Materials and Corrosion 2009, 60, No.

9999

DOI: 10.1002/maco.200805114

Near-wall hydrodynamic effects related to flow-induced localized corrosion

A. Yabuki*
The relationship between near-wall hydrodynamic conditions and flow-induced localized corrosion of copper alloys was investigated. Flow-induced localized corrosion tests for copper alloys were conducted in corrosive solutions using a jet-in-slit testing apparatus, in which various hydrodynamic conditions were reproduced. To evaluate local damage, the profiles of the specimens were measured using a surface roughness meter. In addition, the near-wall hydrodynamic conditions on the material surface in the apparatus were measured by pressure gauges to determine the distribution of near-wall velocities and velocity fluctuations. In the equation used to predict corrosion damage, three types of hydrodynamic conditions were applied as parameters related to the mass transfer coefficient in the film of corrosion products. The damage profiles calculated from the equation could be fitted to that obtained from the corrosion test.

1 Introduction
Localized corrosion frequently occurs near the inlet ends of heat exchanger tubes, which are made of a copper alloy, where seawater is used. Corrosion occurs when the protective film of a corrosion product, formed on the surface of the copper alloy, is broken away by shear stress and turbulence, causing the underlying metal surface to come into direct contact with the corrosive liquid. This phenomenon is known as flow-induced localized corrosion, erosion-corrosion, flow-accelerated corrosion or flow-assisted corrosion (FAC) [13]. Damage to copper alloys by flow-induced localized corrosion largely depends on hydrodynamic conditions such as the flow velocity of a liquid [4]. To predict the extent of damage to copper alloys under a flowing solution, it is imperative to elucidate relationships between damage occurring in the materials and the hydrodynamic conditions of a corrosive solution. Flow-induced localized corrosion of copper alloys proceeds via a diffusion-controlled process in many cases, and the mass transfer equation over the material surface is generally adopted. To apply the mass transfer equation to flow-induced localized corrosion damage, mass transfers in both the concentration boundary layer and the corrosion product film on the material need to be considered, because a corrosion product film formed on the material confers resistance to corrosion [57]. Flow velocity is generally used as a hydrodynamic parameter related to flow-induced localized corrosion damage, because it is quite simple. However, it is
A. Yabuki Graduate School of Engineering, Hiroshima University, 1-4-1, Kagamiyama, Higashi-Hiroshima, Hiroshima, 739-8527 (Japan) E-mail: ayabuki@hiroshima-u.ac.jp

not sufficient for the accurate prediction of damage, since flowinduced localized corrosion frequently occurs in the turbulent portion, such as a pipe bend, an elbow or tee pipe fittings. Several papers have reported that the Sherwood number, a dimensionless number used in mass transfer operations, is useful for the mass transfer coefficient of the concentration boundary layer [813]. Poulson [10] reported that the Sherwood number in many flowing conditions can be estimated through electrochemical measurement. However, the Sherwood number may not be accurate in describing the conditions of a corrosion product film. Schmitt et al. [14] reported that the correlation between turbulent flow intensities as wall shear stress and fracture mechanics properties of corrosion product scales has been theoretical and supported by experimental results. They also reported that the maximum interaction energies between flowing media and solid walls can be quantified in terms of freak energy densities created during singular events of perpendicular impact by near-wall micro turbulence elements with a new model [15]. Nesic and Postlethwaite [16] conducted a numerical simulation of turbulent flow when a rust film is present, and found that the turbulent fluctuations affected both mass transfer through the boundary layer and the removal of the film. A numerical simulation of a pipe flow, to investigate erosion-corrosion, has also been reported [1720]. In this study, flow-induced localized corrosion tests for copper alloys were conducted in corrosive solutions using a jet-inslit testing apparatus. Two types of copper alloys were used as model materials. The damage profile of the specimens was measured by a surface roughness meter to evaluate the local damage. In addition, the near-wall hydrodynamic conditions on the material surface in the apparatus were measured by pressure gauges. The determined hydrodynamic conditions were applied

www.matcorr.com

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Yabuki

Materials and Corrosion 2009, 60, No. 9999

to the equation used to predict the corrosion damage. The relationship between near-wall hydrodynamic effects on the material surface and corrosion of metallic materials under a flowing solution was investigated.

2 Experimental
2.1 Flow-induced localized corrosion test A jet-in-slit testing apparatus was used for corrosion tests under a flowing solution [21]. The apparatus consisted of a test solution reservoir, a pump, a flow meter and a test cell. Figure 1 shows a detailed schematic rendering of the test cell. In this apparatus, the test solution was allowed to flow from the nozzle into the slit between the specimen and the nozzle. The diameter of the specimen was 16 mm. The nozzle was made of a polymethylmethacrylate resin with a bore diameter of 1.6 mm. The gap between the nozzle top and the specimen was 0.4 mm. The impingement of solution brought about high shear stress on the specimen surface. On the other hand, as the solution was injected from the nozzle mouth into the slit, the solution filled the slit and flowed radially over the specimen surface. As it approached the periphery of the specimen, the cross-sectional area of the flow increased, and the flow velocity decreased accordingly. The rapid reduction in flow velocity created an intense turbulence in the flow, similar to what is expected downstream of orifice plates [21]. As a result, localized corrosion damage in the jet-in-slit test can be accounted for primarily by shear stress and the turbulence of the flow. The detailed near-wall hydrodynamic conditions in the apparatus were investigated using a measuring system, as shown in the next section. Brass (58.3Cu-38.2Zn-3.10Pb-0.17Fe- 0.23Sn) and 70CuNi (30.2Ni-Cu) were used as model materials for the corrosion tests. A specimen of each, 16 mm in diameter, was wet-polished with #2000 emery paper, followed by a thorough rinsing in acetone with an ultrasonic bath, followed by drying in air. After masking the side and back of each specimen with a polyester tape, it was used in the corrosion tests. A 3% NaCl solution and a 1 wt%

CuCl2 solution saturated with air were used as the corrosive test solutions. The concentration of dissolved oxygen in the solutions was approximately 6.45 ppm. Cu2 in a 1 wt% CuCl2 solution acted as the oxidizing agent to accelerate the corrosion reaction. The temperature of the test solution was maintained at 40 8C. The flow rate was set at 0.4 L/min. In these conditions, the fluid velocity at the nozzle outlet was 3.3 m/s. The test duration was 24 h in a 3% NaCl solution and 1 h in a 1 wt% CuCl2 solution. Damage depth was determined by comparing the differences in the specimen surface profiles before and after the test, using a surface roughness meter and measurement of the mass loss of the specimen. 2.2 Measurement of near-wall hydrodynamic conditions The near-wall hydrodynamic conditions of each specimen were measured in the jet-in-slit corrosion testing apparatus by using two pressure gauges and a wire. The set-up for the measuring system is shown in Fig. 2. The fluid in a tank flows to a nozzle through a flow meter with a pump, and then returns to the tank. Three types of nozzles were prepared, one the same size as the nozzle of the corrosion testing apparatus, the others were twoand fivefold scale-ups. The fluid velocity at the nozzle equaled that of the corrosion testing apparatus. Two holes 0.3 mm in diameter were bored in the surface of the measurement plate, and a wire 0.05 mm in diameter was set between the holes, as shown in Fig. 3. Pressure gauges (PGM-02KG, Kyowa Electronic Instruments Co., Ltd.) were connected to the holes. The signal from each pressure gauge was input to a personal computer through the sensor interface (PCD-300, Kyowa Electronic Instruments Co., Ltd.). The horizontal velocity Vx (m/s) and the vertical velocity Vy (m/s) were calculated by the differential pressure DP P1 P2 (Pa) and the wall pressure upstream of the wire, P1 (Pa), respectively. Measured pressure was translated to velocity by

Figure 1. Test section in the jet-in-slit corrosion testing apparatus

Figure 2. Set-up for the system to measure the near-wall hydrodynamic conditions of a specimen in the jet-in-slit corrosion testing apparatus

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.matcorr.com

Materials and Corrosion 2009, 60, No. 9999

Near-wall hydrodynamic effects related to corrosion

dependence of horizontal velocity Vx and vertical velocity Vy, respectively. Vxa and Vya are equal to the wall shear stress and the hydrodynamic energy density, respectively, which are proposed by Schmitt et al., [14] and Schmitt and Bakalli [15].

3 Results and discussion

3.1 Damage profile of specimens Cross-sectional profiles of brass and 70CuNi specimens after flow-induced localized corrosion tests at a flow rate of 0.4 L/min are shown in Fig. 5. A thin line indicates the profile before the test, determined as volume loss calculated from mass loss and the density of a specimen coinciding to the volume loss calculated from the surface profile before and after the test. Brass was largely damaged in the central part of the specimen (part A) and in the part approximately 2 mm from the center of the specimen (part B). The damage to the peripheral part was relatively low. This was obviously due to the hydrodynamic effect of the corrosive solution. Similar to the brass, 70CuNi was largely damaged in the central area (part A and B), compared with the damage to the peripheral part. Concerning the damage in the central area, the damage at part B was deeper than that at part A for brass, but the damage at part B was less than that at part A for 70CuNi. The difference should be due to differences in the corrosion product film formed on the surface of each specimen under a flowing solution, because the hydrodynamic conditions near the specimen surface were very similar for each material. The corrosion rate for both materials differed with the concentrations of the oxidizing agents. The flow-induced localized corrosion of copper alloys proceeded through both the initiation step, which occurred following the mechanical destruction of corrosion products film, and the propagation step, which occurred following the repeated formation and breakaway of the products due to the local

Figure 3. Dimensions of two holes connected to the pressure gauges and a wire set on the measuring plate for measuring near-wall hydrodynamic conditions

Equations (1) and (2), given by Bernoullis law, s P1 P2 Vx 2 r s 2P1 Vy r

(1)

(2)

where r (kg/m3) is the density of solution. The measuring plate was moved from the center of the nozzle mouth to the peripheral part to measure each point of the specimen surface. The right half distributions in the apparatus were measured, since the left half distributions should be almost the same as the right side. The sampling time was 10 s and the frequency of the sampling was 2 kHz. The average velocity and the uctuation in the horizontal (x) and vertical (y) directions to the specimen surface, Vxa, Vxf, Vya, and Vyf (as shown in Fig. 4), were calculated from the time

Figure 4. The near-wall average velocity and the fluctuation in the horizontal (x) and vertical (y) directions to the specimen surface, Vxa, Vxf, Vya, and Vyf

Figure 5. Cross-sectional profiles of brass (upper) and 70CuNi (lower) specimens after flow-induced corrosion tests at a flow rate of 0.4 L/min. The thin line is the profile before the test

www.matcorr.com

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Yabuki

Materials and Corrosion 2009, 60, No. 9999

hydrodynamic effect [14, 15]. Since wet-polished specimens were tested in this study, the damage to copper alloys (Fig. 5) was generated mostly in the propagation step, which was equal to the steady state. The porous corrosion products film was recognized by the observation specimen surface after the corrosion test, which agrees with previous research [21]. 3.2 Near-wall hydrodynamic conditions Near-wall hydrodynamic conditions in the jet-in-slit apparatus were measured using three types of nozzles. The distribution of hydrodynamic conditions, measured using the nozzle that was the same as the nozzle of the corrosion testing apparatus, did not appear to be smooth, because the holes for the pressure gauge and wire were too large. The distributions of hydrodynamic conditions measured using the two- and fivefold scale-up nozzles were, however, very smooth, and these data were almost the same. Measurement points were easily adjusted using the fivefold scaleup nozzle. Therefore, that nozzle was adopted to measure the near-wall hydrodynamic conditions. To compare the hydrodynamic conditions with the corrosion damage, the portion measuring hydrodynamic conditions corresponds to the nozzle in the corrosion testing apparatus, that is, the value of the portion was divided by 5. Time dependence of the differential pressure at 1.5 mm from the center of the specimen, DP, and wall pressure at the center of the specimen, P1, are shown in Fig. 6. Fluctuation in DP and P1 was recognized, and the frequencies were approximately 200 and 130 Hz, respectively. The frequencies were not the same in both areas, so the fluctuation was due to the vibration of the fluid rather than to the pulsation of the pump. DP and P1 measured by pressure gauges were translated to horizontal velocity Vx and vertical velocity Vy by Equations (1) and (2), respectively. The horizontal velocity Vx, furthermore, was calculated from the average velocity Vxa and the fluctuation Vxf. The Vxf was obtained as three times the standard deviation of the velocity. The vertical velocity and the horizontal velocity were also

Figure 7. The distributions of near-wall hydrodynamic conditions of specimens, Vxa, Vxf, Vya, and Vyf. Only the right half distributions in the apparatus are shown in the figure

treated to obtain Vya and Vyf. The distributions of the velocities and the fluctuations in the horizontal and vertical directions were obtained by moving the measuring plate. Figure 7 shows the distributions of near-wall hydrodynamic conditions of the specimens, Vxa, Vxf, Vya, and Vyf. Only the righthalf distributions in the apparatus are shown in the figure. The average vertical velocity Vya was highest at the center of the specimen, since the solution was injected toward the specimen vertically. Vya showed negative pressure in the area outward from 1 mm; this was caused by boundary layer separation in the area. The average horizontal velocity Vxa was highest at the portion 1 mm from the center of the specimen, which corresponded to the edge of the nozzle mouse, because the flow direction changed at this location. The fluctuation of vertical and horizontal velocities Vya, Vyf was highest at the portion 2 mm from the center. As a result, the hydrodynamic parameters had different distributions, and the portion showing the maximum differed among these parameters. Thus, we could measure the various hydrodynamic conditions in the jet-in-slit corrosion testing apparatus. In order to relate the hydrodynamic conditions to the corrosion damage, Vxa, Vxf, and Vya were selected as follows. Only positive pressure was used for Vya, because negative pressure in the area outward from 1 mm was caused by boundary layer separation and was related to the fluctuations, Vxf and Vxf, to be substituted for it. The vertical velocity fluctuation Vyf was similar to the horizontal velocity fluctuation Vxf, so only Vxf was used to predict the corrosion damage.

4 Prediction of flow-induced localized corrosion damage by hydrodynamic conditions

Figure 6. Time dependence of the differential pressure at 1.5 mm from the center of the specimen DP and wall pressure at the center of the specimen P1

The corrosion damage of copper alloy in a 1 wt% CuCl2 solution was much larger than the damage in a 3% NaCl solution, as

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.matcorr.com

Materials and Corrosion 2009, 60, No. 9999

Near-wall hydrodynamic effects related to corrosion

shown in Fig. 5. This indicated that the corrosion damage was mainly determined by the cathodic reaction. Thus, the flowinduced localized corrosion of copper alloys mainly proceeds under cathodic control, so that the rate-controlling step in corrosion is the transport of the oxidizing agent from the bulk of the fluid to the metal surface. The corrosion rate should be proportional to the diffusion rate of the oxidant. In the steady state, the corrosion rate, Rc (m/s), can be given as follows [7]. Rc K cb 1=kc 1=kd (3)

where K (m3/mol) is a conversion factor, cb (mol/m3) are the oxidizing agent concentrations in the bulk of the owing uid and kc and kd (m/s) are the mass transfer coefcients in the concentration boundary layer and in the corrosion products lm, respectively. Accordingly, the local damage d (mm) can be given as follows by using the testing time t (h). d 3:6 109 Kcb t 1=kc 1=kd (4)

Figure 8. The experimental and calculated profiles of damage depth for brass tested in 3% NaCl solution for 24 h. The figure show only the right half of the specimen

Assuming that the film of the corrosion products that formed on the copper alloys was thicker than the concentration boundary layer, or that the film was thinner but more dense, it had a structure that served as diffusion resistance to the oxidizing agent. Therefore, the diffusion of the oxidizing agent in the corrosion products film was relatively low for a rate controlling step, namely kc ) kd. The Equation (4) then gives d 3:6 109 Kkd cb t (5)

The corrosion products film formed on the surface was in a steady state, and its thickness and structure were determined by repeated formation and breakaway by the hydrodynamic effects. In the process, the condition of the film should be determined by the mechanical force acting on the film. When the velocity was high, the force acting on the film was large, resulting in the film becoming thinner, so that the mass transfer coefficient in the corrosion products film became larger. Hence, the mass transfer coefficient in the corrosion products film kd (m/s) was assumed to be proportional to the velocities at the near-wall, as follows kd axa Vxa axf Vxf aya Vya (6)

8.9 106 g/m3 is the density of copper), cb 0.20 mol/m3 which corresponds to dissolved oxygen of 6.45 ppm, t 24 h, and each a obtained from fitting to measured data was axa 2.3 107, axf 4.6 105 and aya 1.2 106, respectively. The data used for 70CuNi tested in a 1 wt% CuCl2 solution were as follows: K 7.1 106 m3/mol (63.5/1/8.9 106, where 63.5 g/mol is the molecular weight of copper, 1 is the number of ion-exchanges in anodic and cathodic reactions, and 8.9 106 g/m3 is the density of 70CuNi), cb 74 mol/m3 which corresponds to the concentration of Cu of 1 wt% CuCl2 solution, t 1 h, and each a obtained from fitting to measured data was axa 3.8 106, axf 8.8 106, aya 3.5 106, respectively. The calculated profiles were consistent with the experimental data and with the areas of maximum damage for both copper alloys 2 mm for brass and 1 mm for 70CuNi. Comparing the contributing ratio to each hydrodynamic condition a, the fluctuation of horizontal velocity, axf, was dominant in the corrosion damage of brass. On the other hand, average vertical velocity, axa, also affected the corrosion damage of 70CuNi, in addition to the fluctuation of

where each a is a material-specic constant corresponding to the contributing ratio for each hydrodynamic condition. The profile of damage depth for copper alloys was calculated using equations (5) and (6) as fitted to the experimental damage profile by a trial-and-error method. The calculated and experimental profiles for brass and 70CuNi are shown in Figs. 8 and 9. The figures show only the right half of the damage profile for each specimen. The data used for brass tested in a 3% NaCl solution were as follows: K 3.6 106 m3/mol (63.5/2/8.9 106, where 63.5 g/mol is the molecular weight of copper, 2 is the number of ion-exchanges in anodic and cathodic reactions, and

Figure 9. The experimental and calculated profiles of damage depth for 70CuNi tested in 1 wt% CuCl2 solution for 1 h. The figure show only the right half of the specimen

www.matcorr.com

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Yabuki

Materials and Corrosion 2009, 60, No. 9999

horizontal velocity, axf. Thus, the hydrodynamic effect on corrosion damage for copper alloys did not express a single parameter, such as flow velocity or Sherwood number, but instead was related to multiple effects from horizontal and vertical force and fluctuation. The material-specific constant a in the Equation (6) is related to the mechanical properties of corrosion products film formed on the surface of copper alloys, and consequently the properties are particularly important for the prediction of corrosion damage under a flowing solution.

Acknowledgements: This research was supported in part by the Grant-in-Aid for Scientific Research (C), Japan Society for the Promotion of Science (No. 17560635).

References

5 Conclusions
The relationship between near-wall hydrodynamic effects on the material surface and corrosion damage of material under a flowing solution was investigated using a jet-in-slit testing apparatus, and the following results were obtained. (1) The damage profile for brass and 70CuNi specimens was measured using a surface roughness meter to evaluate local damage. The portion showing maximum damage was different in each material. (2) The near-wall hydrodynamic conditions on the surface of corrosion test specimens were measured by pressure gauges, and the distribution of near-wall velocity and velocity fluctuation was determined. Three types of hydrodynamic conditions were applied as parameters related to the mass transfer coefficient in corrosion products film in the equation used to predict the corrosion damage. (3) The damage profiles calculated from the equation could be fitted to that obtained from the corrosion test using three types of hydrodynamic parameters. (4) The determined material-specific constant agreed with the mechanical properties of corrosion product films against a hydrodynamic effect in corrosive liquid. Consequently, the mechanical properties of corrosion products film formed on the surface of copper alloys are particularly important for the prediction of corrosion damage under a flowing solution.

[1] M. Murakami, K. Sugita, A. Yabuki, M. Matsumura, Corros. Eng. (Zairyo-to-Kankyo) 2003, 52, 155. [2] M. Murakami, A. Yabuki, M. Matsumura, Corros. Eng. (Zairyo-to-Kankyo) 2003, 52, 160. [3] B. Chexal, J. Horowitz, R. Jones, B. Dooley, C. Wood, M. Bouchacourt, F. Remy, F. Nordmann, P. St. Paul, FlowAccelerated Corrosion in Power Plants, EPRI Distribution Center, Pleasant Hill, CA, 1996. [4] B. C. Syrett, Corrosion 1976, 32, 242. [5] B. K. Mahato, C. Y. Cha, L. W. Shemilt, Corros. Sci. 1980, 20, 421. [6] M. Matsumura, K. Noishiki, A. Sakamoto, Corrosion 1988, 54, 79. [7] A. Yabuki, M. Murakami, Mater. Corros. 2008, 59, 25. [8] B. T. Ellison, C. J. Wen, Am. Inst. Chem. Eng. 1981, 77, 161. [9] T. Sydberger, U. Lotz, J. Electrochem. Soc. 1982, 129, 276. [10] B. Poulson, Corros. Sci. 1983, 23, 391. [11] B. Poulson, Corros. Sci. 1993, 35, 655. [12] B. Poulson, Wear 1999, 233235, 497. [13] J. A. Wharton, R. J. K. Wood, Wear 2004, 256, 525. [14] G. Schmitt, C. Bosch, M. Muller, G. Siegmund, CORROSION2000, NACE International, Houston/Texas, 2000, Paper 049. [15] G. Schmitt, M. Bakalli, Mater. Corros. 2008, 59, 181. [16] S. Nesic, J. Postlethwaite, Can. J. Chem. Eng. 1991, 69, 698. [17] B. Bozzini, M. E. Ricotti, M. Boniardi, C. Mele, Wear 2003, 255, 237. [18] Y. M. Ferng, Y. P. Ma, N. M. Chung, Corrosion 2000, 56, 116. [19] A. Keating, S. Nesic, Corrosion 2001, 57, 620. [20] J. Postlethwaite, S. Nesic, G. Adamopoulos, D. J. Bergstrom, Corros. Sci. 1993, 35, 627. [21] M. Matsumura, Y. Oka, S. Okumoto, H. Furuya, ASME STP 1985, 866, 358.

(Received: July 16, 2008) (Accepted: November 9, 2008)

W5114

2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

www.matcorr.com