Corrosion Science 50 (2008) 30053020

Contents lists available at ScienceDirect

Corrosion Science
journal homepage: www.elsevier.com/locate/corsci

Experimental investigation on gasliquid two-phase slug ow enhanced carbon dioxide corrosion in vertical upward pipeline
Donghong Zheng, Defu Che *, Yinhe Liu
State Key Laboratory of Multiphase Flow in Power Engineering, Xian Jiaotong University, Xian 710049, China

a r t i c l e

i n f o

a b s t r a c t
The carbon dioxide corrosion behavior of API N80 grade steel enhanced by gasliquid two-phase vertical upward slug ow has been both mechanistically and experimentally investigated. It is found that the hydrodynamic characteristics of slug ow, such as the direction alternated wall shear stress, the uctuation of wall normal stress, and the mass transfer near the wall, have signicant effects on the carbon dioxide corrosion process. It is difcult to form dense, compact, and protective corrosion product lm in the corrosion process, which is dominated by general corrosion, and can develop into pitting and mesa attack due to localized corrosion. An empirical correlation is suggested to predict slug ow enhanced carbon dioxide corrosion. It is found that the mass transfer in corrosion product lm can be neglected and the slug ow enhanced carbon dioxide corrosion is dominantly controlled by mass transfer or by both of mass transfer and electrochemical corrosion reaction. 2008 Elsevier Ltd. All rights reserved.

Article history: Received 2 January 2007 Accepted 3 August 2008 Available online 22 August 2008 Keywords: C. Slug ow C. Mass transfer C. Wall shear stress C. Carbon dioxide corrosion C. Corrosion rate

1. Introduction An important consideration in oil and gas industry is multiphase transport from remote wells for much more economical transport of oil and gas combined. The multiphase transport pipelines are mostly made of carbon steel and low-alloy steel, which are able to meet many of the mechanical, structural, fabrication requirements and may offer considerable capital savings over the more expensive alloys [1,2]. However, the frequently encountered multiphase uids may contain signicant levels of CO2, which in combination with free water can make the pipeline environment seriously corrosive, resulting in the ow enhanced CO2 corrosion, thus the damage to the interior of carbon steel pipeline walls, and the decrease in pipeline lifetime and even possible shut down of the pipeline. Over the last few decades, with the development of the crude oil exploitation technology and long distance multiphase transport technology, the problem of ow enhanced CO2 corrosion has made the economic losses increasingly serious, and it is especially true for the oil and gas multiphase transport in the offshore and deep-sea oil elds. Corrosion is a surface damage process. Therefore, what is going on at the metal surface has a profound effect on the corrosion [3]. Many aspects of uid dynamics related to or determined by the interactions between uid and metal surface are of importance to corrosion. The changes in uid hydrodynamics, turbulence, wall shear stress, mass transfer, electrochemical corrosion, the forma* Corresponding author. Tel.: +86 29 82665185; fax: +86 29 82668703. E-mail address: dfche@mail.xjtu.edu.cn (D. Che). 0010-938X/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.corsci.2008.08.006

tion and the destruction of the corrosion product lm, are all intimately related to the hydrodynamic boundary layer and the diffusion boundary layer in the vicinity of the metal substrate. It has been found that metallurgical characteristics of pipelines and hydrodynamic characteristics of ow are important factors inuencing corrosion rate. There are remarkably different hydrodynamic characteristics for different ow regimes, resulting in different mechanisms of ow enhanced corrosion. Among the gasliquid two-phase ow regimes, the vertical upward slug ow is highlighted potentially as the most aggressive and commonly encountered regime, which always exists in oil and gas pipelines with higher oil and gas production rates, and causes a high corrosion rate of carbon steel [4,5]. The remarkable hydrodynamic characteristic of the slug ow is the ow intermittence, that is, the ow is pseudo-periodically alternated by Taylor elongated bubble with an annular falling liquid lm around it and a portion of succeeding liquid slug, simply described as a sequence of slug unit. The Taylor bubble occupies nearly the whole cross-section of the tube with a bullet-shaped front and a at tail prole, while the annular falling liquid lm is assimilated by the succeeding liquid slug entraining many small bubbles. Although slug ow appears a well-ordered ow, it is highly complicated with an unsteady nature, which is inuenced by many factors, such as the velocities of gas and liquid, void fraction, pressure shock, density wave, liquid slug frequency and the physical properties of gas and liquid. In this paper, using the limiting diffusion current technology, conductivity probe technology and digital high-speed video system, mass-loss method, electrochemical impedance spectroscopy

3006

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

(EIS), linear polarization resistance (LPR), scanning electron microscopy (SEM), and X-ray diffraction (XRD), the vertical upward slug ow enhanced CO2 corrosion is investigated both experimentally and mechanistically. The major characteristics investigated include the shear stress and mass transfer coefcient in the near wall zone, the effect of hydrodynamic characteristics on the corrosion process, CO2 corrosion rate, the morphological and composition analysis of corrosion product lm, the coupling characteristics of hydrodynamics and electrochemical corrosion, etc. 2. Experimental 2.1. Experimental set-up The experiments are conducted in a two-phase ow set-up manufactured from 316L (UNS S31603) stainless steel, which is shown in Fig. 1. The detailed description of the experimental setup has been reported in the previous paper of the authors [6]. 2.2. Hydrodynamic test section The hydrodynamic test section is a vertical 35 mm inside diameter, 5 m long Plexiglas pipe, schematically shown in Fig. 2. The experimental liquid phase is the synthetic electrolyte of 0.5 mol/L NaOH0.01 mol/L K4Fe(CN)60.01 mol/L K3Fe(CN)6 by dissolving analytical-reagents in distilled water, and the experimental gas phase is N2. The detailed description of the hydrodynamic measurement principles has been reported [6,7]. The pressure differential with a distance of 1570 mm is measured by the 1151DP type capacitance differential pressure transmitter. A pressure transmitter is installed downstream to measure the local absolute pressure. Wall mass transfer probe and wall shear stress probe are used to measure the mass transfer coefcient and the shear stress near the wall zone using the limiting diffusion current technology. A couple of single-sensor conductivity needle probes with the same geometrical structure are used to measure the average void fraction, local bubble size, slug length distribution, slug frequency, Taylor bubble velocity, etc. A digital high-speed video system is installed downstream of test section to directly measure two-phase ow characteristics by shooting the photographs of the ow at a high-speed. An armored copperconstantan thermocouple is installed to measure the uid temperature. The temperature and pressure are

maintained constant at 40 C and 0.2 MPa throughout the experiments. 2.3. Electrochemical corrosion test section The electrochemical corrosion test section is a vertical 35 mm inside diameter, 5 m long 316L stainless steel tube, schematically shown in Fig. 3. The experimental liquid phase is the synthetic solution of 0.5 mol/L NaCl by dissolving analytical-reagent in distilled water, and the experimental gas phase is CO2. The temperature and pressure are maintained constant at 68 C and 0.21 MPa throughout the experiments. The pH value of solution is around 4.5 and the conductivity is about 5.1 X1 m1. Corrosion rates are recorded using electrochemical impedance spectroscopy (EIS) and linear polarization resistance (LPR), double-checked with mass-loss method. The pH is monitored throughout the experiments using standard litmus paper. To measure the electrochemical corrosion characteristics in situ using EIS and LPR, four sets of electrochemical corrosion measurement devices with the same three-electrode conguration are ush-mounted to the wall of the pipe to minimize ow disturbances, as shown in Fig. 4. All coupons are machined from the parent tubular section into cylindrical specimens with 10 mm in diameter, 6 mm in thickness. All of the surfaces are coated with epoxy, except leaving a bottom surface exposed to the corrosive solution with an area of 0.785 cm2. In each three-electrode measurement device, the working electrode (WE) is made of API N80 mild steel, the counter electrode (CE) and reference electrode (RE) are all made of 316L stainless steel. The martensitic microstructure of sampled N80 mild steel indicates that this material has been quenched and tempered. The chemical compositions of N80 mild steel and 316L stainless steel are presented in Tables 1 and 2. Before each test run, anaerobic grade CO2 gas is used to deoxygenate the solution, lowering the dissolved oxygen level to less than 30 lg/L from the formation of oxides. Moreover, the dissolved iron level is controlled below 15 mg/L, avoiding the saturated iron ion effect on the corrosion process. During the experiments, the oxygen and iron ion levels are regularly monitored. The coupons are rstly polished progressively with 400-, 800- and 1200- grit silicon carbide abrasive paper, then degreased with acetone and rinsed with alcohol many times, dried, weighed, and nally ush-mounted. At the end of the electrochemical corrosion mea-

Fig. 1. Schematic diagram of experimental set-up: (1) gas storage vessel; (2) gas supplement device; (3) freezing dryer; (4) liquid supplement tank; (5) secondary separator; (6) secondary pump; (7) liquid storage tank; (8) oil bath heater; (9) primary pump; (10) liquid phase orice owmeter; (11) twin-screw compressor; (12) gas surge tank; (13) gas phase orice owmeter; (14) two-phase mixing device; (15) test section; (16) non-return valve; (17) cut off valve.

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

3007

Fig. 2. Schematic diagram of hydrodynamic test section.

surements in situ, these coupons are carefully disassembled from the test section, rinsed, dried, weighed, stored in a glove bag and sealed in helium gas environment, preventing the corrosion product from oxidization and destruction, to be further morphologically analyzed. The electrochemical corrosion measurements are conducted using an EG&G Princeton Applied Research (PAR) Potentiostat/Galvanostat Model 273A and a 5210 lock-in-amplier. The EIS measurements are carried out at the open-circuit potential with amplitude of 10 mV sinusoidal alternating current (AC) potential in the sweep frequency range of 10 mHz to 10 kHz. The acquired spectra are analyzed based on the principle of an equivalent circuit using nonlinear least-squares curve tting algorithms provided by the accompanying software. The LPR measurements are performed by polarizing the working electrode in the range of 20 mV from open-circuit potential with a scan rate of 0.1 mV/s, and the automatic IR drop compensation is used. In calculating corrosion rates from linear polarization measurements, anodic and cathodic Tafel slopes are assumed to be 120 mV/dec. At the end of experiments, the potentiodynamic sweeps for Tafel polarization curves are done at direct current (DC) potential range 500 mV below and 300 mV over the open-circuit potential with a scan rate of 0.1 mV/s. 3. Results and discussion 3.1. Hydrodynamic characteristics There exist four processes in the ow enhanced corrosion as follows [3,8,9]:

1. The formation of reactants in the bulk solution. 2. The reactants transport from the bulk solution to the metal surface. 3. Electrochemical corrosion reaction at the metal/solution interface. 4. The corrosion product transport from the metal/solution interface to the bulk solution. Since electrochemical corrosion occurs on the metal surface, the hydrodynamic boundary layer and diffusion boundary layer exert a pivotal effect on the ow enhanced corrosion [3,10]. Being decisive for the ow enhanced corrosion, the extremely complicated boundary must be simplied here. Starting from the wall, the hydrodynamic boundary layer can be further subdivided into four regions: a diffusion sublayer, a laminar viscous sublayer, an intermediate buffer layer, and an outer logarithmic turbulent layer. In the innermost diffusion sublayer, the mass is dominantly transported by molecular diffusion as a result of concentration gradient. In the laminar viscous sublayer, the velocity and turbulent stresses of the ow are damped to zero, leaves only the viscous stresses of laminar ow to act on the wall; the random movement is the mechanism of individual molecules and the mass is transported by convection diffusion. In this sublayer, the laminar molecular viscous forces dominate over turbulent forces and the ow is essentially laminar. In the outer logarithmic turbulent layer, turbulence plays a signicant role, and mass is transported by turbulent eddy diffusion where the random movement of macro-volumes of solution controls the diffusion. In the intermediate buffer layer, the effects of both molecular viscosity and turbulence are of the equal

3008

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

Fig. 3. Schematic diagram of electrochemical corrosion test section.

Table 2 Chemical composition of 316L stainless steel (mass%) C 0.02 Si 0.80 Mn 1.10 S 0.02 Cr 17.00 Mo 2.00 Ni 14.00 Fe Balance

Fig. 4. Schematic diagram of electrochemical three-electrode measurement system: (1) supporting seat; (2) coupon; (3) seal packing ring; (4) built-in tting; (5) gland nut.

importance, and mass is transported by both convection diffusion and turbulent eddy diffusion. The mass transport in the boundary layer can be measured by mass transfer coefcient k. The viscous energy loss within the turbulent boundary layer, namely the intensity of turbulence in the uid acting on the wall, can be directly measured by wall shear

stress sw. The wall pressure of uid perpendicular to the pipe wall can be measured by wall normal stress rn. Consequently, the hydrodynamic characteristics, especially the transport characteristics in the near wall zone, are of signicant effect on corrosion. A higher mass transfer coefcient can cause an increase in corrosion rate due to the increased transport of both reactants and corrosion products. A higher wall shear stress can lead to higher corrosion rate by preventing the protective corrosion product lm from formation, or by retarding the lm growth, and by removing or destroying existing lm. A higher wall normal stress uctuations can also exacerbate the corrosion rate by facilitating the turbulence formation, making the corrosion product porous. The failure of the pipelines by ow enhanced corrosion usually originates from pitting and mesa attack, and is aggravated by micro-turbulence further created by uid ows. To better understand the slug ow enhanced corrosion, it is necessary to have a deep insight into the transport characteristics in the near wall zone. As shown in Fig. 5, the Taylor bubble is a large axisymmetric bullet-shaped bubble with spherical cap and

Table 1 Chemical composition of N80 carbon steel (mass%) C 0.24 Si 0.22 Mn 1.19 P 0.013 S 0.004 Cr 0.036 Mo 0.021 Ni 0.028 Nb 0.006 V 0.017 Ti 0.011 Cu 0.01 Fe Balance

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

3009

Acquisition voltage U (V)

5 4 3 2 40 20 0 -20 -40 16 14 12 10 8 6 210500 210400

USG=0.45 m/s USL=0.344 m/s USG=0.45 m/s USL=0.344 m/s USG=0.45 m/s USL=0.344 m/s

n (Pa)

kferro-cyanide /10 (m/s)

-4

w /10 (Pa)

210300 210200 210100 0.0 0.2 0.4

USG=0.45 m/s USL=0.344 m/s

0.6 Time (s)

0.8

1.0

1.2

Fig. 6. Experimental results of wall shear stress and mass transfer coefcient.

Fig. 5. Physical description of a slug unit.

nearly at bottom occupying almost the whole cross-section of the tube. Due to the buoyancy, the Taylor bubble pushes aside the liquid phase toward the tube wall in the Taylor bubble nose zone at point A, and a falling liquid lm starts to form at point B. The annular developing falling liquid lm runs downward with an increased velocity all the way around the Taylor bubble, and thins as it falls, until the wall friction force is able to balance the gravitational force of the lm and a terminal constant thickness and velocity of a fully developed falling liquid lm are attained. When the fully developed annular falling liquid lm ows over the Taylor bubble into the wake, the point C, where the annular falling liquid lm plunges into the succeeding liquid slug, encountering the upward bulk liquid ow, and its velocity is diminishing until the falling liquid lm is fully assimilated by liquid slug at D. The uid ows upward in the zone from the point D to the point A0 at which the succeeding Taylor bubble comes. Fig. 6 shows the instantaneous wall shear stress, wall normal stress, and mass transfer coefcient proles in slug units at supercial velocities USG = 0.45 m/s, USL = 0.344 m/s. The acquisition voltage U is obtained using the conductivity probe, the wall shear

stress sw is gained using the limiting diffusion current wall shear stress probe, the wall normal stress rn is achieved using pressure transmitter, and the mass transfer coefcient k is acquired using the limiting diffusion current mass transfer probe. The limiting diffusion current technology is based on the redox principle of potassium ferri- and ferro-cyanide couple. The high level of acquisition voltage U represents the Taylor bubble, the corresponding duration time is attributed to the length of Taylor bubble; the low level of acquisition voltage U represents the liquid slug, the duration time corresponds to the length of liquid slug. Whereas the high levels of wall shear stress, wall normal stress, and mass transfer are associated with falling liquid lm, the low levels are associated with the liquid slug, and the skew lines are associated with the transition from one zone to another. By comparing the proles of acquisition voltage, wall shear stress, wall normal stress, and mass transfer coefcient, the effects of the Taylor bubble nose zone, the Taylor bubble wake zone and the penetration length of the falling liquid lm in the liquid slug on the wall shear stress and the mass transfer coefcient can also be obtained. As shown in Fig. 6, although the transitions of the wall shear stress and the mass transfer coefcient from the low level to the high level are almost simultaneous, the transitions lag behind the acquisition voltage, resulting from the existence of LA in the Taylor bubble nose zone; and the transitions of the wall shear stress and the mass transfer coefcient from the high level to the low level are also almost simultaneous, the transitions also lag behind the acquisition voltage, resulting from the existence of LC in the Taylor bubble wake zone, as shown in Fig. 5. From the above analysis of hydrodynamic structure for the slug unit, and from the validation of experimental results, the mass transfer for the slug unit can be classied as four parts as follows: 1. Mass transfer in the Taylor bubble nose zone from A to B with coefcient kTBn. 2. Mass transfer in the falling liquid lm zone from B to C with coefcient kTBf.

3010

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

3. Mass transfer in the Taylor bubble wake zone from C to E with coefcient kTBw. 4. Mass transfer in the remaining liquid slug zone from E to A0 with coefcient kLS. Designating the downward wall shear stress positive, the wall shear stress in the slug unit can be divided into two parts as follows: 1. Positive wall shear stress swf from B to D due to the falling liquid lm, and 2. negative wall shear stress swLS from D to B0 due to the liquid slug. Experimental results have been found that both the mass transfer coefcient and wall shear stress in each zone increase with the increase of supercial gas velocity, and decrease with the increase of supercial liquid velocity. For the falling liquid lm zone, as the supercial gas velocity is increased, the falling liquid lm terminal velocity is increased and the falling liquid lm thickness is reduced, which results in the increases of the velocity gradient and the concentration gradient, thus increased mass transfer coefcient and wall shear stress. Whereas as the supercial liquid velocity is increased, the falling liquid lm terminal velocity is decreased and the falling liquid lm thickness is increased, which results in the decreases of the velocity gradient and the concentration gradient. For the Taylor bubble wake zone, increasing the supercial gas velocity leads to an increased falling liquid lm terminal velocity, resulting in more turbulent, disorderly and chaotic wake, and the mass transfer coefcient is increased; the increased supercial liquid velocity leads to a decreased mass transfer coefcient. For the liquid slug zone and the Taylor bubble nose zone, the increase of the supercial gas velocity gives rise to the increase of void fraction, resulting in increased probability of bubble impact on the wall and a larger turbulence of liquid slug, thus increased mass transfer coefcient and wall shear stress; whereas a contrary change is enhanced by increasing the supercial liquid velocity. It can also been found that the mass transfer coefcient kTBf is maximum, the mass transfer coefcient kTBn is minimum; the magnitude of the wall shear stress associated swf with the falling liquid lm is always larger than that of swLS associated with the liquid slug. Due to the complicated hydrodynamic characteristics of slug ow, there is still no widely applied normalized formula valid to accurate determination of the mass transfer coefcient and wall shear stress. Mass transfer prediction is of great importance in computing corrosion rates, particularly for the transport based models, and is also the key for the coupling of CO2 corrosion and hydrodynamics of slug ow. The global formula for mass transfer prediction in fully developed single phase liquid full-pipe turbulent ow can be expressed as

kTP 0:023CRe0:8 Sc0:33  0:5 Frs C 0:27 1 b

        Dd qUL 0:8 l 0:33 Dd 0:023C L l qDd L 3 4

where C is the corrected factor considering the difference of gasliquid slug ow from the single phase ow; kTP the mass transfer coefcient (m/s) in gasliquid two-phase slug ow; Dd the diffusion coefcient (m2/s), for the hydrogen ion (H+) Dd = 9.31 109 m2/s, for the ferrous ion (Fe2+) Dd = 7.12 1010 m2/s; L the characteristic length (m); U the characteristic velocity (m/s); l the characteristic dynamic viscosity (Pa s); q the characteristic density (kg/m3); Frs, (USL + USG)/(gD)0.5 the Froude number for the mixture velocity of supercial gas phase velocity USG and supercial liquid velocity USL; and b the Taylor bubble length fraction, ratio of Taylor bubble length LTB to slug unit length LSU = LTB + LLS. The experimental results show the falling liquid formation distance from the Taylor bubble nose LA is short, and the corresponding mass transfer coefcient kTBn is minimum. Therefore, kTBn can be neglected. The proposed expressions for calculating other mass transfer coefcient are shown in Table 3. Good agreement can be found with maximum derivation within 10%. In Table 3, qL and lL are the physical properties for liquid phase, qG and lG the physical properties for gas phase, D the diameter of the pipe, haTBi the average Taylor bubble void fraction, haLSi the average liquid slug void fraction, hUaxiali the average axial velocity in the Taylor bubble wake zone, the calculation can be referred to [6,7], g the gravitational acceleration. The average mass transfer coefcient in a slug unit can be calculated as follows:

ki fi

where ki is the mass transfer coefcient kTBf, kTBw, and kLS in each part of a slug unit, fi the weight factor based on the length fraction of each part, LTB/LSU, LB/LSU, (LLS LB)/LSU, respectively. Fig. 7 shows the average mass transfer coefcient of H+ for vertical upward gasliquid slug ow at different supercial gas and liquid velocities. The wall shear stress can be calculated as follows:

sw

f qU 2 2
0.2

Sh a Reb Scc

where f = 0.046Re is wall friction factor, Re = qLU/l, U the characteristic velocity (m/s), L the characteristic length (m), q the characteristic density (kg/m3), l the dynamic viscosity (Pa s). The proposed expressions for calculating the wall shear stress swf associated with falling liquid lm and the wall shear stress swLS associated with liquid slug are listed in Table 4. Good agreement can be found with maximum derivation within 10%. In Table 4, UGLS and ULLS are the gas bubble velocity and liquid velocity in liquid slug, respectively. Leaving out the directions of wall shear stress, the average wall shear stress in a slug unit can be calculated as follows:

where a = 0.023, b = 0.8 and c = 0.33 for well-known fully developed single phase liquid full-pipe turbulent ow correlation developed by Chilton and Colburn [11]. Sh is the Sherwood number kL/Dd, Re the Reynolds number qUL/l, and Sc the Schmidt number l/ (qDd). Based on the single phase mass transfer coefcient formula, a corrected normalized mass transfer coefcient formula for gasliquid vertical upward slug ow is suggested as follows:

sw jswf j b jswLS j 1 b
Table 3 The proposed expressions kTBf U (m/s) L (m) q (kg/m3) 9.916[gD(1 haTBi0.5)]0.5 4 0.5D(1 haTBi0.5) kTBw hUaxiali D qL(1 haLSi) + qGhaLSi kLS

qL

Sh C 0:023Re0:8 Sc0:33
namely

l (kg/(ms))

lL

lL(1 haLSi) + lGhaLSi

USG + USL D qL(1 haLSi) + qGhaLSi lL(1 haLSi) + lGhaLSi

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

60
700

3011

USG=0.15 m/s
55 50 45

USG=0.25 m/s USG=0.35 m/s USG=0.45 m/s

Maximum n (Pa)

600

500

kH /10 (m/s)

40 35 30

400

-4

300

USG=0.15 m/s USG=0.25 m/s USG=0.35 m/s USG=0.45 m/s

200

25 20 0.30
100 0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

USL (m/s)
Fig. 7. Average mass transfer coefcient of H+ for vertical upward gasliquid slug ow.

USL (m/s)
Fig. 9. Maximum wall normal stress uctuation values for vertical upward gas liquid slug ow.

Table 4 The proposed expressions for wall shear stress

swf
U (m/s) L (m) q (kg/m3) l (kg/(m s)) 9.916[gD(1 (haTBi) )] 4 0.5D[1 (haTBi)0.5]
0.5 0.5

swLS
UGLS + ULLS D qL(1 haLSi) + qGhaLSi lL(1 haLSi) + lGhaLSi

qL lL

Fig. 8 shows the average wall shear stress for vertical upward gasliquid slug ow at different supercial gas and liquid velocities. The wall normal stress rn, which represents the wall pressure of uid perpendicular to the pipe wall, is constant with small uctuations when the only-liquid ows in a straight smooth pipeline. In such case, wall normal stress has a weak effect on the corrosion. However, in the gasliquid two-phase slug ow, the larger wall pressure uctuations are caused by the pseudo-periodical intermittence of Taylor bubble and liquid slug, exerting a large additional mechanical force onto the metal surface. Thereby, the larger wall pressure uctuation is a very important parameter considered in the slug ow enhanced corrosion. Fig. 9 shows the maximum wall normal stress uctuation values for vertical upward

gasliquid slug ow. It can be seen that the maximum uctuations are insensitive to the different supercial gas and liquid velocities. From the above analysis, it can be seen that the intermittence of Taylor bubble and liquid slug is an essential characteristic. As shown in Fig. 6, when a Taylor bubble and the succeeding liquid slug sequentially pass through a given cross-section of the tube, the direction of wall shear stress changes twice, which also causes larger uctuations of both mass transfer coefcient and wall normal stress. The intermittence characteristic is one of the important factors for the ow enhanced corrosion. The intermittence of slug ow can be characterized by slug frequency FSU,

F SU

NSU Dt

where NSU is the number of slug units at time interval Dt. Fig. 10 shows the slug frequency distribution at different supercial gas and liquid velocities. It is indicated that the slug frequency is generally above 2 Hz, and the slug frequency increases with increased supercial liquid velocity, decreases with increased supercial gas velocity. 3.2. CO2 corrosion characteristics Fig. 11 shows the corrosion rates measured by LPR for CO2 corrosion of N80 steel at different supercial gas and liquid velocities.
6

220

USG=0.15 m/s
200 180 160

USG=0.15m/s USG=0.25m/s
5

USG=0.25 m/s USG=0.35 m/s USG=0.45 m/s

USG=0.35m/s USG=0.45m/s

FSU (Hz)

140

w (Pa)

120 100 80 60

2
40 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70

0.30

0.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

USL (m/s)
Fig. 8. Average wall shear stress for vertical upward gasliquid slug ow.

USL (m/s)
Fig. 10. Slug frequency distribution at different supercial gas and liquid velocities.

3012
30

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

25

USG=0.15 m/s USG=0.35 m/s

25

USG=0.25 m/s USG=0.45 m/s

20

USG=0.15 m/s USG=0.35 m/s USL=0.446 m/s

USG=0.25 m/s USG=0.45 m/s

USL=0.344 m/s

CR (mm/y)

CR (mm/y)
0 20 40 60 80 100 120 140

20

15

15

10 10

5 0 20 40 60 80 100 120 140

Time (h)
20 15

Time (h)

USG=0.15 m/s USG=0.35 m/s USL=0.518 m/s

15

USG=0.25 m/s USG=0.45 m/s

USG=0.15 m/s USG=0.35 m/s USL=0.651 m/s

USG=0.25 m/s USG=0.45 m/s

CR (mm/y)

CR (mm/y)
10 5 0 20 40 60 80 100 120 140

10

5 0 20 40 60 80 100 120 140

Time (h)

Time (h)

Fig. 11. Corrosion rates measured by LPR for CO2 corrosion of N80 steel at different supercial gas and liquid velocities.

The corrosion rates keep constant within the 30-h exposure to corrosion environment in all experiments due to no corrosion product lm. Nevertheless, the corrosion rates remarkably decrease due to the formation of the corrosion product lm for 5085-h exposures. For instance, under the condition of USG = 0.45 m/s, USL = 0.344 m/s, the corrosion rate changes from 19.7 mm/yr at the beginning of the corrosion experiment to the smallest value of 16.11 mm/yr after 75-h exposure. Further exposed to corrosion, under the larger mass transfer and wall shear stress situations, e.g., USG = 0.45 m/s, USL = 0.344 m/s, and USG = 0.45 m/s, USL = 0.518 m/s, the corrosion rates increase to 22 mm/yr and 16.4 mm/yr after 140-h exposure, respectively, higher than that within the initial 30-h exposure, namely 19.7 mm/yr and 15.3 mm/yr, respectively. Whereas under smaller mass transfer and wall shear stress situations, e.g., when USG = 0.15 m/s, USL = 0.446 m/s, and USG = 0.15 m/s, USL = 0.651 m/ s, the corrosion rates decrease continuously from 10.21 mm/yr and 7.9 mm/yr at the initial exposure to 8.51 mm/yr and 5.8 mm/ yr at the end of the experiments after 140-h exposure, respectively. In such situations, corrosion product lm has begun to build on the coupons after 50-h exposure, and the lm steadily grows, making the corrosion rate dramatically decrease. As shown in Fig. 11, it can be seen that the average corrosion rate reaches its maximum when USG = 0.45 m/s, USL = 0.344 m/s, minimum when USG = 0.15 m/s, USL = 0.651 m/s in all experiments. According to the hydrodynamic characteristics analysis, as shown in Figs. 7 and 8, the average mass transfer coefcient and the average wall shear stress are maximum when USG = 0.45 m/s, USL = 0.344 m/s, minimum when USG = 0.15 m/s, USL = 0.651 m/s. It is further validated that the hydrodynamic characteristic plays

a key role in the corrosion process, namely larger mass transfer coefcient and larger wall shear stress lead to a higher corrosion rate. Whereas at smaller mass transfer and smaller wall shear stress, the corrosion rate keeps lower and is prone to form the corrosion product lm on the coupon. Figs. 12 and 13 show the potentiodynamic polarization curves for CO2 corrosion of N80 steel when USL = 0.344 m/s and USL = 0.651 m/s with different supercial gas velocities. It is also can be seen that the polarization currents of N80 steel are dependent on the near wall hydrodynamic characteristics. When USL = 0.344 m/s with USG = 0.15 m/s and USG = 0.45 m/s, except 140-h exposure, the cathodic branches of the polarization curves are similar for 1-h, 50-h and 70-h exposures. The cathodic branches of the curves exhibit limiting current, which is a direct consequence of depletion of H+ ions near the surface and the slow hydration rate of CO2. Right below the corrosion potential, the cathodic reaction process is controlled by activation, and the external power provides extra electrons, making the cathodic reaction accelerated; whereas farther negative to the corrosion potential, due to the adjacent active species depletion, the diffusion transport of electroactive species from the bulk solution to the electrode surface is dominated, and the cathodic reaction is diffusion controlled with the existence of explicit limiting current. However, for the anodic branch of polarization curve, the current density increases with the increase in anodic polarization potential, and unlimited supply of iron atoms is available at the interface. The anodic reaction rates do not depend on the mass transfer process. Except 75-h exposure, the anodic branches of polarization curves are similar for 1-h, 50-h and 140-h exposures. It is found that the polarization

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

-300 -300

3013

1h 50 h -400 USG=0.15 m/s Potential (mV) vs 316LSS

-500 -600 -700 -800 -900 -1000 -5.5

75 h 140 h -400 USG=0.15 m/s Potential (mV) vs 316LSS

-500 -600 -700 -800 -900 -1000 -6.0

USL=0.344 m/s

USL=0.344 m/s

-5.0

-4.5

-4.0

-3.5
2

-3.0

-2.5

-5.5

-5.0

-4.5

-4.0
2

-3.5

-3.0

-2.5

Log Current density (A/cm )

-300 -400 -500 -600 -700 -800 -900 -1000 -1100 -5.5

Log Current density (A/cm ) 75 h 140 h -400 USG=0.45 m/s USL=0.344 m/s
-300 -500 -600 -700 -800 -900 -1000 -1100 -6.0

1h 50 h USG=0.45 m/s Potential (mV) vs 316LSS

-4.5 -4.0 -3.5
2

Potential (mV) vs 316LSS

USL=0.344 m/s

-5.0

-3.0

-2.5

-5.5

-5.0

-4.5

-4.0

-3.5
2

-3.0

-2.5

-2.0

Log Current density (A/cm )

Log Current density (A/cm )

Fig. 12. Potentiodynamic polarization curves for CO2 corrosion of N80 steel at USL = 0.344 m/s.

curves gradually change from general corrosion to pitting corrosion. For 1-h and 50-h exposures, the polarization curves of N80 show the typical shape of general corrosion; whereas for 75-h exposure, the curves show the pitting type, and the small passive current densities should be attributed to the corrosion product lm, although the lm is not-fully covered and not compact; for 140-h exposure, the corrosion product lm is no longer resistant to hydrodynamic characteristics and detaches from the surface. Table 5 shows the shifting trend of corrosion potential and the corresponding corrosion current density when USL = 0.344 m/s with USG = 0.15 m/s and USG = 0.45 m/s. Since the corrosion potential increases in the anodic direction with the formation of corrosion product lm, the corrosion product lm may be considered to have anodic anticorrosive capability. The polarization characteristics when USL = 0.651 m/s with USG = 0.15 m/s and USG = 0.45 m/s are similar to those when USL = 0.344 m/s. However, due to smaller mass transfer and smaller wall shear stress, a corrosion product lm is formed on the coupon when USG = 0.15 m/s, USL = 0.651 m/s for 50-h, 75-h and 140-h exposures, and the tendency of pitting corrosion is greatest. Figs. 14 and 15 show the typical Nyquist and Bode plots of EIS spectra for CO2 corrosion of N80 steel at USL = 0.344 m/s and USL = 0.651 m/s with different supercial gas velocities. When USG = 0.15 m/s, USL = 0.344 m/s, two semicircles appear after 75-h exposure, which means a corrosion product lm has been formed on the coupon. The rst semicircle is so small that it merges into the second semicircle, cannot clearly distinguished from each other, namely Rf is much smaller than the charge transfer resistance Rct. The rst semicircle at higher frequency would be due

to the corrosion product lm because a surface dielectric lm normally has a small time constant and so has a phase angle shift in the high frequency range. The measured results validate the conclusion with Rf = 6.3 X cm2, Cf = 1018 lF/cm2, the value of Rf being one seventh of the Rct. This indicates that the corrosion product lm has non-protective capacity, and further validates the results of corrosion process measured by LPR and potentiodynamic polarization. As shown in Table 6, the charged transfer resistance increases from the 1-h exposure to 50-h exposure, and corrosion product lm starts to form on the coupon. After 75-h exposure, the charged transfer resistance increases to the maximum with corrosion product lm covered; however, the corrosion product lm is not-fully covered and non-compact. The difference for the physical property of corrosion product lm from that of the metal substrate, leads to high tendency of pitting corrosion. In such a case, the corrosion product lm is detached from the metal substrate due to the effect of the hydrodynamics of slug ow. Thereby, the charge transfer resistance Rct decreases to the minimum after 140-h exposure. From the analysis of EIS spectra, it is found that the type of corrosion and the kinetics of the corrosion reaction when USG = 0.45 m/s, USL = 0.344 m/s, and USG = 0.45 m/s, USL = 0.651 m/ s are similar to those when USG = 0.15 m/s, USL = 0.344 m/s. The impedances when USG = 0.15 m/s, USL = 0.651 m/s, are shown in Table 6. Since corrosion product lm starts to build on the coupons at the beginning of the exposure, the angle shifts in higher frequency are very marked. The corrosion product lm resistance Rf is very small, no more than one fth of the charge transfer resistance Rct, which indicates that the corrosion product

3014

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

-200 -400 -300 -400

Potential (mV) vs 316LSS

-5.0 -4.5 -4.0 -3.5
2

-600

Potential (mV) vs 316LSS

-500 -600 -700 -800

-800

-1000

-1200

1h 50 h -1400 USG=0.15 m/s

-1600 -6.0

USL=0.651 m/s
-5.5 -3.0 -2.5 -2.0

75 h 140 h USG=0.15 m/s -1000 USL=0.651 m/s

-900 -1100 -6.5 -6.0 -5.5 -5.0 -4.5 -4.0 -3.5
2

-3.0

-2.5

-2.0

Log Current density (A/cm )

-300 -400 -500 -600 -700 -800 -900 -1000 -200 -300 -400 -500 -600 -700

Log Curent density (A/cm )

Potential (mV) vs 316LSS

1h 50 h USG=0.45 m/s USL=0.651 m/s

-5.5 -5.0 -4.5 -4.0 -3.5
2

Potential (mV) vs 316LSS

95 h 140 h -900 USG=0.45 m/s

-800 -1000 -6.5

USL=0.651 m/s
-6.0 -5.5 -5.0 -4.5 -4.0 -3.5 -3.0
2

-6.0

-3.0

-2.5

-2.0

-2.5

-2.0

-1.5

-1.0

Log Current density (A/cm )

Log Current density (A/cm )

Fig. 13. Potentiodynamic polarization curves for CO2 corrosion of N80 steel at USL = 0.651 m/s.

Table 5 Corrosion potential vs corrosion current density USG = 0.15 m/s, USL = 0.344 m/s Corrosion potential (mV) (vs 316L SS) 1h 50 h 75 h 140 h 600 580 540 560 Log corrosion current density log i (A/cm2) 3.018 3.14 3.27 2.97 USG = 0.45 m/s, USL = 0.344 m/s Corrosion potential (mV) (vs 316L SS) 630 615 560 581 Log corrosion current density log i (A/cm2) 2.77 2.78 2.86 2.72

lm is very porous, its anticorrosion capacity is very poor, despite the corrosion product lm formation on the coupons at the initial exposure. Results of the polarization measurement and impedance measurement show that the corrosion lm has limited protective capacity in the vertical upward gasliquid two-phase slug ow enhanced CO2 corrosion due to the remarkable effect of the peculiar hydrodynamics. It is still difcult for a fully covered, compact, and anticorrosive corrosion product lm to form due to the effect of hydrodynamics of slug ow. The hydrodynamic characteristics of vertical upward gasliquid two-phase slug ow govern the corrosion mechanism, and are intimately related to the overall corrosion process. Mass transfer characteristic has an adverse effect on the corrosion rate, whereas the wall shear stress and wall normal stress are responsible for the nal morphological consequence. Under the peculiar hydrodynamics of pseudo-periodical intermittence of Taylor bub-

ble and liquid slug, the interactions of both the upward and downward wall shear stresses, the uctuation of wall normal stress make the nal morphological consequence of the corrosion extraordinarily serious. Fig. 16 shows the resultant stress and included angle for wall shear stress and the uctuation of wall normal stress. The high level of acquisition voltage U represents the Taylor bubble; the low level of acquisition voltage U represents the liquid slug. When a Taylor bubble passing by a given cross-section of the pipe, the positive wall shear stress swf changes from 0 Pa to the maximum (swf)max, and the uctuation of the wall normal stress changes from the 0 Pa to the maximum (Drn)max as well. Consequently, the resultant stress of the wall shear stress and the uctuation of wall normal stress reaches its maximum, and the included angle between the resultant stress and the wall shear stress increases from 0 to the maximum (+90) as well. When a succeeding liquid slug passes, the direction of the wall shear stress is changed upward, namely the negative wall shear stress swLS changes from 0 Pa to the maximum (swLS)max, and the uctuation of the wall normal stress is changed to 0 Pa again from the maximum (Drn)max. As such, the resultant stress of the wall shear stress and the uctuation of wall normal stress changes to the minimum, and the included angle between the resultant stress and the wall shear stress decreases from +90 to the 0 as well. Due to the existence of transition from the leading liquid slug to the succeeding Taylor bubble, the direction of the wall shear stress changes with stress starting to increase, and the uctuation of wall normal stress increases. as such, the included angle rstly changes to 90, and then dramatically changes to 0, further to +90.

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

-40 100 -60

3015

-30

1h 50 h 75 h 140 h USG=0.15 m/s |Z| (Ohm-cm )

2

1h 50 h 75 h 80 140 h USG=0.15 m/s

60

-50

Im(Z) (Ohm-cm )

-20

-30 40 -20 20

1.8Hz 5.2Hz
-10

5.3Hz
-10

0 0 10 20 30 40
2

0 50 60 -2 -1 0 1 2 3 4

Re(Z) (Ohm-cm )
-20 50

Log Freq (Hz)

-50

-15

1h 50 h 75 h 140 h USG=0.45 m/s |Z| (Ohm-cm )

2

40

1h 50 h 75 h 140 h USG=0.45 m/s USL=0.344 m/s

-40

Im(Z) (Ohm-cm )

30

-30

-10

5.2Hz 6.1Hz 6.2Hz 8.8Hz

20

-20

-5

10

-10

0 0 5 10 15 20
2

0 25 30 -2 -1 0 1 2 3 4

Re(Z) (Ohm-cm )

Log Freq (Hz)

Fig. 14. Nyquist and Bode plots of EIS spectra for CO2 corrosion of N80 steel at USL = 0.344 m/s.

Both experimental results and previous studies by [12,13] show that the main components of the corrosion product lm are iron carbonate FeCO3 and iron carbide Fe3C. The N80 steel is corroded away as Fe2+, which can be precipitated out of the solution on the coupon substrate as FeCO3 when the concentration of which exceeds the super-saturation level in the solution. Fe3C, as the skeleton of the metal in a classical sense, is the remains on the parent metal substrate after the corrosion. Corrosion product lm is brittle, compared to the metal substrate, which has plastic physical property. The brittle material is sensitive for the higher uid impingement angle, such as 7590, and the plastic material is sensitive to the smaller uid impingement angle, for example, 2045. Due to the higher resultant stress with included angle changing from 90 to the +90 accompanied with the frequency more than 2 Hz, the hydrodynamics make the bonding between the metal matrix and corrosion lm diminish, until the corrosion lm is destroyed and peeled off, leading to the exposure of partial or entire metal to the corrosive reactant and occurring localized corrosion or general corrosion. Fig. 17 shows the corrosion product lm thickness on N80 steel coupons vs different exposure time. When USG = 0.45 m/s, USL = 0.344 m/s, and USG = 0.15 m/s, USL = 0.344 m/s, and USG = 0.45 m/s, USL = 0.651 m/s, the thickness of corrosion product lm rstly increases with the increased exposure time, reaching maximum after 75-h exposure; and then decreases with increased exposure time. This can be elucidated as follows: the corrosion product lm starts to form on the coupons, although it is not-fully covered and non-compact; however, the localized corrosion is initiated due to

higher wall shear stress and wall normal stress; with the propagation of the localized corrosion, the micro-turbulent intensity increases dramatically, which sometimes may be ten orders of magnitude higher than that in the smooth pipe; then the localized corrosion further aggravates, resulting in higher corrosion rate, more damage to the corrosion product lm, and more pitting. When USG = 0.15 m/s, USL = 0.651 m/s, due to the smaller wall shear stress and mass transfer of slug ow, there exists a favorable condition for FeCO3 precipitation. A corrosion product lm starts to form on the coupons, builds-up, and further strengthens. This makes a thicker corrosion product lm, and decreases the corrosion rate. Fig. 18 shows the average area fraction covered by corrosion product lm on N80 coupons. It can be seen that the maximum area fraction covered by corrosion product lm is no more than 60% due to the effect of hydrodynamics of slug ow. Fig. 19 shows the area fraction vs corrosion depth for N80 steel coupons after 140-h exposure. It can be seen that the corrosion is dominated by general corrosion with small pitting due to higher corrosion depth with smaller area fraction in each experimental conditions. Under higher wall shear stress and mass transfer of slug ow, there is higher corrosion rate, as well as more numerous and deeper pittings. 3.3. Coupling characteristics of hydrodynamics and corrosion It is essential to develop a mechanistic model based on the simultaneous and interaction of corrosion characteristics and

Phase (degree)

USL=0.344 m/s

Phase (degree)

1.1Hz

USL=0.344 m/s

USL=0.344 m/s

-40

3016
-50

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

100 -60

-40

1h 50 h 75 h 140 h USG=0.15 m/s |Z| (Ohm-cm ) USL=0.651 m/s 2.4Hz 8.8Hz

2

1h 50 h 75 h 80 140 h USG=0.15 m/s

60

-50

Im(Z) (Ohm-cm )

-30 40 -20 20

-20

3.1Hz
-10

-10

0 0 10 20 30 40
2

0 50 60 -2 -1 0 1 2 3 4

Re(Z) (Ohm-cm )
-40

Log Freq (Hz) 1h 50 h 95 h 140 h USG=0.45 m/s |Z| (Ohm-cm ) USL=0.651 m/s
2

80

-60

-30

1h 50 h 95 h 140 h 60 USG=0.45 m/s USL=0.651 m/s

40

-50

Im(Z) (Ohm-cm )

-40

-20

-30

2.4Hz
-10

-20 20 -10

3.1Hz

0 0 10 20 30
2

0 40 -2 -1 0 1 2 3 4

Re(Z) (Ohm-cm )

Log Freq (Hz)

Fig. 15. Nyquist and Bode plots of EIS spectra for CO2 corrosion of N80 steel at USL = 0.651 m/s.

Table 6 Equivalent circuit model analysis USG = 0.15 m/s, USL = 0.344 m/s Rct (X cm2) 1h 50 h 75 h 140 h 28.61 37.77 45.12 26.08 USG = 0.15 m/s, USL = 0.651 m/s Cdl (lF/cm2) 1070 2342 3208 1150 Rct (X cm2) 35.11 36.27 40.12 49.38

Hmax ()
49 51.8 54.2 46.6

Hmax ()
46.04 56.18 57.72 49.16

Cdl (lF/cm2) 1463 1418 451 1345

Rf (X cm2) 2.12 5.39 6.13 6.63

Cf (lF/cm2) 630 417 476 572

Note: Rs = 4.1 (X cm2).

hydrodynamic characteristics, such as the rate equations of both metal anodic dissolution and oxidants cathodic reductions, transport equations of reactants and products. Nevertheless, an effective mechanistic model incorporating seamlessly all of these factors is not available currently. When the corrosion rate is electrochemical reaction controlled, mass transfer characteristic has no effect on the corrosion. However, majority of elds and laboratory investigations have shown that the corrosion are both controlled in gasliquid two-phase slug ow enhanced CO2 corrosion, namely the mass transfer mechanism and the electrochemical reaction mechanism are of equal importance in determining the corrosion rate. In such a situation, the ow enhanced corrosion is not merely the simple summation of the electrochemical corrosion caused by the corrosive reactants and the mechanical losses caused by the gasliquid two-phase slug ow, the two causes are synergistical, namely there is a 1 + 1 > 2 effect. There is a strong nonlinear coupling of three factors so that

the contribution of individual cause to corrosion cannot be distinguished easily. These three factors are the wall shear stress caused by the turbulence in the hydrodynamic boundary layer near the wall of the pipelines, the mass transfer caused by the turbulence diffusion in the diffusion boundary layer near the wall of the pipelines, and the electrochemical corrosion due to the corrosive reactants in the gasliquid two-phase ow. Based on the analysis for the hydrodynamic boundary layer and diffusion boundary layer in Section 3.1, the charge transport ux through the boundary layer will be considered according to the Ficks rst law as follows:

i Dd nF

dC dx

where Dd is the diffusion coefcient of the ionic species (m2/s), n its valence, F the Faraday constant (96,500 C), and dC/dx the concentration gradient of the ion of interest.

Phase (degree)

Phase (degree)

-30

USL=0.651 m/s

-40

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

5
60

3017

Acquisition voltage U (V)

USG=0.45 m/s USL=0.344 m/s

4
50

USG=0.15 m/s USL=0.344 m/s USG=0.45 m/s USL=0.651 m/s USG=0.15 m/s USL=0.651 m/s

Area fraction covered by film (%)

40

2 500 400 300 200 100 0 90 60 30 0 -30 -60 -90 0.0 0.2 0.4

USG=0.45 m/s USL=0.344 m/s

30

Resultant stress (Pa)

20

10

USG=0.45 m/s USL=0.344 m/s

0 0 20 40 60 80 100 120 140 160

Exposure time (h)

Fig. 18. Area fraction covered by corrosion product lm on N80 steel coupons.

Included angle (degree)

USG=0.45 m/s USL=0.344 m/s

0.6 0.8 1.0 1.2

25

USG=0.15 m/s USL=0.344 m/s USG=0.25 m/s USL=0.344 m/s

20

Time (s)
Fig. 16. Resultant stress and included angle for wall shear stress and the uctuation of wall normal stress.

Area fraction (%)

15

10

12

USG=0.45 m/s USL=0.344 m/s USG=0.15m/s USL=0.344 m/s USG=0.45 m/s USL=0.651 m/s USG=0.15 m/s USL=0.651 m/s

10

Film thickness ( m)

0 0.10

0.12

0.14

0.16

0.18

0.20

0.22

0.24

0.26

Corrosion depth (mm)

Fig. 19. Area fraction of corrosion depth for N80 steel coupons after 140-h exposure.

i DH nF
0 0 20 40 60 80 100 120 140 160

Cb Cm kTP nF C b C m dd

10

Exposure time (h)

Fig. 17. Corrosion product lm thickness on N80 steel coupons against different exposure time.

As shown in Fig. 20, there are two scenarios for the coupling of hydrodynamic characteristics and electrochemical characteristics as follows: 1. Coupling without corrosion product lm, and 2. coupling with corrosion product lm. The reduction of H+ is the main culprit for the CO2 corrosion. Hence, only the H+ reduction is of interest in the coupling model for convenience. On the naked metal substrate surface, for Scenario 1 shown in Fig. 20, according to the Nernsts linear concentration diffusion gradient model, the charge transport ux of H+ can be expressed as

where kTP is the H+ mass transfer coefcient (m/s); Cb = 10pH, the bulk concentration, (mol/L); Cm the concentration at the metal/solution interface (mol/L); n = 1 for H+, valence; dd the diffusion boundary layer thickness (m), which differs from the thickness of hydrodynamic boundary layer. The Fe2+ mass transfer has a great effect on the formation of the corrosion product lm on the metal substrate surface. The corrosion product lm will be not only the diffusion barrier between the metal substrate and the corrosive medium, but also give rise to a concentration gradient of the electrochemical reactants. When super-saturation concentrations of Fe2+ and CO2 3 next to the metal substrate surface are achieved, a corrosion product lm forms, as shown in Scenario 2 of Fig. 20. When the corrosion product lm starts to build-up on the metal surface, the corrosion kinetics is changed, the additional mass transport through the corrosion lm must be also considered. The charge transport ux of H+ can be expressed as

i kTP nF C b C f
followed by

11

3018

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

Cb
H 2 CO3
HCO3

Cf

2CO3

H+
OH -

Cm

Fig. 20. Schematics of mass transfer between bulk solution and metal surface.

i kf nF C f C m

12

where Cf is the H+ concentration on the corrosion product lm/solution interface (mol/L), Cm the H+ concentration on the metal/solution interface (mol/L), kf the mass transfer coefcient in corrosion product lm (m/s). In Eqs. (10) and (12), when the cathodic current is limited by the transport of H+ to the surface, then the concentration at the surface is zero, Cm = 0, namely the electrochemical reactions are controlled by the mass transfer of H+. This is valid solely for mass transfer-controlled corrosion. There has been no advanced method to accurately determine the H+ concentration Cm at the metal/solution interface so that it is hard to take into account the coupling of hydrodynamics in the practical corrosion process, especially when the corrosion is controlled by electrochemical reaction or by both of mass transfer and electrochemical reaction. Double-checked by mass-loss method, based on the above analysis on Scenarios 1 and 2, as shown in Fig. 20, and based on the formula for the charge transport ux of H+ controlled by the mass transfer, an empirical correlation for the practical charge transfer

transport ux of H+ is suggested to model slug ow enhanced CO2 corrosion as follows:

i PkTPt nFC b 1 1 1 kTPt kTP kf

13 14

where kTPt is the total mass transfer coefcient of H+; G the empirical coefcient, 0 < G < 1 for the corrosion controlled by electrochemical reaction, G = 1 for the corrosion controlled by mass transfer, and G > 1 for the corrosion controlled by both; Eq. (14) only has a physical rather than mathematical meaning in Scenario 1 without corrosion product lm, namely kTPt = kTP and kf = 0. Fig. 21 shows the total mass transfer coefcient kTPt and the mass transfer coefcient kf of H+. It can be seen that the value of kTPt rstly decreases within the 95-h exposure and then increases after the 95-h exposure when USG = 0.45 m/s, USL = 0.344 m/s, and USG = 0.15 m/s, USL = 0.344 m/s, and USG = 0.45 m/s USG = 0.651 m/ s; the value of kTPt reduces continuously when USG = 0.15 m/s, USG = 0.651 m/s. Nevertheless, the average value of kTPt at different supercial gas and liquid velocities is approximately equal to the

USG=0.45 m/s USL=0.344 m/s

70

USG=0.15 m/s USL=0.344 m/s

USG=0.45 m/s USL=0.651 m/s

2500

USG=0.15 m/s USL=0.651 m/s

60

2000

1500 50

kTPt /10 (m/s)

-4

40

kf /10 (m/s)

1000

-4

300

30 200 20 100

10 0 20 40 60 80 100 120 140 160

0 0 20 40 60 80 100 120 140 160

Exposure time (h)

Exposure time (h)

Fig. 21. Total mass transfer coefcient kTPt and mass transfer coefcient in corrosion product lm kf of H+.

D. Zheng et al. / Corrosion Science 50 (2008) 30053020

3019

value kTP calculated based on the hydrodynamics of slug ow, as shown in Fig. 7. This means that the total mass transfer coefcient kTPt of H+ is dominated by kTP rather than kf. Hausler [14,15] found that the mass transfer coefcient kf in the corrosion product lm is directly proportional to the permeability of the lm and is inversely proportional to the thickness of the lm. From the analysis on the corrosion product lm in Section 3.2, it can been seen that the thickness of the corrosion product lm in the micro-dimension, much thinner than the thickness of hydrodynamic boundary layer; moreover, the maximum area fraction covered by corrosion product lm is no more than 60%. Due to the effect of hydrodynamics of slug ow, the porous, non-compact, and not-fully covered corrosion product lm has poor protection, which can be further validated by the value kf. As shown in Fig. 21, the value of kf is much higher than kTPt and kTP, as high as 15 times kTP when USG = 0.15 m/s, USL = 0.344 m/s, and USG = 0.45 m/s, USL = 0.651 m/s, and USG = 0.15 m/s, USL = 0.651 m/s, even as high as 43 times kTP when USG = 0.45 m/s, USL = 0.344 m/s. This indicates that the mass transfer coefcient kf in the corrosion product lm can be neglected in calculating the corrosion rates. Ikeda et al. [16] found that there is a qualitative change in the corrosion kinetics at temperatures around 60 C. Below 60 C, the corrosion rate increases with increased temperature, accompanied with the formation of the non-protective corrosion product lm iron carbide Fe3C; from 60 C to 90 C, the corrosion rate reaches its maximum with increased temperature, accompanied with the formation of the protective product lm iron carbonate FeCO3; above 90 C, the corrosion rate decreases with increased temperature due to the protective iron oxide Fe2O3, which becomes the main component of the product layer. Therefore, at the experimental temperature of 68 C, the corrosion rate is high, and the corrosion process is controlled mostly by mass transfer or by both, which can also be validated by empirical coefcient G. Fig. 22 shows the empirical coefcient G, it can be seen that it increases with the increased exposure time throughout the experiments, which means that the corrosion process is initially controlled by electrochemical reaction, then by mass transfer, and further by both. At the initial exposure, there is no corrosion product lm on the coupons, and the empirical coefcient is between 0.9 and 1, which further validates the reliability of the model. With increased exposure time, due to the precipitation of corrosion product lm composed of iron carbonate FeCO3 and iron carbide Fe3C, and due to the electronic conductibility of Fe3C, the galvanic corrosion occurs with high potential, leading

to high localized corrosion, further propagating to pitting corrosion and mesa attack; In such conditions, the localized corrosion can bring forth the larger roughness of the wall, making the local turbulent intensity dramatically rise, compared to that in the smooth pipe. Due to the difculty in taking into account the turbulent intensity in the localized corrosion with macro-parameters in a smooth pipe, a large deviation arises, which makes the empirical coefcient G larger than 1, even reach 1.5 when USG = 0.45 m/s, USL = 0.344 m/s. Therefore, it is indicated that the slug ow enhanced CO2 corrosion is dominantly controlled by mass transfer or by both.

4. Conclusions The gasliquid slug ow enhanced carbon dioxide corrosion of API N80 grade steel in vertical upward pipeline has been experimentally investigated. The following conclusions can be drawn: 1. Gasliquid two-phase vertical upward slug ow has peculiar hydrodynamics. When both the Taylor bubble and the succeeding liquid slug of a slug unit sequentially pass through a given cross-section of the tube, the direction of wall shear stress is changed twice, which causes larger uctuations of both mass transfer coefcient and wall normal stress as well. The frequency of the pseudo-periodical alternation by Taylor bubble and succeeding liquid slug is generally above 2 Hz. 2. Experimental results validate that the hydrodynamics of slug ow have signicant effects on the CO2 corrosion. A higher mass transfer coefcient can cause the increase of corrosion rate since the transports of both reactant and corrosion product are increased. The higher wall shear stress can lead to higher corrosion rate by preventing the protective corrosion product lm from formation, or by retarding the lm growth, and by removing or destroying existing lm. The higher wall normal stress uctuations can also exacerbate the corrosion due to the turbulence, making the corrosion product porous. 3. Within the initial 30-h exposure, the corrosion rate keeps constant; within the 5085-h exposure, the corrosion rate decreases due to the formation of corrosion product lm; if the exposure time is further increased, due to the damage of corrosion product lm, accompanied with the galvanic corrosion and pitting corrosion, the corrosion rate is increased, even larger than that at the initial exposure. The hydrodynamic characteristic plays a key role in the corrosion process, larger mass transfer coefcient and/or larger wall shear stress lead to a higher corrosion rate, smaller mass transfer and smaller wall shear stress lead to a lower corrosion rate and is prone to form the corrosion product lm on the coupons. 4. A not-fully covered, non-compact, non-protective corrosion product lm forms on the coupon due to the effect of hydrodynamics of vertical upward slug ow. It is found that the thickness of the corrosion product lm is no more than 12 lm, the area fraction covered by corrosion product lm is no more than 60%. The slug ow enhanced CO2 corrosion is dominated by general corrosion, and the corrosion can further develop into pitting and mesa attack due to localized corrosion. 5. Based on the formula for the corrosion process controlled by mass transfer, an empirical correlation is suggested to model slug ow enhanced CO2 corrosion. Because that there only exists a non-protective corrosion product lm due to the effect of hydrodynamics of slug ow, the mass transfer in corrosion product lm can be neglected. The slug ow enhanced CO2 corrosion is dominantly controlled by mass transfer or by both of mass transfer and electrochemical corrosion reaction.

1.6

USG=0.45 m/s USL=0.344 m/s

1.5 1.4 1.3 1.2 1.1 1.0 0.9 0.8 0 20 40 60 80 100 120 140 160

USG=0.15 m/s USL=0.344 m/s USG=0.45 m/s USL=0.651 m/s USG=0.15 m/s USL=0.651 m/s

Expsoure time (h)

Fig. 22. Values for the empirical coefcient G.

3020

D. Zheng et al. / Corrosion Science 50 (2008) 30053020 [6] D.H. Zheng, D.F. Che, Experimental study on hydrodynamic characteristics of upward gasliquid slug ow, Int. J. Multiphase Flow 32 (2006) 11911218. [7] D.H. Zheng, D.F. Che, An investigation on near wall transport characteristics in an adiabatic upward gasliquid two-phase slug ow, Heat Mass Transfer (2006), doi:10.1007/s00231-006-0193-8. [8] E. Dayalan, G. Vani, et al., Modeling CO2 corrosion of carbon steels, in: Corrosion/1995, NACE International, 1995, Paper No. 118. [9] S. Nesic, L. Lunde, Carbon dioxide corrosion of carbon steel in two-phase ow, Corrosion 50 (1994) 717727. [10] D.T. Wasan, C.L. Tien, C.R. Wilke, Theoretical correlation of velocity and eddy viscosity for ow close to a pipe wall, AIChE J. 9 (1963) 567568. [11] C.H. Chilton, A.P. Colburn, Mass transfer (absorption) coefcient prediction from data on heat transfer and uid friction, Ind. Eng. Chem. 26 (1934) 1183 1187. [12] C. de Waard, Prediction of CO2 corrosion of carbon steel, in: Corrosion/1993, NACE International, 1993, Paper No. 69. [13] R. Zvauya, J.L. Dawson, Electrochemical reduction of carbon dioxide and the effects of the enzyme carbonic anhydrase 11 on iron corrosion, J. Chem. Technol. Biotechnol. 61 (1994) 319326. [14] R.H. Hausler, The mechanism of CO2 corrosion of steels in hot, deep gas wells, Advances in CO2 Corrosion, vol. 1, NACE, 1984. [15] R.H. Hausler, D.W. Stegman, CO2 corrosion and its prevention by chemical inhibition in oil and gas production, in: Corrosion/1984, NACE, 1984, Paper No. 363. [16] A. Ikeda, M. Ueda, S. Mukai, CO2 Behavior of Carbon and Cr Steels, Advances in CO2 Corrosion, NACE, 1984.

Acknowledgments The nancial supports from the Natural Science Fund of China (50231020, 10372077) are gratefully acknowledged. The valuable discussions with Prof. Lin HE of Material Science and Engineering School of Xian Jiaotong University, Prof. Minxu LU of Material Science and Engineering School of Beijing Science and Technology University, Dr. Zhenquan BAI of Tubular Goods Research Center of China National Petroleum Corporation, and the kind help from them are also gratefully acknowledged. References
[1] M.B. Kermani, D. Harrop, The impact of corrosion on the oil and gas industry, J. SPE Prod. Facilities 8 (1996) 186190. [2] M.B. Kermani, A. Morshed, Carbon dioxide corrosion in oil and gas production a compendium, Corrosion 59 (2003) 659683. [3] K.D. Efrd, Flow-induced corrosion, in: R. Winston Revie (Ed.), Uhligs Corrosion Handbook, second ed., John Wiley & Sons, Inc., 2000, pp. 233248. [4] Z.S. Mao, A.E. Dukler, Rise velocity of a Taylor bubble in a train of such bubbles in a owing liquid, Chem. Eng. Sci. 40 (1985) 21582160. [5] W.P. Jepson, Modeling the transition to slug ow in horizontal conduit, Can. J. Chem. Eng. 67 (1989) 234.