Oct-2002

System analysis of process burners

Modelling is now a vital tool in the designing of ultra-low NOx burners, together with advanced problem-solving techniques such as computational fluid dynamics for troubleshooting and solving potential problems

Carol A Schnepper and Joseph D Smith, John Zink Company
L David Wilson, Marathon Ashland Petroleum

Article Summary

Construction of a new coker unit at a petroleum refinery in Louisiana, USA, necessitated NOx reduction throughout the entire refinery. To achieve the required reduction, the refinery planned to retrofit several crude heaters with ultra-low NOx burners. Retrofitting the burners for NOx reduction presented a unique set of challenges for both the end-user and burner supplier. While conventional burners operate satisfactorily in existing heaters, ultra-low NOx burner installations can prove more challenging in the plenum-burner-furnace system due to issues such as burner-to-burner and burner-to-furnace interactions. In addition, these problems are not usually discovered until after the new burners are installed.

Modelling has become a vital tool in the design and development of today’s ultra-low NOx burners in the hydrocarbon and chemical processing industries. As advanced problem-solving techniques, computational fluid dynamics (CFD) and cold-flow physical modelling have been successfully used to troubleshoot and solve potential problems with ultra-low NOx burner retrofits.

In the present study, the burner supplier’s CFD engineers performed pre-installation modelling of the plenum to determine where problems were most likely to occur for typical operating conditions of the end-user’s furnace. After installation, furnace operators encountered unexpected problems with long, pulsing flames intermittently impinging on the process tubes. Unsteady airflow through the burner plenums, burner throats and furnace proved to be among the challenges faced.

The burner supplier’s modelling team used both CFD and physical modelling to help diagnose and eliminate these problems, improving the operation of the end-user’s furnace and providing a clean, blue, steady flame, as seen in Figure 1.

Besides the work presented here, additional modelling work was done to help improve furnace operation; details of the additional work were reported elsewhere [Wilson L D et al, Retrofitting ultra-low NOx burners in a refinery process heater – a case study; Joint International Combustion Symposium, Hawaii, 9–12 September 2001].

Crude heaters
Among the units being modified were two crude heaters. Each unit is a rectangular, cabin-type heater that shares a common air preheat system. Figure 2 shows the general schematic of the crude heater system. Each heater has eight tube passes and may operate either on balanced or natural draught. Each heater has two rows of 12 burners for a total of 24 burners (burner arrangement shown in Figure 3).

The burners are staggered down the centreline of the longitudinal axis of the heater and are in close proximity to each other. Groups of three burners are arranged in a triangular pattern and share a common air plenum (Figure 3). Combustion air enters both sides of each plenum, and there are eight plenums per heater. Each crude heater is capable of processing 125000bpd of crude oil for a total of 250000bpd. Each heater has a maximum fired duty of 281 million Btu/hr with air preheat at 500 °F.

The new ultra-low NOx burners were specified to meet a guaranteed 0.06 lb NOx/million Btu (higher heat value) at 3% excess oxygen (dry basis) under the given air preheat temperature and maximum fired duty.

Pre-installation modelling
The asymmetry of the burner locations and the ductwork supplying the air led to concerns about the air distribution among the burners in a plenum. The air follows a convoluted path from the ductwork, through the plenum and burners, and into the furnace. Therefore, the burner supplier’s CFD modelling group performed a pre-installation analysis of the plenum to reveal the complex flow patterns.

An earlier in-house CFD study had shown that the flow in any one of the eight plenums was independent of the flow in the other seven. For the present CFD study, the flow patterns in all of the plenums were assumed to be, more or less, the same. Therefore, the CFD engineer modelled a single plenum with three burners and the section of the heater directly above the plenum. The “as-modelled” geometry, illustrated in Figure 4, represents a slice of the overall furnace. Although the model did not capture influences from the end walls, it was deemed a good representation of the plenums located toward the centre of the heater.

Because the purpose of the pre-installation simulation was to examine the airflow distribution, isothermal calculations were performed with non-reacting, cold flow (non-combusting flow) held constant at the temperature of the preheated air. Although the calculated pressure drop across the burners would be higher if combustion reactions in the furnace were included, the flow distribution within the plenum would be minimally affected. Also, as shown in Figure 3, the non-reacting model permitted geometrical simplifications.

The gas tips on the burners and the tubes in the furnace did not participate in the non-reacting, isothermal flow and were thus excluded from the model.

In the pre-installation model, the incoming preheated air was divided equally between the two air ducts. The total inlet airflow rate was based on 3% excess O2 (dry basis) with each burner firing at 14.5 million Btu/hr. The simulation showed that the air was evenly distributed between the three burners in a common plenum to within 3% by mass.

However, velocity vectors, shown in Figure 5 on a horizontal plane passing through the openings in the burner registers, showed a strong recirculation zone between the burners on the right and on the lower left. This and other recirculation zones within the plenum resulted in each burner receiving most of its air through one preferred register opening rather than equally through all four openings.

The highest speeds in the velocity vector plot in Figure 5 clearly indicate the preferred openings. So, while the air distribution between the burners was nearly perfect, the air distribution within any single burner was non-uniform. Nevertheless, the pressure drop across the burners was sufficient to dampen the effect of the flow patterns within the burner registers. At the burner throats, the velocity was highest near the centres of the burners, and the velocity profiles were reasonably symmetric about the centre axis of each burner.

Preliminary troubleshooting
The new ultra-low NOx burners were installed and successfully met the refinery’s NOx requirements. However, shortly after startup of the heater, operations personnel became concerned about the flame patterns. At firing rates of 70 to 80% of burner design, the flames of the individual burners in each cluster would merge. When extended, the flames would “lean” towards the process tubes and the “tails” would break off.