Vous êtes sur la page 1sur 40

BINARY DISTILLATION

Introduction In the 19th Century, the various branches of the Chemical Industry grew up independently. Each Industry had its own expertise, which was jealously guarded as a trade secret. Thus, the Alcohol industry had experts on alcohol distillation, the Coal Tar industry had experts on coal tar distillation, the infant petroleum industry (mainly engaged in making kerosene for lamps) had experts on petroleum distillation, and so on. In the late 19th / early 20th Century, particularly as the petroleum industry started to develop, a few far-sighted pioneers, such as George Davis in Manchester, realised that it was possible to study the common features of all these operations to study Distillation itself, independent of what was being distilled. This marked the start of Chemical Engineering as a discipline. The name "Unit Operations" was coined by Arthur D. Little in the U.S. to describe distillation, gas absorption, solvent extraction etc. and the name was used until the 1980s, when it was generally replaced with the more descriptive Separation Processes. In this first module on separation processes, you will study two major processes: Binary Distillation and Gas Absorption. Others, such as Solvent Extraction, Multicomponent Distillation, adsorption, membrane processes, etc. will follow in later modules. We shall only look at the early stages of calculation - later stages, such as the detailed design of the column internals will be covered in other modules. Today, more and more routine design is being done on computers, and you will be learning to use a package called "HYSYS" in the H82CSY Computer Systems module which runs next semester. HYSYS can perform heat and material balances, but also design distillation columns, both binary and multi-component. However, using such packages as "black boxes" without an understanding of what is going on can lead to expensive mistakes, and this module will emphasise the basic principles underlying the design.

DISTILLATION
This is still the largest and commonest Separation Process. Its importance can be seen from the fact that it is estimated that 2-3% of the entire energy consumption of the world goes into distillation. Huge distillation columns dominate the skyline of any large chemical complex or oil refinery. This module will start to look at how they are designed. 1. Basic Principles

The idea behind distillation is simple: if a mixture of two or more miscible liquids is heated to boiling, the vapour formed will be richer in the more volatile component(s). If this vapour is totally condensed, it will yield a liquid of the same composition, richer in the more volatile component(s) than the starting liquid ("feed"). Repeating the process a sufficiently large number of times will give the most volatile component as a pure liquid. The first stage in our design is clearly to quantify this concept of volatility, which is part of the study of vapour-liquid equilibrium. Equilibrium is a dynamic state, reached when the rate of evaporation of each component from a surface is equal to its rate of condensation at the surface. Our starting point is Raoults Law, given by equation 1.1 for any component i
p i = p i* x i
* Where p i is the saturated vapour pressure of pure i

(1.1)

This is combined with Daltons Law for the gas phase:


p A + p B + ........ = P

(1.2)

to give an equation relating the vapour and liquid equilibrium compositions at any temperature:
yi = p i p i* = .x i P P

(1.3)

Raoults Law (equation 1.1) only holds for an ideal liquid mixture basically one in which all the molecules are similar chemically. The worked examples we shall follow are for the system benzene-toluene, for which this assumption is satisfactory. In other cases, especially ones involving water where there is hydrogen bonding, Raoults law does not hold. You will learn how to deal with non-ideal systems in the 4th Semester module Chemical and Phase Equilibria. Daltons Law (equation 1.2) holds for ideal gases, and most vapours will be approximately ideal at pressures of 1 or 2 atmospheres. The saturated vapour pressure, pi*, which appears in equations 1.1 and 1.3. is a function of temperature. This function is often expressed by the empirical Antoine Equation:
log 10 p i* = A B t +C
2

(1.4) where A, B and C are constants. The units of p and t must be specified, as they affect the values of the constants. The table below gives a few examples with t in C and p in mmHg. Component benzene toluene water methanol Carbon disulphide Carbon tetrachloride 1.1 Binary systems A 6.89326 6.96554 7.95864 8.07919 6.94193 6.84083 B 1203.828 1351.272 1663.125 1581.341 1168.620 1177.910 C 219.921 220.191 227.528 239.650 241.534 220.576

When there are only two components, 1 and 2, we can write: x2 = 1 x1 y2 = 1 y1 Writing equation 1.3 for each component, using the above equations and adding, we get:
* * p1 p2 1= x1 + (1 x1 ) P P

(1.5)

At a given Pressure P and temperature t (which fixes p1* and p2* from equation 1.4) we can solve equation 1.5 for x1, the liquid composition which would just boil at the given conditions.
* P p2 x1 = * * p1 p 2

(1.6)

Substitution into equation 1.3 then gives the corresponding equilibrium vapour composition y1. Example At 95C, the vapour pressures of benzene (BPt 80.1C) and toluene (BPt 110.8C) are 1180 and 478 mmHg respectively. What mole fraction of benzene will give a liquid phase boiling at 95C under a pressure of 760 mmHg, and what is the corresponding equilibrium vapour mole fraction?
xB = 760 478 = 0.402 180 478

yB =

1180 0.402 = 0.624 760


3

This calculation can be repeated for a number of different temperatures between the two boiling points (why not outside this range?) and the results plotted as graphs of temperature against mole fraction at constant pressure (a T-x-y diagram). Enter the points from your calculation on the graph below, and sketch in the complete curves. Note that, by convention, for a binary system, we plot mole fractions of the more volatile component.

120 110 100 90 80 70

T/ C

0.2

0.4 0.6 x or y

0.8

Note that, although there is a slight pressure drop up a distillation column it is usually ignored in simple calculations. The pressure of a distillation column is normally controlled at the top, and the whole column is assumed to be at this pressure. HYSYS allows you to specify a pressure drop per plate if required, though for preliminary calculations you may wish to set this at zero.
1 0.8 0.6 y 0.4 0.2 0 0

0.2

0.4 x

0.6

0.8

In binary distillation calculations, we do not normally use the temperatures (except at the top and bottom of the column), as will be explained later, so we plot instead a graph of vapour
4

mole fraction against liquid mole fraction (the x-y diagram). Plot your single point on the graph above, and sketch in the rest of the curve. Note that T-x-y and x-y diagrams can also be drawn for non-ideal systems, either from direct experimental measurements, or by the more complex methods explained in the C&PE module. The binary distillation calculations we will be doing later assume we have a valid xy diagram for the system and pressure in question. The effect of moderate pressure changes on the x-y diagram is not, in fact, very large, so it is not always necessary to have a curve at the exact pressure. 1.2 Multicomponent systems

Here we can no longer use the simple x-y diagram, and we need to work in terms of relative volatility. The volatility of any component i in a mixture can be described by the K-factor, which is the ratio K = yi/xi, and the relative volatility of component i relative to component j is defined as:
i, j =
Ki y /x = i i K j yj / xj

(1.7)

In an ideal liquid mixture, from equation 1.3, we have yi/xi = pi*/P whence

i, j

y i / xi p i* / P p i* = = = y j / x j p *j / P p * j

Both pi* and pj* increase with temperature, but over small ranges, the ratio, and hence i,j, is often approximately constant. In distillation calculations, an average value between top and bottom temperatures is usually used. For binary systems, the subscripts on are not needed, and if it is constant, we can write
=
y/ x (1 y ) /(1 x)

which may be rearranged to give the equation of the x-y diagram:


y =

x 1 + ( 1) x

(1.8)

This technique was often used in the past as a quick way of generating an x-y diagram, but with modern spreadsheets there is little advantage over the more exact method which does not assume that is constant. Most short-cut multicomponent calculations, however, still use . 2. One-Stage (Flash) Distillation

The simplest form of continuous distillation is shown below. The mixture is fed to the still, which is heated. The top product is vapour, which is then condensed in condenser C, and the bottom product is liquid with
5

approximately the same concentration as in the still. The top & bottom products are assumed to be in equilibrium.

condenser

D W

heat

F is the Feed flowrate (molar units) of composition xF D is the top product flow of composition xD (Distillate) - it is yD before condensation. W is the bottom product flow of composition xW (historically Waste", as in ethanol-water) Set up material balance assuming steady state : Overall MVC
F = D +W
x F F = x D D + xW W

(2.1) (2.2)

Eliminating W (2.3)

D x F xW = F x D xW

Now, writing D/F = f, the fraction evaporated, and noting that xD is equal to the vapour composition, y, and xW to the liquid composition, x within the still, we can rearrange equation 2.3 to obtain
y= x f 1 .x + F f f

(2.4)

This is the equation of a straight line, of (negative) gradient (f-1)/f and passing through the point (xF,xF) on the y=x line. Its intercept with the equilibrium curve gives the compositions of top (y) and bottom (x) products. Question
7

What happens to xW when D/F 0? equation will be xF] and when D/F 1?

(zero evaporation of feed) [xW in

(total evaporation of feed) [xD xF] x-y diagrams

Plot the general case, and these limiting results on the below.

D/F 0

0 < D/F < 1

D/F 1

As D/F increases, the top product composition falls. In practice, the amount chosen to be evaporated is a compromise between yield, composition, and cost. In the constant pressure flash described above, the heat input is usually adjusted to give a constant temperature in the still, whence it is called an Isothermal Flash. The temperature determines the composition via pi* (see equation 1.6 for an ideal mixture) which then fixes both K and the amount evaporated.
condenser

F
P1 P2

D W

In the more common Adiabatic Flash the hot liquid feed is let down to a pressure below its bubble point, so that some of it evaporates. The evaporation cools the mixture to a lower temperature than the feed. Since the evaporation is adiabatic, we can add an enthalpy balance to the two mass balances equations 2.1 and 2.2:
hF F = H D D + hwW

(2.5)

This may be combined with equation 2.1 and rearranged to give D hF hW = F H D hW The problem needs an iterative solution:

(2.6)

Guess the temperature. This enables us to calculate the compositions of liquid and vapour xW and yD = xD (from equations 1.6 and 1.3 or using a T-x-y diagram). Substitution into equation 2.3 gives D/F, the fraction evaporated.
9

Then, from the guessed temperature and calculated compositions, we can calculate the liquid and vapour enthalpies hW and HD, and since the feed conditions are known, hF can also be found. Substitution into equation 2.6 then gives another value of D/F. If the two values are not the same, a new temperature must be guessed. This is a tedious calculation, best left to a computer package such as HYSYS. However, if an Enthalpy Composition diagram is available at the flash drum pressure, a graphical solution is possible, as studied in CPP1 in semester 1. In the diagram below, F lies in the two-phase region, while D and W lie on the dew point and bubble point curves respectively (Saturated vapour and liquid). Any straight line through F will give values of D & W which satisfy the mass balance (equation 2.3) and enthalpy balance (equation 2.6) (This is the lever rule). By making sure the line is a tie line we also satisfy the requirement that D & W are in equilibrium.

H or h

x or y The problem for hand calculations is that the number of enthalpy-composition diagrams which have been produced is very limited. Even for systems which have been studied, a different diagram is needed for each pressure.

10

3. Continuous multistage distillation


condenser condenser

heat

heat

The maximum possible concentration of the MVC in a single adiabatic stage is that of a vapour in equilibrium with the feed (when D/F = 0, see page 6). To do better, we must take the liquid top product and feed it to another stage, and repeat as many times as necessary. Question. What shall we do with the bottom product from the second and subsequent stages? Modify the diagram above when you have the answer. Now suppose we want a bottom product richer in the LVC than that from the feed stage. Sketch below how this could be achieved.

We now have a counter-current, multiple stage distillation which can achieve a bottom product as rich as we like in the LVC, and a top product as rich as we like in the MVC. However, each stage has a temperature controller (to fix D/F - see page 6), a heat input and a condenser. Such a device would be extremely costly to build and to run. Question: How can we improve it? [The answer is over the page, but don't cheat!]

11

Obviously, the heat given out by the condensing vapours can be used to boil the next stage, provided the vapour temperature is higher than the boiling liquid. (is it? - think of the T-x-y diagram). Rather than use expensive heat transfer coils, we can bubble the vapour directly through the liquid to be evaporated, as sketched below. There is actually very little heat transferred during the bubbling process: what happens is that some of the LVC in the vapour condenses, and some of the MVC in the liquid evaporates so that the two latent heats balance. It becomes essentially a mass transfer process.

Vapour Question

Liquid

Where does the liquid in the last stage come from? Add this to the diagram.

Finally, the need for pipework and pumps can be diminished by stacking the stages one on top of the other, so that the liquid flows by gravity. This gives the standard trayed distillation column. Each tray has some method of dispersing the vapour bubbles into the liquid; the commonest is the sieve tray, which simply contains a number of holes drilled on a regular pattern. Liquid level is maintained by means of a weir, over which the liquid product from the tray flows, into the downcomer, and so to the next tray. We will look a little at tray design in a later lecture. Because the vapour path through the liquid is quite short (10 cm at most), the vapour and liquid leaving the tray will not have long enough to reach complete equilibrium with each other as we assumed when looking at the single isothermal stage. To cope with this, we design the column assuming that equilibrium is achieved, and then use an empirical stage efficiency. (See section 6.1) The vapour leaving the top of the column passes to a condenser. Some of the condensate is returned to the column as reflux, while the rest forms the top product. The bottom
12

stage of the column is the reboiler, in which vapour is generated, and from which bottom product is withdrawn. 4. Design of a binary counter-current plate (trayed) column. The figure below shows the main components of the distillation unit: column, plates, downcomers, condenser, reboiler, reflux drum. The feed is introduced on a suitable plate (we will show later how to fix this). The section of column above the feed is the enriching (rectifying) section, and that below the stripping section. Mark all these on the figure and show the vapour/liquid traffic

In the single stage isothermal flash we were allowed to choose what fraction of the feed we evaporated (D/F) and this determined the
13

composition of the products. In a multi-stage column, we can choose (within limits) the reflux ratio, R, which is defined as the ratio of the condensate returned to the column as reflux to that taken as top product, and this determines how many stages we need for a given separation. 4.1. Plate-to-plate calculations In the initial design it is assumed that equilibrium is reached on each plate, that is the vapour leaving is in equilibrium with the liquid leaving, and that the liquid on the plate is well mixed and has the same composition as the liquid leaving. So the number of ideal plates is calculated. Then a stage efficiency is assumed (see later) which is used to calculate the actual number of plates required for the specified separation. The first analysis of fractionation was carried out by Sorel in 1899 and is still the basis of rigorous computer based simulations, such as those performed by HYSYS. Start by looking at the top of a column. We will number plates from the top down (both this and the opposite convention are used), and liquid and vapour streams are named after the plate they leave. Since we are dealing with binary systems, and by convention the MVC mole fractions are calculated, we do not need subscripts for the component, so all subscripts in the following refer to stream numbers.

V1

D V2 L1

L0

In this case the reflux ratio R, the top product flow D and the composition xD are fixed. Since by definition of R, we have R = L0 / D And by mass balance V1 = D + L0 Solving L0 = RD (4.1) V1 = (R+1)D so that L0 and V1 are also known. We may also note that y1 = x0 = xD

(4.2)

To determine the conditions on the first tray there are 4 unknowns: x1 the liquid composition on tray 1, y2 the vapour composition entering from tray
14

2, L1 the liquid flow leaving tray 1, V2 the vapour flow entering from tray 2 below. These can be calculated from the following four equations: Equilibrium between streams L1 and V1
y1 = K .x1

(4.3)

Overall mass balance over tray


V2 + Lo = V1 + L1

(4.4)

MVC mass balance over tray


y 2V2 + x o Lo = y1V1 + x1 L1

(4.5)

Heat balance over tray (4.6)

H 2V2 + ho Lo = H 1V1 + h1 L1 + losses

(Heat losses are negligible from large columns with insulation). Solution of these equations requires trial and error as the stream enthalpies are dependent on the compositions (via Txy) and compositions cannot be established without the enthalpies. The calculation is very similar to that involved in adiabatic flash distillation given on page 7. Once tray 1 is fully defined the calculation can be repeated for tray 2 etc down the column. The main difficulty with Sorel= s method is in the estimation of the stream enthalpies, especially when the liquid mixture is non-ideal, and there are heats of mixing (see C&PE, semester 4).. For a binary mixture these are conveniently expressed on an enthalpy composition diagram, as was done above for adiabatic flash. Computer simulations will have a database of physical properties and estimation methods available (eg HYSYS). Sorel, a Frenchman, was mainly interested in distilling alcohol from water - a non-ideal mixture. The Americans began distilling petroleum products in the early 1900's. Hydrocarbon mixtures are reasonably ideal mixtures (no heat of mixing) and the molar latent heat is almost independent of composition. In fact, the molar latent heats of most pure liquids at their normal boiling points are fairly similar, as may be seen from the table below Liquid water methanol Latent heat MJ kmol-1 40.683 35.278
15

benzene toluene CCl4 CS2

30.781 33.201 30.019 26.754

In 1909 Lewis showed that this implies that the liquid and vapour flows either side of a plate are equal even though the composition has changed. Essentially, apart from small temperature changes between incoming and outgoing liquid and vapour streams (and hence mcP T terms) all the heat liberated by condensing some of the LVC from the incoming vapour will be used to evaporate some of the MVC from the liquid on the plate. Since there are no heats of mixing, and the molar latent heats are the same, the moles condensed and evaporated must be equal, so Vn = Vn-1 . Substitution into the overall mass balance (equation 4.4) then leads to Ln = Ln+1 This is termed equimolal overflow or constant molal overflow and removes the need for the heat balance. 4.2 The McCabe-Thiele Method Mc Cabe & Thiele (1925) suggested a very convenient graphical method for design of binary columns. It makes the same assumptions as Lewis: Constant molar latent heat No heat losses No heat of mixing This leads to constant molar overflow as shown above. However, the liquid mixture need not be ideal, provided there is no heat of mixing, since equilibrium is expressed by an x-y diagram, and such diagrams can be drawn for any binary system.

1 2

n-1 n

Ln Vn+1
16

Consider a section of column above the feed, and perform an MVC mass balance over the boundary sketched in the diagram above.
y n +1 .Vn +1 = x n Ln + x D D

(4.7) (4.8) (see equation 4.1) (see equation 4.2)

whence

yn+1 = x n

Ln x D + D Vn +1 Vn +1

But Ln = Ln-1 = = L0 = RD And Vn+1 = Vn = = V1 = (R+1)D Substituting into equation 4.8, we get x R y n +1 = xn + D R +1 R +1

(4.9)

This is the equation of a straight line, of gradient R/(R+1) and passing through the point (xD, xD) on an x-y diagram. It is called the top operating line (tol) of the column, and can be drawn for a required top composition xD when R has been chosen. Now, starting at the top of the column, x1 is related to y1 (which equals xD) by the equilibrium line on the x-y diagram, y2 is related to x1 by the tol, x2 is related to y2 by the equilibrium line, etc. The compositions can be found by "stepping off" on the graph. Try it!
1 0.9 0.8 0.7 0.6 0.5 0.4 0.1

equilibrium line

tol
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x
Note The equilibrium line relates the compositions of liquid & vapour leaving a plate The operating line relates the compositions of liquid & vapour passing between plates As we move down the column, we reach a point where compositions match the feed, and it is sensible to introduce the feed onto this tray. Below the feed, liquid and vapour molar flows will again be constant, but not the same as those above it. Calling the flows above the feed L t
17

and Vt, and those below Lb and Vb, with the feed being F, what can you say about Lb and Vb in the following cases? a) Feed is saturated liquid.
b)

Lb = Lt + F Lb = Lt Lb = Lt + F Lb > Lt + F Lb < Lt

Vb = Vt Vb = Vt F

Feed is saturated vapour

c) Feed is 50% liquid, 50% vapour d) Feed is subcooled liquid e) Feed is superheated vapour

Vb = Vt F Vb > Vt Vb < Vt F

The above results can be generalised quantitatively by defining a quantity q: Question


q = Heat to vaporise 1 mole of feed Molar latent heat of feed

What are the values of q in the five cases above? We can generalise the five results above by writing: Lb = Lt + q.F Vb = Vt +(q-1).F (4.10)

18

Vb Lb
m m1 +

N -1 N

Now consider the bottom of the column, and perform a total and MVC mass balance over the boundary drawn. (Note that L & V have been set to the constant values Lb and Vb.) Overall: MVC: Or ym =
Lb = Vb +W x m 1 = y mVb + x wW

(4.11) (4.12) (4.13)

Lb x W x m 1 w Vb Vb

Equation 4.13 is also the equation of a straight line on the x-y diagram. Putting x = x W, and substituting from 4.11 shows that the line passes through (x W,xW). This is the bottom operating line (bol). Note that the gradient of both operating lines (equations 4.8 and 4.13) is given by the local value of L/V. Where do the two operating lines intersect? Ignoring the subscripts n, n+1, etc, and subtracting (4.7) from (4.12): y(Vb - Vt) = x(Lb - Lt) - (DxD + WxW) Mass balances over the whole column give:
19

(4.14)

Overall MVC

F=D+W FxF = DxD + WxW (4.16)

(4.15)

Now substituting from (4.10) and (4.16) into (4.14) we obtain:


y= q 1 x xF q 1 q 1

(4.17)

This is a straight line of gradient q/(q-1) and passing through the point (xF,xF). It is usually called the q-line.
1 0.9 0.8 0.7

Increasing q

y
0.6 0.5 0.4 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Question Describe the q-line for a saturated liquid feed? Gradient And for a saturated vapour? Gradient 0

Vertical Horizontal

The graph above shows the effect of changing values of q on the q-line. We can now set out the complete McCabe-Thiele graphical method.
1. 2. 3. 4. 5.

From the given information (usually xF, F, xD, xW) solve the whole column mass balances (4.15 and 4.16) for the two unknowns (usually D & W). Calculate q for the given feed conditions, and draw in the q-line as a line of gradient q/(q-1) through (xF,xF). Select a reflux ratio R (see the next section for how to do this), and draw in the tol as a straight line of gradient R/(R+1) and passing through (xD,xD) Mark the intersection of lines (2) & (3), and draw the bol through this point and (xW,xW). The intersection point is often labelled (xq, yq) Starting at point (xD,xD), step down, using the tol until you pass the intersection (4), then switching to the bol until (xW,xW). is reached.
20

6.

Count the number of stages, and locate the feed.

Now try it! On the benzene-toluene x-y diagram provided, enter xF = 0.5, xD = 0.95, xW = 0.05. Use a reflux ratio of 2.5, and assume feed is a saturated liquid. 4.3 Minimum Reflux Ratio A distillation column can be run on total reflux, where all the condensate is returned to the column as reflux. This is often done at start-up, to get the column ready to receive fresh feed. Question What is the value of R in this case? And what does the tol look like?
1 0.9 0.8 0.7 0.6

Gradient of 1 so lies on y=x

equilibrium line

tol

q-line
0.5 0.4 0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

As the reflux ratio is reduced, the gradient of the tol decreases. Eventually you reach a point where the equilibrium curve, tol and q-line all meet at a point, as shown above. In this case, if you try stepping off the plates, the steps get smaller and smaller as this point is approached, and their number becomes infinite. This is the minimum reflux ratio, Rmin, at which the desired separation is possible. For the situation as sketched, the minimum gradient is 0.524, corresponding to a Rmin of 1.103. The situation at R = Rmin, where the tol and equilibrium curve approach each other is called a pinch. Very non-ideal systems can have equilibrium curves with a point of inflexion in them, and a pinch becomes
1 0.9 0.8 0.7 0.6 0.5 0.4

xD

q-line
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

21
1

possible even before the tol touches the intersection of the q-line and the equilibrium curve, as sketched here. 4.3.1 Optimum Reflux Ratio The bigger the reflux ratio, the larger are the vapour and liquid flows within the column for a given top product flow (see equations 4.1 & 4.2). Since the vapour velocity within a column is normally around 1 m/s, a larger vapour flow means a larger column diameter, and hence a more expensive column. More seriously, since the vapour flow is generated in the reboiler, a larger vapour flow means a higher energy consumption, and greater operating expense. On the other hand, the smaller the reflux ratio, the greater the number of plates, which means a taller, and more expensive column. The graph below sketches the variation of capital (construction) and operating costs with R. Note the steep rise of capital costs near to Rmin, due to the rapid increase in the number of plates. The total cost goes through a minimum at some value of R, and it is normal to operate just above this minimum. (why not at the minimum?). Typically, R/Rmin = 1.05 - 1.3, with 1.2 being a good rule of thumb.

Cost

Rmin

4.3.2

Counting the plates.

Each vertical line in the stepping-off graph links values of x and y which are in equilibrium, and hence corresponds to a theoretical stage. The horizontal line which crosses x W represents the bottom of the column, and the fraction of that line to the right of x W is the last fractional stage. Thus, in the diagram overleaf, there are approximately 9.7 theoretical stages needed. The reboiler at the bottom of the column counts as one of these stages, since its liquid composition is xW (but see later, section 5.1), so the column below has 8.7 theoretical stages, plus reboiler.
22

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x
Useful Tip If xD is very close to 1.0 or xW very close to 0.0, it may be difficult to draw the steps, as you are close to a pinch. Try drawing an enlarged version of the graph for the pinch region, or, in extreme cases, use log-log coordinates, plotting y against x for the pinch near x = 0, and (1-y) against (1-x) for the pinch near x = 1. Note that the operating lines are no longer straight in log-log coordinates, and must be plotted from their equations (4.9 for tol, 4.13 for bol).
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0 0.1
xF =0.005

0.1

0.1

el

0.05

bol

bol y= x
y

0.02 0.01 0.005 0.002 0.001 0.0005 0.0002 0.001 0.005 0.02 0.1 0.0005 0.002 0.01 0.05

el y= x

4.3.3

Locating the feed plate

The optimum location for the feed plate is the step crossing xq (stage 5 from the top in the figure above). The vertical before it descends to the tol, and the vertical after it to the bol, as sketched below. However, the column will still operate with other feed plate locations, although the total number of stages required for a given separation will be greater. Sketch on the graphs below what happens when the feed is on a lower and a higher plate than the optimum.

23

Can you see what is the lowest possible position? And the highest possible?

5. Reboilers and condensers 5.1 Reboilers There are two major types of reboiler: the kettle and the thermosyphon. Condensing steam is normally used as heating medium, although hot oil is sometimes used when the material being distilled would react with any water arising from a leak, or where hot enough steam is not available. The oil is heated in a separate furnace.
Vapour to column

Steam Condensate

Steam
Bottom product Liquid from column

Condensate

Bottom product

The Kettle reboiler (sketched above) is a special horizontal shell-and-tube heat exchanger with a shell larger than the tube bundle. The bundle is at the bottom of the shell and is covered by the liquid to be boiled, leaving a vapour space above it. Heating fluid is on the tube side. Liquid from the bottom of the column is passed to the boiler, and the vapour
24

generated returned to the column. The remaining unboiled liquid forms the bottom product, and is normally removed over a weir. Occasionally, the tube bundle can be placed directly in the bottom of the column, as also sketched above. Because the vapour returned to the column is in equilibrium with the bottom product, the kettle reboiler acts as one theoretical stage. In the thermosyphon reboiler, liquid is withdrawn from the bottom of the column, partially vapourised, and both vapour and liquid are returned to the column as a two-phase mixture Phase separation only occurs in the column, above the liquid level in the base. Both vertical and horizontal thermosyphon reboilers are used, and they are sketched below.

Steam Steam Condensate Condensate Bottom product

Bottom product

In the vertical arrangement, the liquid to be boiled is on the tube side and heating fluid on the shell side, whereas in the horizontal arrangement, the heating fluid is in the tubes and the liquid to be boiled in the shell. In both cases, circulation is maintained by natural convection, with the lighter vapour/liquid mixture in the heat exchanger and riser being balanced against the denser pure liquid in the downcomer and column sump. The vertical thermosyphon is more efficient, but requires a greater vertical height to operate, so raising the bottom of the distillation column further off the ground. Bottom product is normally withdrawn from the liquid stream entering the reboiler. This stream is not in equilibrium with the vapour, so the thermosyphon reboiler acts as less than a theoretical stage. 5.2 Condensers The commonest arrangement is a horizontal shell-and-tube heat exchanger with cooling water in the tubes and condensing vapour in the shell. However, if the condensate is very corrosive, or the operating pressure high, the cooling water can be put in the shell and the condensing vapour in the tubes. A separate reflux drum is usually provided to collect condensate and control the split between reflux and top product, although this can be done directly in the condenser shell when condensation is on the shell side.
25

Normally, all the vapour is condensed, but occasionally, a partial condenser is used, in which only enough vapour is condensed to provide reflux, with the uncondensed vapour forming the top product. Since some sort of vapour-liquid equilibrium will be established, the partial condenser acts as an additional stage, or a fraction of a stage. 5.3 Choice of Operating Pressure The pressure at which a distillation column is operated is normally determined by the pressure in the up- or down-stream equipment. A small pressure drop (say 0.5 bar) is normally allowed to drive fluid from one piece of equipment to the next through a control valve, so pressures throughout the system are selected at the same time. Liquids can easily be pumped up to higher pressures between plant items if necessary, but compression of gases or vapours is very expensive and should be avoided when possible. The major effect of increased pressure on a distillation column is to raise the temperature, and this has a particular impact in the reboiler and the condenser. Question What governs the lowest pressure at which you might operate a column? The temperature in the condenser should not be below what can be reached with cooling water. What fixes the highest pressure you would choose for the column? The temperature in the reboiler should not be above that of the hottest heating medium available, not so high that produced decomposition occurs.

Occasionally, these lowest or highest pressures must be exceeded, but it is expensive, and is only done after careful consideration of the alternatives. (What might these be?) For very low pressure For very high pressure Use refrigeration Use hot oil

26

Plate design This section covers the absolute basics only. Further details can be found in C&R Vol 2 (4th ed) pp 494-506, or Kister "Distillation Design" pp 259-363 (descriptive sections only) and you will be expected to show evidence of having read these or equivalent material when answering exam questions on this topic. The commonest type of plate installed in columns today is the sieve plate, which simply has holes (typically 5 or 6 mm diameter) drilled in a regular pattern (usually triangular pitch) over the active area of the plate. There are no holes beneath the downcomer from the plate above, nor close to the edges of the plate, where a support ring will be installed. (See sketch on page 10, where the holes have been drawn on a square pitch). If the vapour flow is not great enough, liquid can drip down through the holes ("weeping"), or, in severe cases, the entire liquid content of the tray could drop through the holes ("dumping"). The fraction of design flow at which weeping starts fixes the possible turndown of the tray, and for sieve trays a turndown of around 2:1 is the maximum possible.

The valve tray has a smaller number of larger holes, each having a captive "lid" which can be raised or lowered by the passage of vapour. This reduces the tendency for weeping, and allows much greater turndown as compared to a sieve tray, at only around 20% more cost. The bubble cap tray also has large holes, each fitted with a short circular "chimney" which is covered with a fixed perforated or serrated cap. The chimney acts as a weir which maintains liquid on the tray even if there is no vapour flow at all, while the serrations help to ensure uniform bubbling. These trays have very high turndown without much loss of efficiency, but they are expensive to build and have higher pressure drops than the other types. Although once very common, they are rarely installed today.

6.1

Plate efficiency

Having found the number of ideal stages required for the specified separation you must then estimate the number of actual plates required in the column. The simplest way is to find a value for the overall plate efficiency in the column, Eo, which is the number of ideal trays divided by the real number of trays (usually expressed as a percentage). 60% - 70% is a typical figure, and C&R p503 has a graph for estimating values in
27

xn-1

y n

n+ 1

xn

particular cases. Different values can be applied to different sections of the column if data are available. Individual plate efficiencies are usually expressed as the Murphree efficiencies EmV or EmL based on vapour and liquid concentrations respectively. They are expressed as the ratio of the actual difference in mole fraction across the tray to that which would have been found if equilibrium had been achieved. Using the nomenclature of the diagram above: EmV = (6.1) where y*n is in equilibrium with xn and x*n is in equilibrium with yn The Murphree efficiencies refer to average concentrations in the vapour and liquid entering and leaving the tray. If liquid on the tray were perfectly mixed, its concentration would equal xn over the whole tray; however, there is usually some sort of concentration gradient, especially on larger trays. We can then define a point efficiency, Ep. At any location on tray n let the liquid composition be xn'. Ep is then given by:

y n y n +1 * y n y n +1

EmL =

x n 1 x n * x n 1 x n

Ep =

y n y n+ 1 y n* y n+ 1

(6.2)

where yn' and yn+1' represent the vapour compositions leaving and entering the selected point on the tray, and yn'* is in equilibrium with xn'. Question Ep is formally analogous to EmV. Is it possible to define a point efficiency analogous to EmL? Explain.
28

No, because liquid flows across a tray are not known. 6.3 Using the Murphree Efficiency

It is possible to re-draw the McCabe-Thiele diagram to allow for known values of the Murphree efficiency. Suppose, for instance, that in the top (rectifying) section of the column, the Murphree vapour efficiency, EmV, is constant from plate to plate at a value of 50%.

y1'* y1 y2' B

D C E A

x1'

x1 el el' F tol G

x
The distillate composition xD is given, and this must equal the vapour composition y1 leaving the top plate (point A). If the top plate were an equilibrium stage, the liquid leaving it would be in equilibrium with this vapour (point B) with a composition of x1. However, since the plate is only 50% efficient, the liquid leaving it will be richer in the mvc, with composition x1'. This composition can be found by locating the point C on the line AB such that the following equation holds:
y1 ' y 2 ' CE 1 (6.3) = = y1 '* y 2 ' DE 2 In the case of a 50% Murphree efficiency, C is the mid-point of DE. The vapour entering the top plate will lie on the standard operating line, since the operating line is based on a mass balance, and is not affected by plate efficiency. This gives point E, with the composition y2'. The process can then be repeated for the next real plate. E mV =

29

The net result is that the tol remains the same, but the equilibrium line, el, is replaced by an "effective equilibrium line", el', situated (in this case) half-way between el and tol. See the series of powerpoint slides titled Murphree efficiency worked example for more details.

30

6.4

Relation between efficiencies

6.4.1 Overall and Murphree efficiencies It is clear from the diagram above that the Murphree efficiency, even when it is constant from plate to plate, is not equal to the overall efficiency in the rectifying section. Two theoretical stages from the top composition (point A) takes us to point F. If the overall efficiency were 50%, there would be four real stages, but in fact four real stages with a Murphree efficiency of 50% only reach point G. To reach point F, the Murphree efficiency would have to be larger than 50%, ie EmV > E0 (6.4)

Repeating the calculations in the stripping section shows that in this case EmV < E0 A theoretical equation was derived by Lewis: (6.5)

E0 =

ln[1 + E mV ( 1) ] ln

(6.6)

where is the ratio of the gradients of the equilibrium line to the operating line. Equation 6.4 results when < 1 (OL steeper than EL) and equation 6.5 when > 1 (EL steeper than OL). This explains the different results in the two sections of the column.
6.4.2

Murphree and Point Efficiencies

For small diameter columns, the liquid on the plates is well mixed and the composition is equal at all points, hence the two efficiencies are equal. On larger plates liquid mixing in the direction of flow is incomplete and a concentration gradient exists across the plate. In the extreme case of plug flow, the concentration starts at xn-1 at the exit from the downcomer and ends at xn at the weir. The vapour at the liquid entrance meets a concentration higher than at the exit.
eqm. vapour compn. vapour compn.

y* n y n
(average value)

xn-1
liquid compn.

xn

31

Liquid flow

The diagram above assumes complete mixing of the vapour below the tray. The point efficiencies are significantly lower than the Murphree efficiency for the tray. Question When the point efficiency is high it is possible for the average vapour composition to be greater than the equilibrium value with xn. What happens to the value of EM in this case?
E M >1

Some progress has been made in predicting point efficiencies in terms of liquid properties using mass transfer models. These may then be entered into the hydrodynamic models to calculate Murphree efficiencies. However, in most practical work, empirical correlations for Murphree or overall efficiencies are used directly. Details may be found in specialised texts 6.5 Entrainment and Weeping

Entrainment occurs when droplets of liquid from a tray are carried with the vapour to the tray above. It occurs at high vapour flows, which also lead to increased pressure drop across the tray, and ultimately cause flooding, when the liquid backs up in the downcomer to the tray above. Weeping, as already mentioned, occurs when liquid drops through the holes in a sieve tray to the plate below. Both phenomena cause a drop in efficiency. More details may be found in C&R Vol 2, pages 497-500. (You may be asked to write brief notes on these in the exam. You will also need to do this for the discussion of the lab experiments on distillation) 7. Sidestreams and Multiple Feeds

A distillation column need not consist of one feed stream with one top and one bottom product. Particularly in multi component distillation (oil refining) product streams can be taken from any tray in the column (usually from the downcomer), although this is not very common in binary systems. However, it is possible to have multiple feed streams, of different compositions, even in binary distillation. This happens, for example, in a column recovering a solvent from a number of aqueous waste streams
32

from around a plant, or in a case where the feed comes from an initial flash unit. Instead of mixing feed streams of different compositions it can be more efficient to enter them on different trays (although this results in a slightly higher fabrication cost of column). Design of such columns can be performed using an extension of the McCabe-Thiels method. The same principles are applicable to both multiple sidestreams and multiple feeds. Essentially, at each point where a stream is added or removed, there will be a change in the liquid or vapour flow (or both). A different operating line will have to be drawn between each feed (sidestream) tray, and extra q-lines will have to be constructed (by the same method as used earlier) to determine their intersection points. Most sidestreams are taken as boiling liquids, which means the q-line is vertical. Stepping-off the plates then proceeds exactly as in the simple case treated above.

8. Multicomponent systems Although the graphical McCabe-Thiele method was designed for binary systems, it was extended to multi-component systems by Hengstebeck (1946), who took the two key components (heavy and light key - see H83MCS module) as an equivalent binary pair. The technique is not part of the syllabus for this module, but may prove useful in design projects. 9. Packed Distillation Columns Up to now we have only considered the use of trays in the column to effect contact between the liquid and vapour so that mass transfer can occur between the phases. The key physical characteristic of a tray column is that the separation takes place over a discrete number of stages. It is possible to achieve the same separation continuously, over the entire column height. For example, a vertical pipe with liquid flowing down the inside walls and vapour flowing up the centre will function adequately as a distillation column if the pipe were tall enough. This is used in several processes and known as a wetted wall column. The vertical pipe represents the simplest case, and as there is limited contact between the vapour and liquid then the column will need to be extremely tall to achieve most industrially relevant distillations. The column height can be reduced significantly if the contact area between liquid and vapour is increased, and this can be achieved with the use of packing materials. Packing affects the liquid flow in two ways to benefit the process: 1. Spread the liquid over a large surface area. 2. Slow down the liquid flow so it is in contact with the vapour for longer.
33

Packing can be random or in highly orientated, structured form. Details are given in Coulson & Richardson (Vol 6, p542), and will be covered in the Humidification section of this module. 9.1 Calculating the height of packed columns There are two methods to predict the packing height: Analogy with plates Transfer unit analysis The height of packing (Z) needed to enrich the vapour by the same amount as an equilibrium stage is called: Height Equivalent to a Theoretical Plate (HETP) Z = N HETP (9.1)

N is the number of theoretical equilibrium stages obtained using the McCabe-Thiele method. HETP values can be obtained from literature for a wide range of wetted wall columns, structured and random packing. The HETP approach can yields a good initial estimate of packed height during the preliminary design phase of a project. When more rigorous and detailed designs are required it is more accurate to use Transfer Units: Z = HOGNOG = HOLNOL (9.2)

See Coulson and Richardson 4th edition p.516 for derivation and definition of terms. NOG and NOL are the number of transfer units and HOG and HOL are the height of a transfer unit (HTU). The subscripts OG and OL refer to transfer units calculated on a gas and liquid basis respectively. The performance of a packed column is commonly represented by a simple term HOG or HOL. A low value of HOG or HOL corresponds to an efficient column, and is primarily influenced by the type of packing used and the vapour flowrate in the column. Further detail and analysis of the Transfer Unit approach will be given in the Humidification section of the course.
10.

Batch Distillation

All the analysis so far has been on continuous steady state systems. This is the basis of most large-scale industrial columns. However, some distillations are operated in batch mode, especially when the material to be distilled is produced in small batches, as often happens in the production of pharmaceuticals and fine chemicals, as well as spirits (whiskey, brandy, etc). (Note that "small" can mean up to a few tonnes). It is also used where flexible operation is required for varying feedstocks, eg solvent recovery.
34

condenser

accumulated top product

heat
The figure above shows a simple case where there is an initial charge in the still, it is boiled and the vapour is condensed. At any instant the top product is in equilibrium with the still composition. As the top product is richer in the mvc, the still composition, and hence also the top product composition, falls during the distillation, although all the top product is often mixed together in the receiver.. Let S be the number of moles in the still at any time, with composition x. If dS is vaporised, with composition y, a mass balance on the mvc gives :
xS ( x + dx )( S dS ) = ydS . ydS (evaporated)

(10.1) (10.2)

whence

( x y )dS = Sdx

Integrating

1 S dx ln( 1 ) = S0 xy x0

(10.3)

This is the Rayleigh equation. Combining this with the equilibrium relationship y = Kx, we can calculate the amount to be distilled if the final still composition is to be x1.

Question How can we then calculate the resulting top product composition for the batch process?

35

Do overall mass balances


D = S 0 S1 x D D = x D S 0 x1 S1

xD =

x0 S 0 x1 S1 S 0 S1

36

10.2

Batch Fractionation

The top product from simple distillation will be relatively weak in terms of mvc. It can be improved (increased) by putting a fractionating column above the still.
condenser

product receiver

heat
Since batch processes are usually small, packing is more common than plates in the column. But since the analysis for packing is more complex, we will look at a plate column. Assuming that the hold-up of liquid in the column is small compared to the still and it remains constant, the Rayleigh equation is still applicable. (How reasonable are these assumptions?)
ln
xS 1 dx S S1 = xS 0 x S0 D xS

(10.4)

xD and xS are not in equilibrium but depend on the separating power of the column. There are two modes of operation : 1. Constant top product composition - as the still composition falls the reflux ratio is increased to maintain a constant top product composition. Distillation stops when R gets too high. Constant reflux ratio - R is set to give a higher top composition than required initially, which then falls as the distillation proceeds. The
37

2.

process is stopped when the mixed top product has the required composition. The analysis for binary batch distillation can be done with a McCabe-Thiele diagram. The usual mass balance around the top of the column gives the same top operating line as in the continuous case. Question What about the bottom operating line? There isnt one there is no feed
10.2.1

Constant Top Product

xS1

xS2 R1

R2

x
Consider a column with 4 ideal stages plus reboiler. Since the top product composition, xD is fixed, we can draw a series of tols radiating from (xD,xD), and step off 5 stages along each of them to find the still bottom composition, xS which would give the required top product with the value of R corresponding to the tol used. Calculate R for each tol, and plot R against xS as sketched below. The initial value of R is the one corresponding to the initial still composition xS0.

38

xS1

xS0

xS

The problem is open ended, as we are told to increase R and stop "when R gets too high". This will be a matter of economics: the cost of the additional heat input needed against the value of the extra mvc recovered, but in an exam you will be told when to stop! Since the top product xD is constant the Rayleigh equation 10.4 can be integrated : x D xS 0 S1 ln = ln S0 x D x S1 (10.5) or
x S0 S 0 = x S1 S1 + x D ( S 0 S1 )

(10.6)

( x D x S1 ) S1 = ( x D x S0 ) S 0

Problem Write down the overall mass balance on the mvc over the entire distillation Initially in still = finally in still plus finally in product x S S 0 = x S S1 + Dx D
0 1

Can you show that this is equivalent to equation 10.6? Why should this be so? See mass balance top of page 32 - Rayleigh equation also driven from a mass balance A local xD=overall xD when xD is constant Constant Reflux Ratio

10.2.2

39

For this mode of operation the top composition falls as well as the still composition. Since xD is not a constant, we can only integrate the Rayleigh equation (10.4) when we have a relationship between xS and xD. Consider the same column with 5 ideal stages. From the McCabe-Thiele diagram pairs of values of xS and xD can be found using a set of operating lines of the same slope (R/R+1), ie, parallel lines. If the required accumulated top product composition, xD1 is known, you should start at a higher value, and select the gradient of the tol (ie the value of R) so that 5 stages reach a point slightly above the given initial still composition. (In an exam, you would normally be told the values of xD1 and R to use.) The Rayleigh equation (10.4) can then be integrated numerically, eg by plotting 1/(xD-xS) vs xS and finding the area under the curve between the initial and final still compositions. Question If we know the specification of the final mixed top product composition, how can we determine the final still composition xS1 ? Guess xS1 and integrate Rayleigh equation to get S1, substitute in overall mass balance on m.v.c to calculate sD and check whether this is the required value. Designing a batch distillation column is an iterative procedure. Because of the flexibility involved - mode of operation, reflux value, number of plates, - there is often a range of possible designs, not just one solution.

40

Vous aimerez peut-être aussi