Fractionators

Introduction................................................................. 60 Diameter of SieveNalve Trays (F Factor) ................72

Relative Volatility ........................................................ 60 Diameter of SieveNalve Trays (Smith) ........... 73
Minimum Reflux ......................................................... 61 Diameter of SieveNalve Trays (Lieberman) ............75
Minimum Stages .......................................................... 62 Diameter of Ballast Trays .......................................... 75
Actual Reflux and Actual Theoretical Stages ..........62 Diameter of Fractionators. General ............. 77
Reflux-to-Feed Ratio ................................................... 63 Control Schemes .......................................................... 77
Actual Trays................................................................. 64 Optimization Techniques ............................................ 81
Graphical Methods ..................................................... 64 Reboilers...................................................................... 84
Tray Efficiency ............................................................ 67 Packed Columns .......................................................... 88
Diameter of Bubble Cap Trays ................. 71

59
60 Rules of Thumb for Chemical Engineers

Introduction
More shortcut design methods and rules of thumb have trated by actual field causes, I recommend Lieberman’s
been developed for fractionation than probably any other design book.’
unit operation. For example the paper’ reprinted in
Appendix 5 on development of shortcut equipment design
methods contains 18 references for fractionation shortcut Sources
methods out of 37 total. Both the process and mechani-
cal aspects of fractionation design have useful rules of 1. Branan, C. R., Development of Short-cut Equipment
thumb. Many of the mechanical design rules of thumb Design Methods, ASEE Annual Conference Proceed-
become included in checklists of do’s and don’ts. ings, Computer Aided Engineering, American Society
For troubleshooting fractionators see the troubleshoot- for Engineering Education, 1985.
ing section in this book. For a general understanding of 2. Lieberman, N. P., Process Design For Reliable Oper-
how trayed and packed fractionating columns work, illus- ations, 2nd Ed., Gulf Publishing Co., 1989.

Relative Volatility
For quick estimates, a relative volatility can be esti- By definition
mated as follows:
The equilibrium vaporization constant K is defined for pi = rIYi
a compound by
where
K. =-Yi
’ xi rI = Total pressure of the system

where so

Yi = Mole fraction of component i in the vapor phase PiXi = rIYi

Xi = Mole fraction of component i in the liquid phase
and
To calculate a distillation, the relative volatility a is
needed, it is defined as

Therefore for systems obeying Raoult’s Law

where i and j represent two components to be separated.

Raoult’s Law for ideal systems is

where Source

pi = Partial pressure of i Branan, C. R., The Process Engineer S Pocket Handbook,

Pi = Vapor pressure of pure component i Vol. 1, Gulf Publishing Co., 1976.
 Fractionators 61

Minimum Reflux
For shortcut design of a distillation column, the cation composition relationship along with enthalpy
minimum reflux calculation should be made first. correction.
For more than two components, calculation is not so
Minimum Reflux introduction by Keith Claibornl easy. Bounds can, however, often be perceived. If in the
feed there is only an insignificant amount of materials of
A quick overview of minimum reflux requirements as volatility intermediate between the light and heavy keys,
well as improved understanding of the meaning of the term the following applies:
can be had by considering separation of a binary feed. Minimum total reflux (lbs or molslhr) corresponding to
For a binary, let's denote as V the fractional molar split given total feed will be greater than if only the actual total
of the feed into overhead product and as L the fractional mols of heavy and light key components were present.
split into bottom product. Calculate compositions of the Reflux need will be less than if the actual total mols of
flash separation of feed into vapor v and liquid 1 to give feed were present, but composed only of light and heavy
vll = VIL. The resulting vapor can be regarded as being keys. The more closely non-keyed components are clus-
composed of a portion d of the overhead product com- tered to volatilities of the keys, the nearer are reflux needs
position and a portion r of the flash liquid composition. to that calculated for the binary and total feed volume.
It is then easy to calculate what the proportions must
be so that d x overhead composition + r x flash liquid Minimum Reflux Binary
composition = flash vapor composition. Thus:
a-b For binary or near binary minimum reflux ratio, use the
r/d = - following Underwood Equations.
b-c
where Bubble Point Liquid Feed
a = fraction high volatile component in overhead
specification
b = fraction high volatile component in flash vapor
Dew Point Vapor Feed
c = fraction high volatile component in flash liquid
The reason for this simple relationship is that the
concept of minimum reflux implies an infinite number of
stages and thus no change in composition from stage to
stage for an infinite number of stages each way from the Minimum Reflux Multicomponent
"pinch point" (the point where the McCabe-Thiele oper-
ating lines intersect at the vapor curve; for a well-behaved The Underwood Method will provide a quick estimate
system, this is the feed zone). The liquid refluxed to the of minimum reflux requirements. It is a good method to
feed tray from the tray above is thus the same composi- use when distillate and bottoms compositions are speci-
tion as the flash liquid. fied. Although the Underwood Method will be detailed
The total feed tray vapor can be regarded as compris- here, other good methods exist such as the ~rown- art in^
ing some material that passes unchanged through the and Colbum4 methods. These and other methods are dis-
upper section into product. The unchanged material can cussed and compared in Van Winkle's A method
then be regarded as being conveyed by other material (at to use for column analysis when distillate and bottoms
the composition of the reflux to the feed tray) endlessly compositions are not specified is discussed by Smith.6
recirculating as vapor to the tray above and, reliquified, The Underwood Method involves finding a value for a
refluxing onto the feed tray. constant, 8, that satisfies the equation
Binary minimum reflux so calculated implies feed
enthalpy just equal to the above started vapor V and liquid
L. Any increase or decrease in that enthalpy must be
matched by increase or decrease in total heat content of The value of 8 will lie between the relative volatilities
overhead reflux. Note that the Underwood2 binary reflux of the light and heavy key components, which must be
equation essentially computes the flash versus specifi- adjacent.
62 Rules of Thumb for Chemical Engineers

After finding 8, the minimum reflux ratio is determined where:

from Ls = Liquid molar rate in the
stripping section
LR= Liquid molar rate in the
rectification section
F = Feed molar rate
R, = Minimum reflux ratio

Nomenclature Sources

ai = Relative volatility of component i verses the heavy 1. Branan, C. R., The Fractionator Analysis Pocket
key component Handbook, Gulf Publishing Co., 1978.
8 = Underwood minimum reflux constant 2. Underwood, A. J. V., “Fractional Distillation of
XiF= Mol fraction of component i in the feed Multicomponent Mixtures.” Chemical Engineering
XiD= Mol fraction of component i in the distillate Progress, 44, 603 (1948).
XLD,XHD,XHB, XLB = mole fractions of light and heavy 3. Martin, H. Z. and Brown, G. G., Trans. A.I.Ch.E.,
components in the distillate and 35:679 (1939).
bottoms 4. Colburn, A. P., Trans. A.I.Ch.E., 37:805 (1941).
q = Thermal condition of the feed 5. Van Winkle, Matthew, Distillation, McGraw-Hill,
Bubble point liquid q = 1.0 1967.
Dew point vapor q =0 6. Smith, B. D., Design of Equilibrium Stage Processes,
General feed q = (Ls - LR)/F McGraw-Hill, 1963.

Minimum Stages
The Fenske Method gives a quick estimate for the LK = subscript for light key
minimum theoretical stages at total reflux. N, = minimum theoretical stages at total reflux
XHK = mol fraction of heavy key component
XLK= mol fraction of the light key component
OILWHK = relative volatility of component vs the heavy key

Nomenclature component

B = subscript for bottoms Source

D = subscript for distillate
HK = subscript for heavy key Fenske, M., Ind. Eng. Chem. 24,482, (1932).

Actual Reflux and Actual Theoretical Stages

The recommended’ method to use to determine the ac- 1.1-1.2 R, for a refrigerated system, and 1.2-1.35 R, for
tual theoretical stages at an actual reflux ratio is the a hot system.
Erbar/Maddox* relationship. In the graph, N is the theoret- Another useful molar ratio is reflux/feed, L/F. In binary
ical stages and R is the actual reflux ratio L/D, where L/D systems. L/F for all practical purposes is unchanging for
is the molar ratio of reflux to distillate. N, is the minimum wide differences in feed composition, so long as the fol-
theoretical stages and R, is the minimum reflux ratio. lowing hold:
The actual reflux ratio that one picks should be opti- 1. The distillate and bottoms compositions, but not nec-
mized from economics data. For a ballpark estimate use essarily the quantities, are held constant.
 Fractionators 63

2. The feed tray is kept matched in composition to the v R/(R+l)

feed (which means the feed tray moves with feed
composition changes). 0.90 -0.90

The reader can verify the above using the Underwood 0.80 -0.80
equations and the tower material balance. The author
once calculated a case where a large feed change would 0.70 -0.70
change L/D by 46%, whereas L/F changed only 1%. The
0.60 -0.60
stability of L/F is well proven in the field and is a good
factor to use in predicting the effect of feed changes for 0.50 -0.50
design and in an operating plant.
0.40 -0.40
The author has curve-fit the Erbar/Maddox curves
since readability for the graph is limited. For simplicity, -__-------
0.30 -0.30--’---
-/--
4

let:
0.20 -0.20--- Based on Underwood R,,
Extropoloted
x = N,/N
y = WR(R + 1)
z = Rm/(Rm+ 1) ! l I I I l I I l l I I I ( I l
0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
NM/ N
For y = A + Bx + Cx2 + Dx3 + Ex4 + Fx5, the follow-
ing table is presented: Figure 1. Plates-reflux correlation of Erbar and Madd
LIVE GRAPH
z A B C D E F Click here to view

0 .00035 .16287 -.23193 5.09032 -8.5081 5 4.48718 Sources

.1 .09881 .32725 -2.57575 10.20104 -1 2.82050 5.76923
.2 .19970 .14236 -.58646 2.60561 -3.1 2499 1.76282
.3 .29984 .09393 -.23913 1.49008 -2.43880 1.79486 1. Fair, J. R. and Bolles, W. L., “Modern Design of
.4 .40026 .12494 -.49585 2.1 5836 -3.27068 2.08333 Distillation Columns,” Chemical Engineering, April
.5 .50049 -.03058 .81585 -2.61 655 3.61305 -1.28205 22, 1968.
.6 .60063 -.00792 .60063 -2.0691 2 3.39816 -1.52243
.7 .70023 -.01109 .45388 -1.25263 1.94348 -.83334
2. Erbar, J. H. and Maddox, R. N., “Latest Score: Reflux
.8 .80013 -.01248 .76154 -2.72399 3.85707 -1.68269 vs Trays,” Petroleum ReJiner, May 1961, p. 185.
.9 .89947 .00420 .38713 -1.1 4962 1.40297 -.54487 3. Equations were generated using FLEXCURV, V.2.0,
1.0 1.0 -0- -0- -0- -0- -0-
Gulf Publishing Co.

Reflux-to-Feed Ratio
Heretofore, the reflux ratio has been defined as The reader can verify the above using the Underwood
reflux/distillate, L/D. Another very useful molar ratio is equations and the tower material balance. I once calcu-
reflux/feed, L/F. In binary systems, L/F for all practical lated a case where a large feed change would change L/D
purposes is unchanging for wide differences in feed com- by 46%, whereas L/F changed only 1%. Several investi-
position, so long as the following hold: gators report that the stability of L/F is well proven in the
field. L/F is a good factor to use in predicting the effect
of feed changes for design and in an operating plant.
1. The distillate and bottoms compositions, but not
necessarily the quantities, are held constant. Source
2. The feed tray is kept matched in composition to the
feed (which means the feed tray moves with feed Branan, C. R., Pocket Guide to Chemical Engineering,
composition changes). Gulf Professional Publishing, 1999.
64 Rules of Thumb for Chemical Engineers

Actual Trays
After actual theoretical trays are determined (see composition. Experience for a given service is the best
Actual Reflux and Actual Theoretical stages) one needs guide for extra trays.
to estimate the actual physical number of trays required
in the distillation column. This is usually done by divid- Source
ing the actual theoretical trays by the overall average frac-
tional tray efficiency. Then a few extra trays are normally Branan, C. R., The Process Engineer's Pocket Handbook,
added for offload conditions, such as a change in feed Vol. 2, Gulf Publishing Co., 1983, p. 4.

Graphical Methods
McCabe-Thiele Diagram Rectifying Section Operating Line

In addition to the previously mentioned shortcut equa- One point is on the 45" line at XD,and the slope is LR/VR,
tions, plotting a McCabe-Thiele' diagram is also a very where VRis the rectifying section vapor rate. Another point
useful tool. The equation for the equilibrium X-Y diagram is on the Y axis above the origin at DXJVR.
and plotting of the operating lines are described next.
Equilibrium Cuwe Stripping Section Operating Line
Xa One point is on the 45" line at XB, and the slope is
Y=
1+(a- l)X Ls/Vs, where S refers to the stripping section. Another
where Y = mole fraction of the light component in the point is the intersection of the rectifying section operat-
ing line with the q line. And still another point is on the
vapor
Y axis at the location below the origin of -(B/Vs)(XB).
X = mole fraction of the light component in the
liquid
Revised McCabe-Thieie Diagram
Another usehl form is (For Relative Volatilities under 1.25)
Y(1- X)
a= For relative volatilities under 1.25 the McCabe-Thiele
X(1- Y)
plot is difficult to use because all the lines are very close
together. Fortunately Ryan3 gives us an alternate proce-
q Line-The Operating Lines Intersect on This Line
dure for relative volatilities in the troublesome area below
1.25. Figure 1 shows the typical McCabe-Thiele diagram
The thermal condition of the feed is designated as q,
compared to Ryan's alternate plot in Figure 2. These
and is approximately the amount of heat required to
figures have been graciously provided by Mr. Ryan from
vaporize one mole of feed at the feed tray conditions,
his Microsoft Excel@files and are printed unchanged.
divided by the latent heat of vaporization of the feed. One
point on the q line is on the 45" line at XF. The rules are as follows:

Bubble point liquid feed, q = 1.0 (q-line vertical) 1. Plot y-x instead of y on the y-axis.
Dew point vapor feed, q = 0 (q-line horizontal) 2. The operating lines begin on the x-axis at XBand XD.
General feed, q = (Ls - LR)/F Both slopes are [(L/V)-I]. The rectifying line slope
Slope of q line = q/(q - 1) will come out negative and the stripping positive.
3. The q-line starts on the x-axis at XF. The value of q
where Ls = liquid rate in stripping section, lb mols/hr is the same as for conventional McCabe-Thiele. The
LR = liquid rate in rectifying section, lb mols/hr slope of the q-line in the Ryan graph is the McCabe-
 Fractionators 65

Thiele slope minus 1. Therefore for bubble point equipment problems, prior to running new cases. In short,
feed the q-line is vertical for the conventional work first with what you already have.
McCabe-Thiele and Ryan. For dew point feed the To effectively analyze computer simulation results, one
slope is 0 for the conventional McCabe-Thiele and needs a basic understanding of the inner workings of frac-
-1 for Ryan. tionation systems. Kiste8 provides these basic needs prior
4. The familiar steps for the theoretical stages on the to discussing the analysis tools themselves and we recom-
McCabe-Thiele diagram are modified on the Ryan mend his book. Kister is a proponent of graphical analysis
plot. The vertical portions remain vertical, but the methods and uses them in his consulting work. Examples
horizontal portions become slanted with a slope of are discussed in this book in section 22, “Troubleshooting.”
-1.00 (see Figure 2). The subsection, “Fractionation: Mechanical Problems,”
has three field examples of the use of graphical methods
titled Pinch, Key Components, and Profiles. Ryan3 is
Computer Simulation Analysis another proponent of these methods and has supplied this
author with several field consulting example diagrams.
Yes, computer simulations need to be analyzed. This is By using the graphical methods, we are allowed to
the case for many kinds of simulation but is discussed “see” into the proposed or existing equipment. The analy-
here for distillation. Computers aid us with speedy and sis procedure presented here improves upon the hand-
accurate math for a particular input case. Optimization generated shortcut methods by starting from a computer
and workability of the equipment are left for human inter- solution. Two McCabe-Thiele methods are presented
vention. Analyze each computer case carefully to under- by Kister and are summarized in the following tabula-
stand its results before making numerous runs. A design tion. Kister prefers Method 1, the simpler of the two.
case can be analyzed when troubleshooting existing Hengstebeck4 developed Method 2.

Multicomponent

Method 1 Method 2
Pseudo Components Equivalent Binary
Find key components. Find key components.
Lump all light non-keys and Add components having volatilities
the light key into a single similar to a key that end up in
light pseudo-component. A the same product to that key. Then
similar procedure is applied flows and compositions are based
to the heavy pseudo-component. upon the equivalent binary.
Binary _________ + Find q Find q
Calculate q by the difference q = (FL -k LzXLNK,lim - LzXHNK,lirn)/Fe
in liquid leaving the feed stage
minus liquid entering the feed Where: FL = Feed liquid, mols/hr
stage divided by the feed rate. F, = Feed light + heavy keys, mols/hr
L = Rectification section liquid,
.1
c) ----------- mols/hr
L‘ = Stripping section liquid, mols/hr
1 XLNK,lim = Sum Of light non-key limiting
liquid composition’s average
value for the middle of the recti-
fication section
66 Rules of Thumb for Chemical Engineers

XmK,lim = Sum of heavy non-key limiting

liquid composition’s average
value for the middle of the strip-
ping section
Each XLNK,lim = [YLNK,D(D/L)~/[(KLNK/KHK)- 1+01
Each XHNK,lim = [XHNK,B(B/L )]/[1.0 - (KHNK~KLK)]

Where: X, Y = Liquid, vapor mol fractions

D,B = Distillate, bottoms flows, mols/hr
K = Equilibrium K-value = Y/X
The methods converge at this point.
Plot the q-line. The q-line intersects the 45’ line at V’ =Vapor flow rate in the stripping section,
the mole fraction of the light key in the feed. Its slope is mols/hr
d(9- 1>. D = Distillate flow rate, mols/hr
B = Bottoms flow rate, mols/hr
Plot the equilibrium curve. The equilibrium curve X = Mole fraction of the light key component in
can be plotted from the y and x compositions of the light the liquid
key (binary, binary equivalent, or pseudo) component in Y = Mole fraction of the light key component in
the vapor and liquid leaving each stage: the vapor
Subscripts refer to:
T n = Rectification section stage number
m = Stripping section stage number
D = Distillate
XU B = Bottoms
Use Y = Step off stages. Use the usual technique.
1+(a- l)X
where Y = Mole fraction of the light component
in the vapor Using the McCabe-Thiele Plot
X = Mole fraction of the light component
in the liquid Detect pinched regions. Pinching occurs when the
a = Relative volatility operating lines approach the equilibrium curve. Minimum
reflux can draw the operating lines to the equilibrium
Plot the operating (component balance) lines. The curve. A feed too high in the column (not at the intersec-
operating lines can be plotted from the vapor composi- tion of the q-line and the equilibrium curve for a binary
tion entering a stage versus the liquid leaving. These distillation) requires extending the stripping section oper-
“lines” often turn out to be curves. ating line beyond the optimum q-line intersection point.
This gives a pinch at the mislocated feed point. A similar
type pinch can occur with a feed too low in the column.
T3. A “tangent pinch” can occur when an equilibrium curve
Use Y,+l = (L/V)(X,) + (D/V)(XD) for the rectification doesn’t have the smooth concave downward shape that
section and we picture in a “well behaved” system. Such a curve
Y m = (L’/V’)(Xm-l) - (B/V’)(XB) for the stripping might dip to almost touch an operating line. This “tangent
section pinch” probably will not occur at the feed point for binary
Where: systems like the other pinch types.
L = Liquid flow rate in the rectification section, Identify excessive traffic. Over-refluxing is indicted
mols/hr by too large a gap between both operating lines and the
L’ =Liquid flow rate in the stripping section, equilibrium curve.
mols/hr Find the chances for heat optimization. When only
V =Vapor flow rate in the rectification section, one operating line has a wide gap, intermediate heating or
mols/hr cooling may be helpful. A large gap for the bottom section
 Fractionators 67

could indicate potential for feed preheat or an interre- Kister says that “d/b plots are primarily used when
boiler. A large gap for the top section could indicate that a there is a tight spec. on a nonkey component or a concern
precooler or intercondenser would be attractive. about the distribution of an intermediate key component.”
Aid optimization. The graphics can show the effects His book shows d/b curves for various feed stage loca-
of parameter changes such as compositions, feed thermal tions on a plot of the mole ratio of a reference component
condition, and multiple feed or draws. in the distillate to the bottom product, versus the relative
Identify mislocated feed. For binary distillation the volatility of each component to this reference component.
feed point should be where the q-line intersects the equi- This plot is made on log-log paper. The optimum feed
librium curve. For multicomponent distillation this may produced a curve closest to linear. The d/b plot is
or may not be the case. So for multicomponent distilla- suggested as a troubleshooting tool in the subsection of
tion feed location, key ratio plots and cL’b plots are pre- the “Troubleshooting” section, “Fractionation: Operating
ferred and discussed next. Problems.”

Multicomponent Feed location

Sources
Key ratio plots are used to identify mislocated feed
stages for multicomponent distillation. When the mole 1. McCabe, W. L. and E. W. Thiele, Ind. Eng. Chem. 17,
ratio of the keys is plotted on the y-axis with a semilog 605, 1925.
scale versus the theoretical stage number on the x-axis 2. Kister, H. Z., Distillation-Design, McGraw-Hill, Inc.,
with a linear scale, a smooth curve can result at high reflux 1992.
ratios. As reflux is reduced, a larger and larger inflection 3. Ryan, J. M., Ryan Consulting, Inc. (jamesryancon
point is observed. The best feed stage falls within this [email protected]) “Replotting the McCabe-Thiele
inflection zone. Kister’s book has a detailed description of Diagram”, Chemical Engineering, May, 200 1, p. 109.
this procedure with plots for several combinations of the- 4. Hengstebeck, R. J., Distillation-Principles and Design
oretical stages, reflux ratios, and feed locations. Practices, Reinhold Publishing, 1961.

Tray Efficiency
Actual stages depend upon the tray efficiency, which Table 1
will probably be the weakest number in the design. Using Fractionator Overall Tray Efficiency, %
operating data from a similar system is certainly best
Gunness and
where possible. Tables 1 and 2 give some shortcut Other Data Plotted Versus Drickamer and Bradford
correlations. Reciprocal Viscosity Correlation Plotted
Ludwig’ discusses new work by the A.1.Ch.E. which in Maxwell. in Ludwig.
Viscosity (Average Viscosity of (Molal average
has produced a method more detailed than the previous Centipoises liquid on the plates) viscosity of the feed)
shortcut methods. He states that some of the shortcut
methods can be off by 15-50% as indicated by the 0.05* ... 98
0.10 104” 79
A.1.Ch.E. work. The spread of the Drickamer and 0.15 86 70
Bradford correlation shown in the Ludwig plot is about 0.20 76 60
10 points of efficiency or f 5 efficiency points around the 0.30 63 50
0.40 56 42
curve. Ludwig states that comparisons between shortcut 0.50 50 36
and A.1.Ch.E. values indicate that deviations for the short- 0.60 46 31
cut methods are usually on the safe or low side. 0.70 43 27
0.80 40 23
Maxwell’s3 correlation was generated from hydro- 0.90 38 19
carbon data only. Ludwig states that the Drickamer and 1.oo 36 17
Bradford correlation is good for hydrocarbons, chlori- 1.50 30 7
1.70 28 5
nated hydrocarbons, glycols, glycerine and related
compounds, and some rich hydrocarbon absorbers and *Extrapolated slightly
strippers. **Maxwell explains how efficiencies above 700% are quite possible.
1.o

0.9

Y
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 .o
X
LIVE GRAPH
Click here to view Figure 1. McCabe-Thiele Diagram.
0.06
DIBSP ITTER EQUILIBF UM

9
0.05

Y-x 17

0.04

0.03

0.02

0.01

0.00
0.0 0.1 0.2 0.3 0.4 X 0.5 0.6
LIVE GRAPH
Click here to view Figure 2. Revised McCabe-Thiele Diagram.
70 Rules of Thumb for Chemical Engineers

Table 2 Ludwig also presents correlations of O ’ C ~ n n e l l He

.~
Overall Tray Efficiency, % (O’Connell) warns that O’Connell may give high results, Ludwig sug-
Absorbers gests using the O’Connell absorber correlation only in
tt areas where it gives a lower efficiency than the fraction-
ator correlation of Drickamer and Bradford or O’Connell.
Correlating Factor Fractionatom*
For high values of a, low tray efficiency usually
0.01 .. 8 results.
0.05 .. 17
0.1 0 87 22
0.1 5 80 23
0.20 74 26
Sources
0.30 67 29
0.40 62 32 1. Ludwig, E. E., Applied Process Design For Chemical
0.50 57 33 and Petrochemical Plants, Vol. 2, Gulf Publishing Co.,
0.60 55 35
0.70 52 36 1979.
0.80 51 37 2. Drickamer, H. G. and Bradford, J. B., “Overall Plate
0.90 49 38 Efficiency of Commercial Hydrocarbon Fractionating
1 48 39
1.5 44 42 Columns,” Trans. A.1Ch.E. 39, 319, (1943).
2 41 45 3. Maxwell, J. B., Data Book on Hydrocarbons, Van
3 37 48 Nostrand, 1965.
4 35 52
5 33 53
4. O’Connell, H. E., “Plate Efficiency of Fractionating
6 32 56 Columns and Absorbers,” Trans. A.I. Ch.E. 42, 74 1,
7 32 57 (1 946).
8 31 58
9 .. 60
~
10 .. 61

*Fractionators = (Relative volatility of key components) (Viscosity of

feed in centipoises), Relative volatility and viscosity are
taken at average tower conditions between top and
bottom.
**H = Henry’s law constant, molslft (ATM).
P = Pressure, ATM
p = Viscosity, centipoises.
p = Density, lbslft’.
K = Equilibrium K of key components.
M = Mol. wf. of liquid.
 Fractionators 71

Diameter of Bubble Caw Travs

Here are two quick approximation methods for bubble The developed equation for the Souders-Brown C
cap tray column diameter. factor is

C = (36.71+5.456T-0.08486T2)ln S-312.9+37.62T
Standard Velocity
- 0.5269T2
The standard velocity based on total cross section is
defined as: Correlation ranges are:

C = 0 to 700
s = 0.1 to 100
The standard velocity is multiplied by a factor depend- T = 18 to 36
ing upon the service as follows:
Service YOUSTO
The top, bottom, and feed sections of the column can
be checked to see which gives the maximum diameter.
Vacuum towers 110* The Souders-Brown correlation considers entrainment
Refinery group** 1oo*
Debutanizersor other 100-250 PSI0 towers 80 as the controlling factor. For high liquid loading situations
DeproDanizers or other high Dressure towers 60 and final design, complete tray hydraulic calculations are
*Applies for 24” tray spacing and above. required.
**“Refinerygroup” = low pressure naphtha fractionators, gasoline split- Ludwig2states that the Souders-Brown method appears
ters, crude flash towers, etc. to be conservative for a pressure range of 5 to 25Opsig
allowing W to be multiplied by 1.05 to 1.15 if judgment
Souders-Brown‘ and caution are exercised.
The Souders-Brown correlation is shown in Figure 1.
The maximum allowable mass velocity for the total
column cross section is calculated as follows: Nomenclature

C = Souders-Brown maximum allowable mass veloc-

ity factor
The value of W is intended for general application and to SsTD= Standard maximum allowable velocity for tower
be multiplied by factors for specific applications as cross section, ft/sec
follows: W = Souders-Brown maximum allowable mass veloc-
ity, lbs/(ft2 cross section) (hr)
Absorbers: 0.55 pL = Liquid density, lbs/ft3
Fractionating section of absorber oil stripper: 0.80 pv = Vapor density, lbs/ft3
Petroleum column: 0.95 T = Tray spacing, in.
Stabilizer or stripper: 1.15 S = Surface tension, dyneslcm
72 Rules of Thumb for Chemical Engineers

-
,

/’/’ r
/
/
/ 9-’
.
200 /
/
.‘ / 4’ /.
/’
A

I-
/#

/7/
1

// d

7
/*
+ 9’
I)

LIVE GRAPH
Click here to view

Sources

1. Souders, M. and Brown, G. “Design of Fractionating

Columns-I Entrainment and Capacity,” Ind. Eng.
Chem., 26, 98, (1934).
2. Ludwig, E. E., Applied Process Design for Chemical
and Petrochemical Plants, Vol. 2, Gulf Publishing Co.,
1979.

Diameter of SieveNalve Trays (F Factor)

The F factor is used in the expression U = F/(p,)’.’ to F = (547 - 173.2T + 2.3194T2)10-6P+ 0.32 +0.0847T
obtain the allowable superficial vapor velocity based on - 0.000787T2
free column cross-sectional area (total column area minus
the downcomer area). For foaming systems, the F factor
should be multiplied by 0.75. Correlation ranges are:
The F factor correlation from Frank’ is shown in Figure
1. F = 0.8 to 2.4
The author has developed an equation for the F factor P = 0 to 220
as follows: T = 18 to 36
 Fractionators 73

2.4 The author has developed the following equation to fit

the correlation:
2.2
2.0 DL = 6.667T + 16.665
1.8
Clear liquid velocity (ft/sec) through the downcomer is
I .6
Lc then found by multiplying DL by 0.00223. The correla-
1.4 tion is not valid if pL - pv is less than 30 lb/ft3 (very high
1.L pressure systems). For foaming systems, DL should be
1.o
multiplied by 0.7. Frank recommends segmental down-
comers of at least 5% of total column cross-sectional area,
0.8 regardless of the area obtained by this correlation.
0 20 40 60 80 100 120 140 160 180 200 220
Column pressure, psia For final design, complete tray hydraulic calculations
are required.
Figure 1. F Factor as a function of column pressure For even faster estimates, the following rough F factor
LIVE GRAPH drop and tray spacing. guidelines have been proposed:
Click here to view
Situation F factor
For estimating downcomer area, Frank gives Figure Fractionating column total cross section vapor velocity 1G 1 . 5
2. Sieve tray hole velocity to avoid weeping >12
Disengaging equipment for separating liquid droplets
from vapor c6
250
Uotes
1. Clear liquid velocity through downcomer
(ftls) = 0.00223 X design liquid

/
traffic Igprnift?) .

.
:
E
200 2. Not valid for very high pressure
systems ( p L - p v < 30 Ib/ft3).
3. For foaming systems. multiply
&i clear liquid rate (ordinate)
Nomenclature
u’
._
c

?, 150 DL = Design downcomer liquid traffic, gpm/ft2

0
.-
3 F = Factor for fractionation allowable velocity
-
.-P
P = Column pressure, psia
m
C
.-
2 100 T = Tray spacing, in.
n
pL= Liquid density, lb/ft3
pv = Vapor density, lb/ft3
50
10 20 30 L
Source
Tray spacing, in

Figure 2. Estimation of downcomer area for a tray-type 1. Frank, O., “Shortcuts for Distillation Design,” Chem-
distillation column. ical Engineering, March 14, 1977.

Diameter of Sieve/Valve Trays (Smith)

Smith uses settling height as a correlating factor which of settling heights from 2 to 30 inches. Here, U is
is intended for use with various tray types. The correla- the vapor velocity above the tray not occupied by
tion is shown in Figure 1. Curves are drawn for a range downcomers.
74 Rules of Thumb for Chemical Engineers

The developed equations for the curves are (Y sub- Smith recommends obtaining the settling height
script is settling height in inches): (tray spacing minus clear liquid depth) by applying the
familiar Francis Weir formula. For our purposes of
Y = A + BX + CX2 + DX3 rapidly checking column diameter, a faster approach is
needed.
A B C D For applications having 24-in. tray spacing, the author
y30 -1.681 97 -.67671 -.129274 -.0046903 has observed that use of a settling height of 18 inches is
y24 -1.77525 -56550 -.083071 +.0005644 good enough for rough checking. The calculation yields
y22 -1 .a9712 -59868 -.080237 +.0025895 a superficial vapor velocity that applies to the tower cross
y20 -1 96316 -5571 1 -.071129 +.0024613
yl6 -2.02348 -54666 -.067666 +.0032962 section not occupied by downcomers. A downcomer area
y1 6 -2.1 9189 -51 473 -.045937 +.0070182 of 10% of column area is minimum except for special
y14 -2.32803 -.44885 -.014551 +.0113270 cases of low liquid loading.
y12 -2.47561 -.48791 -.041355 +.0067033
YlO -2.66470 -.48409 -.040218 +.0064914 For high liquid loading situations and final design,
y6 -2.78979 -.43728 -.030204 +.0071053 complete tray hydraulic calculations are required.
y6 -2.96224 -.42211 -.030618 +.0056176
y4 -3.08589 -.38911 +.003062 +.0122267
y2 -3.22975 -.37070 -.000118 +.0110772

0.03 0.04 0.05 Ob6

0 60 0 60
0.50 0.50

0.40 040

0.30 0.30

Q 0.20 0.20
3
a
I
4
Q
d
\

U
a
c
0 0.10 V.IV
tn

0.08 0.08

0.06 0.06
0.05 0.05

0.04 0.04

0.03 0.03
0.02 0.03 0.04 0.05 0.b6 0.06 0.10 0.2 0.3 0.4 0.5 0.6 0.8 1.0
LIVE GRAPH ( L / G ) ( ~ , /p I I / *
Click here to view
Figure 1. The vapor velocity function is plotted against a capacity factor with parameters of settling heights.
 Fractionators 75

Nomenclature Sources

G = Vapor rate, lb/hr 1. Smith, R., Dresser, T., and Ohlswager, S., “Tower
L = Liquid rate, lb/hr Capacity Rating Ignores Trays,’’ Hydrocarbon Pro-
U = Superficial vapor velocity above the tray not occu- cessing and Petroleum ReJiner, May 1963.
pied by downcomers, ft/sec 2. The equations were generated using FLEXCURV V.2,
pL = Liquid density, lb/ft3 Gulf Publishing Co.
pv = Vapor density, lb/ft3

Diameter of Sieve/Valve Trays (Lieberman)

Lieberman gives two rules of thumb for troubleshoot- jetting factor = U2pv/pL
ing fractionators that could also be used as checks on a
design. First, the pressure drops across a section of trays Where:
must not exceed 22% of the space between the tray decks,
to avoid incipient flood. Mathematically, hold U = Hole vapor velocity, ft/sec
pL = Liquid density
P/(sG)(T,,)(T,)< 22%
pv = Vapor density
where
For a 15-inch spray height, a tray spacing of at least 21
P = Pressure drop in inches of water
inches is recommended.
SG = Specific gravity of the liquid on the tray at the
appropriate temperature
T, = Number of trays
Source
T, = Tray spacing, in.
For sieve trays, a spray height of 15 inches is obtained Lieberman, N. P., Process Design For Reliable Opera-
when the jetting factor is 6-7. tions, 2nd Ed., Gulf Publishing Co., 1989.

Diameter of Ballast Trays

Calculation/Procedure for Ballast Tray Minimum Table 1
Tower Diameter System Factors’

Service System Factor

The following method will give quick approximate
results, but for complete detailed rating, use the Glitsch Non-foaming, regular systems 1.oo
Fluorine systems, e.g., BF3., Freon .90
Manual.’ Moderate foaming, e.g., oil absorbers,
amine and glycol regenerators .85
1. Determine vapor capacity factor CAF. .73
Heavy foaming, e.g., amine and glycol absorbers
CAF = CAF, x System Factor Severe foaming, e.g., MEK units .60
Foam-stable svstems, e.g., caustic regenerators .30-.60
The proper system factor can be found in Table 1.
2. For pv < 0.17 use
36 & 48“ TS: CAF, = (0.579 - 3.29/TS)
CAF, = (TS)0’65~(pV)1’6/12 - [(0.415TS x pv
For pv > 0.17 use - TS)/1560] - 0 . 0 0 9 5 ~ ~
12-30”TS: CAFo =(0.58-3.29/(TS)- 3. The capacity factor CAF, increases with increasing
[(TS x pv -TS)/l560] tray spacing up to a limiting value. Limits reached
76 Rules of Thumb for Chemical Engineers

below 48’’ spacing are not experienced until the 7 . Calculate flow path length.
vapor density exceeds 4 l b ~ / f t The
. ~ author devel-
oped an equation to determine the limiting CAF, FPL = 9 x DTA/NP
value for vapor densities above 4 l b ~ / f t It
. ~should
be used for situations having 18” tray spacing and 8. Calculate minimum active area.
above.
VIoad + (GPM X FPL/13,000)
AAM =
CAFo(limiting)= 0.648 - 0.0492~“ CAF x FF

Absorbers and strippers frequently operate with A flood factor of .65 to .75 should be used for
a liquid having essentially the same physical char- column diameters under 36” to compensate for wall
acteristics regardless of the pressure. An example of effects. Larger columns are typically designed for
this is a gas absorber. The same lean oil is used if about 80% of flood.
the tower is operating at 100 or 1,OOOpsi. This type
of system is excluded from the CAF, limiting value. 9. Obtain downcomer design velocity, VDdsg.Use the
smallest value from these three equations
4. Calculate Vload
VDdsg= 250 x system factor
Vload =C F S d m VDdsg= 4 ldp, - pv x system factor
5. Using Vload and GPM (column liquid loading in VDdsg= 4735x d G x system factor
gallons per minute), obtain approximate tower
diameter for calculating flow path length. Use where

(DTA)2 = 3.025A+ 0.0012AB+ 7.647 x 10-’AB2 VDdsg= Design velocity,GPM/ft2

+ 2.8 1x 10-4 ~ 1 , 7 1 9 For the system factor, use values out of Table 1
where except for the last item in that table (foam-stable
DTA = Approximate tower diameter, ft systems) use 0.30.
A = Vload 10. Calculate downcomer minimum area.
B = GPM
This is the author’s correlation of a nomograph
in the Glitsch Manual. It gives results within 5% of
the nomograph for diameters 4 feet or greater and where
within 15% for smaller diameters. This is adequate
ADM = Minimum downcomer area, ft2
for this first approximation of tower diameter. It
applies for If the downcomer area calculated above is less
a. Single pass trays than 11% of the active area, use the smaller of
b. 24“ tray spacing at 80% flood
C. DT = 2’-10’
ADM = Double that of the equation
d. Vload= 0-30
ADM = 11% of the active area
e. GPM = 0-1,500
6. For 2-pass trays 11. Calculate minimum column cross-sectional area.
a. Divide Vload by 2 Use the larger of
b. Divide GPM by 2
c. Obtain diameter from single pass equation ATM = AAM + 2 x ADM
d. Multiply diameter by 4 Vload
ATM =
For 4-pass trays, replace the 2’s by 4’s. 0.78 x CAF x FF
 Fractionators 77

where Nomenclature
ATM = Minimum column cross-sectional area, ft2.
AA = Active area, ft2
Further detailed design calculations may
AAM = Minimum active area, ft2
result in a change in tower diameter.
AD = Downcomer area, ft2
12. Calculate column minimum diameter. ADM = Minimum downcomer area, ft2
ATM = Minimum column cross-sectional area, ft2
DT = dATM/0.7854
CAF = Vapor capacity factor
CAF, = Flood capacity factor at zero liquid load
Percentage of Flood for Existing Ballast Tray Columns CFS = Vapor rate, actual ft3/sec
DT = Tower diameter, ft
VIoad +(GPM X FPL/13,000) DTA = Approximate tower diameter, ft
%Flood =
AA x CAF FF = Flood factor or design percent of flood, fractional
FPL = Tray flow path length, in.
Minimum diameter for multipass trays is given in GPM = Column liquid loading, gal/min
Table 2. NP = Tray number of flow paths or passes
Holding GPM/WFP below about 8 is preferred, TS = Tray spacing, in.
although liquid as high as 20 GPM/WFP can and have VDdss= Downcomer design velocity, GPM/ft2
been used. WFP is the width of the flow path in inches. Vload = Column vapor load factor
Some companies like to use trays having no more than WFP = Width of tray flow path, in.
two passes. pL = Liquid density, lbs/ft3
pv = Vapor density, lbs/ft3
Table 2
Minimum Practical Diameter for Multipass Ballast Trays

Multipass Tray Minimum Diameter Preferred Diameter Sources

Number of Passes (W (tt)
2 5 6 1. Glitsch Ballast Tray Design Manual, 5th Ed. Bulletin
3 8 9 No. 4900, Copyright 1974, Glitsch Inc.
4 10 12 2. Branan, C, R. The Fractionator Analysis Pocket Hand-
5 13 15
book, Gulf Publishing Co., 1978.

Diameter of Fractionators, General

For checking designs one can roughly relate tower Where
diameter to reboiler duty as follows:
D = Tower diameter, ft
Situtation Reboiler Duty, MMlBTUlhr
Source
Pressure distillation 0.5D2
Atmospheric pressure distillation 0.3D2 Branan, C. R., The Process Engineer S Pocket Handbook,
Vacuum distillation 0.15D2
Vol. 1, Gulf Publishing Co., 1976.

Control Schemes
A good understanding of control schemes is essential to discuss all possible control hookups. The following
for understanding fractionation systems. It is not possible discussions will therefore include many of the major
78 Rules of Thumb for Chemical Engineers

means of control and attempt to provide enough control times control performance, otherwise acceptable, may
philosophy to allow custom design or analysis for indi- interact with measurements to inhibit accurate accounting.
vidual circumstances. The discussions will be limited to All results besides (1) and ( 2 ) are, however, usually of
conventional instrumentation since computer control is more distant concern. Primarily emphasis on understand-
beyond the scope of this handbook. ing and operating controls should be placed on those two
items. It will sometimes be found that no pressure or
Column Control Introduction by Keith Claiborne‘ temperature controls are needed if good, quick sensors
are available to show what overall split is being made
For orientation in trying to control fractionators, it is and what separation is occurring.
well to emphasize the territory within which fractiona-
tors, and therefore control, move. Pressure Control
Fractionators produce two results only: (1) stream
splitting, with so many pounds going out one “end” and Pressure has been classically controlled at a fixed
all other feed pounds going out the other and ( 2 ) compo- value in fractionating columns. Shinskey2 and Smith/
nent segregation toward one or the other of the product Brodmann3 have discussed variable pressure control.
streams, characterized by the Fenske ratio: The author has also been told of unpublished successful
variable pressure applications. Only the classical pressure
mols It in O.H. mols hvy in Btm control will be discussed here.
Y
mols It in Btm
I\

mols hvy in O.H. For partial condenser systems, the pressure can be con-
trolled by manipulating vapor product or a noncondensi-
Result (1) is achieved by control of product flow from ble vent stream. This gives excellent pressure control. To
one end of the fractionator. Result ( 2 ) is achieved by have a constant top vapor product composition, the con-
control of the heat load on the fractionator. denser outlet temperature also needs to be controlled. For
By a combination of the two control handles, one can a total condenser system, a butterfly valve in the column
affect component specification in the product streams, but overhead vapor line to the condenser has been used.
there is always some maximum possible extent of sepa- Varying the condenser cooling by various means such as
ration (value of Fenske ratio) in a given system. Nearly manipulation of coolant flow is also common.
all control measures are designed to permit control of the Pressure can also be controlled by variable heat trans-
two handles. It should be realized that column operation fer coefficient in the condenser. In this type of control, the
needs to be kept reasonably smooth, otherwise separation condenser must have excess surface. This excess surface
already achieved may in the following five minutes (or becomes part of the control system. One example of this
six hours) partially be undone by surging in the system. is a total condenser with the accumulator running full and
As total and component rates vary in system feed, it the level up in the condenser. If the pressure is too high,
becomes necessary to vary controlled parameters to main- the level is lowered to provide additional cooling, and
tain desired properties in product streams. There may also vice versa. This works on the principle of a slow moving
be subsidiary constraints. If column pressure gets too liquid film having poorer heat transfer than a condensing
high, the feed system may be unable to input feed vapor film. Sometimes it is necessary to put a partially
or allowable design pressures of equipment may be flooded condenser at a steep angle rather than horizontal
exceeded. With too low a pressure, product may not flow for proper control response.
from the system. Too high a bottom temperature may Another example of pressure control by variable heat
cause excesses over equipment design or product degra- transfer coefficient is a vacuum condenser. The vacuum
dation. Subsidiary constraints can produce “rapid” varia- system pulls the inerts out through a vent. The control
tions in flows or pressures. Even when these variations valve between the condenser and vacuum system varies
are not excessive for a given fractionator, they may the amount of inerts leaving the condenser. If the pres-
impose impossible control problems on equipment sure gets too high, the control valve opens to pull out
connected in a larger system with that fractionator- more inerts and produce a smaller tube area blanketed by
upstream, downstream, or laterally. inerts. Since relatively stagnant inerts have poorer heat
In the long term, the best understanding and optimiza- transfer than condensing vapors, additional inerts
tion of a system will usually require attention to attaining withdrawal will increase cooling and lower pressure. The
good heat and component balances in operation. Some- reverse happens if the pressure gets too low.
 Fractionators 79

For vacuum applications having column temperature since psv’s can develop leaks if forced to open and reseat
control above the feed point, put the measuring elements frequent1y.
for the temperature and pressure controllers close The fix for the erratic reflux drum pressure problem
together. This will make for good composition control at was to provide for separate pressure control of the frac-
varying column loads (varying column differential pres- tionator column and the reflux drum. A new pressure
sure). In vacuum systems, a slight pressure change will control valve was installed upstream of the condenser
produce large equilibrium temperature changes. and the old condenser outlet control valve was removed.
Another pressure control method is a hot gas A hot gas bypass, designed for 20% vapor flow, was
bypass. The author has seen a retrofitted hot gas installed around the pressure control valve and condenser.
bypass system solve an unsatisfactory pressure control A control valve was installed in the hot gas bypass line.
scheme. In the faulty system, a pressure control valve The column pressure was then maintained by throttling
was placed between the overhead condenser and the new control valve upstream of the condenser. The
the lower reflux drum. This produced erratic reflux reflux drum pressure was controlled by the hot gas bypass
drum pressures. A pressure safety valve (psv) vented control valve and the “psv saver” working in split range.
the reflux drum to the flare to protect against over- The new system is shown in the figure below.
pressure. A “psv saver” control valve, set slightly below Chin4 states that a scheme having only a control valve
the psv relieving pressure, kept the psv from having in the hot gas bypass line manipulated by the column
to operate very often. The “psv saver” is a good idea pressure controller will not work for the case of zero net

REFLUX DRUM
80 Rules of Thumb for Chemical Engineers

vapor product. He cites an actual plant case and offers his Column Bottom Temperature. The bottom temperature
analysis of the reasons. Chin’s article shows 21 different is often controlled on the reboiler outlet line with a
distillation control methods with advantages and dis- control valve in the heating medium line. The control
advantages for each. point can also be on a bottom section tray. Care must be
exercised in location of the temperature control point. It
is recommended, especially for large columns, that a
Temperature Control cascade arrangement be used. The recommended scheme
has a complete flow recorder/controller (FRC) in the
General. With simple instrumentation discussed here, heating medium line including orifice and control valve.
it is not possible to satisfactorily control the temperature The set point of this FRC is manipulated by the temper-
at both ends of a fractionation column. Therefore, the ature recorder/controller (TRC). This eliminates the TRC
temperature is controlled either in the top or bottom from manipulating the control valve directly (recall that
section, depending upon which product specification is temperature is the most difficult parameter to control).
the most important. For refinery or gas plant distillation This makes for smoother control for normal operations.
where extremely sharp cut points are probably not Also, it is handy for startup to be able to uncouple the
required, the temperature on the top of the column or the TRC and run the reboiler on FRC for a period.
bottom is often controlled. For high purity operation, the For a fired reboiler, a pump-around system is used with
temperature will possibly be controlled at an intermedi- an FRC to maintain constant flow. There will be a low
ate point in the column. The point where AT/AC is flow alarm plus fuel shutoff. There will also be a high
maximum is generally the best place to control tempera- flow alarm plus fuel shutoff, since a tube rupture would
ture. Here, AT/AC is the rate of change of temperature reflect itself in a high flow.
with concentration of a key component. Control of tem-
Feed Temperature. This is often allowed to vary. Often
perature or vapor pressure is essentially the same. Manual
a feed-to-bottoms cross exchanger is used for preheat.
set point adjustments are then made to hold the product
Often the amount of heat available in the bottoms is close
at the other end of the column within a desired purity
to the optimum feed preheat. Care must be taken not to
range. The technology does exist, however, to auto-
heat the feed too hot and put inordinate vapor into the top
matically control the purity of both products.
section of the column. This will be detrimental to rectifi-
Temperature is the hardest parameter to control in a
cation. Feed preheat can be controlled by manipulating
fractionation system. It exhibits high process and mea-
heating medium or by bypass for a cross exchanger. In
surement lag. Temperature can also be ambivalent as a
deciding whether or not to control feed temperature, the
measure of composition. Pressure changes are reflected
designer sometimes has to choose between column sta-
quickly up and down the column. Temperature changes
bility versus energy recovery similarly to the case of con-
are not. It is typical to provide three-mode controllers for
stant versus variable pressure control. Feed temperature
all temperature applications.
control becomes more critical as its amount of liquid and
vapor increases relative to that of the traffic produced by
Column Top Temperature. This temperature can be con- the reflux and reboiler. For column analysis, it is good to
trolled automatically or manually by manipulating reflux, check on how constantly the feed temperature is running.
depending upon whether the automatic control point is at
the top or bottom of the column. For partial condenser level Control
applications, it is typical to control condenser outlet tem-
perature instead of column top temperature as discussed For a total condenser, accumulator level is typically set
previously. In this case, a variety of methods are avail- by varying distillate draw. For a partial condenser, it can
able, such as manipulating cooling water flow or bypass, be controlled with a condenser hot gas bypass.
or varying louvers or fan pitch on an air-cooled con- The column bottom level is sometimes controlled
denser. Manipulation of cooling water flow has the draw- by bottoms draw. Varying reboiler heating medium is
backs of heat transfer being relatively insensitive to this another possibility. For some cases, bottoms draw level
variable and possible fouling if flow is not held at control works better and for others, heating medium level
maximum. control. Bojnowski’ gives a case where heating medium
 Fractionators 81

level control was desired for two columns in a plant. One common operating problems produce a decrease in dif-
column had to be put on bottoms draw level control, ferential temperature across a section of the column. A
however, when improper reboiler thermosyphoning differential temperature as well as a differential pressure
caused sluggish response with heating medium level recorder across a troublesome tower section would
control. In the case of variable heat content reboiler steam provide invaluable operator assistance.
supply, use heating medium level control with bottoms Differential temperature as well as differential pressure
draw rate setting composition. This provides the best can be used as a primary control variable. In one instance,
bottoms composition control. it was hard to meet purity on a product in a column having
close boiling components. The differential temperature
Flow Control across several bottom section trays was found to be the
key to maintaining purity control. So a column side draw
The feed flow is often not controlled but is rather on level flow higher in the column was put on control by the crit-
control from another column or vessel. The liquid product ical temperature differential. This controlled the liquid
flows (distillate and bottoms) are often on level rather than “reflux” running down to the critical zone by varying the
flow control. Top vapor product is, however, usually on liquid drawn off at the side draw. This novel scheme
pressure control. The reflux is frequently on FRC, but also solved the control problem.
may be on column TRC or accumulator level.
Sources
Differential Pressure Control
1. Branan, C. R., The Fractionator Analysis Pocket
For column analysis and troubleshooting it is impor- Handbook, Gulf Publishing Co., 1978.
tant to have pressure drop measured with a DP cell. The 2. Shinskey, F. G., “Energy-Conserving Control Systems
differential pressure can also be used to control column for Distillation Units,” Chemical Engineering
traffic. A good way to do this would be to let the differ- Progress, May 1976.
ential pressure control the heating medium to the reboiler. 3. Smith, C. L., and Brodmann, M. T., “Process Control:
The largest application for differential pressure control is Trends for the Future,” Chemical Engineering, June
with packed columns where it is desirable to run at 80 to 21, 1976.
100% of flood for best efficiency. 4. Chin, T G., “Guide to Distillation Pressure Control
Methods,” Hydrocarbon Processing, October 1979, p.
Differential Temperature Control 145.
5. Bojnowski, J. J., Groghan, R. M., Jr., and Hoffman, R.
Temperature sensing points at various points along the M., “Direct and Indirect Material Balance Control,”
tower shell are often a useful troubleshooting tool. Many Chemical Engineering Progress, September 1976.

Optimization Techniques
This section deals with optimization of an existing such as reflux, feed temperature, and reboiler heat input.
column. Higher throughput or improved product quality The first step will be to determine by plant testing,
is balanced against higher operating costs such as labor, and calculations to extrapolate the plant data, the ultimate
energy, or maintenance. If the product purity and pro- limiting capacities of all pieces of equipment in the
duction are fixed, then the optimum is the minimum system such as column shell, reboiler, condenser, and
energy to do the set job. Optimization would involve pumps.
holding the reflux at the minimum to deliver the distillate The search for the optimum can involve many case
purity and boilup at the minimum to deliver the bottoms studies. Often, if time is limited, a rigorous canned com-
purity. puter program is utilized. However, this is expensive. If
Often, there is latitude in production rate and purity time is not as much of a factor, the method prevented here
specifications, requiring optimization calculations to will allow the calculations to be handled conveniently by
determine the best set of column operating conditions hand or computer having limited core.
82 Rules of Thumb for Chemical Engineers

Smith-Brinkley Method the effective top and bottom section molar liquid and
vapor rates to determine S, and S,.
The procedure proposed for the optimization work is
the Smith-Brinkley Method'. It is especially good for the
Building a Distillation Base Case
uses described in this section. The more accurately known
the operating parameters, such as tray temperatures
If data is meager, section temperatures can be simply
and internal traffic, the more advantageous the Smith-
the average of the top and feed temperatures for the top
Brinkley Method becomes.
section and bottom and feed temperatures for the bottom
The Smith-Brinkley Method uses two sets of separa-
section. Also, the molar flows can be obtained, assuming
tion factors for the top and bottom parts of the column,
equal molar overflow. However, usually a rigorous com-
in contrast to a single relative volatility for the
puter run is available for a plant case or a design case.
Underwood Method.* The Underwood Method requires
Having this, the effective values can be determined by
knowing the distillate and bottoms compositions to
averaging values for all stages in a section. Some columns
determine the required reflux. The Smith-Brinkley
can be calculated adequately with the simple averages
Method starts with the column parameters and calculates
first discussed, but in others the stage-by-stage averaged
the product compositions. This is a great advantage in
values from a computer run are necessary for satisfactory
building a model for hand or small computer calculations.
results.
Starting with a base case, the Smith-Brinkley Method can
The Smith-Brinkley Method can therefore be used to
be used to calculate the effect of parameter changes on
generate a hand base case beginning with either a heat
the product compositions.
and material balanced plant case, a rigorous computer
solution of a plant case, or computer solution of a design
Distillation case. Once the hand base case is established, alternate
cases can be done by hand (or small computer having
Smith' fully explains the Smith-Brinkley Method and limited core) using the Smith-Brinkley Method.
presents a general equation from which a specialized
equation for distillation, absorption, or extraction can Alternate Distillation Cases
be obtained. The method for distillation columns is
discussed here. Changing feed temperature can be approximated by
For distillation component i, adding one half of the change in feed temperature to the
top and bottom effective temperatures. For a large feed
(1- S,N-M)+ R(l- S,) temperature change, the feed flash and the column heat
f = \ I

(1 - SnN-")+R(l - S,)+ hS,N-M(l- S,"") balance will have to be redone.

Small changes in reflux can be approximated by
where assuming no change in effective temperatures, thus avoid-
ing trial and error calculation. The extra liquid and vapor
fi = (BXB/FXF)i
are added to S, and S,. For larger reflux changes, the extra
S,i = KiV/L
trial and error of determining the new end temperatures
S,i = KIV'/L'
by bubble point and dew point calculations can be
h = Correlating factor; if the feed is mostly liquid, use
included. The effective section temperatures change only
Equation 1 and if mostly vapor, Equation 2
one half as fast as end temperatures, since the feed
-( ' -)( -)1-S,
hi = Ki L temperature remains unchanged.
The preceding discussion on reflux assumes that the
Ki L' 1-S,
condenser is not limiting when the reflux is raised. For a
severely limited condenser, an evaluation must first be
made of the condenser heat transfer before analyzing the
effect of a reflux increase with Smith-Brinkley. Likewise,
The effective top and bottom section temperatures are a limiting reboiler or trays close to flood would have to
used to determine Ki and K'i. These are used along with be evaluated prior to Smith-Brinkley calculations.
 Next Page

Fractionators 83

A small change in reboiler duty can be approximated Then:

by adding the increased traffic to S, and S,. If the
top temperature is automatically controlled, extra reflux
would also result.
In summary, for each alternate operation, an approxi-
mate net effect or set of effects can be quickly built into
In x] ::!!’
= 0.02T]:
the Smith-Brinkley input to determine new end composi-
tions. The instrumentation scheme will affect the result -6.908 + 3.912 = 0.02T; T = -150 or T = 150 minutes
produced by an operating change. whereas: 1 “turnover” = 2,000/40
= 50 minutes
Collecting Field Data

When collecting meaningful field fractionating column Nomenclature

data, the column must not only have constant flow rates,
but the flows must give a good material balance (no accu- B = Bottoms total molar rate or subscript for bottoms
mulation). In addition, a steady state condition must exist F = Feed total molar rate or subscript for feed
for the given flow rates. fi = Ratio of molar rate of component i in the bottoms
To determine if steady state conditions exist, the tem- to that in the feed
peratures and pressures in the column can be tabulated to Ki = Equilibrium constant of component i in the top
assure that they are reasonably unchanging. Laboratory section equals y/x
analyses are usually too slow and expensive for checking K’i = Equilibrium constant of component i in the bottom
lined out conditions. Monitoring reflux accumulator section equals y/x
boiloff is often an effective way of noting concentration L = Effective total molar liquid rate in top section
changes. Simply let a sample of the accumulator L’ = Effective total molar liquid rate in bottom section
liquid boil at atmospheric pressure in a bottle with a M =Total equilibrium stages below the feed stage
thermometer inserted. This method is limited to light including reboiler
hydrocarbons and is not accurate enough for precision N = Total equilibrium stages in the column including
fractionation. reboiler and partial condenser
The accumulator holdup will often be the limiting item R = Actual reflux ratio
setting time required for samples to be representative of S, = Stripping factor for component i in bottom section
the column operating conditions. One “changeout” or S, = Stripping factor for component i in top section
“turnover” of the accumulator may be insufficient time, V = Effective total molar vapor rate in top section
if test conditions are much different from conditions prior V’ = Effective total molar vapor rate in bottom section
to the test. The following example will serve to show this. X = Mol fraction in the liquid
Assume an accumulator having a liquid holdup of Y = Mol fraction in the vapor
2,000gal. Assume that the liquid leaving as reflux plus
distillate is 40GPM. Further assume that the accumula-
tor liquid starts at 2.0 vol % of a key component and that Sources
the test conditions produce 0.0 vol. % of the key compo-
nent. How long will it take for the concentration to reach 1. Smith, B. D., Design of Equilibrium Stage Processes,
0.1 vol. % (the assumed laboratory test accuracy)? McGraw-Hill, 1963.
Establish the following symbols: 2. Underwood, A. J. V., “Fractional Distillation of
Multicomponent Mixtures,” Chemical Engineering
x = Concentration, vol. fraction Progress, 44, 603, (1948).
dx = Differential concentration element, vol. fraction 3. Branan, C. R., The Fractionator Analysis Pocket
dT = Differential time element, min. Handbook, Gulf Publishing Co., 1978.