Back to Basics
Control Column Pressure via Hot-Vapor Hot-Vapor Bypass Henry Z. Kister Fluor Corp.
Daryl W. Hanson Valero Energy
Hot-vapor bypass employs a flooded condenser located at ground level to control distillation tower pressure. It offers capital cost savings, but can be tricky to implement. This article offers guidance on designing and troubleshooting a hot-vapor bypass system.
P
ressure is the most important variable for controlling distillation columns, because pressure affects every aspect of a distillation system: vaporization, condensation, temperatures, volatility, etc. An unsteady pressure typically results in an unsteady column. There are several ways to control tower pressure, depending on how the tower is congured. If a tower has an overhead vapor product, manipulating the vapor owow rate usually controls pressure. If the tower has no vapor product (i.e., (i.e., it has a total condenser and liquid product only), tower pressure can be controlled by partially ood ing the condenser and manipulating the liquid level in the condenser. Another alternative for either vapor or liquid products is to manipulate manipulate the coolant coolant owrate owrate (or temperatemperature) to control the tower pressure. Coolant manipulation manipulation is popular in refrigerated towers but is usually avoided in cooling-water cooling-water condensers, as it can cause accelerated fouling and corrosion. Among the ooded-condenser control methods, the hot-vapor bypass employing condensers mounted at ground level is one of the most popular for large cooling-water condensers. The popularity of the hot-vapor bypass scheme with large total condensers stems from its major capital savings. Locating large condensers at ground level eliminates the need for massive condenser support structures and for piping piping cooling cooling water water to high high elevati elevations, ons, and and provides provides easy access for maintenance. The piping is simple, the control valve is small, and the response is fast. These advantages can translate into signicant savings in steelwork, platforms, and maintenance. Capital savings can be signicant in large installations, especially where a battery of condensers rather than a single exchanger is used. Like other oodedCopyright © 2015 American Institute of Chemical Engineers (AIChE)
condenser schemes, this arrangement delivers subcooled liquid for the reux and product pumps, maximizing their available net positive suction head (NPSH) and lowering platform platform heigh heightt requirem requirements. ents. A survey of distillation tower failures (1, 2) identied 2) identied the hot-vapor bypass as the most troublesome pressure and condenser control method. About one-third of the pressure and condenser control malfunctions reported in the literature were problems with hot-vapor bypass schemes, many in reneries. Most problems were due to poor conguracongura tion of hot-vapor bypass piping, which evolves from poor understanding of its principles — principles that have been discussed in the literature for decades (3–7). (3–7). However, even if a hot-vapor bypass scheme is congcongured correctly, correctly, you may experience problems with tower pressure pressure control control.. This This article article descri describes bes our our experien experiences, ces, nd nd-ings, and lessons learned that may be valuable for troubleshooting and design of hot-vapor bypass schemes.
Flooded-condenser Flooded-conden ser control schemes The main ooded-condenser control methods (3, 6, 7) manipulate the condenser ooding to control column pres sure using one of the following congurations: • a condenser elevated above the reux drum with a control valve in the condensate line or in the vapor line to the condenser • a ooded reux drum with a control valve at the drum condensate outlet (or no drum) • a condenser located at ground level and a valve in the bypass bypass from the overhe overhead ad vapor vapor line line to the drum drum vapor vapor space (hot-vapor bypass). In all of these methods, the condenser area is partially
CEP
February 2015
www.aiche.org/cep 35
Back to Basics
ooded by condensate. The ooded tubes do not contact the vapor and perform little condensation. Column pressure is controlled by manipulation of the ooded area. Raising the liquid level in the condenser oods additional tubes, which reduces the condensation area, thereby raising tower pressure. Conversely, lowering the liquid level in the condenser exposes more tubes, which increases condensation area, and subsequently lowers tower pressure. Although the ooded area performs little condensation, it serves the vital purpose of subcooling the condensate before it leaves the exchanger. Hot-vapor bypass systems. Figure 1 shows a correctly congured hot-vapor bypass for column pressure control. Unlike other ooded-condenser control schemes (in which the condenser is elevated above the reux drum), in the hot-vapor bypass scheme, the condenser is at ground level and the drum is elevated, often mounted on the lowest platform. The liquid level in the condenser is 10–20 ft below that in the drum (which may be horizontal or vertical). The condensate is lifted by the difference between the vapor pressure at the condenser liquid surface, which is at the bubble point temperature, and the vapor pressure at the drum liquid surface, which is colder due to the condensate subcooling. Even a few degrees of subcooling can raise the condensate 50 ft to 100 ft, or more. To limit the liquid lift to the desired height (i.e., to the reux drum), hot vapor from the tower overhead is used to heat the liquid surface in the drum. The vapor pressure in the drum is set by the liquid surface temperature, not the subcooled temperature. The surface temperature, and therefore the vapor pressure in the drum, are determined by the heat balance between (1) the hot vapor inow and (2) the heat outows from the surface to the bulk Detail of Liquid Surface dPbypass PC P1
Hot-Vapor Bypass Valve P2
Tower H
Condenser
Hot Vapor Bubble Point Liquid Subcooled Liquid
Reflux Drum
LC
Signal to Distillate-Reflux Controller
Product plus Reflux
a typical hot-vapor bypass, tower pressure is controlled by manipulation of the liquid level in a partially flooded condenser located at ground level. The valve in the vapor bypass is opened to raise the liquid level in the condenser. Raising the condenser liquid level floods additional heat exchanger tubes, which reduces the condensation area and raises tower pressure. Closing the valve lowers the condenser’s liquid level, increases condensation area, and lowers tower pressure. p Figure 1. In
36
www.aiche.org/cep
February 2015 CEP
liquid and from the vapor space to the atmosphere. The liquid lift is manipulated to control the tower pressure. Opening the hot-vapor bypass valve heats the drum liquid surface, raising the liquid’s vapor pressure and therefore the drum pressure. This pushes liquid from the drum into the condenser, ooding more tubes, reducing condensation, and raising the tower pressure. Conversely, closing the valve cools the drum liquid surface, reducing its vapor pressure and therefore the drum pressure, and drawing liquid from the condenser into the drum. This exposes additional condensing area and lowers tower pressure.
Configuring a hot-vapor bypass correctly Hot-vapor bypass schemes can be troublesome, mostly due to poor conguration of the hot-vapor bypass piping. The following techniques can help avoid some of the common pitfalls. • Congure the piping correctly (2, 4–8). Bypass vapor must enter the vapor space of the reux drum (Figure 1). The bypass should be free of pockets where liquid can accumulate; any horizontal runs should drain into the reux drum. Liquid from the condenser must enter the reux drum well below the liquid surface, near the bottom of the drum. If the entry nozzle is at the top of the drum, the liquid line needs to be extended so that it discharges near the bottom of the drum. Any other liquid streams entering the drum, such as the reux pump’s minimum ow recycle, must also enter near the bottom of the drum. Figure 2 depicts one case (9) in which violation of this practice led to severe pressure uctuations, inability to control tower pressure, and a capacity bottleneck. In the original scheme (Figure 2, top), subcooled liquid mixed with vapor at its dewpoint, and vapor collapse occurred at the site of the mixing. The rate of vapor collapse varied with changes in subcooling, overhead temperature, and condensation rate. Variation of this collapse rate induced pressure uctuations and control valve hammering. The red piping in the bottom conguration shows the x that eliminated the problem. The liquid and vapor lines were separated, and the vapor line was modied so that it introduced the vapor into the top of the reux drum. The liquid line was extended to discharge below the drum liquid level. After these changes were made, the tower pressure no longer uctuated, and the problem was completely solved. References 2, 5, 7, 8, 10, and 11 report other experiences in which incorrect piping led to instability, poor control, and hammering. • Do not agitate the drum liquid surface. Operation may be troublesome if the drum liquid surface is agitated (5, 6). High-velocity impingement of the hot-vapor jet on the liquid surface or currents introduced by improper design of the liquid inlet can produce agitation. Agitation may also occur if
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
vapor condenses on the ceiling of the drum and condensate drips down onto the drum liquid surface. • Vent any noncondensable gases. Hot-vapor bypass controls are suitable only for total condensers. The liquid leg between the condenser and drum prohibits venting of any noncondensables. To handle small amounts of noncondensables, such as those trapped in the condenser during startups or upstream upsets, a vent (not shown on Figure 1) is required at their accumulation point(s). The condenser vent can be directed to the vapor space of the drum or elsewhere. If a vent line is absent, instability and capacity bottlenecks may result (12). • Insulate the reux drum. Insulating the reux drum vapor space tends to minimize temperature swings due to rain and snow (5–7). This issue is more pronounced for narrow-boiling-range mixtures (5, 6), and at high pressure (5), where small temperature changes have a large effect on the split of overhead ow between the condenser and the bypass. With wide-boiling mixtures, Rayleigh fractionation (preferential condensation of heavy components without mixing with the remaining mixture) can also interfere with PC
LC
Tower
Four Condensers
FC
PC
LC
Tower
FC
p Figure 2. In this hot-vapor bypass scheme, the incorrect piping
configuration at the top experienced pressure fluctuations. Modifying the piping, as shown on the bottom in red, ensures good pressure control. (Based on Ref. 9.)
this control system (4). On the other hand, insulating the drum often raises concerns of water-trapping and corrosion underneath the insulation. • Tune the pressure controller tighter than the drum level controller. Because of the liquid leg between the condenser and the drum, hot-vapor bypass schemes can suffer from interactions between the drum and the condenser liquid levels (7, 13–15) and from U-tube oscillations (7, 14, 15). To minimize such interactions, the pressure controller should be tuned much tighter than the drum level controller (13, 15). This can be an issue if the reux drum is small and the level controller needs to be tuned fast to avoid overow or loss of level. This is uncommon; we have encountered the situation only once, although it was reported in one other case (16). • Ensure that the bypass valve does not leak. Leakage of vapor through the bypass valve when it is closed can substantially reduce condenser capacity (17). • Size the bypass valve correctly. An undersized bypass control valve may not be able to maintain the tower pressure high enough on cold winter days. The effects of undersizing the valve are most severe when the drum is not insulated. In one case (18), poor pressure control due to undersizing was improved by installing a throttling valve in the liquid line from the condenser to the drum. In another case (19), pressure uctuations and instability due to excessive subcooling at low rates, during plant startups, and on cold nights was countered by controlling the cooling water owrates to prevent excessive subcooling. The amount of subcooling and the vapor bypass rates can only be determined empirically, and sizing the valve is difcult. Simplied sizing procedures are available (5, 20), but they are based on heating all the subcooled liquid to its bubble point and are, therefore, grossly conservative. An interesting approach proposed recently (19) models the drum surface as a heat exchanger, but the method is based on variation of surface temperature along the drum length, which has not been observed in well-functioning drums. • Watch out for a possible inverse response. When pressure rises, it closes the bypass valve, which initially increases pressure until the condenser level begins to change (21). • Control tower pressure, not drum pressure. In some cases, the hot-vapor bypass control valve is manipulated by the drum pressure instead of the tower pressure (18). This type of control is dynamically inferior, because the volume of vapor in the drum is much smaller than the vapor volume in the tower, and the drum vapor is more variable in response to ambient changes. The discussion that follows is based on a well-congured, well-designed, hot-vapor bypass scheme that does not violate these piping conguration guidelines. Nonetheless, some of these guidelines will be revisited in the following sections. Article continues on next page
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
CEP
February 2015
www.aiche.org/cep 37
Back to Basics
Hydraulics of hot-vapor bypass schemes For the correctly congured hot-vapor bypass arrangement in Figure 1, assuming negligible line pressure losses, the pressure balance is:
In typical hot-vapor bypass schemes, the liquid level in the reux drum is about 10–20 ft above the liquid level in the condenser. So subcooling of less than 10 F is sufcient in most practical applications. Since the actual subcooling is typically 10–50 F, the vapor pressure difference will tend to pump the condensate to a much higher elevation than needed. The hot vapor provided by the bypass regulates the vapor pressure in the drum by maintaining the liquid surface tem perature at a value that produces the desired liquid head, H . The hot vapor condenses onto the drum liquid surface and heats it up. At steady state, opening the valve adds sufcient vapor to maintain the drum liquid surface temperature at the value corresponding to the desired vapor pressure P 2 that satises Eq. 1. There is a net heat ow from the hot liquid surface to the subcooled liquid below the surface. This heat ow and the atmospheric heat losses need to be matched by condensing the hot vapor. The pressure difference in the bypass can be written in terms of the pressure difference across the bypass control valve, Δ P bypass: °
°
∆
where P 1 is the pressure at the intersection of the line to the condenser and the condenser bypass (psia), P 2 is the pressure inside the reux drum vapor space (psia), H is the head difference between the reux drum liquid level and the condenser liquid level (psi), and Δ P cond is the condenser pressure drop (psi). Because the condenser inlet line contains a static leg of vapor, the head differential H is calculated using the difference between the liquid and vapor densities. The vapor density is based on P 1 and the condenser inlet temperature. The liquid density is best approximated as the density of the subcooled liquid leaving the condenser. According to Eq. 1, the driving head required to pump the condensate into the drum is supplied by the pressure difference between the condenser liquid surface and the drum liquid surface. P 1 is the vapor pressure at the liquid surface in the condenser, which is the bubble point pressure at the condensing temperature. P 2 is the vapor pressure at the reux drum surface. For P 1 to be higher than P 2, the temperature at the drum surface needs to be lower than the temperature at the condenser liquid surface. It does not take much subcooling to signicantly lower the vapor pressure in the drum. Table 1, derived from physical property data for hydrocarbons (22), lists vapor pressure and liquid head variations with temperature for three common applications. Table 1 shows that for a C3-C4 splitter, subcooling of 1 F can lift the liquid 15 ft. A debutanizer needs just over 2 F of subcooling to lift the liquid the same distance. A dehexanizer requires a bit more subcooling (10 F), but even that is well within the subcooling capability of the condenser. A typical subcooling range for a condenser is 10 F to 50 F.
∆
Combining Eq. 1 and Eq. 2 gives: ∆
∆
a
When the line friction losses are signicant, Eq. 3a needs to include them: ∆
∆ ∆
b
-
∆
-
∆
-
where Δ P bypass-line is the pressure drop in the hot-vapor bypass vapor line excluding the pressure drop across the control valve (psi), Δ P liq-line is the pressure drop in the liquid line from the condenser to the reux drum (psi), and Δ P vap-line is the pressure drop in the tower overhead vapor line to the condenser downstream of the point where the hotvapor bypass splits off (psi). For clarity, the following Table 1. Small changes in temperature can have large effects on vapor pressure and discussion is based on Eq. 3a. liquid head in hot-vapor bypass appli cations. Source: (22). However, all of the following Vapor Pressure Liquid Liquid Head equations can be expanded to Application Pressure Temperature Change Rate Density Change Rate include the line pressure drop C3-C4 Splitter 305 psia 126°F 3.2 psi/°F 28.3 lb/ft3 15 ft/°F terms in Eq. 3b. (72%/28% 218 psia 99°F 30.2 lb/ft3 In hot-vapor bypass schemes, Liquid Volume the vapor pressure differPropylene/Propane) ences directly manipulate the Debutanizer 150 psia 175°F 1.5 psi/°F 31.2 lb/ft3 7 ft/°F ooded height in the condenser. (Butane) 76 psia 125°F 33.7 lb/ft3 Equation 1 states that the vapor Dehexanizer 40 psia 220°F 0.4 psi/°F 36.0 lb/ft3 1.6 ft/°F pressure difference is balanced (Hexane) 15 psia 158°F 38.3 lb/ft3 by the liquid head lift plus the °
°
°
°
38
www.aiche.org/cep
February 2015 CEP
°
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
condenser pressure drop. To raise the pressure in the tower, additional tubes in the condenser need to be ooded, so the liquid head lift H needs to be reduced. To achieve this, the vapor pressure differential P 1 – P 2 also needs to be reduced (per Eq. 1). Because P 1 is constant, this is achieved by opening the hot-vapor bypass valve to heat the liquid surface in the drum, which raises the drum vapor pressure P 2. Conversely, to reduce the tower pressure, the liquid level in the condenser is lowered, so the liquid head lift H is raised. This is achieved by closing the bypass valve. The reduction in hot-vapor ow allows the liquid surface in the drum to cool, lowering P 2. The larger P 1 – P 2 difference pulls liquid from the condenser into the drum, exposing more condenser area for condensation.
Hydraulic imbalances According to Eq. 3a, when the liquid head and the condenser pressure drop are small, the bypass pressure drop will also be small. Under these conditions, the hot-vapor-bypass control system becomes vulnerable to hydraulic imbalances. To the best of our knowledge, this vulnerability has not been previously reported. At low heads and small pressure drops, the valve tends to open. If the pressure drop across the bypass valve exceeds the right-hand side of Eq. 3a even at 80–90% open, the valve will tend to open widely and cause a loss of control. To match the large pressure drop on the left-hand side of Eq. 3a, the liquid head H will tend to rise. In some situations it will keep rising until it sucks all of the liquid out of the condenser. Imbalances can also develop because different factors govern the valve pressure drop and the valve opening. The valve pressure drop is largely governed by the drum liquid surface temperature, whereas the valve opening is governed by the tower pressure controller. While normally the two governing mechanisms vary in unison, a strong disturbance that causes sudden cooling of the liquid surface in the drum may throw the mechanisms out of balance. The drum pressure falls quickly, while the tower pressure, and therefore the valve opening, remains temporarily constant. The bypass pressure drop increase renders the left-hand side of Eq. 3a larger than the right-hand side. The higher pressure differential increases the vapor ow through the bypass. Also, per Eq. 3a, H increases, exposing condenser tube area and causing the tower pressure to fall. The reduction in tower pressure opens the valve, further increasing the hot-vapor ow. If the combined increase in hot-vapor ow is sufciently large to reheat the liquid surface and quickly reinstate the drum pressure, the drum will return to steady state following the bump. In contrast, if the increase in ow does not balance the decrease in drum pressure fast enough, or if it interacts in a manner that aggravates the initial surface disturbance, H in
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
Eq. 3a will keep rising, in some situations until it draws all of the liquid out of the condenser. (This is illustrated by the case study discussed later in the article.) A good solution for both imbalance issues is to increase the pressure drop at the outlet of the condenser, such as by adding a throttling valve as illustrated in Figure 3. The addition of this valve requires the addition of the throttling valve pressure drop, Δ P out , to Eq. 1: ∆
∆
Combining Eq. 4 with Eq. 2 (which remains unchanged) gives: ∆
∆
∆
a
Equation 5a can be expanded to include the line friction terms in Eq. 3b: ∆
∆
b
-
∆
∆
∆
-
-
∆
The throttling valve permits a larger pressure drop through the bypass valve. The additional pressure drop retards the sudden movement of the mass of liquid from the condenser to the drum upon a strong pull from a disturbance at the drum liquid surface. If the control valve is close to fully open, the additional pressure drop helps to return the valve to its normal operating range. Finally, the additional pressure drop counters the tendency for U-tube oscillations mentioned earlier. A hydraulic imbalance can develop when the tower’s overhead is condensed by an elevated air condenser followed by a ground-level cooling-water condenser (Figure 4). This system generally works well when the liquid remains in the cooling-water condenser. However, during cold or wet ambient conditions, at low plant throughput rates, or when the condensers are clean (not fouled), the air condenser may easily provide all the needed condensation area. The presdPbypass
PC P1
P2
Tower
LC
Signal to Distillate-Reflux Controller
H
dPcond
dPout
Product plus Reflux
addition of a throttling valve at the condenser outlet reduces the bypass valve’s vulnerability to hydraulic imbalances. p Figure 3. The
CEP
February 2015
www.aiche.org/cep 39
Back to Basics
sure controller will open the bypass, pushing liquid up into the air condenser. When this occurs, Eq. 1 becomes: ∆
∆
and Eq. 3a becomes: ∆
∆
∆
where Δ P cond,air is the pressure drop across the air condenser (psi), Δ P cond,water is the pressure drop across the water condenser (psi), and h is the liquid head between the air condenser and the reux drum (psi). For this arrangement to work, the sum of the condensers’ pressure drops needs to be well above the liquid head h. This is unlikely, because the feed to the water condenser is all liquid, so the condenser pressure drop is low. The air condenser pressure drop will most likely not be high enough to keep Δ P bypass in Eq. 7 within the valve control range, and the bypass valve will open fully. However, even full opening of the valve will not provide enough heat to keep the liquid level in the drum low enough. The drum will ll up and control will be lost. The simplest solution is to ensure that the liquid always remains in the water condenser by reducing the air condenser duty during winter. This is achieved by turning off fans, closing louvers, and controlling the air condenser’s outlet temperature by manipulating the fan motor speed or the pitch of the blades. Alternatively, although not commonly practiced in this scenario, additional pressure drop (e.g., a throttling valve) can be added in the condensate line from the cooling-water condenser outlet or in the condenser inlet line.
Steady-state heat transfer At steady state, the heat supplied by the hot-vapor bypass balances the heat ow from the drum liquid surface to the drum bulk liquid plus any heat losses from the vapor space of the drum:
temperature corresponding to P 2. This hot temperature does not extend deep below the liquid surface (see the enlarged detail in Figure 1). The vapor lm heat-transfer coefcient for condensation at the liquid surface is high (19). Since most process liquids are good thermal insulators, most of the temperature gradients occur within a thin hot-liquid layer, which can be as thin as 1 in. (4). Experience with thermal scans conrms that the hot layer is only a few inches thick in a well-designed system. Below this hot surface, the liquid remains at its subcooled temperature. The heat transferred from the surface to the bulk liquid by conduction and convection through quiescent liquid, Qcond , is:
where hQHL is the heat-transfer coefcient for conduction and convection through the hot-liquid layer (Btu/hr- F per ft2 of drum liquid surface area), AS is the area of the liquid surface in the reux drum (ft2), and T is the temperature ( F) with the subscript surface denoting the drum liquid surface and subcool denoting the subcooled bulk liquid. The heat supplied by the hot-vapor bypass also needs to offset the heat lost from the wall of the drum above the liquid surface to the atmosphere, Qloss. Assuming the drum liquid surface is at about the same temperature as the drum vapor space (usually a good assumption), the loss of heat from the wall of the drum is: °
°
where hatm is the heat-transfer coefcient for convection and radiation from the exposed drum wall area above the liquid to the atmosphere (Btu/hr-°F per ft 2 of drum wall area), A DW is the area of the drum wall above the liquid surface (ft 2), and the subscript ambient denotes the ambient temperature outside the drum. PC
where Q HVB is the total heat supplied by the condensation of vapor supplied by the hot-vapor bypass (Btu/hr), Qcond is the heat transferred from the drum liquid surface to the drum bulk (subcooled) liquid (Btu/hr), and Qloss is the heat loss from the exposed drum wall area above the liquid surface to the atmosphere (Btu/hr). As long as the liquid surface in the reux drum is smooth and unagitated, most of the heat ow from the surface into the subcooled liquid, Qcond , is by conduction. In reality, some convection and bulk movement also occur and provide addi tional heat transfer from the surface to the subcooled liquid. All of the heating, cooling, and vapor pressure adjustment processes take place at the liquid surface. The vapor space and the liquid surface are hot — at the condensing 40
www.aiche.org/cep
February 2015 CEP
dPbypass
P1
dPcond,air h
LC
Signal to Distillate-Reflux Controller
dPcond,water Product plus Reflux
naphtha splitter tower had an air condenser in series with a cooling-water condenser. On cold winter days, the liquid level rose into the air condenser and it became difficult to maintain a stable pressure in the tower. This issue was solved by turning off fans and shutting louvers in the air condenser and raising the setpoint on the air cooler’s outlet temperature control. p Figure 4. A
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
Combining Eq. 9 with Eq. 10 allows for comparison of the two mechanisms:
Due to the drum’s curvature, the wall area above the liquid surface A DW in a horizontal drum exceeds the liquid surface area AS . In a vertical drum, the length usually is about three times the diameter, so unless the liquid level is near the top, A DW again exceeds the liquid surface area. In either case, the area ratio term AS /A DW in Eq. 11 is less than 1. In most situations, the ambient temperature, especially during a cold winter night, can be much lower than the subcooled tem perature, so the temperature difference ratio term in Eq. 11 is less than 1, often by a large factor. At the same heat-transfer coefcient, the atmospheric heat losses usually exceed the heat ow from the liquid surface to the drum liquid, often quite signicantly. For uninsulated vessels, the ambient-loss heat-transfer coefcient, hatm, varies widely with the ambient conditions. Over the temperature range of 100–200°F, heat-transfer coefcients for ambient heat loss typically range from about 2–3 Btu/hr-°F-ft2 in still air to 5–7 Btu/hr-°F-ft 2 on a windy day (23, 24). On a rainy day, the coefcient may be as high as 15–20 Btu/hr-°F-ft2. With still liquid at the drum surface, the heat-transfer coefcient for conduction plus convection at the hot-liquid layer in the drum is on the order of 2–10 Btu/hr-°F-ft2. It follows that for uninsulated drums, the ambient heat losses exceed the heat ow from the surface to the bulk liquid even under favorable ambient conditions. On a stormy day, the ambient heat losses can be more than an order of magnitude higher than the heat ow from the surface to the bulk liquid. Insulating the drum can reduce the ambient heat losses by an order of magnitude compared to the losses in still air. For well-insulated process vessels (with about 3–4 in. of berglass or mineral wool insulation and no large exposed metal surfaces), the ambient-loss heat-transfer coefcient is 0.25–0.3 Btu/hr-°F-ft2 (23). However, many plants prefer not to insulate the drum for fear of water-trapping and corrosion underneath the insulation. There are four implications of ambient heat losses: • For uninsulated vessels, ambient changes, especially sudden rain or thunderstorms, can generate instability in hot-vapor bypass schemes. This problem is most severe in winter and in cold climates. • Bypass sizing should be based on the total heat lost on a cold and rainy winter night. Use Eqs. 9 and 10 with the appropriate heat-transfer coefcients based on the coldest conceivable ambient and subcooled temperatures to avoid gross oversizing of hot-vapor bypass valves. For conser-
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
vative design, the value calculated can be multiplied by a safety factor of two or even three. • Generously sized bypass valves render hot-vapor bypass systems with uninsulated drums more robust to ambient disturbances. The vapor-lm heat-transfer coef cient for condensation at the liquid surface is high (19), so as long as the liquid surface remains undisturbed and the valve is not close to fully open or fully closed, the additional heat supplied upon valve opening can quickly catch up with
Nomenclature A DW AS h hatm
hQHL
= area of the drum wall above the liquid surface (ft2) = area of the liquid surface in the reux drum (ft2) = liquid head between the air condenser and the reux drum (psi) = heat-transfer coefcient for convection and radiation from the exposed drum wall area above the liquid to the atmosphere (Btu/hr-°F per ft2 of drum wall area) = heat-transfer coefcient for conduction and convection through the hot-liquid layer (Btu/hr- F per ft2 of drum liquid surface area) = head difference between the reux drum liquid level and the condenser liquid level (psi) = pressure at the intersection of the line to the condenser and the condenser bypass (psia) = pressure inside the reux drum vapor space (psia) = pressure difference across the bypass control valve (psi) = pressure drop in the hot-vapor bypass vapor line excluding the pressure drop across the control valve (psi) = condenser pressure drop (psi) = pressure drop across the air condenser (psi) = pressure drop across the water condenser (psi) = pressure drop in the liquid line from the condenser to the reux drum (psi) = throttling valve pressure drop (psi) = pressure drop in the tower overhead vapor line to the condenser downstream of the point where the hot-vapor bypass splits off (psi) = heat transferred from the drum liquid surface to the drum bulk (subcooled) liquid (Btu/hr) = total heat supplied by the condensation of vapor supplied by the hot-vapor bypass (Btu/hr) = heat loss from the exposed drum wall area above the liquid surface to the atmosphere (Btu/hr) = ambient temperature outside the drum ( F) = subcooled bulk liquid temperature ( F) = drum liquid surface temperature ( F) °
H P 1 P 2 Δ P bypass Δ P bypass-line Δ P cond Δ P cond,air Δ P cond,water Δ P liq-line Δ P out Δ P vap-line Qcond Q HVB Qloss T ambient T subcool T surface
°
°
°
CEP
February 2015
www.aiche.org/cep 41
Back to Basics
the additional cooling generated by an ambient disturbance. (This is illustrated in the case study discussed later.) • For insulated vessels, a much smaller bypass valve is required.
Loss of condenser heat-transfer capacity References 17 and 25 report cases in which leakage through a hot-vapor bypass, even when the valve was fully shut, signicantly reduced the condenser’s heat-transfer capacity. In one instance (17), Lieberman reported that blocking-in the bypass increased condenser capacity by 50%, eliminating the need for a larger condenser. As stated earlier, the heat supplied to an uninsulated drum, and therefore the size of the control valve, is primarily dictated by the ambient heat losses during periods of cold, rainy, or snowy weather. Most control valves sized for such conditions tend to be oversized for periods of calm and warm weather, when the heat losses may be an order of 100 F 80 ° , e % r u e v t a l a 60 r V e p d m n e a 40 T l t e e v l e t L u 20 O
0 14 d p b M , e t a r w i o s l p F , r O e C u L s s d e r n P a , 3 C , x u l f e R
295
12 10
g i 290 s p , e r u s s e r P 285 p o T
8 6 4 2
280
0 9:28
10:04
10:33
11:02
11:31
Time, pm
Condenser Outlet Temp, °F
Reflux Flowrate, Mbpd
Hot-Vapor-Bypass Valve Opening, %
Light Cycle Oil Flowrate, Mbpd
Drum Level, %
Tower Top Pressure, psig Hot-Vapor-Bypass dP, psi C3 Production, Mbpd
the unstable event depicted in this operating chart, the differential pressure across the hot-vapor bypass valve jumps and the liquid level in the reflux drum increases from 40% to over 80%. p Figure 5. During
42
www.aiche.org/cep
February 2015 CEP
magnitude lower. During these warm and calm periods, the control valve may operate in a nearly closed position, mak ing it prone to leakage and unstable control. Drum insulation helps eliminate the need for control valve oversizing. The leaking valve that limited condenser capacity in Lieberman’s case was a buttery valve (17). This type of valve is prone to leak, especially when the valve is over sized. Lieberman highlights the importance of designing the bypass to prevent excessive leakage. The valves used for hot-vapor bypasses should have a tight-shut-off design.
Case study: Condenser outlet throttling mitigates instability The problem. In a C3-C4 splitter column, the hot-vapor bypass was correctly congured (as in Figure 1). The condenser liquid level was 10–12 ft below the liquid level in the drum, and the drum was not insulated. The control usually worked well when the tower overhead owrate (reux plus product) was less than 25,000 bbl/day. At higher owrates, the drum would often suddenly ll up and the drum level controller would increase the reux rate. The drum level would then dive, the reux rate plummet, and the reux pump cavitate. The only way to restore stability was to cut back overhead owrate. This restricted the column capacity. The problem was more severe in summer. A similar instability occasionally occurred at lower overhead owrates. Figure 5 shows the operating charts for such an event. During this event, the reux-plus-product owrate was 17,000 bbl/day. At about 10:40 pm, the hot-vapor bypass differential pressure (green triangles) jumped from the normal 4.5 psi to 7 psi in about 1 min, and to 12 psi within 5 min. Over the same 5 min, the tower top pressure declined only slightly, from 289.5 psig to 288.5 psig, increasing the valve opening from 50% to 80% (red squares), while the drum liquid level jumped from 40% to 80% (light blue diamonds). Over the next 15 min, the hot-vapor-bypass differential pressure, valve opening, and drum liquid level returned to their pre-event values, while the column pressure rose slightly, to 290.5 psig. Throughout the event, there was little change in the boilup, reux, and product owrates or in the condenser outlet temperature. Testing. Pressure gages were installed to measure the change in pressure across the condenser. Field measurements taken during an instability at high reux and product owrates showed that the condenser pressure drop rapidly decreased (due to reduced head) and condensate temperature quickly rose as the liquid slugged from the condenser to the drum, suggesting that the condensate seal in the condenser could have been broken. Hydraulic analysis. The event in Figure 5 was considered most suitable for hydraulic analysis because the reux and product owrates were low and the condensate seal was not
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
broken, as evidenced by the condenser outlet temperature remaining unchanged. Because the line losses were signicant, Eq. 3b was used for the pressure balance. Table 2 presents the calculated pressure drops based on measurements just prior to the event in Figure 5. The 1.1-psi discrepancy is probably due to inaccuracy in the measurement of the pressure drop across the control valve or in the calculations. Figure 6 is a plot of owrate vs. valve opening at various valve pressure drops based on the control valve characteristics and the physical properties for the vapor. Figure 6 shows that opening the valve from 50% at a 4.5-psi pressure drop to 80% at a 12-psi pressure drop increased the vapor owrate through the valve from 5,500 lb/hr to 31,000 lb/hr. Thus, over the 5-min event, with approximately a linear increase in valve opening and pressure drop, the additional vapor that entered the drum (above the normal rate) was about 1,000 lb. The drum is 9 ft in diameter and just over 20 ft long (tangent to tangent). Raising the drum liquid level from 40% to 80% requires 21,000 lb. Adding this quantity of liquid to the drum over 5 min is a tremendous movement of liquid, increasing the amount of liquid entering the drum by a factor of 3, from 17,000 bbl/day to about 50,000 bbl/day. The key to understanding the mechanism is to determine where this huge amount of liquid came from. Mechanism. The additional vapor entering the drum accounts for about 1,000 lb. This may be augmented by condensation of the vapor present in the drum vapor space, which is replaced by the liquid during the event. Condensation of the drum vapor accounts for another 2,000 lb. That leaves 18,000 lb to be explained. One explanation may be the onset of heavy rain. Concise records of the ambient conditions at the renery at the time of the event were not available, but records from the nearby New Orleans airport show light rain throughout the evening, with periods of easing off and others of intensifying. The ambient temperature was about 63 F and wind was about 4 mph, occasionally gusting to 20 mph. °
Table 2. This table shows the pressure balance, with calculated pressure drops and liquid heads for both sides of Eq. 3b, for the conditions just prior to the disturbance observed in the case study. Term
Left-Hand Side of Eq. 3b
ΔP bypass, psi
4.5
ΔP bypass-line, psi
0.1
Right-Hand Side of Eq. 3b
0.5
ΔP liq-line, psi
0.4
ΔPvap-line, psi
0.5
H, psi
2.1 4.6
°
°
°
50,000 Valve dP, psi 15 40,000 13 11 9 r h 30,000 7 / b l 5 , w 3 o l 20,000 F 1 10,000
0
ΔPcond , psi
Total, psi
The incremental amount of vapor condensation in the drum during heavy rain can be calculated from Eq. 10. As stated earlier, ambient-loss heat-transfer coefcients for uninsulated drums vary from about 2–3 Btu/hr- F-ft2 in dry, windless conditions to about 15–20 Btu/hr- F-ft2 in rainy and windy conditions. Based on a change in heat-transfer coefcient of 15 Btu/hr- F-ft2 upon the onset of a sudden heavy rain, the increase in the condensation rate is calculated to be about 3,000 lb/hr. This would diminish to about 2,000 lb/hr when the drum liquid level rises to 80% and the drum shell area exposed for condensation, A DW in Eq. 10, decreases. Over the 5-min event, heavy rain accounts for additional liquid generation of a mere 200 lb, which does little to explain the 18,000 lb liquid movement. One conceivable source remains: liquid from the condenser was drawn into the drum. There are two condenser shells in parallel, each 46 in. in diameter and 20 ft long. Allowing for the tube volume, the condensers hold approximately 8,000 lb of liquid when full. During the low-owrate operation prior to the event (with a reux-plus-product owrate of 17,000 bbl/day), and at the relatively cold ambient temperatures, condensation occurred over only about 25% of the condenser’s heat-transfer area. So the condensers contained only about 6,000 lb of liquid. By itself, this falls short of explaining the large liquid movement of 18,000 lb. However, there was a large augmenting factor. As liquid is drawn out of the condensers, additional tube area is exposed, increasing condensation. Emptying all of the liquid from the condensers would quadruple the condensation area. The liquid generated by raising the condensation area from 25% at the beginning of the event to 100% 5 min later is calculated to be about 16,000 lb. This liquid together with the 6,000 lb drained from the condenser exceeds the 18,000 lb increase in drum liquid. This means that some
3.5
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
10
20
30
40
50
60
70
80
90
100
Valve Opening, %
plot of flowrate vs. valve opening at various valve pressure drops is based on the control valve characteristics and the physical properties of the vapor in the case study. Opening the valve from 50% at a 4.5-psi pressure drop to 80% at a 12-psi pressure drop increased the vapor flowrate through the valve from 5,500 lb/hr to 31,000 lb/hr. p Figure 6. This
CEP
February 2015
www.aiche.org/cep 43
Back to Basics
liquid remained in the condensers, retaining their liquid seal and the subcooled outlet temperature, as shown in Figure 5. The intensied condensation rate reduced the tower pressure. To reinstate the tower pressure, the hot-vapor bypass valve opened rapidly as the pressure fell. Eventually the hot-vapor owrate caught up with the disturbance. The drum liquid surface heated up, the drum vapor pressure rose, the differential pressure across the valve declined, the liquid returned to the condenser, column pressure went back up, and normal operation resumed. During the disturbance event, the low-owrate operation and the low cooling-water temperature (inferred from the low condenser outlet temperature) created a relatively large liquid inventory in the condensers and sufcient condensation capacity to cushion against total emptying of the condensers. Had the tower overhead owrates been higher or the ambient temperature warmer, the liquid inventory in the condensers would have been smaller, and it would have been much easier to lose the condenser liquid seal, which would have resulted in loss of subcooling and compounded the upset. Initiation. Up to this point, we followed the movement of liquid produced by a strong suction. However, we did not address the cause of this suction. To move 18,000 lb of liquid in 5 min takes a very strong suction force. Figure 5 shows a fast rise in the differential pressure across the hot-vapor bypass control valve (green triangles). This value is the difference between the tower pressure and the drum pressure, and since the tower pressure changed only slightly, most of the pressure change occurred in the drum pressure — 8 psi over 5 min. The drum pressure is the vapor pressure of the liquid surface in the drum. An 8-psi reduction in vapor pressure corresponds to cooling of the liquid surface by only 2.5 F (Row 1 in Table 1). The problem now is to identify the source of that 2.5 F cooling of the drum liquid surface. One possibility is heavy rain. However, as stated earlier, the heavy rain would have increased the condensation rate by only 2,000–3,000 lb/hr, which would be easily offset by the additional 25,000 lb/hr ow through the hot-vapor bypass during the same period. Furthermore, there were many other events (including the tests) that took place under dry conditions. Another possibility is that hot-vapor impingement agitated the liquid surface. However, at the beginning of the event, the vapor velocity was about 6 ft/sec, making ρvV 2 = 100 lb/ft-sec2 — too low to rufe the liquid surface 5 ft below (ρv is vapor density [lb/ft3] and V is vapor velocity at the drum inlet nozzle [ft/sec]). Also, the column was commonly operated at this vapor velocity without any problems. During the rst minute of the event, the valve pressure drop rose from 4.5 psi to 7 psi, corresponding to a drop in drum surface temperature of 1 F. At the drum pressure of °
°
°
44
www.aiche.org/cep
February 2015 CEP
285 psig, the liquid surface was at 125 F, while the bulk subcooled liquid in the drum was at about 66 F. As little as 1.5% of the hot drum liquid surface being replaced by subcooled liquid can explain a temperature drop of 1 F and a consequent pressure decrease of 2.5 psi. This brings us to the liquid entry. The liquid enters via a 6-in. nozzle at the bottom of the drum, discharging upward at 5.7 ft/sec at the beginning of the event, a velocity exceed ing the good design practice of 4 ft/sec maximum. Normally, this would not be an issue, but here, the drum liquid level was only 3.7 ft above the liquid entry. The initial momentum may have carried some of the subcooled liquid jet to the surface and caused it to pierce the hot-liquid surface. Instability may occur when the subcooled liquid reaches the drum surface. Figure 5 shows that just prior to the pressure differential rise, the tower pressure fell by about 0.5 psi and the drum level rose by 2–3%. Both indicate an increase in the condensation rate, possibly initiated by heavy rain hitting the drum above the liquid level (i.e., the part of the drum corresponding to the vapor space). The drum pressure fell, drawing liquid from the condenser. A 2% increase in the liquid level is equivalent to about 1,000 lb. Pulling 1,000 lb from the condensers would raise the exposed condensation area by about 50%, quickly reducing tower pressure. The bypass pressure drop remained constant at that time, meaning that both the drum and tower pressures fell by the same 0.5 psi. The additional liquid ow, about 1,000 lb in 2 min, would increase the drum liquid inlet velocity from 5.7 ft/sec to 7 ft/sec. The intensied jet raised more subcooled liquid to the liquid surface, generating a fountain effect that cooled the surface. The drum pressure fell, more liquid was drawn from the condenser, the jet and the fountain intensied, the surface cooled further, more liquid was drawn from the condenser, and so on. This process was self-accelerating. At the pressure differential peak, the liquid rate had tripled and the liquid jet velocity had exceeded 15 ft/sec, high enough for the jet to break through the liquid surface, even at the higher level of 80%. At the pressure differential peak, the hot-vapor bypass ow reached 31,000 lb/hr, the vapor inlet velocity 34 ft/sec, and ρvV 2 3,300 lb/ft-sec2, while the level rose to 80%, which was 2.1 ft below the vapor inlet. At some time during the event, vapor impingement on the liquid surface caused more subcooled liquid to rise to the surface, augmenting the instability. Solution. The low pressure drop due to friction in the condenser and its inlet vapor and outlet liquid lines, a total of 1.4 psi (Table 2), offers little hydraulic resistance to counter the self-accelerating fountain effect. Adding a throttling valve in the lines leaving the condenser can mitigate the self-accelerating process by creating additional resistance to the liquid ow being drawn into the drum. The line pressure °
°
°
Copyright © 2015 American Institute of Chemical Engineers (AIChE)
drop, which increases with the square of the ow, severely limits the increase in liquid ow to the drum, giving the hot vapor a chance to catch up. A manual throttling valve was installed in the line from the condenser to the drum, as in Figure 3. The differential pressure across the valve is measured locally and maintained at 2 psi. The operators check the differential pressure across the valve and adjust it as needed. This completely eliminated disturbance events. The modied system can now run with a reux-plus-product owrate up to 40,000 bbl/day during both summer and winter with no instability. Adding the throttling valve raised the hot-vapor bypass pressure drop from about 1.4 psi to 3.4 psi. The throttling valve slows the rate of drainage from the condenser, making it more difcult to suck the condenser liquid into the drum. Closing thoughts. This case demonstrates that hydraulic instability can be eliminated by adding a throttling valve in the condensate outlet line. Ideally, the valve should have a pressure drop greater than 3–4 psi and should be installed more than 10 line diameters away from the drum liquid inlet to minimize turbulence at the drum inlet. The hydraulic analysis also suggests that it should be possible to mitigate similar hydraulic disturbances by keeping the liquid inlet velocity low, installing a horizontal bafe or a stilling cham ber above the liquid inlet, and/or extending the liquid inlet into the drum and directing it horizontally. (These changes were not incorporated in this case.) Another good practice is to install a horizontal bafe in front of the vapor nozzle to disperse the vapor ow and prevent it from impinging on the CEP liquid surface upon intensication.
HENRY Z. KISTER is a Fluor Corp. Senior Fellow and Director of Fractionation Technology (Phone: (949) 349-4679, Email: henry.kister@fluor.com). He has over 30 years of experience in design, troubleshooting, revamping, field consulting, control, and startup of fractionation processes and equipment. He is the author of three books, the distillation equipment chapter in Perry’s Handbook, and over 100 articles, and has taught the IChemE course “Practical Distillation Technology” about 450 times in 25 countries. A recipient of several awards, Kister obtained his BE and ME degrees from the Univ. of New South Wales in Australia. He is a Fellow of AIChE and IChemE and a member of the National Academy of Engineering, and he serves on the Fractionation Research, Inc. (FRI) Technical Advisory and Design Practices Committees. DARYL W. HANSON is a technology advisor (Phone: (210) 345-5929, Email:
[email protected]) at Valero Energy and is focused on distillation, fractionation, and separation issues. He is responsible for design, troubleshooting, and operation/startup issues at 14 refineries and many ethanol plants. His previous experience includes positions at Glitsch, Koch-Glitsch, and Process Consulting Services. He has authored over 20 articles and has a BS in chemical engineering from Texas A&M Univ.
Literature Cited 1.
Kister, H. Z., “What Caused Tower Malfunctions in the Last 50 Years?,” Chemical Engineering Research and Design, 81 (1), pp. 5–26 (Jan. 2003).
2.
Kister, H. Z., “Distillation Troubleshooting,” Wiley InterScience, Hoboken, NJ (2006).
3.
Smith, C. L., “Distillation Control — An Engineering Perspective,” Wiley, Hoboken, NJ (2012).
4.
Whistler, A. M., “Locate Condensers at Ground Level,” Petroleum Renery, 33 (3), pp. 173–174 (1954).
5.
Hollander, L. “Pressure Control of Light-Ends Fractionators,” ISA Journal, 4 (5), pp. 185–187 (1957). Chin, T. G., “Guide to Distillation Pressure Control Methods,” Hydrocarbon Processing, 58 (10), p. 145 (1979). Kister, H. Z., “Distillation Operation,” McGraw-Hill, NY (1990). Sloley, A. W., “Effectively Control Column Pressure,” Chemical Engineering Progress, pp. 38–48 (Jan. 2001). Kister, H. Z., and J. F. Litchfeld, “Distillation: Diagnosing Instabilities in the Column Overhead,” Chemical Engineering, 111 (9), pp. 56–59 (Sept. 2004). Schneider, D. F., and M. C. Hoover, “Practical Process Hydraulic Considerations,” Hydrocarbon Processing, 78 (8), pp. 47–54 (Aug. 1999). Duguid, I., “Take this Safety Database to Heart,” Chemical Engineering, 108 (7), pp. 80–84, and the accompanying Case MS16 (July 2001). Sloley, A. W., “Simple Methods Solve Exchanger Problems,” Oil and Gas Journal, 96 (16), p. 85 (April 20, 1998).
6. 7. 8. 9.
10.
11.
12.
13. Shinskey, F. G., “Distillation Control for Productivity and Energy Conservation,” 2nd ed., McGraw-Hill, NY (1984). 14. Lupter, D. E., “Distillation Column Control for Utility Economy,” Presented at the 53rd Annual Gas Processors Association Convention, March 25–27, Denver, CO, (1974). 15. Nisenfeld, A. E., and R. C. Seemann, “Distillation Columns,” Instrument Society of America, Research Triangle Park, NC (1981). 16. Laird, D., and J. Cornelisen, “Control-System Improvements Expand Renery Processes,” Oil and Gas Journal, 98 (39), pp. 71–74 (Sept. 25, 2000). 17. Lieberman, N. P., “Troubleshooting Process Operations,” 4th ed., PennWell Books, Tulsa, OK (2009 ). 18. Hartman, E. L., and T. Barletta, “Reboiler and Condenser Operating Problems,” Petroleum Technology Quarterly, 8 (4), pp. 47–56 (Summer 2003). 19. Souza, L. L. G., “Model Devised for Plant Hot-Gas Bypass Systems,” Oil and Gas Journal, 108 (33), pp. 118–123 (Sept. 6, 2010). 20. Durand, A. A., “Sizing Hot Vapors Bypass Valve,” Chemical Engineering, pp. 111–112 (Aug. 25, 1980). 21. Buckley, P. S., et al., “Design of Distillation Column Control Systems,” Instrument Society of America, Research Triangle Park, NC (1985). 22. The American Petroleum Institute, “API Technical Data Book,” 7th ed., The American Petroleum Institute and EPCON International Publishing Services, Washington, DC (2005).
Acknowledgments
23. Lieberman, N. P., “Process Equipment Malfunctions,” McGrawHill, NY (2011).
The authors wish to thank Chad Jones (presently at Motiva) for his assistance in gathering the field data for this article, Mark Murphy at Fluor for his analysis of control valve characteristics, and Walter Stupin at Fluor for his invaluable input to the hydraulic analysis.
24. Kern, D. O., “Process Heat Transfer,” McGraw-Hill, NY (1950).
25. Lieberman, N. P., “Troubleshooting Process Plant Control,” Wiley, Hoboken, NJ (2009).
CEP
February 2015
www.aiche.org/cep 45