Experimental Study of High-Temperature Chlorine-Induced Corrosion

May 16, 2013 - Experimental Study of High-Temperature Chlorine-Induced. Corrosion in Dependence of Gas Velocity. Gundula Balan,*. ,†. Marco Losurdo,...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/EF

Experimental Study of High-Temperature Chlorine-Induced Corrosion in Dependence of Gas Velocity Gundula Balan,*,† Marco Losurdo,† and Hartmut Spliethoff†,‡ †

Institute for Energy Systems, Technische Universität München, 85748 Garching bei München, Germany ZAE Bayern, 85748 Garching bei München, Germany



ABSTRACT: High-temperature chlorine-induced corrosion is mainly responsible for the limitation of plant availabilities and electrical efficiencies of waste- and biomass-fired boilers. According to the current state of research, it is controversial if the flue gas velocity influences the corrosion on heat-transfer surfaces. References regarding the behavior of corrosion at different gas velocities are scarcely available, and the complexity in chemistry as well as the varying combustion conditions in conventional plants complicates the interpretation of observations. On the basis of this background, this paper aims to discuss the theoretical background of velocity-dependent corrosion and also investigates the influence of gas velocity on corrosion under ideal laboratory conditions. The work includes both experimental and numerical simulations to study the deposition and corrosion behavior at velocities of 0.6, 1.7, 2.8, and 3.9 m/s. The experimental corrosion tests were carried out in an electrically heated laboratory test rig, where a defined mass of KCl was vaporized and transported by different air flows. An air-cooled corrosion probe was inserted into the reaction tube, and after a set exposure time, the oxide layer thicknesses of the corrosion rings were analyzed metallographically. Accompanied by ANSYS simulations, the condensing rates of KCl on the probe surface were calculated for the different velocity cases. The experiments show tendencies that the mean corrosion rates rise with increasing flue gas velocities. From the experiments and simulations, it can be concluded that the gas velocity influences the condensation and deposition mechanisms, which also directly affects corrosion.



INTRODUCTION The combustion of fuels with high chlorine contents, such as biomass and waste, can lead to considerable corrosion problems on heat-transfer surfaces, which limit the plant availabilities and cause high maintenance costs for plant operators.1,2 The high-temperature chlorine-induced corrosion is, in the first instance, dependent upon the fuel composition, because the fuel defines the load of corrosive species in the flue gas. During biomass and waste combustion, the most corrosive components in the flue gas are alkali chlorides or volatile heavy metal chlorides, which can condense or deposit on heat-transfer surfaces and provide the agents for further corrosion reactions.3 A basic research concerning the corrosion mechanisms can be found in refs 2 and 4−10. Second, if a sufficient amount of chloride is provided in the flue gas, the process parameters, such as metal temperature, flue gas temperature, and flue gas velocity, have an important influence on the degree of corrosion.11−13 On the basis of the experiences of German plant operators in the field of waste combustion, the “Flingern corrosion diagram” shows the qualitative probability of high-temperature corrosion in dependence of the metal and flue gas temperatures.14,15 If the temperature-dependent corrosion follows an Arrhenius relation, as reported in the literature,16,17 an increase of corrosion rates with an increasing temperature can be expected according to K = K 0e

−EA / TR

the temperature (K), and R is the universal gas constant (J mol−1 K−1). In addition to the temperature-dependent corrosion, plant operators observed that, in areas of higher flue gas velocities, high-temperature corrosion can be more serious.14,15,18 The reasons for higher flue gas velocities are for example (i) direction change of flue gas flow in the lateral gas pass, (ii) flow skew positions caused by poor adjustments of secondary air, and (iii) flow breakaway edge between fireproof materials and metal tube surfaces. To consider the flue gas velocity in the estimation of corrosion risks, the “Flingern corrosion diagram” has been extended by regions of lower and higher gas velocities, which are shown in Figure 1.18 The qualitative extension is based on corrosion measurements and simulations in waste-to-energy plants. A quantification of the diagram can only be carried out plant-dependently, because of varying fuel compositions and process parameters in the plants. However, first trials of extensive online-corrosion measurements with a corrosion probe in a waste-fired boiler show that a quantification of the diagram is very challenging, because of the limited number of measuring positions and fluctuating parameters.19−21 The mutual interference of different process parameters complicates the investigation of each single parameter, especially the flue gas velocity, on corrosion. Just a few references15,18−22 to scientific Special Issue: Impacts of Fuel Quality on Power Production and the Environment

(1)

Received: March 12, 2013 Revised: May 13, 2013 Published: May 16, 2013

where K is the corrosion rate constant, K0 is the preexponential factor, EA is the activation energy (J/mol), T is © 2013 American Chemical Society

5628

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

rates according to eq 1. On tubes covered with a respective deposit layer, the higher heat-transfer coefficient also leads to higher temperatures of the outer deposit layer, whereupon the heat accumulates in the deposits because of their thermal insulation properties. If the corresponding melting temperature is exceeded, then the deposits start to melt. Because the diffusion coefficients of liquids are much higher than for solids, the corrosion kinetics can increase.24 Chemical Influence. The chemical influence by means of higher gas velocities becomes noticeable under boiler construction aspects. For instance, the direction changes of the flue gas flow in the lateral gas pass or narrowing of the flow channels is mostly coupled with an increase of the mass flux of corrosive aerosols and particles, and therefore, the higher load on corrosive components increases the corrosion risk. Furthermore, the residence time of the ash and gas species in the flue gas channel is shorter at higher gas velocities, such that the achievement of reaction equilibriums is constrained.16 Mechanical Influence. According to the mechanical influence on corrosion phenomena, two different characteristics need to be considered: (A) the mechanical influence of the gas flow and (B) the mechanical influence of the ash particles transported by the gas flow. (A) The mechanical influence of the gas flow appears at the upstream side of the tubes in the form of rising stagnation pressures. However, the stagnation pressure difference between a tube in a 1 m/s and a 10 m/s air flow (ρ = 0.44 kg/m3 at 800 °C) is about 2.18 mbar. Corresponding to thermodynamic equilibrium calculations, the effect on partial pressures of the corrosive species at this low-pressure difference is negligible. If the diffusion fluxes of the corrosive species through the deposits and oxide layers are increased by stagnation pressures, it is very difficult to prove experimentally. (B) The mechanical influence of ash particles probably plays the most important role for high-temperature corrosion at high gas velocities. Dependent upon the size and velocity, particles can have high kinetic energies, which are released during their impacts onto the heat-transfer tubes. Also, the influence of ash particles transported by the gas flow can be divided into two cases: (I) Particles with low up to high kinetic energy can affect the deposit formation in dependence of the gas velocity. (II) Particles with very high kinetic energy can wear down the protecting oxide layers from the tube metals up to the tube material itself (erosion corrosion). With regard to case I, according to the theoretical and experimental work of Ots,25 the deposit formation from solid particles is dependent upon the particle size, the chemical and mineral composition of the gases and particles, the aerodynamic conditions, and also, the flue gas velocity. He describes three groups of deposits: (a) unbound, (b) loose-, and (c) hard-bound deposits. On the condition that smaller particles have low kinetic energy and deposit mainly by diffusion and bigger particles have high kinetic energy and deposit by impaction, because of their inertia, the unbound deposits (a) result mainly from diffusion or impaction of non-reactive particles, which are retained by the convective heat-transfer surfaces. In contrast, bound deposits (b and c) form, if before retained particles are consolidated by reactive particles, such as sulfates and chlorides, which further react with other components in the flue gas or deposits. The qualitative dependence of the deposit growth rate from the flue gas velocity is shown in Figure 2.

Figure 1. “Flingern corrosion diagram” adapted from Warnecke.15

investigations considering this topic are available, and the correlations cannot be understood to date. The goal of this study is to investigate the dependence of high-temperature chlorine-induced corrosion on the gas velocity under simplified reaction conditions in a laboratory test rig. The adjustment of defined test parameters enables exclusive consideration of the velocity dependence, without mutual interferences of temperatures or chemical reactions between the flue gas and ash species. This new method helps to obtain a deeper understanding of corrosion behavior, which is of great importance for the development and layout of waste- and biomass-fired boilers.



THEORETICAL BACKGROUND In general, the influence of the flue gas velocity on corrosion phenomena may be of thermal, chemical, or mechanical origin. Because these three influences take place simultaneously, the complexity of the corrosion processes is high, and therefore, understanding the subject is difficult. The three influences should be discussed separately. Thermal Influence. With rising flue gas velocities, the thermal transfer conditions are improved. The Reynolds number Re is a dimensionless number in fluid mechanics and describes the ratio of inertial forces to viscous forces. Because the mean Reynolds number Rem is proportional to the mean flue gas velocity wm (m/s), the Reynolds number increases with rising gas velocities, where D is the hydraulic diameter of the tube (m) and ν is the kinematic viscosity (m2/s). Rem =

wmD ν

(2)

This means that, for a cross-flow tube bundle heat exchanger, the dimensionless Nusselt number Nu, which describes the ratio of convective to conductive heat transfer, rises mathematically as a power function with an increasing Reynolds number for the boundary condition 2 × 103 < Re < 4 × 104.23 Pr is the Prandtl number; KA is a correction factor for the configuration of the tubes; and KR is a correction factor for the number of tube rows. Nu = 0.32Rem 0.61Pr 0.31KAKR

(3)

Because the Nusselt number is proportional to the heat-transfer coefficient α (W m−2 K−1) and, therefore, to the heat flux density q (W/m2) Nu ≈ α

and

α≈q

(4)

the heat transfer is increased with higher flue gas velocities. This means that, for a clean tube surfaces, higher local surface temperatures can exist, which can result in higher corrosion 5629

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

in the deposits is high and their consistency is hard-bound. From velocities of w2,crit, the deposit formation is impossible. Additional to Ots description about the velocity-dependent deposit growth rate, the surface temperature of the deposits also influences the growth rate.26 If the outer surface temperature exceeds the melting range of the deposits, the stickiness and, thus, the growth rate will increase.27 An overview about the structure and chemical composition of deposits, which have been formed under typical boiler operation conditions in solid fuel-fired power plants, are given in refs 2 and 27−30. As already discussed, Figure 2 shows that deposit formation is very dependent upon the flue gas velocities, but how does this affect corrosion? Hohmann31 describes that the structure of the deposits decides whether high-temperature chlorine-induced corrosion is taking place or not. Deposits and oxide layers have pores, where the mass transport of diffusing chlorine and oxygen is ensured. A higher porosity, such as weakly bound deposits, represents higher diffusion fluxes. Chlorine as well as oxygen can diffuse through open pores until an equilibrium between the outer metal layer and the deposit layer is arranged. However, a corrosion process with a formation of FeCl2 on the outer layer of the corroding metal is only stable if the oxygen partial pressure is very low.6 This means that the deposits need to be very dense and compact, such as hard-bound deposits with a low porosity, so that the mass transfer is the rate-limiting step and corrosion takes place.

Figure 2. Qualitative ash deposit growth rate of bound deposits as a function of the flue gas velocity adapted from Ots.25

At low gas velocities, non-reactive and reactive particles deposit on heat-transfer surfaces and no destruction effects on deposits occur. Because of the low kinetic energy, the contact density of the particles in the weakly bound deposits is low. With rising velocities, the deposit growth rate rises until a maximum is reached at wdepo,max. With further rising velocities, some non-reactive particles, which do not react with other components and are therefore not consolidated, are blown off by large impacting particles. Therefore, the growth rate decreases and the ratio between reactive and non-reactive particles in the deposits increases. At a certain velocity of w1,crit, barely reactive particles are able to remain and form deposits. Because of the proceeding reactions, the density of the particles

Figure 3. Laboratory test rig. 5630

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

Figure 4. Corrosion probe. temperature settings of the heating elements in the evaporation zone and second with the insertion depth of the sample holders. Hence, the same evaporation rate could be ensured at temperatures of 800 ± 20 °C during all tests. Pressurized air was used as a carrier gas for KCl vapors and was added to the evaporation zone, where it has been loaded with KCl. The velocity of the carrier gas has been adjusted by a mass flow controller in the range of 1−7 N m3 h−1. Therefore, the gas velocities were set to 0.6, 1.7, 2.8, and 3.9 m/s at gas temperatures of 800 °C. An air-cooled corrosion probe was inserted through an optical port into the reaction tube of the condensing zone perpendicular to the air flow. On the basis of extensive preparatory studies, the adjustments of the heating elements were chosen, so that the gas temperature at the position of the corrosion probe always reached temperatures of 800 ± 10 °C. Figure 4 shows schematically the corrosion probe, which was used for the tests. The probe was equipped with two corrosion rings of fastcorroding materials, such as S235JR and high-purity ARMCO iron. The surface temperature of the front S235JR rings could be controlled by a proportional−integral−derivative (PID) controller, which adjusted the air valve positions and, therefore, the cooling air flow rate through the probe. In consequence of the temperature gradient along the corrosion probe, the surface temperature of the back ARMCO-iron rings were always lower than that of the front S235JR rings. The temperature of each ring has been monitored by thermocouples. The ring surfaces were cooled in the temperature range of 300 and 500 ± 5 °C and were exposed to the corrosive atmosphere for 4 h. After the exposure time, the rings have been embedded in epoxy resin and cut in metallographic cross-sections and longitudinal sections. Subsequently, the surfaces of the probes have been polished by sand paper. The mean oxide layer thickness from the cross-section has been measured at eight different positions (Figure 5) initiating with the

With regard to case II, particles with sizes of a few micrometers and, therefore, with very high kinetic energy can wear deposits and protecting oxide layers from the tube metals up to the tube material itself. This process is named erosion corrosion. The erosion corrosion is well-established in relation to liquid fluids, for instance water pipes. Kastner et al.32,33 found a linear correlation between water velocity and metal attack. Differences in metal loss could be observed in dependence of temperature, material, and the geometries of the pipes. On the other hand, the erosion in gaseous fluids through incoming particle impacts has been studied several times in connection with the removal and shedding of deposits2,27,30 or the erosion of different metallic or ceramic materials in a wide range of industrial applications.34−41 An extensive overview about the present knowledge regarding shedding of fireside ash deposits is given by Zbokar et al.27 The degree of erosion in gas flows is mostly dependent upon the particle velocity, their size and shape, the angle of impact, the composition of the eroding particles, the properties of the surface being eroded, and the temperature of the system.42 The material loss, caused by erosion of impacting particles, is proportional to the impact energy of the particles and follows a power law.40,34,43 Here, the velocity constitutes if erosion promotes corrosion or if corrosion promotes erosion.44 It can be summarized that erosion corrosion in combustion plants can cause two effects: on the one hand, particle erosion can damp corrosion, because of the removal of the corrosive deposits and salt melts without demolition of the protective oxide layer, and on the other hand, some literature describes that particle erosion could accelerate the corrosion process, because of the destruction of the protective magnetite layer.16



MATERIALS AND METHODS

Experimental Section. In this study, the influence of gas velocity on high-temperature chlorine-induced corrosion has been experimentally investigated in a laboratory test rig, which is shown in Figure 3. The test rig is sectioned in an evaporation zone and a condensing zone. Each zone consists of a horizontal ceramic tube, which is surrounded by heating elements to adjust to a defined temperature profile up to maximum temperatures of 1200 °C. The outer casing of the test rig has been cooled by cooling water. For each test, a mass of 4 g of potassium chloride has been pressed to tablets, which were placed on two horizontally movable sample holders. Because of the varying heat transfer from the heating elements to the gas at different gas velocities, the evaporation temperatures could first be adjusted by

Figure 5. Positions of measurement on the corrosion ring: (left) crosssection and (right) longitudinal section. 5631

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

windward side (positions 1, 2, and 8), over the lateral sides (positions 3 and 7), to the leeward side (positions 4, 5, and 6). At the longitudinal sections, measurements of the windward and leeward sides have been carried out. An optical microscope (Keyence VHX500) and a scanning electron microscope (SEM) with a coupled X-ray (JEOL JSM5900LV) have been used. The geometric data of the test rig are listed in Table 1, and the test matrix is presented in Table 2.

Table 1. Geometric Data of the Laboratory Test Rig description evaporation tube condensing tube corrosion probe

value

length (mm) inner diameter (mm) length (mm) inner diameter (mm) retraction depth (mm) outer ring diameter (mm)

1310 40 2470 50 45 13.20

Figure 6. FactSage equilibrium calculation of the KCl phases at different temperatures.

Table 2. Test Matrixa gas velocity, w (m/s) xKCl (mm) Tgas (°C) cKCl(Tgas) (g/m3) ttest (h) T1 (°C) T2 (°C) T3 (°C) T4 (°C) T5 (°C) TS235JR (°C) TARMCO (°C)

case 1

case 2

case 3

case 4

0.6 1010 813 0.25 4 800 840 780 750 650 500 400 435 309

1.7 1100 805 0.08 4 900 930 790 750 650 500

2.8 1150 805 0.05 4 1030 1030 800 770 650 500 400 428 313

3.9 1180 805 0.04 4 1120 1130 810 780 650 500

410

resis, which would occur at lower gas temperatures, have not been considered within the numerical simulation. A specific user define function (UDF) was developed to model heterogeneous condensation. In the developed UDF, both the partial and saturation pressure calculations of KCl are included. The KCl saturation pressure psat,KCl (bar) at different temperatures T (K) was calculated using Antoine’s law

log(psat,KCl ) = A −

⎛ B ⎞ ⎜ ⎟ ⎝C + T ⎠

(6)

where A = 4.78236, B = 7440.691, and C = −122.709 for KCl [values taken from thermodynamic National Institute of Standards and Technology (NIST) tables45]. The molar saturation fraction xsat,KCl is calculated as a quotient of the saturation pressure and the absolute pressure.

461

⎛p ⎞ sat,KCl ⎟⎟ xsat,KCl = ⎜⎜ ⎝ pabs ⎠

a

xKCl, insertion depth of the KCl tray; Tgas, gas temperature; cKCl(Tgas), concentration of KCl in the gas phase at Tgas; T1−T5, set temperatures of the heating elements; and TS235JR/TARMCO, surface temperature of the different corrosion rings.

(7) −2

The deposition rate, DR (mol m

DR = C DR ρ0 D0(yKCl − ysat,KCl ) With reference to the theoretical background of this paper, the velocity-dependent corrosion can be of thermal, chemical, and mechanical origins. To minimize the complexity of these different effecting parameters, some influences could be excluded for the tests: (i) The thermal influence has been excluded by precise and uniform regulation of the gas temperature and surface temperature of the corrosion rings at different gas velocities. (ii) The chemical influence resulting from different mass flux of evaporated salts has been excluded by always adding the same mass of KCl in the reactor. (iii) The influence of erosion corrosion can be excluded because of missing bigger particles with high kinetic energy and erosive properties. Because just reactive salt vapors were present in the process gas, the deposition behavior is the only mechanism, which is affected by different gas velocities. Numerical Simulation. Computational fluid dynamics (CFD) simulations have been performed, aiming to provide a better prospective of the fluid dynamics influence on the deposition and corrosion processes. According to Table 2, four different cases (case 1, 0.6 m/s; case 2, 1.7 m/s; case 3, 2.8 m/s; and case 4, 3.9 m/s) and grids were set up and used to simulate in ANSYS Fluent KCl deposition. A constant probe surface temperature of 500 °C for both corrosion rings was assumed for the simulations. According to FactSage thermodynamic equilibrium calculations (Figure 6), KCl is present in a gaseous phase at flue gas temperatures of approximately 800 °C; therefore, condensation is the predominant deposition process. Other possible deposition processes of already homogeneously formed fine particles, for example, diffusion or thermopho-

−1

s ), is therefore calculated as

1 1 d MKCl

(8)

where CDR is the calibration coefficient and ρ0 (kg/m ) and D0 (m /s) are the center cell gas density and the diffusion coefficient, respectively. yKCl is the local mass fraction of KCl, while ysat,KCl is the saturation value within the cell. MKCl (g/mol) is the molar mass of KCl, and d (m) represents the distance between the center of the boundary cell to the tube surface. The calibration coefficient also comprises the conversion of kilograms to grams and serves as a numerical adaption, independent of the gas velocity. To prevent the ANSYS Fluent from returning unrealistic results, the gas-phase density, temperature, diffusion coefficient, and concentrations were calculated at the center of each boundary cell instead of at the center of each boundary surface. To calibrate CDR, two sets of simulations were performed, injecting one halves and the full amount of KCl used in the experiments. A quantitative and qualitative comparison was also attempted for the third case of Table 2 (2.8 m/s), calculating the total mass deposited during this experiment. Because of the fact that the KCl tray was shifted to a different position along the main axis at each experiment, each case was modeled using a different grid. The mesh resolution was kept almost constant by applying the same edge mesh to all four cases. The mesh domain spanned from a minimum of 362 000 to a maximum of 363 000 tetrahedral P1 elements. The highest resolution was always achieved and kept around the probe; e.g., for case 1 (0.6 m/s), the maximum resolution (smallest cell volume) was of 1.514 611 × 10−2 mm3, achieved on the probe element, whereas the minimum resolution was of 3.017 170 × 102 mm3 at the outlet location. The standard k−ε incompressible turbulence model was selected, while the 3

5632

2

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

Figure 7. SEM image: (a) typical morphology of deposits and (b) corrosion products from the cross-sections of a S235JR ring at different positions.

Figure 8. SEM image: oxide layer thickness of S235JR material along the longitudinal sections at (a) windward side, TS235JR = 400 °C, Tgas = 805 °C, and w = 2.8 m/s and (b) windward side, TS235JR = 500 °C, Tgas = 805 °C, and w = 2.8 m/s. thermodynamic properties of KCl vapor were taken from the NIST tables.45 As a boundary condition for the perfectly insulated elements, a value of −0.1 W/m2 heat loss was applied to these boundary elements to mimic a realistic minimal heat loss through the insulation material.



The results of all surface-temperature-dependent corrosion tests are given in Figure 9. The mean oxide layer thickness of each data point has been calculated from the measurements at eight positions of the cross-section and two positions of the

RESULTS

Experimental Section. All corrosion rings, regardless of the material, showed a qualitatively similar morphology of the deposits and corrosion products. Mostly, the oxide layers were covered with deposits of KCl, as shown on the left-hand side of Figure 7. Under the deposits, a typical multi-layered oxide layer was visible. The outer oxide layers often seemed to be more porous compared to the inner oxide layers, as shown on the right-hand side of Figure 7. A similar appearance of the corrosion products has been observed in the literature.46 The oxide layer thickness of all rings increased considerably with an increased surface temperature. The SEM images in Figure 8 demonstrate one example of an increasing oxide layer thickness with an increased surface temperature after an exposure time of 4 h.

Figure 9. Mean oxide layer thicknesses dependent upon the metal surface temperature at different gas velocities with corresponding standard deviations. 5633

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

Figure 10. SEM image: oxide layer thickness of ARMCO-iron material along the longitudinal sections at (a) windward side, TARMCO = 435 °C, Tgas = 813 °C, and w = 0.6 m/s and (b) windward side, TARMCO = 428 °C, Tgas = 805 °C, and w = 2.8 m/s.

Figure 11. Mean oxide layer thicknesses dependent upon gas velocities with corresponding standard deviations for (a) S235JR and (b) ARMCO iron.

longitudinal sections, as reported before. The error bars give the standard deviation of the mean oxide layer thickness. A bigger standard derivation indicates higher differences in the oxide layer thickness dependent upon the direction of the gas flow. With rising gas velocities from 0.6 to 3.9 m/s, a small increase in the oxide layer thickness of the corrosion rings could be observed. The SEM images in Figure 10 demonstrate a second example of an increasing oxide layer thickness with increasing gas velocities after an exposure time of 4 h. The results of all corrosion tests dependent upon the gas velocity are given in Figure 11. Because of a more precise control of the surface temperature of the S235JR rings, the results of the oxide layer thicknesses are related to 400 and 500 °C. The surface temperature of the ARMCO-iron rings was influenced by different convective heat-transfer coefficients because of different heating profiles and gas velocities. Therefore, the data points are named with >300 and >400 °C surface temperatures. The mean oxide layer thickness of each data point has been calculated from the measurements at eight positions of the cross-section and two positions of the longitudinal sections as well. The error bars give the standard

deviation of the mean oxide layer thickness. A bigger standard derivation indicates higher differences in oxide layer thickness, dependent upon the direction of the gas flow. In Figure 12, a comparison of the oxide layer thicknesses depending upon windward, leeward, and lateral sides of the corrosion rings are shown. At lower gas velocities, the highest corrosion rates for S235JR can be found predominantly at the windward positions. Only when the gas velocity was 3.9 m/s, which equates to the highest tested gas velocity of this study, was the highest corrosion rate of S235JR clearly found on the leeward side. For ARMCO-iron rings, some differences occurred compared to the S235JR rings. In four of six tests, the highest corrosion rates could be observed at the lateral sides of the corrosion rings. Only at a gas velocity of 2.8 m/s and at surface temperatures of 428 and 309 °C was the highest corrosion rate determined at the windward side of the rings. In terms of an error analysis in measuring corrosion rates at a single position of the cross-section of a ring, the two main sources of errors can be attributed first to the relevant temperature measurements and second to the optical measurement of the oxide layer thickness. (i) Thermocouples (type K) with an error in measurement of 0.4% were used for gas 5634

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

Figure 13. Axial temperature profile of the streaming gas in the test rig (°C) for case 1, 0.6 m/s; case 2, 1.7 m/s; case 3, 2.8 m/s; and case 4, 3.9 m/s.

shown. The position of the KCl source was approximately 1 m behind the air inlet of the reaction tube. To keep the temperature around the KCl tray constantly above 800 °C, it was necessary to shift the tray position forward from case to case. The probe location was always 1.9 m behind the air inlet. To ensure a uniform gas temperature of 800 ± 10 °C in front of the position of the probe, the temperature of each single heating element surrounding the tube has been adapted, which returned non-direct case by case scalable temperature profiles in Figure 13. The inlet air rate increases (from cases 1 to 4), and the peak temperature shifts along the axis. Table 3 gives the numerical predictions of the deposition rates as well as the total amount of deposits after 4 h of exposure. From the experimental point of view, it was not possible to weigh right deposit amounts, because of the additional weight gain caused by oxidation of the materials as a result of the corrosion reaction. Consequently, the weight of the deposition for case 3 has been estimated by weighing the rings and calculating the theoretical deposit amount. The estimated weight for case 3 is 130 mg, which is within the range of the numerical prediction. As reported in Table 3 and Figures 15 and 16, case 2 operative conditions produced the highest deposition rate as well as the most homogeneous profile on the probe, in both the half and full KCl amounts. A direct comparison between Figures 15 and 16 shows that correspondent cases behave similarly and that increasing the turbulence around the probe induces the deposition to take place progressively on the downstream side of the probe, where a turbulent vortex recirculation occurs.

Figure 12. Mean oxide layer thicknesses depending upon the measured positions of the corrosion rings for (a) S235JR and (b) ARMCO iron.

temperature and surface temperature measurements of the corrosion rings. The temperature gradient, in consequence of thermal conduction, between the surface of the outer ring and the thermocouple within the ring holder of the probe was neglected, because of the short distance of 2 mm. (ii) The optical resolution limits of the optical microscope and SEM were 100 and 1.5−3.0 nm, respectively. In general, the biggest source of error can be due to the manual corrosion measurement itself, because of the variability in oxide layer thicknesses within the very small image section of a metallographic cut. To consider these random errors, a total of three thickness measurements per measuring position were conducted and then an average thickness was calculated. The resulting random error of this method is ±3 μm per measuring position on average. Numerical Simulation. In this paragraph, the numerical results concerning the eight CFD simulations are presented. As already mentioned before, the coefficient CDR needs to be trimmed. Because only steady Reynolds-averaged Navier− Stokes simulations (RANS) were performed, such a coefficient does not alter the worth of the deposition rate or the total estimated KCl amount but only the velocity of the reaction, namely, the required number of iterations necessary to achieve a convergent solution. Computationally, it has been noticed that a CDR close to unity induces numerical instability: numerical fluctuation intensities were inversely proportional to the main driven velocity. The authors have found that CDR between 0.05 and 0.01 stabilized the solution. Therefore, in the results presented here, a CDR = 0.01 is assumed. In Figures 13 and 14, the axial temperature profile of the reaction tube and the cross-section temperature contour are



DISCUSSION The experimental part of this work shows a correlation of the oxide layer thickness and surface temperature of the corrosion rings. The higher the surface temperatures, the faster the corrosion process and the thicker the oxide layers of the tested materials. Also, the experiments show tendencies that mean corrosion rates, represented by the oxide layer thicknesses, rise with increasing gas velocities. These observations are in accordance with the experiences of plant operators. To 5635

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

Figure 14. Cross-section temperature contours in the evaporation zone and at the probe location of the test rig (K).

probe in dependence of the gas velocity may play an important role. The lower S235JR corrosion ring showed, at lower gas velocities, higher corrosion rates on the windward side. At the highest tested velocity (3.9 m/s), the highest corrosion rate could be found on the leeward side. Instead of corrosion rates, the numerical simulations show a similar distribution of depositions around the lower part of the probe, which represents the S235JR ring. Rising velocities increase turbulence around the probe and induce the deposition to progressively take place on the leeward side of the probe. For the lower ring, the numerical results on the deposited KCl profile are in agreements with the experiments. For the upper ring, which represents the ARMCO-iron material, the highest corrosion rates could mostly be found on the lateral sides, except for case 3 (2.8 m/s), where the corrosion rate was highest on the windward side. In comparison to the experimental results, the calculated deposition rates on the windward and lateral sides are always similar, except for case 4 (3.9 m/s), where the highest deposition could be predicted on the leeward side. The small differences between the simulation and experiment for the upper corrosion ring could be attributed to variations of the surface temperature gradients, which could not be controlled accurately, or small discrepancies of the flow profile caused by turbulences. In summary, the experimental and numerical results show that gas velocities define the distribution of the deposition around the heat-transfer tubes and, therefore, the locally distributed mass of corrosive species, which can cause high corrosion rates. Also, Ots25 describes an accumulation of submicrometer particles on the leeward side at higher flue gas velocities. Furthermore, the mechanical influence on the formation of denser deposits at higher gas velocities could not be excluded as a reason for the higher corrosion rates at higher gas velocities. Further research should be carried out at higher gas velocities and under long-term test conditions to study the deposit buildup in dependence of the gas velocities. It is recommended to monitor additionally salt evaporation rates in future work. The discussed findings represent the deposition behavior in consequence of condensation of KCl vapor on the surface of the corrosion probe. If already homogeneously condensed fine particles would be present, the deposition behavior could show similarities. For instance, very fine KCl particles, formed by nucleation, follow the flow path of the gas and can deposit by diffusion or thermophoresis around the tube.27,28 If bigger inert particles would additionally be transported by the gas flow, the deposits would grow especially on the windward side of the

Table 3. Numerical Results of KCl Deposition and Full Amount of KCl for Two Rings case 1 2 3 4

deposition rate (mg/s) ≈1.1914 ≈1.7181 ≈1.3184 ≈1.2700

× × × ×

10−2 10−2 10−2 10−2

total amount after 4 h (mg) ≈171.56 ≈247.40 ≈189.85 ≈182.88

minimize the complexity of the influences on high-temperature corrosion caused by gas velocities, some influences, for instance, the thermal, the chemical, and the erosion influences, because of higher velocities have been excluded in the experiments. The higher corrosion rates at high gas velocities could be based on two causes: (i) The initial mass of the evaporated KCl is kept constant for all tests by always adding the same amount of KCl into the test rig. At higher gas velocities, the mass flux of air is increased, which results in a decrease of the average KCl concentration in the air flow. Nevertheless, because of the blow dry effect caused by higher gas velocities, the discharge of evaporated KCl from the tablets could be higher at higher gas velocities than at lower gas velocities. Therefore, the residence time of deposited KCl on the probe could be longer for higher gas velocities, which could affect the degree of corrosion attack. Unfortunately, the amount of evaporated KCl could not be monitored during the tests. (ii) On the other hand, the deposition behavior of KCl on the probe surface could be influenced by the gas velocities. First, the amount of KCl deposits on the probe surface may have an effect on the degree of corrosion. Second, the deposits of KCl at higher gas velocities could be denser with smaller pores because of the mechanical influence of the gas flow. According to studies by Hohmann,31 the denser the deposits, the lower the O2 partial pressure on the covered metal surface, the more stable iron chlorides, and the better the corrosion process can proceed. With respect to the last point, the numerical simulations in this work show better insights into the deposition behavior of KCl on cooled surfaces and at different gas velocities. The amounts of deposits for all four cases are similar. An increase of the deposition rates could be predicted with rising gas velocities until 1.7 m/s (case 2), where deposition rates reach a maximum. A further increase in gas velocity decreases the deposition rates in this given range. In comparison to the experimental results, the numerical simulations show that the total amount of corrosive deposits on a probe is not solely responsible for mean corrosion rates of the corrosion rings. In fact, the distribution of KCl deposits around the corrosion 5636

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

Figure 15. Cross-section temperature profiles (K) and KCl deposition rate on the probe (kg m−2 s−1) at addition of half KCl amounts (2 g addition) for (a) case 1, (b) case 2, (c) case 3, and (d) case 4. The figure shows the retraction depth of the corrosion probe in the gas flow loaded with KCl. In the experiments, the surface of the probe is equipped with two corrosion rings, which are not shown in the figure.

Experimental investigations have been carried out to study the effect of varying gas velocities on oxide layer thicknesses in a laboratory test rig. To reduce the complexity of influencing parameters, only the deposit formation behavior has been affected by different gas velocities, while other influences could be excluded. Also, numerical simulations with ANSYS Fluent have been carried out to obtain better insights into the deposition process under fluid dynamic aspects. According to the literature, a correlation of the oxide layer thickness and surface temperature of the corrosion rings could be observed. Also, an increase in the oxide layer thickness with rising gas velocities could be determined in the experiments. With regard to the numerical simulations, the rising mean corrosion rates could not be attributed to the total amount of KCl deposits. Dependent upon gas velocity, the profile of KCl deposition on the corrosion probe surfaces varies and the distribution of KCl could produce peaks of a locally high deposition on the windward or leeward side. Locally high deposition rates can result in higher corrosion rates. Also, the

tube by inertial impaction, or if the kinetic energy of the particles is high enough, they can erode away the formed deposit layer, according to Figure 2. This interaction between the small reactive particles and bigger inert particles forms the basis for further research, especially the influence of the gas velocity under the same test conditions and with the same load of acting ash fractions.



CONCLUSION On the basis of the “Flingern corrosion diagram”, hightemperature chlorine-induced corrosion is influenced by not only the temperature but also the flue gas velocities. Higher corrosion rates can occur in regions of higher flue gas velocities. In general, the influence of the flue gas velocity on corrosion phenomena may be of thermal, chemical, or mechanical origin. Because these three influences take place simultaneously, the complexity of the corrosion process is high, and therefore, the understanding is made difficult. 5637

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

Figure 16. Cross-section temperature profiles (K) and KCl deposition rate on the probe (kg m−2 s−1) at addition of full KCl amounts (4 g addition) for (a) case 1, (b) case 2, (c) case 3, and (d) case 4. The figure shows the retraction depth of the corrosion probe in the gas flow loaded with KCl. In the experiments, the surface of the probe is equipped with two corrosion rings, which are not shown in the figure.



consistency of deposits may be affected by the flue gas velocities. Denser deposits may offer better conditions for hightemperature chlorine-induced corrosion. With regard to the presented findings, it can be summarized that the flue gas velocity in combustion plants influences the deposit growth rate, the consistency, the distribution of corrosive species around the tubes, and most likely, the composition of the deposits. Besides the corrosion aspects, the resulting deposit buildup also affects the heat transfer on the tubes, which requires consideration in the development and layout of new combustion plants. Further investigations are required to study the influence of the flue gas velocity on deposition mechanisms, especially the velocity-dependent interactions of different ash components and their particle fractions on the deposition behavior. The velocity-dependent deposit formation seems to be strongly associated with corrosion.

AUTHOR INFORMATION

Corresponding Author

*Telephone: ++49-89-289-16276. E-mail: [email protected]. de. Notes

The authors declare no competing financial interest.



REFERENCES

(1) Spliethoff, H. Power Generation from Solid Fuels; Springer-Verlag: Berlin, Germany, 2010. (2) Frandsen, F. J. Ash formation, deposition and corrosion when utilizing straw for heat and power production. Publishable Doctoral Thesis, DTU Chemical Engineering, Havdrup, Denmark, 2011. (3) Hupa, M. Proceedings of the Impacts of Fuel Quality on Power Production and Environment Conference; Puchberg, Austria, Sept 23−27, 2012. (4) Reichel, H.-H.; Schirmer, U. Waste incineration plants in the FRG: Construction, materials, investigation on cases of corrosion. Werkst. Korros. 1988, 40, 135−141. 5638

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639

Energy & Fuels

Article

(5) Schroer, C.; Konys, J. Rauchgasseitige Hochtemperatur-Korrosion in Müllverbrennungsanlagen; FZKA: Karlsruhe, Germany, 2002; Vol. 6695. (6) Nielsen, H. P.; Frandsen, F. J.; Dam-Johansen, K.; Baxter, L. L. The implications of chlorine-associated corrosion on the operation of biomass-fired boilers. Prog. Energy Combust. 2000, 26, 283−298. (7) Grabke, H. J.; Reese, E.; Spiegel, M. The effects of chlorides, hydrogen chloride and sulfur dioxide in the oxidation of steels below deposits. Corros. Sci. 1995, 37, 1023−1043. (8) Skrifvars, B.-J.; Backman, R.; Hupa, M.; Salmenoja, K.; Vakkilainen, E. Corrosion of superheater steel materials under alkali salt deposits. Part 1: The effect of salt deposit composition and temperature. Combust. Sci. 2008, 50, 1274−1282. (9) Skrifvars, B.-J.; Westén-Karlsson, M.; Hupa, M.; Salmenoja, K. Corrosion of superheater steel materials under alkali salt deposits. Part 2: SEM analyses of different steel materials. Combust. Sci. 2010, 52, 1011−1019. (10) Tillman, D. A.; Duong, D.; Miller, B. Chlorine in solid fuels fired in pulverized fuel boilersSources, forms, reactions, and consequences: A literature review. Energy Fuels 2009, 23, 3379−3391. (11) Alonso Herranz, E. Short-term measurement of high-temperature chlorine-induced corrosion and agglomeration during biomass and waste combustion. Ph.D Dissertation, Institute of Energy Systems, Munich, Germany, 2012. (12) Montgomery, M.; Frandsen, F. J.; Karlson, A.; Larsen, O. H. Proceedings of the International Conference on Effects of Coal Quality on Power Plant Management: Ash Problems, Management and Solution; Park City, UT, May 8−11, 2000. (13) Montgomery, M.; Jensen, S. A.; Borg, U.; Biede, O.; Vilhelmsen, T. Experiences with high temperature corrosion at straw-firing power plants in Denmark. Mater. Corros. 2011, 62, 593−605. (14) Kümmel, J. Feuerungs- Verbrennungs-, Vergasungstechniken. Proceedings of the VDI Bildungswerk; Düsseldorf, Germany, Feb 21−22, 1994. (15) Warnecke, R. Einfluss von Strö mung und chemischen Reaktionen im rauchgasseitigen Belag auf Korrosion an Ü berhitzerrohren in Müllverbrennungsanlagen. VGB PowerTech 2004, 9, 52−59. (16) Bürgel, R.; Maier, H.-J.; Niendorf, T. Handbook Hochtemperaturwerkstofftechnik, 4th ed.; Vieweg + Teubner Verlag: Berlin, Germany, 2011. (17) Hendriksen, N.; Montgomery, M.; Larsen, O. H. Korrosion in energieerzeugenden Anlagen. Proceedings of the VDI Tagung; Würzburg, Germany, Sept 18−19, 2002; pp 111−131. (18) Warnecke, R. Korrosion unter Berücksichtigung von Strömungsgeschwindigkeit und Reaktionsenthalpie. In Rauchgasseitige Dampferzeugerkorrosion; Born, M., Ed.; Verlag Saxonia Standortentwicklungs- und -verwaltungsgesellschaft mbH: Freiberg, Germany, 2003; pp 57−78. (19) Maisch, S.; Warnecke, R.; Waldmann, B., Haider, F.; Horn, S. Feuerung und KesselBeläge und KorrosionIn Großfeuerungsanlagen. Proceedings of the VDI Fachkonferenz; Stuttgart, Germany, June 23−24, 2009. (20) Maisch, S.; Waldmann, B.; Warnecke, R.; Haider, F.; Horn, S. Feuerung und KesselBeläge und KorrosionIn Großfeuerungsanlagen. Proceedings of the VDI Fachkonferenz; Frankfurt, Germany, June 22−23, 2010. (21) Maisch, S. Identifikation und Quantifizierung von korrosionsrelevanten Parametern in Müllverbrennungsanlagen mittels Charakterisierung der deponierten Partikel und elektrochemischer OnlineMessungen. Ph.D. Dissertation, Lehrstuhl für Experimentalphysik II, Augsburg, Germany, 2012. (22) Nicholls, J. R.; Saunders, S. R. J. Comparison of hot-salt corrosion behaviour of superalloys in high and low velocity burner rigs. High Temp. Technol. 1989, 7 (4), 193−201. (23) Polifke, W.; Kopitz, J. Wärmeübertragung, 2nd ed.; Pearson: München, Germany, 2009. (24) Cussler, E. L. Diffusion: Mass Transfer in Fluid Systems, 3rd ed.; Cambridge University Press: New York, 2009.

(25) Ots, A. Mechanism of ash deposit formation, corrosion and sulphur by burning calcium and chlorine containing fuels. VGB PowerTech 2001, 10, 114−120. (26) Walsh, P. M.; Sayre, A. N.; Loehden, D. O.; Monroe, L. S.; Beer, J. M.; Sarofim, A. F. Deposition of bituminous coal ash on an isolated heat exchanger tube: effects of coal properties on deposit growth. Prog. Energy Combust. 1990, 16, 327−334. (27) Zbogar, A.; Frandsen, F. J.; Jensen, P. A.; Glarborg, P. Shedding of ash deposits. Prog. Energy Combust. 2009, 35, 31−56. (28) Frandsen, J. F. Utilizing biomass and waste for power productionA decade of contributing to the understanding, interpretation, and analysis of deposits and corrosion products. Fuel 2005, 84, 1277−1294. (29) Robinson, A. L.; Buckley, S. G.; Baxter, L. L. Experimental measurements of the thermal conductivity of ash deposits: Part 2. Effects of sintering and deposit microstructure. Energy Fuels 2001, 15, 75−84. (30) Zbogar, A.; Jensen, P. A.; Frandsen, F. J.; Hansen, J.; Glarborg, P. Experimental investigation of ash deposit shedding in a straw-fired boiler. Energy Fuels 2006, 20, 512−519. (31) Hohmann U. Kinetische Betrachtungen zur chlorinduzierten Hochtemperaturkorrosion. In Rauchgasseitige Dampferzeugerkorrosion; Born, M., Ed.; Verlag Saxonia Standortentwicklungs- und -verwaltungsgesellschaft mbH: Freiberg, Germany, 2003; pp 79−100. (32) Kastner, W.; Riedle, K.; Tratz, H. Experimentelle Untersuchungen zum Materialabtrag durch Erosionskorrosion. VGB Kraftwerkstech. 1984, 64 (5), 452−464. (33) Kastner, W.; Hofmann, P.; Nopper, H. Erosionskorrosion in KraftwerksanlagenEntscheidungshilfe für Maßnahmen zur Schadensvermeidung. VGB Kraftwerkstech. 1990, 70 (11), 939−948. (34) Raask, E. Mineral Impurities in Coal Combustion; Hemisphere Publishing Corporation: Washington, D.C., 1985. (35) Raask, E. Erosion Wear in Coal Utility Boilers; Hemisphere Publishing Corporation: Washington, D.C., 1988. (36) Latella, B. A.; O’Connor, B. H. Effect of porosity on the erosive wear of liquid-phase-sintered alumina ceramics. J. Am. Ceram. Soc. 1999, 82 (8), 2145−2149. (37) Sundararajan, G.; Roy, M. Solid particle erosion behavior of metallic materials at room and elevated temperatures. Tribol. Int. 1997, 30 (5), 339−359. (38) Oka, Y. I.; Matsumura, M.; Kawabata, T. Relationship between surface hardness and erosion damage caused by solid particle impact. Wear 1993, 162−164, 688−695. (39) Oka, Y. I.; Ohnogy, H.; Hosokawa, T.; Matsumura, M. The impact angle dependence of erosion damage caused by solid particle impact. Wear 1997, 203−204, 573−579. (40) Oka, Y. I.; Yoshida, T. Practical estimation of erosion damage caused by solid particle impact. Part 1: Effect of impact parameters on a predictive equation. Wear 2005, 259, 95−101. (41) Oka, Y. I.; Yoshida, T. Practical estimation of erosion damage caused by solid particle impact. Part 2: Mechanical properties of materials directly associated with erosion damage. Wear 2005, 259, 102−109. (42) Tylczak, J. H. Erosion−corrosion of iron and nickel alloys at elevated temperature in a combustion gas environment. Wear 1995, 186−187, 284−290. (43) Levy, A. V. Corrosion and particle erosion at high temperature. Proceedings of the 118th Annual Meeting of the Minerals, Metals & Materials Society; Las Vegas, NV, Feb 27−March 3, 1989. (44) Stephenson, D. J.; Nicholls, J. R. Modelling the influence of surface oxidation on high temperature erosion. Wear 1995, 186−187, 284−290. (45) National Institute of Standards and Technology (NIST). Material Measurement Laboratory: Potassium Chloride; NIST: Gaithersburg, MD, 2011; http://webbook.nist.gov/cgi/cbook.cgi?ID= C7447407&Mask=1A8F (assessed June 12, 2012). (46) Mayer, P.; Westwood, H. J.; Manolescu, A. V. Corrosion related problems in fossil fired boilers. J. Mater. Energy Syst. 1980, 2, 55−64.

5639

dx.doi.org/10.1021/ef400438k | Energy Fuels 2013, 27, 5628−5639