Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries

SINTEF / NTNU Trondheim, Norway

6-8 June 2005

USE OF CFD TO PREDICT THE PERFORMANCE OF

A HEAT TREATMENT FURNACE

Yongxiang YANG1, Reinier A. de JONG1 and Markus A. REUTER1

1
Resource Engineering, Department of Geotechnology, Delft University of Technology
Mijnbouwstraat 120, 2628 RX, Delft, The Netherlands

ABSTRACT turned on by the controller. The controller works with a set

point and a process value, as PID controllers. The result of
In this paper, a mobile heat treatment furnace was simulated the temperature profile is printed on paper. After this a
by using CFD to investigate thermal performance of the hardness measurement is made to check whether the results
furnace and the heating process of the metal pieces. The are all right.
furnace is used to heat treat dredging pumps and impellers to
obtain the required microstructure and mechanical properties At the moment problems with heating of impellers used in
through stress relief, annealing, hardening and tempering. dredging pumps occur. The hardness of the impeller is
Since the temperature evolution inside the metal piece cannot uniform; in the middle the material is much thicker, which
be tracked in practice, CFD simulation provides a useful tool causes it to cool down more slowly. The hardness at this
to predict the temperature evolution within the metal pieces point is lower. To get a better understanding of the heating
during the heat treatment. The current CFD model consists of and cooling of these impellers a computer model would be
turbulent combustion, thermal radiation, and conjugate heat very useful. This will lead to a better controlled process,
transfer. Temperature measurement was carried out to significant energy savings and more environmental friendly
provide thermal boundary conditions and calibrate model production.
parameters. An overall energy balance indicted a relatively
low energy efficiency of the furnace. To improve energy
efficiency design changes were made with the model.
Reducing the amount of excess air will also save a lot of
energy and thus cost in production. Heating of the furnace by
means of radiation plates proved to be a good alternative.

Keywords: dredging impeller, transient heating, heat

treatment, CFD simulation.

INTRODUCTION

Heat treatment is a common practice in metal production and

manufacturing industries. Through heat treatment at certain
temperatures with special patterns, unique mechanical
properties of the metals and the metallic products could be
obtained. At Akkermans Gloeitechniek, a Dutch company (a) The mobile furnace
situated in Kinderdijk, (a sub-contractor of IHC-Holland)
heat treatment of metal components for dredging industry is
conducted for various type of heat treatment such as stress
relieving, soft annealing and hardening, tempering,
transformation and precipitation. At the company 6 furnaces
are in operation and the 7th is being built.
4 x burner
One of the furnaces is a mobile cylindrical one, as shown in Furnace
Figure 1. The process is based on changes in the crystal
structure of the metals. For example, for hardening treatment
material is heated up slowly up to around 1000°C. At this
temperature the structure of steel changes into “austenite”.
After rapid cooling (remove the hood) the steel becomes
much harder. Cooling can be done in air, in water, and in oil Metal

or polymers. The furnace is heated up by oil-burners.

Temperature control is done with thermocouples and (b) burner orientation
controllers. When the temperature becomes too high, the
controller gives a signal to the burners, which turn off at that Figure 1: The mobile heat treatment furnace and the
time. When the temperature becomes too low, the burners are schematic view.
Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

THE HEAT TREATMENT PROCESS 1200

Preset heating profile of the impeller

The heating process starts with placing the metal piece on 1000

supporting bricks. This is done to support the metal so that it

Temperature ( C)
would not plastically deform at high temperatures or even 800

break in two pieces. When the metal piece is placed on the

600
support the contact thermocouples are attached to the metal
and connected to the controller unit and indirectly to the 400
recording unit. The mobile furnace is placed on top of the
metal as a hood. Sand is pushed to the sides of the furnace to 200

seal it better. After the burners are connected to the

0
controlling unit the heating profile is inserted into the 0 5 10 15 20 25

controller. Time and temperature are set to give a heating Time ( hrs)

profile along the heating process, as is illustrated in Figure 2.

Heating rate, intermediate soaking time and end soaking time Figure 2: The preset heating profile of the dredging impeller.
can be set. When all this is done the program can start and the
furnace is in operation. The heating process is automatically package PHOENICS 3.5 (Cham, 2004). The governing
controlled so that no manual changes are necessary. partial differential equations for conservation of mass,
momentum, and energy in a turbulent flow system can be
After the heating up and soaking stages, which take around expressed by time-averaged 2nd order partial differential
twenty hours, the burners are disconnected from the equations as is shown in equation (1), and solved with a
controller and the hood is removed to let the metal cool common numerical algorithm in the code.
down. The metal piece at about 980°C cools down to
environmental temperature in open air. After the metal is ∂
completely cooled down, usually the next day, the hardness ( ρφ ) + div( ρuφ ) = div(Γφ , eff gradφ ) + Sφ (1)
∂t
of the metal is measured at different points to check whether
the process was successful. Table 1 lists the general operation
data of the heat treatment process, including overall mass and Where φ is the general flow variable such as velocity
energy balance. components, temperature and mass fractions of chemical
species; ρ is the fluid density; u is the fluid velocity vector.
The energy distribution in the furnace varies a lot along the The effects of fluid turbulence are accounted for in the
operation as will be illustrated from the model results later.
For instance, the heat loss from the walls and the energy transport coefficient Γφ ,eff by a turbulence model, and Sφ is
carried out from the off-gas are lower in the beginning than in a general source term which could include all terms other
the end, while the energy stored in the metal piece becomes than transient, convection and diffusion terms in the equation.
smaller along the heating process. This is a time-dependent In the current simulation, all the above mentioned transport
dynamic process. processes are modelled as below:

Item Average value • Turbulence: the standard k-ε model

Furnace dimension: • Thermal radiation: IMMERSOL (Immersed solid)
Inner diameter × Height × • Oil combustion: global gas combustion model
Lining thickness 3.76 × 3.50 × 0.12 m (simplified from oil to gaseous fuel)
Burner inner diameter and 0.1 m, 1.5 m from
location furnace bottom The global gas combustion model and the immersed solid
Average fuel consumption 37.18 kg/hr radiation model are described in more detail later. Since the
Average air supply 1327.2 kg/hr general information about the standard k-ε turbulence model
Stoichiometric ratio 17.24 is available for most CFD analysts and process engineers, it is
Average excess air 107% not be described in this paper. Some preliminary simulation
Average wall temperature 50°C results have been published previously (Yang et al., 2004).
Average off gas temperature 668°C The current paper describes new developments and findings.
More details of the results can be found in the MSc. Thesis of
Average weight of the impeller 4,620 kg
de Jong (2004).
Table 1: Operating data of a typical heat treatment operation.
Global Combustion Model (SCRS)

As the first approximation, the oil combustion is simplified

CFD SIMULATION with gas combustion model. Based on the oil properties, the
same amount of gaseous fuel with the same heat of
Simulation Tool combustion is estimated. The oil is assumed to vaporize
immediately upon entering the furnace.
The heat treatment process in the furnace involves turbulent
gas flow, oil combustion, thermal radiation and conjugate In PHOENICS, the global gas combustion model developed
heat transfer within the metal components and furnace walls. by Spalding (1979) was implemented as the Simple
The simulation is conducted by using a general purpose CFD Chemically Reaction System (SCRS). The SCRS involves a
Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

reaction between two reactants (fuel and oxidant) in which be stored by the metal component. The radiance temperature
these combine, in fixed proportions by mass, to produce a T3 is the temperature of the solid (the dredging impeller and
unique product: the brick lining).

1 kg fuel + s kg oxidant = (1 + s ) kg product (2)

∂ ⎛ ∂ T3 ⎞ ∂ ⎛ ∂ T3 ⎞ ∂ ⎛ ∂ T3 ⎞
⎜ k1 ⎟+ + ⎜ k1 ⎟
∂ x ⎝ ∂ x ⎠ ∂ y ⎝⎜ ∂ y ⎠⎟ ∂ z ⎝ ∂ z ⎠
k1
Where s is the stoichiometric oxidant requirement (kg (4)
oxidant/kg mixture), ensuring complete combustion. This
dT
reaction is taken as irreversible. Reaction rates in turbulent +q ' = C p 3
flow situations are often more greatly affected by local dt
turbulence than by chemical factors. For these situations the
Eddy Break-up (EBU) model is provided, which rests on the where: Cp specific heat capacity of the solid
hypothesis that only turbulence and fuel concentration affect k1 thermal conductivity of the solid
the reaction rate having the following source term in the mass q’ the heat source per unit volume
conservation equation: T3 temperature of the solid

EP
Smfu = −CEBU × min {mfu, mox / s} * (3) Within the space between the dredging impeller and furnace
KE
walls, the distribution of radiation temperature derived from
radiosity (E3 = σ(T3)4), can be represented as obeying the
Where CEBU is the Eddy Break-up reaction constant, mfu is
equation:
the mass fraction of unburned fuel, mox is the oxidant
fraction, KE is the turbulent kinetic energy and EP the ∂ ⎛ ∂T3 ⎞ ∂ ⎛ ∂T3 ⎞ ∂ ⎛ ∂T3 ⎞
⎜ k2 ⎟+ + ⎜ k2 ⎟
∂x ⎝ ∂x ⎠ ∂y ⎜⎝ ∂y ⎟⎠ ∂z ⎝ ∂z ⎠
dissipation rate of KE. In the current study it is assumed that k2 (5)
the reaction rate for the combustion of the oil vapor is
turbulent mixing rate limited and the EBU is chosen to = ( a + s )σ (T 4 − T34 )
determine the reaction rate. The reaction rate constant in the
EBU model proved to be an important parameter, and it
generally determines the shape and size of the flame. where: k2 Temperature dependent conductivity of \
Different CEBU values have been tested between 1 and 50. It the fluid medium, and defined as
was found that if the CEBU value is set below 25 no 16σ T33
k2 =
normally observed flame could be formed and the 3( a + s + 1/ WGAP )
combustion takes place in the broader space in the furnace T3 radiation temperature of the fluid
except in the normal flame zone. Thus the value of 25 has σ Stefan Boltzmann constant (5.67×10-8
been finally used in the simulation model based on the W/m2⋅K4)
observation of the flame shape and flame temperature. T temperature of the fluid phase
a absorptivity of the fluid medium
Immersed Solid Radiation Model (IMMERSOL) s scattering coefficient of fluid medium
WGAP distance between adjacent walls or from
Thermal radiation plays a critical role for combustion the metal component
furnaces and the heat transfer between the combustion gases
and the metal components. Many radiation models have been Equation (5) means that the gas radiation is described in
developed in the world, and in PHOENICS three radiation terms of conduction. The radiation temperature between the
models for participating media are available: the 6-flux metal components or refractory walls can be computed at
model, radiosity model, and the immersed solid model each location in the domain. The net contribution due to
(IMMERSOL) [1]. Since its special design to handle thermal radiation to the energy conservation equation (the
conveniently the multi-solid flow domain and the conjugate energy source term in equation (1), Sφ) is the right-hand-side
heat transfer with conducting solids, it has been chosen for term in equation (5): (a+s)σ(T4-T34). Therefore, conduction
the current study. type of equations (4) and (5) could describe the influences of
both conduction in solids and gas-phase radiation between
The IMMERSOL model is a single heat-conduction-type the solids on the immersed-solids temperature T3 throughout
equation that describes the influences of heat conduction the domain, with a position- and T3-dependent conductivity.
within the solids and radiation in fluid domain between the
solids on the distribution of immersed-solids temperature. Geometry and Computational Grid
The model is mainly dependent on position and conductivity
of immersed-solids. This model is very useful for conjugate The model is constructed with Cartesian grid in order to get
heat transfer problems such as the temperature distribution fast convergence. The PHOENICS VR-Editor, the pre-
within the metal components. The IMMERSOL radiation processor provide the flexibility to construct various shapes
model can be described with the differential equation within of objects with reasonable accuracy, without using BFC grid
the immersed solid per unit volume as in Equation (4). or unstructured grid system. In order to construct the furnace
refractory wall and ceiling, high conducting solids embedded
During simulation of the impeller heating there is no internal in refractory blocks (wall and ceiling) are used to form an
heat generation, so that q' becomes zero. Equation (4) means internal cylindrical surface of the furnace and constant outer
that radiation from a furnace wall and combustion gases will surface thermal boundary condition. In order to test grid
Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

independence of the simulation, both coarse and fine grids Besides the discretisation of space coordinates, the time of
are used. For the dredging impeller to be heat treated, both furnace operation is divided into 55 time-steps for the total of
simplified impeller and the detailed impeller supplied with 20 hours. In order catch the sharp flow and temperature
CAD design are used in simulation. Figure 3 shows the photo changes in the starting period, time steps of 2, 10, 30, 60
of the impeller and the simplified impeller as well as the seconds have been used for the first 5 minutes, later 5 and 10
CAD figure used in the model. minutes till 30 minutes, and 30 minutes for the rest period
between 30 minutes and 20 hours. Because of the long
The coarse grid has number of 15,625 cells (25×25×25) and the computational time, finer time-steps could not been tested for
finer grid has 125,000 cells (50×50×50) for the simplified such a long transient operation.
impeller models. The model with the detailed dredging impeller
consists of 187,974 cells (59×59×54). Figure 4 illustrates the Summary of the Key Model Parameters
computational grid for the detailed impeller model. The
simulation results from both coarse and fine grid showed minor Table 2 lists the general model parameters and boundary
difference. In order to ensure better results, however, similar fine conditions for the simulation. Table 3 gives the thermal
grid is used for the detailed impeller model. physical properties of the combustion gas system. The
properties of the fuel are estimated from the oil properties
and under assumption that all the oil has been vaporized upon
entering the furnace.

(a) The impeller

Figure 5: The Cartesian grid of 59×59×54cells.

(b) The simplified impeller (c) detailed impeller (CAD) in

in the model the model

Figure 3: The dredging impeller in practice and in the CFD Parameters Value Parameters Value
model. Fuel/air input at Emissivity of the 0.80
burners impeller
Inlet temperature 30°C Gas absorption 0.10 m-1
coefficient
Chimneyes Inlet velocity 10.12 m/s Gas scattering 0.001 m-1
coefficient
Tangential 1.15 m/s Outer wall 50°C
velocity (6.5°) temperature
Fuel mass flow 0.00255 Inner wall 0.80
rate kg/s emissivity
Air mass flow 0.0908 Log-law
Burners (x4) rate kg/s Wall – gas wall
Fuel mass interactions function
Fan 0.0273
fraction
Total volume 1151 Turbulence 5%
flow rate Nm3/h intensity at inlet
Heat of 43 MJ/kg k-ε model standard
combustion constants
Sand Stoichometric Eddy break-up
15 25
ratio model constant

Table 2: Input and boundary condition

Figure 4: Geometry and the objects in the CFD model.

Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

Properties Value In addition, a wide range of gas absorption coefficient (0.05, 0.1,
Mixture density Ideal ideal mixing of 3 gas 0.2, 0.4) and surface emissivity of the impeller (0.05, 0.075, 0.1,
components 0.2, 0.3, 0.5, 0.8, 1.0) have been tested. As an extreme case, a
Molecular weight very low absorption coefficient and surface emissivity have been
Fuel (estimated from oil) 144 g/mole
tested for simulating a case of no radiation. It is very clear that
Oxidant (air) 29 g/mole
Combustion product 35.7 g/mole radiation heat transfer plays a very important role. However, the
Heat capacity: mixture heat capacity is change of absorption coefficient between 0.05 and 0.4 does not
proportional to the mass change appreciably the heat transfer, while the change of surface
fraction of each species emissivity of the impeller plays a critical role in the heating
Fuel 2471 J/kg/K profile of the impeller.
Oxidant 1110 J/kg/K
Combustion product 1170 J/kg/K The transient simulation has been mainly conducted with HP
Mixture kinetic viscosity 1.544×10-5 m2/s
workstation of 3 GHz running with windows XP. The total
computation time varies between 44 and 60 hours for the fine
Table 3: Thermal physical properties of the combustion gas grid of 59×59×54 (187,974) cells for a total heating period of 20
Model development
hours with 55 time steps.

The heat transfer from combustion gas to the metal surface,

and further to the inner core of the metal, is dominated by SIMULATION RESULTS
many factors such as combustion parameters and burner
settings, fuel supply strategy (constant or varying), the The combustion and metal heating processes are simulated as
quality of radiation model and influence of radiation a time-dependent transient model. In order to compare the
parameters, rate of excess air in the burners, the model end process of the transient heating, steady-sate simulations
development includes the following models: are also conducted. Quite reasonable results are obtained
• Models with constant fuel supply from the simulation. The basic flow feature, the temperature
distribution in the combustion space, and the temperature
• Different fuel supply at different time periods
evolution within the impeller are described below.
• Influence of absorption coefficient and emissivity
• Influence of excess air General Flow Pattern of the Gas Phase

The simulation results are compared with preset temperature As is described earlier, there are four oil burners installed
profile of the metal surface. After many testing models, a best slightly above the top surface of the impeller. The two
set of parameters have been obtained, and the final model is opposing burners have a small angle of 6.5° so that the flame
developed. Three heating periods with different oil supply jets do not meet in the center of the furnace, and more stable
and excess air will predict reasonably well the surface and slightly rotating flames are formed. Figure 7 shows the
temperature evolution of the impeller. The settings are listed general gas flow pattern obtained from the final model at
in Table 4, and the prediction of the surface temperature of three different cross sections. From Figure 6 it can be seen
the impeller is shown in Figure 6. The detailed feature of the that the jets from the four burners form a counter clockwise
final model will be described in the Simulation Results. rotating flow. This rotating flow extends from the burner
level to the top of the furnace upward and to the bottom of
Period Excess Mixture Fuel Energy per
the furnace downwards. Near the center of the furnace
(hour) Air (%) Fraction (kg) burner(kW)
between the two pair of burners, a downward gas flow is
0-8 226% 0.02 214 80
formed, which is ended near the impeller top-surface. Below
8-10.5 160% 0.025 84 100
the impeller the velocity is much lower, which is not efficient
10.5-20.5 85% 0.035 356 85
for the heating process of the lower part of the impeller.
Total - - 654 -
Table 4: Settings for final model. Owing to the technical and economical limitations, the velocity
distribution and the flow pattern have not been validated.
1100
Especially the use of standard k-ε model may bring errors to
1000
the predicted flow pattern and the turbulence level of the flow.
900
Temperature profile final model Therefore, more tests with different turbulence models e.g.
800
RNG k-ε model should be made, if the flow visualization
cannot be carried out. Perhaps a laboratory scale cold physical
Temperature (ºC)

700
600
models could be used for validating the flow pattern.
500
Heat Transfer and Temperature Distribution
400
Temperature impeller measured
300 Temperature impeller final model
200
Based on the combustion heat transfer models with simplified
100
impeller (coarse and fine grid) and the detailed impeller,
0
temperature distribution within the furnace (combustion
0 2 4 6 8 10 12 14 16 18 20 22 space) and within the metal impeller is obtained. Energy
Time (hour) balance is also made for each case to indicate the energy
Figure 6: Comparison of the surface temperature profile of utilization efficiency, heat loss through the walls. Steady state
the impeller measured and predicted in the final model models are also conducted to verify the end operation of the
heat treatment.
Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

a) At height of the burners b) just above the impeller c) Side view from burners
Figure 7: Gas flow pattern in the furnace viewed from different angles

1 hr 5 hr 10 hr

Furnace temp.
across the
burner axis
12 hr 15 hr 20 hr

1 hr 5hr 10 hr

Furnace temp.
across upper
12 hr 15 hr 20 hr part of impeller

Figure 8: Predicted temperature evolution in the furnace (0 – 20 hours).

Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

Temperature evolution In Figure 9 the upper parts of the impeller show higher
Figure 8 illustrates a temperature contours from the final temperatures. After twenty hours the outside temperature at the
transient simulation model. Due to the limited space in the top is already more than 1020°C. These figures emphasize the
paper, only a few time steps were chosen from the 55 time temperature difference between the top and the bottom.
steps (time 0 – 20.5 hours). From these snap-shots, the heating
processes of the furnace space and the impeller can be clearly Surface temperature at the end of the heat treatment
seen along the heating process. To illustrate more detailed At the end the top surface of the impeller has reached to
temperature distribution within the dredging impeller, a few 1050°C whereas the blades are only 970°C (see Figure 10).
temperature maps across middle part of the impeller (blade The surface temperature difference is 90°C. This may cause
zone) is also shown in Figure 8. It is clearly shown that the some internal stresses, which is not favourable for the
temperature from within the impeller develops slowly. process. The difference in surface temperature at the end of
the process is something, which can be measured in practice
To get better information of the impeller temperature the as well. This can be done when it really becomes a problem.
outside of the impeller as function of time is analyzed. The
temperature is from 180°C to 980°C for the first 15 hours. Energy distribution
After sixteen hours the top of the impeller reached the 980°C It is important that the energy could be utilized to heat up the
already. From this point on the temperature scale is set from metal components more efficiently, under the constraint that
920°C till 1020°C, to get a better view of the temperature the preset heating profile could be maintained. Table 4
differences. This is shown in lower part of Figure 8.The illustrates the energy distribution of 4 simulation cases, in
temperature of the impeller is higher on the top. The lower which Case D is the final model. The figures show both the
parts of the impeller have some delay in temperature first half an hour and at the end of the heat treatment period.
increase. As was mentioned before the heating of lower parts Three models show a large difference in the energy carried
is caused mainly by conduction through the metal more than out by the off-gases. The coarse grid model gives the highest
by convection or radiation. The part where the top plate goes heat loss by the wall, while the two fine grid models
over into the blade, the temperature is also slightly lower. The predicted similar figure for the wall heat loss of about 11%.
thickness of the impeller may have caused this. Thick parts Compared to the theoretical calculations for the energy
heat up more slowly than thin parts in general. balance, the fine grid model with the detailed impeller
predicts the nearest figures for the energy balance.

1 hr 5 hr 8 hr

10 hr 12 hr 15 hr

16 hr 17 hr 18 hr

19 hr 20 hr
Figure 9: Predicted evolution of the temperature distribution on the surface of the impeller.
Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

Energy Loss Loss Energy Design alteration: radiation plates

Model Time input to off- to to the
(kW) gas walls metal According to the analysis of energy distribution for the
( %) ( %) ( %) current practice, it is found that majority of the energy is lost
At 30 min. 512 48.5 5.5 46.0 to the off-gases from fuel combustion. Therefore, energy
A At End 512 67.2 27.3 6.6
optimization will be an important aspect to improve. This
At 30 min. 512 47.0 0 53.0 will result in a decrease in fuel consumption to maintain the
B At End 512 77.7 11.7 9.6
required temperature profile, lead to the lower operating
At 30 min. 512 48.0 0 52.0 costs.
C At End 512 83.6 10.9 5.5
At End The following alternatives have been investigated:
D operation 340* 74.0 14.7 11.3
• Radiation plates
Case A: coarse grid with simplified impeller • Heat exchange of off-gas
Case B: fine grid with simplified impeller • Positioning of burners and metal
Case C: fine grid with detailed impeller (CAD) • Reducing the excess air supply.
Case D: Final model with detailed impeller (CAD)
and varying fuel supply. Heating of the furnace by means of radiation plates is a good
* In practice and the final model, the fuel supply to the alternative. Preliminary results show that energy efficiency is
burners is not at constant rate: 0-8 hrs at 320 kW, 8-10.5 hrs much higher and production costs can be reduced. This way
at 400 kW, and 10.5 – 20.5 hrs at 340 kW. of heating the furnace is already used in practice at other
companies. Figure 12 illustrates the tested configuration and
Table 4: Energy distribution of the heating process Figure 13 shows the predicted surface temperature profile in
4 periods of energy settings (see Table 5) in comparison with
Figure 11 illustrates the dynamic change of the energy the preset values.
distribution along the heating process for the final model with
detailed impeller. It is clear that in the initial heating period the
energy from the oil burners is mainly used for metal heating 100%
solids
but more than half of the energy leaves the furnace by off-gas. 90%
out offgas
It is also obvious that the energy used for metal heating 80% Wall heat loss

decreases substantially from about 50% in the beginning to

Energy distribution

70%
about 20% after 5 hours, and at the end of heating only 11% of 60%
energy is used for heating. The wall heat loss increases slowly
50%
within the first 2 hours to about 5%, and reaches to about 15%
40%
at the end of heat treatment. This distribution feature follows
the heat transfer theory, since the energy transported to the 30%

metal (and other solid walls) is driven by the temperature 20%

difference between the solid and the combustion gas. The wall 10%
heat loss is driven by the increasing temperature difference 0%
between the outer surface (constant at 50°C) and the inner 0 2 4 6 8 10 12 14 16 18 20 22
surface (increasing) of the walls. Time (hrs)

Figure 11: Energy distribution of the final model

Radiation plate

Impeller

Figure 10: Predicted temperature distribution on the surface Figure 12: Illustration of the alternative design: radiation
of the impeller at the end of operation. plates.
Fourth International Conference on CFD in the Oil and Gas, Metallurgical & Process Industries
SINTEF / NTNU Trondheim, Norway
6-8 June 2005

With this total energy consumption of 2233 kWh for the with the global turbulent combustion model. The thermal
alternative design, the following rough estimation has been radiation was simulated with the immersed solid
made, in comparison with the fuel combustion furnace. The (IMMERSOL) radiation model supplied in the code.
price of 1 kWh in industry is around € 0.08. The total energy
costs for the radiation plates are then € 178. The price of 1 The heating profile was reasonably predicted by the
litre of red diesel in industry is € 0.67. The total oil simulation. Temperature measurement data was used to
consumption during normal operation is about 774 litre (654 validate the CFD model. Through the CFD simulation, not
kg), and the total oil costs € 520. Thus in stead of € 520 only only the surface temperature could be properly predicted, but
€ 178 for 2233 kWh electrical energy needs to be paid to heat also the temperature within the metal components could be
up the impeller to 980 °C. This is a difference of € 342 per simulated. The latter is crucial for the heat treatment process,
heat treatment. To get a better evaluation a cost-benefit but is not possible to measure directly in practice. Through
analysis has to be made in the future including the purchasing the simulation it is also found that the energy utilization
costs of the radiation plates. The placement of the radiation efficiency is relative low, and the energy is lost substantially
plates has to be examined too to get a more uniform by the outgoing off-gas. This indicates that there is large
temperature distribution, for instance with plates near the room to improve the energy utilization efficiency.
bottom.
Various case studies have indicated that by varying the fuel
supply to the burners and by regulating the air supply,
Time period Hours Energy setting Total energy substantial energy can be saved. Alternatively using radiation
0-2 2 200 constant 400 kWh plates instead of oil burners can save substantial amount of
2-8 6 75 – 105 kW linear 540 kWh energy, which is otherwise wasted by the large amount of off-
8-15 7 105 – 150 kW linear 893 kWh gases. Other options such as the burner orientation and the
15-20 5 80 constant 400 kWh laying position of the metal component could also reduce to a
Total (0-20) 20 2233 kWh certain extent the energy consumption.

ACKNOWLEDGEMENTS
Table 5: Total energy input for radiation plates
The authors would like to thank Mr. J.N.M. Akkermans from
1100 Akkermans Gloeitechniek for his interest in this research and
1000 for his support to provide process data and temperature
900 measurement. The authors are also grateful for Mr. M.F.
800 Mendes de Leon from IHC for his interests in the project and
Temperature (ºC)

700 constructive suggestions to this work.

600

500

400
REFERENCES
300
Temperature impeller
measured
200
Temperature radiation 4
Cham, PHOENICS: Parabolic Hyperbolic Or Elliptic
100 inputs Numerical Integration Code Series, http://www.cham.co.uk
0 (accessed in December 2004)
0 2 4 6 8 10 12 14 16 18 20 22
Time (hour) De Jong., R.A. (2004), Modelling of a Heat Treatment
Figure 13: Predicted surface temperature profile by radiation Furnace. Msc. Thesis, Delft University of Technology.
plates heating furnace (4 heating periods with varying energy GT2004A10. 133 p.
input).
Spalding, D. B. (1979), Combustion and Mass Transfer,
Other findings of the design changes are as follows. Pergamon Press: Oxford,. 409 p.
Installation of a heat exchanger will lead to some energy Y. Yang, R. A. de Jong and M.A. Reuter (2004),
savings but the cost of purchasing the equipment necessary “Simulation of combustion and metal heating in a mobile
will be too high to get a short return of investment. heat treatment furnace”, Multiphase Phenomena and CFD
Positioning of the burners and metal will also improve the Modeling and Simulation of Engineering Processes, eds.
energy efficiency but it is expected not to be very high. Laurentiu Nastac and Ben Q. Li, TMS Annual Meeting,
Finally the reduction of excess air is worth looking into. The March 14-18, 2004 Charlotte, North Carolina. TMS,
reduction of excess air will save a lot of energy and therefore Warrendale. p. 429-439.
reduce the cost in production. More details of the
investigation can be found in the thesis by de Jong (2004).

CONCLUDING REMARKS

This paper presents the CFD simulation of the heating

process of a dredging impeller in a cylindrical mobile
furnace, by using commercial CFD package PHOENICS. The
combustion of the fuel through four oil burners was simulated