Jul-2009

Computational fluid dynamic modelling of refinery combustion processes

Computational Fluid Dynamic (CFD) modelling has been widely used for a relatively short period in the refining and petrochemical industries for modelling process heaters, fired furnaces and other combustion equipment such as thermal oxidisers, elevated flares and ground flares.

Dr Richard Martin and Matthew Martin
UOP Callidus

Article Summary

Following is a discussion of some of the factors that created a need for the CFD modelling of fired furnaces in refinery and petrochemical applications and examples of the use of CFD modelling for a large cracking furnace, a thermal oxidiser and a ground flare.

A driving force for the use of CFD modelling for fired furnaces was environmental regulations that mandated the use of low NOx burners. Traditionally, emphasis for burner design had been small compact flame volumes that improved the combustion efficiency and heat transfer for the fired unit. In the early 1970s, environmental regulations began to limit the quantity of nitrogen oxides produced in the combustion process in fired heaters and furnaces. Burner designs were changed to control fuel-air mixing and achieve stoichiometric conditions that would decrease the formation of NOx. The initial low emission burner designs delayed fuel-air mixing by injecting the combustion air or fuel in a secondary stage. Later burner design refinement included methods of mixing combustion products from the radiant section of the furnace with the fuel-air mixture, further lowering the flame temperature and resulting NOx formation. These newer burner designs created larger flame volumes; flames were longer and in many cases larger in cross-sectional area.

Many early applications for low NOx burners were retrofitted into existing furnaces, and as more staged fuel burners were installed it became apparent that some furnaces were no longer capable of meeting previous process capacities. The longer flames increased the temperature of the combustion products entering the convection section creating high tube metal temperatures making it necessary to reduce firing rates and heater process capacity. The flames of ultra low NOx burners are less well defined than the original staged fuel burners. As a result, combustion product flow patterns in the radiant section of the furnace have a significant impact on flame shape. In addition to combustion products at excessively high temperatures entering the convection section, flame impingement or high temperature gas impingement can occur on wall mounted tubes because of the ill-defined flames.

Another potential issue with ultra low NOx burner designs is a decrease in burner stability. The effort to delay fuel-air mixing and decrease flame temperature has created burner designs that have a narrower range of operating conditions for which the burner is stable. There have been several serious incidents resulting from loss of flame in furnaces that were fired with ultra low NOx burners. The issues of flame impingement, reduced furnace capacity, and safety created a need for a better understanding of process burner design and the burner-furnace system. CFD modelling has developed into a useful tool, when properly used, for the advancement of knowledge in these areas. When CFD modelling was first applied to process heater burners, the use was primarily for government funded projects to develop burners that could provide lower NOx emissions. Initially the complete burner-furnace system was not modelled nor was there extensive use of the tool by burner manufacturers for their normal application and design process. With the increased use of ultra low NOx burners, it was soon discovered that the furnace environment plays a significant role in successful burner operation. Over the past ten years CFD modelling for burner-furnace systems that are to be equipped with ultra low NOx or ‘Next Generation’ burners has become common practice. In addition, combustion equipment manufacturers have extended the use of the modelling to other products such as thermal oxidisers and flare systems.

Cracking furnace application
A computational fluid dynamics study of a large cracking furnace was performed to determine any consequences for the installation of 96 UOP Callidus ultra low NOx burners (model: CUBLF-3W-HV). The goal of the study was the evaluation of the combustion-side performance of the furnace; in particular, flame patterns, burner flame to burner flame interaction, burner flame to tube interaction, and large-scale heater flow. There are distinctive flame and heater flow patterns that distinguish between successful and unsuccessful applications of ultra-low NOx burners. Most of these patterns can be seen in the results of a CFD study.

The primary “process” failures of ultra-low NOx burner technology can be grouped into four general categories:
1. Flame collapse The burner flames coalesce to the centre of the heater creating an extended region of poor mixing that increases the burner flame length and creates a local region of increased temperature typically increasing NOx emissions by 20-30%.

2. Tube impingement/flame rollover The burner flames contact the heater tubes. Most low NOx burners have “soft” flames when compared to the flame of a conventional burner. These flames are more easily caught in furnace currents and carried into the tubes.

3. High heat density/plug flow Ultra-low NOx burners rely on combustion product gas re-circulation into the base of the flame to lower NOx emissions. High heat density furnaces may exhibit “plug-flow” with limited recirculation of combustion products to the heater floor. This elevates NOx levels and may cause flame impingement on the process tubes.

4. Loss of flame Ultra-low NOx burners can be less stable than a conventional burner. The steady-state CFD models used to analyse large-scale devices such as heaters are ill equipped to predict flame-out, but if the modeller has a combustion background he is able to recognise certain model results such as flow patterns, component concentrations and temperature profiles that can be indicative of possible burner stability issues.

Model description
Fluent CFD software was used to perform the study. The following models were used in the Fluent simulation:
• The Reynolds Averaged Navier-Stokes equations were used to model flow.
• The standard k-❏ model was used to model turbulence.
• A two-step eddy-dissipation/finite rate oxidation model was used to model the chemistry. Hydrogen combustion was modelled with one step. This model has excellent qualitative and quantitative comparison with single burner data from UOP Callidus’ test facility.
• The discrete ordinates radiation model was used, as it is the only applicable radiation model for the geometry under consideration.