May-2022

Use of tube inserts in fired heaters

In 1896 Whitham reported the successful use of twisted tapes (originally called retarders, and now also called turbulators) to increase heat transfer in boiler fire tubes; the effectiveness for increased heat transfer is well known (Whitham). The predominant use of twisted tapes is to increase heat transfer in laminar flows, but their use in turbulent flows has also been studied extensively (Manglik).

Matthew Martin
XRG

Article Summary

Fired heaters are common in refining and petrochemical plants where the process requires high-intensity heat. State-of-the-art fired heaters are some of the most fuel-efficient devices in use, with efficiencies over 92%. It is not readily apparent how the use of turbulators, which engineers have long applied to increase heat transfer, might benefit such highly efficient systems. The process flow in many fired heaters vaporises as it travels through the heater. By applying purpose-built inserts that outwardly resemble turbulators, one can improve heat transfer characteristics of multiphase flows by addressing non-homogeneity in the process flow regime.

Background: Fired heaters, heat transfer, and heat flux uniformity
Figure 1 shows a typical fired heater. Burners fire into a ‘radiant section’ transferring heat from the flue gas to the process fluid which flows through pipes commonly called ‘tubes’. The process fluid typically receives 60-70% of the heat within the radiant section. The flue gas then flows into the ‘convection section’ which transfers 15-20% of the remaining heat to the process fluid. As indicated by the name of each respective section, the principal mode of heat transfer in the radiant section is through thermal irradiation; the principal mode of heat transfer in the convection section is convective. Tubes in the convection section usually have extended surfaces in the form of fins or studs to provide additional heat transfer from the hot flue gas flow. 

The process flow in practical fired heaters is turbulent, with Reynolds numbers on the order of 106. Most heat transfer occurs within the radiant section where radiation accounts for 90% of the heat transferred. The convection section compensates for small reductions in radiant section heat transfer due to fouling and non-ideal flames; a higher radiant section exit-temperature results in more heat transfer within the convection section. Because the process flow is turbulent resulting in a high tube-side convection heat transfer coefficient, the principal heat transfer mechanism is radiation, and the convection section is designed to result in a specified flue gas temperature exiting stack, there has been little incentive for fired heater designers to increase heat transfer intensity in the fired heater using devices such as twisted tapes. However, the benefit of in-tube mixing enhancement goes far beyond the benefits of increased heat transfer.
Reformers and pyrolysis heaters fall among a class of heater where intentional and valuable reactions take place inside the heater tubes. In reactor charge heater tubes, such as those used in propane dehydrogenation units, undesirable chemical reactions cause in-heater feed conversion and reduce unit selectivity, which reduces the ultimate yield. Many other heaters experience unintended reactions that reduce the value of the product, usually oil, passing through the heater tubes. Heaters used in certain services, such as crude distillation, vacuum distillation, or delayed coking, have both unintended reactions and phase change within the heater tubes. Similar devices to heaters, such as once-through steam generators (OTSGs), do not have chemical reactions within the tubes but do exhibit phase change. In all cases, not only is the total absorbed heat important, but also the location of the absorbed heat — particularly when considering yield, reliability, profitability, and safety.

To see why the variation in temperature along the outside coil surface is critical in heaters with multiphase process flow, consider the idealised graph of heat transfer coefficient versus temperature difference between wall and fluid in Figure 2. Beneath the chart is a representative picture of the liquid/vapour composition within the process coil corresponding to the heat transfer coefficient. The highest temperature at any point in the flow occurs at the boundary between the process coil wall and the process flow. When the combustion heats the process flow, vapour first forms at the interior wall of the pipe. Gases have a significantly lower overall convection heat transfer coefficient when compared to liquids, so the process flow transfers less heat away from the pipe wall. In this way, deleterious feedback ensues wherein the high wall temperature begets more boiling which in turn reduces the inside convection heat transfer coefficient, thus increasing the wall temperature and the resulting boiling. In this way, “hotspots” can form on the heater tubes given an initial slight difference in heat transfer.

Gravity further exacerbates phase non-uniformity inside horizontal sections of process coils. Gravity pulls liquid, being denser than vapour, to the bottom of the pipe, resulting in a strong tendency toward stratified flow with resulting higher temperature on the upper pipe surface.

What is the source of temperature non-uniformity? The major sources result from the unavoidable combination of geometry and physics. As the burner flames and hot flue gas transfer radiant heat to the process coil, the flame-facing surface of the coil receives more heat. This is true for both ‘single-fired’ process coils where the burners are mounted on one side of the coil and ‘double-fired’ process coils where the burners are mounted on both sides of the coil. This maldistribution is known as the circumferential flux factor and has been well-characterised. As combustion releases heat from the burner flame and radiant and convection transfer heat from the hot flue gas there is an additional longitudinal flux factor. Engineers have not been able to characterise the longitudinal flux factor in a general sense because it is heater and burner dependent.  Figure 3 shows an example of single-fired tubes with high longitudinal variation flux and the resulting peak-to-average flux ratio of 2.3. This difference in flux ratio translates to a 61 °F higher film temperature and a 100 °F higher tube metal temperature in example calculations.

Variation in heat transfer to the process fluid comes from both within and without the process coil.  So, what is to be done? Variation in the physical properties of the fluid which are driven by vaporisation inside the coil play a strong role in the local tube metal temperature.  It is also apparent that physics and geometry of the flue gas side drive non-uniformity from the outside the coil. A common method of increasing homogeneity in a process is to increase the mixing of the constituents. One can show through simulation that one can precisely achieve that — increased process homogeneity with the antecedent improvements in reduced tube metal temperature, increased yield, and increased capacity.

Simulation results: Improved area goodness factor for tube inserts
Increasing the mixing inside a process coil comes at a cost — the pressure drop through the system also increases.  One can estimate the process fluid diffusion inside the heater tube by using heat transfer as a surrogate measure.  There are potentially an infinite variety of tube insert designs that one could place within a process coil to increase the heat transfer. To measure the relative effectiveness of various inserts we simulated step-change designs in combination with automated optimisation for thousands of combinations of geometric parameters. We judged the relative merit of each design by comparing the area goodness factor, or the Colburn factor (j) divided by the Fanning friction factor (f). The final optimised design has a 30% increase in area goodness factor compared with traditional twisted tapes. The practical implication of this is that by using an optimised design one can achieve more process mixing with lower pressure drop when compared with traditional twisted tape designs. A key parameter in twisted tape design is the twist-pitch, or the number of pipe diameters required for the twisted tape to make a 180° (or 360° depending on the source) helical revolution, within a length of the tape. Figure 4 shows a comparison in the simulated area goodness factor for both traditional twisted tapes and the optimised design versus increasing twist pitch.

Simulation results: Multiphase flow through a horizontal return
Simulations comparing the optimised design to an empty tube demonstrate that the in-coil mixing translates to increased process homogeneity for multiphase flows. Figure 5 shows the predicted convection heat transfer coefficient over the entire tube surface and liquid volume fraction at the inlet and outlet of the tube for a simulation with 80% liquid and 20% vapour by volume. There are inserts both before and after the return, but in this case the flow requires one straight section of tube to establish the mixing motion provided by the insert. With the tube insert in place the convection heat transfer coefficient is more uniform when compared to the empty tube. Near the inlet, at point ‘A’, gravity stratifies both flows. At point ‘B’ the return has temporarily changed the stratified flow to annular flow, producing a more uniform heat transfer coefficient at the tube surface for both cases. At point ‘C’ the stratified flow returns in the empty tube, but the liquid remains adhered to the tube wall when using the insert.