You are on page 1of 8

Degrees of Freedom in Dynamic and Static Systems

D. Christopher Dixon
School of Chemical Engineering, University of N e w South I17ales, Kensington, S.S.IT7. 2033, Australia

Provided that the physical relationships governing a system are understood, the number of degrees of freedom can be determined by counting the boundary and geometrical variables and subtracting the number
of independent relationships among the boundary variables. The dynamic number of degrees of freedom
can exceed the static, the difference being equal to the number of independent hold-ups which are undetermined at steady state. Addition of control loops can make dynamic and static numbers of degrees of fteedom equal, if they were not previously, but not every control system changes the number of degrees of
freedom. A distillaticn column i s used as a special example, and the results of applying the present analysis
are compared with those of previous writers.

T h e r e has been some discussion in recent years of the determination of the number of degrees of freedom available in
steady-state design problems (Forsyth, 1970; Hoffman, 1964)
and in cont'rol syst'em design problems (Howard, 1967;
Yurrill, 1965). The accent has been on distillation columns,
aiid other multistage separation processes, since these are
among the more complex of commoii processes. The following
discussion will concentrate 011 the control syst,em design case,
which has some aspects which are not involved in the st'eadystate case and which does iiot appear to have been properly
treated in the literature. However, the approach to be presented is also applicable to steady-state design problems and
may be of interest as an alternative to previous methods in
this contest.
The number of degrees of freedom in a given problem is
equal to the number of variables involved minus the number
of independent equations (constraints) relating them. This
is the number of variables whose values may be arbitrarily
specified (within certain ranges aiid as functions of time in an
unsteady-state situation), thereby determining the values
of the other variables. The phase rule gives the number of
degrees of freedom required to fix the intensive state of a
tem a t (mechanical, thermal, aiid chemical) equilibrium. However, in the more general cases to be considered
here, open systems are involved and several types of constraint other than equilibrium. Thus, although the analysis
could be based on the phase rule with additional const'raints
(Gillilaiid aiid Reed, 1942), it seems simpler t'o start' from
first principles, lisbing all constraints without special attention
to those considered by t,he phase rule.
The determination of the iiumber of equations relating the
variables of a system requires a thorough understanding of
the physical relationships involved. It does not seem likely
that any method for determining the iiumber of degrees
of freedom of a system can be applied without this knowledge.
Hence, the best t,hat' can be expected is a method n-hich does
not require a detailed formulation of the syst'em equations.
Basic Analysis

The number of degrees of freedom equals the number of


variables minus the number of independent equations relating
them. The number of variables involved in a system can,
however, be very large, and in a control problem one usually
restricts attention to boundary variables. The reason for this
198 Ind. Eng. Chem. Fundam., Vol. 1 1, No. 2, 1972

is, of course, that the response of the system to externally


produced disturbances is the major concern, and every entering disturbance must disturb some of the boundary variables.
Thuq, an independent set of variables, equal in number to
the number of degrees of freedom, is chosen from among the
boundary variables, aiid these are commonly referred to as the
input variables.
Using this approach, the number of degrees of freedom,
or variance, V , is given by

where l y b a is the number of boundary variables and h-be is


t'he number of independent equations containing boundary
variables alone (i.e., iiot involving internal or est,ernal variables).
d boundary variable is a quaiit,ity which has a value a t a
point on the system boundary and which can vary under the
conditions being considered. d boundary quantity which is
physically capable of being varied (such as the temperature of
the feed to a reactor), but which is iiot varying under the
specified conditions, is not counted among Xbv.Thus, - Y b v is
not solely a property of the system but depends also on the
conditions being considered. If this constraint lvere iiot placed
on . v b v it' would be very large even for a simple
chemical engineering systems, the variables a t a point on the
bouiidary are usually some or all (depending on which are
varying) of those listed in Table 1.
: v b e does iiot include equations which involve internal
variables. These place no constraint8 on the boundary variables, since it has been decided not, to choose the degrees of
freedom from among the internal variables. External variables, also, are not included among the degrees of freedom,
since the system is t,he only portion of space which is beiiig
considered, by definit,ion. The t,ypes of equation which
commonly describe chemical engineering pyst,ems are also
listed in Tahle I. Relationships established by controllers
will be excluded for the present. Table I is iiot intended to
imply that a given equation, which is one of t,he t'ypes listed,
necessarily relates boundary variables alone. It often happens
that two equations have to be combined, thereby eliminating:
an internal variable, to obtain one boundary equation. This is
discussed furt,her belox.
The input variables are defined above as a set of iadependent boundary variables equal in number to the number of

degrees of freedom. The word "independent" is included in


this definition because it is not always possible to choose any
set of V boundary variables as the input variables. The
formal criterion is that there must be no equation Jvhich
cont'aiiis only input variables. For example, in a system having
three degrees of freedom and three material streams crossing
the boundary, the mass flow rates of the three streams cannot
be chosen as the input variables a t steady st,ate since they are
related by t'he total-mass balance equation.
There is a further aspect of the iiat'ure of degrees of freedom
which must be considered in some cases. Having chosen a
set of input variables, their values determine the values of t,he
internal variables and other boundary variables. However, it
is oft'en the case that the values of only some of the dependent
bouiidary variables are of interest. I n such cases, it can happen
that some of the input variables have no effect on these dependent variables. These "neutral" input variables represent degrees of freedom of the system since they are in fact
free t'o be arbit'rarily specified, but if one is only interested
in the number of ways in which the result of the process
can be influenced t'hey are not' relevant and would not be
counted.
A given input variable is neutral if it is possible to find a set
of independent boundary equations not containing t'his variable, but containing all the dependent boundary variables
of interest and containing a total number of dependent boundary variables equal to the number of equations. An input
variable which appears in none of the boundary equat,ions is
always neutral, no matter which boundary variables are of
interest.
I n a control problem, the geometrical parameters of bhe
system (e.g., number of plates and plate design in a distillation
column) are usually fixed. However, this is not necessarily so,
and the more sophisticat'ed control systems have variables
of t'his type.
This makes no difference to eq 1 and adds no additional
types of equation to Table I ; it simply provides additional
boundary variables and degrees of freedom. I t is quite natural
to think of such geometrical variables as boundary variables
for a control system, since some externally operable mechanism must be provided for performing the adjustments.
I n a steady-state design problem, however, all the geometrical paramet'ers are available for specification but would
not normally be considered bo be boundary variables. I n
this case, N b v in eq 1 equals the number of boundary variables
plus t'he number of variable geometrical parameters.
A further consideration, which arises in a control problem,
is that there are, in fact, two numbers of degrees of freedom
which need to be considered: st'at'ic, V,, and dynamic, V d .
The static (or steady-state) number of degrees of freedom
is the value of V under the constraint' that the system is a t
steady state, while this constraint' does not apply in evaluating Vd. Thus, eq 1 can be replaced by t'wo equations

v,=
v d

Nbv
S b v

- Arbes

(2)

(3)

S b e

where N b e s is the value of Aibe when the syst'em is a t steadystate.


v d can be greater than V,: that is, S h e can be less than Shes.
This can arise because the balance equations contain accumulation terms under dynamic conditions, but not a t steady
state. If a variable which determines the accumulation term
in a balance equatioii is not related to the boundary variables,
then the balance equation will be couiit'ed among A'ber,
but not among N b e .(Balance equations are the only equations

Table 1. Usual Boundary Variables and Equations


in Chemical Engineering Systems
a . Boundary Variables
Intensive properties

Composition
Temperature
Pressure
Or another set of (W 1) intensive properties, where A' is the
number of chemical compounds present.

Rates of Flow across the Boundary

Material
Heat
Shaft work ( L e , , all work except flow work).
(Flow work is not listed separately, because it ib determined
by the flow rate and pressure of the material.)
b. Equations
1.

Balance

input
output
consumption
accumulation
rate
rate
rate
rate
nhere t = time
2 = amount in the system = hold-up
(a) Material
(b) Energy (total)
(c) llechanical energy or momentum
2.

Equilibrium

Rote

or

(a) Mechanical

hfechanical energy consumption


(friction, etc.) or momentum
consumption (all opposing forces)
(b) Thermal
Heat transfer
(c) Phase
Ilaterial transfer
(d) Reaction
Reaction
(At a given point, for a given type of process, either an
equilibrium or a rate equation applies, bnt not both.)

in Table I which are differential in time, and so are the only


ones which contain terms under dynamic conditions which
are absent a t steady state. Reaction rate equations, for example, are often stated to be differential equations in time,
but they are not in fact (Dison, 1970).)
Any relationship which exists between a hold-up determining variable and the boundary variables will exist under
both dynamic and static conditions. The accumulation term
in a dynamic balance equation gives the change in the hold-up
(of whatever entity the balance is accounting for) from some
initial value. Thus, if the system is imagined to start from a
steady state, the hold-up is subsequently determined only
if it was determined a t steady state, that is, if the hold-up
has a unique value a t steady state for a given set of values
of the boundary variables. (A further point, which is required
later, is that the only variables which can be undetermined
a t steady state are hold-up variables. Every variable must
be affected by the boundary variables under dynamic or
static conditions, if not both; otherwise it would not vary
and so would not be a variable. Since the only differences
between dynamic and static equations are the accumulation
terms in the balance equations, hold-up variables are the
only ones which can vary under dynamic conditions while
having no unique steady-state values.)
Hence the simplest approach appears to be to determine
V , from eq 2, and then lid from
v d

v,= 5,"

Ind. Eng. Chem. Fundam., Val. 1 1 , No. 2, 1972

(4)
199

a;.voi. f l o w

constant pressure

cd oe n ssti at y
I rquld

I
I
I

L - - - -

syslem

boundory 2 - - - - J

00' Qo(h)

' b . P u m p e d - d i s c h a r g e Tonk

liquid

c o n s t a n t - speed

0, i n d e p e n d e n l o f h

Heat Exchanger

constant
comp0:ition

Liquid A

d. Isothermal Reactor
Feed

product
overf lo w

perfectly mixed

Figure 1.

Simple examples

n-here S,,is the number of independent hold-ups which are


undetermiiied a t steady state. K h e n V dexceeds V , the number of input variables is usually taken to be the former.
I n a system for which Vd = V,, a steady state exists for
any set of constant values of the input variables (within
operating ianges). Such a system nil1 be referred to as sdfregulatory, although there is no guarantee, of course, that any
such steady qtate is stable. For a system which is riot selfregulatory, there will be no steady state corresponding to an
arbitrary set of values of the input variables, but only for
certain discrete sets Such a system would not normally be
coiisidered satisfactory for contiol purposes.
I t must be pointed out here that the possibility of Vd
being greater than V,ariyes from the way in k\hich the problem has been approached i n termi of boundary variables.
When
exceeds V,, there are in fact A,' additional degrees
200

of freedom available a t steady state, b u t they cannot be


fixed by boundary variables. These degrees of freedom can
be fixed by a set of Ai,, internal variables, one for each independent undetermined hold-up. However, it appears t h a t in
no case is any of these degrees of freedom of any use or relevance. They do not affect the relation between entering and
leaving material streams and they cannot be directly fixed
from outside the system. Hence, the best approach is to not
count these degrees of freedom, which is what is done b y the
above method of analysis.
Having decided that only the number of steady-state
boundary equations, .?-be%, need be directly determined, and
not h i b e , some further remarks can be made about the equations in Table I. I n balance equations, the accumulation term
will be zero, since only the steady-state case is considered.
Since the input and output rates will always be described
by boundary variables, every balance equation in which
the consumption term is zero will be counted among x b e s .
Thus, the total-mass balance, mass balances on nonreacting
compounds, and the energy balance will always be boundary
equations.
I n balance equations in which there is a nonzero consumption term, the appropriate rate equation will be required to
determine the consumption. If the rate is determined by the
boundary variables then the combination of the balance and
rate equations will be a boundary equation. For example, in a
mass balance equation for a reacting compound, the reaction
rate expression is substituted into the consumption term.
If the concentrations, temperature, etc., which determine the
reaction rate are related to the boundary variables, then
the material balance-reaction rate combination is a boundary
equation.
Figure 1 shows four simple examples which have been
chosen for illustration. The analyses are given in Table 11.
For Figure la, it is easily seen b y inspection that the result
Vd = V , = 1 is correct. Qlcan be varied arbitrarily (within
limits imposed by the height of the tank) under both dynamic
and static conditions, and this determines the variation in
Qo. The mechanical energy balance, with appropriate friction
loss expressions substituted into the consumption term,
does not relate boundary variables in this case. The friction
losses depend on Qi and Q o (boundary variables) but also on
h. There is no independent equation relating h to the boundary
variables, and the mechanical energy balance equation reduces,
in fact, to the relation between Qo and h. I n effect, the mechanical energy of the inlet stream is dissipated by impact on the
surface in the tank, no matter what the value of h, and so
there is no connection between inlet and outlet mechanical
energies.
Another point about this example is that it contains a iiumber of irrelevant degrees of freedom. The liquid density is
constant and Q o depends only on h, which implies that the
viscosity is also constant. However, subject to these two
constraints, the composition and temperature could vary,
1 - 2 = (.V - 2) additional boundary
giving (*V - 1)
variables for both inlet and outlet streams. The energy
balance equation plus (S - 3) compound mass balance
equatioiis (t1F-o of the (S - 1) compound balances are dependent on the rest, because of the density and viscosity
constraints) provide additional boundary equations, leaving
(A' - 2 ) additional degrees of freedom. These additional
degrees of freedom are easily seen to be neutral, if Qo is the
only boundary variable of interest, and they are not listed
in Table I. However, this can also be shown by the formal
procedure given above.

nT/-f+FL/'o

a. G r a v i t y - d i s c h a r g e T o n k

Ind. Eng. Chem. Fundam., Vol. 11, No. 2, 1972

Firstly, a permissible choice of the input variables is Q L


plus inlet temperature and ('V - 3) inlet composition variables, since there is no equation relating these. With this
choice of input variables, it is found that the total mass
balance equation (Q1 = Qo) contains only one dependent
variable (Q,,), and this is the boundary variable of interest,
but the only input variable contained in it is Ql. Hence, the
other input variables are neutral.
Figure l b illustrates what appears to be the only common
situation where Vd exceeds VB;that is, the situation where the
in- and out-flows of a tank containing a phase boundary are
unaffected by the position of the boundary. Compared to
Figure l a , the mechanical energy balance-friction loss equation now relates boundary variables a t steady state, since
there is now, in effect, an independent equation for h (k.,
h
may be arbitrarily specified a t steady state). This extra equation (which reduces to Qo = constant, as shown in Table 11)
reduces V , by one to zero. However, since h is undetermined
a t steady state, v d - V , = 1, so that Vd = 1, the same as
for Figure la. It is easily seen t h a t Q , can be arbitrarily
varied under dynamic conditions but can only have one value
a t steady state.
If the pump in Figure l b were centrifugal, rather than a
positive displacement type, Q,, would be affected to a small
extent by h and so the system would be equivalent to Figure
la. However, the range over which Q i could be varied, while
still allowing a steady state, would be small; that is, the system
would be self-regulatory over a small range only. I n the limit
as the effect of h on Qo approaches zero, the self-regulatory
range approaches zero and Vd - V , becomes unity. Thus, if
one says that the effect of h on Qo, while not zero, is negligible,
which makes v d - Vs = 1, this is equivalent to assuming
that the self-regulatory range of the system is negligibly
small. Even for Figure l a the self-regulatory range could be
negligible under the conditions being considered. If the range
of variation of Qo, which is allowed by the height of the tank,
is very small compared to the variations in Qlwhich are expected, then the effect of h on Q o is negligible, in effect, and the
system is equivalent t o Figure l b .
T h e heat transfer rate equation is included among Nbe.
for the heat exchanger example. The total rate of heat tiansfer
is given by the inlet and outlet temperatures and flow rate
of either stream, and the log mean temperature difference
is given by the inlet and outlet temperatures. Thus, the total
heat transfer rate equation can be expressed in terms of
boundary variables.
I n the reactor example, the volume of the reactor contents
has been assumed constant. Hence, the total rate of reaction
for each compound depends only on the composition of the
contents, which is the same as the exit concentration (boundary variable). Thus, the S compound balance-reaction rate
equations are boundary equations. If the variation of the
volume with flow (because of the overflow product-removal
system) were taken into account, the relation between depth
and outflow (as in Figure l a ) would determine the contents
volume in terms of boundary variables, and would be used
to eliminate the volume from each compound equation,
giving the same result for N b e s .
Effect of Control loops

When a control loop (feedback or feedforward) is added


to a system (which means that it is placed inside the defined system boundary), it establishes a relationship
between a valve setting or other manipulable variable and
another variable (the controller actuating variable), if the

Table II. Analysis of Figure 1 Examples


Gravity-Discharge Tank

a.

Qo (valve setting does not vary, because the stem does not cross the
boundary)
Nbej
1
Material balance
V s = 2 - 1 = 1 E.g., inlet flon- rate
Liquid hold-up is determined by h , which
Vd - V , = 0
is related to Qo
=

s b v

Qi,

b.

Sb,.=

-1=0

V,

I'd

- V, = 1

Pumped-Discharge Tank

Q i (Qo is constant)
Material balance
Q i must equal Qo a t steady state
Liquid accuniuhtioii is determined by h ,
which is not fixed by the boundary
variables
C.

xb,. =

A-beS

12
6

V 3 = 12 - 6
Vd

- T.', = 0

d.
I\lhv

2 5

Shes =

+1

a\-

Vq = 2 5

+1-

I'd - T',

s = *\- + 1
0

Heat Exchanger

Flow, pressure, and temperature for each


entering and leaving stream
2 x material balance
1 X energy balance
2 X mechanical energy balance-friction
loss
1 X heat transfer rate
6 E.g., floxv,pressure and temperature of
each inlet stream
Temperatures determine energy hold-up,
p r e s u r e s determine liquid hold-up;
all are fixed a t steady state
Isothermal Reactor

Compo..ition and flow rate of feed and


product; temperature
llaterial balance-reactioii rate for each
compound
E.g., compohition, flon- and temperature
of the feed
Material accumulation is dfterniined by
reactor composition, which equal3
product composition

adjustable parameters of the controller (set-point, etc.)


are fixed. The general method of analysis given above can,
of course, be applied direct'ly t,o systems containiiig coiitrol
loops. However, to examine specifically the effect of control
loops, t,he change in the number of degrees of freedom resulting from their addition will be considered.
For the purposes of t,his discussion, it will be assumed
that the manipulable variable exists in the system before
the control loop is added. Thus, the manipulable variable
is always a boundary variable since, before the controller is
added, it can only be varied from outside the system. If the
manipulable variable does not already exist, it must be added
to the system, as a preliminary to adding the control loop.
If t,his can be accomplished by providing for the direct variation of some quantity which was previously fixed, then this
adds one extra degree of freedom. The most common case is
the addition of a control valve to a pipeline. This gives the
line a variable, instead of fixed, flow resistance.
I n the case of a feedback control loop containing integral
action (it is assumed that integral action is never used in
feedforward loops), the controller itself does not determine
the manipulable variable as a function of the actuating
variable a t steady state. However, it ensures that steady
state can only exist when the actuat,ing variable has a particuInd. Eng. Chem. Fundam., Val. 11, No. 2, 1972

201

a. G r a v i t y - d i s c h a r g e T a n k , w i t h L e v e l C o n t r o l

oi

_ _ _ _ _ _ _ _ _ _ _ _1

system b o u n d a r y /f

b. G r a v i t y - d i s c h a r g e T a n k , w i t h F l o w C o n t r o l

%---

-,<,p;

c. P u m p e d - d i s c h a r g e Tank, w i t h L e v e l C o n t r o l

Figure 2.

Simple control systems

lar value, and this places a constraint on the manipulable


variable through the system relationships. Since the effect
of integral action is to give the controller an infinite steadystate gain, which removes offset in the controlled (actuating)
variable, the simplest approach to an integral controller
appears to be to determine its effect without integral action
and to then deterniiiie if any further constraint results from
letting the steady-state gain become very large (or constraining the steady-state values of the actuating variable to
a very small range).
Thus, the problem is to determine the effect on the number
of degrees of freedom of making the manipulable variable a
function of the actuating variable, with a large gain a t steadystate when integral action is used. The result depends on
whether the actuating variable is an undetermined hold-up
variable or iiot 111 the original system.
If the actuating variable is undetermined a t steady state
ill the original system, then the manipulable variable after
coniiection through the controller is also undetermined, with
finite or infinite controller gain. Therefore, V . is not changed.
However, the manipulable variable is a boundary variable
and the actuating variable is related to it in the new system
so that Sa,is reduced by one. Therefore, vd is reduced by one.
If the actuating variable is determined (related to boundary
variables) 111 the original system, then the controller fixes the
maiiipulable variable as a function of these boundary variables for both dynamic aiid static conditions, aiid so V.
and Vd are both reduced by one. When integral action fixes
202 Ind. Eng. Chem. Fundom., Vol. 1 1 , No. 2, 1972

the actuating variable at steady state this can have a further


effect by making some degrees of freedom neutral. Depending
on the circumstances, these degrees of freedom may be
irrelevant, and so V , is further reduced. There does not appear
to be any specific rule for determining whether degrees of
freedom are made neutral by integral action, and each case
must be analyzed by the general method given above.
These effects caii be illustrated for the two liquid-holding
tanks of Figure l a and b. Figure 2a shows the gravitydischarge tank (Figure la) with a level controller. The
system boundary has been drawn to cross the valve stem,
indicating that the valve setting caii now be varied arbitrarily
before addition of the controller. The controller removes
this new degree of freedom, since the actuating variable (h)
was previously determined, and v d = 7,= 1, the same as
for Figure la. Closely controlling h has no additional effect.
Q , is still a permissible choice as iiiput variable and will have
an effect even if h is fixed. Thus, adding the control valve
and controller has iiot affected the degrees of freedom in this
case. It has, however, increased the self-regulatory range of
the system. A steady state can exist for a wider range of
values of Q>than previously. (The arrow indicating the controller set-point (sap.)does not cross the system boundaries,
which indicates that the set-point is fixed.)
Figure 2b shows another control system for the gravitydischarge tank. Again the actuating variable (Qo) was previously determined a t steady state and so the controller
reduces Vd and V , by one, canceling the extra degree of
freedom from the control valve. However, in this case the
self-regulatory range has been reduced, and if Qo is closely
controlled Q , has only one value for steady state, and the
result is Vd = 1, V , = 0. Looking a t this in another way,
with close control Q I is no longer a permissible choice for the
input variable a t steady state, but the valve position can be
chosen. The valve position may be arbitrarily specified a t
steady state. This merely determines the value of h required
so that Qo has the required value. However, the valve position
then represents an irrelevant degree of freedom since it only
affects h , and not Q,,,leading again to the result v d = 1,
V 8 = 0. With close control of the flow, the system is equivalent to Figure l b .
Figure 2c shows a level controller applied to the pumpeddischarge tank (Figure lb), which originally had Vd = 1,
V, = 0. The addition of the control valve has added an
extra degree of freedom. Since the actuating variable was
previously undetermined a t steady state, adding the controller does not affect V. but reduces Vd by one. The filial
result is Vd = Ti, = 1, and the system is now self-regulatory
and equivalent to Figure l a .
In the above discussion, adjustable controller parameters
have been assumed fixed. If any one of these is variable, this
adds an extra degree of freedom in all cases.
The above discussion has also been limited to control loops
which are added to the system. A control loop which is external to the system (because of the defiiiition of the system
bouiidary) can reduce the number of degreeg of freedom by
removing (nearly) the variatioii in a boundary variable.
Vd and V. are reduced by one, simply by reduciiig lb,.
From the above d i x u w o n , the statement (lIurri11, 1965)
that the number of iiideloendeiitly-actiug controllers which
may be added to any system cannot eweed the number of
degrees of freedom which are iiiherent in the system requires some qualification. K h e n Vd exceeds
the former is
the number of degrees of freedom which applies to this rule.
Assuming that the final Eystem is required to be self-regu-

v.,

latory, S,,loops must be actuated b y previously undetermined hold-up variables. A maximum of V , other loops may
be added which are either external loops holding boundary
variables constant or internal loops actuated by previously
determined variables. The manipulable variables for all
iiiternal loops must exist before the rule is applied.
Distillation Columns

As remarked in the introduction, distillation columns are


amoiig the more complex systems and have received considerable attention in connection with the present subject.
There is some disagreement between different writers.
Figure 3 shows a basic two-product distillation column,
as has been considered by several writers. Showing the heat
transfer rat.es, for condenser and reboiler, esplicity by arrows
implies that the system boundary is being taken as coincident
with the heat transfer surfaces, rather than as passing through
the water and steam inlet and outlet pipes. The degrees-offreedom analysis of this system is summarized in Table 111.
I n Table 111, S phase equilibrium relations ( L e . , equality
of chemical potentials between phases for each compound)
are shown. The top and bottom products are not, of course,
in equilibrium with each other (unless the column contains
only one ideal stage). However, the equilibrium relationships
between the phases on each plate through the column give,
overall, one set of iV equations relating top and bottom
compositions. A similar comment applies to the mechanical
and thermal equilibrium relationships listed. The assumption
of equilibrium is only an idealization, of course. However,
even if this assumption is not made, the same number of
rate equations will be obtained, instead of the equilibrium
equations.
The inclusion of the mechanical equilibrium and mechanical energy balance-friction loss equations is only permissible
because of the assumed poor self-regulation associated with
the two accumulators. The levels in these accumulators are
assumed to have negligible effect on product flows ( c j .
discussion of Figures l a and b) so that the friction losses in
the accumulators are related to product flows, and the discharge pressures of the two products are related to the reboiler and condenser pressures, and hence to each other via
the equilibrium relations through the column. The liquid levels
on the trays are assumed sufficiently variable to give adequate regulation, and these are determined independently
of the overall mechanical energy balance-friction loss equation by the internal refluxing arrangements.
I n the above analysis, the number of plates in the column
has been taken as fixed, which is the usual situation in a
control system. In a design problem, the numbers of theoretical plates i n the rectifying and stripping sections are not
fixed and provide two additional degrees of freedom. If
real plates are considered, there are numerous additional
degrees of freedom in the plate geometrical details, unless a
fixed plate design and spacing is used.
The analysis in Table 111can be compared with that of other
writers, for the design problem where the numbers of theoretical plates are variable. I n this case, the present analysis
gives V , = N
6. If the feed composition, temperature,
flow rate, a i d pressure (or the column pressure) are specified, four degrees of freedom remain, which agrees n-ith the
result of Kwauk (1956) for a biliary system and of Forsyth
(19i0) for a quinary system. (Xeither of these writers states
esplicit,ly lvhet,her the accumulators are self-regulatory or not.
However, by implicat,ion they consider them nonself-regulatory, which is the present case. They make the usual assump-

Table 111. Analysis of Distillation Column


of Figure 3
Boundary variables

S
S
2
2

+2
+2
+2

35

Feed composition, temperature, presure, flon


Top composition, temperature, pressure, flow
Bottom composition, temperature, pressuie, f l o ~
Reboiler heat flow and steam temperature
Condenser heat flow and steam temperature
Total

+ 10

Steady-state boundary relations

s
1
1
1
1

s
2

Xaterial balance
Energy balance
l\Iechanical energy balance-friction loXeciianical equilibrium
Thermal equilibrium
Phase equilibrium
Heat transfer rate-reboiler and condenser
Total
3-Y 10 - ( 2 s 6) = S 4
T, = 2 Lelels in reboiler and condenser are assumed to h a r e negligible effect on product
flow rates

cq
5.
+
TVd

=
=

s y s t e m boundary

condenser

I
I
I

reflux

t o p product

feed I
N compounds

negligible
heat loss

>

I
I
I

Figure 3.

Basic distillation column

tion of uniform column pressure, or all pressures determined.


If the accumulator levels have a significant effect, this assumption cannot be made.)
Hoffman (1964), however, maintains that there are only
three degrees of freedom for this system, but this apparent
disparity (Forsyth 1970) can be explained. First, Hoffman
does not include the two numbers of theoretical plates among
the degrees of freedom, since the numbers of plates must be
integral. Strictly speaking, this is correct, since a variable
which is a degree of freedom should be continuously variable
over its allowable range. However, this is really only a
matter of definition, the practical point being that the numbers of plates are variable. K i t h this restriction, the number
of degrees of freedom is reduced from 4 to 2 .
Secondly, Hoffman does not include the pressure in specifying the feed. This gives a n extra degree of freedom, and a
final number of 3. There is, of course, no reason why the
feed pressure should be specified. This simply leaves some
other variable (such as the pressure a t the top of the column)
free for specification. Thus, the only way in which Hoffmans
analysis differs from others is in not counting the numbers
of plates as degrees of freedom.
Ind. Eng. Chem. Fundam., Vol. 1 1 , No. 2, 1972

203

system b o u n d a r y

there are four additional boundary relations; the material


balance and mechanical energy balance-friction loss equations for water and steam. Thus, Figure 4 has 13 - 4 = 9
additional degrees of freedom.
T o simplify the analysis, the inlet temperatures and pressures, and discharge pressures, will be taken as constant for
water and steam. There are then three additional degrees of
freedom in Figure 4 (three of the five valves, the other two
replacing q. and pc, in effect), g i v i n g v , = $
7 and V d =
N 9.
Since Vd is greater than V., the first priority usually is to
reduce V d so that it equals V,. From the previous discussion
it is seen immediately that this can be done by actuating two
of the valves from the levels in the condenser and reboiler.
One such set of connections is shown in Figure 5 , together
with other control. loops considered, as an example, by Murrill
and by Howard. The two level controllers are not required
to closely control the levels, but only to keep them within
limits, thus achieving material-balance control (Buckley,
1964).
If the top product valve, say, was used in a product flow
control system, instead of in the material-balance system
shown, then Vd and V , would both be reduced by 1 and their
difference would remain unchanged. As was found in some
early installations, such a system does not work satisfactorily.
The other control loops affect V d and V , equally, each overall
loop reducing Vd and Ti, by one. I n particular, the cascade
loop (TRC2 and FRC3) which adjusts the reflus valve to
control the top temperature reduces the number of degrees
of freedom by one only, not two, as has been stated sometimes.
The inner loop (FRC3) relates the reflux valve setting and the
reflux flowv,but introduces an additional variable in the set
point. Hence, the inner loop does not affect the degrees of
freedom. The outer loop relates the top temperature to the
inner controller set point, and so reduces the degrees of freedom by one. Thus, the complete cascade system reduces
the degrees of freedom by one. Thus, the controllers, other
than the material balance controllers, reduce the static degrees of freedom by 5, leaving VI^ = V . = A
7 - 5 =
N
2.
This is two less than the number of degrees of freedom for
the column without control valves and controllers (Figure 3).
A reduction of four is due to the fact that the feed flow and
temperature and the steam and water flows are controlled,
and an increase of t x o is due to the addition of the two valves
in the material balance loops (these loops do not affect X b e r ) .
The reflux control loop has no net effect on the number of
degrees of freedom compared to Figure 3, because a degree of
freedom has been added via the valve, but removed by the
control loops. The remaining ( N
2) degrees of freedom
could be specified, for example, by the feed composition and
pressure and the pressures a t the points of discharge of the
two products. I t can be seen intuitively that the feed composition and pressure can, in fact, be varied arbitrarily.
Also, if the discharge pressure of the top product is changed,
the material-balance control will adjust the valve to restore
the product flow, and nothing else in the system vi111 be
affected (except an adjustment of the reflus valve to compensate for the effect of a change in accumulator level).
A similar situation applies for the bottom product discharge
pressure, and so both discharge pressures may be arbitrarily
specified.
The above analysis disagrees entirely iyith that of AIurrill
(1965), who concludes that Figure 5 has zero degrees of freedom. The differences in his analysis are: (a) he does not recog-

Figure

4.

Distillation column with control valves

m-I-,

--r

~l I !

U
s y s t e m boundary

L---_/---

Figure

5. Distillation column with control system

Although Kwauk states in the introduction to his paper that


there are four degrees of freedom for the present case, his
following analysis actually gives the answer five (as obtained
by Smith (1963), using Kwauks method). The answer four
is oiily obtained when q, is filed, which is a different case.
The extra degree of freedom results from overlooking a
mechanical equilibrium condition which must be satisfied
a t the point TT here the feed enters the feedplate. K h e n two
fluids mix (in a T-section, say), their pressures are equal a t
the point where they meet. Thus, if both flow rates and one
inlet pressure are specified, the inlet pressure of the other is
determined. Hence, a feed tray has (S I ) , not (h
2),
degrees of freedom more than an ordinary tray, resulting from
the addition of the extra inlet stream. This omission also
occurs in the analyses of Gillilaiid and Reed (1942) and Howard (1967). For dynamic conditions, Howard counts two
extra degrees of freedom, in agreement with Table 111.
The effect of adding a standard set of control loops will
nom be considered. Usually, five control valves are added, as
shown in Figure 4. Compared to the column in Figure 3,
Figure 4 has 13 additional boundary variables. q., qr, and the
mater and steam temperatures have been removed, but 17
boundary variables have been added in the fire valve settings,
and 4 x 3 = 12 variables in the pressure, temperature, and
flow a t inlet and outlet for the water and steam. However.

204 Ind. Eng. Chem. Fundam., Vol. 11, No. 2, 1972

nize that the addition of the two control valves in the material balance loops increases V , by 2; (b) he concludes that the
reflux controller plus valve removes -4 degrees of freedom,
instead of zero. H e states that TRC-2 establishes a relationship between the feed composition and temperature (Ar - 1
degrees of freedom) and the inner loop FRC3 also removes
one degree of freedom.
Howard (1967) obtains AT degrees of freedom. He counts
the seven controllers as removing seven dynamic degrees of
freedom, in agreement with the present analysis, but does
not count the three extra degrees of freedom from the reflux
valve and the material balance valves. Allowing for the extra
degree of freedom counted before adding the valves aiid
controllers (from omitting the feedplate mechanical equilibrium condition), the net result is two degrees of freedom
less than the present analysis.
Conclusions

trollers: (a) adding a control valve adds a degree of freedom;


(b) adding a control loop reduces V d and V, by one, if the
set point is fixed, and if the controller actuating variable is
not an accumulation variable previously undetermined a t
steady state; if the controller actuating variable was previously undetermined a t steady-state, then the control loop
reduces Vd by 1, but leaves V , unchanged; (c) integral action
can also make a degree of freedom neutral, in some cases,
when the actuating variable !vas originally determined a t
steady state.
literature Cited

Buckley, P. S., Techniques of Process Control, p 99 ff, Wiley,

V
P_
Ynrk
~_.
l\j. Y
1964
. _
.
---Dison, D. C.Cheni. Eng. Sci. 2 5 , 337 (1970).
Forsyth, J. S.> IND.
EKG.CHEX, FCSDIM. 9, 307 (1970).
Gilliland, E. R., Reed, C. E., Znd. Eng. Chem. 34,551 (1942).
Hoffman. E. J.. Azeotrooic and Extractive Distillation.
pp 10-15, Interscience, Ke& York, N . Y., 1964.
Howard, G. AI., IKD.ENG.CHEY.,FCNDAM.
6, 86 (1967).
Kwauk, M., AI.Z.Ch.E.J . 2 , 240 (1936).
AIurrill, P. W., Hydrocarbon Process. 44, 143 (1963).
Smith, B. I)., Design of Equilibrium Stage Processes, p 84,
~

~~

From the above discussion, it is concluded that the following procedure can be used for determining the number of
degrees of freedom of a system, for both steady-state design
problems aiid control problems. (I) 1-iicontrolled system:
(a) steady-state degrees of freedom, V,, from eq 2 ; (b) dynamic degrees of freedom, Vd, from eq 4. ( 2 ) Effect of con-

McGraw-Hill, Ne%-York, N . Y., 1963.

RECEIVED
for review February 1, 1971
;ICCEPTI:D
November 27, 1971

Dissolution of a Porous Matrix


a Slowly Reacting Acid

by

William E. Sinex, Jr.,l and Robert S. Schechter*


Department of Chemical Engineering, The C-niversity of Texas, Austin, Texas 78712

I. Harold Silberberg
Texas Petroleum Research Committee, The Vniversity of Texas, Austin, Texas ?87lS

The acid treatment of an oil well to increase its productivity i s commonly practiced; however, at the present
time there i s nc proven method to guide the design of such a process. This research examines the ability of a
previously proposed model to predict the changes in a porous matrix when invaded by a slowly reactive
fluid which dissolves a portion of the solid. The model i s shcwn to predict a relationship between the increase
in porosity and the permeability which i s not precisely unique, as it depends to some extent on the initial pore
size distribution, but for the initial distributions tested the permeabilities were found to lie in a narrow band.
These results are independent of any parameters defining the kinetics except that the reaction be slow. It i s
shown experimentally that the reaction of ferric citrate in the presence of citric acid with porous bronze disks
satisfies the condition of being a slow reaction. The permeability change of the porous bronze disks i s found
to agree closely with the theoretical predictions.

o n e method of stimulating oil wells to greater production


is to dissolve a portion of the oil-bearing rock with an acid,
thereby decreasing the resistance offered by the rock to the
flow of oil. About 8 i million gallons of hydrochloric acid are
used ailnually to stimulate oil viells in carbonate formations
Present address, Fluor Corporation, P.O. Box 35000, Howt,on, Tesnr.

(Hurst, 1970). I n addition, many gallons of hydrofluoric,


acetic, formic, aiid other special purpose acids are also used.
The process of matris acid treatment is basically a simple
one. An acid is pumped down the wellbore of an oil well a t
rates 1Thich are slow enough to aroid fracturing the rock. The
acid invades the oil-bearing formation, displacing the resident
fluids and a t the same time dissolving a portion of the rock.
The distance that the acid penetrates depends on the flow
Ind. Eng. Chem. Fundom., Vol. 1 1 , No. 2, 1972

205