Professional Documents
Culture Documents
DISTILLATION
on-line
www.aiche.org
Optimize
Distillation
Columns
Application of mass-transfer
and pressure-drop
D
istillation is the dominant pro-
fundamentals can
cess for separating large multi- lead to improved
component streams into high pu-
rity products. So, the chemical designs for both trayed
process industries’ ongoing and packed columns.
quest to improve energy utilization, reduce
capital costs, and boost operating flexibility
is spurring increasing attention to distillation A distillation column can use either trays
column optimization during design. Design- or packings. Their mechanisms of mass
ers often approach column optimization in an transfer differ, but the key for both is a good
iterative manner, heavily relying on vendor approach to equilibrium through the genera-
experience and information. A good under- tion of large amounts of interfacial area. This
standing of mass-transfer and pressure-drop interfacial area results from the passage of
fundamentals, as they relate to optimization, vapor through the perforations of trays, or
will enable the column designer to indepen- the spreading of liquid on the surface of
dently judge vendor offerings and effectively packings.
determine the optimal equipment design. First, we will discuss the underlying phe-
©Copyright 2000 This article will address the following op- nomena for trayed columns and the design
American Institute
timization goals: (1) maximizing theoretical approaches that can be used to meet the three
of Chemical Engineers.
All rights reserved.
stages per height of section or column, (2) optimization goals. Then, we will address the
Copying and minimizing pressure drop per theoretical mechanisms and approaches for packed
downloading permitted stage, and (3) maximizing the operational columns. Finally, we will consider the selec-
with restrictions. range, turn-down, or turn-up. tion of trays vs. packing.
I
flows down the column
through downcomers and
then across the tray deck,
while vapor flows upward
through the liquid inventory
on the tray. Tray designs can be di-
vided into cross-flow and parallel-
flow types. Figure 1 illustrates the
conceptual differences. Cross-flow
trays are the most common and least
expensive, but parallel-flow trays, if
properly designed, can provide an ef-
ficiency that is 10% or more higher.
Figure 1 depicts a single-pass cross-
Parallel-Flow Trays Single-Pass Cross-Flow Trays
flow tray. By “pass,” we mean the
number of downcomers per tray. As
column diameter increases, the ratio ■ Figure 1. Types of trays.
■ Figure 2.
Types of tray-deck
mass-transfer
Fixed Valves devices.
Subscripts
L = liquid
ML = liquid molar Flooding
Dumping
MV = vapor molar
V = vapor
Opt = optimum Column Throughput
AH AH 1
= = 2 KS gins to dominate. Figure 5c gives a mean that more stages can be
AB Opt
AB Fr L = 0.5
g hL
plot of calculated optimum weir achieved within a given column
(6) height vs. effective froth density, height. In Figure 6a, typical values
When maximizing ηSECT/hT, where effective froth height is for this parameter are plotted vs.
there also is an optimal weir height. given by: fraction open area. As fraction open
This optimum can exist because, at area decreases, vapor velocity
larger values of outlet weir, pres- hL through the perforation increases,
sure drop increases at a higher rate h Fe = with K S in m/s giving more interfacial area. At ex-
K s 0.91
than efficiency. At very low values e – 12.55 tremely small values of open area,
of outlet weir height, liquid inven- (7) vapor velocity through the perfora-
tory is less, which cuts tray effi- The parameter ηSECT/TS is the tions is very large — this promotes
ciency, but tray pressure drop de- number of theoretical stages per entrainment and thus degrades
creases less rapidly, because hD be- height of column. Higher values overall tray efficiency. In Figure
300
iso/normal Butane enhancement of 10% or more in the
per Meter
Sieve trays normally have a better ef- 3. Maximizing the tray spacing valves, because their minimum flow
ficiency than cap trays; so, unless ex- will raise maximum throughput and area varies with vapor throughput,
treme turndown is particularly impor- not adversely impact weeping, but can have a wider operating range —
tant, the added cost of small bubble- also will lower the number of stages about a factor of three — with a
cap trays is not justified. within a given column height. lower maximum pressure drop than
To minimize the pressure drop per 4. Decreasing the liquid inventory sieve trays, but often with a lower ef-
theoretical stage: by increasing the number of tray ficiency and a higher cost. Small-size
For sieve trays: passes or lowering the weir height fixed-valves generally will have a
1. Use the largest fraction open will enhance turn-down and generally greater maximum range than large-
area that will not result in weeping. will cut entrainment for a given col- size fixed or floating valves when tray
Based on the literature for cross-flow umn throughput. Reducing the outlet count per section height is also im-
trays, this maximum occurs around: weir height below 25 mm, however, portant and a larger pressure drop at
offers little advantage. The lower liq- maximum rates is acceptable. Their
(8) A H 1 uid inventory also will decrease tray ability to turn down, however, is not
= 2 KS (8) efficiency. The spray-like regime as great as floating valves. Small-size
AB Fr L = 0.5
g hL
must be avoided. bubble caps, although expensive, can
The open literature does not address 5. Some high-capacity tray designs have substantial turn-down capability
this for parallel-flow trays. with downcomers that allow liquid to and, with special design considera-
2. Use the smallest practical perfo- drop onto the tray deck below can re- tions to prevent liquid/vapor bypass
ration diameter. sult in weeping through the tray at extreme turn-down, can give an
3. Specify a low outlet weir, gen- openings under the downcomer. For operating range greater than a factor
erally less than 50 mm. such designs, the turn-down capabili- of five. CEP
T
he wide range of com- of section or column, (2) minimizing lustrates some of the common types
mercial packings and pressure drop per theoretical stage of of random packing. Both rings and
their variations in ge- separation, and (3) maximizing the saddles typically have features that
ometry introduce sig- operating range of the column that re- are formed into the packing element.
nificant empiricism into sults in reasonable performance. These features and other details can
the design of packed Columns can be divided into two provide surface area within the struc-
columns. Application of fundamental major groups based on the type of ture, increase resistance to deforma-
mass-transfer and pressure-drop prin- packing used — either random (or tion, and also prevent elements from
ciples, however, still can lead to the dumped) packing, or structured (or nesting.
identification of general optimization ordered) packing. Random packings Within a packing type, elements
rules. As we did for trayed columns, often are either cylindrical (ring) are available in a variety of sizes and
here we will develop rules for the fol- shaped, or half cylinder (saddle) materials of construction; elements
lowing optimization goals: (1) maxi- shaped. We will call a single ring or fabricated from a given material look
mizing theoretical stages per height saddle a packing element. Figure 1 il- essentially alike except in their size.
X
Primary Flow
Direction of Vapor
Primary Vapor
Flow Direct
X
Fold
Angle
X X
X γ
Primary Flow
Direction of Liquid
X
Corrugation
δP Angle
This means that they are geometrical- aS can be calculated to be about characteristics can be controlled by
ly similar and differ only by a charac- 7.25/δP. Packing elements installed in changing the corrugation angle, γ,
teristic dimension, δP. We will find a random or dumped fashion orient the fold angle, α, and the height of
this very useful as we develop our themselves in an irregular manner, the corrugation. We will define the
theoretical understanding. For ran- and this impacts the constant relating characteristic dimension of struc-
dom packings, we will define δP as aS and δP. The constant also can de- tured packing, δP, to be the height
the nominal packing size or diameter. pend upon the ratio of δP to the verti- of the corrugation. Most such pack-
As an example, a 25-mm Pall Ring cal height of the element. In general, ings have a corrugation angle of
has a nominal diameter and value for for both random rings and saddles, aS 45° and a fold angle of 90°; the
δP of 25 mm. = 5.7/δP within a band of about ±10%. height of the corrugation, therefore,
Mass transfer occurs at the inter- The second general category of is the primary geometry parameter
face between vapor and the liquid film packing commonly is referred to now used by the column designer for
on the packing surface. The mass as structured packing. It can be made optimization.
transfer is significantly impacted by of woven or solid material, usually The constant relating aS to δP de-
the specific area of the packing, aS, metal, which typically is corrugated pends upon α. If we neglect the sheet
and this is controlled by the packing’s and bundled into segments that are thickness and any change in surface
nominal size. As long as the diameter placed layer by layer into the column area resulting from surface texture or
of the column is very large compared shell. Most commercially available perforations, aS = (2/sin (α/2))/δP.
to δP, aS is inversely proportional to structured packings have perforations Thus, for the typical fold angle of
δP. The constant relating the two de- or texturing through stamping to help 90°, aS = 2.8/δP. This shows that the
pends upon the packing type and how promote mixing and liquid spreading, specific area for structured packing is
the elements are arranged. For exam- as well as to modestly increase sur- less than that for random packings for
ple, if we assume that the packed col- face area. Figure 2 illustrates the gen- a given value of δP. Keep in mind,
umn is filled with thin-walled cylin- eral geometry characteristics of the however, that we define δP as the cor-
drical elements stacked in an ordered type of structured packing often used rugation height for structured packing
fashion with 60° pitch and the ele- today. and the nominal packing-element di-
ments are just touching, the value for Pressure-drop and mass-transfer ameter for random packing.
HETP
stage of separation. Figure 3 shows
the general relationship between
HETP and column throughput. Per-
formance remains relatively constant Poor Heavy
Enhanced
and stable except at both low and Wetting
Mass Transfer
Entrainment
high rates. At low liquid rates, sheets
of liquid are not stable and rivulets do Column Throughput
not spread sufficiently to wet the en-
tire packing surface; so, mass transfer
is poor. As column throughput in-
creases, there is a large stable region
where packing is well wetted and streams into a combination of Stable operation
HETP is relatively constant. At the rivulets and films that fully migrate Most basic mass-transfer text-
higher-rate portion of the stable re- throughout the packed column cross- books (for example, Ref. 1) develop
gion, mass transfer improves (HETP section. The liquid flow rate per wet- equations for continuous distillation
gets lower). This improvement is ted perimeter controls liquid-film hy- assuming a downward flowing liq-
caused by an increase in interfacial drodynamics. We will designate Γ as uid phase in contact with an upward
area due to liquid waves and entrain- the average value for liquid flow rate flowing vapor phase. A rearrange-
ment. At higher rates, performance per wetted perimeter, assuming all of ment of these equations leads to Eq.
decreases rapidly due to substantial the liquid is uniformly distributed. 1 for the HETP vs. the appropriate
liquid entrainment in the vapor being By mass balance, Γ = L/aS, where L mass-transfer coefficients (see box
carried up the column. This entrain- is the liquid mass flux flowing down below), which can be rearranged to
ment degrades the composition pro- the column. For a given packing type Eq. 2 (see box) where VS is the su-
file and can result in liquid flow and and distillation system, a minimum perficial velocity of vapor up the
vapor flow redistribution. value for Γ is required for the pack- column. The term in the bracket is
Optimization of the packing for ing to be substantially wet. We will the increase in HETP resulting from
(1) maximizing theoretical stages per designate this minimum value as ΓC; liquid-phase resistance. Frequently,
height of section or column, (2) mini- if Γ is below this value, wetting will however, vapor-phase resistance
mizing pressure drop per theoretical be poor, resulting in poor mass trans- dominates. In such cases, this
stage of separation, and (3) maximiz- fer and a large HETP. ΓC would be bracketed term is approximately
ing the operating range of the column expected to be a function of contact unity and assuming that kV can be
that results in reasonable performance angle and physical properties, and correlated using the format of Eq. 3
often requires the column to operate may depend upon packing surface- we get: Eq. 4 (see box on next page).
at the extremes of the stable operating texture details. It should be a con- Packed column data can be used to
range. We need, therefore, an under- stant, however, for a given distilla- arrive at values for p and n, but, by
standing of the phenomena that con- tion system (composition and pres- analogy to heat transfer and wetted-
trol each of the three regions: poor sure) and packing type. wall-column mass transfer, p is ex-
wetting, stable operation, and heavy
entrainment.
Packing Factor, FP
weakly depends upon vapor flow 100
(about VS0.2), but strongly depends
upon the characteristic dimension of
the packing (about δP0.8). This weak
dependency of HETP on VS is why 10
the concept of HETP is so useful.
We also will need a correlation for
pressure drop. One based on a signifi-
cant amount of experimental data is
the Eckert Generalized Pressure Drop 1
0.1 1.0 10
Correlation for structured packing,
this approach defines two axis: Characteristic Dimension, δP, in.
ρ 0.5
µ 0.05
Y = V S ρ –Vρ FP0.5 ρ L
L V L
cant for values of X exceeding 0.20, stack in a more open manner. For
(5) and for values of pressure gradient structured packing, for equal values
and more than 0.5 in. of water pressure of δP, FP is about one-fourth of the
drop/ft of packing height. FP is purely minimum value for random packing.
ρ 0.5
X = L ρV (6) an experimental constant determined But, because the definition of δP for
V L
by minimizing the error between the random and structured packing differ,
This correlation is graphically repre- data and the correlation, and is a a comparison of values of FP at
sented by a plot of Y vs. X along with function of packing type and charac- equivalent specific area is more ap-
lines of constant pressure gradient. A teristic dimension. propriate. We find that structured
key test of the validity of this ap- In Figure 4, we have plotted FP packing values for FP are about one-
proach is whether a single value of values from Kister (3) vs. δP for sev- half the lowest value for random
the packing factor, FP, adequately eral commercial packings. We find packings at equal specific area.
agrees with all of the constant-pres- that, for a given value of δP, FP for For low and moderate loading,
sure gradient lines. Kister and Gill (2) Pall Rings is less than half that of pressure drop is directly proportional
evaluated structured-packing pres- Raschig Rings. This is attributable to to the value for FP; therefore, the
sure-drop data and found that a con- the openness of the Pall Rings com- pressure drop for structured packing
stant value of FP for structured pack- pared to Raschig Rings. The other is about one-half that of the lowest-
ing could not be found; they, there- random packings show an additional pressure-drop random packings at
fore, redrew the constant-pressure one-third to one-half reduction in FP, equivalent values of specific area and
gradient lines for structured packing. resulting from further increases in equal column throughput.
The difference in the random-packing perforations of the element and, for With reasonable accuracy, a curve-
and structured-packing constant-pres- some elements, a reduction in ele- fit of published values for FP is:
sure gradient lines becomes signifi- ment height that allows elements to FP = C3 (δP)-1.1 with δP in inches
(7)
Alternatively, because aS is inversely
proportional to δP,
k V δP ρ V δ µV
p n
= C1 V µ S P (3) FP = C4 (aS)1.1 with aS in ft2/ft3
DV V ρ V DV (8)
The values of C3 and C4 are given in
C VS δ P
1– p
µV Table 1 and are important when we
ln λ
p–n
HETP = a 2 (4) look at optimization for different
i DV ρ V DV λ–1
types of packings.
The Eckert-type correlation does
not allow easy identification of the
∆Z δP ρ L g δP (9) variation in the constants C3 and C4 ly, due to more liquid entrainment in
that relate the packing factor to δP vapor and a greater vapor velocity re-
The pressure gradient ∆P/∆Z is ex- and aS. Manufacturing techniques sulting from higher liquid holdup. In
pressed in units of height of liquid for metal, plastic, and ceramic mate- addition, mass transfer begins to drop
per height of packing, εP is the height rials differ, however, resulting in as increased liquid entrainment flows
of liquid-flow-induced roughness, some differences in element geome- up the column. At very high through-
and g is acceleration due to gravity, try even if elements are part of the put rates, substantial quantities of liq-
which is required as we are express- same packing type. Thus, A1 will uid flow up portions of the column.
ing pressure drop in terms of liquid vary depending upon material of Because packing is relativity open to
height. A and B are constants that we fabrication. This expression for the cross-flow, liquid and vapor tend to
will borrow from rough tube data. We pressure gradient is not a complete redistribute themselves; so, local val-
will assume that εP can be approxi- representation of pressure drop, but ues of L/V vary substantially, result-
mated from the average height of the still has significant similarities with ing in a sharp dropoff in column mass
liquid film; this can be calculated the Eckert-type pressure-drop corre- transfer. In packed columns, there is
from a simple force balance, yielding: lation, especially in the nonloaded no discrete flood point, but operation
region (specifically at values of X of a column anywhere within this
3 µL Γ
0.33
less than 0.5, where pressure drop is very heavily loaded region is undesir-
δF =
ρL ρL – ρV g (10) largely a function of Y2). We expect able. A force balance between the up-
that, for this non-heavily-loaded re- ward pressure drop force on a liquid
where Γ is the average mass-flow rate gion, there is a weak dependency on film and the downward gravity force
of liquid per wetted surface. From be- liquid mass flux (about L0.08). Also, gives at flood:
fore, we know that Γ = L/aS, and, be- for geometrically similar packings,
cause aS and δP are inversely propor- FP is approximately inversely pro- δ P ∆P = A2 δF
∆Z Flood (12)
tional, Γ ∝ L δP. Substitution gives: portional to the characteristic dimen-
Eq. 11 (see box below). Based on sion (about δP-1.15). These trends will If we define fLFlood to be the liquid
rough-tube pressure-drop data, B is be useful as we draw conclusions void fraction at flood, then fLFlood = aS
about 0.25. The constant A1 is ex- about the optimization of packed δF. Because aS is inversely propor-
pected to vary for different types of columns. tional to δP, we have the interesting
relationship:
∆P ∝ fL Flood (13)
3 µL L
0.33B
ρ V V S2 ∆Z
∆P = A Flood
∆Z 1
δP ρ L ρ L – ρ V g
2 ρ L g δP (11) for any particular packing type.
This force balance, coupled with
the equation for δF, and our under-
standing that Γ is proportional to L
δP, predicts that the pressure drop re- to δP-1.0. We also will assume that the packing family, for example, Pall
sulting in flooding is proportional to values for p and n are 0.8 and 0.3, re- Rings or structured packing with uni-
δP-0.8. It also indicates that film thick- spectively. Therefore, we get Eq. 15 form corrugation and fold angles, A1
ness plays some role. In support of (see box below). and C5 are constant. Therefore, within
this force-balance approach, we can Most vendors of structured pack- a packing family and for a given dis-
use the empirical Kister and Gill cor- ing have standardized on corrugation tillation system:
relation for packing flooding (4), Eq. angle and fold angle, namely, γ = 45°
∆P
14 in box below. This correlation and α = 90°; therefore, in general, ∆Z ∝ a 2.35 V 1.8 L 0.08 (17)
does not agree with the dependency metal structured packings from dif- HETP S S
on liquid rate (the dependency on ferent vendors are very similar with
δF) predicted by the force balance, the possible exception of surface de- Reducing the vapor throughput will
but, because we have shown that FP tails. These surface differences can decrease pressure drop, but will sub-
is proportional to δP-1.1, the Kister have some impact on the liquid resis- stantially increase capital investment
and Gill correlation for flooding tance, but we are assuming for this and is rarely justifiable. Selecting a
does indicate that the pressure-drop analysis that vapor-phase resistance lower value of aS to give sufficient
gradient at flood is proportional to dominates. For structured packing HETP, but at an acceptable pressure
δP-0.77, which is identical to our when α = 90° and γ = 45°, this equa- drop, is the typical optimization.
force-balance-approach prediction. tion for HETP shows that, for a given To maximize the stable operating
separation system and column range:
The approach used throughput, the only geometry vari- For packed columns, the range of
We now can combine these equa- able that significantly impacts HETP stable HETP performance is bounded
tions derived from mass-transfer and is δP. Smaller values of δP give at high throughputs by high pressure
pressure-drop fundamentals and look greater values of specific area, aS, and drop and poor mass-transfer perfor-
at their implication on column opti- lower section heights or more mass mance as flooding is approached, and at
mization. Due to decreasing cost and transfer within a given section height. low throughputs by poor performance
several performance advantages, This equation predicts that HETP de- resulting from inadequate wetting. The
structured packing is becoming more pends upon δP1.2 (or aS-1.2), which is characteristic dimension impacts
popular. We, therefore, will address very close to the results from pub- packed column flooding in two ways.
column optimization for structured lished data. First, pressure drop for a given vapor
packings and then discuss, for each To minimize pressure drop per throughput is approximately inversely
optimization goal, the approach for theoretical stage of separation: proportional to δP. Second, as δP gets
random packings. We observe the relationship given smaller, the pressure drop that will re-
To minimize HETP: in Eq. 16 (see box below) from the sult in flood is approximately inversely
We will assume that gas-phase re- equations for pressure drop and proportional to δP. The net result is that
sistance dominates and packing is HETP. for geometrically similar packing:
fully wet such that ai is proportional As discussed before, within a KVFlood ∝ δP (18)
For a given distillation system and
design L/V, there is a minimum ΓC.
We can show that:
Trays vs. packing 4. For applications requiring low promote high entrainment and low
Clearly, both trayed and packed pressure drop per theoretical stage, tray efficiency. This can be mitigat-
columns will continue to have signifi- packing has a significant inherent ad- ed with small perforations, but an
cant roles to play in distillation. Un- vantage, because the interfacial area alternative is to opt for packed
derstanding the factors involved in for packed columns is generated columns if liquid rates are sufficient
optimizing each provides a basis for through liquid spreading on the pack- to obtain good wetting. An added
choosing between trays and packing. ing surface — this is a low-pressure- advantage for packing is its inher-
Here are some general guidelines: drop phenomenon compared to the ently lower pressure drop, which is
1. When the service is nonfouling, mechanism required to generate high particularly important for low-pres-
either type of column can be designed mass-transfer efficiency within a sure applications.
to yield comparable theoretical stage trayed column. 6. Trayed columns are intrinsically
counts per section height. Packing 5. For low-pressure distillation lower cost than packed columns, be-
has the disadvantage of requiring re- applications, liquid flow rates tend cause far less surface is needed for
distributors if large stage counts with- to be very low. The resulting modest trays than for packings, and trays re-
in a column section are required. liquid inventory fosters spray-like quire far-lower-cost internals than
2. Fouling can pose problems for conditions with trays that, in turn, packings. CEP