uncorrected proof

Applied Energy xxx (2018) xxx-xxx

Contents lists available at ScienceDirect

Applied Energy

PR OO F

journal homepage: www.elsevier.com

Investigating the flue-wall deformation effects on performance characteristics of an open-top aluminum anode baking furnace Mouna Zaidania⁠ , Abdul Raouf Tajika⁠ , Zahid Ahmed Qureshia⁠ , Tariq Shamima⁠ ,⁠ b⁠ , Rashid K. Abu Al-Ruba⁠ ,⁠ ⁎⁠ a b

Department of Mechanical Engineering, Masdar Institute, Khalifa University of Science and Technology, Abu Dhabi, P.O. Box 127788, United Arab Emirates Mechanical Engineering Program, University of Michigan-Flint, Flint, MI 48502, USA

ABSTRACT

Keywords: Anode baking furnace Deformation Aging Baking uniformity Flue-wall CFD modelling

The carbon anode baking is typically the bottleneck in the production of anodes for the aluminum industry. The challenge is to produce high quality baked anodes while keeping the energy consumption, environmental emissions, and cost to a minimum. Anode baking homogeneity is an important consideration in the design and operation of the anode baking furnaces. The flue-walls, into which the firing takes place, are the heart of the furnace. The flue-walls must be well designed and well maintained in order to be able to regulate them properly. However, during the service life of a baking furnace, flue-walls deform which affect the baking uniformity and consequently result in over-consumption energy and reduction in carbon anode quality. Studying the effects of flue-wall deformation by plant tests is highly challenging and expensive. Hence, this study aims at investigating this phenomenon by developing a three-dimensional (3D) model which includes many physical phenomena and parameters that play vital roles in the baking process. It was observed that indeed the flue-wall deflection has a significant impact on heat transfer characteristics of the anode baking process and furnace energy consumption. In specific, the flue-wall deformation results in an increase in temperature gradients within the anode pack such that the differences between anode pack minimum, average and maximum temperatures lead to overbaking or underbaking of anodes. In fact, this non-uniform baking gives rise to the evolution of non-homogeneous carbon anodes material properties, which is the main reason for excess energy consumption and various instabilities in the aluminum reduction cell. The insights obtained in the present study can be employed in modifying the furnace geometrical and operational parameters with the deformed flue-walls.

UN CO RR EC TE D

ARTICLE INFO

Nomenclature

identity matrix specific heat (J/kg·°C) hydraulic diameter inside the flue (m) turbulent kinetic energy (m2⁠ /s2⁠ ) temperature (°C) the static pressure inside the flue (Pa) time (s) turbulent dissipation (m2⁠ /s3⁠ ) coefficient of heat transfer (W/m2⁠ ·°C) heat loss to the atmosphere from bottom and top of the pit (W/m3⁠ ) C constant x, y, and z coordinate axes (m) CP⁠ Dh⁠ k T p t ε h Qs

⁎

Greek symbols ρ density (kg/m3⁠ ) Δ difference operator (·) μ dynamic viscosity (Pa·s) μT effective viscosity (Pa·s) ω emissivity (·) σ Stefan-Boltzmann constant = 5.67* × 10−⁠ 8 (W/m2⁠ ·°C4⁠ ) η the absorptivity of the flue gas (·) λ thermal conductivity (W/m·°C) αT⁠ thermal expansion coefficient (1/°C) Subscripts/superscripts ∞ ambient (·) C convective g gas

Corresponding author. Email address: [email protected] (R.K. Abu Al-Rub)

https://doi.org/10.1016/j.apenergy.2018.09.197 Received 26 December 2017; Received in revised form 18 September 2018; Accepted 23 September 2018 Available online xxx 0306-2619/ © 2018.


UN CO RR EC TE D

PR OO F

M. Zaidani et al.

Fig. 1. A general overview of a typical anode baking ring furnace [20].

Fig. 2. General view of (a) the flue-wall deflection [21] and (b) heavily slagged flue-wall covered with metallurgical coke [22].

R S T w

1. Introduction

radiative solids (anode, packing coke and flue-wall) transpose operator (·) wall

The anode baking is a very expensive step in the aluminum production. Fuel consumption and refractory maintenance contribute significantly towards the total cost of anode production. The anode baking is carried out in an anode ring furnace that is composed of a number of preheating, firing, and cooling sections. Anodes are placed between the flue-walls into a pit. The fire group equipment (the exhaust manifold, burner ramps and the cooling covers) moves and the anodes remain stationary. Fig. 1 shows a typical view of an open top furnace and a three-dimensional (3D) view of a section in the ring furnace. Usually, each furnace has two to four fire groups, where each fire group is com

Dimensionless numbers Re Reynolds Number (·) Pr Prandtl number (·)

Abbreviations CFD Computational Fluid Dynamics MP Multi-Physics

2


UN CO RR EC TE D

PR OO F

M. Zaidani et al.

Fig. 3. Schematic representation of the solid and fluid domains.

Fig. 4. The flue and pit sub-models coupling.

posed of typically 10–16 sections: three to four anode preheating sections, three to four firing sections and four to nine cooling sections. Anodes are heated at a certain rate from room temperature to about 1100 °C and then cooled slowly. The process starts by placing green anodes in the pit at the ambient temperature. The anodes temperature will reach up to 600 °C by the end of preheating and up to 1100 °C in the fire sections. Finally, the baked anodes cool down in the cooling sections, heating the flue gas before it enters the fire sections. The entire baking process takes about 240–360 h. The green anodes are baked

in furnace sections with a certain number of pits (max 8 pits/section). The incoming combustion air, pushed in by the first cooling ramp is preheated as it passes through flue sections of adjacent flues in the cooling zone. The pits are separated by the heated flue-walls. The heat generated by combustion is transferred through these flue-walls that are kept under negative pressure to retain and draw the fumes. The focus of the current study is on investigating the effect of deflected flue-walls on the temperature uniformity within carbon anodes through the complicated baking process. To the authors’ best-knowledge, detailed

3


UN CO RR EC TE D

PR OO F

M. Zaidani et al.

Fig. 5. Adopted boundary conditions for the flue and pit sub-models.

pots will greatly outweigh the gains of the energy saved. Poor furnace design, operation and/or maintenance, resulting in a higher than optimum waste gas quantity may increase specific energy consumption by up to 1 GJ/tons of anodes. Due to the huge size of the baking furnace and its very large time constant (in the order of weeks), it is very challenging to conduct physical experiments. Plant trials on such furnaces are quite costly and may cause the loss of production. In general, it is difficult to make detailed measurements around the furnace due to limited accessibility or costs involved. The necessity in investigating the effect of different operational and geometrical parameters on anode baking furnace energy consumption has heightened the need for developing computational tools with varying levels of complexities. The active work on mathematical modeling of the anode baking furnaces started at the beginning of the 1980s with relatively simple approaches. With the developments in numerical modeling and computation capacity (memory and speed), more sophisticated models have appeared [1–3]. Some of these models aim at the furnace operation, they are one or two dimensional and simulate the dynamics of the process. These models could be resolved with fewer details based on simpler approaches, by eliminating such details as 3D velocity distribution or by neglecting gradients in certain directions and reducing the number of dimensions in the solution of equations. Later on, many models of varying complexity, but similar in nature to these early works have been published and are referred to as process models that assist in optimizing the furnace operation parameters for practical furnace optimization and control. However, process models are highly simplified and may not be of great use in investigating anode baking furnaces design characteristics. Other more detailed models are called anode baking furnace design/ CFD modeling. El Ghaoui et al. [4] developed a 3D numerical model for furnace design improvement and demonstrated the use of baffle-less flue-wall design in anode baking furnace. They have shown that baffle-less flue-wall results in a better baking homogeneity and at the same time a higher thermal efficiency. Johnson et al [5] presented a simplified baking furnace model for improving the flue-wall design by solving the conjugate heat transfer problem in the anode baking furnace. They have shown that the problem can be simplified by considering

Fig. 6. The grid independence test.

investigation of the effect of flue-wall deflection, which increases with time and furnace operation, is absent from the published literature. The anode quality has a significant impact on the aluminum production cost. Stringent anode quality and process control are required to improve the anode performance in the pots and hence reduce the smelter costs. The theoretical consumption is given by the stoichiometric equation according to the Faraday law is that to produce 1 ton of aluminum, 334 kg of carbon are needed, assuming a current efficiency of 100%. However, in practice due to excess consumption, carbon net consumption reaches more than 400 kg per ton of aluminum. Anode baking is an energy-intensive process and; therefore, continually reducing the specific energy consumption as far as possible is more important today than ever before. Great progress has been made in this respect in the last twenty years. Today, a modern anode baking furnace has an energy consumption of 1.8 GJ/ton of the baked anodes. In the mid-1990s the figure was about 2.5 GJ/ton, or over a third more. The furnace specific energy consumption must never be optimized at cost of the resulting anode quality, as the inferior anode performance in the

4


PR OO F

M. Zaidani et al.

UN CO RR EC TE D

Fig. 7. Schematic representation of the flue-wall deflection modes.

Fig. 8. (a) The flow pattern (velocity in m/s) on the symmetry plane in a flue (the middle of the flue cavity) and (b) the pressure (Pa) distribution inside the flue cavity.

the processes in the flue as stationary and solve the transient thermal conductivity equation only in the pit. They confirmed that the temperature distribution in the pit filled with anodes is largely dependent on temperature distribution in flue gases and is defined by the result obtained after solving the problem of conjugate heat transfer from gas to the anodes in the pit via the flue-wall. Fan et al. [6] have studied the thermal stress and strain distributions of a wall-fired furnace, they developed a steady state unified numerical approach which integrates the fluid mechanics-based combustion process and the solid mechanics-based stress analysis to obtain steady-state temperature profile in the furnace and the thermal stress and strain fields on the furnace wall. They concluded that the back wall (and its edges) exhibits extremely high temperature, thermal expansion, and stress concentration that may cause thermal fatigue or cracking. Baiteche et al. [7] developed a 3D mathematical model to simulate the different stages of the baking process in the furnace, which was used to determine the impact of the deformation of a flue-wall on one side of a pit while keeping the flue-wall on the other side unchanged (straight). By comparing the temperature profiles on a line in the pit transverse direction for straight and deformed flue-walls, it was observed that after flue-wall deformation, the temperature profile is no more symmetric which indicates a non-uniform baking process. The model developed was used for a deformed

flue-wall, comparing the flow field in the flue cavity between a straight and a deformed flue-wall, it was observed the velocities on one side are higher than the other side creating differences in the baking of the anodes in two adjacent pits. Grégoire and Gosselin [8] compared three different combustion models to simulate the anode baking furnace. They concluded that the hot air jet approximation can be a good alternative in case the combustion models are not available. Tajik et al. [9–11] investigated the effects of flue-wall design modifications on the anode baking furnace and proposed optimized flue-wall designs for both new and old furnaces. In the anode baking furnaces, during the firing sections, the heat provided by the flame jets are responsible for the heating (baking) process. The research has been conducted numerically and experimentally on investigating the effect of various operation parameters on flame jets heating performance and homogeneity [12–17]. In the case of anode baking, homogeneity is an important consideration in the design and operation of the anode baking furnaces. The flue-walls, into which the firing takes place, are the heart of the furnace. The flue-walls must be well designed and well maintained in order to be able to regulate them properly. As shown in Fig. 2, during the service life of a baking furnace, flue-walls deform which affects the baking uniformity and consequently results in over-consumption of energy and reduction in

5


PR OO F

M. Zaidani et al.

UN CO RR EC TE D

Fig. 9. Temperature (°C) mapping distribution at (a) flue cavity and (b) the flue-wall-cavity interface.

Fig. 10. Anode map temperature (°C) at (a) the middle plane and (b) the packing coke interface.

carbon anode quality. Studying the effects of flue-wall deformation by plant tests is highly challenging and expensive. To the best knowledge of the authors, a detailed numerical model on the flue-wall deformation modes and their effect on the baked anode quality has not been published to date. Most of the research work in this area has been focused on straight flue-wall without considering the effect of flue-wall deformation due to furnace aging. Hence, in this work, a fully coupled thermo-hydro-mechanical simulations were done by the finite element multi-physics commercial software COMSOL 5.1. Such a 3D multi-physics modeling can be used as a powerful method for predicting the effect of flue-wall deformation on anode temperature distribution and homogeneity. The present study also gives useful insights about temperature distribution adjustment as a function of the flue-wall and furnace design. The anode temperature distributions for straight flue-wall (undeformed) and deformed flue-wall (10–99 mm of deflection) are compared. A parametric study has also been conducted to investigate the effects of varying velocity and temperature of both hot air-jet and

mainstream flow on the anode temperature for straight and deformed flue-walls. Moreover, the effect of flue-wall deformation on the anode baking homogeneity is also investigated. 2. Numerical methodology A 3D CAD model was developed comprising a typical baking furnace firing section between the centerline of a flue-wall cavity and the centerline of a pit. In the present model, for simplicity tie bricks are not considered. In the pit, 15 anodes (each of 1600 × 600 × 800 mm size) are placed. There are two domains; a solid domain composed of flue-wall, packing coke and anodes, and a fluid domain for the gas flow (see Fig. 3). The thickness of the solid domain material layers, flue-wall, packing coke and anodes in the direction of the pit length is 100 mm, 100 mm and 300 mm, respectively. Symmetry plane is assumed at the centerline of the flue and the pit. This model is applied to the firing section of the furnace in order to explore the effect of the

6


PR OO F

M. Zaidani et al.

UN CO RR EC TE D

Fig. 11. Downstream and upstream thermocouples arrangement inside the pit (at EGA).

anode baking quality. Different physical phenomena occur in the two parts of the furnace (solids and gas). Hence, specific equations are solved in each part of the model through a virtual numerical interface. Thus, the global model of a baking furnace is divided into two sub-models: gas and pit (brick wall, packing coke, and anodes). These sub-models are developed separately and then coupled through an interface at the flue-wall surface on the flue (gas) side (see Fig. 4). This modular approach makes it easier to develop and modify each part. Also, if needed, each sub-model may be used exclusively for testing case scenarios in one part of the furnace (such as the flow distribution in the flue for a given geometry). On the gas (flue) part, the flow, heat and mass transfer equations are solved. On the solid part, heat transfer by conduction is solved, considering the thermal conductivity of various solids (flue-wall, packing coke and the anodes). The model is a steady-state and simulates one heating section. In an anode baking horizontal flue ring furnace, the temperature distribution is one of the key factors influencing the quality of baked anodes and is closely related to the gas flow. For the gas flow distribution in the flue cavity, Navier-Stokes equations with the eddy viscosity turbulence model k - ε are adopted. The description of a turbulent incompressible gas flow is governed as follow:

Fig. 12. Validation with the experimental measurements (at EGA).

flue-wall deformation on the anode baking quality and temperature distribution homogeneity within this critical section. To reduce the computational time, the combustion process is simulated using a simplified hot air jet model, which is a widely used simplified approach in which the combustion process and its thermal energy released are modeled by considering an inlet of hot air [8]. In order to inject an amount of energy at the burners inlets equivalent to that injected with natural gas, the air is injected at the same temperature as the maximum flame temperature. The developed computational model will assist in understanding the effect of flue-wall deformation on the anode baking quality and temperature uniformity. 2.1. Multi-physics mathematical formulation

(1)

(2)

where the viscosity μ depends only on the physical properties of the fluid, while the μT is the turbulent eddy which is supposed to emulate the effect of unresolved velocity fluctuations. The standard k - ε model is based on the assumption that , where k is the turbulent kinetic energy and ε is the turbulent dissipation rate. Hence, the above PDE system is to be completed by two additional convection-diffusion-reaction equations for computation of k and ε, such that:

The flue-wall deformation has a two-way link with temperature and flue gas flow. A change in temperature during baking, for instance in the firing section, would result in the development of thermal stresses in the brick wall that may cause undesired deformations or cracks in the flue-wall and consequently alter its microscopic structure and thermal properties. This will not only progressively change the strength of the flue-wall but will also cause changes in the flow regimes and the

7


PR OO F

M. Zaidani et al.

UN CO RR EC TE D

Fig. 13. The flow pattern (velocity in m/s) on the symmetry plane in a flue (the middle of the flue) as a function of the flue-wall deformation.

Fig. 14. The pressure (Pa) distribution inside the flue cavity as a function of the flue-wall deformation.

by conduction from the flue-wall surface to the bricks, and to the packing coke, then finally to the anodes. Heat transfer through the solid is represented by the transient heat conduction equation:

(3)

(4)

where

(5)

where Ts is the solid (anodes, packing coke and flue-wall) temperature, λs is the thermal conductivity of the solids, and Qs represents the thermal loss to the atmosphere by coke and the foundation. At the boundaries, where solids and flue gases are in contact, heat is exchanged by radiation and convection following the relationship:

and ε are responsible for production and

dissipation of turbulent kinetic energy, respectively. The parameters σk , σε, pk, Cε1 and Cε2 are constants. Heat from the gas in the flue cavity is transferred to the flue-wall surface by convection and radiation, then

(6)

8


UN CO RR EC TE D

PR OO F

M. Zaidani et al.

Fig. 15. Anode map temperature (°C) (a) at the packing coke interface and (b) the anode middle plane as a function of the flue-wall deflection (temperature dependent material properties).

At the brick wall surface, where solids and flue gases are in contact, heat is exchanged following the relationship:

transfer coefficient is given by [18]:

(7)

where q′ is the heat flux through the fluid-solid interface; hC is the convection heat transfer coefficient between the gas and flue-wall; hR is the radiation heat transfer coefficient between the gas and flue-wall; Twis the flue-wall temperature at the boundary; Tg is the gas temperature. The convection heat transfer coefficient hC is determined by the Dittus-Boelter correlation [18]:

(9)

where σ is the Stephan-Boltzmann constant; ωg and ηg are the emissivity and the absorptivity of the mixed gas, respectively. 2.2. Boundary conditions and grid generation As shown in Fig. 7, at symmetry planes, the applied boundary conditions are:

(8)

• Heat transfer symmetric boundary conditions; heat flux equals to zero:

where λg is the thermal conductivity of gas; Dh is the representative hydraulic diameter inside the flue, Pr is Prandtl number of gas; γ is an exponent, Re is the Reynolds number of the flow. The radiation heat

(10) 9


PR OO F

M. Zaidani et al.

Fig. 16. The temperature variation as a function of flue-wall deflection of (a) the anode volume average, the anode middle plane average and the anode-packing coke interface average, and (b) the anode’s volume average, minimum and maximum temperatures.

tion of temperature, based on experimental data. Using linear extrapolation, the values of anodes and packing coke thermal conductivity are extracted corresponding to a temperature value below 20 °C and above 1000 °C for the packing coke and 1200 °C for the anodes. In case 2, thermal conductivity and specific heat for the anodes and the packing coke are assumed to remain constant by considering the experimental average value for the temperature above 500 °C corresponding to the temperature in the firing section [19].

UN CO RR EC TE D

• Fluid flow symmetric boundary conditions; no flow normal to the boundary: (11) (12)

3. Results and discussions

where n is the unit outward normal vector, Ts is the solid temperature and ug is the gas velocity. Viscosity is assumed to remain constant at values corresponding to air at the inlet. Thermal conductivity and specific heat are assumed to be constant in one case and then as a function of temperature in another case. A hot jet air approach is considered to include the energy generated by the natural gas combustion in order to reduce the computational burden of simulating anode baking process and is achieved simply by replacing the burner by an inlet of hot air at −20 Pa. The hot air is injected at the same temperature as the maximum flame temperature. Velocity inlet B.C. is considered for both mainstream flow as well as hot air jet. Fig. 5 provides information about boundary conditions (B.C.) used in the present study. Several grid configurations are studied to make sure that the obtained results are grid/mesh-independent. All grids were generated in COMSOL 5.1 with the use of a local grid refinement towards inlets (“Fluid dynamics” built-in mesh generation option in COMSOL). Grid independence check is carried out such that the difference in anode average temperature as a function of a number of mesh elements is presented in Fig. 6. It can be seen that after increasing the number of elements from 30,000 to 410,000 the difference in anode average temperature is found to be less than 1% for gird size above 150,000 elements. Hence, for the present study, the selected mesh is constituted of 179,669 grid elements. In practice, the generated grid is done at “fine” level (in terms of COMSOL internal grid definitions) such that the average element quality being on the order of 0.7. Finite element-based software COMSOL 5.1 Multi-physics model is used to analyze the turbulent convection heat transfer. To solve the governing equations, segregated solvers are used to calculating the fluid flow (velocity and pressure) and heat transfer (temperature) variables. Generalized minimal residual (GMRES) iterative method solver is used to calculate the parameters with a tolerance of 10−⁠ 3 and a factor in error estimate of 20. As a preconditioner, Geometric Multigrid solver is used with PARDISO (Parallel Sparse Direct Linear Solver) fine solver. In this study, two cases are presented. In case 1, the thermal conductivity and specific heat of the solids (anodes and packing coke) are a func

Firstly, the results for a straight (un-deformed) flue-wall, 0-mm deflection, using the temperature dependent material properties are presented. It can be seen that the fluid flow streamlines inside the flue cavity depend on the baffles position inside the flue-wall. The baffles create low-velocity zones behind them due to the recirculation of flow. The flow distribution through the cavity thickness shows that there is a poor flow distribution zone behind the baffles and high flow distribution zone, around 2 m/s and 5 m/s, respectively. Fig. 8b presents the resulting pressure field inside the flue cavity. The pressure drop across the flue is also important for the calculation of blower power requirement to circulate the gas. The resulting pressure drop value corresponding to the flue-wall geometry is 25 Pa. Fig. 9 presents the temperature distribution within the flue cavity and at the flue-wall-flue cavity interface. Fig. 9a shows the hot air jet extends inside the flue channel. As can be clearly seen from the temperature mapping, the maximum temperature in the flue cavity is between 1500 and 1800 °C near to the burners inlet. The minimum temperature in the flue cavity is around 677 °C, the low-temperature zones are between 677 and 850 °C. These zones are located near to the air inlet and correspond to poor flow distribution zone (Fig. 8a). The medium zone temperature is between 850 and 1050 °C and corresponds to the high flow distribution area near the outlet. The heat transfer towards the flue-wall has smoothed the temperature distribution on the wall-flue cavity interface (Fig. 9b). The hotspot near the burners in the flue cavity has smoothed from 1600 °C in the flue cavity to 1300 °C at the wall–flue cavity interface. The flue-wall temperature distribution is more homogenous and is ranged between 900 °C and 1300 °C (though the difference between the flue-wall maximum and minimum temperature is still significant). Fig. 10 shows the temperature contours at the anodes middle-plane (Fig. 10a) and at the anodes-packing coke interface (Fig. 10b). As expected from the temperature contour at the wall-flue cavity interface (Fig. 9b), the hottest zone in the anode-packing coke interface is located at the top of the pit near to the outlet corresponding to the high temperature zone

10


UN CO RR EC TE D

PR OO F

M. Zaidani et al.

Fig. 17. Anode map temperature (°C) at the packing coke interface (a) and the anode middle plane (b) as a function of the flue-wall deflection (constant material properties).

Fig. 18. The temperature variation as a function of flue-wall deflection of (a) the anode volume average, the anode middle plane average and the anode-packing coke interface average, and (b) the anode’s volume average, minimum and maximum temperatures (constant material properties).

11

M. Zaidani et al.


Table 1 Summary of the anode temperature values for material properties that vary with temperature (values between brackets shows the anode temperature for constant material properties).

Maximum Minimum Maximum - Minimum Volume average Average at packing coke-anode interface Average at middle plane

Case 1 (0 mm)

Case 2 (10 mm)

Case 3 (30 mm)

Case 4 (60 mm)

Case 5 (90 mm)

Case 6 (99 mm)

992 (976) 717 (714) 276 (262) 891 (887) 892 (889) 890 (886)

996 (978) 721 (716) 276 (262) 892 (888) 893 (889) 890 (887)

1004 (988) 722 (719) 282 (269) 892 (889) 894 (891) 891 (888)

1021 (1007) 728 (726) 293 (281) 894 (891) 895 (893) 893 (890)

1053 (1032) 735 (731) 317 (301) 897 (895) 898 (897) 895 (894)

1057 (1047) 738 (736) 319 (311) 899 (897) 900 (898) 898 (896)

PR OO F

Anode temperature distribution (°C)

UN CO RR EC TE D

pack. The measurements are carried out during the flue-gas soaking period (34 h), in 10 min’ intervals. The experimental measurements are compared with the numerical results presented in Fig. 12. Due to a confidential agreement with the Emirates Global Aluminum (EGA), only the normalized temperature is depicted. It can be clearly seen that the numerical model shows a good concordance with the experimental data. The lowest temperature is collected in the top of the anode pack (under-baked area) for both experimental and numerical data, corresponding to US6 and DS6 thermocouples. In the case of the numerical results, the temperatures are slightly higher because of considering a hot air jet instead of fuel injection and a complete combustion assumption. 3.1. The effects of flue-wall deflection As emphasized in the introduction section, in an actual furnace, after a certain number of fire cycles, the flue-wall bricks undergo a deformation leading to an increase or decrease in pit width. Thus, the pit sub-model is used to investigate the effect of the change in pit width on the anode temperature distribution. In order to study this effect, the results are compared with the results of the straight un-deformed flue-wall. The pit width was reduced as a function of the deflection level (10, 30, 60, 90 and 99 mm). The comparison between the results for the straight and deformed flue-walls are presented in Figs. 13–16 or temperature-dependent material properties and Fig. 17 for constant material properties. Generally, the solid temperatures were found to be higher than those of a straight flue-wall, which present a good concordance with the observations for a reduced pit size presented in the literature. This is mainly attributed to packing coke thickness change. Reducing (or increasing) the pit width includes increasing (or decreasing) of the packing coke thickness that constitutes a thermal barrier due to its lower diffusivity compared to anode blocks. To avoid this problem, and to assure the uniformity of the anode temperature as much as possible, the deformed flue-wall must be straightened or replaced regularly. The type of material of the refractory bricks also affects the anode baking. Figs. 13 and 14 present the flow patterns at the flue cavity symmetry plane, and the pressure distribution inside the flue cavity, respectively, as a function of the flue-wall deformation. Compared to the straight flue-wall, very slight variation in the flow and the pressure distribution can be observed as a function of the flue-wall deformation. Similarly, very little variation is observed for the temperature distribution in the flue cavity and at the wall-flue cavity interface (see Fig. 18. Fig. 15 shows the temperature distribution for the case of temperature-dependent material properties at the anode-packing coke interface (Fig. 15a) and at the anode middle plane (Fig. 15b) as a function of the flue-wall deflection. It can be clearly seen that the temperatures of the anode-packing coke and the anode middle plane increase with increasing flue-wall deflection through the development of clearer hotspots

Fig. 19. The anode volume average temperature as a function of the flue-wall deflection for constant and variable material properties as a function of temperature.

in the flue-wall (Fig. 9b) and to the high flow distribution zone (Fig. 8a). The maximum temperature in the hotspot is around 1025 °C. The low-temperature zone is located near the inlet and the coldest zone near the anode top surface, the low-temperature zone is ranged between 825 °C and 716 °C. Moving from the packing coke-anode interface (Fig. 10b) to the anode middle-plane (Fig. 10a), the heat diffusion in the solids has smoothed further the hot and cold spots present in the flue-wall and the packing coke; as expected, the hot spot in the anode middle-plane is located near to the outlet and the temperature is around 925 °C. With the exception of the hotspot in the anodes located in a very narrow zone near to the second burner position, the rest of the anode temperature distribution is homogenous. The anode volume average temperature is around 891 °C. More details concerning the anode temperature distribution for a straight flue-wall and as a function of the flue-wall deflection are presented in the next section. The computational model has been validated versus an experimental data based on comparing the anode temperature distribution. In doing so, the numerical model used in the present study is initially validated for a straight flue-wall using a temperature dependent material property with the experimental work carried out considering a soaking temperature of 1170 °C. The results are validated with the anode temperature collected in the pit. Anodes temperature distribution varies between downstream and upstream. Hence, the thermocouples are positioned in the downstream and upstream position inside the pit at different heights. As depicted in Fig. 11, six thermocouples are placed in the anode pack downstream (DS1, DS2, DS3, DS4, DS5 and DS6) and the other six are positioned in the anode pack upstream (US1, US2, US3, US4, US5 and US6), in total twelve thermocouples in one anode

12


UN CO RR EC TE D

PR OO F

M. Zaidani et al.

Fig. 20. Comparison of the present numerical simulation of the anode temperature distribution for new and aged furnace with the experimental measurements (at EGA).

Fig. 21. Binary color anode mapping temperature at the packing coke interface and the anode middle plane as a function of the flue-wall deflection for a constant and variable material properties.

and over-baking of the anodes in certain zones. This temperature evolution as a function of the flue-wall deflection can be explained by changing the packing coke thickness between the flue-wall and the anodes as a function of the flue-wall deflection. From these results, it can be concluded that the packing coke thickness has a very important role in an efficient heat transfer from flue-walls to anodes. The warmest

zone is located near the outlet and the second burner position for all cases, and the coldest zone is near the inlet. Moving from a straight flue-wall to a deflected flue-wall, it can be clearly seen that the warmest zone is getting further hotter at the packing coke-anode interface (Fig. 15a) and development of a hotspot at the anode’s middle-plane (Fig. 15b). At the packing coke-anode interface (Fig. 15a), the in

13

M. Zaidani et al.


PR OO F

ode quality has not been presented until now. The numerical model used in the present study is validated qualitatively with the experimental work carried out at Emirates Global Aluminium on the aged furnace. Fig. 20 presents a qualitative comparison of the present numerical simulation of the anode temperature distribution for new and aged furnaces with experimental measurements. It can be clearly seen that the temperatures of the anode increase with the increase of the furnace age through the development of clearer hotspots and over-baking of the anodes in certain zones near to the second burner. The warmest zone for numerical and experimental anode temperature distributions are located near the outlet near the second burner position for new and aged furnaces. Moving from a straight flue-wall to aged flue-wall, it can be clearly seen that the warmest zone is getting warmer as a function of the flue-wall aging, and the hot spot in the anode is getting larger by the expansion of this area. The coolest part near the top of the pit is also getting warmer as a function of the flue-wall aging, leading to a very poor anode baking homogeneity for the aged furnace. The obtained results are summarized in Table 1 and plotted in Fig. 16. Fig. 16a shows the evolution of the anode average temperature as a volume average and at the anode middle plane and anode-packing coke interface as a function of the flue-wall deflection. It can be observed that the anode volume average temperature increases with increasing the flue-wall deflection from 891 °C to 899 °C for a straight and a 99-mm deflected flue-wall, respectively. Also, as expected, the average temperature at the anode-packing coke interface is higher than the temperature at the anode middle-plane. At the anode packing coke-interface, the anode average temperature increases from 892 °C for a straight flue-wall and reaches up to 900 °C for a 99-mm deflected flue-wall. The anode’s middle-plane temperature increases from 890 °C to 898 °C for a straight flue-wall and 99 mm deflected flue-wall, respectively. Therefore, the more the flue-wall is deflected, the higher is the anode’s average temperature, leading to an overbaking of the anode in certain zones that consequently leads to an inhomogeneous anode baking. Fig. 16b depicts the volume average, minimum and maximum temperatures as a function of the flue-wall deflection. The anode’s maximum temperature increases from 992 °C to 1057 °C for a straight flue-wall and a 99-mm deflected flue-wall, respectively. The anode’s volume temperature is seen to be almost constant; however, the homogeneity of the temperature can be presented by the difference between the maximum and minimum temperatures of the anode. It can be seen clearly that the temperature difference increases from 276 °C for a straight flue-wall to 319 °C for a 99 mm deflected flue-wall leading to inhomogeneous baking. This difference is significant and cannot be ignored. Also, the anode’s minimum temperature increases from 717 °C

Fig. 22. The ratio of hotspots formation in the anode as a function of the flue-wall deformation (Fct: temperature dependent material parameters; Cst: constant material parameters).

Scenario 1

UN CO RR EC TE D

Table 2 The values of the operational parameter. Scenario 2

Scenario 3

Scenario 4

Mainstream velocity (m/s)

Hot air jet velocity (m/s)

Hot air jet temperature (°C)

Air temperature (°C)

1 1.5 2 2.5 3 3.5 4

8.5 9 9.5 10 10.5 11 11.5

1500 1600 1700 1800 2000 2100 2200

500 600 700 800 900 1000 1100

terface average temperature is around 892 °C for a straight flue-wall and increases to reach 900 °C for a 99-mm deflected flue-wall. Moving from the packing coke-anode interface to the anode middle-plane (Fig. 15b), the heat diffusion in the solids has further smoothed the hotspot. At the anode middle plane, the middle plane average temperature is around 890 °C for a straight flue-wall and increases to reach 898 °C for a 99-mm deflected flue-wall. To the best knowledge of the authors, a detailed numerical model on the flue-wall deformation modes and their effect on the baked an

Fig. 23. Anode volume average temperature as a function of (a) the fuel temperature and (b) the fuel velocity for 0 and 99 mm deflection modes.

14


PR OO F

M. Zaidani et al.

Fig. 24. Anode volume average temperature as a function of (a) the air temperature and (b) the air velocity for 0 and 99 mm deflection mode.

of the refractory bricks also affects the anode baking. The anode baking homogeneity can be optimized by improving the flue-wall design using different thermal conductivity bricks to improve the baking of the critical zone of the anode and improve the baking homogeneity. A high thermally conductive brick can be used on the right side of the flue-wall to improve the under-baked area, and lower thermally conductive bricks can be used on the left side of the flue-wall to avoid the hot spot creation and ensure a homogeneous baking. A flue-wall design improvement should also be conducted in order to avoid the under-baking or the overbaking of the anode for a deflected flue-wall, which depends mostly on the baffles openings. The temperature distribution at the flue-wall may be improved by closing the upstream baffle in order to give a more symmetric and uniform flow distribution and consequently a more uniform temperature distribution, which leads to improved heat exchange between the gases and flue-wall. Further analysis should be conducted in the future to improve the design of the flue-wall by varying the thickness of the flue-wall, the thickness of the packing coke in order to identify the best strategy to follow for improving the flue-wall design and achieve a more homogeneous baking.

UN CO RR EC TE D

to 738 °C for a straight flue-wall and a 99-mm deflected flue-wall, respectively. In reality, it is very difficult to obtain the evolution of the materials properties as a function of temperature especially those of the packing coke and anode materials. Therefore, in order to study the effect of materials properties on the anode temperature distribution as a function of the flue-wall deflection, Fig. 17 presents the anode temperature distribution for the case of constant material properties for different flue-wall deflection levels where the results are summarized in Table 1. As can be clearly seen in Fig. 17, similar qualitative results are obtained from the anode temperature distributions for the case of variable materials properties (see Fig. 10). The anode temperature is lower when considering constant material properties that do not vary with temperature. However, the difference is small except for the maximum anode temperature. For example, for un-deformed flow-wall, the anode’s maximum temperature when considering constant properties is 16 °C lower than for variable properties, which is not negligible (see Fig. 18). Fig. 19 shows a comparison between the anode volume average temperature as a function of the flue-wall deflection for constant and temperature-dependent material properties. It can be observed that there is a deviation in the results of the anodes average temperature in these two cases. The range of deviation is largest for un-deformed flue-wall and decreases as the deflection increases. This behavior could be attributed to the fact that in the present study the constant material properties are obtained by averaging the values above 500 °C which gives more accurate results for the higher temperature values obtained as flue-wall deflection increases. However, it should be mentioned that using constant material properties leads to misleading conclusions regarding the anode temperature distribution and the creation of hotspots in the anodes. When temperature-dependent properties are used, which is closer to the actual situation, the solid temperatures are higher, and the temperature gradients are lower compared to those of the case with constant properties. Recently, the effect of variable material properties on the solid temperature has been reported. It is noteworthy that in many computational studies reported in the literature, the solid material properties are taken as constant. It can be concluded that the aging of the furnace, affect the anode baking temperature distribution and homogeneity and leads to an under-baking or an overbaking of the anode in certain areas, sometimes, to a redistribution of the anode temperature based on the original anode temperature distribution for a straight flue-wall and on the degree and the shape of the flue-wall deflection. Therefore, the expected anode temperature distribution and homogeneity for a deflected flue-wall depends on the temperature distribution for a straight flue-wall before deflection. To avoid this problem, and to assure the uniformity of the anode temperature as much as possible, the deformed flue-wall must be improved, straightened or replaced regularly. The type of material

3.2. Baking uniformity In this section, for a finer evaluation of the previously stated results, the possibility of utilizing image processing methods to evaluate the hotspot formation on the anode is investigated. The hotspot formation is potential for an overbaked zone with lower baking quality. Here, an image processing technique is applied to the anode temperature distribution images obtained from the previous numerical simulations. Firstly, the colored contours of the anode temperature distribution (Figs. 15 and 17) are turned into grayscale pictures. Secondly, the pictures are scanned such that important statistical values of the images are obtained (such as the mean value of the color codes of pixels, and a number of total pixels). Thresholding is the simplest method of image processing and mainly it means turning a grayscale image (having 255 colors) into a binary image (having only 2 colors) (see Fig. 21). After applying the thresholding method, within 5% of the maximum temperature is obtained for the straight flue-wall (images have only two colors; 255 for black and 0 for white). In this way, hotspots are represented by black pixels. Finally, the last step is to calculate the number of black pixels. The ratio of the number of black pixels to the total number of pixels represents the ratio of hotspots creation in the anode. Fig. 22 depicts the ratio of hotspots creation in the anode as a function of the flue-wall deflection for both constant and variable material properties. One can clearly see from Fig. 17 that the hotspot regions where overbaking occurs are close to the flue-wall outlet at the top of the anode. This is also consistent with the regions where maximum anode temperature is observed. Fig. 22 presents the percentage of the hotspot

15

M. Zaidani et al.


creation at the packing coke-anode interface and at the anode middle plane as a function of the flue-wall deflection. These percentages indicate the total area of the overbaked anode. It can be observed that the percentage of the hotspot creation increases with the increase in the flue-wall deflection for temperature-dependent material properties from 1.94% to 5.65% at the packing coke-anode interface, and from 3.65% to 7.83% at the anode middle plane. For constant material properties, the hotspots percentage in the anode is increasing from 0.08% to 4.68% at the packing coke-anode interface and from 1.97% to 6.3% at the anode middle plane. Employing constant material properties, which is the common practice in the literature when simulating the carbon anode baking process, underestimates the percentage of overbaked anodes.

4. Conclusions

PR OO F

Flue-wall deformation is one of the most crucial aging modes of carbon anode baking furnaces. Such deformation does not only create difficulties in loading and unloading carbon anodes in the pits but also leads to inhomogeneous anode baking that affects the quality of the baked anodes and furnace energy efficiency. The current study develops a computational CFD 3D simulations on the effect of flue-wall deformation on the anode baking and creation of hotspots using a coupled heat and mass transfer models considering constant and temperature-dependent material properties. By the initial assessment, it is perceived that in the case of anode baking furnace modeling, it is essential to use temperature-dependent material properties for the accurate estimation of anode temperature distribution. Hence, the developed computational model is employed to investigate the effects of flue-wall deformation using temperature dependent material properties. It is observed that the maximum anode temperature which occurs near the flue-wall outlet (at the top of the anode pack) increases significantly with the increase in the level of deflection. This is due to the fact that the variation of side packing coke thickness increases by the increase in flue-wall deformation. This implies that as the flue-wall deflection increases, temperature inhomogeneity and over-backing zones increase. This non-homogenous temperature distribution within the anode pack is not only the main cause of excess consumption of the carbon anodes but also overconsumption of energy in both the aluminum reduction cell as well as the anode baking furnace. Moreover, the effect of some key operational parameters such as hot air jet (fuel) and mainstream (oxidizer) velocity and temperature are also investigated. It is observed that for both the straight and deformed flue-walls, these parameters have a significant impact on anode temperature distribution and should be selected in a view to improving the baking uniformity. The insight obtained in the present study can be effectively employed in modifying the furnace geometrical and operational parameters with the deformed flue-walls.

3.3. The effect of varying operational parameters

UN CO RR EC TE D

The effects of the velocity and temperature of air and fuel are also presented in this work. In this parametric study, an investigation of the effect of these parameters on the anode volume average temperature is conducted. Here, a straight un-deformed flue-wall and a 99-mm deflected flue-wall (worst case scenario) are considered for temperature dependent material properties. Table 2 summarizes the parameters used in this parametric study such that one parameter is changed at a time and the rest are set constant. The range of these parameters has been selected close to what is usually measured in real furnace operations. Fig. 23a depicts the anode volume average temperature as a function of the fuel temperature (hot air jet) for straight and 99 mm deflected flue-walls. The range of the fuel temperature is between 1500 °C and 2200 °C. It can be observed that the anode volume average temperature increases nonlinearly with the increase of the fuel injection temperature. For a straight flue-wall the anode volume average temperature increases from 869 °C to 911 °C for 1500 °C and 2200 °C fuel temperatures, respectively. For a 99-mm deflected flue-wall, the anode volume average temperature increases from 875 °C (for 1500 °C fuel temperature) to 918 °C (for 2200 °C fuel temperature). The anode volume average temperature increases by almost 7 °C from the straight flue-wall to the 99 mm deflected flue-wall case. Fig. 23b depicts the anode volume average temperature as a function of the fuel temperature for 0 and 99 mm deflection modes. The fuel velocity is ranged between 8.5 and 11.5 m/s. It can be observed that the anode volume average temperature increases linearly with the increase in the fuel injection velocity. For a straight flue-wall, the anode volume average temperature increases from 873 °C to 906 °C for velocity values of 8.5 and 11.5 m/s, respectively. The anode average temperature is higher in the case of a 99-mm deflected flue-wall by 6 °C as compared to the straight flue-wall. Fig. 24a presents the anode volume average temperature as a function of the air inlet temperature for a straight flue-wall and 99 mm deflection mode. The air inlet temperature is ranged between 500 °C and 1100 °C. The anode volume average temperature increases linearly with the increase in the air inlet velocity. The anode average temperature for a straight flue-wall is lower by 4 °C as compared to the 99-mm deflected flue-wall. Fig. 24b shows the anode volume average temperature as a function of the air inlet velocity, where the air inlet velocity is ranged between 1 and 4 m/s. In contrast with the other cases, it can be observed that the anode temperature decreases with the increase in the air injection velocity. For a straight flue-wall the anode temperature decreases from 949 °C to 858 °C for air injection velocities of 1 and 4 m/s, respectively. Also, the anode temperature is higher in the case of a 99-mm deflected flue-wall by approximatively 6 °C as compared to the straight flue-wall. Apparently, the increase in anode’s average temperature due to deformation of flue-walls is higher when varying Fuel flow rate and temperature than varying gas flow rate and temperature.

Acknowledgment This research paper is made possible through the help and support from the Emirates Global Aluminium (EGA). We are also very thankful for the carbon anode area representatives at EGA for providing the needed support. References

[1] N. Oumarou, D. Kocaefe, Y. Kocaefe, An advanced dynamic process model for industrial horizontal anode baking furnace, Appl Math Model 53 (2018) 384–399. [2] N. Oumarou, D. Kocaefe, Y. Kocaefe, B. Morais, Transient process model of open anode baking furnace, Appl Therm Eng 107 (2016) 1253–1260. [3] A.R. Tajik, T. Shamim, R.K.A. Al-Rub, M. Zaidani (Eds.), Performance analysis of a horizontal anode baking furnace for aluminum production, ICTEA: International Conference on Thermal Engineering, 2017. [4] Y. El Ghaoui, S. Besson, Y. Drouet, F. Morales, A. Tomsett, M. Gendre, et al., Anode baking furnace fluewall design evolution: a return of experience of latest baffleless technology implementation, In: Light metals 2016, Springer, 2016, pp. 941–945. [5] J.A. Johnson, A.V. Rozin, A.P. Skibin (Eds.), Light metals-Warrendale-proceedings, TMS, 2004. [6] Q. Fan, S. Hui, S. Zhao, Q. Zhou, X. Chen, Q. Zhao, Thermal stress and strain distributions of a laboratory scale wall fired furnace: a numerical study and experimental verification, Eng Failure Anal 25 (2012) 227–237. [7] M. Baiteche, D. Kocaefe, Y. Kocaefe, D. Marceau, B. Morais, J. Lafrance, Description and applications of a 3D mathematical model for horizontal anode baking furnaces, In: Light Metals 2015, Springer, 2015, pp. 1115–1120. [8] F. Grégoire, L. Gosselin, Comparison of three combustion models for simulating anode baking furnaces, Int J Therm Sci 129 (2018) 532–544. [9] A.R. Tajik, T. Shamim, M. Zaidani, R.K.A. Al-Rub, The effects of flue-wall design modifications on combustion and flow characteristics of an aluminum anode baking furnace-CFD modeling, Appl Energy 230 (2018) 207–219. [10] A.R. Tajik, T. Shamim, R.K.A. Al-Rub, M. Zaidani, Two dimensional CFD simulations of a flue-wall in the anode baking furnace for aluminum production, Energy Procedia 105 (2017) 5134–5139.

16

M. Zaidani et al.

Applied Energy xxx (2018) xxx-xxx [17] J. Oh, D. Noh, E. Lee, The effect of CO addition on the flame behavior of a non-premixed oxy-methane jet in a lab-scale furnace, Appl Energy 112 (2013) 350–357. [18] T.L. Bergman, A.S. Lavine, F.P. Incropera, D.P. DeWitt, Incropera's principles of heat and mass transfer, Wiley Global Education, 2017. [19] M. Zaidani, R.A. Al-Rub, A.R. Tajik, T. Shamim, 3D Multiphysics model of the effect of flue-wall deformation on the anode baking homogeneity in horizontal flue carbon furnace, Energy Procedia 142 (2017) 3982–3989. [20] B. Friedherz, G. Frank, Ring pit furnaces for baking of high quality Anodes – an overview Riedhammer, Aluminium Int J Aluminium Essen Fair (2006) 1–15. [21] A. Yurkov, Refractories for anode baking furnaces. Refractories for Aluminium, Springer, 2015245–249. [22] H. Lenz, F. Brunk, Improved anode baking furnace cover material, Light Metals (2002) 629–632.

UN CO RR EC TE D

PR OO F

[11] A.R. Tajik, R.K.A. Al-Rub, M. Zaidani, T. Shamim, Numerical investigation of turbulent diffusion flame in the aluminum anode baking furnace employing presumed PDF, Energy Procedia 142 (2017) 4157–4162. [12] S. Gövert, D. Mira, J.B. Kok, M. Vazquez, G. Houzeaux, Turbulent combustion modelling of a confined premixed jet flame including heat loss effects using tabulated chemistry, Appl Energy 156 (2015) 804–815. [13] S. Li, X. Wei, L. Yu, Numerical study on NOx/CO emissions in the diffusion flames of high-temperature off-gas of steelmaking converter, Appl Energy 88 (4) (2011) 1113–1119. [14] I.A. Ramadan, A.H. Ibrahim, T.W. Abou-Arab, S.S. Rashwan, M.A. Nemitallah, M.A. Habib, Effects of oxidizer flexibility and bluff-body blockage ratio on flammability limits of diffusion flames, Appl Energy 178 (2016) 19–28. [15] Y. Choy, H. Zhen, C. Leung, H. Li, Pollutant emission and noise radiation from open and impinging inverse diffusion flames, Appl Energy 91 (1) (2012) 82–89. [16] A.R. Tajik, V. Hindasageri, A numerical investigation on heat transfer and emissions characteristics of impinging radial jet reattachment combustion (RJRC) flame, Appl Therm Eng 89 (2015) 534–544.

17