FLARE SYSTEMS-CONCLUSION SAFETY, NOISE, AND EMISSIONS ELEMENTS ROUND OUT FLARE GUIDELINES

Olavo Cunha-Leite
T-Thermal Inc.
Conshohocken, Pa.

Gas flare systems include many features that enhance safety and reduce atmospheric emissions.

This concluding article in a two-part series on gas flares looks at the proper use of liquid knockout drums to increase flame stability, and liquid-seal drums to prevent flame propagation (OGJ, Nov. 23, p. 70).

Also covered are flare ignition devices-another important safety feature-and vapor recovery systems, which reduce emissions of carbon dioxide and unburned hydrocarbons.

Included in this article are equations for predicting flare noise, and design and operating guidelines for ground flares.

KNOCKOUT DRUMS

Liquids can fuse flare refractory linings and create difficulty in maintaining flame stability. Knockout, or KO, drums are used to separate and collect liquid before the vapor is sent to the flare (Fig. 1).

The flare can handle small liquid droplets, but a KO drum is needed to separate droplets larger than 300-600 m, (typically, larger than 400 m). KO drums should be provided with groups of process units that are to be shut down together.

KO drums are either horizontal or vertical, and can have a variety of configurations and arrangements, depending on economics. The drums must be of sufficient diameter to effect the desired liquid/vapor separation. On a horizontal drum, a split entry or exit is often used to reduce the size of the drum for large flows (Fig. 2).

Vertical drums come with a tangential inlet nozzle and a cylindrical baffle, giving a swirling effect that improves separation (Fig. 3). These drums can be incorporated at the base of the stack.

A reduced-size KO drum should be provided close to the flare when the flare line serves more than one unit, or when the distance from process unit to flare exceeds 600 ft.

Generally, KO drums are designed based on American Petroleum Institute (API) Recommended Practices (RPs) and should have a design pressure of 50 psi to resist explosion. A minimum corrosion allowance of 1/16 in. should be used with carbon steel vessels.

Typically, a storage capacity of 500 gal, plus a liquid holdup capacity of 20-30 min release, should be provided in any drum. The drain system should be equipped with two pumps. Each should be sized to empty a half-full drum within 2 hr and to accommodate a minimum capacity of 25 gpm.

HORIZONTAL DESIGN

The liquid dropout velocity is given by Equation 1 (see equations and nomenclature boxes). The value of the constant "k" is primarily a function of liquid particle size and drag coefficient.

For a 400-m particle, API RP 521 uses a k value of 0.236/C0.5, where C is the drag coefficient. A widely used k value, however, is 0.417. For critical service, the value 0.21 is widely used, regardless of the droplet size. This results in drums very close to API design.

The vapor velocity, Vx, that allows the droplets enough time to settle out before the vapor leaves the drum is determined by Equation 2. The gas residence time is calculated as shown in Equation 3.

The time required for the droplet to settle out must be at least equal to the residence time (Equation 4). Solving Equation 4 for Vy yields Equation 5.

Replacing Vx by its expression (Equation 2) and solving for Dd produces Equation 6. The Ln/Dd ratio for this type of application is generally close to 2.5. Consequently, Dd can be defined as shown in Equation 7, using fy = fa = 0.5 as a good start for this trial-and-error sizing.

When the vessel diameter exceeds 10-12 ft, a split entry or a split exit should be considered. This will reduce the vessel diameter by a factor adequate to reduce the flow rate by half.

The two inlet (or outlet) nozzles should be placed a distance at least 1.25Dd away from the outlet nozzle. Additionally, if there is a secondary drum at the base of the flare stack (i.e., water seal or secondary KO drum), the KO drum diameter is reduced again by another factor corresponding to double the design velocity.

VERTICAL DESIGN

The dropout velocity is given by the same equation as for the horizontal design, although the value of k is different for some cases. API RP 521 uses the same k value (0.21) for both horizontal and vertical designs with the inlet stream baffled to direct the flow downwards and tangential. Both values are based on a 400-m droplet size.

The diameter, Dd, of the vertical KO drum is defined as shown in Equation 8.

For a centrifugal-type separator, the gas enters tangentially and spins around a long center tube-extended 2 1/2 nozzle diameters below the inlet nozzle-and a "turn-around" section between the center tube and the maximum liquid level of at least three nozzle diameters.

The vessel diameter ranges from 2.8 to 4.2 times the diameter of the inlet nozzle, and the center tube diameter ranges from 1 to 2 times that diameter.

For this design, k is 0.4 for normal service and 0.2 for critical service. For designing the inlet nozzle velocity one can use a k of 3.6, and for the center tube, 0.9 to 3.6. The higher the centrifugal velocity, the higher the separation force will be.

To avoid reentrainment, velocity should remain low in the turn-around section.

If there is another KO drum upstream, the velocity can be doubled, resulting in a reduction in the drum diameter. Additionally, if the upstream drum is a split-entry (or exit) type, the velocity will be increased again by the same factor. This consequently , results in another reduction of KO drum diameter.

LIQUID-SEAL DRUM

Flame arresters are subject to plugging, in addition to not being effective, when exposed to any liquid that might congeal. The most effective method to prevent flame propagation into the flare system is the installation of a liquid seal drum (Fig. 4).

A liquid seal can also be used as a back-pressure device to maintain positive pressure in the flare header. The main danger in the use of liquid seals is the possibility of freezing, which blocks the flare systems. Steam coils are often applied to prevent this blockage.

In normal operation, the waste gas bubbles up through the "V" notches at the dipleg bottom. The back pressure in the vent header must be higher than the dipleg pressure so that the waste gas will flow through the seal.

The liquid-seal drum is designed to act as a final or secondary KO drum for separating liquids from gases. Also incorporated are special antisloshing baffles which provide plenty of viscous damping. Notches are also used on the end of the dip pipe to increase the flow area and thus minimize surge.

Continuous flow of water at 20 gpm should be maintained to constantly skim the hydrocarbons and maintain the seal-water level. Seal drums generally are designed based on API RPs, with a design pressure of 50 psi to resist explosions and a minimum corrosion allowance of 1/16 in. when carbon steel is used.

Another type of liquid seal is the displacement type. In this design, the waste gas forces the liquid level down to the dipleg notches. The liquid then bubbles up through the center leg.

The displacement seal is not affected by the water sloshing. This results in pulsation-free combustion. With a ground flare and elevated flare system, a single drum with different dipleg depths can provide the required double seal and staging (Fig. 5).

LIQUID-SEAL DESIGN

The maximum distance, L, that the dipleg is submerged is proportional to the maximum exit back-pressure allowable in the vent (Equation 9). Typical seal depths are 24 in. for elevated flares and 6 in. for ground flares, measured from the top of the V notches to the liquid level.

The diameter of the vertical drum should be at least twice the inlet or dipleg diameter. This ensures that the free area above the liquid will be at least triple the inlet cross-sectional area to prevent surges of gas flow to the flare.

To act as a final or secondary KO drum, the liquid-seal drum diameter should be adequate to provide enough disengaging space. The height of the vapor space should be at least one vessel diameter or a minimum of 3 ft.

The dropout velocity, Vd, is defined by Equation 10, where, for this case, k is 0.9 for normal service and 0.45 for critical service. Also, a constant vapor velocity of 12-15 fps is widely used for all conditions.

The vapor velocity through the dipleg should be based on k = 3.6. Occasionally, it is necessary to increase the dipleg diameter to reduce the gas velocity and allow enough room for the slots.

IGNITION SYSTEMS

Manual ignition is achieved by manually operating a pushbutton to energize the transformer and spark plug fitted to the ignition tube.

During ignition only, 150 scfh of ignition gas is required, plus approximately 1,500 scfh of air and 0.3 kva of electric power.

FLAME-FAILURE PANEL

Each pilot carries a separate thermocouple which is attached to the pilot line and enters the pilot nozzle base. The pilot flame-failure panel monitors the signals from the thermocouple installed in each pilot, on a continuous basis, by sensing the circuitry. Should any pilots fail, a local panel alarm lamp will illuminate. Usually, a remote alarm can be connected.

Electrical circuitry typically is housed in a weatherproof (NEMA 4) enclosure. An explosion-proof enclosure (Class 1, Gr. D, Div. 1) can be supplied, as well as an air-purge enclosure, to meet Class 1, Gr. D, Div. 2.

AUTOMATIC IGNITION

In case of pilot flame failure, the signal from the thermocouple energizes the relay, which in turn traps the on/off timer and solenoid valves on the fuel gas and air supply lines. The ignition transformer will be energized about 4 times/min, providing 15 sec for purging the ignition line between attempts.

This sequence will continue until all pilots are lit, thus closing the solenoid valves, stopping ignition, and turning on the green lights. If any pilot does not light, the time cycle ends (1-5 min) and an alarm signal is sent to the control room or a remote alarm.

OPERATION

The flare tip is fitted with continuous pilot burners to ensure ignition, regardless of wind conditions. The pilots are ignited by a remote front-flame generator (Fig. 6). ("Front flame" refers to a moving ball of fire that is used to remotely ignite the flare pilot.)

Fuel and air are fed via needle valves, nonreturn valves, and restriction orifices, to a mixing igniter tube, where a spark ignites the mixture. The mixing igniter tube is connected to the pilots by the ignition line, in which the front flame travels and lights each of the pilot burners. The pilot flames are blue and stable, sometimes making it difficult to see if they are lit in daylight. Each pilot can be provided with type-K thermocouples to monitor pilot flames and warn of pilot flame failure.

To operate the ignition system safely:

1. Completely purge the flare system with natural gas or nitrogen.

2. Open the air and gas needle valves and set the pressure of both at 10 psig.

3. Open the pilot fuel gas valves.

4. Purge all lines for 2-3 min.

5. Spark to light the mixture. The front flame will be seen as a blue-yellow flash in the igniter sight port.

6. If the pilot does not light, purge and spark again.

7. If the pilot does not light, adjust air needle valve and repeat Steps 4 and 5.

8. When all pilots are lit, close the air and gas needle valves, keeping the pilot gas open. It is good practice to mark the pressure setting because it depends on fuel gas gravity and heat content.

9. After all pilots are lit, open the flare gas valve.

Although these systems do not require extensive maintenance, the following procedures should be performed:

1. Drain the condensate from all ignition, air, and pilot lines.

2. Check and reset the igniter electrode.

3. Check the fuel gas and air supply pressures.

4. Clean the igniter tube.

5. Clean the air inspirator jets on pilot assemblies.

6. Check the thermocouples and wiring.

GROUND FLARES

Ground flares are used to conceal the flame and reduce combustion noise, both of which are frequent sources of complaints from neighbors. Ground flares use single or multiple burners inside a steel refractory-lined open enclosure (either a round or rectangular open area) designed to operate smokelessly at only a fraction of the elevated flare maximum capacity, typically about 10% (Fig. 7).

Larger relief flows normally are diverted to the elevated flare by the use of a liquid seal. The elevated flare is used mainly for emergencies. The ground flare capacity should be designed to reduce elevated flare use to 10-20%.

Ground flares usually consist of several manifold-mounted flare burners placed near the bottom of the enclosure shell. Some designs use a larger center flare burner-air or steam-assisted-to achieve smokeless combustion.

When using multiple burners, staging helps maximize the turndown while maintaining smokeless operation. A burner is staged by using liquid seal diplegs with different submergences or by mounting pressure-switch/control-valve assemblies.

A staged ground flare divides the large mass flow through the multiflare burners, achieving better mixing with atmospheric air. This results in a short, smokeless flame.

The staged system requires as much as 15 psig available pressure to achieve high smokeless turndown capacity without the assistance of steam, water, or forced-draft air.

A three-stage multiburner typically is created by the use of pressure-switch/control-valve assemblies. The first stage is always open without automatic valve control. As the flow increases, the pressure switch will activate the solenoid to open the second-stage control valve.

At this point, all first and second-stage burners become operational. As the back pressure builds up, the pressure switch will energize the solenoid to open the control valve on the third stage.

Each control valve should have a bypass line with a rupture disc for safety. Generally, these valves are butterfly valves with an actuator mounted for fail-open position, thus providing safety upstream on the waste gas header.

Mounting the actuator for the fail-closed position will close the valves, and all pilots will fail, thus ensuring that unburned gases do not create an explosion when ignited by the hot refractory.

The average height of a multiburner ground flare enclosure is about 32 ft, mounted 3-5 ft off the ground. This generates low dispersion, which increases the grade-level concentration of pollutants, especially if some sulfur is present in the waste gas.

At least one pilot should be provided near the burner of each subheader stage. Pilots need to be designed for continuous service, incorporating a type-K thermocouple to send a signal activating the pilot failure alarm.

If all pilots fail, a signal is sent immediately to close the upstream valve mounted on the main ground-flare header, thereby shutting down the system for checking the pilots.

The air-assisted ground flare system is based on a center air-assisted flare burner, which ensures proper mixing of gas and air to give complete, smokeless combustion in a short, vertical flame with low heat radiation. Both high and low-pressure gases can be handled in this combustion system.

The waste gas is connected to the flare burner outer body through a side inlet. A fraction of the air required for combustion is introduced by a forced-draft fan connected to the center body of the flare burner. This ensures a very stable flame and creates a highly turbulent mixing zone.

To reduce power consumption, the air blower should be equipped with a two-speed motor, using only one eighth of the connected horsepower at low speed. A pressure switch in the flare header can be used to change the fan speed to high or low, according to flow rates. The remainder of the air necessary for complete combustion is introduced by natural draft through the existing gap on the enclosure bottom.

For this type of ground flare, the shell diameter and height are both a function of total heat released.

The major shortcoming of the ground flare, aside from its high cost, is its difficulty in controlling large variations of waste-gas flow. This makes it necessary that the conventional elevated flare work together with the ground flare, on demand, for release of larger flows.

Pulsation can cause the flame to go out with low flow. To prevent this, a positive back pressure can be generated by the use of a liquid-seal dipleg. This auxiliary equipment will also prevent flashback.

The enclosure shell is lined on the entire length with refractory and insulating material, keeping the steel skin temperature slightly above the dew point. A 5-8 ft high windshield fence should be placed near the enclosure. Usually, it is octagonal, with louvers that avoid direct wind impingement of the flare burners.

Flame lift-off is prevented by providing the burner tips with flame-retention devices, which also improve mixing of the air and waste gas.

Care should be taken to disentrain any liquid slugs from the gas stream because they can fuse the refractory. Liquids in the burner header also can create serious problems in maintaining flame stability.

Combustion efficiency is high in ground flares, and fluid gas analysis has shown an average of 165% excess combustion air. Ground flares also achieve combustion noise reduction of up to 15 db.

GROUND FLARE DESIGN

The inside open area of a multiburner ground flare is based on the total heat release at the design flow rate, It is common to design the enclosure according to the following rule: 3-4 MMBTU/hr-sq ft of enclosure open area.

For a multiburner ground flare, the height of the enclosure is independent of the total heat release. It is about 32 ft, mounted 3-5 ft off the ground.

For a cylindrical shell, the inside diameter of the refractory lining, D, (ft), is given by Equation 11.

The multiple flare tips discharge vertically from the subheaders (manifolds), which are connected just outside the enclosure to a large header. The number of flare tips, N, is based on the maximum heat release-as much as 5 MMBTU/hr for each burner (Equation 12). Multiple flare tips are placed on a 12 in. minimum pitch, and with 12 in. clearance from the inside wall.

To calculate N for a cylindrical shell, Equations 11 and 12 can be combined to produce Equation 13.

SINGLE TIP DESIGN

In a single air-assisted flare tip design, the inside diameter and the length of the ground flare refractory are based on the flame length and width. The flame dimensions are a function of the heat release, Q, expressed in BTU/hr (Equations 14 and 15).

The air-assisted flare tip is designed according to smokeless flare design procedures. Clearance should be given between the flame and wall to avoid any flame impingement from the refractory.

VAPOR RECOVERY SYSTEM

A vapor recovery system should be considered when continuous flow exceeding the purge requirements is flared. The exception to this is when there is a feasible change in the process that would eliminate the need to send the waste gas to the flare.

Continuous wasting of gas is expensive. The costs of cleaning, storing, and recompressing the gas are also high; however, the long-term payback of vapor recovery (approximately 2 years) is worthwhile.

The system also improves the environment by preventing the release of carbon dioxide and unburned hydrocarbons. The compression station generally utilizes a screw compressor, a water seal drum, a knockout drum, and an automated load-control system. The vapor recovery system must be designed to avoid interference with the flare relief which is basically a safety system.

A liquid seal should be used on the suction side of the compressor to increase inlet back-pressure, thus preventing the compressor from sucking down air from the flare stack. Also, its dipleg depth should be designed to relieve excess waste gas by breaking the seal and directing the gas to the flare stack tip. The liquid seal also will prevent flashback and protect the upstream system.

The flare stack should be continuously purged to prevent air ingress in the flare system. A gas seal, fluidic or molecular-type, should be installed to reduce the amount of purge gas (Figs. 8 and 9).

Because waste gas flow rates and compositions are variable, the vapor recovery system must be designed to handle the varying conditions of normal operation.

Gas recovery is a simple process (Fig. 10). The gas is aspirated downstream of the flare liquid seal. It then runs through a KO drum to remove liquid. Dry gas goes to the compressor inlet after passing through the inlet control valve.

Compressed gas is discharged and sent to cooling heat exchange, then to a second KO drum to further separate liquid particles before the gas is stored. If the gas contains a significant amount of hydrogen sulfide, it can be blended with other sour gases and fed to a gas treater unit.

Coolant should be added to the compressor to prevent the compressed gas from reaching temperatures higher than the compressor will allow. The coolant is condensed in the heat exchanger downstream of the compressor after separation in the KO drum. A bypass control loop should be installed to avoid high suction-pressure variations caused by variable waste gas discharges. The operation of a vapor recovery system requires minimal involvement of both process and maintenance personnel. Under normal operation, there is not enough pressure to break through the liquid seal, consequently all gas is being recovered.

Under upset conditions waste gas flow exceeds compressor capacity, breaking the liquid seal. Excess waste will be flared, which keeps the compressor recovering at its design flow. The liquid seal will reset itself while the compressor keeps operating.

To have a high recovery rate, the compressor station should have a design capacity two to three times the average normal flare flow.

COMBUSTION NOISE

The prediction of flare combustion noise is important in determining placement of the flare.

The thermal power of the source, WT (in watts), is defined by Equations 16 and 17. The acoustic efficiency for burning a typical hydrocarbon on a standard flare tip is estimated to be approximately 5 x 10-8.

The sound power level, PWL (db), of the combustion noise can be calculated by Equation 18. Assuming free field (i.e., spherical propagation of sound waves), the sound pressure level, SPL (db) at distance Rs from the sound source is given by Equation 19.

For distances greater than 100 ft, SPL needs to be corrected by the atmospheric absorption. Consequently, the combustion noise sound pressure level, in db, will be estimated by Equation 20, at a distance Rs ft.

The heat value can be assumed constant-approximately 20,000 BTU/lb-for a mixture of hydrocarbon gases. The noise level expressions can be simplified as shown in Equation 21.

VENT NOISE

Vent noise is caused by turbulent mixing of waste gas with atmospheric air. For a typical hydrocarbon, the acoustic efficiency for venting at a velocity of 0.5 Mach or less on atmospheric air, is estimated to be approximately 3 x 10-6.

The sound power level is estimated as shown in Equation 22. Applying 90 directivity (-6 db), the sound pressure level on free field can be approximated by using Equation 23.

FLARE NOISE SPECTRUM

The noise level at each octave band frequency is shown in Table 1 for the resultant of the combustion noise and the vent noise. The resultant noise spectrum was calculated by logarithmic addition of both noise source spectra.

The spectrum on the different octave band frequencies can also be established based on the combustion noise, as shown in Table 1, because the overall resultant noise has the same value as the combustion noise. Although the combustion noise is 16 db higher than the vent noise, because of the spectrum and peak frequency, the vent noise is predominant at frequencies greater than 500 hz.

The values calculated using the combustion noise levels and spectrum described here are very close to some experimental data and published papers.

A-weighted frequency correlations were designed to approximate the loudness sensitivity of the human ear. Adding the A-weighted correction factors (Table 2) to the noise spectrum calculates the SPL at each frequency, in db(A). The combination of two or more noise levels is done by logarithmic addition.

Combining the A-weighted octave band frequency noise levels results in the level defined by Equation 24. By comparing Equation 24 with the equivalent expressions for linear db (not A-weighted), the spectrum shape allows the relation shown in Equation 25 to be established.

STEAM INJECTION NOISE

Any steam-assisted flare generates noise caused by the high-pressure steam jets and injectors, aside from the combustion noise of the hydrocarbons.

High-pressure steam generates a typical high-frequency jet noise with or without combustion. Steam injection also increases combustion efficiency, consequently increasing the energy release and burning rate, which contribute to a higher combustion roar.

The overall noise level of a steam-assisted flare, therefore, has three components:

• Low-frequency combustion roar (SPL combustion)

• High-frequency steam jet noise (SPL steam)

• Change in combustion roar with steam flow at constant hydrocarbon gas flow (SPL relative).

Generally, the noise spectrum below 355 hz is dominated by the combustion noise plus the steam effect on the combustion roar.

STEAM JET NOISE

Above 355 hz, the noise spectrum follows only the steam jet noise values, without any influence from the combustion noise. The acoustic efficiency that affects the mechanical stream power is 3 x 10-3 , assuming 100 psig steam pressure.

Using multipart injectors with smooth angular drilling (which generate less turbulence), the acoustic efficiency value becomes approximately 2 x 10-4).

STEAM NOISE SPECTRUM

To find the overall resultant noise spectrum for a steam-assisted flare, both steam and combustion noise spectra were combined (Table 3). The overall sound pressure level on the "A" scale is calculated by combining the values of all octave band frequencies, after being A-weighted.

Generally the steam-togas weight ratio varies between 25% and 100%. If a ratio of 50% is assumed, the expressions below 355 hz can be replaced with steam weight as the new variable, giving a small error of +/- 1.5 db, and the overall noise level can be estimated by Equation 26. Applying the A-weighted scale results in Equation 27. Comparing both O.A. SPL equations, Equation 28 can be established to calculate the A-weighted sound pressure level.

The noise-prediction calculations provided here can help determine the least intrusive location for the flare, which may result in fewer complaints from plant neighbors.

BIBLIOGRAPHY

American Petroleum institute, "Manual on Disposal of Refinery Wastes," API931, 1977.

API Recommended Practice 521, 2nd edition, September 1982.

Alcazar, C., and Amillo, M., "Get Fuel from the Flare," Hydrocarbon Processing, July 1984.

Allen, G., et al., "Flare-gas recovery success at Canadian refineries," OGJ, June 27, 1983, p. 79.

BASF trade literature.

Brzustowski, T.A., "Flaring in the Energy Industry," Proc. Engr. Combust. Sci., Vol. 2, 1976.

Clark, L., Manual for Process Engineering Calculations, McGraw-Hill Inc., 1975.

Flare-gas Corp. trade literature.

Gerunda, A., "How the Size Liquid Vapor Separators," Chemical Engineering, May 4, 1981.

Kaldair Ltd. trade literature.

Kent, G.R., "Practical Design of Flare Stacks," Hydrocarbon Processing, August 1984.

Miller, P.D., et al., "The design of Smokeless, Non-Luminous Flares," API Proc., 1958.

Leite, O.C., "Predict Flare Noise and Spectrum," Hydrocarbon Processing, December 1988.

Leite, O.C., "Practical Flare Sizing," Pollution Eng., June 1989.

Leite, O.C., "Smokeless, Efficient, Nontoxic Flaring," Hydrocarbon Processing, March 1991.

Powell, D.T., and Schwartz, R., "Operating Experience with a Low Level Flare," API Proc., 1972.

Seebold, J., "Practical Flare Design," Chemical Engineering, December 1984.

Seebold, J., "Flare Noise: Causes and Cures," Hydrocarbon Processing, October 1972.

Swithenbank, J., "Ecological Aspects of Combustion Devices," AIChE Journal, May 1972.

Smolen, W.H., "Smokeless Flare Stacks," Hydrocarbon Processing, September 1951.

Straitz, J.F., "Make the Flare Protect the Environment," Hydrocarbon Processing, October 1977.

Tan, S.H., "Flare Systems Design Simplified," Hydrocarbon Processing, January 1967.

Vanderlinde, L.G., "Smokeless Flares," Hydrocarbon Processing, October 1974.

Copyright 1992 Oil & Gas Journal. All Rights Reserved.