“The liquid distributor is often considered to be the most important part of a packed bed design. A good distributor cannot make a bad packing perform well, but a bad distributor can certainly make a good packing perform poorly”

—(Dhabalia and Pilling, Ref. 83)

4.1 DIAGNOSING PACKING MALDISTRIBUTION: AN OVERVIEW

Figure 4.1 shows two of the most common basic types of liquid distributors.

An orifice pan distributor (Figure 4.1a) is a pan containing evenly spaced floor perforations, which may be equipped with short tubes, through which the liquid is distributed to the packed bed below, and with round or rectangular riser tubes through which vapor from the bed below passes without contacting the liquid. Liquid is usually fed to the distributor by an elbowed-down pipe above its center or by a pipe with downward-pointing perforations above its center.

In a trough distributor (Figure 4.1b), liquid is fed to a rectangular box called a parting box. The parting box contains holes that apportion the liquid to the troughs below, such that the larger troughs receive more liquid and the smaller ones receive less. The troughs contain evenly spaced openings, either at the bottom (visible in the three troughs in the bottom left of the photograph), or in the side of the troughs near the bottom with downward-deflecting baffles or downflow tubes, through which liquid descends onto the packed bed below. Liquid is fed to the parting box by a central sparger pipe along its length with downward-pointing perforations which may be equipped with downflow tubes.

Many additional features are often incorporated to optimize the performance of either distributor type. Additional common distributor types are notched distributors (Section 4.14) and spray distributors (Section 4.15).

Troubleshooting Our malfunctions survey (198, 201) found liquid distributors to be the second most troublesome internal in distillation towers. The number of malfunctions was higher in chemical than in refinery towers, probably due to the wider spread application of packings in chemical towers. The survey identified liquid distributor issues as the prime malfunction in chemical towers.

Plugging and overflow were the two major liquid distributor issues. Good filtration and fouling-resistant distributor designs were effective cures. While overflows were often caused by plugging, many more resulted from distributor overloading caused by excessive liquid flows, insufficient orifice area, and hydraulic problems with the feed entry.

Poor irrigation quality, fabrication/installation issues, and poor feed entry configurations were other common issues (198, 201, 239). Poor irrigation quality, which the literature has focused on, accounted for only 20% of the reported liquid distributor malfunctions. The more troublesome items, such as plugging, overflows, and feed entry, received disproportionally less attention.

Feed entry issues, such as excessive velocities, splashing, and poor orientation of the feed pipes, frequently caused poor packing efficiencies. Flashing feed entry into a liquid distributor has been especially troublesome and a fertile source of malfunctions.

Inadequate hole pattern, distributor damage, infrequent redistribution, out-of-levelness, insufficient mixing, and interference with hold-downs constitute the remaining common issues (198, 201, 239). Surprisingly, distributor out-of-levelness was low on the list. Insufficient mixing was mostly troublesome in larger diameter (>10 ft ID) towers and in certain applications like extractive distillation.

Olsson’s (346) classic article has been the leading reference for distributor troubleshooting. His key advice is “ASSUME NOTHING.” Olsson advocates critically examining the fouling potential and possibility of vaporization in streams entering the distributor, testing distributors by running water through them at the design rates, either in the shop or in situ, and finally, good process inspection. Following Olsson’s guidelines would have prevented more than 80–90% of the reported liquid distributor malfunctions in Refs. (198) and (201).

Vapor is easier to distribute than liquid, but since common vapor-distributing devices are far less sophisticated, it is not trouble-free. The prime source of vapor maldistribution has been undersized reboiler return or gas inlet nozzles, which shoot high-velocity jets into the tower (198, 201). These jets persist through low-pressure drop devices like packings. Interference with packing support beams has also been troublesome. Installing vapor distributors, improving vapor distributor designs, and even using inlet baffles have alleviated many of these problems.

This section describes key techniques for diagnosing the nature and cause of maldistribution. Gamma scans (Chapters 5 and 6) and thermal wall measurement by pyrometers or thermal cameras (Chapter 7) provide additional excellent diagnostics. Drawing to-scale sketches is powerful for diagnosing point of transition issues (Chapter 8) at feed and vapor entry and redistribution. For maximum effectiveness, the troubleshooter is advised to combine these techniques.

4.2 EXPECTED PACKING HETPS

A rule of thumb for packing HETP (in inches), as a function of a_p, the packing surface area per unit volume (m²/m³), gives:

For organic and hydrocarbon systems, surface tension <25 dynes/cm:

• HETP = 93/a_p Metal random packings, 1″ or larger
  
• HETP = 100/a_p + 0.1 Y-type or third-generation structured packings
  
• HETP = 145/a_p + 0.1 X-type structured packings, a_p < 300 m²/m³

Multiply the above values by two for water due to its higher surface tension (70 dynes/cm). For intermediate surface tensions, interpolate.

These rules of thumb are generally plus or minus 25% and assume good industrial-grade liquid and vapor distribution (not perfect distribution like in test columns).

Measured HETP values that are significantly higher indicate possible or likely maldistribution.

4.3 SMALL-SCALE VERSUS LARGE-SCALE MALDISTRIBUTION: DO THEY EQUALLY RAISE HETP?

This topic is discussed at length with extensive literature citation in Kister’s book (193) and Perry’s Handbook (245). Below is a summary focusing on diagnostics rather than design:

1. Three factors define the effect of maldistribution on efficiency:
   
   
   
   1. Pinching. Zonal variations in L/V ratio cause zonal composition pinches. In the rectifying section, the liquid-deficient zones have low L/V ratios, and their composition approaches a pinch. Upon reaching a pinch, composition change stalls. The converse occurs in stripping zones; here local pinches occur in high L/V zones, with little composition change above.
      
   2. Lateral mixing. Liquid and vapor are deflected laterally by packing particles or corrugated layers, promoting mixing of vapor and liquid and counteracting the pinching.
      
   3. Liquid nonuniformity. Liquid exits commercial distributors with some nonuniformity. Liquid flows through the packing tend to concentrate at the wall.
   
   
2. At small ratios of tower-to-packing diameter (D_T/D_p < 10), the lateral mixing offsets the pinch, and a greater degree of maldistribution can be tolerated without a serious efficiency loss. As the D_T/D_p ratio increases, the lateral mixing becomes progressively less effective to offset the pinch, so the loss of efficiency to maldistribution escalates. In large-diameter columns and small-diameter packings, when D_T/D_p > 100, lateral mixing becomes ineffective to counter the pinch.
   
3. Wall flow effects become large only when D_T/D_p < 10.
   
4. Maldistribution tends to be a greater problem at low liquid flow rates than at high liquid flow rates (521). The tendency to pinch and spread unevenly is generally higher at the lower liquid flow rates.
   
5. Packing HETP has reasonable tolerance for a small-scale maldistribution, i.e., local maldistribution over a small area, and for a random variation. HETP is intolerant of large-scale maldistribution, i.e., large regions deficient of liquid, overirrigated, or zonal variations (267–269, 521). Tests in a commercial-scale column (268, 269) show no significant efficiency loss due to random, small-scale unevenness in the liquid load. The local pinching of small-scale maldistribution is evened out by lateral mixing, and therefore causes few ill effects. In contrast, the lateral mixing is powerless to rectify large-scale maldistribution, and efficiency is lost.

4.4 BY HOW MUCH DOES MALDISTRIBUTION REDUCE PACKING EFFICIENCY?

Figure 4.2 shows the Moore and Rukovena (326) empirical correlation for efficiency loss due to liquid maldistribution in packed towers containing Pall® rings or IMTP® packing. This correlation is simple to use and was shown to work well for several case studies (Figure 4.2), making it valuable, at least as a preliminary guide.

The quality of liquid irrigation is expressed by the Moore and Rukovena distribution quality rating index (Section 4.10). Typical indexes are:

• 10–40% for poor distributors,
  
• 40–70% for most standard commercial distributors,
  
• 75–90% for intermediate-quality distributors,
  
• 90% for high-performance distributors.

Locckett and Billingham (36, 301) modeled maldistribution as two columns in parallel (Figure 4.3a), one receiving more liquid (1 + f)L and the other less (1 − f)L, with the vapor assumed to be equally split between the columns. f is the maldistribution fraction, and L is the liquid flow rate in molar units. At the different L/V ratios, the overall separation is inferior to that at uniform L/V. As the maldistribution fraction f increases, the effective number of stages in the combined two-column system declines (Figure 4.3b). The decline is minimal when the maldistribution fraction f is small or in short beds (e.g., 10 theoretical stages), as noted earlier by Strigle’s extensive experience (462).

Figure 4.3b shows that beyond a limiting maldistribution fraction f_max, the required separation cannot be achieved even if the bed is tall enough to provide many stages. Physically, at f_max the liquid-deficient column becomes pinched. Figure 4.3c shows that the steep drop in packing efficiency due to this pinch occurs earlier when the bed contains many stages and that the efficiency loss can be tremendous. For f_max in a binary system, Billingham and Lockett derived the following equation:

(4.1)

In Eq. (4.1), x and y are mole fractions of the more volatile component in the liquid and vapor leaving a stage, respectively, and the subscripts 0, 1, N, and N + 1 signify stage numbers, with 0 being the lowest stage and N + 1 the top stage. This equation permits the calculation of f_max directly without the need for a parallel column model. The terms in Eq. (4.1) can be readily calculated from the output of a steady-state commercial simulation. This method can be extended to multicomponent systems either by lumping components together to form a binary mixture of pseudolight and pseudoheavy components or by normalizing the mole fractions of the two key components. Once f_max is calculated, Billingham and Lockett propose the following guidelines:

• f_max <0.05, extremely sensitive to maldistribution. The required separation will probably not be achieved.
  
• 0.05 <f_max <0.10, sensitive to maldistribution, but separation can probably be achieved.
  
• 0.10 <f_max <0.20, not particularly sensitive to maldistribution.
  
• f_max >0.20, insensitive to maldistribution.

Figure 4.3c shows that shortening the bed can increase f_max. Relative volatility and L/V ratio also affect f_max. The bed length and L/V ratio can often be adjusted to render the bed less sensitive to maldistribution.

Analysis of field data using similar principles was discussed by Krishnamoorthy and Yang (264).

4.5 DIAGNOSING PACKING AND DISTRIBUTOR PLUGGING

The following steps have been recommended:

1. Consider all possible sources of solids and identify those likely to be active in your tower (346, 347). Common sources include solids in the feed, catalyst fines, rust in the tower or piping, polymerization, precipitation, crystallization, reaction, tars, antifoam breakdown, gasket disintegration, dirt, debris, poor turnaround cleaning, and commissioning operations. Some solids are living (slime, bacteria). Fouling mechanisms may influence each other; for example, catalyst fines can induce polymerization. Check if you can make a call whether the solids are formed internally in the tower or externally.
   
2. Check the potential for solids in the feed. Do not just ask Operations. If possible, sample the feed and closely examine it. Murkiness or coloration may be due to suspended solids. Allow the sample to settle for a few days and observe if it becomes clear. Alternatively, pass the sample through fine filters and see if it clears.
   

   Figure 4.4 shows samples from the feed stream to a chemical column in which trays were replaced by packings (201, 346, 347). Prior to the retrofit, Operations reported no solids in the feed. Samples were drawn and showed coloration (similar to samples 1, 2, and 3 in Figure 4.4) but no solid deposits. After startup, the feed distributor plugged up, causing massive maldistribution, poor packing efficiency, and a capacity bottleneck.

   
   

   Feed samples 1, 2, and 3 show colorations, but there is no telling whether these indicate suspended solids. The clear sample on the left, drawn after a fine filter was installed on the feed stream, clearly shows that the colorations were due to suspended solids. Deposits were also observed when the color samples were allowed to settle over a week or two. The filter eliminated the plugging problem.

   
   

   When Operations say they have no problem with solids in the tower, their concise statement should be interpreted as “we have not seen solids in the feed, nor have we experienced fouling issues with our existing (at the time, trayed) internals.” Packing distributors and many packing types are far more prone to solids than trays, so a solids issue hidden in a tray tower may become apparent following a packing retrofit.

   
   
3. Good filtration should use two full-sized filters in parallel, operating one on one off, without a bypass. The filter openings should be at least 10–20 times smaller than the smallest hole in the distributor. The author is familiar with cases of distributor plugging when the filter openings were only four times smaller than the distributor openings, as in Case 6.4A in Ref. (201). Distributors usually have dead spots where solids can agglomerate, later spalling off and plugging the distributor holes.

   

   
   

   Have Mechanical check that the filter screens and baskets are adequately mounted. Liquids can easily find a path around an improperly mounted basket. Check the filter pressure drop and log it over a time period. Loss of pressure drop across a filter may signify a damaged or removed basket.

   
   

   Check the frequency of filter cleanings. Expect a high frequency during startup and initial run, as the feed moves along residual solids lodged in the line. Closely watch the action of the Operations team. When cleaning is too frequent, especially during startup, baskets are often completely removed, resulting in solids ending up in the distributors or packings (e.g., 216, Case 1). Line up additional personnel to help during these busy periods.

   
   

   Refer Section 10.3.11 for additional recommended filter checks.

   
   
4. It is essential to thoroughly flush all lines and auxiliary equipment from high to low points before connecting to the column to avoid introduction of construction debris, dirt, corrosion products, and other plugging materials. Verify that this has been done. Cases 6.4C, D, and E in Ref. (201), as well as Ref. (282), describe incidents in which construction debris and rust caused distributor plugging.
   
5. Avoid water entry into carbon steel towers processing water-insoluble components. Water circulating in such towers can become acidic and generate corrosion products that plug distributors.
   
6. Distributors with small holes often have the holes at the wall of the distributor troughs or in downflow tubes, typically about 1″ or 2″ above the floor. A downflow tube (Figure 4.5) or a baffle directs the liquid vertically downward into the packings. Elevating the holes resists plugging by sediment settling on the floor. In fouling services, distributor run length with wall perforations is typically double that of floor perforations. Perforations of 0.5″ and larger are fouling-resistant, and are usually located in the floor.
   

   With reactive components at high temperatures, elevating holes needs a careful review. Below the perforations, the liquid residence time is high, and when “cooked” by hot vapor, it can undergo undesirable reactions, polymerize, or turn into tars or sludge.

   
   
7. Do not underestimate the fouling potential of a service. “A cup of scale can stop up a distributor” (25). Plugging has taken place in seemingly clean systems. In one pan distributor with tubes containing two rows of holes (Figure 4.5), the holes were plugged, the distributor overflowed as evidenced by water marks, and column performance was poor. The amount of solids that plugged the holes was minute, would not fill a cup of coffee, and was mainly rust particles. In another case (413), less than 1 lb of solids plugged 80% of the perforations.

   

   
   
8. Plugging-resistant features can be incorporated, often at the price of inferior distribution under clean conditions or at higher costs. Increasing hole diameters greatly improves plugging resistance, usually at the expense of fewer distribution points and lower liquid heads. Experience showed (218, 320) that a lower-quality distributor that works beats a high-irrigation quality distributor that plugs. “Tea strainers” hole-opening guards have successfully prevented plugging in at least some cases. Two-stage distributors, some with rectangular notches instead of holes, are also effective for improving plugging resistance while maintaining irrigation quality (69, 168, 188), but at a higher cost. Notched-trough distributors (Section 4.14) have been successful in fouling services (387, 523), but often provide a lower irrigation quality (192, 245, 267).
   
9. “Newer is not always better” (282). New state-of-the art distributors can be far more prone to plugging than their low-quality predecessors. In one case, a lower-quality ladder pipe distributor replacement by a “newer and better” trough type introduced severe plugging that was later eliminated by returning to the proven ladder pipe distributor.
   
10. Corrosion products, scale and salts formed at the top head of a tower can spall off into the distributor troughs below as in Case 6.4B in Ref. (201). Well-designed trough covers or screens can eliminate this.
    
11. In many services, polymerization inhibitor is injected into the reflux. Outer distribution points are usually about 1–1.5″ away from the tower wall. The open lattice structure of modern random packings does not spread the liquid well, so it may take a foot or two for the liquid to reach the wall. This region remains dry, receives no inhibitor, and may polymerize. First- and second-generation packings are much better liquid spreaders, shortening the dry length at the wall. In one retrofit, changing second- by fourth-generation random packings caused polymer plugging not previously experienced. See Figure 4.16 and Section 4.10 on identifying underirrigated zones near the wall.
    
12. Beware the possibility of commissioning operations introducing solids that plug distributors. Examples include flushing feed or reflux lines into the tower, washing the tower with solid-containing water, opening the tower to the atmosphere after a water wash without adequate drying (rust). Consider temperature differences between operation and commissioning; in one case (3), residual solids were not removed during a cold flush, but were dislodged later when the system was heated.
    
13. Consider the possibility of component freezing and plugging distributors when cold reflux or feed enter the column, either during normal or during commissioning operations. O’Brien and Porter (343) provided valuable guidelines.
    
14. Consider injecting a solvent, either continuously or batchwise, to dissolve the solids. Typical examples include the on-line caustic wash of an aldehyde scrubber in Case 12.3 of Ref. (201), or the pyrolysis gasoline washes to dissolve polymer in olefins plant caustic towers, as described in Ref. (259).
    
15. Consider the possibility of changing operation conditions to dissolve solids. Examples are raising boilup and cutting reflux in a methanol–water tower, thereby inducing water to the top section to dissolve solids that built up there (285), or using water washes to dissolve salts from the top of refinery towers, described by many case studies in Ref. (201).
    
16. In vacuum towers processing chemicals that easily polymerize or tar, perform extensive leakage tests before startup to minimize air leakage.

4.6 TROUBLESHOOTING FOR DISTRIBUTOR, COLLECTOR, AND PARTING BOX OVERFLOWS

Overflows are the most common distribution issues, often referred to as “Death to packed beds.”

4.6.2 Overflows in Distributors, Redistributors, and Parting Boxes

In distributors and redistributors Overflow-induced flooding (Section 4.6.1) that propagates up the tower is most severe with pan distributors due to their small riser areas. Trough distributors have much larger vapor passage area, which often allows channeling of vapor through some of the vapor passages with the overflowing liquid finding its way down through the others. Sheet or rivulet formation by the overflowing liquid makes it easier to descend. This drains the accumulating liquid and alleviates the flood, at the expense of severe maldistribution and efficiency loss in the bed below. The overflow process may unpredictably alternate between flooding, flooding and dumping, and the channeling modes, generating tower instability.

The troughs in some distributor designs contain slots near the top that direct overflowing liquid into guide tubes, that in turn take this liquid to the bed, out of the path of the vapor. These also minimize maldistribution. The effectiveness of these designs varies with the amount overflowing and the guide tube design.

Figure 4.6a is a photograph taken during a turnaround in situ water test of a redistributor on the verge of overflowing. At that liquid flow rate, the tower worked well, but when raised beyond this point, an overflow occurred. Figure 4.6b is a close-up of two risers experiencing overflow in the water test. During normal operation, vapor flowing through the risers prevented liquid descent. Beyond the overflow rate, flooding initiated, propagating through the bed above and causing liquid carryover into the tower overhead product stream.

Diagnosing overflows The techniques below, as well as gamma scans (Chapters 5 and 6), and to a lesser extent, thermal scans (Chapter 7), have been effective for diagnosing overflows.

An invaluable tool for diagnosing overflows is the head–flow relationship for packing orifice distributors, given by the Torricelli Orifice Equation:

(4.2)

where Q is the liquid flow rate, gpm; K_D is the orifice discharge coefficient, with a recommended value of 0.707 (68, 161, 192, 245); n_D is the number of holes; d_h is the hole diameter, inches; and h is the net liquid head, inches. The actual liquid height in the distributor, H_actual, can be obtained by applying buoyancy and pressure drop corrections to the net liquid head, both of which will tend to raise the liquid height in the distributor:

(4.3)

where ρ_L and ρ_V are liquid and vapor densities, lb/ft³, and ΔP_{vapor rise} is the vapor pressure drop through the distributor risers (pan distributors) or the vapor rise areas (trough distributors), inches of liquid. In most pan distributors, ΔP_{vapor rise} is of the order of 0.25 in. of liquid (161, 192), and is less in trough distributors. For vacuum and atmospheric systems, the buoyancy correction is negligible, but it becomes significant at higher pressures (>50–100 psia). The vapor rise pressure drop can be estimated from Eq. (4.4)

(4.4)

where V_R is the vapor velocity through the risers or vapor passages, ft/s. K_R is the hydraulic constant representing the number of velocity heads lost through the risers. As a conservative estimate, K_R = 2.5, based on the head loss through an orifice. In reality, the riser is a short pipe with entrance and exit losses (hydraulic K’s of 0.5 and 1, respectively), giving K_R = 1.5. The correct value is probably somewhere between, depending on the riser length, with K_R = 1.5 more appropriate for long risers (>18″).

Applying Eq. (4.2) for diagnosing overflows Consider a distributor with wall perforations about 1 in. above the floor. For turndown, a minimum net head of about 2 in. above the perforations is required. Turning up the flow to twice the minimum will raise the net head to 2 × 2² = 8 in. per Eq. (4.2). Since the perforations are already 1 in. above the floor, the liquid height will be 9 in. above the floor, assuming negligible buoyancy and pressure drop corrections in Eq. (4.3). If one wants to turn up the flow to quadruple of the minimum, the required liquid head above the perforations by Eq. (4.2) will need to be 2 × 4² = 32 in., which is 33 in. (=32 + 1) above the floor. To safeguard from overflow, the troughs need to be about 40 in. high. The author had never seen a commercial distributor with even close to 40 in. tall troughs. Typical trough heights in the industry are 8–12 in., mostly closer to 8 in.. Of these, the bottom inch is the elevation of the perforations above the floor, and the top inch is often slotted to permit orderly overflow should it occur. With 8 in. tall troughs, a liquid height of 9 in. above the floor will be 2 in. above the bottom of the overflow slots and will overflow into the vapor passages. With 12″ tall troughs, the liquid level will be 2 in. below the bottom of the overflow slots.

This example illustrates why overflow is such a severe problem. Per the orifice Eq. (4.2), the liquid head rises parabolically with the liquid flow rate and very easily reaches the overflow. Taller troughs or risers will make the distributor much more robust for overflow prevention.

Do not overlook the pressure drop term. In two small-diameter towers (1.25–1.5 ft), the vapor riser areas in pan distributors were about 10% of the tower cross-section areas, and the number of holes was excessive. The liquid head was nearly equal to the vapor pressure drop through the distributors. This is believed to have induced vapor to break through the distributor liquid, causing frothing, turbulence, and the observed poor efficiency (433).

In parting boxes Overflows occur not only in the distributors but also in their parting boxes, which are (Figure 4.1) the boxes that split the incoming liquid to the various toughs of a distributor. Figure 4.7 shows a photograph of a liquid overflow in a water test of a model parting box. The consequences are much the same as those of a distributor overflow.

In parting boxes, there is often considerable aeration, which drives up the froth level, resulting in an early overflow (247, 248). Water tests with the sparger-fed model parting box, as depicted in Figure 4.8, showed froth heights typically 20–30% taller than the liquid heads calculated using the orifice Eq. (4.2). Using dip tubes that bring liquid from the feed pipe to near the parting box floor eliminates the frothing, but generates a horizontal momentum effect that also causes the liquid head in the parting box to exceed the liquid head calculated using the orifice equation by 20–30% (249). Dip tube designs that break the momentum can reduce the parting box liquid heads to the values calculated using the orifice equation (249).

4.6.3 Overflows in Collectors

Packed tower collectors are common overflow producers. From the collector, liquid descends to a redistributor either by downcomers, downpipes (most common), or by perforations. If the descent is by perforations, the orifice Eq. (4.2) can be used to check for overflow, with any possible aeration taken into account. If the descent is by downpipes or downcomers, the residence time of liquid on the collector between the overflow level and the top of the downpipe should be calculated. If this time is less than 30–60 seconds, the liquid is likely to be aerated and the downpipes need to be sized for self-venting flow, i.e., a flow pattern in which liquid descends while any trapped vapor bubbles disengage back up.

Figure 4.9 shows an excellent correlation for self-venting flow, initially established 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, causing the liquid to back up the top of the risers, generating an overflow. The overflow liquid is maldistributed and may generate flooding just like other overflows. The correlation is often expressed (439) as maintaining the liquid Froude Number in the downpipes less than 0.31.

The cases described in Refs. (416) and (417) are classic examples of painful collector overflow experiences. To improve the yield of the valuable jet fuel, trays in the jet fuel–diesel separation section (as well as other sections) of an atmospheric crude tower were replaced by structured packings. Upon restart, the separation worsened instead of improving, and the jet fuel yield was reduced by half. A superb troubleshooting investigation diagnosed the root cause to be collector overflow generating maldistribution in the jet fuel–diesel separation packed bed below. The liquid residence time in the collector was low, and the downpipes transporting its liquid to the distributor below were undersized for self-venting flow. Level measurements and gamma scans showed liquid level reaching the overflow, as well as pulsations indicating frequent intermittent dumps.

Two years later, the same refinery performed a similar revamp in a parallel (second) crude tower of similar design. Incorporating the lesson learned, a chimney tray collector with downpipes sized for self-venting flow was specified to the supplier. Upon restart, jet fuel–diesel separation was not better, with low jet fuel yield again. Tests, simulations, gamma scans, and computer-aided tomography (CAT) scan pointed again to a collector overflow. This time, the overflow was caused by excessive hydraulic gradients on the chimney tray, incurred by chimneys perpendicular to the liquid flow, similar to the problem described in Section 8.10 and Figure 8.11a. Additional obstruction to liquid flow was due to triangular chimney reinforcements that severely hindered flow.

Learning from these experiences, when the first unit was turned around, the collector was replaced with a chimney tray containing smaller chimneys that occupied less of the cross-section area and allowed liquid movement. The downpipes were replaced by down boxes with ample area for self-venting flow. This time, excellent separation and the desired yield were achieved.

The second unit was less fortunate. After its revamp, it performed even worse than before (417). An in situ water test during an outage showed skewed liquid distribution. Visual observations and CFD showed that the long chimneys on the collector and their positions too close to downpipes and down boxes created “shadow” regions (Figure 8.22, Section 8.18, Case 2). The long chimneys obstructed liquid flow to these regions, so their down boxes received much less liquid than other unobstructed down boxes. Success was achieved by replacing with a new chimney tray in which liquid movement to the down boxes was unobstructed by chimneys.

Reference (416) states that overflow from collectors that had low residence times and whose downpipes were undersized for self-venting flow occurred in two other refineries by the same company. All these towers were made to work by good chimney tray design.

Case 7.1 in Ref. (201), as well as Refs. (129, 275), and (482) describe case studies of very poor packed tower performance resulting from collector overflows. The author is familiar with many more.

A summary of the lessons learned for collectors is expressed in Ref. (417) “If it starts the wrong way, it will keep that way.” If the collector delivers an uneven feed to the distributor, distribution to the bed is likely to remain poor.

4.7 TROUBLESHOOTING MALDISTRIBUTION AT TURNDOWN

Liquid maldistribution tends to lower packing turndown (192, 268, 462)

The orifice Eq. (4.2) is also powerful for addressing distributor turndown. In the example in the previous section, the liquid head at maximum flow was 8 in. above the perforations. At half that flow (a turndown of 2:1), the head declined to (8/2²=) 2 in. Often, 2 in. is considered the lowest head needed to ensure a reasonable distribution. If instead of a turndown of 2:1 it is desired to go to a 3:1, the liquid head per Eq. (4.2) will be (8/3²=) 0.9 in. At such a low head, the liquid is likely to be maldistributed. Also, at this low head the distribution is likely to be sensitive to even small degrees of out-of-levelness, waves, turbulence, and uneven frothing. The turndown on most orifice distributors is about 2:1, which is limited by the orifice equation.

There are certain features that packing suppliers often incorporate to enhance turndown. A common feature is using two perforations, one above the other, as shown in Figure 4.5. Often the perforations are of different sizes, the lower smaller ones active at low liquid rates, and both active at high liquid rates. This is not entirely trouble free. It may experience a “bump” when the liquid level just rises past the upper perforations, especially when these upper perforations are larger (which is often the case), as the liquid head building above the upper perforations will be too small to evenly distribute the liquid issuing from them. In a recent case, we identified this issue as one of the causes of poor packing efficiency experienced in a tower. Also, with two perforations along the wall, the lower ones usually need to be smaller, and therefore more prone to plugging, than with a single perforation.

Figure 4.10 illustrates that packing turndown is drastically reduced by maldistribution. The “standard distributor” curve depicts typical variation of HETP with vapor or liquid load at a constant L/V ratio. The two upper curves depict a progressively lower quality of initial liquid distribution. The unevenness produced by maldistribution is amplified when the liquid flow rates and the liquid heads in the distributors are lower. A curve like the upper curve in Figure 4.10 is a clear indication of poor distribution.

4.8 TROUBLESHOOTING DISTRIBUTOR OUT-OF-LEVELNESS

Out-of-levelness in distributors has been a common issue causing poor efficiency in packed towers, as demonstrated in numerous case studies (137, 201, 450, 483, 508).

The orifice Eq. (4.2) is also powerful for addressing issues related to distributor out-of-levelness. In the example in the previous section, the liquid head at a turndown of 3:1 was about 1 in. Consider a tilt of 0.5 in., such that at the liquid head is 1 inch at the shallow end and 1.5 in. at the deep end. If the liquid flow rate at the shallow end (1 in. head) is 100 gpm, then at the deep end (1.5 inch head) the flow per Eq. (4.2) will be (100 × 1.5^0.5=) 122 gpm, a difference of 22%. Since most towers are designed to operate at 20–30% above minimum reflux (193), a 22% flow variation can easily get the bed section beneath the shallow end into a pinch, approaching or even falling below the minimum reflux. At this low liquid head, even a small amount of out-of-levelness will be disastrous to packing efficiency.

Furthermore, at such low liquid heads, part of the distributor where the liquid heads are low may approach drying up, making the problem even greater.

Consider the same tilt when the liquid head in the distributor is at 8 in. The shallow end will have a liquid level 8 in. above the perforations, and the deep end will have a level of 8.5 in. If the liquid flow rate at the shallow end (8″ head) is 100 gpm, then at the deep end (8.5 in. head), the flow rate per Eq. (4.2) would be (100 × [8.5/8]^0.5 = 103 gpm, a difference of only 3%. So at high liquid heads in the distributor, out-of-levelness is a much lesser issue than at low liquid heads.

The author was involved with tilt testing on a specialty distributor. The tests found that a 0.5 in. tilt made little difference to packing HETP. This raised a few eyebrows until it was appreciated that the test was performed with a 10″ liquid head in the distributor. At such tall heads — and only at such tall heads — the distributor becomes tolerant to out-of-levelness.

Figure 4.11 shows FRI test data (268) for separating C₆ from C₇ hydrocarbons in a 4 ft tower, packed with 1 in. Pall rings operated under different types of simulated maldistribution. The y-axis is the ratio of measured HETP to the base (excellent distribution) HETP. The two lower curves simulate distributor tilt, expressed as the ratio of liquid flow at the deep end to that at the shallow end (for instance, 1.25 means that the flow at the deep end was 1.25 times that at the shallow end). The x-axis is the percent of usable capacity at a constant L/V operation, so the higher this percent is, the higher the liquid flow rate and the higher the liquid head in the distributor. The diagram echoes the above message, showing a much greater effect of tilt at turndown, when the liquid head is lower.

Closely review the out-of-levelness of the parting box. Parting box tilt will induce preferential higher flows to the troughs beneath the deep end and lower flows to those beneath the shallow end, as reported in one case (409).

Hydraulic gradients, especially in narrow troughs, can also be troublesome. The hydraulic gradient in a trough can be calculated from (40, 192, 326)

(4.5)

where

• h_hg = hydraulic gradient head, inches of liquid
  
• v_H = liquid horizontal velocity in the trough, based on the wetted trough area, ft/s.

Distributor troughs with floor perforations have less wetted area than those with wall perforations and, therefore, larger hydraulic gradients. It was recommended to keep the maximum transverse liquid velocity in the troughs less than 1 ft/s (40). The maximum value is at the point of liquid supply to the troughs.

4.9 TROUBLESHOOTING DISTRIBUTOR FEEDS

Feed entry ranks among the most common causes of maldistribution and capacity bottlenecks in packed towers. The following guidelines are recommended:

1. Draw a sketch to scale showing the inlet arrangement (Chapter 8). This is the most powerful technique for identifying these problems. Sections 8.13 and 8.18–8.23 contain cases involving packed tower feeds.
   
2. Check for the presence of vapor in all liquid feeds. Liquid feed distributors cannot handle vapor, even in seemingly small quantities. Note that a small quantity of vapor by weight, say 1%, can mean 90%+ vapor by volume due to the lower vapor density. If any vapor is present, refer to the guidelines in Section 4.13.
   
3. Beware of the following common situations when unexpected vapor may be present in the feed. Closely check with a flash calculation at column entrance conditions. Critically review simulation results; many flash a feed to the tower pressure without reporting the vapor/liquid split resulting from this step, creating a false impression that the feed is all liquid at the tower entrance.
   
   
   
   • Liquid feed flashed from a higher pressure source to a lower pressure column.
     
   • Feeding the bottom stream from one column to another.
     
   • A preheater on the liquid feed stream.
     
   • Liquid feed to the tower is steam-traced or heat-traced.
     
   • Liquid feed to the tower contains a highly volatile component.
     
   • Feed mixing two liquid streams at different temperatures.
     
   • Alternative feed that is only occasionally used.
   
   
4. Use the sketch in item 1 above to closely examine the mixing between the feed and the tower liquid. Case 4 in Section 8.18 (and Figure 8.24) is an excellent example. This guideline is imperative when dealing with extractive distillation. The success of extractive distillation depends on good mixing between the solvent and reflux, and between the feed and solvent. Both must be closely reviewed to ensure good mixing. The cases in Section 8.19, and Figures 8.25–8.27, are excellent examples.
   
5. Closely check the dimensions in the sketch. Look for any restrictive passages for liquid flow.
   
6. When feeding via one or more parting boxes, check the following (248, 249):
   
   
   
   • Apply the orifice Eq. (4.2) to calculate the liquid head in the parting box. If the calculated head is more than 70–80% of the parting box height, check with the detailed guidelines in Refs. (248) and (249) for the possibility of overflow due to frothing or horizontal momentum effects.
     
   • Perform a Gamma scan of the parting box, shooting chords parallel to the length of the box, to measure the liquid level. Request the scanners to shoot every 1 inch along the box elevation to get a good handle on whether the box is overflowing.
     
   • Calculate the liquid velocities in the sparger holes or dip tubes and sparger pipes. Hole/dip tube velocities should be 4–5 ft/s or less. The hole/dip tube velocities should not be less than the sparger hole velocity.
     
   • Closely compare the liquid entry to the parting box to the various alternatives in Refs. (248) and (249). Many common sparger arrangements, with or without dip tubes, give mediocre, even poor split of liquid to the distributor troughs below, especially at lower liquid heads.
     
   • Verify that the parting box liquid ends up in the troughs and does not miss them.
     
   • Test the tower at higher or lower liquid flow rates. If the packed bed produces more stages at higher liquid flow rates, one possibility is poor liquid split to the troughs at turndown. If the packed bed produces more stages at turndown, one possibility is an overflowing parting box. Keep in mind that improvements at lower or higher liquid loads may also be due to many other factors such as distributor and packing performance.
   
   
7. When liquid from a collector or parting box descends through several downpipes or down boxes, closely examine (with the aid of a sketch to scale, see Chapter 8) whether there is an even liquid split to the downpipes. In many cases (275, 390, 417), uneven split caused poor efficiency in the bed below. In Case 1 in Section 8.18, downpipes from a collector discharged into regions in the distributor almost totally enclosed within chimney walls, which obstructed flow to the rest of the distributor (Figure 8.21). In Cases 2 and 3, obstruction by chimneys or down boxes generated an uneven liquid split, causing liquid maldistribution and poor separation.
   
8. Parting boxes that handle highly subcooled (>50–70°F) feeds are prone to damage. Upon an abrupt feed rate step-up, vapor from the tower collapses onto the subcooled liquid, exerting strong forces that can deform troughs and parting boxes. Figure 4.12 is one example.

4.10 EVALUATION OF DISTRIBUTOR IRRIGATION QUALITY

An excellent evaluation method was developed by Moore and Rukovena (326). This method has gained wide acceptance (40, 192, 275, 336, 346, 347). Over the years, the author and his colleagues successfully extended this method to situations not addressed by the original Moore and Rukovena paper, and these extensions are incorporated in the following discussion.

The Moore and Rukovena method calculates a distribution quality index D_Q in percent, given by

(4.6)

Typical values of D_Q range from <40% for poor distributors, 40–70% for standard distributors, 75–90% for intermediate distributors, and >90% for high-performance distributors (326). To determine D_Q, each distributor drip point is represented by a circle. The center of each circle is where the liquid from this point strikes the top of the bed. The total area of all the circles is set equal to the column cross-section area, so when all the drip points have the same diameter, the area of each circle equals the tower cross-section area divided by the number of drip points. An extension by the author for distributors containing drip points of different diameters still sets the total circle area to equal the tower cross-section area, but proportions the circle area for each hole size to its hole area.

The terms in Eq. (4.6) are evaluated as follows (Figure 4.13a):

• A is the percent of the column cross-section area not covered by the circles. It provides a direct measure of the unirrigated area at the top of the bed.
  
• B is evaluated by selecting the most underirrigated or overirrigated continuous area at the top of the bed occupying one twelfth of the tower cross-section area. This is the area where the deviation from the average flow is maximum, and is a measure of the large-scale maldistribution. If this area is underirrigated, B is the ratio of the circle area in this region to the area of the region (1/12th of the tower cross-section area). If overirrigated, B is the ratio of the area of the region to the area of the circles. The lowest value of B anywhere at the top of the bed, expressed as a percent, is used in Eq. (4.6).
  
• C is the total area of overlap of adjacent drip point circles. It is expressed as a percent of the column cross-section area and represents the area that spreads the liquid unproductively.

Figure 4.13b–d illustrates the application of this technique to evaluate irrigation quality for various distributors. Figure 4.14a shows an application (346) in which this method diagnosed poor irrigation (D_Q = 45%) as the root cause of high packing HETP. The distributor preferentially sent liquid to the outer regions at the top and bottom of the diagram. Replacement by a high-performance distributor (D_Q = 90%; Figure 4.14b) improved staging by 40%. Application of the Moore and Rukovena method to trough distributors with side holes and external drip guides directing the liquid into the packings (336) is illustrated in Figure 4.15.

The Moore and Rukovena technique can be applied at three different levels:

• Basic, just drawing the circles per rules above. In all the cases the author experienced, this gave a good qualitative assessment of the distribution quality. This can readily be seen by inspecting Figures 4.13–4.15.

  

  
  
• Intermediate, also calculates the D_Q. This gives an excellent quantification, which the author found very reliable.
  
• Advanced, D_Q can be applied together with Figure 4.2 to quantitatively estimate the stages loss to maldistribution. The author’s experience with this advanced level is limited.

The author highly recommends this method for evaluating distributor irrigation at the basic and intermediate levels.

With small-diameter towers, B may be difficult to define. In this case, only the basic level can be applied with confidence. With a small number of drip points, double and triple overlaps may incur difficulties (40).

For structured packing, the Moore and Rukovena method may not fully characterize the distribution quality because it does not credit the liquid spreading performed by the top layer of packings, and therefore may penalize distributors that irrigate the packings in a drip line. Spiegel (457) proposed an alternative “wetting index,” which accounts for the liquid spread at the top layer of packing, but unfortunately did not provide generic guidelines for the application of this method. The author has had good experience applying the Moore and Rukovena method also to distributors that irrigate structured packings in a drip line. It works at least on the qualitative level, as long as one recognizes that the liquid will spread parallel to the layers. The author concurs with Spiegel’s point that the quantitative value of D_Q may penalize such distributors.

Perry et al. (362) presented a useful method, which they conceptually credit to Moore, for troubleshooting irrigation in the tower wall region. In modern designs, drip point patterns are computer-generated. Due to mechanical constraints and support considerations, the computer often locates outer drip points some distance from the tower wall, producing underirrigated zones, which go undetected. The Perry et al. method divides the column into three concentric zones of equal areas, and the number of drip points in each zone is counted by hand. This procedure can readily detect underirrigated zones near the tower wall, as illustrated in Figure 4.16, which is its main application (336, 362). For simplicity, Figure 4.16 does not specifically show the vapor risers, but these need to be taken into consideration when applying this method.

When reviewing the irrigation pattern, ensure that the liquid drip points are not too close to the vapor passages to be entrained by the vapor. Also, check that any drain holes do not pass quantities of liquid large enough to generate maldistribution. Finally, confirm that the distributor supports and packing hold-downs do not interfere with liquid distribution. In one case (261, 257), the distributor was supported by 4″ × 8″ plates that covered 8% of the packing, leading to poor separation.

4.11 TROUBLESHOOTING FOR VAPOR MALDISTRIBUTION

High-velocity jets entering the tower due to undersized gas inlet and reboiler return nozzles are the prime source of vapor maldistribution. These jets persist through low-pressure-drop devices such as packings. Vapor distributors, even well-designed inlet baffles, have alleviated such problems. Vapor distribution is most troublesome in large-diameter columns. Below are some guidelines for troubleshooting vapor maldistribution. In addition, many of the guidelines in the next section dealing with flashing feeds also apply for vapor distributors.

1. Strigle (462) recommends a vapor distributing device whenever the vapor nozzle F-factor [u_N is the vapor velocity at the narrowest area of the inlet pipe or nozzle, ft/s, and ρ_G is the gas or vapor density, lb/ft³] exceeds 22 ft/s (lb/ft³)^0.5, or the kinetic energy of the inlet gas exceeds eight times the pressure drop through the first foot of packing, or the pressure drop through the bed is less than 0.08 in. water/ft of packing. Cases of poor packing performance at high F_N values have been reported (69)
   
2. An alternative criterion proposed by Moore and Rukovena (326) replaces the F_N < 22 rule in item 1 above by a rule to keep F_N < 52.4 ΔP^0.5, where ΔP is the bed pressure drop in inches of water per foot of packing.
   
3. When none of the criteria in items 1 and 2 is violated, vapor maldistribution is unlikely to be troublesome, so there is no need for a vapor distributor (326). This applies to tower diameters less than 20 ft.
   
4. For nozzle F-factors in the range 52.4 ΔP^0.5 < F_N < 81.2 ΔP^0.5, Moore and Rukovena (326) recommend a simple vapor distributor such as a sparger pipe. At higher F_N, a sophisticated vapor distributor should be used (326).
   

   Kabakov and Rozen (186) measured severe gas maldistribution in 15-ft diameter CO₂ amine absorbers containing 50 ft of random-packed beds. They did not report values for F_N. The efficiency in these columns was about half that measured in another column that had a specially designed inlet gas sparger. Schafer et al. (415) report a case where adding a multi-vane vapor distributor greatly improved aromatics stripping from water.

   
   

   Using CFD, Wehrli et al. (499) found that a very simple device such as the V-baffle (Figure 8.14b) gives much better distribution than a bare nozzle. Both Wherli et al. and Schafer et al found that a more sophisticated sparger or vane distributor was even better.

   
   

   With a V-baffle, use a sketch to troubleshoot for possible adverse effects produced by the baffle. Some cases are discussed in in Sections 8.13 and 8.16. It is important to verify that vapor from the V-baffles does not impinge on the liquid below or interact to generate vapor maldistribution, as explained in Section 8.13 and in Case 3 in Section 8.16.

   
   

   In one case (271), with F_N ≫ 81.2 ΔP^0.5, the entering vapor jet repeatedly damaged the bottom of an 11-ft tall bed containing sturdy grid packing. The entry was similar to that in Figure 4.17a, but there was no chimney tray between the inlet and the bed. The damaged packing was opposite the vapor entry (Figure 4.17a). High vapor velocity, possibly with entrained liquid, was “literally ripping the bed apart from the bottom (271).” In another case (158), some coking was found in the bottom (slurry) grid bed of an fluid catalytic cracking (FCC) fractionator, most pronounced on the side of the column opposite the vapor feed inlet nozzle (again, similar to Figure 4.17a but no chimney tray). The authors (158) comment that “this type of slurry bed coking is not uncommon in the industry.”

   
   
5. The best vapor distributor in distillation columns is pressure drop. A high-pressure-drop vapor distributor provides good distribution, while a low-pressure-drop distributor does little to improve vapor distribution, as demonstrated (69) and illustrated in Figure 4.17. Schafer et al. (415) report a case in which adding a high-capacity tray between the stripping steam distributor and the bottom bed, which improved vapor distribution by increasing pressure drop, greatly improved aromatics stripping from water. In another case (415), replacement of ceramic packings with high-efficiency metal packings lost stripping efficiency because the lower pressure drop of the metal packings was less effective in correcting the vapor maldistribution. Improving vapor distribution solved this problem. In Case 6.1 in Ref. (201), increasing the vapor distributor pressure drop helped eliminate vapor maldistribution.

   

   
   

   To evaluate the effectiveness of the vapor distributing device, calculate its pressure drop and compare it to the inlet nozzle velocity head. Pressure drops through vapor distributors typically range from 1 to 8 inches of water (326) and can be calculated using Eq. (4.4). Strigle (462) recommends that the pressure drop through the vapor distributor should be at least equal to the velocity head at the inlet nozzle.

   
   
6. In the loading region, i.e., at high gas rates, liquid and vapor maldistribution interact (460). A poor initial liquid distribution may channel vapor and even induce local flooding. Similarly, liquid distribution pattern in the bottom structured packing layers is influenced by strongly maldistributed inlet vapor (350) or by local liquid accumulation (local flooding) induced by high vapor loads (Section 5.5.15). The author had experiences where increasing liquid loads helped alleviate vapor maldistribution in the bed.
   

   Outside the loading region, the influence of the liquid flow on gas maldistribution is small or negligible.

   
   
7. The effect of gas maldistribution on packing performance is not well understood. FRI’s commercial-scale tests (55) showed little effect of gas maldistribution on both random and structured packing efficiencies. In one case, blocking the central 50% or the chordal 30% of the 4 ft tower cross-section area beneath a 5 ft-tall bed of 250 m²/m³ structured packing had no effect on packing efficiency, pressure drop, or capacity. The blocking did not permit gas passage but allowed collection of the descending liquid. Simulator tests with similar blocking (350) differed, showing that a 50% chordal blank raised pressure drop, gave a poorer gas pattern, prematurely loaded the packing, and generated maldistribution that penetrated deeply into the bed.
   
8. Midspan I-beams can be deep, sometimes ranging from 1 to 2 ft. When the bottom of the I-beam is too close to the liquid level, it may interfere with the vapor distribution to the bed. For good vapor distribution, it was recommended that the pressure drop for half the vapor flow passing under the I-beam (between the bottom of the support and the liquid level) should be less than 10% of the pressure drop of the packed bed above (286). Troubleshoot this possibility by drawing a to-scale sketch as shown in Section 8.12, Figure 8.13a.
   
9. I-beam interference can be troublesome also on chimney trays, when the bottom of the I-beams gets close to the chimney hats. Check this by drawing a to-scale sketch as shown in Section 8.12, Figure 8.13b.
   
10. Chimney trays with risers ending too close to the bed above may initiate vapor maldistribution. For good distribution of vapor to a packed bed above chimney tray riser hats, Lieberman (285) and Lee (275) advocate a hypothetical angle of 30°, as illustrated in Section 8.12, Figure 8.13c. One case was presented (275) in which a 70° angle was too wide, generating vapor maldistribution and poor packing efficiency in the bed above.
    
11. Item 10 above extends to vapor inlet distributors discharging too close to the bed above. Hanson et al. (155) reported a case of severe coking in the grid wash bed of a coker main fractionator with a 54″ feed nozzle discharging via a specialized vapor feed distributor. The top of the distributor was only 9 in. beneath the bed. The successful solution relocated the grid bed higher in the tower.
    
12. CFD has been powerful for analyzing the effects of gas inlet geometry on gas maldistribution in packed beds. Ali et al. (5) showed that the gas velocity profiles from a commercial CFD package compared well with those measured in a 1.4-m simulator containing structured packing with commercial distributors and collectors. Their CFD model guided them to a collector design that minimized gas maldistribution. The application of CFD for evaluating vapor maldistribution is discussed at length in the next section with special reference to troubleshooting refinery vacuum tower feeds maldistribution.

4.12 VAPOR MALDISTRIBUTION IN THE FEED ZONE TO REFINERY VACUUM TOWERS

This is one application in which CFD technology has been extensively applied.

This tower is unique in that the feed to the tower, typically 30–70% vapor by weight, becomes >99.9% vapor by volume under the deep vacuum at the tower inlet. Physically, the feed enters at close to sonic velocity with the liquid dispersed in the vapor as a fine mist. This feed often enters via a tangential nozzle and a vapor horn device (Figure 4.18) that flings the feed against the tower wall. The wall catches the liquid and runs it down the wall, while the vapor expands to the tower diameter. An alternative arrangement uses a radial nozzle and a vane distributor (12, 245, 499). In either case, the vapor from the feed zone rises through a chimney tray (Figure 4.18, also Figure 7.2a) to a packed bed (termed “the wash bed”), which is the most critical bed in the tower. Economic considerations necessitate minimizing reflux to this bed. Maldistribution of vapor or liquid may render the distillate from the top of the bed (heavy vacuum gas oil, or HVGO) to be off-spec or form dry spots in the bed that can severely coke, limiting the capacity and run length of the tower, and often of the entire crude unit. These make vapor distribution to the bed critical.

Several studies were reported on how the vapor inlet geometry affects vapor distribution to the wash bed. Lin et al. (296) tested a 3-ft ID air–water scaled-down model of the flash zone of a large vacuum tower, transfer line, feed nozzles, and slop wax chimney tray (no vapor horn). They observed the vapor preferentially rising at the opposite side of the feed inlet (similar to Figure 4.17a), with some feed liquid deflecting into these chimneys. Vaidyanathan et al. (487) and Torres et al. (482) applied CFD to examine the effect of the geometry of a chimney tray above the inlet on vapor distribution and liquid entrainment. Paladino et al. (355) demonstrated that the presence of liquid in the feed affects the gas velocity profile and must be accounted for in modeling. Paladino et al. (355) and Waintraub et al. (497) used two-fluid models to explore the velocity distributions and entrainment generated by different designs of vapor horn distributor designs. Wherli et al. (499) and Kolmetz et al. (260) presented CFD-produced flow patterns for vacuum tower inlets, including radial bare nozzle, V-baffle, vane inlet device, and tangential entry with a vapor horn. Wehrli et al. (500) produced pilot-scale data simulating a vacuum tower inlet, which can be used for CFD model validation. Remesat and Riha (394) describe at length the application of CFD for vacuum tower feed inlet design.

For a refinery vacuum tower fed via a vapor horn, Figure 4.18 (75) shows CFD-generated horizontal (“x–z”) vapor velocity profiles at different elevations between the vapor horn and the wash bed. The arrows, slightly challenging to see, indicate the direction of flow. At plane 0, situated below the horn, Figure 4.18 shows a high swirl velocity of 70–80 m/s, directly beneath the region where the horn terminates. This swirl may generate a vortex at the bottom sump. If there are stripping trays in the smaller-diameter section, this swirl may upset the top tray or the chimney tray that directs the liquid to the trays. Plane 1, which slices the vapor horn, shows very high horizontal vapor velocities at the horn (as could be expected) and much lower velocities in the vapor rise region. Keep in mind that the velocities shown are in the horizontal plane. More important in this region are the vertical (“y”) velocities, which determine the potential for excessive entrainment and a premature system limit. It was recommended (394) to keep the peak C-factor (Eq. 2.1) based on the area open to vapor ascent below 1 ft/s.

Plane 2, right above the horn, shows the first transition from the feed zone profile to the flow profile entering the chimney tray and is useful for evaluating the effectiveness of the vapor horn as a vapor distributor. Plane 3 is just below the chimney tray. Comparing Planes 2 and 3 shows that the vapor velocities have significantly calmed down in the vertical space between the top of the horn and the chimney tray. This is well in line with the industry experience (155) that keeping the feed inlet nozzle top at least 5 ft below the chimney tray effectively retards entrainment from the flash zone. Finally, Plane 4, above the chimney tray and just below the wash bed, shows further slowing down of the horizontal vapor velocities. Unevenness on this plane is likely to persist in the wash bed. The reduction in velocities from Plane 3 to Plane 4 indicates that this chimney tray improves vapor distribution.

More important than the horizontal vapor velocities below and above the chimney tray are the vertical velocities. Figure 4.19a shows CFD-obtained vertical (“y”) velocity profiles at and above the chimney tray of a different vacuum tower (in this case, the feed enters via a vane distributor, not a vapor horn). The diagram on the right shows high-velocity peaks directly above the vapor outlets from the chimneys as well as around the periphery. These high-velocity peaks would persist into the wash bed and explain the reduced product quality, premature damage to the packing, and high slop wax production experienced (391). Similar peaks were reported earlier by Vaidyanathan et al. (487), who attribute them to vapor from two neighboring chimneys merging to form an upward jet that is dependent on the hat design. Torres et al. (482) also observed these peaks. The lower diagrams in Figure 4.19 show that these issues can be mitigated by an enlarged feed nozzle (68″ instead of 52″) and better vane distributor and chimney tray designs.

CFD analysis can be invaluable for interpreting field data. In one case (335), a high pressure drop was measured between the flash zone and the top of the wash bed, suggesting wash bed coking. A CFD analysis found that high pressure drop was due to the flash zone pressure measurement, which included a dynamic velocity head component, and not due to coking.

For the results of the CFD analysis to be valid, an accurate representation of the feed properties and velocity profile is necessary (394). A feed profile is created for the CFD analysis using a 3-dimensional mixture of turbulent compressible vapor flow, with liquid droplets moving with the vapor. It is equally important to correctly represent the geometry, and with an existing tower, to validate the CFD model with plant data (394, 506). There are several assumptions involved in CFD modeling, and if these deviate from reality, so will the results and conclusions.

4.13 TROUBLESHOOTING FLASHING FEEDS ENTRY

Liquid entry is typically designed for an inlet velocity of less than 4–5 ft/s. Figure 4.21 shows two nightmare scenarios of flashing feed entering liquid distributor parting boxes. In Figure 4.21a, water at 95°C entered a parting box of a bitumen stripper at a velocity of 3 ft/s. To improve performance, a consultant proposed reducing tower pressure from slightly above atmospheric to vacuum. Upon drawing vacuum, the feed flashed, and due to the low steam density, the feed velocity increased to 100 ft/s, which sprayed the feed liquid everywhere, precluding any decent distribution.

In Figure 4.21b, a flashing feed entered via a side nozzle into a V-notch trough that crossed the tower diameter. The design intention was for the flashing liquid to fall into the parting box and descend via the notches into the troughs of the distributor below, and for the vapor to disengage and rise to the bed above. Tests at the Raschig shop (420) (Raschig were not involved in the original design, only in the troubleshooting) showed that the high momentum of the incoming flashing feed pushed the liquid to the end of the parting box opposite the inlet nozzle, with some of it missing the parting box altogether (photograph in Figure 4.21b). The result was terrible distribution to the packing.

Other cases of flashing feeds entering liquid distributors causing hydraulic and separation problems were reported (71, 239, 240, 433). In one case in which a flashing feed entered a false downcomer, the downcomer was found repeatedly blown apart and lying on the distributor (271).

With flashing feeds, the vapor and liquid need to first be adequately separated, then the liquid descend onto a packing liquid distributor. The vapor must not enter the liquid distributor. There is a variety of flash box designs that achieve this, as described in several literature references (199, 245, 462). One of the most popular types is gallery distributors (Figure 4.22). The feed nozzle is located well above the upper plate. The gallery wall facing the inlet nozzle deflects the incoming vapor upward and the incoming liquid downward. The liquid thus separated from the feed joins the liquid raining from the peripheral zones of the bed above, and both descend through the holes in the upper plate. The lower plate is a liquid distributor to the bed below. Vapor from the risers of the distributor below rises through the hollow inside of the gallery.

In many galleries or other flashing feed distributors, liquid from the gallery descends to the distributor below using downpipes instead of floor perforations. To prevent overflow, these downpipes should be adequately sized for self-venting flow (Figure 4.9). The downpipes, either from the gallery or from a collector above, should enter the distributor below without restricting liquid equalization in the distributor, as occurred in Case 1 in Section 8.18.

The impact of the feed on the gallery wall tends to deflect the incoming jet upward. The top of the gallery in this region should be covered to prevent entrainment. Likewise, at the inlet, the feed splits into two streams, with the momentum pushing each stream horizontally along each wall of the gallery. Right across from the inlet, the remnants of the two streams collide. This region too can produce entrainment, which can be prevented by a gallery cover at that location.

When the feed enters too close to the gallery floor, it can entrain the gallery liquid, as described in the case study in Section 8.13. Reference (37) describes severe entrainment from a high-velocity flashing feed that entered via a 10″ inlet. The bottom of the inlet was 10.7″ above a shed deck that held some liquid.

When calculating liquid heads in the gallery using Eq. (4.2), account for liquid raining into the gallery from the bed above, in addition to the feed liquid. Also, there is likely to be some frothing of the liquid, and this needs to be considered in the evaluation but cannot be calculated.

Gallery distributors are prone to damage at the feed inlet point due to severe hydraulic pounding of the gallery wall by the entering feed. There have been cases where sudden increases in the quantity of entering vapor have flattened galleries. Such upset conditions should either be eliminated or catered for by mechanical strengthening.

It is important to ensure that packed tower feed pipes operate outside of the slug flow regime. Cases 6.8 and 6.9 in Ref. (201) describe experiences with packed tower instability, premature flooding, and poor separation resulting from slug flow in the feed.

4.14 TROUBLESHOOTING NOTCHED DISTRIBUTORS

V-notches, Y-notches (Figure 4.23), and rectangular slots are successfully used in high fouling rate services such as slurries, vent scrubbers, quench towers, and where other distributors experience plugging (40, 83, 192, 258, 260). Rectangular notches are also used in some high-quality two-stage distributors. Notched troughs are insensitive to plugging, corrosion, and erosion. They minimize stagnant zones, keep particulates moving, and allow the liquid to sweep the trough clean. The notches typically discharge the liquid directly into the up-flowing vapor.

The recommended equations for flow through notch is from Perry’s Chemical Engineers’ Handbook (5th ed., p. 5–17; some later editions are not clear on the units). For rectangular notches:

(4.7)

For narrow rectangular notches, Eq. (4.7) simplifies to

(4.8)

For triangular notches, not fully covered by liquid:

(4.9)

where

• q = volumetric flow rate, ft³/s
  
• L = width of rectangular notch (Figure 4.24a), ft
  
• h_o = height of liquid above bottom of notch (Figure 4.24a,b), ft
  
• g = gravity acceleration, 32.2 ft/s²
  
• Φ = angle in degrees, measured up from a horizontal line drawn at the apex (Figure 4.24b)

Per Eqs. (4.7)–(4.9), the flow rate rises to the power of 1.5 and 2.5, respectively, for rectangular and triangular notches, in contrast to flow rates rising to the power of 0.5 with orifice distributors (Eq. 4.2). This means a much lower range of liquid heads. A typical V-notch distributor will have a liquid head of 1–3 in. over its entire operating range (83). If at a head of 1 in. it handles 100 gpm, Eq. (4.9) gives that increasing the head to 3 in. will raise this flow rate to (100 × 3^2.5=) 1600 gpm. This is a huge dependence on liquid head. On the credit side, this gives good turndown. On the debit side, this makes the distribution sensitive to small changes in head and to fabrication irregularities from notch to notch, and makes it difficult to incorporate a large number of drip points, which inherently reduces irrigation quality. This strong dependence on the liquid head also makes V-notch (and to a lesser degree, rectangular notch) distributors far more sensitive to out-of-levelness, hydraulic gradients, and surface turbulence, and these could adversely affect their performance. The hydraulic gradient can be calculated using Eq. (4.5).

At low rates, the issuing liquid may wrap itself around the troughs (Figure 4.25a). At high rates, liquid streams from adjacent troughs may join each other into a single stream, thereby halving the number of distribution points (Figure 4.25b). Both of these were observed by Olsson (346) in distributor water tests. The solution was to add drip guides (Figure 4.25c) that direct the liquid to the desired points.

At high liquid rates, the flow down the length of the troughs is typically very high, and the flow out of each “V” is no longer uniform due to variance in the liquid momentum (Figure 4.26). Furthermore, at high liquid rates, the liquid may submerge the notches and maldistribute from then on.

Ensure that the liquid level does not reach the top of the notches. Once it does, the notches become triangular orifices, and the mode of operation totally changes, producing maldistribution and/or overflow.

4.15 TROUBLESHOOTING SPRAY NOZZLE DISTRIBUTORS

In spray distributors (Figure 4.27), a header pipe and laterals supply liquid to spray nozzles arranged in a pattern, typically elevated 18″ to 36″ above the packed bed. From the nozzles, liquid is sprayed as a wide-angle (typically 120°, but sometimes 90°) full cone. The cone is seldom homogenous. The bed coverage is determined by the number of nozzles, spray angle, and height above the packed bed.

It is a good practice to restrict the spray nozzle pressure drop to 10–15 psi, but up to 20 psi is used. This minimizes the entrainment and also minimizes the likelihood of a spray angle collapse. FRI’s experiments with wide-angle (120°) sprays (107) showed the spray cone angle to collapse from 120° to about 45° from the same spray nozzles when the spray nozzle pressure drop was increased between 10 and 20 psi with subcooled liquid. This spray angle collapse is closely related to the collapse of condensing vapor onto subcooled feed liquid (107).

Entrainment Depending on the nozzle pressure drop and vapor superficial velocity in the tower, spray distributors can generate a significant amount of entrainment (up to 25% or more). Pigford and Pyle (364), and later Lin et al. (296, 297) and Trompiz and Fair (484), measured entrainment in pilot-scale simulators using the air–water and air–Isopar systems. All these tests found entrainment to strongly increase with the gas velocity (C-factor, Eq. 2.1) and with the nozzle pressure drop. Some dependence on the nozzle type was observed. An excellent correlation for calculating the entrainment is provided by Trompiz and Fair (484).

Spray footprint diagram The spray circles overlap, so some areas at the top of the packing are overirrigated while others are underirrigated. The objective is to keep this maldistribution small-scale, but this objective is not always met. Spray patterns tend to be empirical. Generally, the distribution quality of spray distributors using established spray patterns is not as good as that of orifice distributors, and can be a lot worse when using nonstandard patterns.

A good way to check the irrigation quality is to prepare a spray footprint diagram. The diameter of the footprint of each spray circle on top of the bed is approximated by

(4.10)

where

• H = height of bottom of the spray nozzle above the bed (Figure 4.28a), inches
  
• D = diameter of circle drawn by the spray footprint on the bed, inches
  
• Φ = spray angle in degrees, measured up from a horizontal line drawn at the apex (Figure 4.28b)

The correction factor of 0.92 is an empirical factor accounting for gravity, suitable for 18″–30″ height between the bottom of the nozzles and the top of the packing.

The footprint of each spray nozzle is drawn as shown in Figure 4.28b. The pattern in Figure 4.28b is a common pattern that minimizes large-scale distribution with small circle overlaps. This pattern has worked successfully in many installations. There are many other patterns, some with much more extensive circle overlaps, that have also worked well. The key is to prepare a diagram like Figure 4.28b and to critically review it to minimize large-scale maldistribution. To optimize the pattern, the locations of some of the nozzles or the height above the bed can be adjusted on the diagram, and if warranted, a new header can be designed.

Spray pressure drop test This should be the very first test when dealing with a spray distributor. Make sure the flow to the sprays is correctly monitored. Check it against pump curves and control valve settings. Then install a simple, calibrated pressure gauge in the liquid line to the sprays, downstream of any control valve, filter, or heat exchanger. The measured spray nozzles pressure drop equals the pressure gauge reading, minus the tower pressure (inferred from the tower pressure transmitter reading), minus the liquid static head in the line to the sprays (calculated from the elevation difference), and minus the friction loss in the liquid line to the sprays (calculated from the liquid flow rate and line size, and is often negligible). Compare the measured spray nozzles pressure drop thus obtained to the pressure drop across the spray nozzles predicted from the spray vendor catalog at the measured flow rate. Do not forget to apply the density correction as prescribed in the spray vendor catalog, as the catalog flows are based on water.

Typically, pressure drop across the sprays at design flow rates is 10–20 psi. For a given flow rate, a measured sprays pressure drop significantly above that calculated from the catalog means plugged spray nozzles; a measured sprays pressure drop well below that calculated from the catalog means broken header or spray nozzles. Measuring the sprays pressure drops at several different liquid flow rates and comparing to the sprays vendor catalog is ideal and more informative. From the comparison of measured versus calculated pressure drop, the number of plugged nozzles can be estimated.

Several case studies where this technique diagnosed the root cause of poor performance of a wash bed in a refinery tower were described in Ref. (130). In one, the pressure drop was zero because a header flange disintegrated due to rigid bolting of the header to the tower wall with no allowance for thermal expansion. In another, the pressure drop was 0.2 psi due to disintegration of a thin-gauge flange. In a third case, a pressure drop of 94 psi was measured where 11 psi was expected, indicating severe plugging. In another case (443), oversized nozzles and header leaks due to missing gaskets led to a pressure drop of 1 psi. The author can add many more to the list, including one instance (209, 211) when the pressure drop was 150 psi; most of the nozzles were plugged solid.

Turndown Spray nozzles typically have a 2:1 turndown ratio. A high nozzle pressure drop produces fine spray droplets that can be entrained into the vapor. At very low pressure drops (typically <3 psi) the sprays droop, or even come out as “pencils” from the nozzles.

Fouling or plugging can be a concern with spray nozzles, but less so than for gravity distributors. The number of spray nozzles per square foot of bed cross-section area is much smaller than the number of drip points per square foot with a gravity distributor. This permits larger spray nozzle openings compared to gravity distributors, which reduces their plugging tendency. On the debit side, the tortuous path inside the spray nozzle, which generates the spray cone, produces a multitude of low-velocity spots where solids can be trapped and accumulate.

Filters, strainers, and/or screens should be used in the outside liquid pipes to the sprays to minimize the fouling per the guidelines in Section 4.5. Spray headers should be checked for the absence of, or at least the minimization of, stagnant pockets where solids can accumulate or liquid can overheat. Solids accumulating in such pockets often spall off and plug the spray nozzles. Lines from the filters to the tower should be fabricated out of corrosion-resistant metallurgy.

Flow tests It is recommended that all spray nozzle systems be flow tested inside and/or outside the towers with water (if possible) to check the coverage. Where liquid distribution is critical, such as for wash sections of refinery vacuum towers, the sprays should be tested at every turnaround. If one spray nozzle fails, a significant cross-section area of the column will not be wetted properly. Tests have uncovered plugged nozzles (e.g., Figure 4.29), leaks, liquid shooting right down, loose nozzles, corrosion or erosion damage, and debris. Prior to the tests, the lines leading to the spray headers need to be flushed without going through the nozzles (505) to prevent the introduction of debris. The quality of water supply needs to be monitored. Dirty water aggravates plugging, and high-chloride water causes stress corrosion of stainless steel.

Testing inside the tower is ideal but is sometimes disallowed due to various safety practices and concerns. In this case, the header should be removed and tested outside. Also, the testing may reveal a need for tack or seal welding, which may mandate removal and outside testing.

Chapter 23 in Lieberman’s book (287) describes his personal experience with spray testing inside one tower, which highlights some potential safety challenges. These include narrow access that slowed emergency exit, release of pockets of hydrocarbons and H₂S trapped behind plugged nozzles when the nozzles were loosened by hand, and excessive water pressure due to poor communication from inside the tower to outside. The high water pressure created noise (making communication even more difficult), mist that fogged safety glasses, and flushed out trapped hydrocarbons and H₂S into the tower.

When liquid distribution is critical, like in sprays to wash beds in refinery vacuum towers, the flow test should include obtaining a pressure drop versus flow curve. Control valve settings and valve openings should also be recorded. This test provides a useful baseline reference for future troubleshooting when plugging or breakage of the distributor is suspected.

Other than the pressure drop versus flow curve, a visual observation is usually sufficient in spray tests. Should a close evaluation of distribution quality be desired, area samples can be collected to study the flow variation. The visualization also gives insight into the droplet size range and whether collision of sprays from different nozzles creates dry spots or overirrigated spots underneath. Spray nozzle suppliers can often provide drop size distribution data determined from water tests. One watchout is that liquids that have lower surface tensions and densities than water produce smaller drop sizes.

Watch out for drain holes. These are seldom needed in spray distributors. With a high pressure drop across the spray nozzles, drain holes can generate powerful, undesired jets that damage packing.

For high-flow systems, such as pumparounds, a water test may require too much water to be practical. The distribution in pumparound sections is usually far less critical than in the wash section, so the test is often skipped. However, if practical, the test can still be performed by circulating water through the pumparound. Another way of testing nozzles for high-flow systems is by building a rig outside the column that can test one to three nozzles at a time. A less effective test is to run air through the spray header. It can identify nozzles that are plugged or shoot right through.

In some small towers, a single-spray or a few-spray distributor is used. Due to the non-homogeneity of the spray, these may give a very poor distribution. These should be carefully water-tested, and the nozzles should be replaced or adjusted in situ to improve the spray pattern as needed. The tests should be observed via a view port (if available) or an open manhole (505).

4.16 DISTRIBUTOR WATER TESTS

4.16.2 Gravity Distributor Water Tests In situ

Water testing of distributors in situ is invaluable, both for new distributors and those that have been in operation.

For new distributors, people often raise the question whether it is necessary to retest in situ after the distributor has already been tested in the shop. The merit of the in situ testing is to reveal issues that cropped up during installation and remained unnoticed. The piping in the plant differs from the setup in the supplier’s shop; the distributor supports in the plant differ from those in the shop; assembly errors could have occurred; debris and dirt could have lodged in some holes and plugged them; new leaks could have developed during assembly; feed pipe flanges may leak and damage could have occurred during transportation and mounting. Each of these can make a difference to the liquid distribution.

There are excellent examples of the merit of in situ testing. Bouck (43) found that at the minimum design liquid rate, the feed pipe cycled between being covered and uncovered by liquid in the distributor, changing the pressure drop across the control valve, which made it difficult to hold the setpoint flow rate. An in situ water test by Schnaibel et al. (417) provided an unexpected finding that liquid flow to some down boxes in a collector was obstructed by long chimneys, so these boxes received much less liquid than unobstructed down boxes. Another in situ test (264) identified maldistribution due to partial plugging of distributor holes by rust and debris as the root cause of poor performance. Diaz et al. (84) discovered a flange leak during an in situ water test, that contributed to the poor packing efficiency in a chemical vacuum batch distillation column. In a case the author is familiar with, a distributor that tested well in the shop showed maldistribution when tested in situ due to a sag in the middle. The sag did not occur in the supplier’s shop because it was better supported.

For distributors that have already been in service, the above holds, and in addition, residual dirt, corrosion of holes, damage to delicate parts, bending and sag, loosening of nuts, bolts, or clamps, damage to flanges and gaskets, and unexpected splashing could all adversely affect the distribution.

In situ tests with the packings in place afford the additional merit of showing the liquid distribution pattern as it comes out of the bed. Poor patterns may develop due to an anomaly or plugged or damaged section in the packed bed, or the development of excessive wall flow.

Although in situ water tests usually lack the sophisticated instrumentation and measurement equipment available for supplier shop tests, they usually provide the information needed for troubleshooting. An attempt should be made to run the test at the normal liquid flow rate and test turndown as necessary. The water for the distributor can be provided from the water mains via the distributor feed pipes, pumped from a specially installed water tank, or simply provided by a water hose. Good solid-free and noncorrosive water quality is mandatory for the test, and in the case of stainless steel, also low in chlorides. The quantity of water needed for the tests should be estimated ahead of time, and special pumps should be ordered as needed. The setup used must be approved by the health, safety and environmental personnel. Water exiting the bed can be sampled by a cup that is tied to a rod and is moved around to different points, with its filling time at each point recorded.

It is useful to watch the water emptying out of the distributor after the water supply is stopped. The last streams of water should come out randomly or reasonably evenly. If the last ¼–½ in. of water come out from one area, a levelness issue is implicated.

Figure 4.30c shows an in situ water test that explained a very poor water-ethylene glycol separation (346, 347). All the distributor liquid leaked at the tray floor through ¾ in. weep holes and through gaps where the weir risers fitted through the deck. Figure 4.30d shows a baffle deformed in service, deflecting the water issuing from distributor holes in an undesired direction. Figure 4.30e shows a 12-ft long parting box receiving downcomer reflux liquid from the trays above as well as fresh liquid feed. A water test showed a hydraulic gradient in the box with preferential flow and overflow at the downcomer end of the box. Section 9.2.8 describes how an in situ water test showed maldistribution that gave poor separation for decades and was not picked by gamma scans.

A caustic absorption tower with four circulating pumparound beds was water-tested in situ by circulating demineralized water and observing the action through the manholes (77). The tests revealed plugged perforations and a need for mechanical cleaning in one of the distributors. In addition, they revealed overflow at 75% of the design liquid flow rates, which explained the carryover they were experiencing during operation and forced operation at reduced circulation. In another caustic absorption tower, a water test on the collector from a pumparound circuit found no leaks, but circulating water through the circuit with the manhole open found that some of the collector hats overflowed into the risers.

In small towers, it may be possible to test the vapor distribution. An air blower can be connected to the vapor entry nozzle and operated at normal flow. If safe and practical, it may be possible to measure the flow distribution in the tower using an anemometer. Another test advocated by Willard (505) is to run water for several minutes until the packings are fully wetted, then turning the water off while allowing the air to flow. Packings moist at the center and uniformly wet at the walls indicate good vapor distribution.