DStext 2015

DEQ/FEUP
INTRODUCTION TO SYSTEMS DYNAMICS
J. M. LOUREIRO
(Prof. Associado, FEUP)
LSRE - Laboratory of Separation and Reaction Engineering
Department of Chemical Engineering, FEUP
University of Porto, Portugal
February 2015
Introduction to Systems Dynamics / J. M. Loureiro - 2015 2
1. INTRODUCTION
The systematic study of the dynamic behavior of systems is the object of Systems
Dynamics. Since the dynamic behavior is to be studied, in Systems Dynamics the
accumulation terms (derivatives in order to time) are never dropped from the model equations
(balances) and then differential equations (ordinary or partial) are obtained. The main
application of Systems Dynamics is Process Control; the effects of perturbations (on input
variables) in the systems responses are studied, in order to devise control actions which are
able to react to undesired perturbations on the system.
According to their complexity (which is related with their dynamic behavior) systems
are first classified into categories. The main division is into two classes of systems:
• Lumped parameter systems - the intensive properties are supposed to be independent
of position within an extensive property (commonly the volume) of the system, i.e., intensive
properties within the system change only with time. In this case, ordinary differential
equations (ODE's) are obtained. According to the order of the model ODE's, the system itself
is called of first order, second order, etc.
• Distributed parameter systems - the intensive properties change with position, as well
as with time, within the system. In this case, partial differential equations (PDE's) are obtained
for the system model.
In the following, some examples will be given of both systems classes with the main
objective of clarifying what is their expected dynamic behavior.
Before going any further, let us define a general framework for balances. For any
conservative property (mass, energy, momentum, etc.) in any system it can be said that:
(amount in) = (amount out) + (accumulation) (1.1)
The foregoing equation is valid for total mass; we know (since Lavoisier) that, although
total mass is conserved, the same is not always true for the mass of a given component, since
if a chemical reaction is going on inside the system, the component can be consumed (if it is a
reactant) or produced (if it is a product). Then, a more useful way of writing the general
conservation balance, including component mass balances when chemical reactions are
present, is the following:
(amount in) + (production) = (amount out) + (accumulation) (1.2)
This general equation will be applied throughout in the following, wherever a system
model is needed, since models are nothing else but the summing up of several equations,
mathematically translating the physical phenomena involved (with given approximation):
• conservation equations;
• equilibrium relations;
• kinetic (physical, chemical) rate expressions;
• initial and boundary conditions.
2. LUMPED PARAMETER SYSTEMS

In the following, first and second order lumped parameter systems will be introduced
through examples.
2.1. FIRST ORDER SYSTEMS

A system is said to be of first order when its model equation is of first order. Let us
consider a general linear first order ordinary differential equation(1) with constant coefficients:
d y( t )
+ a y( t ) = b x ( t ) (2.1)
dt
In this equation, assume that x(t) and y(t) are deviation variables (see Appendix 2) and
that y(t) is the response of a (first order) system to a given perturbation x(t) in the system inlet,
where t is time; a and b are constants.
Taking the Laplace transform of eq. (2.1), we obtain:
s y( s) + a y( s) = b x( s) (2.2)
It is then easy to obtain the transfer function G(s) (see Appendix 2) of the (arbitrary)
first order system:
y ( s) b b/a
G( s ) = = = (2.3)
x( s) s + a s / a + 1
The degree of the polynomial in the denominator is always equal to the order of the
system. In this form (when the independent term is equal to one - it can always be made one),
the denominator of the transfer function is said to be in canonical form; then, for a first order
system, the coefficient of s (1/a) is called the time constant of the (first order) system and the
numerator of the transfer function (b/a) is the steady state gain of the system. Incidentally, eq.
(2.3) also shows that the transfer function G(s) is in fact independent of the system inlet
perturbation, since no perturbation was introduced to the system so far.
In order to clarify the meaning of the time constant and (steady state) gain of the system,
let us introduce an unitary step perturbation in the system inlet variable, i.e., let us assume
that:
1
x(t) = H(t) so x( s) = (2.4)
s
where H(t) is the Heaviside step function (see Appendix 3). The response of the system in the
Laplace domain is then:
b b / a −b / a
y( s ) = = + (2.5)
s( s + a ) s ( s + a)
(1)If the model equation is not linear, in order to apply the Laplace transform, a linearization is needed. In
Systems Dynamics the behavior of nonlinear systems can also be studied without linearization, but then the
Laplace transform cannot be applied. This methodology will be shown when studying second order systems,
where a nonlinear system will be analyzed with and without linearization.
In order to invert the system response to the time domain, the obtained rational fraction
was already expanded into partial fractions (see, for example, N. Piskounov, MIR, 1966). The
time domain response can then easily be obtained (Tables of Laplace transforms are available
in standard references):
b
y( t ) =[1 − exp(− a t )] H( t ) (2.6)
a
The H(t) is used in the response in order to reinforce that the response is given by that
expression only for t ≥ 0 , i.e., before t = 0 the expression is not valid. In the following draw,
the response of the system is represented:
The final response of the system (at t = ∞ ) will be y( ∞ ) = b/a; the system gain (often
designated by K) is exactly the ratio between the final values of the outlet and inlet variables,
y( ∞ )
i.e., gain = . The time constant (commonly designated by the symbol τ) is the time after
x( ∞ )
which the response of the system is (1- e-1) ≈ 63.2% of the final response.
Let us now look at some examples of first order systems:
2.1.1. CSTR (Continuous Stirred Tank Reactor)

a) CSTR with no reaction:
Let us start with a CSTR with no reaction, i.e., a mixer; the CSTR has volume V and a
solute at concentration cE is being continuously fed at flowrate Q. The concentration inside
the CSTR as well as at the outlet (at the same flowrate Q) is c:
A material balance to the solute gives:

d c( t )
Q cE ( t ) = Q c( t ) + V (2.7)
dt
The system is of first order, since the material balance is a first order ODE. In steady
state, the balance becomes:
Q cEs = Q cs (2.8)
This means that, in steady state, the outlet and inlet concentrations are equal (remember
that there is no reaction). Subtracting the steady state equation (2.8) from the unsteady state
one (2.7) and introducing the deviation variables ĉE ( t ) = cE ( t ) − cEs and ĉ( t ) = c( t ) − c s , the
unsteady state mass balance (in terms of deviation variables) becomes:
dĉ( t )
ĉE ( t ) = ĉ( t ) + τ (2.9)
dt
where τ = V/Q is the space time of the reactor (equal to the mean residence time of the
molecules inside the reactor). Remembering that, according to the definition of the deviation
variables, ĉ E ( 0 ) = ĉ( 0 ) = 0 , the Laplace transform of equation (2.9) gives:
cE ( s ) = ( 1 +τ s ) c ( s ) (2.10)
The transfer function for the system is then:
c( s) 1
G( s ) = = (2.11)
cE ( s ) τ s +1
Again, since the denominator of the transfer function is a first degree polynomial the
system is of first order and, since G(s) is already in the canonical form, it is easy to see that
the system has an unitary gain and that its time constant is nothing else but the space time.
b) CSTR with first order reaction:
If a first order reaction A → B is supposed to occur in the isothermal CSTR, with

r = k c, the material balance to the reactant becomes:
d c( t )
Q cE ( t ) = Q c( t ) + V k c( t ) + V (2.12)
dt
And, in steady state:
Q cEs = Q cs + V k cs (2.13)
Subtracting the steady state from the unsteady state equation and introducing deviation
variables:
dĉ( t )
ĉE ( t ) = ( 1 + k τ ) ĉ( t ) + τ (2.14)
dt
Transforming and rearranging, the transfer function finally becomes:
c( s) 1 1 /( kτ + 1)
G( s ) = = = (2.15)
cE ( s ) τ ( s + k ) + 1 τ s /( kτ + 1) + 1
i.e., the transfer function with a first order reaction can be obtained from the transfer function
without reaction by just changing s in s + k, where k is the reaction rate constant. Note that the
gain of the system is now 1/(kτ+1) and the new time constant is τ/(kτ+1), where kτ = Da is the
Damköhler number (also called number of reaction units NR, it is the ratio between the CSTR
space time τ and the time constant for the first order reaction τR = 1/k).
2.1.2. Storage Tank:

Consider now the following storage tank with cross sectional area Ω (V = Ω h):
The valve at the outlet is supposed to be linear, i.e., ∆P = ρ g h = R Q, where R is the

resistance of the valve (see Appendix 5 - Ohm's law). A material balance to the system in
unsteady state, gives:
dV ( t ) ρ g d h( t )
QE ( t ) = Q( t ) + = h( t ) + Ω (2.16)
dt R dt
where ρ is the stored liquid density and g is the acceleration due to gravity. In steady state:
ρg
QEs = Qs = hs (2.17)
R
Subtracting (2.17) from (2.16) and introducing the deviation variables Q̂E ( t )=QE ( t )−QEs
and ĥ( t ) = h( t ) − hs , the first order linear ODE in time becomes:
ρg dĥ( t )
Q̂E ( t ) = ĥ( t ) + Ω (2.18)
R dt
Remembering that the deviation variables are zero at time zero, the Laplace transform of
this equation gives the transfer function G(s) for the system, relating any perturbation Qˆ E (t)
with the corresponding height of fluid in the tank:
h( s) 1 R /( ρ g )
G( s ) = = = (2.19)
QE ( s ) Ω s + ρ g / R Ω Rs /( ρ g ) + 1
i.e., the system time constant is Ω R/(ρ g) and its steady state gain is R/(ρ g).
2.1.3. Mercury thermometer:

Suppose that you use a mercury thermometer for measuring the temperature Tb of an
"infinite" bath. The thermometer is initially at the ambient temperature Ta and, at time zero,
its bulb is immersed into the bath:
The actual temperature of the mercury (given by the

thermometer) is T(t). Suppose that the mass of mercury is m,
and that its specific heat is Cp; neglect the heating of the glass
(it will be considered just as a resistance between the bath and
the mercury). A heat balance to the system gives:
dT
h A ( TE − T ) = m C p (2.20)
dt
where h is the heat transfer coefficient between the fluid in the
bath and the mercury, and A is the (average) area of the bulb.
Note that in this case a step input T̂E = Tb −Ta was made in the "inlet" temperature. We
will not calculate the response of the system (the response of a general first order system to
such an input was already considered). The steady state equation (previous steady state) is:
h A (TEs - Ta) = 0  TEs = Ta (2.21)
Subtracting, introducing deviation variables and transforming, we arrive finally to the
transfer function G(s) for the thermometer:
T ( s) 1
G( s ) = = (2.22)
TE ( s ) mC p
s+1
hA
i.e., the steady state gain of the thermometer is 1 (hopefully!) and its time constant is
mC
τ = p (only after 3τ the response is about 95% of the final response…). This example
hA
shows one of the advantages of knowing the dynamic behavior of systems: in this case, if the
operator does not wait time enough he will not measure the true temperature of the bath; but,
if he knows the time constant of the system, he does not need to wait, since the final response
(actual temperature) can be calculated after one single point is known …
2.2. SECOND ORDER SYSTEMS

When the model equation is a second order ODE, the system is said to be of second
order. Let us consider the general linear second order ODE with constant coefficients:
d 2 y( t ) d y( t )
2
+a + b y ( t ) = c x( t ) (2.23)
dt dt
Assuming that x(t) and y(t) are inlet and outlet deviation variables and that t is time
while a, b and c are constants, the Laplace transform of equation (2.23) gives:
( s2 + a s + b) y ( s ) = c x ( s ) (2.24)
i.e., the transfer function G(s) is:
y( s) c c/b
G( s ) = = 2 = 2 (2.25)
x( s) s + a s + b s / b + s a / b + 1
Written in the canonical form (independent term equal one), the denominator of the
transfer function has the form: τ2 s2 + 2 ζ τ s + 1, where τ is the time constant of the system
and ζ is the damping factor; as in the first order system, the numerator of the transfer function
is the steady state gain. For this (general) second order system, the gain is then K = c/b, the
time constant is τ = 1/ b and the damping factor is ζ = a /( 2 b ) = aτ / 2 . The damping
factor plays a very important role in second order systems dynamics; as a matter of fact,
according to its value, the response of the system can be oscillating even for non-oscillating
perturbations at the inlet. Negative values for ζ will not be considered (they give rise to
unstable systems - with unbounded response).
In order to have an idea of the (possible) responses of the system, let us consider an
unitary step input to this system, x(t) = H(t), i.e., x ( s )=1 / s ; the response in the Laplace
domain becomes:
K
y( s) = (2.26)
s (τ s + 2ζ τ s + 1)
2 2
In order to calculate the response in the time domain, the response in the Laplace
domain should be inverted; this inversion imposes the expansion of the rational fraction in
simple fractions; the roots r1 and r2 of the second order polynomial have to be calculated; they
are:
1 
−ζ ± ζ − 1 
2
r1 , r2 = (2.27)
τ  
These roots are real if ζ ≥ 1. In this case, the denominator is simply a product of three
first order polynomials with real coefficients. When ζ < 1, the roots are complex; in this case,
the denominator cannot be written in the form of a product of three first order polynomials
with real coefficients. The expansion in simple fractions greatly depends on the form of the
denominator, as well as the inversion will depend on the type of the simple fractions obtained.
Let us consider the three possible cases:
a) ζ > 1 (two distinct real roots) - overdamped system
In this case, the response in the Laplace domain and the respective expansion in simple
fractions, taking into account that 1/τ2 = r1r2, can be written:
K /τ2 K K r2 /( r1 − r2 ) K r1 /( r1 − r2 )
y( s) = = + − (2.28)
s( s − r1 )( s − r2 ) s ( s − r1 ) ( s − r2 )
The inversion is straightforward now (see a table of Laplace transforms). The response
in the time domain is:

y( t ) = K 1 +
1
[r2 exp( −r1t ) − r1exp( −r2t )] H( t ) =
 r1 − r2 
  
   t ζ  t  
= K 1 − exp( −ζ t / τ ) cosh ζ − 1  +
2
sinh ζ − 1   H( t )
2
(2.29)
   τ ζ 2
− 1  τ  
 
This response y(t)/K is represented against t/τ in the following figure, for several values
of the damping factor ζ.
1,2
1,0
1
y(t)/K
1.5
0,8
2
0,6 3
4
ζ
0,4
0,2
0,0
0 5 10 15
t/ τ 20
b) ζ = 1 (two equal real roots) - critically damped system

The equal roots are r = r1 = r2 = - ζ/τ = - 1/τ; the response in the Laplace domain
becomes:
K /τ 2 K /( rτ ) 2 K /( rτ 2 ) K /( rτ ) 2 K K /τ K
y( s) = = + − = − −
s( s − r ) 2
s (s − r) 2
(s − r) s (s + 1 /τ ) 2
(s +1 /τ )
(2.30)
The response in the time domain is then:
[ ]
y ( t ) = K 1 − ( 1 + t / τ ) e −t / τ H ( t ) (2.31)
In the previous figure the response of a critically damped system (ζ = 1) is also
represented.
c) ζ < 1 (two complex conjugate roots) - underdamped system

In this case, the polynomial in the denominator cannot be decomposed in simple
binomial factors with real coefficients through the use of its roots since they are complex. The
approach is then different: the trinomial is first written in the form of a sum of squares. As a
matter of fact, it is easy to show that:
 2
b   c 
2
2   b2  
ax + bx + c = a  x +  +
  −  (2.32)
 2a   a 4a 2 
   
The response in the Laplace domain can then be written as:
K /τ2
y( s) = =
  
2
  1 −ζ
2 2
 ζ  
s s + +
   
τ   τ
   
 
 ζ 1 −ζ 2
Ks + 
K  τ  Kζ τ
= − − (2.33)
s  2  
2
1 −ζ 2  2  
2
 ζ   1 −ζ  ζ   1 −ζ
2 2
   
 s + τ  + τ    s + τ  + τ  
 
    
  
With the help of a table of Laplace transforms, the inversion is now straightforward; the
response in the time domain is:
  
  t ζ  t 
y( t ) = K 1 − e −ζt / τ cos 1 −ζ 2  + sin 1 −ζ 2   H( t ) (2.34)
   τ 1 −ζ 2  τ  
 
Due to the trigonometric functions, and depending on the value of the damping factor ζ,
this response is oscillatory; the amplitude of the oscillations is attenuated with time by the
factor e- ζt/τ. In the next figure, responses y(t)/K vs. t/τ are shown for several values of ζ.
0.1
1,6 0.2
y(t)/K
0.4
1,2 0.6
0.8
0,8
ζ
0,4
0,0
0 5 10 15 t/ τ 20
The underdamped second order system is commonly described by a series of

characteristics of its step response (overshoot, rise time, decay ratio, period of the natural
oscillation, etc.). The respective values can be obtained from equation (2.34); they can also be
found in any standard textbook on Systems Dynamics and/or Process Control.
Apart from natural second order systems, the association of two first order systems is
itself a (commonly overdamped or critically damped) second order system. In the following
two examples of second order systems resulting from such an association, with interaction and
without interaction, will be treated; then, the case of a tank with a discharge tube in laminar or
in turbulent regime will be tackled with the search for underdamped operating conditions,
since a TP is envisaged based on this system; it will be shown also that if the tube is packed
(with sand, for example) the system will become overdamped, whatever the regime.
2.2.1. Tanks in Series (without interaction):

Let us consider the two following tanks in series (note that there is no interaction since
the first tank is not aware of what is happening in the second):
The cross sectional areas of the tanks are Ω1 and Ω2 respectively. As before, we will
assume that the resistances are linear, i.e., ∆P1 = ρ g h1 = R1 Q1 and ∆P2 = ρ g h2 = R2 Q2. A
material balance to each tank gives, in unsteady state:
dV1( t ) ρ g dh ( t )
QE ( t ) = Q1( t ) + = h1( t ) + Ω1 1 (2.35)
dt R1 dt
ρg dV ( t ) ρ g dh ( t )
Q1( t ) = h1( t ) = Q2 ( t ) + 2 = h2 ( t ) + Ω 2 2 (2.36)
R1 dt R2 dt
In steady state the material balances just become:
ρg ρg
QEs = Q1s = h1s = Q2 s = h2 s (2.37)
R1 R2
Subtracting the steady state equations from the unsteady state ones and introducing the
deviation variables Q̂E ( t )=QE ( t ) −QEs , ĥ1 ( t )=h1 ( t )−h1 s and ĥ2 ( t )=h2 ( t )−h2 s the differential
equations in terms of deviation variables are:
ρg dĥ1( t )
Q̂E ( t ) = ĥ1( t ) + Ω1 (2.38)
R1 dt
ρg ρg d ĥ2 ( t )
ĥ1 ( t ) = ĥ2 ( t ) + Ω 2 (2.39)
R1 R2 dt
Transforming both equations, the following transfer functions g1(s) for the first tank and
g2(s) for the second are obtained:
h ( s) 1 R1 /( ρ g )
g1( s ) = 1 = = (2.40)
QE ( s ) Ω1 s + ρ g / R1 Ω1R1 s /( ρ g ) + 1
h ( s) ρ g / R1 R2 / R1
g2 ( s ) = 2 = = (2.41)
h1( s ) Ω 2 s + ρ g / R2 Ω 2 R2 s /( ρ g ) + 1
And the transfer function G(s) for the system, relating the height of the fluid in the
second tank ĥ2 ( t ) to any perturbation in the inlet volume flowrate to the first tank Q̂E ( t ) is
simply obtained through the product of the previous individual transfer functions, i.e.:
R2 /( ρ g )
G( s ) = g1( s )⋅g2 ( s ) = (2.42)
Ω1R1Ω 2 R2 2 Ω 1R1 +Ω 2 R2
s + s+1
( ρg )2 ρg
G(s) is already in the canonical form; by comparison with the standard second order
system, the time constant for this association of two non-interacting first order systems is then
Ω1R1Ω 2 R2 Ω R + Ω 2 R2
τ= and the damping factor is ζ = 1 1 .
ρg 2 Ω1R1Ω 2 R2
2.2.2. Tanks in Series (with interaction):

If the first system is aware of what happens in the second system, i.e., it is also disturbed
by the behavior of the second, we have an association with interaction; let us treat the case of
the two tanks in series (with interaction) represented in the following figure:
The interaction is expressed by the equation for the first valve; as a matter of fact, the
pressure drop across this first valve is now the difference between the pressures at the bottom
of both tanks, i.e., the equation for the first valve is now ∆P1 - ∆P2 = ρ g (h1 - h2) = R1 Q1; the
linearity of the second valve is still expressed by the previous equation: ∆P2 = ρ g h2 = R2 Q2.
The material balances for both tanks are, in unsteady state:
dV1( t ) ρ g
QE ( t ) = Q1( t ) + = [h1( t ) − h2 ( t )] + Ω1 dh1( t ) (2.43)
dt R1 dt
ρg
Q1( t ) = [h1( t ) − h2 ( t )] = Q2 ( t ) + dV2 ( t ) = ρ g h2 ( t ) + Ω1 dh2 ( t ) (2.44)
R1 dt R2 dt
After writing the steady state mass balances, subtracting them from the unsteady state
equations, introducing deviation variables, transforming and rearranging, the following
transfer function for the interactive association, relating the height of the fluid in the second
tank ĥ2 ( t ) to any perturbation in the inlet volume flowrate to the first tank Q̂E ( t ) is obtained:
R2 /( ρ g )
G( s ) = g1( s )⋅g2 ( s ) = (2.45)
Ω1R1Ω 2 R2 2 Ω 1( R1 + R2 ) +Ω 2 R2
s + s+1
( ρg )2 ρg
Comparing with the transfer function for the non-interactive association, we can see that
only the coefficient of s in the transfer function is altered, i.e., the time constant is maintained
Ω ( R +R ) + Ω 2 R2
and the damping factor is now ζ = 1 1 2 .
2 Ω1R1Ω 2 R2
2.2.3. Tank discharging through a (long) tube:

Within the framework of the analogies between systems (see Appendix 5), a long thin
tube can be considered as a resistance; then, if a (long) thin tube is used, this problem reduces
to the already tackled problem of the storage tank (section 2.1.2) - it is a first order system.
When a long tube with a large cross section is used, the framework of the analogies tells us
that the tube itself is the association of a resistance with an inductance(1) (the tank itself is a
capacitance). In this last case, a second order system can result. In the following, the general
equations will be established for each situation (laminar and turbulent regimes in empty tubes
and packed tubes respectively, without the use of the equivalences), and then operating values
for the parameters will be searched in order that the resulting second order system can,
eventually, be underdamped. When a nonlinear equation is obtained, the solution will be
searched for its linearized counterpart (using Laplace transforms) and for the nonlinear
equation itself (by direct integration) as well.
Let us then establish the general equations for the problem, represented in the following
figure:
The tank (of internal diameter DT and cross sectional area Ω = π DT2 / 4 ) receives liquid
(of density ρ) at the volume flowrate QE(t); the volume of liquid inside the tank is V(t) = Ω
h(t), where h(t) is the height of the liquid in the tank; the tube has length L, internal diameter
D, cross sectional area A = π D2/4, and through it liquid is discharged at the volume flowrate
Q(t). We have already made the material balance for such a tank (section 2.1.2):
(1)The three basic electrical elements used in the framework of the analogies are the resistance R, the capacitance
dV dI
C and the inductance L with the following definition equations: V = R I, I = C and V = L .
dt dt
dV ( t ) dh( t )
QE ( t ) = Q( t ) + = AV(t ) + Ω (2.46)
dt dt
As for the tube, a force balance will be made, stating that the variation of the velocity of the
fluid in the tube (its acceleration) is a consequence of the balance between two counteracting
forces: the hydraulic forces at the bottom of the tank (pushing the liquid in the tube to its
outlet) and the friction forces, exerted by the moving fluid in the opposite direction (in the
tube walls). Using Newton's second law (F = m a, where F is force, m is mass and a = dv/dt is
acceleration), the force balance around the tube becomes:
dv( t ) 1
AL ρ = A g ρ h( t ) − ρ v 2 ( t ) Aλ f (2.47)
dt 2
where v(t) is the liquid velocity in the tube. In this equation Aλ = π D L is the lateral (internal)
area of the tube, f is the friction factor (see Appendix 5) and g is the acceleration due to
gravity. Rearranging both equations (2.46) and (2.47) we obtain the following model
equations for the system:
d h( t ) 4 D2
= QE ( t ) − v( t ) (2.48)
dt π DT2 DT2
dv( t ) g 2
= h( t ) − f v 2 ( t ) (2.49)
dt L D
Depending on the expression for the friction factor, different situations can arise; in the
following, three cases will be considered: empty tubes with flow in laminar regime and in
turbulent regime (within the validity of the Blasius correlation - see Appendix 5) and the case
of packed tubes (covering all the regimes - Ergun equation).
a) Empty tube with flow in laminar regime:

ρvD
In the laminar regime ( Re = < 2100 , where Re is the Reynolds number), the
µ
Hagen-Poiseuille equation applies. Using the friction factor for laminar regime (f = 16/Re -
see Appendix 5), the model equations become:
d h( t ) 4 D2
= QE ( t ) − v( t ) (2.48)
dt π DT2 DT2
dv( t ) g 32 µ
= h( t ) − v( t ) (2.50)
dt L ρ D2
Note that in this (laminar) case, the model equations are linear. And, in steady state:
4 D2
0= QEs − vs (2.51)
π DT2 DT2
g 32 µ
0 = hs − vs (2.52)
L ρ D2
Subtracting the steady from the unsteady state equations, introducing the deviation
variables Q̂E ( t ) = QE ( t ) − QEs , ĥ( t ) = h( t ) − hs and v̂( t ) = v( t ) − v s transforming and manipula-
ting, the following transfer function linking the height of the liquid in the tank ĥ( t ) with the
volume flowrate at the inlet Q̂E ( t ) is obtained:
b1
( s + a2 )
h( s) a1 b2
G( s ) = = (2.53)
QE ( s ) s2 a s
+ 2 +1
a1 b2 a1 b2
where, for convenience, the following names were given to the parameters:
g 32 µ 4 D2
a1 = a2 = b1 = b2 = 2 (2.54)
L ρ D2 π DT2 DT
It is seen that the time constant for this system is τ = 1/ a1b2 and that the
corresponding dumping factor is ζ = a2/(2 a1b2 ) = a2τ/2. Note that in the numerator of the
transfer function a s is present; this means that a derivative action is present (multiplying by s
in the Laplace domain corresponds to deriving in order to time in the time domain - see
properties of the Laplace transform, transform of a derivative, Appendix 1). This derivative is
a contribution from the inductive part of the tube (if the tube were only a resistance, the
derivative action should not appear).
In terms of the real parameters of the system, the time constant and damping factors are
respectively:
DT L µ DT L
τ= ζ = 16 (2.55)
D g ρ D3 g
This means that both τ and ζ increase with DT and L and decrease with D; a convenient
situation for an experimental demonstration of an underdamped system should be a large
value of τ (in order that the registration of the eventual variations is possible) and a very small
ζ (in order that the system oscillates the most); from the relations in (2.55), it is seen that this
is contradictory and a compromise has to be made - based on the fact that D influences much
more ζ (as a cube) than τ (linear dependence). In the EHICS, the average temperature of the
available water is 12 °C; the viscosity of water at 12 °C is approximately 1.24 cp; taking µ =
1.24 x 10-3 kg/(m s), ρ = 1000 kg/m3, g = 9.8 m2/s, and fixing the tank diameter DT = 6 cm
(in order that ζ does not increase, a small diameter has to be chosen), the following table
shows the values of τ and ζ for two tube diameters (D = 2 cm and D = 3.5 cm); also shown
are the steady state heights of the liquid in the tank, velocities in the tube and volume
flowrates for a (arbitrarily chosen) Re = 1000 < 2100; these values are calculated with the
steady state equations (2.51) and (2.52).
D = 2 cm D = 3.5 cm
hs, vs, Qs, hs, vs, Qs,
L, m ζ τ, s ζ τ, s
cm cm/s l/min cm cm/s l/min
10 0.150 3.03 0.63 6.20 1.17 0.028 1.73 0.12 3.54 2.05
20 0.213 4.29 1.26 6.20 1.17 0.040 2.45 0.23 3.54 2.05
30 0.260 5.25 1.88 6.20 1.17 0.049 3.00 0.35 3.54 2.05
40 0.301 6.06 2.51 6.20 1.17 0.056 3.46 0.47 3.54 2.05
50 0.336 6.78 3.14 6.20 1.17 0.063 3.87 0.59 3.54 2.05
REFERENCES
Churchill, R.V., Operational Mathematics, McGraw-Hill, New York, 1972.
Coghanowr, D.R., Process Systems Analysis and Control, 2nd ed., McGraw-Hill, New York,
1991.
Luyben, W.L., Process Modeling, Simulation and Control for Chemical Engineers, 2nd ed.,
McGraw-Hill, New York, 1990.
Piskounov, N., Calcul Différentiel et Intégral, 2e ed., Éditions MIR, Moscou, 1966.
Seborg, D.E., Edgar, T.F. and D.A. Mellichamp, Process Dynamics and Control, John Wiley,
New York, 1989.
Stephanopoulos, G., Chemical Process Control – An Introduction to Theory and Practice,
Prentice-Hall, Englewood Cliffs, New Jersey, 1984.
APPENDIX 1
A.1.1. THE LAPLACE TRANSFORM
The Laplace Transform is an integral transform which operates on a function through
the following integral:
∞
£{f(t)} = F(s) =  e −st f ( t ) dt (A1.1)
0
in which the complex number s is the Laplace parameter (if s is a pure imaginary number, the
Fourier transform is obtained); for physical integrity of eq. (A1.1), s has the inverse
dimensions of t. Since the transform is defined only to the right of t = 0, the independent
variable t is commonly associated to time (but, in fact, it can be any variable).
Since the operation of integration is linear (as a matter of fact it is an area), the Laplace
transform is a linear operator in the sense that it is a function for which the principle of
superposition is valid:
£{a f(t) + b g(t)} = a F(s) + b G(s) (A1.2)
The main advantage of the Laplace transform is that ordinary differential equations
(ODEs) in the independent variable t are transformed in algebraic equations in the parameter
s; furthermore, partial differential equations (PDEs) in n independent variables are
transformed in PDEs in (n-1) independent variables (or, if n=2, in ODEs). This greatly
reduces the complexity of the solution of the corresponding equations: the solution is first
searched in the Laplace domain and it is then "inverted" to the time domain again.
Since this is only an introduction (there are excellent books on the subject - see for
example the book of Churchill, 1972), let us calculate analytically only the Laplace transforms
of some very simple functions (tables of Laplace transforms are available, e.g., in the above
referred book): the Dirac delta function δ(t) and the Heaviside step function H(t), specially
important in the domain of Systems Dynamics (for the definitions of these functions see
Appendix 3).
To calculate the Laplace transform of the Dirac delta function, we have to resort to a
general theorem related to this "strange" function:
∞
 f ( t ) δ ( t − t o ) dt = f(to) 0 ≤ to < ∞ (A1.3)

0
This result can be easily obtained integrating by parts this improper integral (of the 1st
and 2nd kinds!):
∞ A
lim
 f ( t ) δ ( t − t o ) dt = A → ∞  f ( t ) δ ( t − t o ) dt =
0 0
 A 
= Alim
→∞
[ f ( t ) H ( t − t )
o 0]A
−  f ' ( t ) H( t − t o ) dt  =
 0 
 A 
= Alim
→∞
f ( A ) − 0 −  f ' ( t ) dt  = f(to) (A1.4)
 to 
Given this result, it is easy to show that:
£{δ(t - to)} = e-sto (A1.5)
and, if to = 0:
£{δ(t)} = 1 (A1.6)
For the Heaviside step function, we have:
∞ ∞
e − sto
£{H(t - to)} =  e − st H( t − t o ) dt =  e −st dt = (A1.7)
0 to
s
and, if to = 0:
1
£{H(t)} = (A1.8)
s
Incidentally, equations (A1.5) and (A1.7) also show the result of the application of the
2nd shift theorem (see in the following some properties of the Laplace transform).
A.1.2. PROPERTIES OF THE LAPLACE TRANSFORM

In the following, some properties of the Laplace transform are given; the listing is not
exhaustive and the interested reader should consult a standard reference for a complete listing
of properties and for the transforms of some common functions.
1. Final value theorem:
lim f ( t ) = lim [s F ( s )] (A1.9)
t→∞ s→0
2. Initial value theorem:

lim f ( t ) = f ( 0 ) = lim [s F ( s )] (A1.10)
t→0 s→∞
3. Transform of an integral:
t 
£  f ( ξ ) dξ  =
F( s)
(A1.11)(1)
 0  s
(1)Since a definite integral is a function of its limits but not of the integration (mute) variable, this integration
variable was here represented by ξ (and not by t) in order that it cannot be confused with the independent
variable t of the function to be transformed (the definite integral). This procedure will be discarded in the rest of
these small notes (i.e., t will be used as the mute integration variable in the following).
4. Transform of a derivative in order to time:

d n f (t ) 
£ n 
= sn F(s) - sn-1 f(0) - sn-2 f(1)(0) - … - s f(n-2)(0) - f(n-1)(0) (A1.12)
 d t 
where f(i)(0) means the ith derivative of f(t) in order to time evaluated at t = 0. In particular for
first derivatives:
d f (t ) 
£  = s F(s) - f(0) (A1.13)
 dt 
5. First shift theorem (translation in the Laplace domain):
£[e-atf(t)] = F(s+a) (A1.14)
6. Second shift theorem (translation in time):
£[f(t - to) H(t - to)] = e-sto F(s) (A1.15)
In the above equations, F(s) stands for the Laplace transform of f(t), i.e., F(s) = £{f(t)}.
7. The Van der Laan Theorem:
Although the Van der Laan theorem can hardly be classified as a property of Laplace
transforms, it will be included here, given its utility and its relation with the Laplace
transforms. The theorem states that the moments of the impulse response of a system (or of its
residence time distribution, which is the same) can be obtained directly from the transfer
function of the system (for the definitions of the impulse and of the transfer function see
Appendices 3 and 2, respectively), through the relations:
d n G( s)
µ n = (−1) n (A1.16)
d sn s =0
where G(s) is the transfer function, dnG(s)/dsn stands for the nth derivative of G(s) in order to s
and µ n is the nth moment of the impulse response of the system:
∞
t
n
E ( t ) dt ∞
µn = 0
∞
=  t n E ( t ) dt (A1.17)
 E( t ) dt
0
where E(t) is the residence time distribution (RTD) for the system (equal to the impulse
response), whose integral between zero and infinity is equal to one.
APPENDIX 2
THE TRANSFER FUNCTION
One of the main advantages of the Laplace transform related to the study of Systems
Dynamics is the so-called transfer function of the system under study; this function relates the
output with the input to the system for whatever input, as far as the system is linear.
In order to simplify the calculations, "deviation" (or "perturbation"(1)) variables are first
introduced; these are defined by the difference between the actual values of the variables and
their values in the precedent steady state (before time zero). In this way, the deviation
variables are always equal to zero at time zero (if you look to the transform of a derivative,
you can better appreciate the motivation). In doing so, we then study only the result of the
perturbations introduced to the system at a given instant which, in terms of the analysis of the
system behaviour, is t = 0; in the end we are always able to revert to the original variables (by
simple summation of the previous steady state value).
Let us consider a system with an inlet variable x(t) and an outlet variable y(t):
The deviation variables are then defined as:

x̂( t ) = x(t) - x(0) and ŷ( t ) = y(t) - y(0) (A2.1)
Note that the system is supposed in steady state before any perturbation is introduced;
the values x(0) and y(0) are then the previous steady state values and the result [ ŷ( t ) = y(t) -
y(0)] of any perturbation introduced to the system [ x̂( t ) = x(t) - x(0)] will be studied. Those
steady state values can easily be obtained by the solution of the system model equations in
steady state.
In general, the output is a function of the input; in the time domain we have:
ŷ( t ) = f[ x̂( t ) ] (A2.2)
If we call x ( s ) and y ( s ) to the Laplace transforms of x̂( t ) and ŷ( t ) respectively, it can
be proved that there exists a function G(s) in the Laplace domain so that:
y( s)
G(s) = (A2.3)
x( s)
This is the so-called transfer function of the system. It relates y ( s ) with x ( s ) for
whatever perturbation, i.e., for whatever the value of x ( s ) . This function is in general
obtained from the system model. And the independence of its value from the system input,
enables the study of the responses of the system to several perturbations without the need to
develop the equations each time.
(1) Or "disturbances", since they disturb the previous steady state.

APPENDIX 3
SOME COMMON INPUTS
In order to study the dynamic behavior of systems, some inputs are commonly used and
the system responses to those inputs analyzed. The most common inputs used in Systems
Dynamics, in view of their application to Process Control, are the impulse, the step, the ramp
and the sinusoidal inputs. In the following, a brief description of each of these perturbations to
the system steady state entry is made, together with the respective mathematical definition and
Laplace transform.
1. The impulse input:

The impulse input is a mathematical abstraction of what occurs when the system inlet is
suddenly perturbed by a given amount during a very small (mathematically zero) time
interval; to give an example: imagine a tank to which a constant water volume flowrate QE =
Q is being fed; the tank has an output tube and, in steady state, the outlet is also QS = Q;
imagine that, suddenly, at time to, someone drops Vo l of water in the tank; although
physically some time is needed to complete the operation, mathematically we say that the time
elapsed was zero and that an impulse input was made to the system; we can then write for the
inlet flowrate:
QE = Q + Vo δ(t - to) (A3.1)
In this equation δ(t - to) is the Dirac delta function; it is defined as:
1 ∞ t = to
δ(t − t o ) = lim =  (A3.2)
t →t o (t − t o ) 0 t ≠ to
∞
 δ ( t − t o ) dt = 1 0 ≤ to < ∞ (A3.3)
0
Note that, according to its definition, δ(t - to) has the inverse dimensions of the variable
t; namely, if t is time, it has dimensions of (time)-1.
Since the deviation variable Q̂ E ( t ) = Vo δ(t - to) (see Appendix 2) is infinity at t = to, its
graphical representation is generally made by an arrow located at the respective time; in the
same draw, the area under the impulse (in this case the value of the indetermination 0 x ∞ = Vo
is known) is shown by the inclusion of its value with another (curved) arrow:
The Laplace transform of the Dirac delta function located at time zero is the unity, i.e.,
^ (t) = V δ(t - t ), applying the second shift theorem
£{δ(t)} = 1; then, for a general input Q E o o
(see properties of the Laplace transform, Appendix 1) we have:
£{ Q̂ E ( t ) } = Q E (s) = £{Vo δ(t - to)} = Vo £{δ(t - to)} = Vo e-sto (A3.4)
The impulse input is experimentally simulated by a square pulse (as fast as possible) on
the desired entry variable.
Finally, it can be said that in the framework of residence time distributions, the response
of a system to a Dirac delta input δ(t) in the concentration of a tracer at the system inlet is
exactly the residence time distribution E(t) of the system. Furthermore, given that £{δ(t)} = 1,
the system response to such an input (in the Laplace domain) is equal to the transfer function
of the system G(s); then, the residence time distribution E(t) is also equal to the inverse
transform of the transfer function G(s), i.e., E(t) = £-1{G(s)}.
2. The step input:

Another common input used in Systems Dynamics is the so-called step input; it
corresponds to a sudden change on the system entry from a steady value to another steady
value (if this new value is lower than the previous one, we have a negative step, sometimes
called purge since, in a sense, it corresponds to the purging of the system). Choosing again the
tank example, imagine now that, at time to, the steady entry Q is suddenly changed to Q + qo.
Mathematically, the system entry can be represented by the following expression:
QE = Q + qo H(t - to) (A3.5)
where H(t - to) stands for the Heaviside step function. This function is defined as:
0 t < to
H(t − t o ) =  (A3.6)
1 t > to
The value of H(t - to) at time to is somehow arbitrary; 0, 1/2, 1 can be found in the
literature. Nonetheless, when solving one given problem, the same value for the Heaviside
step function at time to should be used throughout in order to maintain coherence and to
preclude errors in the analysis. In this small introduction, the convention that at time t = to the
value is already unitary will be used, i.e., the following definition for the Heaviside step
function is valid:
0 t < to
H(t − t o ) =  (A3.7)
1 t ≥ to
The function H(t - to) is dimensionless.
In the following draw the deviation variable Q̂E ( t ) = qo H(t - to) is represented. The
Laplace transform of the Heaviside step function located at time zero is 1/s, i.e., £{H(t)} =
1/s; then, for a general input Q̂E ( t ) = qo H(t - to), using the second shift theorem, we obtain:
e − sto
£{ Q̂E ( t ) } = QE ( s ) = £{qo H(t - to)} = qo £{H(t - to)} = qo (A3.8)
s
Note that the Heaviside step function H(t - to) can be viewed as the integral of the Dirac
delta function δ(t - to):
t
d H( t − t o )
H(t − t o ) =  δ(t − t o ) dt and δ (t − t o ) = 0 ≤ to ≤ t (A3.9)
0 dt
This relation between both functions has important consequences in the analysis of the
dynamic behavior of linear systems; as a matter of fact, given that relation and the
superposition principle, the same information can be extracted from each of the perturbations
and the response of the system to one of them can be obtained from the response of the system
to the other: the response of the system to a step input (height qo) is simply the integral of its
response to an impulse response (amount injected qo); and the response to an impulse (amount
Vo) is simply the derivative of the response to the corresponding step (height Vo).
3. The ramp input:

When the entry to the system drifts linearly with time, we have the ramp input; taking
the same system as before, it occurs when the flowrate changes linearly with time from t = to
on, starting at the steady value Q:
QE = Q + a (t - to) H(t - to) (A3.10)
Graphically, the deviation variable Q̂E ( t ) = a (t - to) H(t - to) can be represented by:
Given that the Laplace transform of t H(t)(1) is 1/s2, i.e., £{t H(t)} = 1/s2, the Laplace
transform of the general input Q̂E ( t ) = a (t - to) H(t - to), using again the second shift
theorem, is:
(1) Note that H(t) is used throughout when dealing with system entries; this is so in order that the Laplace
transform (defined between times 0 and infinity) can be applied; as a matter of fact, transforms of functions for
times before time zero do not exist. If the response to such an entry is desired (the negative step or purge is an
example) the function should be defined in terms of alternative functions, starting at time zero; for example the
unit purge is written as: P(t) = 1 - H(t) and the system response is obtained, applying the superposition principle,
as R(t) = 1 - F(t), where F(t), known as the Danckwerts F curve in the framework of residence time distributions
(when the perturbation is applied to the concentration of a tracer at the system inlet), is the response to the unit
e − sto
£{ Q̂E ( t ) } = QE ( s ) = £{a (t - to) H(t - to)} = a £{(t - to) H(t - to)} = a
(A3.11)
s2
The ramp input (t - to) H(t - to) is the integral of the Heaviside step function H(t - to);
then, the comments made above for the impulse and step inputs still apply, mutatis mutandis,
to the step and ramp inputs, of course.
4. The sinusoidal input:

A sinusoidal change in one entry variable to a system is sometimes made; the study of
the system behavior to this kind of perturbation is known in Systems Dynamics as frequency
response analysis.
Taking again the tank example, the entry volume flowrate for a sinusoidal perturbation
is written:
QE = Q + a sin[w(t - to)] H(t - to) (A3.12)
where a is the amplitude and w is the angular frequency of the sinusoidal perturbation; the
period is T = 2π/w.
The graphical representation of the deviation variable Q̂E ( t ) = a sin[w(t - to)] H(t - to) is
made in the following draw:
w
The Laplace transform of [sin(wt) H(t)] is £{sin (wt) H(t)} = ; then the Laplace
s + w2
2
transform of the general input Q̂E ( t ) = a sin[w(t - to)] H(t - to) is:
£{ Q̂E ( t ) } = QE ( s ) = £{a sin[w(t - to)] H(t - to)} =
w e − sto
= a £{sin[w(t - to)] H(t - to)} = a (A3.13)
s2 + w 2
5. Random inputs:
Sometimes a mathematical expression able to simulate a random input is needed;

random inputs can, in general, be decomposed in small pieces, each of them with its own
mathematical representation; the response to the global input can then be obtained by the
summation of the system responses to each part of the input (principle of superposition). In
step (Heaviside) function. Of course, the Dirac delta function δ(t) does not need to be multiplied by H(t), given
that it is zero to the left of t = 0.
the following it is shown how are the pieces assembled in order to have a unique
mathematical expression for an arbitrary random input.
Consider the following random input x̂ E ( t ) ; consider also that you know the
mathematical expressions for each of the functions f1, f2, f3, f4 and f5:
The general input x^ E(t) can then be represented by:

x̂ E ( t ) = f1 [H(t - to) - H(t - t1)] + f2 [H(t - t1) - H(t - t2)] + f3 [H(t - t2) - H(t - t3)] +
+ f4 [H(t - t3) - H(t - t4)] + f5 [H(t - t4) - H(t - t5)] (A3.14)

With the help of the Heaviside step function you should be able to prove that it is
exactly like this. This representation is surely not the only possible, but it is surely also the
simplest way to cope with random inputs.
APPENDIX 4
SYSTEMS LINEARIZATION
In the framework of Systems Dynamics with control purposes, nonlinear systems are
commonly linearized around a given steady state (generally the design steady state, where the
system is supposed to work) and the linearized system studied instead of the original one; this
is so because of the simplicity associated with linear systems and the possibility of the
application of the Laplace transform to their study; in general this enables a much faster
control action if the system is perturbed and, given that the linearization is around the design
steady state, the linearized solution is normally a reasonable approximation.
Linearization is made through the Taylor series expansion until the first (linear) term; as
you know the Taylor series expansion of a given function has exactly the same value of the
function if the infinite number of the series terms is conserved; as this is not possible, the
quality of the approximation decreases with the number of terms conserved. Conserving only
the linear term, we have for the value of a function f(x) at (x+a):
f(x+a) ≈ f(a) + x f '(a) (A4.1)
where f'(a) stands for the derivative of f(x) at x = a.
The meaning of the approximation can be graphically visualized (x = b in the drawing):
Equation (A4.1) substitutes the true value f(a+b) of f(x) at x = a+b by the approximated
value y. The drawing also shows that generally (always for monotonic functions), the larger b,
the worse the approximation. This is one of the reasons why the linearization is made around
the design steady state.
When dealing with transfer functions we said that deviation variables relative to the
previous steady state are normally used; the Laplace transforms of derivatives in order to time
of a given function f(t) become simpler in this case if we suppose that perturbations occur at
time zero, since ^f (0) = 0. Remember that the Taylor series around zero is also called the
MacLaurin series. Then, when the linearization is made around the previous steady state
(which should be the design one for any perturbation, at least if the system is working
properly), and time zero is selected for the perturbations, the deviation variable is simply
given by the MacLaurin expansion:
^f (t) = f(t) - f(0) = t f '(0) (A4.2)
APPENDIX 5
FLOW IN TUBES
Flow in tubes is but a small part of Momentum Transport; in this appendix only a brief
review of the subject, in view of its application to some designed experiments is covered (see,
e.g., the excellent book of Bird, Stewart and Lightfoot, John Wiley, 1960).
Three different situations are considered: flow in open tubes (laminar and turbulent) and
flow in packed tubes. The general framework of the equivalences between systems will be
used in the following approach. In the field of electricity, Ohm's law, stating that the electric
potential difference (V) between the terminals of an electrical conductor is proportional to the
current intensity (I) through the conductor, is well known:
V=RI (A5.1)
The proportionality constant R is called the resistance of the electrical conductor. This
law is in fact general (similar laws are, e.g., Fick's law of diffusion in mass transfer, Fourier's
law of conduction in heat transfer, Newton's law of viscosity in momentum transfer, etc.), and
it can be applied (with reasonable accuracy) to many fields in physical sciences, provided the
convenient equivalence is chosen; when the system is nonlinear, a linearization is needed (for
parameter identification) before the equivalence can be established. In fluid flow, the pressure
drop is generally chosen as the potential difference and the volumetric flowrate as the current
intensity (as a matter of fact the equivalence can be reversed - duality - with no loss of
generality).
1. Laminar flow in open tubes:

In the laminar regime, the flow in open tubes is modeled by the Hagen-Poiseuille
equation:
128 µ L
∆ P = Po - PL = Q Re < 2100 (A5.2)
π D4
In this equation Po and PL are the pressures at the tubes inlet and outlet respectively, µ is
ρvD
the viscosity of the fluid, L is the tube length and D its (internal) diameter; Re = is the
µ
Reynolds number, ρ the fluid density and v the average fluid velocity. Since, in laminar flow,
the relation between ∆ P and Q is linear, the equivalence between the Hagen-Poiseuille
128 µ L
equation and Ohm's law is straightforward: the resistance is simply R = .
π D4
A common approach in the study of flow in tubes is the introduction of friction
factors(1). The friction factor f is defined by the following equation:
Fk = A K f (A5.3)
(1)The definition of friction factors is not unique; published definitions vary by constant factors. When using
tables or graphics from a reference, the definition used in the reference should be used. In these notes, the
Fanning friction factor is used.
In this equation, Fk is the force exerted in the tube walls due to the fluid movement
1
(kinetic energy), A is the wetted surface area and K = ρ v2.
2
If, instead of the model equivalence approach, one wishes to use the friction factor
approach, it is easy to show that for laminar (Poiseuille) flow:
Fk ( ∆P )( π D 2 / 4 ) 16
f= = = (A5.4)
A K ( π D L )(1 / 2 ρ v ) Re
2
2. Turbulent flow in open tubes:

Turbulent flow is normally approached through the use of friction factors. Dimensional
analysis shows that, for fully developed flow, the friction factor is only a function of the
Reynolds number. Then, experimental results of friction factors are represented against the
Reynolds number. From these graphics (sometimes called Moody diagrams) Blasius
established the following formula:
0.0791
f = 2100 < Re < 105 (A5.5)
Re 0.25
Given the dependency of the friction factor on Re- 0.25, the equivalence with Ohm's law
can only be applied after linearization, i.e., the equivalent resistance can only be calculated
after linearization.
3. Flow in packed tubes:

As well as for flow in open tubes, two extreme regimes (laminar and turbulent) can be
found in flow through packed beds. The respective equations are normally called Blake-
Kozeny and Burke-Plummer equations. The respective sum gives the Ergun equation, which
correlates with reasonable accuracy the experimental data, even in the transition regime:
∆P 150 µ ( 1 − ε ) 2 1.75 ρ ( 1 − ε ) 2
= vo + vo (A5.6)
L Dp ε
2 3
Dp ε 3
in which the first term accounts for the laminar and the second for the turbulent contributions
for pressure drop. In this equation ε is the bed porosity (void volume/bed volume), Dp is the
equivalent mean particle diameter (Dp = 6 Vp/Ap; Vp is the particle volume, Ap is the particle
surface area) and vo = εvi is the superficial velocity (vi is the interstitial velocity, i.e., the
actual average velocity of the fluid in the bed voids).
Friction factors can also be calculated for both regimes, using a similar approach to the
case of flow in empty tubes (with the necessary adaptations, naturally); this will not be made
here, since for the Systems Dynamics approach, the Ergun equation is already in a convenient
form. It shows that the pressure drop is proportional to velocity in the laminar regime (just
like in open tubes), in which case the equivalent resistance is immediately evident; it further
shows that in turbulent regime the dependency is quadratic, which means that for turbulent
regime (and for intermediate regime, since then both terms should be considered),
linearization is needed.

DStext 2015

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

DStext 2015

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DStext 2015

Uploaded by

Copyright:

Available Formats

DEQ/FEUP

INTRODUCTION TO SYSTEMS DYNAMICS

2. LUMPED PARAMETER SYSTEMS

2.1. FIRST ORDER SYSTEMS

2.1.1. CSTR (Continuous Stirred Tank Reactor)

A material balance to the solute gives:

b) CSTR with first order reaction:

If a first order reaction A → B is supposed to occur in the isothermal CSTR, with

2.1.2. Storage Tank:

The valve at the outlet is supposed to be linear, i.e., ∆P = ρ g h = R Q, where R is the

2.1.3. Mercury thermometer:

The actual temperature of the mercury (given by the

2.2. SECOND ORDER SYSTEMS

b) ζ = 1 (two equal real roots) - critically damped system

c) ζ < 1 (two complex conjugate roots) - underdamped system

The underdamped second order system is commonly described by a series of

2.2.1. Tanks in Series (without interaction):

2.2.2. Tanks in Series (with interaction):

2.2.3. Tank discharging through a (long) tube:

a) Empty tube with flow in laminar regime:

 f ( t ) δ ( t − t o ) dt = f(to) 0 ≤ to < ∞ (A1.3)

A.1.2. PROPERTIES OF THE LAPLACE TRANSFORM

2. Initial value theorem:

4. Transform of a derivative in order to time:

The deviation variables are then defined as:

(1) Or "disturbances", since they disturb the previous steady state.

1. The impulse input:

2. The step input:

3. The ramp input:

4. The sinusoidal input:

Sometimes a mathematical expression able to simulate a random input is needed;

The general input x^ E(t) can then be represented by:

+ f4 [H(t - t3) - H(t - t4)] + f5 [H(t - t4) - H(t - t5)] (A3.14)

1. Laminar flow in open tubes:

2. Turbulent flow in open tubes:

3. Flow in packed tubes:

You might also like