-HQVHQ-07XPPHVFKHLW+

0RYLQJ%RXQGDU\0RGHOVIRU'\QDPLF6LPXODWLRQVRI7ZR3KDVH)ORZV
QG,QWHUQDWLRQDO0RGHOLFD&RQIHUHQFH3URFHHGLQJVSS

3DSHUSUHVHQWHGDWWKHQG,QWHUQDWLRQDO0RGHOLFD&RQIHUHQFH0DUFK
'HXWVFKHV=HQWUXPIU/XIWXQG5DXPIDKUWH9'/52EHUSIDIIHQKRIHQ*HUPDQ\
$OOSDSHUVRIWKLVZRUNVKRSFDQEHGRZQORDGHGIURP
KWWSZZZ0RGHOLFDRUJ&RQIHUHQFHSDSHUVVKWPO
3URJUDP&RPPLWWHH
 0DUWLQ 2WWHU 'HXWVFKHV =HQWUXP IU /XIW XQG 5DXPIDKUW H9 '/5 ,QVWLWXW IU
5RERWLN XQG 0HFKDWURQLN 2EHUSIDIIHQKRIHQ *HUPDQ\ FKDLUPDQ RI WKH SURJUDP
FRPPLWWHH
 +LOGLQJ(OPTYLVW'\QDVLP$%/XQG6ZHGHQ
 3HWHU )ULW]VRQ 3(/$% 'HSDUWPHQW RI &RPSXWHU DQG ,QIRUPDWLRQ 6FLHQFH /LQN|SLQJ
8QLYHUVLW\6ZHGHQ
/RFDORUJDQL]HUV
0DUWLQ2WWHU$VWULG-DVFKLQVNL&KULVWLDQ6FKZHLJHU(ULND:RHOOHU-RKDQQ%DOV
'HXWVFKHV=HQWUXPIU/XIWXQG5DXPIDKUWH9'/5,QVWLWXWIU5RERWLNXQG
0HFKDWURQLN2EHUSIDIIHQKRIHQ*HUPDQ\

Jensen J.M., Tummescheit H.

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

Moving Boundary Models

for Dynamic Simulations of Two-Phase Flows
Jakob Munch Jensen and Hubertus Tummescheit

Department

Department of Mechanical Engineering

Technical University of Denmark
jmj@mek.dtu.dk

of Automatic Control
Lund University, Sweden
hubertus@control.lth.se

Abstract
Two-phase flows are commonly found in components
in energy systems such as evaporators and boilers. The
performance of these components depends among others on the controller. Transient models describing the
evaporation process are important tools for determining control parameters, and fast low order models are
needed for this purpose. This article describes a general moving boundary (MB) model for modeling of
two-phase flows.
The new model is numerically fast compared to discretized models and very robust to sudden changes in
the boundary conditions. The model is a 7th order
model (7 state variables), which is a suitable order for
control design. The model is also well suited for open
loop simulations for systems design and optimization.
It is shown that the average void fraction has a significant influence on the system response. A new method
to calculate the average void fraction including the influence of the slip ratio is given. The average void
fraction is calculated as a symbolic solution to the integral of the liquid fraction profile.

Introduction

First principle mathematical models of dynamical systems are made for a range of purposes, but one of
the most common ones is to develop and verify controllers. The complexity of the model should be in accordance with the purpose of the model and this simple principle suggests that models for control design
should be of low order and preferably easy to linearize.
Unfortunately, physical systems are not sticking to this
class of models, on the contrary: most mathematical
first principle models are of distributed nature. The
natural way to describe such a model is partial differ-

The Modelica Association

235

ential equations (PDE). PDE are infinite dimensional

and their common numerical approximations, spatially
discretized PDEs using one of the many possible discretization schemes, are of high order and without further model reduction not well suited for control design. The problem of control-oriented modeling is to
derive a model which at the same time fulfills the requirements of control theory and characterizes those
features of the system which are needed to satisfy the
controller specification.
Moving boundary models for two phase flows in heat
exchangers are a good example of low order control
design models. They can be used for evaporators, condensers and steam generators. Their only disadvantage is that a number of mathematically rather different
models arise depending on the operating conditions of
the heat exchanger and the fluid conditions at the inlet
of the equipment.
The model presented in this paper covers the most general case of two-phase heat exchangers with subcooled
liquid at the inlet and superheated vapour at the outlet.
This flow configuration is commonly found in thermal
power systems, and the model can easily be extended
to condensers and heat exchangers with subcooled liquid at the inlet and two-phase at the outlet. The special
case of dry-expansion evaporators for refrigeration has
been derived in [6].
The idea of a moving boundary model is to dynamically track the lengths of the different regions in the
heat exchanger: the length from the inflow to the onset of boiling and the length of the two phase region.
Simulation results are given for an evaporator in an
organic rankine cycle, which utilizes the waste heat
from a gas turbine in a small power plant. Other references to MB models include B.T. Beck that describes
a MB-model for incomplete vaporization [2], a two region MB-model by He [4] and a three region model by
Willatzen [7].

Modelica 2002, March 1819, 2002

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

Equations (2), (4) and (5) are the differential balance

equations, which will be integrated over the three regions to give the general three region lumped model
for a two-phase heat exchanger. A schematic of the
model is given in Figure 2. It is assumed in the following analysis that the change in pressure along the
evaporator pipe is negligible.

Roman and Greek Letters

A
Cv
Cw
D
L
S
Vcyl

area
nozzle coef.
heat cap. of wall
diameter
length
slip ratio
cylinder volume

h
m
q
t
v
x
z

enthalpy
mass flow
heat flux
time
velocity
mass fraction
length coordinate

heat transfer coef.

liquid fraction
volumetric efficiency
void fraction
density ratio

density
pump speed
dissipation function
vapour generation

i
in
l
o
out
r
w

inner
inlet
saturated liquid
outer
outlet
refrigerant
wall

Jensen J.M., Tummescheit H.

Subscripts
1
2
3
12
23
amb
c
g

subcooled
two-phase
superheated
interface 1-2
interface 2-3
ambient
condensation
saturated gas

Figure 1: Schematic of the Three Region MB-model.

1 : subcooled, 2 : two-phase and 3 : superheated.

Superscripts

2.1

flux per length

flux per volume

flux per area

Mass Balance for the Subcooled Region

Integration of the mass balance (2) over the subcooled

region gives

Table 1: Notation used in the Moving Boundary Model

L1


A
dz
t


Governing Equations

The general differential mass balance is

v


(1)

which for the one-dimensional case can be written as

A
t

m
z


hv


Dp
Dt


m
dz
z

(6)

Applying Leibnizs rule (see Appendix A) on the first

term and integrating the second term give for a constant area pipe:
A

d
dt

L1


dz

A L1

dL1
dt


m 12


The general differential energy balance is

h
t

m in

0


(7)

(2) The density at the interface L1 is equal to the saturated liquid density l . Pressure and mean enthalpy h 1
define the state of the subcooled region where

L1


1
hin
2

h 1

(3)

hl


(8)


which can be simplified by neglecting the axial conThe inlet enthalpy hin is known from the boundary
ductivity, radiation and the viscous stresses and assumconditions and hl is a function of the pressure. The
ing one dimensional flow:
mean density in the subcooled region is approximated
by
mh

Ah Ap
D Tw Tr
(4)
1 L1
t
z
dz p h 1
(9)
1
L1 0
A simplified differential energy balance for the wall is
achieved by setting all flow terms in (3) equal to zero The mean temperature is calculated from the same
states as T1 T p h 1 . The mass balance for the suband neglecting the axial conductivity.
cooled region can be rewritten as
Tw
i Di Tr Tw o Di Tamb Tw
Cw w Aw
dL1
d 1
t
L1
m in m 12
(10)
A 1 l
(5)
dt
dt


Modelica 2002, March 1819, 2002

236

The Modelica Association

Jensen J.M., Tummescheit H.

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

L1
where 1 h 1 1 h1
0 hdz. The above equation
(15) is inserted into the energy balance (14), which
gives the final energy balance for the subcooled region:

The term d 1 dt is calculated using the chain rule:

d 1
dt

1 d p 1
p h dt h 1
1
1 1
p h 2 h 1
1 1 dhin
2 h 1 p dt


d h 1
p dt
dhl d p
p d p dt


(11)

1
A
2

1 hin


hl


dL1
dt

2l hl


1
dhin
h p dt
dhl
hin hl
1
dp


1 L1

The term dhin dt is determined from the boundary

conditions to the heat exchanger model. The expression for d 1 dt is inserted into the mass balance (10),
such that the final mass balance for the subcooled region reads:

L1


dL1
1
1 1
L1
A 1 l
dt
p h 2 h 1

1 1 dhin
m in m 12
L1
2 h 1 p dt


dhl d p
p d p dt

2.3




dhl
p dp


m 12 hl


dp
dt

2





Di L1 i1 Tw1


Tr1


Mass and Energy Balances for the TwoPhase and Superheated Regions

1 1
2 h

m in hin

(16)

1
p

(12) The mass and energy balances are integrated over the
two-phase region and the superheated region using the
2.2 Energy Balance for the Subcooled Region same procedure as for the subcooled region. The equations are derived in detail in Appendix B.
Integration of the energy balance (4) over the sub- The flow in the two-phase region is assumed to be hocooled region gives
mogeneous at equilibrium conditions with a mean den1 l , where the void fraction is
sity of g
L1 Ah Ap
L1 mh

A
defined
as

g A. The average void fraction is dedz

dz
L1 L2
1
t
z
0
0
dz and is assumed to be invari(13) fined as L2 L1
L1
ant
with
time.
A
detailed
model of the calculation of
D Tw Tr dz
0
the void fraction is derived in section 2.5. The mass
balance for the two-phase region is
Applying Leibnizs rule on the first term and integrating the other terms give for a constant area pipe and a
L1
dL2
1 l g
A l g
constant heat transfer coefficient
dt
dt
d
d
d
p
L
1
g
l
dL1
dp
d
L2
1
m 12 m 23
hdz A L1 h L1
AL1
A
dp
d p dt
dt 0
dt
dt
(14)
(17)
m 12 hl m in hin Di i1 L1 Tw1 Tr1


and the energy balance for the two-phase region is

The first two terms are evaluated as

d
dt

L1

0

hdz

dL1
dt
dL1
l h l
dt

L1 h L1


d
1 h 1 L1
dt
1
dL1
l h l
1 hin hl
2
dt
1
1
1
hin hl
L1 1
2
2
h1
1
dhl
hin hl
L1 1
2
dp
1
1 1 dhl
dp

p h 2 h1 p d p
dt


p


dhin
dt

(15)

dp
dt

(18)

Tr2


237

The Modelica Association

The derivative of the properties at the phase boundaries are written in a short notation and can be rewritten as e.g. d g hg d p hg dg d p g dg d p .
Both d g hg d p and d l hl d p can be calculated

d g h g
dp


d l h l
dp
dL1
g hg 1 l hl
dt
dL2
1 l hl g hg
dt
m 12 hl m 23 hg Di i2 L2 Tw2

A L2

Modelica 2002, March 1819, 2002

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

Jensen J.M., Tummescheit H.

from the pressure. The mass balance for the super- The wall temperature in the model is discontinuous at
L1 giving
heated region reads:
1 3
2 h 3

dhg 3
dp
A L3
p dp
p h dt
dL1
dL2
g 3
g 3
dt
dt
1 3 dhout
m 23 m out
L3
2 h 3 p dt


Tw L1


(19)

dL1
dt
dL1
Tw1 for
dt

Tw L1


Tw2 for

(24)

0


Similar expressions are derived for the walls adjacent

to the two-phase and the superheated regions see Appendix B. Typically in the literature a simplified mean
value for Tw L1 has been used, which seems attractive in order to simplify the model see e.g. [4] and [7].
Simulations show that the response times for the system for some test conditions depend significantly on
the expression for Tw L1 , and the full equations given
by (23) and (24) should therefore be used.
The general three region moving boundary model is
described by the mass and energy balances for the
flow stated in equations (12), (16), (17), (18), (19) and
(20) and the energy balances for the wall regions as
stated in equations (23), (49) and (51). In addition the
two discontinuous equations for the wall temperatures
as stated in (24) and (50) are needed. This equation
system contains 9 equations with the 7 state variables:
L1 L2 p hout Tw1 Tw2 and Tw3 . The variable hin ,
which also appears differentiated, is calculated as a
boundary condition and is thus not included in the
state variables for the MB-model. Dependent variables
can be calculated from the state variables and include:
1 l g 3 1 h 1 p 1 p h dl d p dg d p
3 h 1 p 3 p h h 1 hl hg h 3 dhl d p dhg d p Tr1
Parameters
are conTr2 Tr3 m 12 m 23 .
stant
during
simulation
and
include:
A Di Do i1 i2 i3 o L Tamb Cw w Aw . The
boundary models calculate the variables m in m out hin
and dhin dt , which are boundary conditions to the
MB-model.

The energy balance for the superheated region is given

by

A


1
3 hg
2

dL1 dL2
g h g
hout
dt
dt
1
1 3 dhg 3
L3
hg hout
2
2 h 3 p d p
p h
1 dhg
dp
3
1
2 dp
dt
1
1 3
dhout
hg hout L3
3 L3

2
4 h3 p
dt
m 23 hg m out hout Di i3 L3 Tw3 Tr3


(20)

The mean properties of the superheated region are

calculated in the same way as in the subcooled region. Thus h 3 0 5 hg hout , 3 p h 3 and
Tr3 T p h 3 .

2.4

Energy Balance for the Wall Regions

The energy balances for the walls are derived. Integration of the wall energy equation (5) from to gives

Cw w Aw

Tw
dz
t

i Di Tr


Tw dz


o Di Tamb

(21)

Tw dz


Applying Leibnizs rule, assuming constant wall properties give and rearranging gives the general energy
balance for a wall region:
Cw w Aw


Tw

dTw
dt
Tw

Tw

Tw


d
i Di
dt
Tamb Tw


2.5
Tr

Tw


o Do


dTw1
dL1
Tw1 Tw L1
dt
dt
i 1Di L1 Tr1 Tw1
o Do L1 Tamb

Modelica 2002, March 1819, 2002

Calculation of the Average Liquid Fraction

The liquid fraction in the two phase region z is related to the void fraction z via the equation
z


z


(25)


The same equation holds for the average values and

over the whole region, which are the parameters of
interest. It is computed as the integral over the normalized profile. For the derivation of a z profile, a
Tw1
(23) couple of assumptions are necessary:


For the wall region adjacent to the subcooled region

0 and L1 , which gives
Cw w Aw L1

(22)

d
dt

238

The Modelica Association

Jensen J.M., Tummescheit H.

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

1. All assumptions made in the derivation of the where

g

moving boundary model apply also to the derivaand

A
l
l
tion of the liquid fraction profile, in particular that
The boundary conditions at the length coordinates z
a constant pressure is assumed along the pipe.
0 0 and z 1 0 are
2. The profile can be evaluated under steady state
conditions. For the purpose of slow, start-up tran1
1
0
(30)
0
sients as well as for linearization purposes this
does not pose any restrictions. This means in par- From (28), (29) and the boundary conditions, the following function for z can be derived:
ticular that the pressure is in steady state.


3. The vapour generation rate is uniform over the

evaporator length.

1


z


z


1
S

(31)

1


4. The slip velocity ratio S ug ul between the The influence of the slip ratio S on the amount of satgas and the liquid velocities is constant along urated liquid in the evaporation region, can seen in
the evaporator length and a known function of Figure 2. z can be integrated symbolically to give:
the model states that also can be evaluated under
Liquid Fraction over z for S  1, 3, 5, and 7 and  0.01
steady state conditions1.


Al ul
z
Ag ug
g
z
l

0.2

0.4
0.6
DimensionlessLength z

(27)

0.8

Figure 2: Liquid Fraction z along the normalized

evaporation region.


u
z
1 u
S
z


z dz


S S


1 ln S
S 1 2


(32)

This can only be used together with the dynamic

model from the previous section when the time derivative of can be neglected. This holds for slow pressure
transients. The density ratio is a unique and simple
function of the pressure, but for the slip ratio S a number of empirical correlations are available to choose
from. Because of the assumptions made above, we
have to choose a slip ratio which is independent of the
(28) local void fraction or mass fraction. A simple and appealing correlation is the one from Zivi (1964) cited
(29) in [9] which minimizes the total kinetic energy flow
locally at each position z along the pipe:

1 Remark: It is possible to weaken this assumption and use a

slip ratio S z which is a function of the length coordinate. Many
of the rather complex empirical slip correlations depend on the lo as well and in this case the profile and
cal mass fraction x m g m
the integral over the profile can only be solved numerically. For
certain applications this may result in the most accurate approximation of the mean void fraction.


S  7.0

S  5.0
0.4

(26)

is the net generation of saturated vapour per unit

length kg ms , Al and Ag are the cross sectional areas taken up by liquid and vapour respectively and the
densities are independent of the length coordinate because we assumed no pressure loss and steady state
conditions for the pressure. This equation is normalized by setting A Al Ag 1 and letting the length
of the evaporation region run from 0 to 1 so that the
cross section area Al z is now equivalent to the liquid
volume fraction z . Then, replacing ul with u and ug
with Su and dividing by l the following normalized
equation is obtained:


S  3.0
0.6

0.2

S  1.0
0.8
liquid fraction

A similar derivation but assuming a slip velocity ratio

of 1 has been done in [3]. Under the above assumptions, the following coupled ODE boundary value
problem holds:

S


ug
ul


l
g

1 3

(33)

The Modelica Association

239

Using this slip correlation, the average liquid fraction

in the pipe becomes a function of only one variable,

Modelica 2002, March 1819, 2002

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

the density ratio .

1


z dz


1

2 3

2 3 ln 1

2 3

one for each of the lengths of the subcooled and

the two-phase regions. These are a combination of the mass and energy balances for the respecitive region.


Jensen J.M., Tummescheit H.

(34)
Three more eigenvalues come from the energy
Both the density ratio and the slip S approach 1 when
balances of the evaporator walls.
the pressure is rising toward the critical pressure. In
the limit, the liquid and vapour densities are equal as In [1] Bauer derived a more detailed, distributed model
well as the flow speeds, so that a mean liquid fraction of heterogeneous flow2 and validated it against meaof 0 5 is expected, compare the plot of in Figure 3. surement data for the refrigerant R22. According to
[1], the advantage of the heterogenous model over the
0.5
homogeneous one is that the void fraction turns out to
0.4
be more realistic. Therefore, the dominant time constants are modeled more accurately. The average void
0.3
fraction 1 has a strong influence on the total fluid
mass in the evaporator, as can be seen clearly from
0.2
Figure 4. It can be concluded from this argument that


liquid fraction

0.1

0.45

0
0.2

0.4
0.6
Density ratio

0.8

1
0.4

Figure 3: Average Liquid fraction as a function of

the density ratio .

Dominating Time Constants of the

Linearized Model

In this section the influence of some model parameters on the eigenvalues of a linearization of the system derived in section 2 is investigated. Models for
fluid flow exhibit two types of time constants: fast, hydraulic time constants for disturbances traveling with
the speed of sound and much slower thermal ones,
whose disturbances move at the flow speed. In two
phase flows, the coupling between thermal and hydraulic phenomena is much tighter than in one phase
flows, because a change in the hydraulic pressure is
tightly coupled to a change in the temperature. The
eigenvectors reveal that the 7 eigenvalues are tightly
coupled, but roughly their physical interpretation is as
follows:

Mass in the 2 phase zone

0.35

0.3

0.25

0.2

Velocity Slip

Figure 4: Total mass in the two-phase region as a function of velocity slip.

the void fraction is a crucial parameter in the moving

boundary model. Using a good approximation of the
void fraction, which may be obtained from a detailed,
distributed model like in [1], is important for obtaining
realistic dynamic behaviour.
The slow modes in the overall system are mostly influenced by the wall temperatures: higher heat capacities
and smaller heat transfer coefficients result in slower
modes. This means that the slowest mode usually is
governed by the pipe wall in the liquid region. Two
model parameters with a large influence on the slow
one mode comes from the overall mass balance time constants are the void fraction and the ratio
of the evaporator which depends on the ratio be- of the total heat capacities of fluid and pipe walls of
tween the total mass and the sum of the mass the evaporator pipes. The latter depend on the system
flows in and out of the evaporator and
pressure and the pipe diameter. Correct estimation of
one for the overall energy balance which depends
2 Heterogeneous flow means that the flow speeds of the gas and
on the ratio of the total heat capacity to the sum liquid phases can be different. A homogeneous flow assumption
of convective and heat transfer energy flows,
is equivalent to the same flow speed for both phases.

Modelica 2002, March 1819, 2002

240

The Modelica Association

Jensen J.M., Tummescheit H.

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

the void fraction gets more important at lower pressures because the slip increases due to smaller density
ratio and the heat capacity of the pipes is usually
smaller due to thinner pipe walls.
The root locus plot in Figure 5 shows the slow eigenvalues of the system, which are the dominating ones
for control design purposes. These vary significantly
when the slip ratio S (and thus the void fraction) is
varied from 1.0 to 8.0. In the example with approx.
31 bars the pressure is relatively high for the working
fluid R22 and therefore the slip ratio is not very far
from 1. Nonetheless, the slow eigenvalues move considerably on the root locus. The change in the model
dynamics will be larger at lower pressures.
0.2

0.1

The model derived above is still too complex and has

too many states for some purposes, e. g., online dynamic optimization as it is done in Model Predictive
Control (MPC). There are several ways to reduce the
number of states in the moving boundary model. One
possibility is to assume that the 2-phase heat transfer
coefficient is much higher than the outer heat transfer
coefficient so that the wall temperature and the fluid
temperature in the evaporation region are equal. This
assumption may also be extended to the subcooled
and superheated regions. The model will loose accuracy in the high frequency range but will be very
similar to the full model at low frequency range. Another possible simplification is to get rid of the states in
the superheated region, because it is usually short and
contains only few percent of the fluid mass. The dynamic model for the region can be replaced by a semiempirical algebraic relation for the superheat temperature, see [4]. Investigation of these options for model
reduction is the goal of future work by the authors.

4.1

Boundary Models

0.1

The test simulations for the heat exchanger are performed for a simple cycle containing a pump that sup0.2
plies the liquid flow into the evaporator and a nozzle
0.8
0.7
0.6
0.5
0.4
(turbine) at the end of the evaporator. The pump model
Figure 5: Root Locus for the slow eigenvalues. Dia- is defined by a simple expression for the mass flow
monds mark a slip ratio of 1.0, stars a slip ratio of the
(35)
m pump v pumpVcyl
test case of 1.7. Slip ratios vary from 1.0 to 8.0.


where v is the volumetric efficiency, pump is the inlet

density to the pump, Vcyl is the cylinder volume and
is the number of revolutions per second. The specific
4 Model Variants
enthalpy at the inflow of the evaporator is hin and is a
constant
below the saturated liquid enthalpy.
The main effect of using a velocity slip estimate to calculate the average void fraction is an improved model The model of the nozzle is computed as:
of the fluid mass in the evaporator. In a model for con(36)
m nozzle Cv out p pc
trol design around a narrow operating pressure the
short transient case a constant void fraction based
where Cv is a coefficient, out is the outlet density from
on the profile derived above, will give sufficiently acthe evaporator, p is the pressure in the evaporator and
curate results. Different slip correlations than the one
p is a constant pressure lower than p.
from above and numerical quadrature can be used to c
find a good estimate of the mean void fraction. A fixed
void fraction will be less adequate when long tran- 5 Simulation Result
sients over wide pressure ranges are to be simulated.
During the simulation of the start-up of a near- or su- The evaporator in an organic rankine cycle (ORC)
percritial once-through boiler, which is a classical case is simulated using the three region moving boundary
for a moving boundary model (see [3]), the density ra- model and the models for the pump and the nozzle.
tio will change by 3 orders of magnitude. In that The simulation program Dymola [5] has been used to
case the simple slip correlation S 1 3 works well to perform the simulations. The ORC is used to conmodel the fluid mass in the evaporator.
vert thermal energy to electric energy in applications


The Modelica Association

241

Modelica 2002, March 1819, 2002

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

3.25

3
10

10

20

30

40

50

60

70

80

10

20

30

40

50

60

70

80

90

L1 [m]
L2 [m]

3
2.5

L3 [m]

20

30

Time

40

50

60

70

80

90

The length of the subcooled and the two-phase region

is seen to increase as the pump speed is increased (at
t=0s). Also the pressure and the heating effect is increased which is expected. An increase in outer heat
transfer (at t=30s) results in shorter two-phase and superheated regions as well as in increased heating effect. The larger nozzle coefficient (t 60 s) results in
a decrease in pressure. The reduced pressure lowers
the boiling point and thus the fluid temperature in the
evaporation region. The length of the subcooled region
is therefore shrinking. The length of the two-phase region grows in this case but this trend depends on the
conditions. The lower evaporating temperature tends
to decrease the length of the two-phase region and the
larger latent heat increases it. The heating effect rises
in this case, but this trend depends on the conditions
as well. The model gives the right trends even though
no experimental data has been available to validate the
model.

8.5

1.5
10

10

90

8
10

Figure 7: Pressure in the evaporator

9.5

10

20

30

40

50

60

70

80

90

Time

Figure 6: Lengths of the three regions

Pump and nozzle parameters
v 0 6 [-]
60 [r ps]
Vcyl 1 30E


5 [m3 ]

Cv
pc

3 76E 5 [m2 ]
1 4 [M pa]

Evaporator steady state results



Q
m
p
Tw1
Tw2
Tw3
Tr1
Tr2
Tr3
L1
L2
L3


122

163 [kW ]
0 54 [kg s]
3 6 [M pa]
388 8 [K]
371 7 [K]
449 5 [K]
306 0 [K]
352 3 [K]
384 0 [K]
39m
87m
24m


120

Total heating effect in kW

L 15 [m]
Di 0 020 [m]
Do 0 022 [m]
Cw 385 [J kgK]
w 8 96E3 [kg m3]
i1 2451 [J m2 K]
i2 11404 [J m2 K]
i3 2071 [J m2 K]
o 500 [J m2 K]
Tamb 573 1 [K]
S 1 67 [-]
0 665 [-]


3.1

3.5
10

3.15

3.05

4.5

x 10

3.2

Pressure [MPa]

with small temperature differences between the high

and the low temperature heat sources. The ORC can
thereby be used to improve the energy efficiency in
gas turbine power plants by converting the waste heat
energy in the exhaust gas to electricity. Simulation results for a test case of an evaporator pipe are shown in
Figure 6 to Figure 8.

Jensen J.M., Tummescheit H.

118
116
114
112
110

108
10

Table 2: Parameters and steady-state results

10

20

30

40

Time

50

60

70

80

90

Three experiments are performed with parameters and

Figure 8: Heating effect to the evaporator
initial steady state results in table 2. At t 0 s the
pump speed is increased by 5%, at t 30 s the outer
heat transfer coefficient o is increased by 10% and at
5.1 Conclusions
t 60 s the nozzle coefficient Cv is increased by 10%.
Figure 6, 7 and 8 show the transient response of the A new moving boundary model has been presented
describing the dynamics of two phase heat exchangsystem.


Modelica 2002, March 1819, 2002

242

The Modelica Association

Jensen J.M., Tummescheit H.

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

ers with liquid at the inlet and vapour at the outlet.

The new model is numerically fast compared to discretized models and very robust to sudden changes in
the boundary conditions. The model is a 7th order
model (7 state variables), which is a suitable order for
control design. The model is also well suited for open
loop simulations for systems design and optimization.
It is shown that the average void fraction has a significant influence on the system response. A new method
to calculate the average void fraction including the influence of the slip ratio is presented. The average void
fraction is computed from the symbolic solution to the
integral of the liquid fraction profile.

Energy Balance for the Two-Phase Region

The energy balance (4) is integrated over the twophase region from L1 to L1 L2 . Applying Leibnizs
rule gives for a constant area pipe

d
dt

L1 L2


hdz

L1

A L1 h L1

dL1
dt
L2

d L1
dt
Di i2 L2 Tw2

A L1


L2 h L1

m 23 hg

m 12 hl


L2

dp
dt
d p (41)
AL2
dt
AL1

Tr2


The first term is evaluated as

d L1 L2
d L1 L2
hdz
g hg 1 l hl dz
dt L1
dt L1
d
g hg 1 l hl L2
Appendix A: Leibniz's Rule
dt
d g h g
d l h l d p
Leibnizs rule for differentiation of integrals with time
L2
1
dp
dp
dt
varying limits reads ([8]):
dL2
g hg 1 l hl
dz2
dz1
d z2
dt
f z t dz f z2 t
f z1 t
dt z1
dt
dt
(42)
(37)
z2 f z t
Inserting (42) into (41) gives the final energy balance
dz
t
z1
for the two-phase region as state in (18).


Appendix B: Derivation of the Model Mass Balance for the Superheated Region
Equations
Mass Balance for the Two-Phase Region

The mass balance (2) is integrated over the superheated region from L1 L2 to L which for a constant
area pipe gives

The mass balance (2) is integrated over the two-phase

region from L1 to L1 L2 Applying Leibnizs rule
gives for a constant area pipe

L


A
dz
L2 t

d
A
dt

L1 L2


dL1
dz A L1
dt
d L1 L2
A L1 L2
dt
m 23 m 12 0

L1

(38)

L1

L1 L2

m
dz
z

0


(43)

Applying Leibnizs rule on the first term and integrating the second term give for a constant area pipe
d
A
dt


d L1 L2
L2
dt
m out m 23

L1 L2

dz

A L1

L


0 (44)

The flow is assumed to be homogeneous at equilibrium The mean density in the superheated region is 3
conditions with a mean density of g 1 l . 1 L
p h3 , which inserted in the mass
L3 L1 L2 dz
The mass balance for the two-phase region becomes
balance (44) gives
d
dL1
dL2
d3
dL1
dL2
2 L2
l g
g
A
g 3
g 3
A L3
dt
dt
dt
(39)
dt
dt
dt
m 12 m 23
m 23 m out (45)
1 l . The time derivative of 2
where 2 g
The derivative of 3 is calculated as
is
dg
dl d p
d2
d3 3 d p 3 dh
(40)

1
dt
dp
d p dt
dt
p h dt
h p dt
which inserted into the mass balance (39) gives the fi1 3 dhg 3
d p 1 3 dhout
nal mass balance for the two-phase region as stated in
2 h3 p d p
p h dt 2 h3 p dt
(17).
(46)


The Modelica Association

243

Modelica 2002, March 1819, 2002

Moving Boundary Models for Dynamic Simulations of TwoPhase Flows

Jensen J.M., Tummescheit H.

3
The expression for d
dt is inserted into (45), which For the wall region adjacent to the superheated region
gives the final mass balance for the superheated region L1 L2 and L, which inserted in (22) gives
as stated in (19).
dTw3
dL1
Tw L1 Tw2
Cw w Aw L3
dt
dt
Energy Balance for the Superheated Region
dL1 dL2
Tw L1 L2 Tw3
dt
dt
The energy equation 4 is integrated over the superi 3DiL3 Tr3 Tw3
o Do L3 Tamb Tw3
heated region from L1 L2 to L. Applying Leibnizs
(51)
rule gives for a constant area pipe

d
dt

L


hdz

L1 L2

A L1

L2 h L1

L2

d L2
dt


dp
m out ho ut m 23 hg
dt
Di i3 L3 Tw3 Tr3

AL3


L
L1 L2

hdz

d
3 h 3 L3
dt


References


[1] Olaf Bauer, Modelling of Two-Phase Flows with

Modelica, Masters Thesis ISRN LUTFD2/TRFT
5629SE, Department of Automatic Control,
Lund University, November 1999.
(47)
[2] B.T. Beck and G.L.Wedekind, A generalization of
the system mean void fraction model for transient
two-phase evaporation flows, Int. J. of Heat Transfer 103 (1981), 81 85.

The first term is calculated as

d
dt

1
d L1 L2
3 hg hout
2
dt
1
d 3
L3 hg hout
2
dt
1
dhg d p dhout
3 L3
2
d p dt
dt

(48)


[3] S. Bittanti, M. Bottinelli, A. De Marco, M. Facchetti, and W. Prandoni, Performance Assessment

of the Control System of Once-Through Boilers,
Proceedings of the 13th Conference on Process
Control 01, June 2001.

[4] X.D. He and S. Liu, Multivariable Control of

where h 3 12 hg hout and 3 p h 3 . Equation
Vapor Compression Systems, HVAC Research 4
d 3
(1998), 205 230.
(48) and the expression for dt from equation (46) is
inserted into (47), which after some rearranging gives
[5] http://www.dynasim.se.
the final energy balance for the superheated region as
stated in (20).
Jensen and H.J. Hoegaard Knudsen, A new
[6] J.M.
moving boundary model for transient simulations of dry-expansion evaporators, Proceedings
Energy Balance for the Walls
of the 15th International Conference on Efficiency,
For the wall region adjacent to the two-phase region
Costs, Optimization, Simulation and Environmen L1 and L1 L2 , which inserted in (22) gives
tal Impact of Energy Systems, July 2002.

dTw2
dt

Cw w Aw L2


Tw L1

Tw2

Tw2


Tw L1


o Do L2 Tamb

Tw L1 is given by (24), and Tw L1



Tw L1

L2

Tw L1


L2

L2

dL1
dt


i 2Di L2 Tr2


Tw2


[8] N.E. Todreas and M.S Kazimi, Nuclear Systems

I, Thermal Hydraulic Fundamentals, Taylor and
Francis, 1993.

Tw2


L2 is given by

dL2
Tw3 for
dt
dL2
Tw2 for
dt

Modelica 2002, March 1819, 2002

dL2
dt

[7] N.B.O.L. Pettit M. Willatzen and L. PlougSrensen, A general dynamic simulation model for
evaporators and condensers in refrigeration, Int.
(49)
J. of Refrigeration 21 (1998), 398 414.

0


[9] P.B. Whalley, Boiling, Condensation and GasLiquid Flow, 1987.

(50)

0


244

The Modelica Association