Startup Stability Diagrams One situation where the author and others (26, 28) found the stability diagrams particularly useful is for addressing startup conditions. Here, the above-listed limitations are usually not an issue, as the limits are few and high accuracy is not required.

In many startups, vapor is introduced to the tower before the liquid. This vapor will flow both through the trays and the downcomers. If the vapor velocity through the downcomers exceeds the system limit (based on the downcomer area), it will not allow liquid descent into the downcomer and cause flooding in the tower. This is a downcomer unsealing flood, as described in Section 2.11. An attempt to introduce liquid before the vapor is unlikely to help, because in the absence of vapor the liquid will dump through the tray openings and bypass the downcomers without sealing them. The startup stability diagram, showing both the critical vapor velocity (downcomer unsealing flood limit) at which the vapor will prevent liquid descent in the downcomer and the liquid dumping limit, is invaluable for defining the stable region on the vapor–liquid map at which neither limit will be restrictive and the tower can be started up. Once started up, the tray liquid will seal the downcomers, and the critical vapor velocity will no longer be a limit. Successful applications of this technique have been reported (26, 28, 190).

Excessive Column Pressure Drop The liquid accumulation that characterizes flooding incurs high pressure drop. In general, a normal tray pressure drop is 4–5 in. of liquid. For most organic and hydrocarbon systems whose specific gravity is around 0.7–0.8, this gives 0.1–0.14 psi per tray. If the measured pressure drop per tray is 0.15 psi per tray with an organic or hydrocarbon system, possible flooding should be suspected. Once the measured pressure drop exceeds 0.2 psi per tray and the measurement is correct, flooding is highly likely.

For aqueous systems, as well as other higher-density systems like chlorosilanes with a specific gravity of about 1.0, the normal pressure drop can be somewhat higher (about 0.18 psi/tray), possible flooding occurs when the measured pressure drop is above 0.2 psi/tray, becoming highly likely when the measured pressure drop exceeds 0.25 psi per tray, and the measurement is correct.

With packed towers, the flood pressure drop is given by the Kister and Gill Equation (193, 220, 245, 462)

(2.2)

where

• ΔP_flood = flood pressure drop, inches water per ft of packing
  
• F_p = packing factor, ft⁻¹

The values of packing factors to be used in Eq. 2.2 should be taken from the 9th Edition of Perry’s Handbook (245). Packing factors originating from other sources may not be compatible with Eq. 2.2 and can lead to inaccurate or even incorrect predictions. Measured pressure drops significantly higher than the value predicted from Eq. 2.2 indicate flooding.

While a high pressure drop (per the above criteria) almost always signifies flooding, many towers flood, yet the pressure drop remains low. The pressure drop rise is due to liquid accumulation. When the liquid accumulation is small, the pressure drop rise is small. Typical scenarios include flooding near the top of a tower (only a short packed length or a few trays accumulate liquid), flooding with channeling (the channeled vapor bypasses the accumulated liquid), flooding in vacuum-packed towers (accumulation is channeled and the vapor bypasses the accumulation zone), and flooding at low liquid rates (slow rate of liquid accumulation).

Techniques for pressure drop measurements are discussed elsewhere (4, 54, 192, 410).

Watch out for the dynamic velocity head of a vapor feed or reboiler return adding to the pressure drop reading. In Case 4 in Section 8.16, the dynamic velocity head of the vapor–liquid reboiler return mixture converted into static pressure at the tap, giving a reading of 165 psig, compared to 160 psig at a neighboring tap. Likewise, the pressure drop across the wash bed in a refinery vacuum tower (335) declined from 3.1 mmHg (indicating flood) to 0.7 mmHg (indicating normal operation) when the pressure measurement tap was switched from the feed (flash) zone to a point above the chimney tray just below the bed. The high pressure at the feed zone tap was due to impingement of the high-velocity vapor/liquid feed.

Watch out for reranged differential pressure transmitters in towers where trays were replaced by low pressure drop internals like packings (166). The accuracy of a reading is a fixed percentage of the range of the sensing element. Reranging the output signal to go full scale at a lower value increases the fractional error. Lower-range transmitters should be used instead.

Closely audit the location of the pressure drop points, especially when trays are replaced by packings. In one revamped tower (442), a pressure drop of 1.1 psi was measured across the top packed bed, more than double the design. The lower pressure tap was in the overhead vapor line. The real pressure drop across the bed was only 0.4 psi, while the remaining 0.7 psi was the entrance loss into the overhead vapor pipe.

In high-pressure (>50–100 psig) packed towers, depending on the measurement method, the vapor static head between the lower and upper taps may need to be subtracted from the pressure drop reading. This pressure drop may exceed the packing dynamic pressure drop (4, 54, 512). In dP cells, the gas filling the measurement leg may have a different density than the vapor in the tower, and the difference would affect the pressure drop measurement. A detailed procedure allowing for both of these effects is very well illustrated by Xu et al. (512) and by Cai and Resetarits (54).

Sharp Rate of Rise of Pressure Drop A sharp rate of rise of pressure drop with vapor load is a more sensitive flooding indicator than the magnitude of pressure drop. Pressure drop rises with increasing vapor loads. Upon flooding, the pressure drop rise is largely enhanced by the liquid accumulation. In many cases, once started, the pressure drop will keep rising even when vapor loads are no further raised.

The flood point can be inferred from a plot of pressure drop against vapor flow rate. It is the point where the slope of the curve steepens (Figures 2.5 and 2.6). In tray columns, the slope change can be relatively mild (curve 1), or relatively steep (curve 2), as illustrated in Figure 2.5. A mild slope is more common for jet (entrainment) flooding or a small number of flooded trays, while downcomer flooding or flooding that propagates throughout several trays tends to give a steeper slope. It is not unusual to find a vertical slope once the flood point is reached (119, 185).

In packed towers, inferring the flood point from a pressure drop versus load curve (Figure 2.6) is generally more difficult, because the slope begins changing at the loading point (point B), and the change may be continuous (curve BCD), rather than sharp. Further, many packed towers experience a rapid drop in efficiency well before reaching the hydraulic flood point. Here, the throughput limit is the loss of separation, and the hydraulic flood point may be of lesser practical significance. There are some cases of flood, especially in vacuum distillation, where no point of inflection is observed (232).

For the best definition of the rate of pressure drop rise, it has been recommended (187, 192, 319) to install differential pressure recorders across each section of the tower prior to troubleshooting and flood testing. This technique (Figure 2.7) can clearly identify the region of flooding initiation. In one case (204), applying this technique to identify the flood initiation zone prevented misdiagnosis and unnecessary shutdown. In another case (363, 383), it conclusively confirmed that the flood initiated in the bottom section of a column.

The open circles in Figure 2.8 show the multitude of points where pressure transmitters can be installed. Pressure drops between judiciously selected pairs of points can concisely identify the flood location and nature. In one case (21), at upset conditions, the top packed bed pressure drop fluctuated between 0.24 and 1.75 psi, while the pressure drop between the bottom tray and the top of the tower cycled at the same frequency between 3 and 4.7 psi, roughly the same amount. This proved that the pressure drop increase was in the top packed bed, not on the trays below. There was a pressure measurement point (not marked on Figure 2.8) in the top P/A (pumparound) liquid draw line from the collector tray, which gave a pressure drop of over 2 ft of liquid across the collector tray (the risers were only 10 in. tall), proof that the collector tray was liquid-full and backing liquid into the packing. The collector tray would fill up only if the downcomers of the upper trays below were plugged and backing up. This concise identification of the flood location and nature led to a solution that eliminated the flood without tower shutdown.

A similar technique using a multitude of pressure measurement points enabled Öztürk et al. (354) to diagnose a flood zone between the kero draw and the top in another atmospheric crude tower. Similarly, Barletta and Kurzym (20) used two high-accuracy electronic gauges to identify the top pumparound and the heavy naphtha pumparound to be the two flooded zones out of six zones in an atmospheric crude tower. They took pressure readings simultaneously to ensure column pressure variations would not influence the measured pressure drops.

If it can be afforded, a multichannel pressure drop recorder has been highly recommended (166, 319) and successfully applied (119, 319) for flood testing and diagnosis. This instrument can trace the sequence of events, zero in on the location of flood initiation, the conditions when it occurs, how the flooding propagates, and which remedial action is working. This instrument is invaluable for identifying flooding at the top tray (e.g., due to plugging), when the pressure drop rise may be too small to be noticed on the section or tower differential pressure. Figure 2.7 (319) shows an application of this technique, depicting step-by-step initiation and propagation of flood and, later, of the tower getting out of flood, clearly identifying the location and timing of each step.

Onset of Fluctuations in Pressure Drop, Bottoms Level, and Tray Temperature This symptom is less common than the previous two, and may be caused by factors other than flooding. It signifies liquid accumulation above a certain point, and when enough liquid builds up, the liquid rapidly dumps. In the Figure 2.8 case cited above (21), at stable operation the top packed bed pressure drop was 0.2 psi, but upon an upset fluctuated between 0.24 and 1.75 psi. The dumping step is usually faster than the buildup step and is accompanied by cooling down below the flooded zone.

Figure 2.9 shows pressure drop fluctuations due to foam flooding of a chemical tower (516). The tower contained a packed bed above the feed, with trays below the feed. Both the trays and packing pressure drops fluctuated, the fluctuation amplitudes varied 6–12 mbars (0.1–0.2 psi). The average tray pressure drop aligned with the design, while the packing pressure drop was higher than expected. Gamma scans showed intermittent loading and unloading of the packed bed, no tray flooding, and signs of foaming throughout. A thorough investigation determined definite foaming in the tower (516). Senger and Wozney (422) reported similar pressure drop behavior “pulsating flood” in some of their tests with another chemical foaming system in a 12 in. ID tower that flooded from the random-packing support up.

Pressure drop fluctuations are not unique to foam-flooding. Hanson and Lee (148) report cycles of pressure drop dumping and building resulting from partial trays plugging in a non-foaming service. These were accompanied by temperature excursions and bottom level loss. The author experienced similar cycles in a similar service.

The pressure drop fluctuations are often accompanied by bottom level fluctuations that tend to be out-of-phase with the pressure drop fluctuations. A pressure drop peak represents liquid accumulation and will correspond to a bottom level valley. Similarly, a pressure drop dive represents a liquid dump and will coincide with a bottom level rise. In packed towers, the pressure drop peaks tend to occur simultaneously with the bottom level valleys, and vice versa. Trays have time constants of the order of 0.1 minutes per tray, so the pressure drop and bottom level peaks tend to be out of phase.

Figure 2.10 shows operating charts from a large chemical column (298). Here, the bottom sump level and tray temperature 10 trays above it (“tray N”) showed very repeatable oscillations at a fixed amplitude and a period of several minutes. The amplitude was small at incipient flood, steeply rising upon the onset of full flood. The plant added an incipient flooding indicator that detected the small-amplitude fluctuations and warned the operators. Adding this indicator reduced the average number of flood episodes per year from 12 to 2 and raised tower capacity by 5%.

Reduction in Bottoms Flow Rate Reduction in bottom flow is a common symptom of flood, and a primary criterion to determine the flood point (192, 293, 437). Upon flooding, liquid accumulates in the tower, so less liquid reaches the bottoms, and the bottom level falls. Most frequently, the bottom level is controlled by manipulating the bottom flow rate, so the level may stay constant while the bottom flow rate is reduced.

While a reduction in bottom flow indicates flooding, many towers flood with no significant decline in bottom level or flow rate. For instance, if the feed is mostly liquid and flooding occurs above the feed, the bottom section may continue to operate normally with no significant decline of bottom level or flow rate. Also, if the flood location is well above the bottom, the time delay from the onset of flooding to the reduction in bottom level or flow may make accurate measurement of the flooding conditions difficult.

Generally, reduction of bottoms flow rate is a good indicator of flooding in towers that flood near the bottom and in towers that are relatively short (437).

Rapid Rise in Entrainment A rapid rise in entrainment is another common flood symptom (4, 192, 284). The liquid accumulating in the tower builds up to the top and is entrained in the tower overhead vapor stream.

When the tower overhead vapor stream goes to a knockout drum (e.g., absorbers), or to the bottom of another tower, this entrainment can be recognized as buildup or rapid rise of a liquid level in the drum or bottom of the downstream tower. In most distillation towers, the overhead vapor stream goes to a condenser, and from there to a reflux drum. The reflux drum level control usually manipulates either the distillate rate (Figure 2.11a) or the reflux rate (Figure 2.11b).

When the drum level manipulates the distillate rate (Figure 2.11a), the entrainment rise is often indicated as a significant increase in distillate rate for no apparent reason. When the drum level manipulates the reflux rate (Figure 2.11b), the entrainment is often indicated as an increase in reflux for no corresponding increase in boilup rate. The flooding prevents most of the increased reflux from descending down the tower, so it entrains back into the tower overhead, further increasing the drum level and reflux rate. The reflux flow meter reads high and the reflux valve often opens widely due to the recirculation of entrainment around the tower overhead loop. Alternatively, the entrainment is often indicated as a rise in reflux flow rate that does not raise the heat input required to maintain the same tower temperatures. In one case (280), the primary flood symptom was a normal reflux rate even though the boilup rate was extremely low.

In some cases, as in many refinery fractionators, tower overhead temperature (instead of drum level) manipulates the reflux rate (Figure 2.12). Upon flooding, entrainment from the top raises the heavies content of the tower overhead, and therefore the overhead temperature. The temperature controller raises reflux. The flooding prevents the increased reflux from descending down the tower, so it entrains back into the overhead, further increasing the overhead temperature. Here too, the reflux flow meter reads high, and the reflux valve often opens widely due to the recirculation of entrainment around the tower overhead loop (upper operating chart in Figure 2.12). Alternatively and less frequently, with highly subcooled refluxes, the cold reflux entrainment reduces the tower overhead temperature, and the controller cuts back reflux, which in turn heats up the overhead, causing the controller to increase reflux. In this case, the entrainment rise shows up as temperature and reflux flow rate fluctuations (lower operating chart in Figure 2.12).

The entrainment from the tower overhead may overwhelm the condenser gravity drain lines, which are generally sized for the normal process flow rate (reflux plus distillate) and not for the massive entrainment incurred by flooding. Unable to drain, the entrained liquid builds up in the condenser, partially flooding it, thus reducing the area available for condensation. This in turn causes the column pressure to rise. Pressure differential builds up between the column and reflux drum, and eventually becomes large enough to quickly push and empty the liquid out of the condenser. The area available for condensation quickly rises, and with it the condensation rate, causing the tower pressure to dive. Following the fast liquid-emptying step, drainage from the condenser returns to a gravity mode, liquid accumulates, and the process repeats. The end result is severe pressure fluctuations in the tower, which can be as high as 7 psi in a few minutes. So a scenario of steady tower pressure breaking into severe fluctuations is often a symptom of flooding near the top with the entrainment interacting with the condenser drainage.

The rapid rise in entrainment indicator is particularly useful when flooding occurs near the top. However, this indicator will fail to indicate a stripping section flood that does not propagate to the rectifying section.

Loss of Separation As flooding approaches, the amount of liquid entrained by the vapor sharply rises. At high pressures and/or high liquid rates, the amount of vapor entrained in the downflowing liquid also rises. As either type of entrainment rapidly escalates in the vicinity of the flood point, efficiency and separation decline (Figure 2.13). The decline in efficiency tends to occur closer to the flood point with downcomer floods than with entrainment floods. In Figure 2.13, the limiting mechanism for the isobutane–normal butane system is believed (513) to have been downcomer flood, while that for the cyclohexane–heptane system was entrainment flood. In packed towers, the decline in efficiency tends to take place closer to the flood point with smaller packings than with larger packings. With large packings (e.g., 2 in. or larger random packings, and of surface area per unit volume below 60 ft²/ft³ for structured packings), and/or under vacuum, the decline in efficiency may begin at rates well below the hydraulic flood point.

Since loss of separation starts before the tower is hydraulically flooded, using it gives a lower flood point than other indicators. This is hardly a disadvantage, because the point at which separation begins to rapidly deteriorate is of greater practical importance than the hydraulic flood point. The point at which the separation begins to deteriorate is often referred to as “the maximum useful capacity,” or “maximum operational capacity,” and the vapor loads at which it occurs are typically about 0–20% below the hydraulic flood point. In most atmospheric and superatmospheric tray towers, this point occurs at flow rates of 5% or less below the hydraulic flood point.

The loss of separation is usually inferred from laboratory analyses of column products. A plot of tray or packing efficiency (Figure 2.13) against flow rate at a constant reflux ratio can identify the point where the separation begins to deteriorate as flood is approached.

The tower temperature profile is an alternative indicator of the loss of separation. Temperature often rises above the flooded region because the accumulating liquid rising from below is richer in heavies, and also because the flooded region no longer performs an efficient separation. Temperature may also rise below the flooded region because the smaller amount of liquid downflowing from the flooded region leads to heating up below, and also because the higher pressure drop raises the boiling points below.

A good knowledge of the normal and flooded temperature profiles under similar feed conditions is invaluable for the successful application of temperature profiles for flood indication. Figure 2.14 shows temperature profiles under normal and flooded conditions (231). The points are pyrometer measurements of wall temperatures taken from the access ladder of an uninsulated tower. All the temperatures are for the even-numbered trays, as the ladder was on the left side of the tower in Figure 2.14, and the side downcomers obscured the odd-numbered trays. The crosses represent the normal temperature profile, showing a discrete reduction in temperature every two trays as one ascends the tower. The circles represent the temperature profile when the bottom three to four trays were flooded. The temperature spread across the bottom four trays completely disappeared, indicating loss of separation. The circles also show hotter temperatures above, due to movement of heavier components up induced by the poor separation on the bottom three to four trays. A similar application of wall-measured temperature profiles to successfully identify flood was reported by Lieberman (284).

Section 2.9.1 presents an example of a packed tower under deep vacuum in which neither the pressure drop nor gamma scans provided conclusive evidence for flooding. It was the loss of temperature spreads in the packed bed that provided the conclusive flooding diagnosis.

A shortcut for the Figure 2.14 procedure is to look at the differential between two suitable temperatures in the relevant tower section. For instance, in the tower in Figure 2.14, the normal temperature difference between tray 10 and tray 6 was (140 − 108) = 32 °F. Upon flooding, this temperature difference fell to (147 − 137) = 10 °F. A differential temperature indicator between trays 10 and 6 will indicate the onset of flood. It is important to select the correct trays in this procedure. For instance, the temperature differential between trays 6 and 4 in Figure 2.14 will be a poor indicator of flood. A flood-detection method using temperature differentials changing with time was proposed by Dzyacky et al. (91, 92). Generally, temperature differentials change faster than differential pressure and can give an earlier detection of the onset of flood.

Caution must be exercised with temperature profile interpretation, because unchanging temperatures may also indicate a “pinch” condition (i.e., poor separation due to insufficient reflux or reboil). In order to tell the difference, reflux and reboil can be raised. If this improves separation, pinching is indicated. If this worsens separation, or does not affect it, flooding is indicated. Other flood symptoms can help the interpretation. In one tower (24), flood was noted by a rise in pressure drop, but the separation stayed good. Reducing reflux got the tower out of flood, but the separation deteriorated. The tower had more trays than needed and operated near a pinch, so its separation was unaffected by the loss of stages upon flooding but deteriorated when reflux was cut.

Caution is also required when interpreting packed tower temperature profiles, as it may be difficult to distinguish poor separation due to vapor or liquid maldistribution from one due to flooding. Again, tests of raising reflux and reboil can shed light. If separation improves, it will argue against flooding.

In order to reliably establish the column temperature profile, a large number of temperature measurement points is needed. If unavailable, a vertical temperature survey can be conducted. Temperature surveys are discussed in detail in Chapter 7.

Temperature profiles provide an effective, low-cost method of detecting flood, but its success depends on the existence of a large enough number of measurement points and on the tower having a large enough temperature gradient under normal operating conditions. If the normal temperature difference from tray to tray is small, as in close separations, the flooded temperature profile does not vary significantly from the normal profile, and temperature profiles will be poor indicators of flooding.

Gamma Scanning Gamma scanning is one of the best techniques for detecting flood. It is powerful for diagnosing flood, identifying flooded regions, and gaining insight into the nature of the flood. A detailed discussion is in Chapters 5 and 6.

Bleeders Use of vapor bleeders was found effective for flood testing (187) in atmospheric and pressure towers. Each bleeder is located in the vapor space above a tray, or more frequently in the overhead line upstream of the condenser. Upon bleeder opening, only vapor coming out indicates no flooding; liquid spraying out indicates flooding. When the tower is above its atmospheric boiling point, bleeding flashes and chills the liquid. The presence of liquid in the bleeder can therefore be detected as a sharp temperature drop of gas issuing from the bleeder (187), or “weeping” or icing up downstream of the bleeder.

The bleeder technique is not very popular and is often hazardous or environmentally unacceptable. Other disadvantages are a shortage or lack of valved bleed points inside the tower, the need to know which trays are most likely to flood, and a lack of indication of an approaching flood condition. Of its few applications, most are bleeders in tower overhead lines that discharge into closed systems.

Sight Glasses (Viewing Ports) These provide superb visual indication of flooding (187, 192). Sight glasses are expensive, increase the leakage potential, and upon glass breakage may lead to a chemical release. Supplying a light source that permits adequate viewing can also be an issue. For these reasons, this technique is uncommon in commercial towers, mainly used for towers that process nonhazardous materials at near ambient pressure.

Noise Increase This technique was demonstrated in a pilot packed column at the Separation Research facility at the University of Austin, Texas, but not commercially. A sharp increase in the noise signal from the packing indicates the onset of flooding (365), remaining high when the packing is flooded. This is analogous to the fizzling noise of a carbonated drink.

Pattern Recognition In many cases, as flood is approached, certain variables were observed to oscillate in a brief, repeatable, systematic pattern (91, 92). For example, liquid inventories on trays rise momentarily, resulting in a temporary decrease of the bottom level simultaneously with a temporary increase in the tower dP, or the onset of erratic movement of either or both. Since prior to flood these variables remained within the normal operating limits, it may be difficult to recognize the pattern from observing the operating charts. In contrast, it was reported (91, 92) that this pattern can be readily recognized empirically using first and second derivatives with time, which change a lot more than the variables. Operators often have expertise in identifying such patterns, and should be consulted. This field-tested pattern recognition technique can provide early detection of flood and permit tower operation closer to the flood limit (91, 92).

Flooding Near the Top of the Tower A small pressure drop rise, often of the order of 0.3–0.5 psi, is typical when the flooding occurs near the top of a tower. Since only one or very few trays flood, the liquid accumulation is not high, which incurs only a small pressure drop rise. When the tower dP is 3–4 psi, a small dP rise of 0.3–0.5 psi remains indistinguishable from the general noise in the dP measurement. Figure 2.22 shows actual operating charts, with Figure 2.22a being the most common. The chart in Figure 2.22b is from a tray tower in Figure 2.4 with vapor flood at the top. An increase in vapor rate generates flood (the vertical rise in dP). Once liquid accumulates at the top, the liquid downflow drops, and the control temperature rises. The controller cuts back on boilup, reducing vapor upflow, and the upper trays gradually unflood (the sloped downward trend in dP). The unflooding re-instates the normal liquid downflow, the control temperature declines, increasing boilup, and the flood returns (the vertical rise in dP). The typical cycle amplitude is about 0.3–0.5 psi with a period of three to five minutes.

The ability of the dP versus vapor rate plot (as described in Section 2.8) to decouple the dP changes due to flooding from regular dP changes makes this plot invaluable for diagnosing flooding near the top of the tower. One 20-trays high-pressure (200 psig) hydrocarbon absorber in foaming service experienced premature flooding at about 67% of the design lean oil rate. Flooding was identified by liquid carryover from the tower top to the knockout drum downstream. The drum normally operated with no liquid in it. The plant always identified flooding by the appearance of a liquid level in the drum. When the operators were asked whether they saw a high dP upon flooding, their answer was “Never. The tower operates with 2 psi pressure drop both under normal and flooded conditions.”

Figure 2.23 plots data collected over a two-month operating period. The hollow circles are periods of four hours or longer in which there was no liquid level in the drum, i.e., definite no-floods. The filled circles are periods of four hours or longer in which there was continuous level in the drum, i.e., definite floods. The points fall on two distinct lines, with the difference between them being 3 in. of water (0.11 psi) pressure drop. So the dP versus vapor load curve was sensitive enough to identify such a small pressure drop change.

An interesting feature of Figure 2.23 is absence of a point of inflection like in Figures 2.5 and 2.16–2.19. The absence of a point of inflection indicates a flood that is insensitive to vapor load, i.e., a liquid flood. Indeed, in this case it was a downcomer choke flood.

Channeling Allows Liquid Accumulation Without Interfering with Vapor Rise Another situation where a tower floods with little or no pressure drop rise is when channeling occurs, and the channeling permits vapor to rise, bypassing the liquid accumulation. This sometimes occurs in packed vacuum towers. One such case is described in detail in Section 2.9.1. In another vacuum tower (65), a CAT scan (Section 6.4) showed that upon flooding a bed of structured packings, liquid accumulated along the periphery, allowing vapor to channel through the center.

Grobenrichter and Stichlmair (142) studied pressure drop rise due to crystallization of salt by passing dry air through saturated aqueous solutions in a 6-in. column. With wire gauze packings, there was always a large pressure drop increase due to liquid accumulation. With corrugated sheet packings, the pressure drop rise was high at higher liquid rates, but much smaller at low liquid loads (<12 gpm/ft² in their tests), suggesting vapor channeling around zones of local flooding.

A similar scenario may also occur in tray towers. In the tower described in Section 2.8 (Figure 2.19), after the popped-out valve floats disappeared, the pressure drop came down (the winter 2016–2017 curve). However, the flood experienced earlier (the early autumn 2016 curve) did not go away, as evidenced by gamma scans and off-spec products. Popping out the valve floats induced vapor channeling, which generated a rise path for the vapor without incurring a high pressure drop. A similar experience of flooding without pressure drop rise was reported (226) in a chemical column whose manways were left uninstalled at a turnaround, allowing vapor to channel through the open spaces.

Symptoms A foam flood gives similar symptoms to other floods, namely high pressure drop, entrainment from the top of the tower, reduction in bottom flow rate, and poor separation. However, in addition, it may give one or more of the following symptoms:

1. An erratic rise in pressure drop, sometimes with no apparent reason (Figure 2.26a).
   
2. Fluctuating tower pressure drop (Figure 2.9).
   
3. Nonreproducible flood data. The points in Figure 2.26b depict different vapor and liquid loads at which flooding occurred at different times. All occurred prematurely.
   
4. Flood sensitive to temperature (Figure 2.26c; also, Ref. 469). At the hotter temperature, the foaming did not occur.
   
5. An abnormal temperature profile. For example, in amine absorbers foaming will cause the reaction to occur further up in the tower (15, 325). The rich solution will become cooler, while the lean gas will warm up. A significant drop in temperature difference between the inlet and outlet gases, as well as between the rich and lean amine, will suggest foaming.
   
6. Flooding goes away when antifoam is added.
   
7. Gamma scans and downcomer neutron scans effectively identify foams (Section 5.5.5).
   
8. Downcomer inlet velocities exceeding the criteria in Section 2.16.

Where to Look for Foaming A survey of published case histories (198, 201) found foaming to be a service-specific phenomenon. Ethanolamine absorbers that absorb acid gases such as H₂S and/or CO₂ from predominantly hydrocarbon gases, and their regenerators, account for about one-third of the published cases. Another 10% were also in acid gas absorption/regeneration service, but using alternative solvents such as caustic, hot potassium carbonate (hot pot), and sulfinol. Another 10% were in absorbers of gasoline and LPG from hydrocarbon gases using hydrocarbon solvents.

Case histories of foaming were also reported (198, 201) in each of the following services:

• Chemical: aldehydes, soapy water/polyalcohol oligomer, solutions close to their plait points (solubility limits), extractive distillations, solvent residue batch still, ammonia stripper, DMF absorber (mono-olefins separation from diolefins), cold water H₂S contactor (heavy water GS process). Recent work added to this list include distillation of organic acids and alcohols (369), heterogeneous azeotropic systems (369), and butanol–water (422).
  
• Refinery: crude preflash, crude stripping, visbreaker fractionator, coker fractionator, solvent deasphalting, hydrocracker depropanizer.
  
• Gas: glycol contactor.
  
• Olefins: high-pressure condensate stripper.

There is a rich literature on experiences with foaming problems, with most of the pre-2006 reported cases described on pages 543–557 of Distillation Troubleshooting (201). The discussionbelow briefly describes some of the highlights. The reader troubleshooting a potential foaming problem is strongly encouraged to explore these experiences and see whether one or more can be beneficial to the problem at hand.

In about a third of the reported cases (198, 201), suspended solids, corrosion products, or polymers catalyzed foaming. Although not specifically reported, solids could have catalyzed foaming in many of the other cases. The survey by Le Grange et al. (272–274) of 108 foaming incidents in ethanolamine absorber–regenerator systems found that half of them had foam promoting contaminants entering the plant with the feed. Good filtration can often alleviate, even mitigate the foaming, as described by many of the cases abstracted in Ref. 201 and detailed in Refs. 98, 100, and 316. However, solvents may react with the material used in filters, and the reaction products may cause foaming (15, 81, 341, 357).

In ethanolamine service, or other alkaline absorber–regenerator systems, chemical analysis or coloration of the lean solvent showing a high solids or degradation product content indicates a high potential for foaming. Solids and degradation products catalyze foams. A qualitative guide correlating appearance with foaming tendency in ethanolamine solutions was presented by Lieberman (286). Engel (98, 100) provides an in-depth discussion of the various root causes of foaming in ethanolamine services.

In many cases, the foaming was caused or catalyzed by an additive such as a corrosion inhibitor, in some cases in ppm concentrations. In one reported case (452), as well as many others experienced by the author, severe foaming occurred in amine or caustic absorbers due to a corrosion inhibitor contained in the steam condensate used for solution makeup. Another case was reported (516) of a heavy contaminant coming into the feed of an extractive distillation tower and causing foaming. The author is also familiar with a case where an anti-foulant induced foaming.

Ross and Nishioka (403) found that the potential for foaming is maximized near the plait point (near the solubility limit), i.e., where a solution is still homogenous but is near the point of breaking into two liquid phases. Once the plait point is reached, and two phases are formed, one phase may serve as a foam inhibitor and destroy the foam. Alternatively, they may form a foaming emulsion. Hydrocarbon condensation into aqueous solutions (6, 81, 181, 273, 274, 331, 341, 452, 473), or small amount of water in a hydrocarbon system (Case 16.7 in Ref. 201, also Ref. 265, both Olefins condensate strippers), is known to promote foams. One refinery preflash tower (287) foam flooded because water was refluxed into the tower due to malfunction of the reflux drum boot. The flooding stopped when the boot again removed the water. A similar foam flood was experienced in a butyraldehyde column (469) due to refluxing small quantities of water; the problem was solved by adding a water-removal boot to the reflux drum.

High pH stabilizes foaming emulsions. The author experienced foaming in one depentanizer and one de-isobutanizer, normally non-foaming services, when small quantities of caustic entered the towers.

A foaming problem may occur in one installation but not in another even though both handle the same materials. Often a foaming problem remains “hidden” because a tower or downcomers are oversized, or may not occur in one installation because a surface-active impurity is absent. In one case (39), a new absorption unit experienced foaming while a similar older unit did not. This was because the older unit had oversized downcomers that adequately handled the foam. In another case (1), foaming was observed only after a capacity expansion. Tests by Senger and Wozney (422) showed that foaming with the same chemical system may proceed differently depending on the liquid flow rates and the packings

Avoiding local high-turbulence regions may help prevent foaming. In one case (250), a tower foamed when the reflux entering via a downward elbow dropped 12 in. into a liquid pool behind an inlet weir (Figure 2.27a). A gamma scan showed foam on a few trays, but severe foaming only on the top tray. Extending the pipe into the pool (Figure 2.27b) eliminated the foaming. In another case (250), liquid to the regenerator of a hot pot absorber entered via a pipe distributor with holes pointing downwards into a sump behind an inlet weir (Figure 2.27c). Again, gamma scans showed foam on the top tray. To cure, the pipe was fitted with dip tubes that were submerged in the sump liquid (Figure 2.27d). There was no more foaming after this.

In one amine absorber (223), a retray with low pressure drop trays led to channeling of the inlet gas jet to the region opposite the gas inlet nozzle, generating a high-gas velocity region, and the increased turbulence produced premature foam flood. Figures 2.28 and 2.29 show how increased gas velocities raise foam height and foam decay time in test equipment.

Testing Foaming tests were reported in many literature sources (1, 32, 39, 79, 122, 192, 201, 225, 422, 438, 490, 516), and these have been applied with various degrees of success. Experience has shown (39, 122, 222) that when not performed under actual plant operating conditions, foam tests can be inconclusive. Generally, methods 3–7 below reproduce actual plant conditions, and at least one of them is often simple to set up. Common test methods include:

1. The “bottle shake” test. A clean bottle is charged with a sample of the solution. It is left open for some time to allow all the entrapped gas to escape. The bottle is capped and shaken vigorously. It is then set on a table. The foam height and the time it takes for the foam to settle to the liquid level are recorded. Settling times exceeding five seconds indicate foaming; the longer, the more severe. Successful application of this test for amine absorbers was reported (15, 81). This method is easy, simple, and reliable when it indicates foaming. However, it often fails to detect foaming.

   

   

   
   
2. Dispersing gas (e.g., air, nitrogen) into a solution sample in a small vessel. Foam height and foam decay time give a measure of the foaming tendency. This is again a simple method and reliable when it indicates foaming. One watchout is that foaming cases were reported, with this method failing to indicate foaming (39, 192, 201, 225, 222). It was recommended to conduct this test at the process temperature, but sometimes the temperature makes little difference (452). Refs. 452 and 504 provide detailed procedures that have been successfully applied for ethanolamine absorber–regenerator systems.
   
3. An apparatus that disperses gas (e.g., air, nitrogen) through a solution sample (e.g., Figures 2.28a and 2.29). To achieve good dispersion, the gas is often dispersed via a fritted filter. Details are described elsewhere (1, 122, 452). In principle, this test may suffer from the same problem as in 2 above, but the experience reported with it has been excellent (1, 122, 273, 438, 452, 516). The good experience and its simplicity have made this test the most popular.
   

   Figures 2.28b and 2.29 show how this method should be applied. For best results, the foam heights and decay times should be plotted against the superficial gas velocity. The test should first be performed on a non-foaming solution like demineralized water (blue dots in Figure 2.29d). The test is then repeated with the test solution (orange dots in Figure 2.29d). Different antifoam agents, at different concentrations, can then be tested (Figure 2.28b; yellow and dark brown dots in Figure 2.29d).

   
   
4. Oldershaw or pilot-scale towers have been successfully used for diagnosing foaming (490, Case 2.23 in Ref. 201). With glass Oldershaws or pilot towers with observation windows, this method can also provide invaluable information on the nature of the foam and where in the tower it is likely to foam. This technique is most suitable for atmospheric and vacuum towers but may have difficulty dealing with solids and high-pressure applications.
   
5. An armored level glass (Figure 2.28c) can be set either in the field (222) or in the laboratory (39), and nitrogen is dispersed into the sample at actual plant temperature and pressure. This procedure has special merit when column temperatures and pressure cannot be reproduced by other tests and has been successful where methods 1 and 2 failed (39, 222).
   
6. Antifoam tests. If a tower responds to antifoam injection, it means there is foaming. However, if the tower does not respond to antifoam injection, it does not necessarily deny foaming. Some foams respond to one antifoam agent but not to others. The antifoam concentration also matters, too much antifoam may produce more severe foaming than none (177, 201). In one case (81), an antifoam was able to break down the foam, but the foam would return less than an hour later due to degradation by heating and cooling cycles. Excessive use of some antifoams can produce fouling and accumulation (6, 473).
   

   When the tower responds to antifoam, these tests can help find the optimum agent and concentration, as described in one case (81), see also Figure 2.29. To rule out foaming, try a variety of agents and concentrations. It would be a good idea to experiment with them in the lab before injecting into the tower to determine what the best agents and concentrations are to try. If several agents and several concentrations have been tried and the tower has not responded, one can conclude that foaming is unlikely, as described in one case (225) and many other experiences by the author.

   
   
7. Field tests. Testing a system where foaming is suspected with a non-foaming solution can conclusively confirm or deny foaming. In one amine absorber (225), where foaming was on the suspect list, a field trial was conducted in which the amine solution was replaced by clean water. While an amine system often foams, a clean water system does not. The water trial showed an identical capacity limitation as experienced with the amine solution, conclusively denying foaming.
   

   In the abovementioned case, it was necessary to interrupt operation and perform the water test. In other cases, the trial requires no process interruption. The author is familiar with cases in which a steam condensate makeup was used for caustic or amine solutions, and foaming was experienced. Foaming disappeared within a few days after a trial in which demineralized water replaced the condensate makeup.

In addition, there are tests that can help in specific circumstances:

1. Foam tests on different samples. In one ethanolamine absorber (273), foaming caused severe carry-under of gas into the rich amine, resulting in a venting problem downstream. The foaming source diagnosis was established by performing foam tests on both the lean amine and rich amine (Figure 2.30a,b). The foaming was induced by contaminant-containing liquids in the inlet gas line (Figure 2.30c).

   

   
   
2. Stepping up the bottoms flow rate can provide inconclusive supporting evidence for diagnosing foaming. A large step-up in bottoms flow rate that does not produce a significant change in bottoms level supports a foaming diagnosis. When the foam height exceeds the upper tap of the level transmitter, the transmitter measures the foam density, not the liquid level. Detecting liquid when opening the vent valve on the upper tap of the level glass or transmitter also supports foaming. A step-up in bottoms flow accompanied by a drop in bottoms level does not deny foaming; it just means that the level has not risen above the upper tap.
   
3. A discrepancy between staggered level gauges in the bottom sump can indicate foaming (504). Observation of the level glass can also show a hydrocarbon layer above an aqueous solution (6), which may be linked to foaming.

Cures Three cures have been successful for foaming problems: eliminating the foam-causing chemical (e.g., by eliminating the additive or by filtering out the solids that catalyzed the foams), injecting antifoam, and debottlenecking downcomers (either by using larger downcomers, or by reducing downcomer backup, or by replacing trays by random packings). Injecting antifoam has not always been effective, and some experimentation with the inhibitor type and concentration are often required.

These cures are discussed at length in Ref. 192. Cures specific to ethanolamine absorber–regenerator systems are discussed by Le Grange et al. (273, 274).

Reliable Level Monitoring Some of the lessons learned are as follows:

1. Do not rely on a single level indicator. Always have a cross-check by another, ideally using a different technology (if reliable). In key fractionators, some plants add a third level indicator with a voting mechanism (2/3).
   
2. Ensure that the upper tap of the level transmitter is beneath the reboiler return or vapor inlet. Figure 2.34a illustrates the reason. Once the reboiler vapor blows into the liquid, the liquid becomes a frothy mixture, typically with half the specific gravity of the pure liquid. Therefore, a column of froth that is x ft tall will only show up as x/2 ft of liquid in the level transmitter. The tower can be liquid-full, with the transmitter only reading 70% level.
   
3. Ensure that the reboiler return or gas inlet does not impinge on the level taps or the liquid level.
   
4. Closely check that the correct liquid density is used in the calibration. An out-of-calibration level transmitter has been one of the most common causes to false level indications.
   
5. Under startup and abnormal operation conditions, liquid compositions and temperatures, and therefore liquid densities, are likely to differ from those of normal operation. Guide the operators on how to interpret the level transmitter readings (calibrated for normal operation) to the startup and abnormal operation.
   
6. Pay attention to control during startup. Most towers manipulate the bottoms flow rate to control the base level. However, at startup the bottom flow rate is yet to be established. Work with the operating team on how to control the level at startup without encountering high liquid level issues. This may require rigging lines from the tower bottom to an off-spec tank, or from a product tank to the tower base, to supply liquid for the reboiler startup.
   
7. Ensure that the dP transmitter in the tower is operational. A flood initiated by a high base level will show up as an increase in dP.

Figure 2.34b shows a good level measurement practice. It features two independent level transmitters mounted on two separate bridles. The upper tap of each transmitter is beneath the reboiler return nozzle. There is a third level transmitter coming from one of the upper level taps and going to another tap just below the bottom tray or packed bed. This transmitter should always read zero. A reading on this transmitter (which can be set to trigger an alarm) indicates a rise of the liquid level above the reboiler return inlet and likely flood initiation.

Good Sump Design Some of the lessons learned are as follows:

1. Never introduce the reboiler return or vapor feed below the liquid level (see Ref. 192 for a possible, very infrequent exception).
   
2. Always have the bottom of the reboiler return nozzle at least 12 in. above the high liquid level (HLL), as shown in Figure 2.34b. In one case (141), the bottom of the nozzle was only 4 in. above the HLL, causing liquid entrainment from the bottom sump into the first tray, which not only reduced tower capacity but also perpetuated foaming and severe fouling on the tray.
   
3. Avoid impingement.

Impingement by the reboiler return or incoming gas on the liquid level generates entrainment that can travel up the tower and flood trays or packings. Reboiler return or inlet gas impingement on instruments gives false readings, which may cause high liquid levels to go unnoticed. Premature floods can also occur due to impingement of the reboiler return vapor on the bottom tray, the seal pan overflow, and the inlet from a second reboiler. Some case studies (201) also reported severe local corrosion due to gas flinging liquid at the tower shell in alkaline absorbers fed with CO₂-rich gas, mostly in ammonia plants.

Figure 2.35 shows poor arrangements that generated impingement of reboiler return or inlet gas on the liquid level, leading to premature floods or entrainment that traveled right up the tower. Sections 8.14–8.16 contain more cases.

Diagnosis The following can help diagnose flood due to a high liquid level or an impingement issue:

1. Check the level measurement by an alternative technique. This may involve another transmitter. Gamma scanning or neutron backscatter is particularly powerful and informative for this purpose. Section 5.5.18 shows the application of gamma scans to detect flooding due to excessive pressure drop in a kettle reboiler circuit.
   
2. A useful test is increasing the bottoms flow rate to see if the level comes down (like it should). If the level does not come down, liquid is likely stacked above the upper level transmitter tap, in which case the transmitter measures the froth density rather than the liquid level. Keep raising the bottom flow rate until the level starts declining or another constraint is reached. Has this alleviated the flood? Observe the tower dP when performing this test. If the flood is caused by a high liquid level, the tower dP should decline as the bottom flow rate is increased. Incorporating gamma scans or neutron backscatter with this test can make its results even more conclusive.
   
3. A useful check is to open the vent valve at the upper level tap (if this can be done safely). Normally, it should vent vapor. Liquid coming out is an indication of a liquid level above the transmitter.
   
4. Draw a to-scale sketch of the bottom sump. Look for impingement of the reboiler return vapor on the liquid level, instrument taps, and the seal pan overflow. See Sections 8.14–8.16 for many cases in which such a sketch was key to a correct diagnosis.
   
5. If the reboiler is a kettle reboiler, compile a pressure balance on the kettle circuit. Reference 217 provides a case study of column flood due to excessive pressure drop in a kettle reboiler circuit. The reference contains a detailed description of the pressure balance that was central to the diagnosis.

Hiccups (Also Referred to as Burping, Puking, Cyclic Slugging) A typical symptom of nonsteady state accumulation is cyclic slugging, which tends to be self-correcting. The intermediate component builds up in the column, typically over hours or days. Eventually, the column floods, or a slug rich in the offending component exits either from the top or from the bottom. The column end from which the slug leaves often varies unpredictably or according to the operator’s action. Once a slug departs, normal column operation resumes fairly quickly, often without operator intervention. The cycle will then repeat itself. Several experiences have been reported in the literature, many of them described in Refs. 94, 172, 199, 201, 389, and many others cited by the mentioned references.

Intermediate component accumulation may interfere with the control system. For instance, a component trapped in the upper part of the column may warm up the control tray. The controller will increase reflux, which pushes the component down. As the component continues to accumulate, the control tray will warm up again, and reflux will increase again; eventually, the column will flood. Cases describing a similar sequence have been reported (Case 2.13 in Ref. 201, also 94, 389).

One of the most fertile breeding grounds for intermediate component accumulation is a condition in which at least some section in the tower operates close to the point of separation of a second liquid phase (201). Either the accumulation itself, a condition in which phase separation equipment in the feed route (a decanting drum, a coalescer) malfunctions, or simply a change in feed composition can induce separation of a second liquid phase inside the tower. The introduction of a second liquid phase changes volatilities and azeotroping, which can aggravate hiccups.

Hiccups can cause major upsets in downstream equipment. A slug of lighter material escaping from the bottom of one tower can quickly vaporize in the hotter downstream tower, causing a pressure surge there and possibly lifting its relief valve.

Diagnosis Periodic cycling that tends to be self-correcting, with a long time period, is the key symptom for hiccups. The time period is seldom less than one hour, which distinguishes hiccups from other, shorter cycles such as those associated with flooding, hydraulic, or control issues, whose period is typically a few minutes. Typical cycle period for hiccups ranges from about once every two hours to once every week. The cycles are often regular, but if the tower feed or product flows and compositions vary, the cycles may be irregular.

Drawing internal samples from a tower over the cycle (or over a period of time) is invaluable for diagnosing accumulation. Such samples will show high and increasing concentrations of the intermediate component over the cycle. Even a single snapshot analysis can detect accumulation of a component. Unfortunately, safety, environmental, and equipment constraints often preclude drawing internal samples.

Closely tracking and monitoring temperature changes is also invaluable. Temperature changes reflect composition changes. Accumulation of an intermediate component tends to warm the top of the tower and cool the bottom of the tower. The response of the control system to temperature changes and any rise in pressure needs to be considered when interpreting temperature trends. One symptom is common: at the initiation stage, temperature deviations from normal are small, often negligible. As the accumulation intensifies, systematic temperature deviations (“excursions”) from normal become apparent. Close to the hiccup point, temperature deviations become relatively large.

Figure 2.36 illustrates the variations in the temperature profile. The “0.5% decant” curve shows a temperature profile close to normal, practically unaffected by the accumulation. As the component builds up, the curves move closer to the “no decant” curves, first at stage 11 and then progressively to stage 17. In this case, the temperature declines by a sharp step due to the formation of the second liquid phase on these stages. Had there been no second liquid phase, the accumulation would have lowered the temperatures in the accumulation zone below the feed in a more gradual (rather than stepped) manner.

Steady state bulges can be readily simulated and recognized on a plot of component concentration against stage numbers, as in Figures 3.1 and 2.36. The simulation needs to be able to correctly simulate a second liquid phase, which may form during component accumulation. With hiccups, steady state simulations are likely to run into convergence problems, reflecting the reality that steady state is not reached. To get around this, “sneak into the solution” by studying a related system that can converge, e.g., input less intermediate component in the feed, or draw a side product. Once a converged solution is reached, slowly proceed toward the actual system (by raising the intermediate component in the feed, or cutting back on the side draw). Track the changes by plotting a concentration versus stages diagram for each step. In the case of a second liquid phase, many stages in the simulation may alternate between a single liquid phase and two liquid phases, making convergence particularly problematic. Either increasing or decreasing the concentration of the second liquid phase can help reach a solution.

Gamma scans and dP measurements can detect the intermittent flooding. Gamma scans typically show normal operation at the beginning of the cycle and mid-column flooding when a hiccup occurs. If enough pressure taps are available on the tower, dP transmitters can be just as informative. Finally, sight glasses are extremely useful when safety requirements permit.

Cures There are three cures that can effectively mitigate the hiccups and two “band-aid” solutions that can reduce the frequency and severity of the hiccups. These cures are discussed in detail with several examples elsewhere (199, 201, 419):

1. Reducing the column temperature difference. This allows the accumulating component to exit with one of the products but will increase the impurity in this product. It is low-cost, easy to implement, and therefore often used.
   
2. Removing the accumulated component from the tower using a side draw in the accumulation zone. This is often accompanied by using decanting or another separation technique to separate the intermediate component from others. Figure 2.36 shows the effectiveness of this technique for mitigating the accumulation. The bulge went down from 40% to 5%, and the number of stages with two liquid phases went down from 7 to 2.
   
3. Removing the accumulated component from the feed.
   
4. Modifying tower internals—this is a band-aid solution.
   
5. Improving controls and living with the problem—another band-aid solution.

Downcomer Draw Boxes Column flooding is likely to initiate in the draw arrangement in Figure 2.37a whenever the liquid flow rate to the draw exceeds the draw hydraulic limit. The excess liquid has nowhere to go, so it will stack up in the downcomer and flood the next tray. From there, the flood will propagate upward. Column flooding is unlikely to initiate in the draw arrangement in Figure 2.37b because the generous clearance under the downcomer will allow excess liquid to drain onto the tray when the draw encounters a hydraulic limit.

Depending on the dimensions, column flooding may or may not initiate in the draw arrangement in Figure 2.37c when the draw reaches its hydraulic limit. At the limit, liquid will stack up in the seal pan and in the downcomer up to the seal pan overflow elevation. If the overflow weir is tall enough to back up liquid in the downcomer all the way to the tray above, flooding is likely. It is important to note that the seal pan liquid is likely to be degassed, while the downcomer liquid is likely to be frothy and therefore have a significantly lower specific gravity, typically about half, so the seal pan can back downcomer liquid to about twice the submergence of the downcomer. On the other hand, if there is ample vertical height to allow overflow without excessive downcomer backup, flooding is unlikely.

A potential problem with the arrangement shown in Figure 2.37b is vapor entry into the downcomer during excessive draw rates. Should the draw rate exceed the rate of incoming liquid, vapor will be sucked into the downcomer and the draw via the downcomer clearance. The entering vapor may choke the downcomer, the draw, or both. Intermittent flooding or cycling may initiate, producing instability. With arrangements shown in Figure 2.37a and c the downcomers are always liquid-sealed, so vapor entry is unlikely.

Should a side draw bottleneck be suspected, the following checks and field tests can help diagnose:

1. Check for self-venting flow in the draw line. Liquid in the column is aerated, i.e., contains vapor bubbles, unless it has been degassed. Degassing a liquid takes a minimum residence time of 30–60 seconds based on the sump volume and the total flow rate through it, from the point of liquid collection to the top of the draw nozzle. Downcomers are usually designed for a residence time of 3–7 seconds, which may increase to about 10 seconds after adding the volume of the sump underneath. So the fluid leaving in the draws from any of the arrangements in Figure 2.37 is usually not liquid; it is aerated liquid.
   

   Aerated liquid does not follow the liquid laws; it follows the aerated liquid laws, which state that all the pipes downstream of the draw, all the way to the lowest point, including any horizontal sections, and including the draw nozzle, need to be sized for self-venting flow. A self-venting flow is where liquid descends and any entrapped vapor bubbles rise back up to be vented back through the line and draw nozzle. Just like they do in a downcomer. If the line is not large enough for self-venting flow, the bubbles will accumulate in the line and choke it, initiating flooding or instability.

   
   

   Figure 4.9 is a graphical version of an excellent correlation for self-venting flow by Simpson (439) and later by Sewell (425). Above the diagonal, the flow is self-venting. Below the diagonal the downpipes are likely to choke, backing up liquid in the tower. The correlation straight line extrapolates to larger and smaller pipe diameters.

   
   

   Cases 10.1 and 10.2 in Distillation Troubleshooting (201) describe case studies of premature tower floods due to product lines undersized for self-venting flow. Many more were described in the literature (8, 18, 198, 280, 489). Some of these resulted in poor product quality or instability rather than flood.

   
   

   There are a number of situations where one can often get away with draw lines undersized for self-venting flows:

   
   
   
   
   • When the line originates in a sump with more than 30–60 seconds residence time. This is a common situation when the line originates in a tower bottom sump (typical residence time three to five minutes) or in a chimney tray that holds liquid level above the nozzle (typical residence time of about one minute).
     
   • When the line feeds a low-head centrifugal pump. A typical example is pumparound pumps. It appears that these pumps suck in the bubbles. This may not extend to high-head pumps in which vapor bubbles may lead to cavitation.
     
   • When a subcooled stream enters the sump and quenches the bubbles.
   
   
2. Observe what happens when increasing the draw rate (e.g., by opening the control valve). If opening the control valve does not increase the flow rate, or if the draw pump cavitates, it means that no more liquid is available from the draw sump to the draw. For best results, install a static pressure gauge at the bottom of the line, just before the draw pump (if pumped), or just before the control valve if by gravity. The difference between the pressure gauge reading and the column pressure at the draw point (which can be estimated from the column pressure transmitters) should equal the static head in the line. From this reading one can infer how liquid-full the line is at different moments during the test. Likewise, gamma scans (Chapter 5) or neutron backscatter measurements (Section 6.3) can estimate the fluid density in the line. When practical, gamma scan or neutron scan the line.

Chimney Tray draws are not immune to side-draw bottlenecks. When they hold liquid, chimney trays usually, but not always, provide sufficient degassing time to degas the liquid (see criterion in item 1 above). In Figure 2.38, the pumpdown draw will benefit from the degassing time on the chimney tray, and if sufficient (>30–60 seconds), there will be no need for self-venting flow. Not so with the side draw in Figure 2.38; the cascade over the chimney tray weir re-entrains vapor into the side draw takeoff sump (250). If the side draw line is undersized for self-venting flow, liquid will back up onto the chimney tray, possibly resulting in instability, and/or overflow and/or a premature tower flood.

In another case (150), liquid to circulating thermosiphon side reboiler was drawn from a chimney tray, with the reboiler effluent returning to the same chimney tray. Overflow pipes transported the reflux to the tray below. Circulation through a thermosiphon reboiler generated a very large liquid flow through the chimney tray, depriving it from the residence time needed for degassing. The reboiler draw lines were undersized for self-venting flow of the huge recirculating liquid flow and choked. The excess liquid overwhelmed the overflow pipes, so it accumulated in the tower and flooded it, with gamma scans showing 15 ft of liquid piling up above the chimney tray. This scan observation was the key to the diagnosis. The problem was resolved primarily by returning the reboiler effluent to the tray below the reboiler (in essence, converting the circulating thermosiphon reboiler to a once-through thermosiphon reboiler), which eliminated the excessive circulation through the chimney tray and the reboiler draw line.

Chimney tray overflow or premature flood due to downpipes undersized for self-venting flow are discussed with reference to packed tower case studies under “collectors” in Section 4.6. Tray case studies are found in Section 9.2.5 and in Refs. 118 and 388.