TITLE : Smoke exhaust system of a heating furnace

BY : Dr A V Ivanov

DATE : Autumn, 2002 - Spring, 2003

FOR : Demonstration and educational purposes

PHOENICS version : 3.5.0.


A smoke exhaust system is essential for heating furnaces to work efficiently. To put it more exactly for heating furnaces which use gaseous, liquid, powdered fuel or their combination.

In turn, a chimney stack plays a key role in overall performance of a smoke exhaust system. The stack causes smoke to pass through the flues to the stack and discharge to the atmosphere.

This entry is announcing the new project concerned with simulation of the smoke exhaust system, smoke and ambient air flow patterns, heat and mass transfer throughout the smoke exhaust system including heating furnace, flues, stack damper, chimney stack.

The project is relatively large. It has been finished yet. The project data are treated now.


The project was divided into three stage.

Stage A

On stage A the 2D chimney stack was simulated. In fact, the chimney stack represented the rectangular channel. The model implemented in this stage was mostly aimed at the demonstration of main principles of chimney stack working, i.e. it was the model for educational purposes.

Besides various turbulence models were estimated and appropriate model was chosen (K-L model). The estimation of turbulence models was performed by means of comparison of the computed stack base suction with one obtained on the basis of the design technique. This commonly used technique employs a formula deduced from Bernoulli equation.

Turbulence models were estimated under different values of ambient air temperature, smoke temperature, stack height, stack roughness length and stack damper height. Grid independence test has been carried out.

Stage B

Stage B devoted to the simulation of 2D chimney stack with converging channel. The same words concerning educational model are also applicable here.

The interesting point of the case is that there is the overpressure zone within the stack under certain conditions. This zone usually located near the stack mouth. The zone is very harmful for stack work because of combustion materials seepage through the stack brickwork and its consequent failure. Hence natural engineers' wish is to define conditions at which the zone occurs. Some criteria, so called stack aerodynamics operating mode criterion, were suggested to solve the problem. But they were not general enough reflecting mostly particular cases.

K-L turbulence model, KLMODL, as established in stage A, was used to calculate the distribution of the turbulent viscosity on stage B.

Stage C

On stage C the 3D model of the hypothetical smoke exhaust system of the heating furnace was considered. The following entities were included in the simulation:

The task was to compute the flow, pressure, and smoke temperature fields inside the flue system and the chimney stack as well as to reveal fields alterations at stack draught changing.

Along with the main task the following standalone problems were considered and solved:

Stack draught strongly depend on the stack damper position (i.e. on pressure loss induced by stack damper drag). Fine adjustment of the stack damper position would require the very fine grid in the vertical direction within the stack damper travel. So the stack damper was introduced by means of pressure loss source located in the flue near the stack base.

It should be noted that the stack damper is cut in a channel of the furnace heat work automatic control system. This channel is responsible for keeping of zero pressure into the furnace at its floor level. It is necessary to prevent both combustion products escape from the furnace (if pressure is above zero) and cold air suction into the furnace (if pressure is below zero). First one makes worse conditions of work within the workshop, second one increases processing time and makes worse the quality of material heating. Stack damper is allow to set off the effects of draught uncontrolled variations caused by change in outdoor air temperature, fuel composition etc.

K-L turbulence model, KLMODL, as established in stage A, was used to calculate the distribution of turbulent viscosity on stage C. In estimating of the length scale on all stages the distance from the nearest wall (DISWAL) was used.

The anisotropic effects of atmospheric turbulence were taken into account. Neutral atmosphere was examined as an example.

A Disperse Phase Mass Conservation, DPMC, model was used to simulate water vapours condensation within the stack plume. The additional information about DPMC model can be found here.

On this stage conjugate heat transfer was carried out to provide the detailed temperature field within the chimney stack walls. The chimney stack walls were filled with brick.

Buoyancy source term in the airflow equation was introduced through density-difference option. On all stages considered the buoyancy effects within the flue system were neglected.

The heating furnace overall dimensions were as follows: 6.4×3.9×2.7 m.

The chimney stack had the following main dimensions: height - 35 m; outer diameters - 2.75/1.8 m; inner diameters - 2.25/1.5 m.

The flue system length from the furnace to the chimney stack was equal to 113 m.

The cross-section of each flue was rectangular with dimensions 1.0×2.0 and 1.5×2.0 m. The flue system had uniform depth of occurrence - 1.0 m (with regard to the flue symmetry axis).

The workshop overall dimensions were 30.0×15.0×15.0 m, and its gate had average dimensions 3.0×4.0 m.

The flame length of each burner was equal to 1.95 m, with heat rating being 2×2.0 MW.

Combustion products generation rate was equal to 2×2.38 kg/s.

The workshop, the chimney stack and associated entities were exposed to the steady oblique atmospheric airflow with the log-law wind profile and wind speed 1.2 m/s.


Stage A

The computational domain dimensions were 100.0×100.0 m. The basic grid was nonuniform one and it had in total 1763 cells (43 cells in horizontal direction and 41 cells in the vertical direction).

Stage B

The computational domain dimensions were 100.0×100.0 m. The basic grid was nonuniform one and it had in total 2255 cells (55 cells in horizontal direction and 41 cells in the vertical direction).

Stage C

The computational domain dimensions were 100.0×155.0×200.0 m. The basic grid was nonuniform one and it had in total 120120 cells (56 cells high-low, 55 cells east-west and 39 cells in the vertical direction).


The sample plots show the distributions of pressure, velocities, temperature and the other related fields within and outside the smoke exhaust system of the heating furnace.

Pictures are as follows:

Stage A

Stage B

Stage C


Stage A

Overall performance of all considered turbulence models was very good. The discrepancies between values of suction did not exceed 15%.

It was found that neither built-in heat sources switched on not the buoyancy source term activated in the turbulent energy equation had significant influence on the final results.

Stage B

PHOENICS has aided in demonstration and explanation of the physical mechanism of the stack overpressure zone formation.

On the basis of PHOENICS calculations and theoretical considerations a new stack aerodynamics operating mode criterion was suggested and written in the general form.

The critical parameters of the proposed criterion for 2D case considered were established using numerous PHOENICS calculations.

It should be noted, although the proposed criterion can be an indicator of the overpressure zone presence, the exact location of the zone, its extension and even the number of zones can be predicted by means of CFD codes only. Using PHOENICS is a good way to do it.

Application of a "cut-off" technique, PARSOL, for the representation of the chimney stack shape on the Cartesian grid has not revealed significant qualitative and quantitative improvements in computation results compared with those without PARSOL using.

Stage C

Wet plume was regarded as visible if the liquid water specific humidity within it was not less than 10-4 kg/kg.

It was originally assumed that the smoke exhaust system worked properly when the stack damper pressure loss factor was such as MF (mass flow through the flue system, kg/s) and MS (combustion products generation rate, kg/s) to be equal. Considerable efforts were required to gain equality between MF and MS. The results of this series of computations can be sketched as follows:

a) MF<MS; MF+M1+M2=MS

b) MF>MS; MS+M1+M2=MF

c) MF=MS; M1=M2

Outline flow patterns for case a (stack damper pressure loss is large) and for case b (stack damper pressure loss is small) are predictable and understandable. Case c requires some comments. It would seem neither cold air suction into the furnace nor combustion products escape from the furnace occur in this case. However one can see in the picture both processes simultaneously and they compensate each other. Instead of best performance of the smoke exhaust system, twice as deteriorated situation was found.

This effect can be accounted for dissymmetry of the pressure field around the furnace. In turn, this dissymmetry is caused by the asymmetrical flow around the workshop (oblique wind direction) but indirectly through the complex pattern of the workshop natural ventilation. Therefore all factors influencing on the workshop natural ventilation also have an indirect influence on the pressure field around the furnace. Among these factors most important ones can be mentioned: wind speed and wind direction, area and location of workshop intake/outtake openings, outdoor air temperature, equipment lay-out within the workshop. Moreover the number and location of furnace technological apertures are of importance.

Thus, the choice of optimal stack damper position is very complex problem depending on many factors. PHOENICS computations of the smoke exhaust system of the heating furnace can greatly help to determine system best performance, including help in selection of seating of furnace heat work automatic control system sensors.


All model settings have been made in VR-Editor of PHOENICS v.3.5.0.


Prof. B S Mastryukov and Dr S V Zhubrin are gratefully acknowledged. Prof. B S Mastryukov for the long-term scientific supervising and Dr S V Zhubrin for the special attention to the author on initial stages of author's activity as a PHOENICS user.

[Back to the Top]

Rambler's Top100
Сайт управляется системой uCoz