Column Efficiency What to Expect and Why

Mark Pilling
Sulzer Chemtech, USA



















1999, Mark Pilling
Prepared for Presentation at 4
th
Topical Conference on
Separations Science and Technology, November 1999
Session T1006 Distillation Hardware and Application I
Introduction
This paper will deal with efficiencies of process columns in a practical manner for the average engineer
rather than the distillation expert. Processwise, the focus will be mainly on distillation, although stripping
and absorption processes will also be discussed. With respect to internals, the focus will be on trays,
although references to packing and HETP will be made.

Much research has been done to predict and understand efficiencies in process columns over the years.
However, even with this excellent work, nagging uncertainties still exist regarding what efficiencies can be
expected, especially in processes that are not well proven. Due to the fact that most column designs are a
balance of efficiency, capacity, and energy consumption, an improvement in the certainty of tower
efficiency would lead to improvement in the process design and the design of the internals for the column.

Process Design
This paper will present a rather one-sided view of process design, where all efforts are solely aimed at
producing the optimum tower design for the optimum tower operation. During the design stage, the
process is simulated in order to determine what operations are needed to transform the feedstock into the
required end products. A significant portion of this design will be the evaluation of the towers with respect
to efficiency (number of theoretical stages required) versus energy consumption (reflux rate, stripping rate,
absorption rate). It is at this point where one must decide what efficiency is to be expected for a given set
of internals and how sensitive this efficiency is to energy input. The process engineer must ask, What is a
typical efficiency for these internals with these process conditions? It is at this point where a great deal of
uncertainty can creep into the design of process columns.

The problem with process design simulations and their ability to properly predict the true process
conditions is well documented.
1,2
In particular, it has been demonstrated that calculated efficiencies can
vary substantially when using different process simulation packages with an identical set of process input
data.
3
One of the important values calculated by the process simulation is relative volatility. Kister
4
states
that, Errors in relative volatility are the most underrated factor that affects both tray and packing
efficiency. This is especially true when the relative volatility is low. Therefore, it is critical to have an
accurate simulation of the process when calculating the number of theoretical stages that can be achieved in
a particular design.

Efficiency Nomenclature
Tray efficiencies are generally classified as either overall efficiency (Fenske), point efficiency, or average
tray efficiency (Murphree). The overall efficiency term is quite straightforward. It is the number of actual
stages achieved versus the number of trays in the tower or section of the tower. Point efficiency and
Murphree tray efficiency are similar. They represent the ratio of the actual compositional change and the
theoretical compositional change at equilibrium. The compositional change is usually measured in the
vapor phase but can be measured in the liquid phase. The difference between the point efficiency and
Murphree tray efficiency calculation is the reference point. Point efficiency is measured at a specific point
and the Murphree tray efficiency is measured across a complete tray. Therefore, the compositional
gradients normally found on a tray will affect the Murphree tray efficiency but will not affect the point
efficiency. When the liquid and vapor both have homogeneous compositions, point efficiency and
Murphree tray efficiency will be equal. In practical terms, trays with little or no liquid flow path length
will essentially achieve a point efficiency while trays with conventional flow path will achieve a higher
Murphree tray efficiency due to the compositional gradient of the liquid flowing across the tray deck.

Calculation of Efficiency
When calculating efficiencies, the engineer will typically first calculate the point efficiency and then
modify this calculation to represent an average tray efficiency, which is a more applicable term. When
working with packings, efficiency is usually dealt with in terms of theoretical stages per unit depth of
packing. The commonly used terms are HETP (height equivalent per theoretical plate) and NTSM (number
of theoretical stages per meter).

As far as calculation of efficiency is concerned, most equations can be considered semi-empirical. Most
efficiency correlations in use today are based upon the AIChE model that was published in 1958 in the
AIChE Bubble Tray Manual. This correlation is based on the two-film theory, which accounts for
resistances in both the vapor and liquid phases. For typical distillation systems, the resistance to mass
transfer is highest in the vapor phase. However, an in-depth review of efficiency correlations is beyond the
scope of this paper. For those who would like to pursue this topic, Kister and Lockett have given excellent
reviews of published correlations.
4,5
A brief summary of their findings is as follows:

Generally, the closer a correlation is to actual theory, the more difficult it is to apply to actual column
operation. Conversely, very empirical correlations are easier to apply but risk errors due to the
likelihood of the correlation being applied outside of the original data used to create the correlation.

Of the more theoretical efficiency correlations, the Chan and Fair correlation is the most widely used
and accepted. This is based on a froth regime model and should not be used for spray regimes that are
found at lower pressures.


Empirical Correlations and Data
As far as empirical data and correlations are concerned, there are several sources of information that can be
used. These sources are listed below in order of preference.

1. The best source of data is nearly always a similar process with similar internals. Barring bad field data
or misinterpretation of those data, this is always the most reliable source of information. The engineer
should obtain quality test data from a similar column, simulate the process, and then compare the
predicted efficiency versus the actual efficiency.

2. Other good sources of information are tables found in literature that show typical efficiencies that can
be expected for certain processes. For example, the GPSA Data Book
6
has tables that show typical
efficiencies for trays and packings in gas plants. Table 1 shows typical efficiencies for common
processes. The main problem with this type of information is that it is rather generic. When using this
information, the engineer generally wont know many specifics about the operations that were used to
derive these efficiencies and therefore must make somewhat conservative assumptions when
calculating the efficiency for their particular tower.

3. The next best choice would then be experimental data for a similar process that can be scaled up.
There is always some risk in scaling up data. There are various sources available that tabulate
published experimental efficiency data.
4,7
There are two main problems to be accounted for when
using this type of data. First, these data sets are usually limited to standard test systems that may or
may not have commercial significance. Second, these data have usually been taken with experimental
scale columns. Small column effects and scale up factors must be taken into consideration. However,
if the correct precautions are taken, this can usually be done reliably.

4. Of the more empirical efficiency correlations, the OConnell correlation is the most widely used and
accepted. It is based on various sets of actual plant operating data. It predicts point efficiency as a
function of liquid viscosity and system relative volatility. There is also a modified OConnell plot for
use in absorption systems. In this plot, the efficiency is calculated as a function of the Henrys law
constant, the system pressure, and the liquid viscosity. Even when using one of the more preferred
empirical sources, it is a good idea to check the OConnell correlation to ensure that there is a general
agreement.

Effect of Physical and Transport Properties on Efficiency
Physical and transport properties of the system play a significant role in the efficiency of a particular
process. The OConnell correlation uses liquid viscosity and relative volatility. The AIChE (two-film)
correlation uses vapor and liquid diffusivities. When comparing overall column efficiency to Murphree
tray efficiency, the slope of the equilibrium line, which is a direct function of relative volatility, is used.
Other correlations use a liquid surface tension term in the calculation of efficiencies. Below, each of these
terms is discussed with respect to how they affect efficiency.

Liquid Viscosity: This is the first term in the OConnell correlation. As pointed out by Lockett
5
, the main
influence of the liquid viscosity term is that an increase in liquid viscosity is also usually associated with a
decrease in liquid phase diffusivity. As liquid diffusivity decreases, the tray efficiency will decrease
accordingly. The overall effect is that as liquid viscosity increases, efficiency decreases.

Relative Volatility: This is the other term used in the OConnell correlation. When relative volatility
increases, efficiency decreases. Low relative volatilities can have a significant effect on efficiency. More
importantly, errors in calculating low relative volatilities (<1.5) can create dramatic errors in calculated
efficiency.

Vapor Diffusivity: By its nature vapor diffusivity has a direct influence on efficiency. Increased
diffusivities will increase efficiency. Both vapor and liquid diffusivities are seldom reported and are
difficult to obtain with any degree of certainty. However, when reviewing vapor diffusivity data versus
system pressure, there is a very clear logarithmic relationship between the two. As pressure increases,
vapor diffusivities decrease.

Liquid Diffusivity: Most of the comments above for vapor diffusivity apply also for liquid diffusivity.
Higher liquid diffusivities will result in higher efficiencies. Liquid diffusivities generally increase with
pressure. This will directionally tend to offset the change in vapor diffusivity. However, in most
distillation systems the influence of the vapor diffusivity will dominate the influence of the liquid
diffusivity.

Liquid Surface Tension: There has been some question as to what influence liquid surface tension really
has on efficiency. Chen
8
has published data that confirm that surface tension does have an effect on tray
efficiency. The paper states that surface tension was found to have a significant effect on tray efficiency
through its effect on bubble sizes and interfacial area in froths. The data show that, in the froth regime, a
decrease in surface tension alone will cause a resulting increase in efficiency. However, in high pressure
hydrocarbon systems, the effect of surface tension appears either to be minimal or it is compensated for by
other factors in typical correlations.

Effect of Tray Design and Operation
There are various aspects of equipment design that can affect efficiency. Any time a device can maximize
the vapor/liquid contact while maximizing the compositional approach between the vapor and liquid, that
device will maximize the efficiency of the tower. Conversely, any device characteristics that limit contact
or compositional approach will lessen the efficiency of the tower. Characteristics that may affect efficiency
are discussed below.

Weir Height: With trays operating in the froth regime, an increase in weir height will directionally increase
the efficiency. Kister
4
has noted that the removal of even a small outlet weir can noticeably decrease the
effective tray efficiency. Weir height is especially important in liquid limited systems or systems where a
slow chemical reaction is taking place.

Flow Path Length: Directionally, an increase in flow path length will increase efficiency. This was
discussed earlier in the difference between point efficiency and Murphree tray efficiency. This holds true
unless the length of the flow path creates anomalies in the tray operation such as liquid backmixing or
vapor cross flow channeling.

Liquid and Vapor Maldistribution: As would be expected, vapor and liquid maldistribution will cause
decreases in efficiency. This is especially true when dealing with packings. Generally, maldistribution
problems are generated by the distribution of feeds to the columns rather than by the contacting devices
themselves.

Weeping and Entrainment: Weeping and entrainment will also directionally cause decreases in efficiency.
When considering the effects of weeping, it is important to differentiate between inlet side weeping and
outlet side weeping. With inlet side weeping, the liquid will effectively miss two tray decks and the effects
can be substantial. With outlet side weeping, only a small portion of the deck is missed and the effects
should be minimal. The effects of entrainment are generally more detrimental that weeping. This is
because entrainment carries liquid upward and decreases the vapor-liquid equilibrium approach. Weeping
actually increases the equilibrium approach at the expense of a decrease in contacting time.

General Guidelines and Procedures
Now returning back to the question of process design. A recommended process for estimating efficiencies
would be as follows.

1. Identify your process and simulate it as carefully as possible.

2. Find a similar process application and simulate it with carefully obtained field test data. Obtain the
efficiency from the resulting number of stages and the actual number of trays in the column. Be sure
to account for theoretical stages created by the reboiler and condenser. If actual plant operating data
are not available, look in the open literature to try and find efficiency data that will match closely with
the process. If this is not available, and the process is unknown, pilot plant testing may be a good
option. If none of these options provide an acceptable solution, use an empirical correlation to
calculate the expected efficiency.

3. Depending upon the degree of confidence in the calculated efficiency and the similarity between the
test data and the actual process, take into account the differences between the process and equipment
that was used to calculated the expected efficiency and the actual process that will be installed.
Differences in physical and transport properties must also be accounted for. Differences in equipment
must be evaluated. The more uncertainty there is in these evaluations, the more conservatism should
be used in the final design.

4. Utilize any cross checks or substantiating calculations that may be available. If possible, perform the
more theoretical efficiency calculations to verify that they agree with the proposed design.

5. Design the column with some operating flexibility (degrees of freedom). When reviewing the required
efficiency, it is usually a good idea to review the sensitivity of the product purity to losses of efficiency
in the tower. One way to do this is to construct plot of required stages versus reflux ratio. As an
example, a stages versus reflux ratio plot for a C
3
splitter is shown in Figure 1. The most balanced
design point for this tower would be with about 110 theoretical stages. At this point, the efficiency is
not overly sensitive to variations in reflux ratio and vice-versa. A good design practice is to set the
desired efficiency and then calculate what increase in reflux ratio would be required to offset a 10%
loss in efficiency. This is shown in Figure 2. The next step is to rate the tower internals and auxiliary
equipment to verify that they can accommodate the increased flows and duty that correspond to the
increased reflux rate. If the internals and auxiliary equipment will not handle these increased rates,
then the design may be too inflexible for services where efficiency is not well defined.


Conclusions
Although the concept and calculations of column efficiency can sometimes be quite daunting, there are
methods available to take some of the risk out of the process. Obviously, taking advantage of similar
process designs can very beneficial. The important thing is to carefully consider what you actually know
versus what you dont know or what is uncertain. By using the wide variety of efficiency correlations and
information available along with a large dose of common sense, the process engineer should be able to
provide a functional design for all but the most unknown processes.


Bibliography




1. Schad, R.C., Chemical Engineering Progress, p. 21, January 1998
2. Sowell, R., Hydrocarbon Processing, p. 102, March 1998
3. Yang, N.S., Z.P. Xu, K.T. Chuang, M. R. Resetarits, Same Distillation Column Field Data Simulated
by Different Software Packages Yields Different Observed Efficiencies, paper presented at AIChE
Annual Meeting, November 1997.
4. Kister, H.Z., Distillation Design, McGrawHill, 1992
5. Lockett, M. J., Distillation Tray Fundamentals, Cambridge University Press, 1986
6. GPSA Engineering Data Book, 10
th
Edition, Gas Processors Association, 1987
7. Walas, S. M., Chemical Process Equipment, Butterworth Publishers, 1988
8. Chen, G.X., A. Afacan, K.T. Chuang, The Canadian Journal of Chemical Engineering, p. 614, Vol. 72
August 1994.
Figure 1




Figure 2

Theoretical Stages vs. Reflux Ratio
Constant Product Purities
50
100
150
200
0 5 10 15 20 25 30
Reflux Ratio
T
h
e
o
r
e
t
i
c
a
l

S
t
a
g
e
s
Based on McCabe Thiele
For Commercial C
3
Splitter
Optimized Design
Detail of C
3
Splitter Plot
80
100
120
140
8 10 12 14 16
Reflux Ratio {RJD)
T
h
e
o
r
e
t
i
c
a
l

S
t
a
g
e
s
10% Loss of
Theo. Stages
10% Increase
in RJD
Min. No. of Trays


Table 1


Process Typical Efficiency, %
Demethanizer 50
Deethanizer 60
C
2
Splitter 85
Depropanizer 85
C
3
Splitter 85
C
3
/C
4
Splitter 80
Debutanizer 90
Deisobutanizer 85
Deisopentanizer 85
Dehexanizer 80
Naphtha Splitter 80
Crude Atm Tower:
Overflash/Wash
50
Mid Frac Hvy Naphtha &
Dsl
75
Top Frac Kero and
Naphtha
80
Side Strippers 40
Crude Stabilizer 30
FCC Main Fractionator:
Wash Section Above
Quench
50
Mid Frac HCO/LCO 60
Top Frac LCO HC
Naphtha
75
FCC Absorber 70
Amine Absorber 30
Amine Regenerator 30
Glycol Contactor 30
Glycol Regenerator 30
Sour Water Stripper 40
Alcohol/Water 65
Benzene/Toluene/Xylene 75
EB/Styrene 90