reynolds number (Re). For rectangular channel flow, C_f can be calculated by the Patel correlation presented in Eq. (14):

For rectangular channel, D_h from Eq. (1) could be used to calculate Re. The WSS was estimated to 137.2 Pa, at maximum flow velocity for PS of 7 m/s. According to the literature [41,42], strengths of integrated corrosion scales composed of FeCO₃ might reach 1-10 MPa. However, the strengths of loose corrosion scales remained undetermined. Here, the strength of loose and porous scales composed of Fe₃C and FeCO₃ was greater than 137.2 Pa.

3.2.2. Specimens with arc defect

3.2.2.1. Low-degree turbulence

The EIS spectra of SAD at flow velocity of 3 m/s are presented in Fig. 6(b). During the corrosion tests, a capacitive semicircle in the high and medium frequencies range, and an inductive loop in the low frequency range were identified. An electrochemical equivalent circuit was employed to fit the experimental impedance data are shown in Fig. 7(c), where L is the inductance, R_L is the inductive resistance (Table. 3). During the whole corrosion period, Q_dl enhanced with the corrosion time but R_ct declined, which would mainly be caused by presence of Fe₃C in carbon steel [12,37,43]. Compared to PS, the inductive loops of SAD at low frequencies showed that local flow field turbulence caused by surface defects still negatively impacted the corrosion scales even at low flow velocities. Under local flow field induced by local defects, the loose sediment formed by Fe₃C and FeCO₃ on steel surface was continuously stripped away and failed to form a complete corrosion scales. The impedance fitting data showed a decrease in R_L with the corrosion time, and R_L was found directly proportional to FeOH_ads coverage of the surface. According to the anodic reactions shown in Eqs. (7) and (8), the dissolution of FeOH_ads increased the concentration of Fe²⁺ at the interface and promoted the formation of FeCO₃ deposition. This led to formation of incomplete porous mixed scales during the corrosion. Due to low rate of stripping than that of deposition under low flow velocities, porous mixed scales would gradually form by extending the corrosion time.

The SEM micro-morphologies of the SAD after the corrosion test are shown in Fig. 10, and the CFD simulation results of turbulent kinetic energy near the arc defect area are provided in Fig. S2. The local turbulence induced by defects induced different micro-morphologies of corrosion scales at different locations. The region (a) located upstream of local defect was less affected by the turbulence, leading to integrated porous corrosion scales similar to that of the PS corrosion test at position (a-1). Near the defect, the position (a-2) led to formation of incomplete bulk corrosion scales. Due to turbulent flow (Fig. S2(a)), the porous corrosion scales became stripped off from the surface. The region (b) showed maximum turbulence degree at positions (b-1) and (b-3) but minimum turbulence degree at position (b-2). This induced corrosion scales with various micro-morphologies at different positions. The region with maximum turbulence depicted loose scales stripped off from the surface to form complete and compact corrosion scales, especially the right side of position (b-1) and left side of position (b-3). By comparison, the position (b-2) at the bottom of arc defect with minimum turbulence displayed formation of porous corrosion scales. On the other hand, region (c) located downstream of the defect was affected by defect induced turbulence, leading to formation of damaged and incomplete corrosion scales at positions (c-1) to (c-4). The large distance from defect to the positions (c-5) and (c-6) led to gradual reduction in turbulence effects, yielding complete porous corrosion scales.

The compositions of the corrosion scales were analyzed by XRD and EDS, and showed similar elemental compositions as the PS corrosion test. However, the arc defects caused higher WSS and turbulence, leading to strong strip of loose scales. The molecular ratio of Fe₃C: FeCO₃ was calculated as 9: 12. At flow velocity of 3 m/s, CFD simulations estimated the maximum WSS on the surface of SAD to 184.2 Pa. When combined with analyses results of the PS corrosion test, the strength range of mixed porous corrosion scales composed of Fe₃C and FeCO₃ was identified as 137.2 Pa-184.2 Pa.

3.2.2.2. Medium-degree turbulence

The EIS spectra of the SAD at flow velocity of 5 m/s are depicted in Fig. 6(c). During the first 9 h of the corrosion, a capacitive semicircle in the high and medium frequencies range, and an inductive loop in the low-frequency range were identified. An electrochemical equivalent circuit was employed to fit the experimental impedance data are gathered in Fig. 7(c). The impedance parameters obtained from 1 to 9 h of the corrosion test are listed in Table. 4. During the first 9 h of the corrosion, Q_dl and R_ct rose with the corrosion time, indicating an increase in charge transfer resistance with time at high flow velocity. As local flow field turbulence increased, the effect of fluid on corrosion scales rose as well, especially for loose sediment Fe₃C. As a result, FeCO₃ was identified as the corrosion product deposited on the surface under high flow velocities. On the other hand, the deposition of FeCO₃ hindered charge transfer between the solution and steel surface, which increased the R_ct value. At the corrosion time of 10 h or longer, two capacitive semicircles were identified. These features were related to formation of complete corrosion scale on the specimen surface after 10 h of the corrosion test. An electrochemical equivalent circuit was utilized to fit the experimental impedance data (Fig. 7(d)), where R_df is the resistance of the dense scales and Q_df is CPE of the dense scales. The impedance parameters obtained between 10-16 hours of the corrosion test are listed in Table. 5. As flow velocity reached 5 m/s, dense corrosion scales was formed on the surface of the SAD, which looked pretty resistance to the corrosion process.

The SEM surface micro-morphologies of the SAD subjected to a flow velocity of 5 m/s after the corrosion are presented in Fig. 11. As flow velocity rose to 5 m/s, the turbulent kinetic energy of upstream of defects reached 0.5 (m²/s²) (Fig. S2(b)), greater than most positions obtained at flow velocity of 3 m/s. The turbulence stripped off the loose deposits to form dense corrosion scales in region (a). In region (b), loose deposits were also stripped off, leading to formation of complete and compact corrosion scales at positions (b-1) and (b-3) under maximum turbulence. On the other hand, porous corrosion scales were induced at position (b-2), located at bottom of defects under minimum turbulence. By comparison, region (c) formed complete and compact FeCO₃ corrosion scales due to elevated turbulence intensity at positions (c-1) and (c-2), leading to formation of partially damaged dense corrosion scales.

The XRD analysis showed that the corrosion scale was FeCO₃ at flow velocity of 5 m/s (Fig. 9(b)). The atomic of Fe: C: O ratio from EDS analysis confirmed FeCO₃ as the corrosion scale (Fig. S3) [44]. At 5 m/s, the maximum WSS determined from CFD simulations was 364.5 Pa, which looked far from 1-10 MPa level required to destroy dense corrosion scales [45,46]. However, the local dense corrosion scales at positions (c-1) and (c-2) were partially damaged.

3.2.2.3. High-degree turbulence

The EIS spectra of the SAD subjected to flow velocity up to 7 m/s are gathered in Fig. 6(d). A capacitive semicircle at high and medium frequencies, as well as an inductive loop at low frequencies were identified. An electrochemical equivalent circuit was utilized to fit the EIS data (Fig. 7(e)), and the impedance parameters are listed in Table. 6, Q_dl decreased with the corrosion time but R_ct increased. This suggested rapid strip off of loose corrosion scales formed on the surface by local turbulence at high flow velocities, and led to formation of compact FeCO₃ corrosion scales. However, the dense corrosion scales could not be maintained on the surface under high flow velocities. The local turbulence possessed sufficiently energy to destroy the densely formed FeCO₃ corrosion scales, inducing local damage of the scales, which could explain the presence of inductive loop at low frequencies.

The SEM micro-morphologies of the SAD at flow velocity of 7 m/s after the corrosion are shown in Fig. 12. The WSS of region (a) estimated from CFD simulations was 148.3 Pa. The porous corrosion scales looked stripped off entirely, meaning lower strength than 148.3 Pa. According to Section 3.2.2.1, the strength range of porous scales can further be reduced to reach values between137.2 Pa and 148.3 Pa. The same phenomenon was observed for region (b), formed porous scales. It means that the corrosion process at the bottom of the defect was promoted by the porous corrosion scales, which might be explained by the occurrence of deep corrosion in small holes. However, the corrosion scales in region (c) showed damages under high flow velocity. The formation of complete corrosion scales was observed in position (c-1). But in positions (c-2) to (c-6), the dense corrosion scales looked obviously damaged. As distance in downstream of defect increased, the single damage area of corrosion scales became small but number of damaged spots rose.

The elemental compositions of the corrosion scales at flow velocity of 7 m/s were also analyzed by XRD and EDS, and the results were shown similar results to the SAD at flow velocity of 5 m/s, confirming FeCO₃ as corrosion product after the corrosion. The maximum WSS at flow velocity 7 m/s obtained from CFD results was 593 Pa, which was lower than the strength required for destroying dense corrosion scales (1-10 MPa).

The Eqs. (9) and (10) shown that with the reaction of carbon steel in CO₂-saturated solution, FeCO₃ will precipitates from solution when the Fe²⁺ and CO₃^2- (or HCO3-) concentrations reaching the solubility product, K_sp,FeCO3, which is a function of temperature and ionic strength, and can be describes as [47]:

where the ${{\left[ F{{e}^{2+}} \right]}_{eq}}$ and ${{\left[ CO_{3}^{2-} \right]}_{eq}}$ are the equilibrium aqueous concentration of Fe²⁺ and $CO_{3}^{2-}$. The activation energy of FeCO₃ precipitation [48] $\Delta {{G}_{FeC{{O}_{3}}}}$ = 73.7 kJ/mol.

The total energy E of the flow exert on the corrosion scale can be calculated, which includes the thermal energy K_t and kinetic energy E_k. The thermal energy K_t is only related to the temperature of the solution, and can be expressed as [49]:

where R is the gas constant, T is the absolute temperature (K). In this work, the temperature of the solution is 313 K and the thermal energy is calculated to be K_t = 3.9 kJ/mol.

The kinetic energy is produced by the orderly motion of the fluid, which is related to the mean flow velocity, and can be calculated as:

where m is the molecules molar mass (kg/mol), v is the flow velocity (m/s). In this work, the NACE solution is mixed by three components (water, sodium chloride, acetic-acid), the molar mass of the solution is obtained according to the weighted average, and can be calculated as

where ${{m}_{{{H}_{2}}O}}$, ${{m}_{NaCl}}$ and ${{m}_{C{{H}_{3}}COOH}}$ is the molar mass of H₂O, NaCl and CH₃COOH, respectively, kg/mol. The average molar mass is calculated to be m= 20.235 kg/mol. Under the 7 m/s flow condition, the kinetic energy is calculated to be E_k= 0.05 J/mol. The results illustrate that the energy produced by the flow is far below the activation energy of FeCO₃ (73.7 kJ/mol).

For uneven surfaces, sudden changes in pressure field near the defects might lead to vaporization of the liquid phase. The CFD simulation results of absolute pressure near the arc defect area is shown in Fig. 13. As distance downstream of defect increased, cavitation will occur at the pressure recovery point, with enough energy to destroy the dense corrosion scales [50]. However, the lowest pressure points in flow field appeared between positions b-3 and c-1, and minimum absolute pressure at flow velocities of 5 m/s and 7 m/s were 91,564 Pa and 83,404 Pa. These values were larger than the vaporization pressure of water under experimental conditions.

The NACE solution remained CO₂-saturated throughout the whole experiment. Therefore, the reason for cavitation would unlikely be related to vaporization of water but rather to release of CO₂ gas at low pressure. The CO₂-saturated solution was very unstable and could easily be released when the flow field decreased by small margin [51,52]. The potential energy (E_p) of a single spherical cavitation bubble could be calculated as Eq. (19) [53]:

where r is the bubble radius (m), P_∞ is the environmental pressure at infinite distance from the bubble (Pa), and P_V is release pressure of CO₂ (Pa), and. The potential energy of the cavitating bubble is calculated to be about 75 kJ/mol, enough to destroy the corrosion scale.

When CO₂-saturated solution flow moved through the lowest pressure point, large amounts of CO₂ were released, forming several big-sized bubbles. As the bubbles continued to move downstream with flow field to reach the lower point, the turbulence would cause some big-sized bubbles to collapse and form small-sized bubbles. The cavitation and fragmentation of big-sized bubbles have enough energy to destroy the dense corrosion scales, resulting in the damage seen in Fig. 12(c-2) to (c-4). The resulting small-sized bubbles from fragmentation will then move with the flow, further collapsing and breaking at the downstream region. This caused the damage observed in Fig. 12(c-5) and (c-6).

After the corrosion test, all the corrosion scales of specimens were removed. First of all, using SEM to analysis the X80 corrosion in different areas of each specimens, and located the localized corrosion areas. Second, using 3D measuring laser microscopy to characterize micro localized corrosion pits.

After removed the corrosion scale, the SEM of PS removed scale shows uniform corrosion under the uniform flow condition (Fig. S4). The 3D surface morphology and profile of the PS at flow velocity of 3 m/s is shown in Fig. S5. Because the plane flow is was insufficient to destroy the loose and porous corrosion scale, which cause the whole test surface was covered evenly. Therefore, the corrosion of PS surface was uniform and without localized corrosion pits. The 3D surface morphology results were basically the same when flow velocity reached 5 m/s and 7 m/s for PS.

The SAD surface has different flow field characteristics in different area. The area (a) of the defect, because it is not affected by defect, shows the same SEM micro-morphologies and 3D surface morphology as the PS (Fig. 14). The area (b) also formed a complete loose and porous corrosion scale due to the low turbulent. As shown in Fig. 14 and Fig. S6, the corrosion in the area (b) was uniform corrosion, and without obvious localized corrosion pit.

The SEM of area (c) of SAD shows different localized corrosion characteristics at different flow velocity (Fig. 14), and the 3D surface morphology and profile shows that the localized pits were different. At flow velocity of 3 m/s, the loose corrosion scale was damaged by the turbulent, and the loose and porous corrosion products are dispersed on the steel surface. The corrosion process of area (c) covered by the loose and porous corrosion products was promoted, because the products act as ion microchannel. As shown in Fig. 15(a), there was a localized corrosion pit on the surface. However, because the stripped rate of the corrosion scale by turbulence was lower than the deposit rate of the corrosion scale at low flow velocity, the localized corrosion process lasts for a short time and the localized corrosion pit was shallow. When the flow velocity was 5 m/s (Fig. 15(b)), the loose corrosion scale at the area behind the defect was completely stripped off after 9 h, and the protective corrosion scale is formed. Therefore, the localized corrosion occurs only in the first 9 h of the corrosion test. The protective corrosion scale at the area (c) was damaged at flow velocity of 7 m/s (Fig. 15(c)), the steel was directly exposed to the solution and lead heavy local corrosion. The localized corrosion rate was calculated by [54]:

where $Co{{r}_{L}}$ is the localized corrosion rate (mm/y); d is the depth of localized corrosion pits (μm).

The localized corrosion rate and the depth of localized corrosion pits at the area behind the defect were presented in Fig. 16. It is proved that at low turbulence, the localized corrosion occurs at the area covered by the dispersed loose corrosion scale. At medium turbulence, the localized process was not obvious. At high turbulence, the protective corrosion scale was damaged, and the localized corrosion process was promoted.

The development mechanism of local corrosion in X80 pipeline is very complex under flow condition. Because the corrosion process involves the coupled mechanisms of electrochemical reactions and fluid dynamics. In this paper, the development mechanism of local corrosion in X80 pipeline was investigated by the preparation of special morphology samples, and by using electrochemical test, CFD simulation and surface testing technology. The following conclusions are drawn from the study:

1) The high turbulent kinetic energy and shear stress induced by defect could effectively strip off the loose corrosion scales (Fe₃C), leading to formation of dense corrosion scales (FeCO₃). The strengths of the obtained loose corrosion scales ranged from 137.2 Pa to 148.3 Pa.

2) Due to low turbulent kinetic energy and WSS, the loose mixed corrosion scales could easily form at the bottom of defects (Fe₃C), promoting the corrosion at the bottom and deepening the defects.

3) Under high flow velocities, the defects induced high shear stress (593 Pa) but insufficient to achieve the strengths of dense corrosion scales (1-10 MPa).

4) The damage to the dense corrosion scales (FeCO₃) downstream of defects was caused by cavitation and fragmentation of bubbles. The bubbles formed in CO₂-saturated solution were identified as CO₂ gases released under pressure fluctuations.

This work was supported by the National Natural Science Foundation of China with (No.51774314), Natural Science Foundation of Shandong Province with grant number (No. ZR2018MEM002), and the Fundamental Research Funds for the Central Universities (No. 19CX05001A) for financial support.

Supplementary material related to this article can be found, inthe online version, at doi:https://doi.org/10.1016/j.jmst.2019.10.023.