I was working on a particularly problematic heater in 2001 that had been retrofitted with some of the most advanced low NOx burners of the time. Our previous installation made 6 PPM of NOx in the fielded application and the flames looked great. This installation produced 6 times as much NOx and the flames were massed into a large fireball that suddenly pulsed every 30 seconds. The burners were the same, and at the time the root cause was a real mystery. I was tasked with fixing the problem and set our comparatively massive CFD resources to solve the problem: All 12 CPUs strained for days to show that indeed the flames looked bad in the CFD model as well. Now what?
I had a conversation with a Ph.D. candidate at the local university that has stuck with me to this day. It perfectly encapsulated the difference between theory and practice, new knowledge and experience, analytical and empiricism. Upon hearing of the flame problems often experienced in vertical cylindrical heaters he said, “This sounds like an easy problem to solve.”
These words reverberated with me every late night and weekend morning spent modeling heaters for the next 10 years. He proposed that a purely analytical solution could be found from round jet theory. The work of others in the intervening years seems to suggest that part of the solution may lie down this path inquiry, but I contend that there remains more complexity. I have seen repeatedly that many seemingly minor details can make a large difference in the outcome: specific features of the burners, location of the coil inlets and outlets, and convection section offtake shape all seem to appear on the list of design choices with high impact.
The careful work surrounding burner spacing rules that have gone into the forthcoming API 560 standard should vastly improve the chance of successful application by removing the most common issues with burner-flame/heater-coil interaction. For 10 years or more, the better CFD practitioners in the industry can solve most of the flame and heat transfer issues related to steady-state flow patterns using their simulations and a relatively reliable set of design rules. So, what is left?
I am not convinced the steady-state rules are entirely sufficient; they do form an excellent basis from which to proceed to a more detailed study. One of the top fired equipment experts in the industry has suggested for years at AFRC and API meetings that the combustion companies provide transient CFD simulations, a reasonable request given the problems faced in the industry that has fallen on deaf ears. Most fired heater experts seem to recognize that in most instances we are using steady-state approximations to try to solve transient problems. This level of abstraction in our modeling leads to a “reading of the tea leaves” at the conclusion of every study to determine the best course of action.
Transient simulations that calculate the less diffusive turbulence models, combustion, and radiation are still extremely computationally expensive to solve. Depressingly small time-steps are required for time accurate simulations given the mesh requirements to resolve the geometry. Faced with running these models on a realistic project schedule, many wisely choose to run a steady-state model and “read the tea leaves”. Fortunately, since I joined XRG Technologies, I have been able to spend some time pursuing innovation rather than the myriad of other tasks that often consume our day. One of the current innovations is a fast screening tool for fired heaters.
Figure 1 - A vertical cylindrical heater.
Figure 1 shows approximately 10 seconds of animation from a 90-second simulation of a vertical cylindrical heater. By adding to an existing code base, I can produce the simulation input from XRG’s proprietary heater modeling code automatically. There is no manual mesh, so the end-to-end setup time takes as long as it takes me to read the drawings. The total wall-clock time to a solution is about 45 minutes on my laptop.
There is a large amount of geometric simplification for these models, but the large-scale transient problem is solved directly. The tool can be useful for screening heater designs for proper burner placement as well as filling times for furnace flooding events. The heater dimensions are captured with reasonable accuracy, so the effect of the convection section transition shape and burner layout is represented.
This is the first generation of this tool. There is a clear product road map towards improvement – increase the geometric accuracy, and reduce the run times. I also have a few tricks planned to get to the optimal design in a rigorous manner. Hopefully, those of you in the industry can see the utility of such a tool. I am looking for validation cases and any chance to improve the tool – so if you have a particularly difficult heater and want to take a crack at solving flame related issues, give me call so we can look at it…you know, now that we have solved all the easy problems…