TALLINN UNIVERSITY OF TECHNOLOGY

School of Science
Department of Cybernetics

Kaspar Selke 212094YAFM

S IMULATING P RESSURISED WATER R EACTOR USING

M ODELICA

Master’s Thesis

Supervisor: Marti Jeltsov

PhD

Tallinn 2024
 TALLINNA TEHNIKAÜLIKOOL
Loodusteaduskond
Küberneetika instituut

Kaspar Selke 212094YAFM

S URVEVEEREAKTORI MODELLEERIMINE M ODELICAGA

Magistritöö

Juhendaja: Marti Jeltsov

PhD

Tallinn 2024
 Author’s Declaration of Originality

I hereby certify that I am the sole author of this thesis. All the used materials, references
to the literature and the work of others have been referred to. This thesis has not been
presented for examination anywhere else.

Author: Kaspar Selke

01.01.2024

1
 Supervisor’s Approval

The thesis complies with the requirements for Master’s theses.

Supervisor: Marti Jeltsov

01.01.2024

2
 Abstract

In this thesis we created a preliminary pressurised water reactor model. The model was
developed using Modelica language with the Dymola interface. The main library used for
the reactor components was the TRANSFORM library. The final system reaches steady
state on its own and three different transients tests were performed and analyzed. The
reactor model was a success and the goal of the thesis was achieved although further
validation and verification is needed. Multiple steps have to be taken before real world
decisions can be made based on the model.

3
 Annotatsioon
Surveveereaktori modelleerimine Modelicaga

Antud töös lõime esialgse surveveerreaktori mudeli. Mudel valmistati kasutades Modelica
programeerimis keelt koos Dymola liidesega. Pealmine teek mida kasutati reaktori kom-
ponentideks oli TRANSFORM. Lõplik süsteem saavutab stabiilse seisundi enda pea ja
kolm erinevat katse olukorda testiti ja analüüsiti. Reaktori mudel oli edukas ja töö eesmärk
saavutati, kuigi järgnevat validatsiooni ja tõestamist on tarvis. Mitu sammu peab veel
võtma enne kui mudel on valmis päris maailmas kasutamiseks ja otsuste võtmiseks.

4
 Table of Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1 Basic Nuclear and Reactor Physics . . . . . . . . . . . . . . . . . . . . . 10
3.1.1 Fast Fission Factor . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.2 Resonance Escape Probability . . . . . . . . . . . . . . . . . . . 12
3.1.3 Thermal Utilization Factor . . . . . . . . . . . . . . . . . . . . . 14
3.1.4 Reproduction Factor . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1.5 Fast Non-leakage Probability . . . . . . . . . . . . . . . . . . . . 14
3.1.6 Thermal Non-Leakage Probability . . . . . . . . . . . . . . . . . 15
3.1.7 Point Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.1 Pipes and Flow Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.1.1 Pipe Parameter Selection . . . . . . . . . . . . . . . . . . . . . . 18
4.2 Pressurizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.2.1 Pressurizer Parameter Selection . . . . . . . . . . . . . . . . . . 24
4.3 Core and Fuel model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3.1 Core and Fuel Model Parameter Selection . . . . . . . . . . . . . 27
4.4 Heat Exchanger Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4.1 Heat Exchanger Model Parameter Selection . . . . . . . . . . . . 29

5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.1 Transients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.1.1 Core Inlet Temperature Change . . . . . . . . . . . . . . . . . . . 34
5.1.2 Loss of Secondary Loop . . . . . . . . . . . . . . . . . . . . . . 35
5.1.3 Leak in Primary Circuit . . . . . . . . . . . . . . . . . . . . . . . 38
5.2 Future works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

7 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5
Appendix 1 – Non-Exclusive License for Reproduction and Publication of a
Graduation Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

Appendix 2 – Dymola tips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

6
1. Introduction

Computer modelling is a powerful process to forecast and understand the events that
haven’t yet taken place in the physical world. It allows us to learn about and design
systems without having to spend the materials and facing the difficulties of building a
physical prototype. As nuclear systems are large and complex, it is challenging to perform
physical, real world experiments and computer modelling has become the main design
and analysis tool. However, computer modelling isn’t perfect and comes with its’ share
of issues. The most well-known being its dependence on human knowledge of physics,
human error and lack of validation data. Most of these are inherent to computer modelling.
Within this field the things we can optimize are accuracy with better physical models,
computational time with faster code and for large models the large overhead can be reduced
by using more efficient systems or languages. These three are what Modelica is trying to
achieve. To reduce the cost of building and maintaining models while also making them
reasonably fast a library of compatible components with well defined physical assumptions
is provided. This thesis aims to test Modelica and its development environment Dymola
by developing a Pressurized Water Reactor (PWR) model using the TRANSFORM library.
This research aligns with Estonia’s considerations on deployment of nuclear power to
meet its power system and climate goals and supports the plans by developing reactor
technology and modelling related local expertise. The reason for selecting a pressurized
water reactor over a boiling one is that the PWR core that consists of single-phase, all liquid
water is somewhat simpler to model compared to very non-linear turbulent two-phase
in case of a BWR. Since the purpose of the thesis is to gain the knowledge of how to
work with Modelica and develop a full reactor model, the PWR type was chosen. Since
currently there is no prevailing reactor modelling framework or tools in Estonia, options
like Modelica might have use in the future while also giving students the possibility to
practise and learn nuclear computer modelling.

7
2. Background

In this section a brief history of Dymola and Modelica is presented and their main work-
ing principles described. Also the main collection of components used in the model is
introduced.

Modelica is a freely available modelling language developed and maintained by the

Modelica Association [1]. It was first released in 1999 alongside Dymola which was
the first commercial tool that supported the language. Dymola is a component based
tool for modelling and simulation [2]. The original idea behind Dymola dates back to
1970’s and were developed by Dr Hilding Elmqvist as a PhD thesis. Modelica has a free
editing tool as well called OpenModelica, but due to differences in how they compile the
models the library required for building a nuclear reactor was only available using Dymola.
Later Dymola became a part of Dassault Systèmes. It solves the problems of overhead
and computational efficiency in many ways. First of them being a component-oriented
approach allowing for components of different domains such as electrical, hydraulic,
thermal etc. to be combined and done so with efficiency since all components are checked
beforehand. The second is an equation based approach meaning all models are described
mathematically using differential, discrete, and algebraic equations. Third and possibly
the most useful feature is Symbolic Manipulation. Most modelling and programming
languages use the form variable = expression and code is read procedurally. This
means that first variables are given values and then a computation can be made based on
an expression with those variables and the value given to a new variable. This requires
that the model developer manually rearranges the equations to suit the problem for every
possible way of calculating a given necessary value. For example, having acceleration as
input but also sometime requiring that instead of acceleration, speed or position can be
given. This usually causes one model to have multiple versions causing large overheads
due to maintenance. The work required increases with rising complexity. Dymola uses
expression = expression form so that the program itself can arrange the equations as
needed to suit a particular problem. For example we might have an line that computes
x = y ∗ z. Traditionally if we had x, z given and wanted the value of y instead we would
need to write a separate line y = xz . With more complex equations this adds up to multiple
lines of code and can even create compounding amounts of work required. Dymola on
the other hand can do this automatically if at any point any of the values in an expression
need to be calculated it can rearrange the equation as required. In practice this Symbolic
Manipulation also gives us an increase in speed due to less equations required. Another

8
reason why it was decided to use Modelica and more specifically Dymola in this thesis
work is due to the TRANSFORM library. It is a library of components designed for
building nuclear reactors. Although most of the focus has been on molten salt reactors, it
has the necessary components for more common reactor types such as pressurized water
reactors. The library is developed by Scott Greenwood [3]. The library was developed
with the idea that the user learns through examples that have been included. This means
that it has minimal documentation and the learning curve for new users is extremely steep,
especially those who have no previous experience with nuclear modelling or modelling
in general. Although active development of TRANSFORM has stopped it still receives
bug fixes semi regularly. An additional reason for choosing Modelica is it’s successful and
wide use in variety of fields. This gives the students better opportunities in the future while
also giving them the skills to focus further into other nuclear modelling systems.

9
3. Theory

In the thesis we are assuming that the reader has basic knowledge of modelling and nuclear
physics ideally up to the level of cross-sections. If the reader wants to model their own
reactor, then basic knowledge of Modelica language is also required. In this section we
will go over the basic nuclear reactor physics required to understand how a reactor core
operates, the main value used to describe the activity of the reactor and the nuclear kinetics
model this thesis uses.

3.1 Basic Nuclear and Reactor Physics

These sections and related subsections about reactivity and multiplication factor were
written based on U.S Department of Energy nuclear energy handbook volume 1 [4] and
volume 2 [5].

Reactors work by using nuclear reactions with fissile material to be in a constant critical
state. The critical state is a state of self-sustaining reactions where all excess neutrons
either escape or are captured by something other than the fuel. Typical fission reactions
have two fission products, some number of prompt neutrons, prompt gamma and then
released energy. The probability of this reaction happening is dependent on the energy
of the neutron as seen in figure 1. Here a larger cross-section can be viewed as a higher
probability of the event happening.

10
Figure 1. U-235 fission cross-section dependence on neutron energy. The blue indicates
energies of thermal neutrons and the green is the resonance region. Source: OECD NEA
Janis database [6].

From figure 1 we can also see why we want to reduce the energy of the neutrons. It is
because neutrons of lower energy have a higher probability to produce fission reactions
leading to more useful energy and we also want to avoid the resonant part which can be
see in the graph. It is also important to know that the delayed fission reactions can produce
neutrons upon decay and these are important in keeping the reactor stable and have to
therefore be considered in modelling. Furthermore the heat produced by the decay of these
materials continues on for a long time after the main fission reaction has stopped. The
number of neutrons in a reactor in time can either decrease, increase or stay the same. This
"trend" is represented in nuclear reactor physics by the effective multiplication factor kef f
which is relevant for finite reactors.

k < 1 → subcritical
k = 1 → critical
k > 1 → supercritical

For reactors which are very well known and defined we can use the six-factor formula to
calculate kef f . In our case the calculation for kef f is more complex and will be explained
later into the chapter. It is still useful to explain six-factor formula as it gives us a good
overview how the reactor stays at a critical state and other factors that help maintain it.

To understand kef f we must first understand k∞ , which is known as the infinite multiplica-

11
tion factor. Both tell us how many neutrons were produced from fission in one neutron
generation compared to how many neutrons were lost in the previous generation. This
means that if k∞ = 1 then the same amount of neutrons were produced as were lost,
meaning there is no change in the neutron population and the reaction is self-sustaining.
The formula for calculating k∞ is called the four-factor formula. One has to consider an
infinitely large system which has no neutron leakage due to being infinite. In this system
four parameters define the multiplication ability of the system. These four parameters
are the fast fission factor, resonance escape probability, thermal utilization factor and
reproduction factor.

3.1.1 Fast Fission Factor

The fast fission process ϵ is the first process that a new generation of neutrons takes part in.
It is the fission of neutrons at much higher than thermal energies. In figure 1 the energy
of fast neutrons is around 1 MeV. This process has a lot lower probability of happening
compared to fission at thermal energies as can be seen in figure 1 so it only creates a
small amount of fission but this also depends on the concentrations of the fuel and the
arrangement of the core.

3.1.2 Resonance Escape Probability

The way for neutrons to reduce in energy is by colliding with a moderator like water. Each
time they collide with moderator molecules they loose some amount of energy. Ideally
we would want them to end up at thermal energies which is noted in blue in figure 1.
Sometimes they lose a certain amount of energy and end up colliding with fuel while in
the resonance region seen in the same figure as green. The resonance escape probability
p is the probability of neutrons reaching thermal energy without being captured in this
resonance zone and is defined as the ratio of neutrons that reach thermal energies compared
to the number of fast neutrons that are slowed down. Elements in the core such as 238 U
and other isotopes have wider resonance bands and there for contribute strongly to the
absorption of neutrons in this region. The value of resonance escape probability thus
strongly depends on the fuel composition and also the geometry of the core. If neutrons
manage to escape the fuel matrix they have much higher chance of slowing down without
reacting to 238 U or other isotope atoms resulting in a higher resonance escape probability.
Resonance escape probability is also one of the main safety features and contributes to
reactor stability.

12
Figure 2. Effective multiplication factor (kef f ), resonance escape probability (p), and
thermal utilization factor (f) dependence on moderator to fuel ratio. The y-axis is the
neutron generation multiplication factor. [7]

These days most light water reactors are designed in the under moderated region seen in
figure 2. This means that the reactors aren’t actually designed at the optimal point. As
the moderator temperature increases the density decreases causing shift in figure 2 to the
left meaning that more neutrons are absorbed in the resonance region, that is if the reactor
is operating in the under moderated zone. This decreases the amount of neutrons taking
part in fission reactions and in turn lowers the power output, which will under normal
circumstances reduce the temperature of the moderator and the core will slowly stabilize.
This is also known as the moderator temperature coefficient (MTC). The second way it
helps to keep everything stable is Doppler broadening. This is the fastest feedback we have
since it takes place within the fuel and is considered almost instant. The aforementioned
MTC takes time since the water has to warm up. The neutron production of the fuel
changes with temperature and in most reactors this change coefficient is negative meaning
higher temperatures will reduce the neutron production rate. The simplified theory behind
the Doppler effect is that when the fuel nuclei lattice, that is the atomic structure of the
fuel, starts to vibrate the relative velocity between neutron and nucleus changes. The effect
of this change in relative velocity is the broadening of the resonant capture area and a
shortening of its cross-section. Meaning neutrons can now be captured at higher energies
by the resonance region, thus less neutrons are left for fission processes.

13
3.1.3 Thermal Utilization Factor

Next come the neutrons that didn’t get absorbed into the resonance region. If they remain
in the core they will reach thermal energies, but before fission they also have the probability
of being absorbed by the various other materials inside the core such as cladding, control
rods, moderator and so on. The thermal utilization factor f there for is the ratio of neutrons
that get absorbed into the nuclear fuel divided by the number of neutrons that are absorbed
into all the materials in the core. This process of non-fuel materials absorbing thermal
neutrons is also referred to as parasitic absorption.

3.1.4 Reproduction Factor

The last step the neutrons go through is reproduction. This is where finally the neutrons
reach the fuel and induce a fission reaction. To consider the reproduction factor in fresh
uranium fuel only three uranium isotopes have to be considered but as the fuel burns
up, different isotopes are created which then also have to be taken into consideration
when calculating the reproduction factor. There is also a chance that even if the neutrons
get absorbed they don’t produce a fission reaction but rather result in radiative capture,
creating a new compound nucleus and not producing any neutrons or a different amount.
The reproduction factor η is there for the ratio of fast neutrons produced by thermal fission
divided by the number of thermal neutrons absorbed into fuel.

These four together form the four-factor formula also called the infinite multiplication
factor k∞ :
k∞ = ϵ p f η (3.1)

From this we can get kef f by adding two more factors: the fast non-leakage probability
and the thermal non-leakage probability. When before we considered an infinite reactor
now we are considering a finite one. Due to this we have to add the two factors.

3.1.5 Fast Non-leakage Probability

When we were assuming an infinite reactor leakage could be neglected but now moving to
a finite reactor we can’t anymore. The fast non-leakage probability Pf characterizes the
amount of fast neutrons that do not leak out of the reactor. It is a ratio of the number of
fast neutrons that do not leak over the number of fast neutrons that were produced.

14
3.1.6 Thermal Non-Leakage Probability

Very similar to fast non-leakage probability, only this time its the leakage of thermal
neutrons PT and is defined as the ratio between the number of thermal neutrons that do not
leak from the reactor core and the number of neutrons that reach thermal energies. Both of
these probabilities are influenced by the moderator temperature. An increase in moderator
temperature results in an increase of leakage. This is one of the main ways the moderator
temperature coefficient is kept negative in PWRs. These probabilities can sometimes be
combined into one called the total non-leakage probability.

Having added the two factors we have now reached an adequate level of understanding of
how the core stays stable. The six-factor formula there for is:

kef f = ϵ p f η Pf PT (3.2)

kef f is one of the most important parameters and usually tells us a lot about the status
of the reactor. Instead of kef f however a different parameter called reactivity ρ is more
commonly used which is a measure of a reactors relative departure from criticality.

kef f − 1
ρ= (3.3)
kef f

When changing to rho the critical state is then determined as:

ρ < 0 → subcritical
ρ = 0 → critical
ρ > 0 → supercritical

As was discussed earlier the computation for kef f in a less well defined/studied reactor is a
lot more complicated. It usually requires the solving of an eigenvalue problem where kef f
is the eigenvalue of that system. The explanation of how this works is outside the scope of
this thesis.

3.1.7 Point Kinetics

This subsection was written based on the book "Dynamics and Control of Nuclear Reac-
tors"[8].

To know how much power our reactor is making we need a set of ordinary differential

15
equations that describe the time evolution of neutron density. Dynamic equations can be
formulated in the following logic:

Rate of change of quantity = rate of production – rate of losses (3.4)

Using this logic the equations for a point kinetics model can be presented as follows:

∂n
= Fp PT + Pd − La+L (3.5)
∂t
∂Ci
= Fd PT − Ldecay , i = 1, 2, . . . , 6 (3.6)
∂t

Here n is the neutron density per unit volume, Fp is the fraction of neutrons released
promptly, PT is the total rate of neutrons produced by fission, Pd is the rate of release for
delayed neutrons, La+L is the rate at which neutrons are lost by absorption and leakage,
Fd is the fraction of neutrons that are released by delayed neutron precursors, Ci is the
concentration of i-th delayed neutron precursor, and Ldecay is the rate of decay of the
delayed precursors.

From here we can define β which is defined as the total fraction of neutrons produced that
are delayed and βi will be the fraction of fission reactions that result in delayed neutron
precursors Ci being produced. We get the loss of precursors Ci by using a radioactive
decay constant λi as λi Ci . Typically precursors are grouped into six groups. This is mainly
done because if we do it by isotope we won’t know if we have missed a specific precursor.
It has been shown to work well and determining all delayed neutron precursors is a difficult
challenge. Also there is the added benefit of speeding up the model computation. With
these additions we can write the previous equations as:

6
∂n X
= (1 − β)PT − La+L + λi Ci (3.7)
∂t i=1
∂Ci
= β i PT − λ i C i i = 1, 2, . . . , 6 (3.8)
∂t

We can write eq 3.7 as:

6
∂n La+L X
= PT [(1 − β) − ]+ λi C i (3.9)
∂t PT i=1

We can now see that LPa+L

T
is actually kef f since the fraction is the ration of total neutrons
produced by fission divided by the total amount of neutrons lost. Using this we can turn

16
the equation into the following:

6
∂n 1 X
= PT [(1 − β) − ]+ λi C i (3.10)
∂t kef f i=1

From earlier when we defined reactivity we can use equation 3.3 and in turn we get:

6
∂n X
= PT (ρ + β) + λi C i (3.11)
∂t i=1

The total neutron production PT can also be written as:

PT = νΣf nv, (3.12)

where ν is the average number of neutrons produced per fission, Σf is the macroscopic cross
section for fission, and v the velocity of neutrons. It can be shown that the multiplication
of νΣf v is the time between the production of a neutron and the production of the next
generation of neutrons by fission called Λ. After replacing it in we get:

6
∂n (ρ − β) X
= n+ λi Ci (3.13)
∂t Λ i=1
∂Ci βi
= n − λi C i i = 1, 2, . . . , 6 (3.14)
∂t Λ

These are the final equations for point kinetics.

17
4. Model

This section will explain the different components and options we choose for the model.
In addition we will explain the different helpful options for working with Dymola. For
extra tips about working with Dymola refeer to the appendix 7.

4.1 Pipes and Flow Model

Pipes are the fundamental component in Modelica and TRANSFORM for building hydro-
dynamic models. Most components are created on the basis of pipes or volumes. Realistic
flows of water are complex and difficult to simulate. Because of this we make certain sim-
plifications that don’t compromise too heavily on the accuracy of the simulation. The pipes
used in the model are native pipes to the TRANFORM library. They use a one dimensional
Finite Volume Method flow of water with mass, momentum and energy partial differential
equations. The medium parameters were set to follow the IF97 standard [9] and were
explicit in pressure and enthalpy. Trace substances and heat dispersion along the circuit
weren’t considered although the possibility to do that should be theoretically available.
Simple mass flow pumps were used to drive the flow of water at a steady rate. These
pumps ignore cavitation and other issues (e.g. multi-dimensional flow characteristics) that
do take place in real pumps. The use of more complex pumps caused numerical instability
and are not the focus of this work.

4.1.1 Pipe Parameter Selection

Since in reactors there are pipes with many different shapes, sizes and properties we
need to understand how select proper parameters to best express what happens in the
reactor. When we mention pipe we usually mean the component named GenericPipe_-
MultiTransferSurface, there are of course other pipes for different use cases that work
more or less the same way with added components such as heat capacity. We can ignore
heat capacity of the pipe because compared to the heat capacity of the circulating water it
is small and is something that can easily be added later when a working model is achieved.
Adding the pipe component to the diagram it will have an arrow indicating flow direction
and two ports. The ports will look different, one will have an outline of a circle around
it and the other won’t. These can be changed under the Advanced options using the
exposedState setting. Further explanation of these can be found in the TRANSFORM
library -> User’s Guide -> Connectors. The pipe has many properties which can be changed.

18
First one being nParallel which allows us to basically have the same component in parallel.
When changing this parameter, it is important remember that the flow coming in is then
divided between the pipes or components. Next is Medium, which allows us to change the
medium of the system. It is best to start off with “Water using the IF97 standard, explicit in
p and h, recommended for most applications”, later one can add trace elements or change
the medium as needed. Going on down the list we will see Geometry which is a drop down
menu with many options. The generic options are mostly partial models and require the
user to complete the equations themselves. In the beginning it is easier to use StraightPipe
option. After choosing the option next to the dropdown arrow there is a gray box. Clicking
on the box brings up a submenu with multiple options. While this box is small and easily
missable, it is important to define the actual parameters for the geometry of the pipe. Other
components will have similar unnoticeable gray boxes so it is important to pay attention
to these. Within the Geometry submenu the first option that one might need to change
is nV. This is the number of volume nodes in the pipe, one can imagine the pipe being
cut into multiple volumes on each volume calculations are made, the more volumes we
have the more “accurate” the pipe simulation will be at the cost of simulation runtime.
Functionally this is the resolution of the pipe. Depending on different uses the number
required will change. For example, heating a pipe and having the water turn into vapor at
some point might require more nodes, more accuracy. For simple flow 4 volume nodes
will be enough. The nSurfaces setting can be kept at 1 for most uses. The other primary
settings to change are dimension, length, and angle. Dimension assumes that the pipe is
circular if one has a different pipe they might have to change the crossArea parameter
instead. It is a good idea to familiarize oneself with the concepts of wetted perimeters
and hydraulic diameter. The angle parameter is very important to pay attention to as this
dictates the direction of the flow. -90 degrees corresponds to a downwards pipe while
90 degrees is upward. The mistake of forgetting to change this parameter can be made
easily upon copying pipes, changing their directions, and then receiving anomalous results.
With this we can close the submenu and move on to the flow model. Most of the times we
want to leave this on default unless it is certain that a different flow model is necessary. In
the submenu minor loss coefficients can be added but are not necessary and can be left
default. Heat transfer options are quite self-explanatory. Turning use_HeatTransfer option
to true adds a red connection node to the pipe in diagram view. Different heat sources
can now be connected to the pipe. In the settings there is also the option to change the
coefficient of heat transfer to different values but for simple models this can be left as
Ideal in the beginning. We also have the option for InternalHeatGen if we want to for
example, simulate the heat produced by a reactor core. The GenericHeatGeneration option
is sufficient for this. The submenu allows to add per volume/node heat generation. If
one knows the heat generation of the whole reactor a simple trick is to divide the value
by number of nodes using [Name of pipe].geometry.nV. A useful tip for navigating the

19
different variables is to press Ctrl+Spacebar to bring up a menu of paths for different
components and commands, this can only be accessed if there is no text after the cursor.
We also have the same options for trace substances as we did for heat, which work in a
similar manner. Trace components are outside the scope of this thesis. Next tab we take
a look at is the Advanced tab. Here many features can be found like the aforementioned
exposedState option. Most noteworthy category is Dynamics. Each of the submenus will
have the same four options. The detailed description of the different options can be found
here [10]. In general, we use the FixedInitial option. If we know our system is in steady-
state and have good steady-state parameters we can use SteadyState. To find SteadyState
parameters we can initialize the system using FixedInitial or DynamicFreeInitial and seeing
at what values the system is in a steady state and then changing the parameters to those.
Although it sounds simple it often isn’t and requires a lot of tweaking to initialize properly
especially with components like the heat exchanger. In the initialization tab is where the
user can input the initial values of the simulations. All options will have an a and b point.
The rest of the intermediate values will be by default calculated as a linear spacing between
those values. It is important to note the units which can be changed by clicking on them. If
the temperature option isn’t available, and enthalpy isn’t known, then the enthalpy can be
found by using [Name of component].Medium.specificEnthalpy pT(p,T). This pressure
(p) and temperature (T) need to be in Pascals and Kelvins respectively. One can find a
function for conversions in Modelica.Units.Conversions. The Species Mass Fraction and
Trace Substances sections can be ignored.

4.2 Pressurizer

One of the most important parts of the pressurized water reactor is the pressurizer. This
component keeps the water pressure at a pre-determined high enough level so it doesn’t
boil and evaporate in the core, ensuring stable single-phase heat transfer from the core. The
component has 3 ports: 1 connected to the main loop, 1 for the spray systems, and 1 for
the pressure relief valve. There are also options to add heating to the vapor and fluid, with
additional heat dispersion options. Only the fluid heating system was used in this model.
The spray and pressure relief systems were driven using a proportional–integral–derivative
(PID) controller and the heating was controlled by a hysteresis component, which was
used to keep the pressure between certain values so the heater isn’t constantly on. The
balance equations used in the pressurizer were split into two, one for vapor and the other
for liquid. The mass balance equations are

ṁliquid = ṁcond − ṁevap + ṁsurge + ṁspray + ṁcSpray + ṁwl (4.1)

20
and
ṁvapor = ṁevap − ṁcond + ṁsteam − ṁcSpray + ṁvapSpray − ṁwl (4.2)

ṁsurge , ṁsteam , and ṁspray are the values of mass flow rates from the three ports. ṁcond
and ṁevap are the mass flow rates of condensation and evaporation respectively. ṁcSpray
and ṁvapSpray are the condensate and vaporization mass flow rates of when water is injected
into the spray port. These values are calculated differently based on the enthalpy of the
spray compared to the vapor and liquid inside the tank. ṁwl is the flow rate of vapor into
liquid at the interface of the two phases. In Modelica code it looks like this:

mb_f low_liquid = W _cBulk − W _eBulk + W _surge

+ W _liqSpray + W _cSpray + W _vl, (4.3)

and

mb_f low_vapor = W _eBulk − W _cBulk + W _steam

− W _cSpray + W _vapSpray − W _vl. (4.4)

The third port value ṁspray is renamed to W _liqSpray since is value changes depending
on the enthalpy of the spray along with ṁcSpray and ṁvapSpray in the following ways. If
the enthalpy of the spray Hspray is smaller or equal to the enthalpy of the saturated fluid
Hf sat then the equations are formulated in the following way:

Hf sat − Hspray
ṁcSpray = ṁspray ∗ ( ), (4.5)
Hvapor − Hf sat
HcSpray = ṁcSpray ∗ Hf sat , (4.6)
ṁvapSpray = 0, (4.7)
HvapSpray = 0, (4.8)
ṁliqSpray = ṁspray , (4.9)
HliqSpray = ṁspray ∗ Hf sat . (4.10)
(4.11)

21
If Hspray is greater or equal to the enthalpy of saturated gas Hgsat then the following
equations are used:

ṁcSpray = 0, (4.12)
HcSpray = 0, (4.13)
ṁvapSpray = ṁspray , (4.14)
HvapSpray = ṁvapSpray ∗ Hspray , (4.15)
ṁliqSpray = 0, (4.16)
HliqSpray = 0. (4.17)
(4.18)

When the value of Hspray is in between Hf sat and Hgsat , the final set of equations is used:

ṁcSpray = 0, (4.19)
HcSpray = 0, (4.20)
ṁvapSpray = ṁspray ∗ Xspray , (4.21)
HvapSpray = ṁvapSpray ∗ Hgsat , (4.22)
ṁliqSpray = ṁspray − ṁvapSpray , (4.23)
HliqSpray = ṁliqSpray ∗ Hf sat . (4.24)
(4.25)

Here Xspray is introduced which is the thermodynamic quality of the steam from the spray
port and is defined as follows:

Hliquid − Hf sat
Xspray = . (4.26)
Hgsat − Hf sat

In addition to mass balance we also have to consider the energy balance for which the
liquid part of the equations are:

Ḣliquid = ṁcond ∗ Hf sat − ṁevap ∗ Hgsat

+ ṁsurge ∗ Hsurge
(4.27)
+ ḢliqSpray + ḢcSpray
+ ṁvl ∗ Hgsat ,
Qliquid = QlHeater + Qvl + Qwl , (4.28)
d
Wliquid = −p ∗ , (4.29)
d Vliquid

22
and for vapor:

Ḣvapor = ṁevap ∗ Hgsat − ṁcond ∗ Hf sat

+ ṁsteam ∗ Hsteam
(4.30)
− ḢcSpray + ḢvapSpray
− ṁvl ∗ Hgsat ,
Qvapor = QvHeater − Qvl + Qwv , (4.31)
d
Wvapor = −p ∗ . (4.32)
d Vvapor

The first equation Ḣ denotes the enthalpy flow rate in which ṁ is for mass flow and H
is enthalpy of the liquid or vapor. The second equation Q is for heat flow. lHeater and
vHeater are the values of a connected heater component that by default is 0 if not defined
for liquid and vapor. Qwl and Qwv are the liquid and vapor heat transfer values with the
wall or ambient temperature if wall heat capacity is unspecified. Qvl is for heat transfer
between liquid and vapor with positive direction into the liquid. The third equation W is
for work done by pressure on the liquid or vapor. In the Modelica language the equations
take the following forms:

Hb_f low_liquid = W _cBulk ∗ h_f sat − W _eBulk ∗ h_gsat

+ W _surge ∗ h_surge
(4.33)
+ H_liqSpray + H_cSpray
+ W _vl ∗ h_gsat,
Qb_f low_liquid = liquidHeater.Q_f low + Q_vl + Q_wl, (4.34)
W b_f low_liquid = −p ∗ der(V _liquid), (4.35)

for vapor they take the form:

Hb_f low_vapor = W _eBulk ∗ h_gsat − W _cBulk ∗ h_f sat

+ W _steam ∗ h_steam
(4.36)
− H_cSpray + H_vapSpray
− W _vl ∗ h_gsat,
Qb_f low_vapor = vaporHeater.Q_f low − Q_vl + Q_wv, (4.37)
W b_f low_vapor = −p ∗ der(V _vapor). (4.38)

Depending on the selected energy dynamics the complete energy balance equation is

23
formed differently. If steady state option is selected the equations will be formed as:

0 = Ḣliquid + Qliquid + Wliquid , (4.39)

0 = Ḣvapor + Qvapor + Wvapor . (4.40)

Otherwise the following equation set is used:

d
= Ḣliquid + Qliquid + Wliquid , (4.41)
d Uliquid
d
= Ḣvapor + Qvapor + Wvapor , (4.42)
d Uvapor

where U designates the internal energy of the liquid or vapor. These are the equations of the
TRANSFORM default pressurizer. They actually have slight issue in that they overestimate
the energy going into the fluid which instead of a pressure decrease while pumping in
cold water result in an increase instead. This problem seems to have been faced by others
using the pressurizer component as well to the point where the decision to model their own
pressurizer was taken [11]. The main issue seemed to be in the assumption that when the
spray reaches the fluid that it has reached saturation. In the thesis model this assumption
was changed to the specific enthalpy of the spray, which is not physically rigorous but a
numerical approximation that ensures model stability. In the future an analytical solution
should be attempted if the component will be used for real experiments.

4.2.1 Pressurizer Parameter Selection

The first thing to consider when selecting parameter values for the pressurizer is the shape
of the pressurizer which can be selected from DrumType. The height and radius of the
component can be set in the submenu. If one has selected the Traditional drum type then
the V_total can be set to

0.5*4/3*Modelica.Constants.pi*(drum2Phase.drumType.r_1)^3
+0.5*4/3*Modelica.Constants.pi*(drum2Phase.drumType.r_3)^3
+drum2Phase.drumType.h_2
*Modelica.Constants.pi*(drum2Phase.drumType.r_2)^2

where drum2Phase can be set to the name of the component. All the closure models can
be set to the Constant form. For BulkEvaporation the time was set to 15 seconds. This
is a time delay for the mass flow of the evaporation. BulkCondensation was set to 15
seconds. Same as before but for condensation. MassTransfer_VL was set to 1e − 3 s mkg2 K .
This parameter describes the amount of mass transported between the vapor-liquid surface.

24
HeatTransfer_VL was set to 100 W. This describes the amount of heat transfered between
the vapor-liquid surface. Since in this model no heat transfer between wall was defined the
last two options were set to 0. In the initialization tab Vfrac_liquid_start can be set to 0.5
meaing that the pressurizer is half full and p_start to the pressure of the core.

4.3 Core and Fuel model

An important part of any reactor model is the core and fuel. This is what defines our power
and heat outputs. Different core arrangements and sizes are possible and some can be
found online. In this thesis, a large PWR core was defined according to the values provided
by MIT [12].

The core model uses a popular and efficient point kinetics based model for neutron
population calculations. The energy balance equation for the core is:

j
∂Qf (ρ − β) X Qext
= + λi Ci + , (4.43)
∂t Λ Qf i=1
Λ

In this equation Qf is the amount of energy generated by fission reactions, Qext is additional
added external energy and j is the amount of precursor groups selected. In our model this
was taken to be 6 groups and no source of external energy was added Qext = 0. In code
the equation looks like:

der(Q_fission) =(rho - Beta)/Lambda*Q_fission

+ sum(lambdas .* Cs[:])
+ Q_external/Lambda;

The values relating to delayed neutrons and precursor groups are taken from TRANSFORM
library 6 group precursor data set. The value of total reactivity rho is given by the following
equation:
ρ = ρinput + ρmod + ρdop + ρf is (4.44)

The reactivity from fuel ρdop and moderator ρmod were considered using Doppler and
moderator feedback coefficients, which can be found in Table 1. Control rod reactivity
ρinput was adjusted using a PID-controller, which was simplified to a positive or negative
insertion of reactivity to the core by using the rho_input parameter. The thermodynamic
effects of the control rod were not considered only it’s impact on the reactivity. ρf is is

25
calculated by using:
j
X −σA (i)NCi (i) V
ρf is = (4.45)
i=1
νΣF

In this equation the sum is over all the delayed precursor groups. σA is the microscopic
cross-section for absorption and ΣF is the macroscopic cross-section for fission reactions.
ν is the amount of neutrons released per fission reaction. NCi is the amount of fission
product atoms. In code the two equations look like:

rho = rho_input + sum(rhos_feedback[:])

+ (if toggle_ReactivityFP then -sum(fissionProducts.rhos_start)
+ sum(fissionProducts.rhos[:]) else 0)
+ sum(fissionProducts.rhos_add[:]);

and

rhos[j] =-sigmasA[j]*mCs[j]/(nu_bar*SigmaF)/V;

In addition to these we have decay heat so our final equation for the total power produced
is:
X k
Qtotal = Qf + Qdecay (i). (4.46)
i=1

Decay heat was considered in 11 groups k = 11.

Fuel pins are modelled as 3 regions: fissile material (fuel), gap, and cladding. Heat transfer
is considered only in radial direction. The number of radial nodes for each section was 3
and axial direction was 4. The fuel pellet was from Uranium dioxide, the gap was taken to
be Helium, and the cladding a Zirconium alloy ZrNb_E125. The inlet temperature was
held constant using a PID-controller that adjusted the flow rate of the water in the cooling
loop. Heat transfer effects between moderator and gap were calculated using laminar and
turbulent Nusselt numbers. Laminar Nusselt number was taken to be 4.36 and turbulent
number was calculated using the Dittus–Boelter equation, these are default values for
Nusselt numbers in TRANSFORM.

26
 Table 1. Core parameters

Parameter Value
No. Fuel Assemblies 193
No. Fuel Rods in Assembly 264
Lattice Configuration 17 x 17
Active Fuel Length 365.76 cm
Assembly Lattice Pitch 21.50364 cm
Pin Lattice Pitch 1.25984 cm
Total Core Power 3411 MW
Core Operating Pressure 156 bar
Core Flow Rate 17083 m/s
Core Inlet Temp 294.22 ◦ C
Doppler Feedback Coef. -2.5e-5 1/K
Moderator Feedback Coef. -20e-5 1/K
Fuel Reference Temp 705.3 ◦ C
Coolant Reference Temp 314.242 ◦ C

4.3.1 Core and Fuel Model Parameter Selection

Before starting to parameterize the core component it is a good idea to set the input
temperature into the core to be constant. In a test scenario with the core separated this
can be easily done but in a loop test case this should be achieved with a PID-controller
that controls the amount of cooling happening in the cold leg of the loop. One has the
option of cooling the loop by using a heat exchanger and controlling the flow speed of the
second loop or by cooling a pipe with heat flow ports and having the PID control the heat
flow. In this model the speed of the simple pump on the secondary side was controlled
using a PID type contoller with an initial output of 3000, k=1, k_s = 1, k_m = 1, Ti =
0.5 sec, Td = 0.1 sec, wp = 1, wd = 0, Ni = 0.9, and Nd = 10. To control the cooling by
using a pipe with heat flow enable and a Heatflow_multi boundary the PID would need
to be configured with different settings and might also require that the inputs, u_s and
u_m, are switched. There are three types of cores in the TRANSFORM library, 1-,2-,
and 3-region cores. The regions refer to the fuel, gap and cladding. So depending on the
resolution required the choice can be made. This model used the 3-region version. The
core component is actually one assembly that is where the "subchannel" in the name comes
from. So in the parameters window nParallel is the total amount of assemblies in the whole
reactor. Most of the parameters within geometry are self-explanatory, it is important to pay
attention to angle again. For better accuracy the HeatTransfer can be set to “Specify Nus

27
2 region laminar and turbulent”. In the Kinetics tab the Data_PG can be set to “Default
delayed-neutron groups from TRACE Manual Table 9-1”. Other values and parameters
can be adjusted based on the reactor one is trying to model but first time it’s recommended
to model using the MIT data [12]. When using this data set the Teffref_fuel to 705.3 C
and Teffref_coolant to 314.242 C. Fuel element initialization will depend on the reactor
one is trying to model and the amount of negative rho_input one is starting with, which
will be explained later. These values can also be found by waiting until the reactor is
at steady state. If one has trouble running the simulation due to large spikes the size of
the pressurizer can be set to very large e.g 100m height and then slowly reduced to more
normal values as better core parameters are found by letting the system stabilize. In coolant
initialization it is a good idea to initialize inlet and outlet at their respective temperatures.
Inlet temperature can be found in specification sheets and outlet temp can be calculated
based on core power and flow rate. It might be helpful to give steady state temperature
values for each node as an array in Ts_start. Steady state initialization is not necessarily a
good thing as this usually means that the solution to the system is known. Usually it is
more important that the system converges to a steady state. This usually can be achieved by
parameter calibration. If one wishes to start from steady state they can set the energy and
mass balance of the coolant to SteadyState. This can be quite difficult to achieve however.
KineticDynamics and FissionProductDynamics can be set to FixedInitial. One of the more
important values to set is at the bottom of the General tab rho_input. Each reactor has
some negative rho_input value that they are operating under coming from the boron in the
moderator or the control rods. Additionally, there is some negative rho from the design of
the reactor that the core model doesn’t take into account. For this it is best to use a PID
controller. The inputs are the nominal and Q_total.y values for the coresubchannel. One
has to carefully tune the controller to achieve reasonable results, if the controller is too
fast or too slow it can make finding the stable value impossible. This requires for some
experimentation on the users’ behalf. It is also a good idea to set a maximum and minimum
output, within reasonable limits, when the stable values have been found. In this model
the controller was set to PI type, k_s and k_m were set to 1/coreSubchannel.Q_nominal,
yMax to 0.012. Under tuning controls k = 5e-4, Ti = 0.1, wp = 1 and Ni = 0.9. Under
initialization one should set the controller to have an initial output otherwise they will see
large jumps and spikes in power in the beginning potentially causing the system to reach
undesirably high pressures and temperatures. In this model the initial output was set to
-0.009. This was determined empirically by looking at the initial rho and Q_total values
and adjusting the input so they are closest to steady state values.

28
4.4 Heat Exchanger Model

Each reactor has some way of dissipating heat from the primary loop. Usually this is
done by way of heat exchangers. In PWR the heat exchanger is also known as the steam
generator since the water is turned into steam in the secondary loop. TRANSFORM offers
multiple different types of heat exchangers to choose from. This model uses a shell and
tube type of heat exchanger. It is basically a number of multiple tubes coming from the
reactor that are directed in to a large vessel with flowing water. The thermodynamic effect
of the wall was also modelled in the heat exchanger. The material of the wall was taken to
be Inconel 690. Thermal transfer coefficients in the primary circuit side were calculated
using single-phase Nusselt numbers and Dittus-Boelter equation. The secondary side
was modelled using two-phase three-region alpha coefficients. This means that thermal
coefficients were calculated separately for single-phase liquid, single-phase vapor, and
two-phase saturated liquid, since the water turns into steam and under goes phase changes
while going through the component. For single-phase liquid and vapor turbulent part
Nusselt numbers were used with Dittus-Boelter equations. For the two-phase saturated
liquid Chen’s correlation was used.

4.4.1 Heat Exchanger Model Parameter Selection

The heat exchanger (HX) component is probably the hardest to integrate into one’s model,
due to it causing issues with all other parts. Depending on the reactor type different
geometries can be selected for the heat exchanger. For a pressurized water reactor, we
used the shell and tube HX. One has to be careful not to select GenericHX since it is a
partial model and realistically shouldn’t be in the options, this is mainly used if there is
a need to build a different kind of heat exchanger using some established base. In the
geometry submenu, we used 20 volume nodes. This is taken from one of the examples
included and is most likely to do with the phase change in the second loop. Having less
nodes and a phase change can lead to issues in the simulation. Most of the values for
the heat exchanger can be taken from the NSSS example reactor which is included in the
TRANSFORM library under Examples/LightWaterReactor_PWR_Westinghouse. A few
other things to point out are heat transfer models and shell initialization. For heat transfer
on the shell side it is a good idea to use two-phase 3-region alphas model, because we
have a phase change in the HX component meaning liquid turns into steam. So we need to
tell the model that it has to be two-phase also the different phases have differing thermal
diffusivity, alpha, so we need to calculate them in 3 regions them being single-phase liquid,
single-phase vapor, and two-phase saturated liquid. For the tube side single-phase 2-region
Nus can be used. If there is a need 5-regions can also be simulated adding turbulent and

29
laminar regions for liquid and vapor, in this model such accuracy wasn’t necessary. Since
we know the primary circuit should be liquid we can use a single-phase model here and the
2 regions we calculate the Nus, Nusselt numbers, for are the laminar and turbulent regions
of the liquid. It is important to use two-phase models if one has both vapor and liquid.
The other initialization tip would be to use enthalpy on the shell side for initialization and
then to initialize the medium at vapor saturated vapor enthalpy or a little bit less than that
if it has trouble initializing at that value. The rest of the heat exchanger works similarly
to regular pipes. If one is having issues with the shell and tube type then it a good idea
to make sure if the pipes on the tube side can actually fit in the shell and that they don’t
cause too much flow constriction. An important factor to consider is when testing the heat
dissipation of primary loop one has to be careful when setting static values as this can lead
to cooling or heating over long periods of time. Thus it is best to use a PID or some sort of
dynamic cooling to adjust the rates.

30
5. Results

The final model can be seen in figure 3 consists of 21 components excluding sources and
on compilation has 15883 equations. The solver used for these simulations was Esdirk45a.

Figure 3. Component diagram of reactor model

On initialization one numerical Jacobian and three nonlinear systems are present. For
faster computations these should be minimized or removed. As of writing this paper the
source of the nonliearities and Jacobian are unknown but most likely the heat exchanger
component is the one that should be looked at. A 1000000 second simulation runs for
71.719 seconds on a mid-tier laptop. Most of the time is spent in the first 5000 seconds
which takes 62.269 seconds, when the system has yet to reach steady state.

31
Figure 4. Results of base model. The line represents the measurement point of steady state
values which can be seen in the box next to it

Figure 5. Zoomed in results of base model

As seen in figure 5 it takes about 4000 sec in order to converge to a steady state. In figure
4, it can be seen that a completely steady state is achieved at around 70000 s. Without the
heat exchanger component, figure 6, it is also possible to start directly from steady state.

32
 Figure 6. Component diagram of reactor model without heat exchanger

We can do a simple calculation for how much we can expect the water to heat up when it
passes through the core using the equation:

Q = ṁ Cp ∆T, (5.1)

which we can convert to get temperature delta between core inlet and outlet

Q
∆T = . (5.2)
ṁ Cp

We can use the values provided in Table 1 and take inlet pressure to be 158 bar from
simulation and calculate Cp = 5.318 kgkJK . We get the value of

∆Tc = 37.546 ◦ C. (5.3)

Comparing this to our simulation where we have

∆Ts ≈ 35 ◦ C, (5.4)

we get a difference of
∆Tc − ∆Ts = 2.546 ◦ C. (5.5)

This difference while it may seem small is in quite large. This might be simply due to the

33
naivety of our theoretical value since the heat capacity Cp of water changes as the water
goes through the core. The core has 4 volume nodes meaning 4 separate calculations with
Cp should be made and could give a more accurate theoretical value.

Figure 7. Pressure across the primary loop

It can also be interesting to look at the pressure drop across the system seen in figure 7.
Since the tee for the pressurizer doesn’t use any flow model and the resistance components
between the core and the pressurizers are there for numerical stability, they are negligible
and because of this the core outlet pressure and the heat exchanger inlet pressures are the
same.

5.1 Transients

5.1.1 Core Inlet Temperature Change

The first test we ran was a 2% change in the inlet temperature of the core. This can happen
if for some reason the water temperature or flow rate in the secondary loop side of heat
exchanger changes. At first we waited until 7000 seconds for the system to reach more or
less steady state. Then we dropped the target value for the PID that controls the temperature
going into the core by 2%. The results can be seen in the charts below

34
 Figure 8. Results of decreased core inlet temperature

First thing we notice in figure 8 is a large power spike caused by a change in the reactivity
ρ. This change in ρ is caused by a drop in water temperature. Since the water temperature
is lower it decreases the energy of the neutrons that are reflected causing their fission
cross-sections to be larger as we saw in Figure 1, resulting in an increase in the neutron
population. Due to the release of more power from more fission reactions the outlet of
the core is actually hotter than it was previously. This causes an increase in temperature
that the PID systems on the pressurizer try to manage. After around 5000 seconds the
core stabilizes at a new higher power output and larger temperature differential. This
demonstrates that the model is able to capture the very important temperature-power
feedbacks.

5.1.2 Loss of Secondary Loop

The second test we performed was a pump failure test in the secondary loop, which
can happen if the secondary side pumps and backup pumps loose power. We wait till
10000 seconds have passed and then quickly ramp down the secondary pump speed. The
simulation fails at 35823 seconds.

35
 Figure 9. Results of secondary loop pump failure

Here in figure 9 we can see multiple things happening we see a sharp increase in core
pressure followed by a slower decrease. We also see a sudden power drop and a drop in
reactivity. Finally we see a sudden rise in temperature as we would expect when there
is no cooling. The sudden loss in power can be explained by the sudden increase in
temperature. Due to the increase in temperature of the moderator neutrons have a harder
time reaching thermal energies, thus resonance escape probability decreases. The fuel also
warms up causing Doppler broadening to increase further increasing resonance capture.
Quite quickly there aren’t enough neutrons being generated each generation and the core
loses it’s criticality and nuclear fission stops. From the chart we can still see that even as
fission has stopped there is still some energy being generated. This is decay heat from the
decay of different nuclear isotopes.

36
 Figure 10. Decay heat of secondary pump failure

From the figure 10 we can see that the decay heat left over is actually quite large. Some
isotopes decay in a matter of seconds while others decay for hundreds of years. We can
also glean more information if we take a look at the pressurizer.

Figure 11. Decay heat of Secondary pump failure

In these graphs 11 we can see why the simulation fails. During the simulation the pressur-
izer tries to keep its designed pressure level but has a hard time doing so because of all the
extra heat. It slowly fills up to try to reduce the pressure by spraying cold water but when

37
there is no more room in the pressurizer and since the system can’t overflow the simulation
fails. The reason for the spike in pressure is due to a sudden spike in temperature in the
cold side of the loop seen in the figure 9 where inlet and outlet temps become basically the
same. The reduction in pressure happens because fission chain reaction stops in the core
after that pressure keeps building due to decay heat.

5.1.3 Leak in Primary Circuit

The third test we performed was a leak test in the primary circuit of the reactor. Leaks can
happen because of the harsh condition in the reactor pipes can become more brittle over
time. A leak of 0.2 kg/s was introduced right before the heat exchanger at 10000 seconds
into the simulation when the system was stable.

Figure 12. Core results for the leak test

From the core values in figure 12 we can see that it looks mostly normal except that the
pressure is going up and down periodically and the simulation suddenly ends at 83500
seconds.

38
 Figure 13. Pressurizer results for the leak test

Here in figure 13 we can see why the simulation ends, the pressurizer runs out of water.
Here we see also the water temperature in the pressurizer doing the same periodic move-
ment while in the rest of the loop it didn’t show up including the vapor of the pressurizer.
Using this lead we can take a look at what the liquid heater is doing.

Figure 14. Pressurizer heater values for the leak test

This graph 14 gives us the reason for why we see this periodic jumping. It comes from the
heater which is controlled by a PID to keep it above a certain threshold pressure level. The

39
leak in the system causes a pressure drop which the heater then tries to compensate for by
vaporizing more water and creating a pressure increase. In this case the system might have
failed earlier in reality because the physical position and size of the heaters isn’t modelled.

5.2 Future works

While this work serves as a first test to develop a full nuclear reactor simulator in a new
modeling language, there are several areas of the modeling that call for improvements in
the future. More detail should be added for pressure losses in the system due to friction
from the piping. The pressurizer component also needs to be verified and validated more
to make sure the numerical set up is as intended and the model represents also reality i.e.
is valid. This work has provided first experience and feeling on how the model works
and which parameters have which importance on the results, there is room for sensitivity
analysis to understand the assumptions made for each component and how it affects the
model as a whole. Whole another work direction is the BWR modeling. Currently there is
no such component which can be used as the BWR core. More research is also required
into the solvers and what short comings they have if the results are to be used to base real
world decisions on. The solvers might have different idiosyncrasies’ which might lead to
having to change the model to fit the solver better. All in all, the model requires integral
verification, validation and uncertainty quantification (VVUQ) before it can be used for
any real nuclear engineering purpose.

40
6. Summary

A working preliminary model of a pressurized water reactor and the aim of the thesis were
achieved. The model was developed using Modelica and Dymola, with the main library
being TRANSFORM. The final reactor consisted of a nuclear core with a 3 region fuel
model, a pressurizer and a heat exchanger. The model was shown to be able to reach steady
state and 3 different test transients scenarios were also shown. In addition parameterization
of different components were also shown and explained.

In conclusion developing a more complex model in Dymola using TRANSFORM seems

promising but will require a lot of work to try and understand the different components on
a more fundamental level so errors could be minimized, since TRANSFORM has minimal
documentation. To develop a boiling water reactor further components will need to be
developed since TRANSFORM natively doesn’t have them. A better understanding of
Modelica is also require especially on the front of solvers and what assumptions and short
comings the make or have.

41
7. Acknowledgments

The author is grateful to his supervisor Marti Jeltsov for suggesting the thesis topic, as
well as for his timely advice and support. The author is also thankful to everyone who has
guided him along the way and given their input and wisdom.

42
References

[1] Homepage of the Modelica Association. Visited on 31/12/2023. URL : https :

//modelica.org/.
[2] Dassault Systems Dymola homepage. Visited on 31/12/2023. URL : https://
www.3ds.com/products/catia/dymola.
[3] Github link to TRANSFORM library. Visited on 31/12/2023. URL: https://
github.com/ORNL-Modelica/TRANSFORM-Library.
[4] Nuclear Physics and Reactor Theory. DOE Fundamentals Handbook Volume 1.
Visited on 20/12/2023. 1993. URL: https://www.standards.doe.gov/
standards- documents/1000/1019- bhdbk- 1993- v1/@@images/
file/.
[5] Nuclear Physics and Reactor Theory. DOE Fundamentals Handbook Volume 2.
Visited on 20/12/2023. 1993. URL: https://www.standards.doe.gov/
standards- documents/1000/1019- bhdbk- 1993- v2/@@images/
file/.
[6] OECD NEA Janis database. Visited on 18/12/2023. URL: https://www.oecd-
nea.org/jcms/pl_39910/janis.
[7] Factor dependance on moderator-to-fuel ratio. Visited on 20/12/2023. 1993. URL:
https : / / www . standards . doe . gov / standards - documents /
1000/1019-bhdbk-1993-v2/@@images/file/.
[8] Belle R. Upadhyaya Thomas W. Kerlin. Dynamics and Control of Nuclear Reac-
tors. Elsevier, 2019. URL: https://www.sciencedirect.com/book/
9780128152614/dynamics-and-control-of-nuclear-reactors.
[9] IAPWS Industrial Formulation 1997 for water. Visited on 26/12/2023. URL: http:
//www.iapws.org/relguide/IF97-Rev.html.
[10] Energy dynamics in Modelica. Visited on 16/12/2023. URL: https://build.
openmodelica . org / Documentation / Modelica . Fluid . Types .
Dynamics.html.
[11] Idaho laboratory new pressurizer components. Visited on 26/12/2023. URL: https:
/ / github . com / idaholab / HYBRID / tree / devel / Models / NHES /
Fluid/Vessels.

43
[12] Core parameters. Visited on 12/12/2023. URL: https://crpg.mit.edu/
sites/default/files/css_injector_images_image/BEAVRS_2.
0.2_spec.pdf.

44
Appendix 1 – Non-Exclusive License for Reproduction and
Publication of a Graduation Thesis1

I Kaspar Selke

1. Grant Tallinn University of Technology free licence (non-exclusive licence) for

my thesis “Simulating Pressurised Water Reactor using Modelica”, supervised by
Marti Jeltsov
1.1. to be reproduced for the purposes of preservation and electronic publication of
the graduation thesis, incl. to be entered in the digital collection of the library
of Tallinn University of Technology until expiry of the term of copyright;
1.2. to be published via the web of Tallinn University of Technology, incl. to
be entered in the digital collection of the library of Tallinn University of
Technology until expiry of the term of copyright.
2. I am aware that the author also retains the rights specified in clause 1 of the non-
exclusive licence.
3. I confirm that granting the non-exclusive licence does not infringe other persons’
intellectual property rights, the rights arising from the Personal Data Protection Act
or rights arising from other legislation.

01.01.2024

1
The non-exclusive licence is not valid during the validity of access restriction indicated in the student’s
application for restriction on access to the graduation thesis that has been signed by the school’s dean,
except in case of the university’s right to reproduce the thesis for preservation purposes only. If a graduation
thesis is based on the joint creative activity of two or more persons and the co-author(s) has/have not granted,
by the set deadline, the student defending his/her graduation thesis consent to reproduce and publish the
graduation thesis in compliance with clauses 1.1 and 1.2 of the non-exclusive licence, the non-exclusive
license shall not be valid for the period.

45
Appendix 2 - Dymola tips

There are a multiple useful features that aren’t enabled in Dymola by default, these can
simplify the process of debugging and building models by a noticeable margin. Within
Dymola there are a few useful settings that help one analyse the model they have built.
In the Simulation tab under Setup -> Translation many useful options can be found. One
of the first to turn on is “List continuous time states selected”. This will allow one to see
what variables the model actually selected for use in the simulation. Ideally the variables
should all be similar for example, pressure and enthalpy for all fluid parts with no pressure
and temperature variables. This can help to identify if the system is actually using the
values one wants it to use and not ignoring them. Another useful setting is “List non-linear
iteration variables”. Since non-linear systems cause our simulations to potentially run a
lot slower we should try to keep these to a minimum. By default, Dymola doesn’t tell us
what variables cause said non-linearities but using this setting it gives us the components
involved under the Translation Logs -> Statistics -> Variables appearing in the nonlinear
systems of equations. Next in the Debug tab it is useful to turn on “Detailed logging of
failed nonlinear solutions”. This setting helps one to troubleshoot nonlinear equations on
which the simulation stops. Often it will give the names of the components related and
while it might not be those components directly causing the issue, it narrows down the
search by a large amount. Another useful setting within the Debug tab is “Provide variable
dependencies and equation incidence”. With this active one can right click variables and
click “Plot dependencies” which will open up a separate window showing what other
variables the selected variable is dependent on. This can help narrow down unreasonable
values the user has selected or help spot cases where the compiler has selected a different
value than the one entered. There are other potentially useful settings as well depending on
the situation, it’s always worth taking a look when dealing with non-linearities or variable
initialization. When starting with a new language or tool it’s best to start simple and look
at documentation. The second best incase of missing documentation is to look at other
examples one might find. The TRANSFORM library doesn’t contain much documentation
by design. The idea of the author is that one can learn the library by exploring the various
examples. While this idea has its’ merits it can cause an extremely steep learning curve and
lengthen the process. Best practice is to start with each component separately and create
test scenarios to see if they behave as expected. A good idea would be to try out different
solvers within the simulation tab and find one that is most suited for what one is trying
to accomplish. If the end goal is not clear, then either one of the solvers can be chosen.
The reason for choosing one and sticking with it is that the different solvers have varying

46
idiosyncrasies and getting used to one can help speed up the learning process. The more
popular one’s for nuclear simulations seem to be Esdirk45, Dopri45, Cerk34 or Cerk45.
This model shown in this thesis is made using Esdirk45.

47