478

Chem. Eng. Technol. 20 (1997) 478-484

Effect of Two-Pass Trays on Distillation Efficiencies

Kimmo T. Klemola, Jarno K. Ilme* Two important structural factors that affect distillation efficiencies, the outlet weir height and the liquid flowpath length, are investigated. Performance and efficiency data of an industrial scale i-butane/n-butane distillation column equipped with two-pass trays are used as a basis for the calculations. A mathematical development for a new method for predicting the numbers of vapor and liquid phase transfer units is given. This method together with some other NTU calculation methods is used to assess the effect of the outlet weir height on efficiencies. The effect of outlet weir height on the Murphree tray efficiencies is investigated using the observed point efficiencies and different point efficiency to the Murphree tray efficiency relation methods. The effect of varying liquid flowpath lengths on efficiencies is studied by calculating the Murphree tray efficiencies for one-pass and twopass trays. The results obtained using the NTU calculation method presented in this study show that a certain outlet weir height point efficiency reaches its maximum. Most of the other methods give opposite results giving a minimum for point efficiency at a certain outlet weir length. The results also show that the Murphree tray efficiencies for one-pass trays are higher than for two-pass trays. Obviously, this is caused by the longer liquid flowpath length of one-pass trays. The Murphree tray efficiencies are also calculated for an industrial-scale MTBE purification column. The column is equipped with two-pass trays in the stripping section and with one-pass trays in the rectifying section. The Murphree tray efficiencies of one-pass trays are considerably higher than the two-pass tray Murphree tray efficiencies.
1 Introduction 2 Efficiency Definitions

The best possible separation on a distillation tray is achieved when the vapor phase and liquid phase are in equilibrium, This is called a theoretical tray or an ideal stage. In practice the equilibrium condition between vapor and liquid is seldom achieved. Efficiencies are used to compare the separation of a real tray to the separation of a theoretical tray. In recent years, the new calculation methods in distillation have become capable of handling industrial-scale distillation columns. Such methods are the rate-based and the matrix models, which can be used to simulate a column with real trays. In both of these methods mass transfer is described using the number of transfer units (NTU) correlations. In the matrix models NTU correlations are used to determine point efficiencies. The parameters used in NTU correlations are the physical properties, the flow rates, and the tray design parameters. Almost all NTU correlations have been developed using a small diameter column data and the correlations have not been verified for industrialscale columns. Point efficiencies are related to the Murphree tray efficiencies using the correlations that take into account the fluid flow phenomena on the tray. One important tray design parameter that affects efficiencies is the liquid flowpath length.

The overall column efficiency is the ratio of the number of theoretical trays to the number of real trays in the column.

lYrc

The Murphree tray efficiency is the ratio of the actual change in the average vapor composition to the change that would occur if the vapor leaving the tray was in equilibrium with the liquid leaving the tray. The Murphree vapor phase tray efficiency is the most commonly used efficiency definition, and it is the only one with practical importance. The Murphree tray efficiency is defined as follows in terms of vapor composition. Em,=
Y n-Yn -1

Yn -Yn-1

where y* = mx, + b and x, is the average composition of the liquid leaving tray n. For a cross-flow tray, the Murphree tray efficiency can exceed 100% due to concentration gradients in liquid. Point efficiency describes the approach to equilibrium at a point on the tray. Point efficiency is not necessarily constant on the tray. Point efficiency is defined as follows. Eov =
Yn -YnY*-Yn-l
0930-7516/97/0709-0478 $17.50+ SO10
1

* K. T. Klemola (corresponding author), Helsinki University of Technology, Department of Chemical Engineering, Kemistintie lM,

02150 Espoo, Finland, J.K. Ilme, Neste Oy, Neste Engineering,

POB 310, FIN-06101 Porvoo, Finland.

(3)

0 WILEY-VCH Verlag GmbH, D-69469 Weinheim, 1997

Chem. Eng. Technol. 20 (1997) 478-484 where y* = mx+ b and x is the local liquid concentration at the point on the tray.
3 Factors Affecting Efficiencies

479 (Eq. (4)). Integration through the two-phase dispersion is expressed as follows.

NTU,,= SdNTU,,

Yn

dy -

(4)

Yn-lY*-Y

The factors that affect efficiencies can be divided into three groups: the structural factors, the functional factors and the system factors, and physical properties. The functional factors are, for example, flow regimes, liquid entrainment, liquid weeping, and channeling and flooding. The system factors and physical properties include, for example, surface tension, liquid and vapor density, diffusion coefficients, concentration, viscosity, relative volatility, pressure, and temperature. The most important structural factors that affect efficiencies are flow patterns, trays, tray spacing, outlet weirs, and downcomers. The effect of flow patterns on efficiencies is discussed in more detail in this study. Cross-flow trays are the most common industrial trays. Their turndown ratio and tray efficiency are high. Liquid flows downward through downcomers to the tray below and then horizontally through the vapor-liquid contact area on the tray over the outlet weir to the next downcomer. One-pass tray is the most commonly used liquid flowpath arrangement. To reach higher capacity 2-pass, 3-pass, 4-pass or radial flow type trays can be used. Multipass trays are especially suitable for large column diameters and high liquid loads. Multipass trays have shorter liquid flowpaths and hence lower Murphree tray efficiencies. Lim et al. [ 2 1 studied the effect of liquid channeling on twopass tray efficiencies and found that efficiency is decreased because of liquid channeling on the trays with side-to-center flows. On these trays, liquid flows into a diverging cross section, and stagnant liquid regions, where separation is negligible, are formed. Stagnant regions d o not exist on the trays with center-to-side flows and hence liquid channeling is avoided on these trays. Bolles [2] has proposed a method of how to design a multipass tray with optimum flow distribution and tray efficiency.

F~~plug flow of vapor and with y* constant, integration of E ~(4)gives Eq. ( 5 ) . .

NTU,,
= - In

(5)

B combining Eqs. (3) and (5) the following equation beY tween the Point efficiency and the number of overall vapor phase transfer units is obtained.

E,, = 1- e-NTuov

(6)

The number of overall vapor phase transfer units can be expressed as a function of the number of vapor phase transfer units, the number of liquid phase transfer units, and the stripping factor. Eq. (7) is based on the two-resistance theory of mass transfer and on the steady state mass balance on the liquid.

NTU,, =

1
1

NTU, NTUl
The stripping factor S is defined as follows.

+-

(7)

m S=- V L
NTU,,, NTU,, and NTU, are defined as follows, NTU,,
= K,,at,

(9)

4 Point Efficiency Prediction Methods

4.1 Basic Equations for Theoretical Efficiency Prediction Methods

There are a number of correlations to calculate NTU, and NTUl. These are found in for example, the following correlations: AIChE [3, 41, Chen and Chuang [ 5 ] , Chan and Fair [ 6 ] , Hughmark [7], Zuiderweg [S], and Stichlmair [9].

4.2 New Method for Determining the Number

The basis for the theoretical prediction methods for point efficiencies is in the two-film or two-resistance theory of mass transfer. It is assumed that there is plug flow of vapor with n o horizontal mixing in the two-phase dispersion. The number of differential overall vapor phase transfer units is defined as the ratio of the vapor concentration change over the differential element to the driving force causing the change

of Vapor and Liquid Phase Transfer Units

A new semiempirical method for predicting the number of vapor and liquid phase transfer units was developed. The resulting correlations were given in an earlier article by the authors [LO]. A detailed mathematical development is given here. In the model development the contributions of the mass transfer coefficients k, and kl, the interfacial area a and the vapor and liquid residence times t , and tl are taken

480
into account. The NTU correlations are based on the fundamental Eqs. (10) and (11). The mass transfer coefficients k, and kl are determined using Higbie's [I 11 penetration theory as given recently by Chen and Chuang [ 5 ] . According to Higbie [Ill vapor and liquid phase mass transfer coefficients are given as follows.

Chem. Eng. Technol. 20 (1997) 478-484 Zuiderweg [8] plotted the h f a values as a function of the Weber number and he observed the following dependence in the emulsion regime.

(22)
The following assumptions are made.

kv=

(2)
0.5 0.5

If it is assumed that

The horizontal liquid velocity can be expressed as follows.

&at;

(15)

Combining Eqs. (22) to (25) gives uvev

0.5

then the following equation is obtained by combining Eqs. (12), (14), and (15).
k, tva (D, t :)0.5

el0.5 QI

h:lWa

O3

(16)

Combining Eqs. (lo), (II), (14), (16-18), and (26) gives

0.5
0.5

According to Chen and Chuang [5], Eq. (17) can be written for kl tl.

N T ~ v = C v ~hcl Wea, Q,

0'5

(27)

where

Combining Eqs. (7), (8), (27), and (28) gives the following equation for the overall number of vapor phase transfer units.

Zuiderweg's [8] correlation for the interfacial area a in the emulsion regime was used in the model development. Zuiderweg used surface tension and the vapor and liquid momentum fluxes as the main variables. At first he defined a mixed Weber number as follows. The two constants C, and C1 are obtained by fitting Eq. (29) to the experimental tray efficiency data. Using the experimental data of Sakata and Yanagi [I21 for the system i-butane/n-butane the constants C, and Cl were found to be 49.22 and 80.85, respectively. The method can be applied to hydrocarbon mixtures operating in the froth regime.
5 Results and Discussion
5.1 Column and Performance Data

Zuiderweg [8] calculated hfa values from Eq. (21) using the FRI data [12, 131 for the systems cyclohexane/n-heptane and i-butane/n-butane. Eq. (21) can be derived from Eqs. (7), (8), and (20).

hfa=uv

(L-+- 3 v

NTU,,

The performance testing of an industrial scale i-butane/ n-butane fractionator was carried out. The compositions of the feed and product flows were analyzed and at the same time operating data were gathered from the automation system including flow rates and temperature and pressure da-

Chem. Eng. Technol. 20 (1997) 478-484 ta. The state of the column was observed during a few days to make sure that the column was operating in steady state. The column specifications are given in Tab. 1 and the column performance data in Tab. 2.
5.2 Column Performance Simulations

48 1 The Murphree tray efficiencies were calculated for the onepass and two-pass trays. First, the point efficiencies were calculated for each tray using theNTUcorrelations (27) and (28) and Eqs. (6 - 8). The point efficiencies were then related to the Murphree tray efficiencies using the method of perfectly mixed pools [14]. Molnars [15] eddy diffusivity correlation was used in the Peclet number calculation. For the calculation of one-pass tray Murphree tray efficiencies the geometry of a two-pass tray was converted to the geometry of a one-pass tray by keeping the active tray area and the column diameter constant. The liquid flowpath length for the one-pass trays was 1.81 m (0.97 m for two-pass trays) and the outlet weir length was 2.27 m (average 2.37 m for two-pass trays). The different liquid distribution phenomena of one-pass and two-pass trays were not taken into account. The Murphree tray efficiencies for one-pass and two-pass trays obtained using the method above and the observed two-pass tray efficiencies [lo] are given in Fig. 1. The Eqs. (27) and (28) were found to predict slightly conservative efficiencies [lo]. However, the one-pass tray efficiencies of Fig. 1 are considerably higher than the observed and calculated two-pass tray efficiencies. The results show that the longer liquid flowpath length of one-pass trays corresponds to higher Murphree tray efficiencies. In the i-butaneln-butane column all the trays were two-pass trays and the calculations for one-pass trays could not be tested against the performance data from actual one-pass trays. In order to get actual performance data from both one- and two-pass trays the performance testing of an industrial-scale MTBE purification column was carried out. The column was equipped with 14 two-pass trays in the stripping section and with 14 one-pass trays in the rectifying section. The liquid flowpath length for the one-pass trays was 1.350 m and for the two-pass trays 0.567 m. The simulations were carried out for real trays using a simulator which can update efficiencies while solving the column model [16]. Point efficiencies were calculated using the NTU correlations of Chan and Fair [6] and the point efficiencies were related to the Murphree tray efficiencies using

Ideal tray distillation simulations were carried out by trial and error until the calculated and measured key component productions were equal. The feed was composed of nine components and the key components were i-butane and nbutane. The liquid phase activity coefficients were calculated using the UNIFAC group contribution method. The Antoine equation for vapor pressures was used. Vapor phase fugacity coefficients were computed using the original Soave modification of Redlich Kwong equation (SRK). Eqs. (30) and (31) were used to calculate the composition of the light key component i-butane in the pseudobinary i-butaneln-butane mixture. Pseudobinary i-butane composition is needed in the calculation of the slope of the equilibrium curve m from Eq. (32). YLK=
YLK YLK + YHK

m=

YLK,n+l-

YLK,n

XLK,n+l-XLK,n

5.3 Two-Pass Trays vs. One-Pass Trays

Efficiencies were predicted from the column composition profiles and temperatures, physical properties, and some column performance data of each tray using different prediction methods. These results have been given in an earlier article by the authors [lo].

Table 2. Performance data (product flow rates adjusted).

Table 1. Column specifications.
Column height [m] Column diameter [m] Number of trays Type of trays Weir length (side) [m] Weir length (center) [m] Liquid flow path length [m] Active area [m] Downcomer area (side) [m] Downcomer area (center) [m] Tray spacing [m] Hole diameter [mm] Total hole area [m] Outlet weir height [mm] Tray thickness [mm] Number of valves per tray Free fractional hole area [%I 51.8 2.900 74 2-pass Ballast V-1 valve 1.859 2.885 0.967 per pass 4.9 0.86 0.86 0.600 39 0.922 51 2 772 18.82 Feed tray Feed flow rate [kg/h] Bottoms flow rate [kg/h] Reflux flow rate [kg/h] Reflux temperature [C] Column top pressure [kPa] Feed pressure [kPa] Boiler duty [MW] Compositions [wt%] Propane i-Butane n-Butane I-Butene i-Butene Trans-2-butene Neopentane i-Pentane n-Pentane 37 26234 18119 92838 18.5 658.6 892.67 10.240 Feed 1.5 29.4 67.7 0.2 0.2 0.1 0.1 0.8 0.1 Top 5.3 93.5 0.2 0.4 0 0.6 0 0 0 Bot 0.3 0.3 98.1 0.1 0 0.1 0.2 1.1 0.1

482

Chem. Eng. Techno]. 20 (1997) 478-484

130

I
t

5.4 Effect of Outlet Weir Height on Efficiencies

one-p,as tray

two-pass tray

two-pass tray (observed)

Seven NTU prediction correlations were used to calculate point efficiencies for the different outlet weir heights. These correlations were: AIChE [3, 41, Chen and Chuang [5], Chan and Fair [6], Hughmark [7], Zuiderweg [8], Stichlmair [9] and the correlations (27) and (28) presented in this study, The outlet weir heights studied were 0, 0.05,. . . , 10cm. The point efficiency for a given outlet weir height was calculated for each tray of the column and an average point efficiency value was taken. The average point efficiencies for each outlet weir height using different point efficiency prediction methods are given in Fig. 3. The point efficiencies were related to the Murphree tray efficiencies using the method of perfectly mixed pools [15]. Eddy diffusivities were calculated using Molnar's [ 161 method. The average Murphree tray efficiencies as a function of outlet weir height are given in Fig. 4. The behavior of the efficiency curves for the correlations (27) and (28) is opposite to the curves given by other methods. When the outlet weir height is increased from 0 to 1.5 cm, point efficiency is increased, and at higher outlet weir heights it decreases. For most of the other methods the situation is opposite. Efficiency is first decreased when the outlet weir height is increased and at higher weir heights efficiency steadily increases. The effect of outlet weir height on the Murphree tray efficiencies was investigated using the observed point efficiencies for each tray [lo] and different point efficiency to the Murphree tray efficiency relation methods. An average

80
1

17

25

33

41

49

57

65

73

tray
Figure 1. Murphree tray efficiencies of one-pass and two-pass trays calculated using the NTU correlations presented in this study. The observed two-pass tray efficiencies are from Klemola and Ilme [lo].

the mixed pool model of Gautreaux and O'Connell [14]. The eddy diffusivities were calculated using the method of Welch et al. [17]. The simulated composition profiles were in good agreement with the measured composition profiles, which indicates that the calculated Murphree tray efficiencies are correct. The individual component Murphree tray efficiencies for each tray are given in Fig. 2. The efficiencies for one-pass trays (trays 1 - 14) are considerably higher than the efficiencies for two-pass trays (trays 15 - 28).

400T A

'"1 1

this study

Chan and Fair

A,ChE

Hughmark

50

-Ibutane

'
I
25
30

0 0
5
10

04
0

: :::: : :
0.02
0.04

: : :: : : ;:
0.06
0.08

:-I
0.1

15

20

tray number Figure 2. Murphree tray efficiencies of MTBE column with one-pass
I

weir height, m
Figure 3. Average point efficiencies as a function of outlet weir height using different point efficiency prediction methods. _ _ -

trays above the feed and two-pass trays below the feed.

Chem. Eng. Technol. 20 (1997) 478-484

this study
I

483
AlChE
I

Chan and Fair

140 T

120

1
Hughrnark Zuiderweg

plemented in the assessment of the effect of outlet weir height on the Murphree tray efficiencies. The outlet weir heights studied were in the range from 0 to 10 cm. The eddy diffusivity correlations used were: AIChE [3], Barker and Self [19], Bennett and Grimm [20], Gilbert [21], Molnar [16] and Welch et al. [18]. The point efficiency to the Murphree tray efficiency correlation of AIChE [3, 41 was tested for different outlet weir heights too. Eddy diffusivities needed also in this correlation were calculated using the method of Molnar [16]. The effect of outlet weir height on the Murphree tray efficiencies given by various eddy diffusivity correlations is shown in Fig. 5 . The point efficiency to the Murphree tray efficiency relation methods which were found to give the most consistent results with the observed Murphree tray efficiencies [lo] show that liquid mixing on the trays equipped with higher outlet weirs is more complete and consequently the Murphree tray efficiencies are lower.

20a
0 0

0.02

0.04

0.06

0.08

0.1

weir height, m
Figure4. Average Murphree tray efficiencies as a function of outlet weir height.

7 Conclusions

point efficiency value for the column studied was found to be 8 1.9% [ 101. The mixed pool model [151 was used to convert point efficiencies to the Murphree tray efficiencies. TO determine the number of perfectly mixed pools on the tray eddy diffusivity is needed. It is a measure of liquid mixing on the tray. Six eddy diffusivity correlations were im-

Mathematical development of a new number of transfer units correlation for the prediction of point efficiencies was presented. Some existing distillation data are required for the determination of the two empirical constants of the correlation. Such data were available, and the constants were calculated and the correlation applied to the column studied. The industrial scale i-butaneln-butane distillation column, for which the performance data were available, was equipped with two-pass trays. Rigorous simulations were repeated until the calculated results equalled the column performance. The data from each tray were used in the calculations to study the effect of two-pass and one-pass trays and the outlet weir height on efficiencies. As expected onepass trays had higher Murphree tray efficiencies than twopass trays due to longer liquid flowpath lengths. This result was supported by the efficiency data from an industrial-scale MTBE purification column equipped with both one-pass and two-pass trays. For the effect of outlet weir height on efficiencies the results were not that clear. Five of the point efficiency prediction methods showed a minimum for efficiencies at the outlet weir height of 1.5 cm. The correlations presented above gave just the opposite results and showed a maximum for efficiencies at the weir height of 1.5 cm. One method showed almost no effect of outlet weir height on efficiencies. The results from the different point efficiency to the Murphree tray efficiency relation methods show that higher outlet weirs correspond to higher degrees of liquid mixing and lower Murphree tray efficiencies.

Bennett and Grirnrn

AIChE

\,

Barker T d Self

%
c L

'I5

I I . AlChE correlabon
Moinar

0.02

0.04

0.06

0.08

0.1

weir height, m
Figure 5 Effect of outlet weir height on the Murphree tray efficiencies . given by various eddy diffusivity correlations.

484
Acknowledgment

Chem. Eng. Technol. 20 (1997)478-484

composition of vapor leaving a point on tray n- 1 average composition of vapor leaving tray n

This work was financially supported by the Neste Oy Foundation and the Neste Oy Educational Foundation to which the authors are grateful.

Greek symbols
liquid density vapor density surface tension liquid exposure time vapor exposure time

Received: October 18, 1996 [CET 8671

Symbols used
effective mass transfer area, m2m-3 contact zone intercept of equilibrium line constant in Eqs. (27) and (29) constant in Eqs. (28) and (29) liquid phase diffusion coefficient vapor phase diffusion coefficient Murphree vapor phase tray efficiency overall column efficiency point efficiency for a tray clear liquid height froth height overall vapor phase mass transfer coefficient liquid phase mass transfer coefficient vapor phase mass transfer coefficient liquid flow rate slope of equilibrium curve molecular weight number of real trays in the column number of theoretical trays in the column number of liquid phase transfer units number of vapor phase transfer units number of overall vapor phase transfer units liquid flow rate stripping factor liquid residence time on the tray liquid residence time on the tray defined in Eq. (18) vapor residence time in the two-phase dispersion vapor residence time on the tray defined in Eq. (14) superficial vapor flow rate per active contact area horizontal liquid velocity vapor flow rate outlet weir length Weber number liquid composition average composition of liquid leaving tray n liquid composition of light key component in pseudo binary mixture vapor composition of light key component in pseudo binary mixture vapor composition composition of vapor in equilibrium with the liquid at a point on the tray composition of vapor in equilibrium with the liquid leaving tray n composition of vapor leaving a point on tray n

Subscripts
1
V

HK LK

liquid phase vapor phase heavy key component light key component

References
(11 Lim, C.T., Porter, K.E., Lockett, M.J., Trans. Instn Chem. Engrs 52 (1974) pp, 193 - 201. [2] Bolles, W.L., AIChE J. 22 (1976) No. 1, pp. 153 - 158. [3] AIChE, AIChE Bubble Tray Design Manual, AIChE, New York 1958. [4] Gerster, J.A., Hill, A.B., Hochgraf, N.N., Robinson, D.G., Tray Efficiencies in Distillation Columns, University of Delaware and AIChE, New York 1958. [5] Chen, G. X., Chuang, K.T., Ind. Eng. Chem. Res. 32 (1993) No. 4, pp. 701 -708. [6] Chan, H., Fair, J.R., Ind. Eng. Chem. Process Des. Dev. 23 (1984) NO. 4, pp. 814-819. [7] Hughmark, G.A., AIChE J. I7 (1971) No. 6, pp. 1295- 1299. [8] Zuiderweg, F.J., Chem. Eng. Sci. 37 (1982) No. 10, pp, 1441- 1464. [9] Stichlmair, J., Die Grundlagen des Gas-Fliissig-Kontaktapparates Bodenkolonne, Verlag Chemie, Weinheim 1978. [lo] Klemola, K., Ilme, J., Ind. Eng. Chem. Res. 35 (1996) No. 12, pp. 4579 - 4586. [ l l ] Higbie, R., Trans. A m . Inst. Chem. Eng. 31 (1935) pp. 365 - 389. [12] Sakata, M., Yanagi, T., Inst. Chem. Eng. Symp. Ser. 56 (1979) pp. 3.2/21- 34. Yanagi, T., Sakata, M., Ind. Eng. Chem. Process Des. Dev. 21 (1982) NO. 4, pp. 712-717. Ilme, J., Klemola, K., Aittamaa, J., Nystrom, L., unpublished work. Gautreaux, M.E., O'Connell, H.E., Chem. Eng. Prog. 51 (1955), pp. 232 - 237. Molnar, K., Periodica Polytechnica Mechanical Engineering I8 (1974) NO. 2-3, pp. 155-166. Aittamaa, J., Kemia-Kemi 8 (1981), pp. 295-317. Welch, N.E., Durbin, L.D., Holland, C.D., AIChE J. I0 (1964) NO. 3, 373 - 379. Barker, P.E., Self, M.F., Chem. Eng. Sci. I7 (1962) pp, 541 - 544. Bennett, D.L., Grimm, H.J., AIChE J. 37 (1991) No. 4, pp. 589 - 596. Gilbert, T. J., Chem. Eng. Sci. I0 (1959) No. 10, pp. 243 -253.