This article describes a computational fluid dynamics (CFD) model used to evaluate furnace performance of the #8 furnace at Roth Bros./Philip Services.

Fig. 1 Reverbatory aluminum furnace with continuous charging procedure.
Oxygen based combustion was tried in the aluminum industry to increase furnace production as early as 40 years ago. These experiments gained a lot of negative publicity. Most of the trials ended with a high rate of aluminum oxidation and furnace refractory damage. The loss in yield and expensive furnace relining forced the aluminum industry to reject the use of pure oxygen. Today, the industry is re-evaluating the concept of oxygen firing in aluminum furnaces because of the growing demand to increase production without investing significant capital and because of advancements in oxygen-enhanced combustion technology.

Air Products and Chemicals, Inc. (APCI) was chosen by the DOE as the project managing organization because of their significant experience in oxy/fuel combustion in aluminum and other industries. The scope of the project included the development and subsequent demonstration of a low cost, high efficiency, and low NOx technology for the aluminum industry. The Roth Bros. Smelting Corp. was chosen as the host site for implementation of the new technology.

A sketch of the furnace is shown in Fig. 1. This furnace consists of three main elements: furnace hearth, pump well, and charging well. Aluminum scrap is constantly delivered into the charging well during the melting cycle. A pump circulates molten metal from the furnace hearth through the pump and charging wells back into the furnace hearth. The speed of the metal circulation and the metal temperature dictate the charging rate. The desired average metal temperature at the inlet to the pump well is usually 1000 to 1035 K (1350 to 1400 F). The melt temperature decreases in the charging well as a function of the amount of scrap charged. Thus, rapid temperature recovery in the furnace hearth is of great importance and the rate of energy delivered to the melt is directly linked to the production achieved.

The dross formation rate is determined by oxygen availability at the melt interface. Therefore, another important parameter is the oxygen concentration throughout the combustion space of the furnace. As long as this concentration is kept low, excess dross is not formed and as a result yield and heat transfer are maximized inside the furnace.

Fig. 2 Model outline.

Model Description

A computational fluid dynamics (CFD) model was developed based on the commercially available FLUENT code. The model solves two sets of equations simultaneously, one for the turbulent reacting flow of gases in the combustion space, the other for the laminar flow of the molten aluminum bath. Such an approach provides the flexibility to evaluate furnace performance at a variety of production and firing rates. The interactions of furnace production, melt rate, metal pump speed, metal level, burner orientation, and firing mode can easily be changed and studied with this model. For example, a change in the melt height automatically recalculates velocity, temperature, and species concentration everywhere in the domain.

In this particular case, the furnace production rate and molten bath height were fixed. In addition, the firing rate of the burners and orientation were also fixed for three furnace firing modes:

  • Air/Fuel (AF)
  • Oxy/Fuel (OF)
  • Air-Oxy/Fuel (AOF) with 35% total oxygen enrichment (TOE)

An outline of the model is shown in Fig. 2. One should note that the model does not include the pump well and the charging well. By using the average inlet velocity of the molten metal through the inlet arch, the metal circulation through the wells and the pump speed were approximated. The temperature at this location was calculated based on an energy balance around the well. Except for the above assumption, the rest of the model has the exact furnace geometry. The elimination of the well in the model helped to reduce the computational domain and to obtain the results faster without sacrificing accuracy. It would be a simple matter to include the well if necessary.

All three cases were built with the same 90,000-cell grid and with the only difference in the number of cells assigned to represent the burners. For example, the AF burner has the highest number of cells since the burner has the largest cross-sectional area. The OF burner has the smallest number of cells. The inlet velocity and temperature were specified for each burner for natural gas, oxygen, and air. The firing rate of each burner in each case was kept the same. This firing rate was 3.5 x 106 Btu/hr, so the total firing rate for the furnace was 14 x 106 Btu/hr regardless of the firing mode (AF, OF, AOF).

The molten aluminum bath height was chosen at 25 inches from the furnace floor. This metal height represents an intermediate point during the furnace melting cycle. It is possible to change the metal line and study the process at different metal levels. Ideally this approach gives the flexibility to simulate quasi steady-state solutions, the combination of which should represent the overall melting cycle. For a 25 inch metal level, the pump speed was chosen to circulate 8800 lb of aluminum per minute. The scrap charging rate of 12000 lb per hour was specified. Therefore, the aluminum mass and energy balance through the furnace wells was defined. The following boundary conditions were applied at the arch between the charging well and the furnace hearth: velocity of 6.2 x 10^-2 m/s, temperature of 976 K (1297 F).

The important part of the modeling work was model validation. The model temperature prediction was compared with the temperature measured in the pump well, in the flue, and in the roof refractory. The agreement of the temperatures was within 10%. The relative comparison of the cases is now valid since all the cases were built with the same grid and assumptions.

Fig. 3 Horizontal temperature distribution through burner level for the Air-Oxy/Fuel case.

Results and Discussion

Some results of the model predictions are shown in Figs. 3 through 12. Each figure has three sub-figures, a, b, and c and representing the AF, AOF, and OF cases, respectively. The same color scheme ranges are used for each sub-figure. This way it is possible to make a visual comparison without closely focusing on the numbers displayed on the left hand side of each sub-figure. However, if more accurate information is needed, it can be obtained from the scale provided. It is also important to note that the same colors are used regardless of the parameters presented (temperature, mole fraction, velocity, etc.). It is important to look at the values and the range of the displayed parameter. For example, compare Figs. 3 and 5. The temperature field in Fig. 3 is presented in K with a temperature variation from 1000 K (1340 F) to 2500 K (4040 F). The temperature distribution inside the melt, Fig. 5, is also presented in K, but the temperature range variation is only 7 K (45 F) from 975 K (1296 F) to 982 K (1308 F). The white spots indicate that the values are outside the chosen range, for example, the hot flame temperature in Fig. 3c and Fig. 4c.

Fig. 4 Vertical temperature distribution for 1st and 3rd burners for the Air-Oxy/Fuel case.
The temperature distribution in the furnace combustion space is shown in Figs. 3 and 4. The OF case shows the hottest calculated peak flame temperature of about 2760 K (4508 F). The AOF case has a calculated peak flame temperature of about 2500 K (4040 F), while the calculated peak flame temperature in the AF case is only about 1680 K (2565 F). The vertical distribution of burners 2 and 4 was similar. It is also shown that the calculated average gas temperature in the furnace combustion space in the OF case is significantly hotter than the AF case. The calculated average gas temperature in the AOF case is shown to be much closer to the OF case than to the AF case. From these results, one can expect the following:

Since the AOF case provides just enough energy to melt 12000 lb of the charged scrap, the AF case does not supply enough energy to the melt at the same firing rate. A significantly higher firing rate would be required for AF operation. At the same time, it is not obvious what firing rate the furnace can sustain without overheating or excessive refractory wear.

Slightly more than the required energy is supplied to the melt in the OF case, however, the refractory is a concern in this case since temperature exceeds 3000 F in several places. Additional work is needed to understand if lower firing rate can provide enough energy to the melt with a simultaneous reduction of the refractory temperature.

Fig. 5 Horizontal temperature distribution through the melt for the Air-Oxy/Fuel case.
The melt temperature distribution is shown in Fig. 5. An analysis of the results shown here leads to the conclusion that the melt temperature at the pump door in both the AOF and OF cases is sufficient to melt the scrap charged at the specified rate of 12000 lb/hr with the melt circulation of 8800 lb/min. The AF case shows that the melt temperature is actually lower at the pump-well arch than at the charge-well arch. This is obviously impossible in the actual furnace. However, in the model this leads to an important conclusion. One should recognize that the boundary conditions at the charge-well arch are fixed. Then, the temperature distribution shows that the melt temperature cannot be recovered to melt the aluminum scrap at the charging rate of 1200 lbs/hr with AF firing of 3.5 x 10^6 Btu/hr per burner. Consequently, the firing rate needs to be increased in the AF case to be able to melt the scrap. The only question is what the firing rate should be. An additional study is planned. Increasing the firing rate obviously will increase natural gas consumption. This matter should be taken into consideration when the economics of all three technologies are evaluated. A field trial has confirmed that an increase in firing rate is required for the AF case compared to the AOF case. The burners were fired 30-40% higher in the AF case, however, the same production as with the AOF burners was not achieved.

A hot spot is shown in the melt in Fig. 5 for the AOF and OF cases. This hot spot is more pronounced in the OF case. Once again, the scrap charging rate was kept the same for all three cases. These results lead to the conclusion that even a slight increase in the firing rate in the OF case might overheat the aluminum surface. Such overheating can result in faster aluminum surface oxidation, yield loss, excessive drossing, and reduced aluminum alloy quality. It is predictable that the hot spot location is at the metal stagnation point if it exists in the melt. The location of the hot spot does not change with the melt depth, but the melt temperature rises towards the melt surface. The melt temperature in the side-well furnace is usually kept around 1000 to 1035 K (1350 to 1400 F) at the pump-well arch. The deeper the melt bath, the less efficient the furnace operation. Some furnaces do not have a metal pump at all. A "puddler" is used instead to stir the metal in the furnace charging well. A furnace of this kind is less efficient than a furnace which operates with a metal pump. Indeed, heat transfer in a molten aluminum bath occurs due to conduction through the metal depth and convection which is defined by the melt motion. Therefore, the melt circulation is critical for overall furnace thermal efficiency. At the same time any stagnation zone could cause surface overheating.

Fig. 6 Horizon velocity vector field in the melt for the Air-Oxy/Fuel Case.
The metal flow field distribution is shown in Fig. 6 at the horizontal plane 12 inches above the furnace floor. This plot confirms that the melt hot spot is indeed located at the stagnation zone, where the melt is re-circulating between two submerged furnace arches. Since the pump speed was kept the same, one would expect the same melt flow pattern for all three cases. Fig. 6 suggests that the attempt to move the hot spot towards the pump-well arch, would lead to a furnace production increase.

In Part II of this article, important issues in process and furnace control will be addressed including control of refractory temperature, firing rate, and oxidation of the aluminum.