Frank U.

Rückert, Michael Sauer, Tuomo Liimatainen and Dirk Hübner

Digital Twin Development

An Introduction to Simcenter Amesim
Frank U. Rückert
Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany

Michael Sauer
Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Tuomo Liimatainen
Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland

Dirk Hübner
Lightweight Construction, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany

ISBN 978-3-031-25691-2 e-ISBN 978-3-031-25692-9

https://doi.org/10.1007/978-3-031-25692-9

© The Editor(s) (if applicable) and The Author(s), under exclusive license
to Springer Nature Switzerland AG 2023

This work is subject to copyright. All rights are solely and exclusively
licensed by the Publisher, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any
other physical way, and transmission or information storage and retrieval,
electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.

The use of general descriptive names, registered names, trademarks, service

marks, etc. in this publication does not imply, even in the absence of a
specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice
and information in this book are believed to be true and accurate at the date
of publication. Neither the publisher nor the authors or the editors give a
warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The
publisher remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

This Springer imprint is published by the registered company Springer

Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham,
Switzerland
Preface
This book is dedicated to our families who gave us time together for writing
it. We worked on it from winter semester of 2020 until summer of 2022 at
the University of Applied Sciences Saarbrücken in Germany. Additionally, a
seminar on this topic was conducted at the LAB University of Applied
Sciences in Lappeenranta, Finland, during the summer session in order to
get feedback from international English-speaking students. To assess in-
course progress, students were asked to provide feedback, suggest outcome
improvements, and avoid any possible misunderstandings.
The authors have many years of experience in creating sophisticated
simulation models for complex, technical solutions in automotive industry
and power plant research. In order to improve further development in the
field of renewable energies, we added examples of this subject area. For us,
it was important that the students could easily start building their own
digital twins for complex technical systems. It was important that students
explore the various parts of the model or initially develop uncomplicated
systems to examine the behavior in a user-friendly environment thereby
learning a step-by-step approach. The graphical visualization of the
innovation helped to analyze and solve technical problems much faster.
The software tool, Simcenter Amesim, was provided free of charge by
Siemens for study purposes, increasing time efficiency and minimizing
challenges through modeling. Networking and collaboration during the
development process facilitated the exchange of information effortlessly. As
model-based engineering continues to grow, it is important to start
educating young ones early on.
Although our book was written for students of natural sciences,
engineering, or information technologies, the language is deliberately kept
easy to understand for anyone who is interested in the development of
digital twins for physical problems and/or wants to start modeling.
Especially for students, the early possibility of a free, easy-to-understand
simulation environment can enrich physics lessons at school and contribute
significantly to the understanding of technical relationships of the models.
Our book takes into account the feedback given to us by the students
and will help especially novices to find alternative ways to use the
simulation tools. We expect that the book will help to make the subject easy
to understand and that the reader will find it educational and enjoyable
when developing new technical solutions and innovations.
Frank U. Rückert
Saarbrücken, Germany
October 2022
Acknowledgements
It is anticipated that interested readers will have as much fun reading this
book and playing with the models as we had during writing the chapters and
compiling the simulation examples. Our aim is that students will have fun
while modeling easy solutions without struggling because of antiquated and
unnecessary mathematical doctrine and quibbles.
We would like to thank our long-time industrial companions Stephan
Wursthorn, Christian Steinbrecher, Heiko Roth, Gerhard Sünderhauf,
Christoph Magel, Thilo Klingel, Martin Katz, and Sibel Yilmaz from Robert
Bosch GmbH. We also want to thank Rüdiger Thieman and Rolf Hartge for
their support. During many discussions, they taught us a lot, while we spent
hours working together with Simcenter Amesim on technical problems and
innovations, we drank a cup or two of tea or coffee.
We thank our faithful and hardworking assistants and students at the
University of Applied Sciences Saarbrücken, namely Philipp Spindler,
Benjamin Allweyer, Sebastian Grün, Emile Degro, Xenia Hülsmann, Tarek
Khiar, Daniel Lehser-Pfeffermann, Alexander Hamman, Johannes König,
Yannick Planta, Badre Ait Amghar, Tim Dennemärker, Barbara Kaiser, Nils
Kleiber, Jan Molter, Christian Boiger and Tim Breuer, who created many
simulation models and examples for this work. Tim Breuer passed away
during the Corona pandemic, and we will always remember him and his
service fondly. I want to thank the students who participated in our lectures
about Digital Twins during Summer School 2021 and 2022 at LAB
University of Applied Sciences. We would like to thank the staff of the
International Office in Saarbrücken and Lappeenranta, namely Doris
Kollmann, for the financial support within the Erasmus+ program and the
European Union, we hope this will be possible again in the near future. We
want to thank Michael Sandoval for reading the manuscript. His positive
feedback was an enrichment. We would also like to thank Claudio Santarelli
and Helge Tielbörger from Siemens AG for their kind support, help, and
good advice.
Last but not least, a big thank you to our friends at Lake Saimaa in
Lappeenranta, Finland, who were such good hosts during wartime. Thanks
to Tuomo for giving me the chance for holding my lecture at LAB. It means
a lot to me that you gave my mother a moose antler. Finally, we would like
to thank Michael Sauer for repairing the boat.
Contents
1 Introduction
1.​1 Motivation from History
1.​2 What Is a Digital Twin?​
1.​3 How to Create a Twin?​
1.​4 Adding Physics
1.​5 Analysis of Digital Twins
References
2 Mathematics, Signals and Control Library
2.​1 The First Steps
2.​2 How to Build a Simple Calculator
Problems
References
3 The Mechanical Twin
3.​1 What Is Mechanics?​
3.​2 The Model of a Bouncing Ball
3.​2.​1 Simulation Model
3.​2.​2 Submodels and Parameters
3.​2.​3 Results and Analysis
Problems
3.​3 The Mechanical Rocker
3.​3.​1 Simulation Model
3.​3.​2 Submodels and Parameters
3.​3.​3 Results and Analysis
Problems
3.​4 How a Row on a Pulley Works
 3.​4.​1 Simulation Model
3.​4.​2 Submodels and Parameters
3.​4.​3 Results and Analysis
Problems
3.​5 Damper of a Driving Car
3.​5.​1 Simulation Model
3.​5.​2 Submodels and Parameters
3.​5.​3 Results and Analysis
Problems
3.​6 How to Use 3D Geometries
References
4 The Thermal Twin
4.​1 Examination of Heat Transfer
4.​2 Heat Transfer in Electric Generators
4.​2.​1 Simulation Model
4.​2.​2 Submodels and Parameters
4.​2.​3 Results and Analysis
Problems
4.​3 Site Selection for a Solar Collector
4.​3.​1 Simulation Model
4.​3.​2 Submodels and Parameters
4.​3.​3 Results and Analysis
Problems
Reference
5 The Hydraulic Twin
5.​1 What Is Hydraulics?​
 5.​2 Two Fuel Oil Tanks and a Pump
5.​2.​1 Simulation Model
5.​2.​2 Submodels and Parameters
5.​2.​3 Results and Analysis
Problems
5.​3 How Does a Hydraulic Jack Work?​
5.​3.​1 Simulation Model
5.​3.​2 Submodels and Parameters
5.​3.​3 Results and Analysis
Problems
References
6 The Pneumatic Twin
6.​1 Pneumatic, Fluid Flow, and Turbulence
6.​2 Safety Valve for a Biogas Tank
6.​2.​1 Simulation Model
6.​2.​2 Submodels and Parameters
6.​2.​3 Results and Analysis
Problems
6.​3 Ventilation System of a Building
6.​3.​1 Simulation Model
6.​3.​2 Submodels and Parameters
6.​3.​3 Results and Analysis
Problems
Reference
7 The Electric Twin
7.​1 Permanent Electric Motor with Load
 7.​1.​1 Simulation Model
7.​1.​2 Submodels and Parameters
7.​1.​3 Results and Analysis
Problems
7.​2 How an Asynchronous Motor Works?​
7.​2.​1 Simulation Model
7.​2.​2 Submodels and Parameters
7.​2.​3 Results and Analysis
Problems
7.​3 Electric Generator with Resistor and Heat Transfer
7.​3.​1 Simulation Model
7.​3.​2 Submodels and Parameters
7.​3.​3 Results and Analysis
Problems
Reference
8 Analysis of Complex Technical Systems
8.​1 How Does the Liquid Piston Compressor Work?​
8.​2 Design and Function of a Liquid Piston Compressor
8.​3 Liquid Piston Compressor with One Cylinder for One Stroke
8.​3.​1 Simulation Model
8.​3.​2 Submodels and Parameters
8.​3.​3 Results and Analysis
8.​4 Liquid Piston Compressor with One Cylinder and Multiple
Strokes
8.​4.​1 Simulation Model
8.​4.​2 Submodels and Parameters
 8.​4.​3 Results and Analysis
8.​5 Liquid Piston Compressor with Two Cylinder and Multiple
Strokes
8.​5.​1 Simulation Model
8.​5.​2 Submodels and Parameters
8.​5.​3 Results and Analysis
Problems
References
9 Digital Twins and Artificial Intelligence
9.​1 Neural Networks in Nature
9.​2 Neural Networks and Digital Twins
9.​3 The Artificial Frog Model
9.​3.​1 Submodels and Parameters
9.​3.​2 Optiflow Neural Network (ONN)
9.​3.​3 Results and Analysis
Problems
References
10 Conclusions
10.​1 Overview of This Textbook
10.​2 What You Can Take Away
10.​3 Teaching Methods and Gamification
10.​4 Our Outlook for the Future
10.5 Important: Disclaimer for Our Work
References

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_1

1. Introduction
Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

1.1 Motivation from History

What could early humans have done to plan ahead for their problem-solving
to save resources and energy? One first step could be to draw a situation in
order to become familiar with different aspects of the problem. Figure 1.1
shows such a situation. It’s a picture painted after early petroglyphs located
at Bryce Canyon National Park in Utah (USA). It shows a hunter on his
horse and a deer [1]. Similar illustrations can be found all over the world. If
you look closely at the scene, you can almost imagine the arrow moving.
The purpose of problem visualization was used to describe and solve a
specific situation. Additionally, the picture could be used to teach younger
hunters with less experience. The petroglyph also depicts technical
equipment. Drawing on a wall can be considered as an early kind of
simulation or analysis of a specific situation. Today, the first thing students
are advised to do is to draw a sketch of a technical device. Visualizing a
problem at least helps to analyze and solve it. We believe that the
petroglyph scene was for teaching or self-reflection. Knowing what to do is
as important today as it was in the past.

Fig. 1.1 Sketch after a scene from petroglyphs in Bryce Canyon National Park, USA [1]
What we don’t see in Fig. 1.1 is that there are connections between the
different objects. They are not painted, but in fact, there is a story and a
connection between the different symbols and people. Two larger persons
with antlers on their heads could also be identified as supervisors evaluating
the situation and the quality of the hunting scene. We want to demonstrate
the idea behind simulation models for problem-solving processes and how
we can create artificial twins for any kind of problem. The core problem is
that today’s technical solutions are becoming more and more complicated.
More details are included, leading to unexpected behavior. Simulation
models can help understand your work, but—beware—they can also make
it worse because they make the system more complex. Too much ingenuity
can very quickly lead to failure. An engineer has to understand the aim and
purpose of his simulation to answer the question why. After that, answer to
question how is often more trivial. At the end of this book, in the
Conclusions chapter, we will discuss further methods for teaching and
introduce an additional new idea on how to teach complex technical issues
very quickly by introducing a role play method.
During creating the so-called digital twins, we wanted to introduce the
simulation tool Simcenter Amesim, which can be used to simulate the
behavior of systems. The software we use is from Siemens PLM Software
[2]. A free-of-charge demo version can be installed without much effort on
Microsoft Windows or Linux systems. Working on Apple operation systems,
one can use the program via a virtual private network (VPN) connection. Of
course, this is less convenient. The software tool contains a ready-made
library, including smaller models to simulate pneumatic, hydraulic, thermal,
electrical, and mechanical behavior. It additionally includes a theoretical
library for signals, mathematics, and artificial intelligence, e.g. artificial
neural networks (ANN) or reduced order modeling (ROM).
We will explain the basic functions and how to start modeling with this
software tool. It is considered as a multi-domain software, meaning that all
the different physical domains can be linked with each other or with the
signal library. The software offers the possibility of simulating systems
without the need for computer-aided design (CAD) geometry, but 3D
geometries can be added and controlled by the models if this is needed.
These models can also be printed with a 3D printer. In former times, small
models made of wood or clay were built to study the behavior of systems.
Today, our students are fascinated by 3D printing, which plays a similar
role. The driver is similar to that of the petroglyphs, and is why we use it for
a clearer understanding.
The usable symbols for the different libraries in Simcenter Amesim are
partly standard symbols defined and therefore easy to recognize. The
software allows it to interact with other programs such as Excel or
MATLAB/Simulink. One big advantage is, if compared, for example, with
Simulink. Many industrial models are already finished and can be used as
reference cases from the onset.
 In order to bring the basics closer in the study, it was necessary to plan a
lecture with the contents, e.g. the artificial model construction and the
practical application of the used software. As previously described, the
fascination lies not only in the manual creation of the simulation model but
also in the possibilities that arise when one prototype has to be constructed.
With a digital twin, technical, physical, or biological components can be
tested long before they are used. The digital twin must reproduce certain
properties of its realistic model to allow its functions and behavior to be
scrutinized.
The earlier in the development process the engineer has the opportunity
to make quantitative statements about the function of a component in the
system, the more effective a later design or development process can be.
The digital twin is the image of its real counterpart and represents the first
step in the development process. In our daily work experience, there is a
difference between creating and using a numerical model. Even though the
creation of the twin requires an expert who must know the process to be
mapped down to the smallest detail, the user of the digital twin does not
need to understand everything in detail in order to work with the twin. We
use the example of the light switch. You do not have to understand how it
works to use it. The gap between creation and use should be bridged by this
book. Possibly this was also one idea behind the petroglyph scene in
Fig. 1.1.
The relevance of this topic in today’s world is obvious. Also in the most
diverse areas of product development. Today, the topic of sustainability also
encompasses the behavior of a product over its lifetime cycle. This means
that a simulation model should not only make a statement about how a
technical product works and which functionality in the various phases of its
life but also how large an effect really is and what its consequences are is
only known to a very limited extent. Today, for example, a very small
amount of a powder containing anthrax virus can start a war. Quantities of
harmless gas that are only in the ppm range (ppm = parts per million) can
bring an entire industry to a standstill. On the other hand, sometimes small
amounts of a virus are enough to have a very large effect. In such a case, the
complexity of the overall system is crucial. When we build systems more
complex, they get more susceptible to small quantities.
1.2 What Is a Digital Twin?
When creating the model or digital twin from reality, it should be noted that
the original cannot be reproduced in its entirety. Rather, certain properties
are to be reproduced in isolation, i.e. in digital form as in Fig. 1.2. What is
meant by this? We can explain it by the example of a human body or better
with a prometheus, as described in [3].
In the case of this human being, like in Fig. 1.2, it can be possible to
digitally reproduce only the blood vessels and thus create a kind of
hydraulic twin. Such a twin would not provide any information about the
bone structure. To represent the statics of the bones of the human being, one
would have to create a mechanical twin. The original can therefore have
different digital twin brothers, which only partially reflect the physical or
geometric properties of their real counterpart.

Fig. 1.2 Digital twin is the counterpart to the real twin, but with limited capabilities

A thermal twin could, for example, allow conclusions to be drawn about

whether the subject has a fever or how quickly he cools down in winter, but
not whether there is a fracture in the thigh. The pneumatic twin could, for
example, provide conclusions about lung function or movements of other
compressible fluids inside the body. For the human nervous system, there is
a so-called signal database for neural networks. To control the digital twin,
the nervous system is needed, just like in real life. This shows how
important the idea of artificial intelligence (AI) is. To explain in detail how
to work with it in Amesim, we will first start in the following chapter by
rebuilding a simple calculator with the signal database. At the end, we
explain what a neural network is and how it can be used to control our twin.
As can be seen in Fig. 1.3 the digital twin is a virtual replica consisting
of data and algorithms of a real object or product. This replica can be a
model, a simulation, or even an algorithm that emulates the behavior of real
objects. this offers the advantages of testing and simulating systems and
products in advance. Parameters of the virtual image are set and tested, the
results can be output and parameters can be adjusted quickly. The goal is to
simplify planning and thus save time and costs.
Fig. 1.3 Different use cases can lead to different digital twins
Application types of a digital twin can be distinguished depending on
the product life phase. In the planning phase, the digital product twin is the
preliminary product a computer-aided design (CAD) or 3D model. The
model helps to examine the behavior of the product in different phases to
make forecasts. A virtual model can be quickly adapted and also respond to
customer requests. The digital production twin is used to forecast efficiency
and quality in manufacturing. Furthermore, the virtual variant can be used
to test deliveries and the availability of materials in advance.
All digital twins can be divided into several archetypes. Van der Valk et
al. divided them into five different categories or so-called archetypes to
compare the degree of evolution [4]. In Table 1.1, all archetypes are
presented.
Table 1.1 Definitions of the archetypes of digital twins [4]

Archetype Description
Basic Digital This digital twin provides an human-machine interface (HMI). It therefore
Twin extends the digital twin by a single feature and is therefore considered
fundamental
Enriched Based on basic archetype, this digital twin enriches its database by preprocessed
Digital Twin data from supplementary systems. Many objects in this cluster describe the
possibility of semi-manual data collection
Autonomous This twin is an advancement from enriched digital twin. It offers autonomous
Control Twin control, but at the same time, it contains an HMI for the option to intervene. As
direct communication with another virtual or physical machine, this archetype
needs at least interoperability via a translator interface
Enhanced The digital twin offers autonomous control over a physical asset while integrating
Autonomous external, downstream data processing systems. The interoperability needs to be
Control Twin secured at least over a translator as well. On the downside, this archetype does not
offer an HMI
Exhaustive A digital twin with exhaustive data acquisition options, data processing, and
Twin control over a physical asset. This archetype provides the user with all options.
The twin is able to work and control autonomously. At each point, humans have
the ability to intervene or to enrich the database and, hence, provide a semi-
manual data linkage. This archetype demands a fully interoperable data linkage to
downstream systems as well as to the physical asset itself

1.3 How to Create a Twin?

As described above, we provide the reader with the opportunity for an easy
introduction to the topic. As shown in Fig. 1.4, various technical systems
are, e.g. simple fluid energy machines, which can initially be understood as
a system model. Fig. 1.4 illustrates that components from different
disciplines can be combined and thus built up into complex, cyber-physical
systems. Thus, the technical system can then include thermal, pneumatic,
mechanical as well as hydraulic or electric components. Once the system
has been set up, it is possible to study the behavior of, for example, a fluid
energy machine over a period of time. The goal is to be able to perform a
virtual investigation in an isolated space with it.
Fig. 1.4 Example of combined models: propeller, mechanical friction, solar collector, and hydraulic
pump
As already described, no programming skills are required for the
creation of the twins, and we will not dwell on mathematical subtleties.
Mathematics should be subordinated to the development process as an
auxiliary science and not pursued as an end in itself. We want to get results
as quickly as possible. The creation of the physical model is facilitated in
the platform by numerous well-documented examples. A very detailed
illustrated documentation of the theoretical basics is also available. The
company Siemens decided to make its development tool Simcenter Amesim
available free of charge. It gives the possibility to create digital twins as
preliminary stage of the design of a new product with little effort. The
program can be used by university staff and students free of charge.
Integration into the teaching process and promotion of young academics is
thus facilitated. A license must be purchased as soon as Amesim is used
commercially.

Important: You can test the program before you buy it, find a free
student version of Simcenter Amesim at the following link:

https://​www.​plm.​automation.​siemens.​com/​plmapp/​education/​simcenter/​
en_​us/​free-software/​student
 Paths for software can change, and books are usually more persistent
than websites on the Internet. Therefore, the specified path may no longer
be correct. In this case, try to find the described software using a common
search engine. The free-of-charge version from Siemens has some
restrictions such as limitation of the library. With this student version, the
tasks were designed for a lecture plan of this textbook. Only the available
libraries are explained in more detail below and all examples could be done
with the free version.

Fig. 1.5 Graphical user interface of Simcenter Amesim offers a clear overview

The user interface of Simcenter Amesim illustrated in Fig. 1.5 is mostly

operated graphically with the mouse. Here, the graphical program interface
is illustrated with the digital twin of a bouncing ball. The program offers a
tidy interface, which is initially divided into a workbench, as well as
different library windows. On the workbench, different items can be
dragged from the libraries with the mouse and connected with each other in
the so-called sketch mode. The time period over which the digital twin is to
be observed in a virtual laboratory is then defined.
1.4 Adding Physics
All subsequent chapters of this book are structured in the same way for
better clarity and how to work with it. The chapters are always divided, first
into the section Simulation model where the technical system has to be
sketched by the students with the mouse on the workbench. This first step
should be followed by a section on the text page Submodels and
Parameters where the exact physical quantities for the simulation model
must be specified. Here, you need to go into detail. Finally, there is always
a section presenting the simulation results and a reference to further work
proposals, where modeling, simulation, monitoring and mirroring are
essential [5]. To create the model or simulation network, Simcenter Amesim
provides an integrated simulation platform for cyber-physical problems and
multidisciplinary system simulation at different stages of innovation and
life cycle and could be divided into autonomous, federated, and digital twin
maturity models.

Important: Please do not focus too much on detailed physical quantities

and variables at the beginning. The overview of the problem is more
important. Parameters and boundary conditions can be changed later.

We have several categories of libraries available to share below.

Modeling with Simcenter Amesim takes place in four steps. First, a closed
model must be created using the components from the existing libraries. For
this purpose, it is also important to define used fluids or media. In the
Submodel mode, the physical model is selected, which is assigned to each
component. The third step is processed in the Parameter mode as given in
Table 1.2.
Between the step from Submodel mode to Parameter mode, a
compilation of the model takes place. The individual mathematical models
become an overall model description. In Parameter mode, the required
parameters for the physics of the components are set. Here, you have to take
care about the quantity of the models. For the simulation, appropriate
settings can be made such as the time, resolution, or whether the
calculations should be dynamic or stable behavior. After that, the simulation
run can be started in the Simulation mode and the results can be analyzed.
Table 1.2 Definitions of different modes in Simcenter Amesim

Mode Description
Sketch mode Drawing the model on the worksheet
Submodel mode Selection of the different submodels from the libraries
Parameter mode Detailed definition of the physical parameters
Simulation mode Run and review the simulation and graph the results
To add physics for the technical problem we have to choose the right
physical library. Each library has its own color. The signal library is usually
illustrated in red color. We want to take the chance to introduce the
functions of Amesim in detail.
Signals and Control library The signal library contains different
mathematical functions and is used to control the other libraries. Basic
mathematical functions are available. We will explain in the next chapter
how to use them. This library contains artificial intelligence models.
Mechanical library This category contains linear and rotary motion
elements. It is often used to complement other library categories to
simulate mechanical ratios. Components such as masses, springs,
dampers, friction elements, and also transmission ratios can be found in
this section.
Hydraulic library This section contains general hydraulic components
such as pumps, pipes, and valves for the simulation of the dynamic
behavior of fluids.
Pneumatic library It contains components that are used to model
compressed air-based systems. Typical components include pressure and
flow sources, valves, or gas properties.
Thermal library With the thermal library, heat states and interactions
can be simulated. The thermal components such as convection,
conduction, or even radiation elements enable calculations of heat flow
and losses.
Electric library In this library are electrical components with which
systems can be reproduced, especially for automotive electronics.
Components are resistors, inductors, capacitors, batteries, and also
already prefabricated motors.
The applications of the presented program are described in more detail
in further sections. The color of the items can also be changed by right-
clicking it with the mouse. But at the beginning, we should leave it in the
default color for better understanding.

1.5 Analysis of Digital Twins

Various classical mathematical methods can be used for the analysis of
digital twins. A 3D geometry helps to understand the large amount of data
through visualization. An example of such a digital twin coupled with a 3D
geometry is illustrated in Fig. 1.6. The figure shows the model of a
transporter chassis driving over an obstacle. For this purpose, these
hydraulic and mechanical models must be further developed and certain
skills must be trained by algorithms. The algorithms can be used to control
the 3D models in a virtual space, e.g. the transporter has a drive unit and a
hydraulic damping system. It has the task of driving over a stone sidewalk
which produces a lot of data. It should be mentioned that the amount of data
can be high, but not so high that we should call it Big Data [6]. The figure
shows how the chassis drives over the stone, and its functionality can be
tested under real conditions.
Fig. 1.6 3D geometry and damping of a transporter chassis can be tested under realistic conditions
Due to its hydraulic suspension and the pneumatic model of the tires,
the transporter succeeds in doing so and is able to drive independently from
one side of the slab to the other. The physics of the hydraulic damping
system is realistically reproduced according to physical principles. Using
the data for training certain skills such as structural thinking is covered by
the terms machine learning or deep learning. These are two sub-fields of
artificial intelligence (AI). This would allow the chassis model to learn how
best to drive over the obstacle. While machine learning is the generic term
for allowing an algorithm to learn further by feeding it data, deep learning
is a method of machine learning that builds on artificial neural networks.
We will describe this in detail in the last chapter of this book. Machine
learning, by feeding structured data, can create an algorithm that can
analyze new information based on the training. Deep learning can handle
large unstructured data but requires powerful computers to do so. Artificial
intelligence technologies offer a wide range from weak AIs that are used to
solve limited tasks to strong AIs that emulate human intelligence [7]. Their
use is intended to be a support for humans in the industrial, or also in the
domestic area, to deal with information overload and to work
autonomously. In the production process, for example, there is potential for
this in research and development, in-process optimization, or also for
maintenance systems for machine maintenance [7].
To create an accurate virtual image and thus minimize subsequent
production errors, it is advantageous to enter as much data as possible.
Digital twins could assist to generate this data. To deal with a large amount
of data, data processing techniques for Big Data will become necessary [6].
It is used to present the information so that it is available quickly and can be
used effectively by companies. It must be able to quickly capture and
collect the data before it can be processed further.

References
1. Hiser, D.: Records of the Environmental Protection Agency, The Environmental Protection
Agency’s Program to Photographically Document Subjects of Environmental Concern, 1972–
1977. https://​catalog.​archives.​gov/​id/​545671
2.
Siemens AG: Simcenter Amesim. https://​www.​plm.​automation.​siemens.​com/​
3.
Shelley Wollstonecraft, M.: Frankenstein, or, The Modern Prometheus. Oxford University Press,
New York (1818)
4.
Hendrik van der Valk, H., Hasse, H., Möller, F., Otto, B.: Archetypes of digital twins. Bus. Inf.
Syst. Eng. 64(3), 375–391 (2022)
5.
Kim, Y.-W.: Digital Twin maturity model, WEB 3D 2020 industrial use cases workshop on digital
twin visualization. https://​doi.​org/​10.​13140/​RG.​2.​2.​28750.​48967 (2020)
6.
Weisberg Business Consulting GmbH: Einfach erklärt: Was ist Big Data? https://​weissenberg-
solutions.​de/​einfach-erklaert-was-ist-big-data/​
7.
Smith, S.E.: Neuromuscular junction. In: Zaimis, E. (ed.) Handbook of Experimental
Pharmacology, vol. 42, p. 593. Springer, Heidelberg (1976)

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_2

2. Mathematics, Signals and Control

Library
Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

2.1 The First Steps

Before we get into the interesting, physical issues, let’s briefly revisit the
topic of signals and mathematics [1]. This rather dry topic is necessary
because we will come across the use of so-called signals again and again in
the further course of modeling. To show how Simcenter Amesim works, we
will first create a simple calculator. After starting Amesim, the first thing
that opens is the program’s interface. The first step to create a model is to
drag single items from the model library to the drawing layer of the
program shown in Fig. 2.1.
In this book, all examples and exercises were created under the
operating system Microsoft Windows. But there is also the possibility to run
Amesim under the operating system Linux. This can be advantageous if you
want to couple the program with other simulation tools or if you want to
automate the models. Here, however, we want to operate Amesim
exclusively via the graphical user interface (GUI). We start with the first
example in the following chapter, to perform a simple calculation like:
(2.1)
We want to build up the model for this mathematical calculation by trying
to recreate a calculator with the signal library (red library).
Fig. 2.1 Graphical user interface of Simcenter Amesim after startup—no model is added yet
Fig. 2.2 Example: Creating a simple calculator; Step 1: Creating a constant

2.2 How to Build a Simple Calculator

We start with the model for a simple calculator. To do this, we use the
mouse to drag the item for a constant from the Signal library onto the
drawing area. Make sure that you are in Sketch mode. Otherwise, you will
not be able to add anything to the drawing area shown in Fig. 2.2.
We can activate this item for the constant with the mouse and then, as is
common in Windows, copy it with the key combination Ctrl + C, and paste
it again with the key combination Ctrl + P. With the key combination Ctrl
+ R, we can rotate the items. Further shortcuts will be given by a menu if
you press the right mouse button. The equation 1 + 2 = 3 is to be solved
with the calculator; therefore, one must drag a sum item with the mouse
onto the drawing area from the library shown in Fig. 2.3.
After the two constants and the sum item have been copied or dragged
on the workbench, they must be connected with the mouse as shown in
Fig. 2.4.
The items must be connected to each other as shown in Fig. 2.5. To do
this, you have to move the mouse pointer to the so-called ports of the items
until they are displayed in green. Now, you can connect the ports of the
items with each other.

Fig. 2.3 Example: Creating a simple calculator; Step 2: Copying constants and adding a sum
function
Fig. 2.4 Example: Creating a simple calculator; Step 3: Connection of the two constants and the
sum function
Fig. 2.5 Connecting an item for constant value with the item for summation
When all ports of a model are properly connected, the color of the items
changes. The model is then no longer highlighted in dark color and can be
used. If the connecting does not work, there is usually a thinking error in
the design. This can be checked if you compare the units of the two sides of
the connections.

Important: Always make sure that the elements are connected correctly
and highlighting has been changed. It is also possible that the elements
cannot be connected if a physical model is illogical, e.g. if the units do
not fit. In this case, you should question if your model is logical.

In general, you can connect the elements from the signal library with all
other libraries. A sink from the signal library, as shown in Fig. 2.6, must
therefore always be compatible with the other libraries and can then be used
accordingly if you cannot find a suitable connection or continuation of the
model for a particular simulation model in a physical library.

Fig. 2.6 Example: Creating a simple calculator; Step 4: Closing the model with a sink item
 Once the model is complete for addition, no item will be darkened. All
items should then be displayed in normal color. Now, you can switch from
Sketch mode to Submodel mode as shown in Fig. 2.7.
Switching to the Submodel mode is possible only if the model has been
drawn correctly before. In order to enter individual values in the graphical
model, it is then necessary to switch further to the parameter mode as
shown in Fig. 2.8.
In the mode parameter it is now possible to enter the values to solve our
equation . You can see exactly where to enter the values in
Figs. 2.9 and 2.10. Additionally, in Figs. 2.11 and 2.12, you can see how to
switch to simulation mode and then read the result of the calculation.

Fig. 2.7 Example: Creating a simple calculator; Step 5: Change from Sketch mode to Submodel
mode
Fig. 2.8 Example: Creating a simple calculator; Step 6: Change from Submodel mode to Parameter
mode
Fig. 2.9 Example: Creating a simple calculator; Step 7: Entering a value for the first constant
Fig. 2.10 Example: Creating a simple calculator; Step 8: Entering a value for the second constant
Fig. 2.11 Example: Creating a simple calculator; Step 9: Switching to Simulation mode and starting
the simulation
Fig. 2.12 Example: Creating a simple calculator; Step 10: The simulation is finished and the result
can be read off

Problems
For better understanding, you can change the signal model now and try out
any other calculation [2].

2.1 Sum up three or four different constants to one result.

2.2 Use another operator like, e.g. multiplication or division.

2.3 Copy your model with Ctrl+C, paste it somewhere on the same
workspace, and use two different models in one workspace.

2.4 Plot a diagram of your results. Try to plot two different results from
different models in one diagram.
References
1. Günther, M., Velten, K.: Mathematische Modellbildung und Simulation, Wiley-VCH (2014)
2.
Bossel, H.: Modellbildung und Simulation. Springer Vieweg, Wiesbaden (1992). https://​link.​
springer.​com/​book/​10.​1007/​978-3-322-83658-8

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_3

3. The Mechanical Twin

Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

3.1 What Is Mechanics?

The word mechanics is derived from the ancient Greek word mechané. In
physics, mechanics is usually understood as the classical mechanics of
moving bodies and particles. It deals with the motion of bodies and the
forces acting on them. It is needed for the planning of machines or
buildings. In engineering, this includes static and dynamic mechanics. Let
us solve a easy physical problem first. We want to examine a mechanical
system and have chosen the example of a bouncing ball. The necessary
mechanical library can be found on the right side of the interface in the
library tree. The colour of the mechanical library is green.

3.2 The Model of a Bouncing Ball

The idea of what exactly happens when a ball falls to the ground and
bounces up again is relatively easy to understand and will not be explained
again here. We want to start right away within the symbolism and workflow
of Amesim. Therefore we should create the bouncing ball model directly in
the Sketch mode. To do this, we first open a so-called simulation model for
the mechanical process and create our twin as shown in Fig. 3.1. The sketch
already contains all the essential functionalities. When creating the
simulation model, however, please do not yet worry too much about how
long, large or heavy a body is and exactly what forces are acting. First,
draw the model as you imagine it. The details will then be clarified
afterwards. After that all other quantities, such as the weight of the ball or
the acting forces, are entered in section Submodels and Parameters.

Fig. 3.1 Mechanical simulation model for a falling ball that is bouncing on the ground
 Important: Focus on a correct sketch of the model in the beginning and
try if it can be compiled and the simulation runs. For this, estimate the
physical parameters and boundary conditions first. After that step you
can refer the correct quantities and add them. This will help you to keep
track.

3.2.1 Simulation Model

So for, it doesn’t really matter how heavy or how big the ball is or from
which height it hits the ground. Let’s start with the sketch of the functional
model. We represent our sphere with a so-called mass symbol [MAS001].
We drag it with the mouse onto the empty workspace in Sketch mode.
Now it gets a bit more complicated. We will need a symbol for the
height from which the ball falls to the ground, or the better say the gap
between the ball and the ground. The impact of the elastic ball should be
damped when it hits the ground. To represent these effect with our twin, we
choose an elastic contact symbol from the mechanical library. In addition to
the mass item, that represents the ball, we drag a symbol for the elastic
contact [LSTP00A] and the ground [V001] onto the drawing area. You can
see how the finished simulation model should look like in Fig. 3.1.
Please also make sure that all symbols are properly connected and no
longer have a dark background. Only then does the model work. When all
the symbols are connected, your first bouncing ball twin is ready. Now
comes the final touch and we need to look a little closer. Switch to the
parameter mode and enter the parameters for our twin.
Fig. 3.2 Press button to set run parameters—specify simulation times and time step size

3.2.2 Submodels and Parameters

Once the graphical creation of the simulation model is complete, the
physical quantities for the different Submodels can be selected on the
Submodel tab. At first, not much changes in the presentation. The current
parameters can be specified under the Parameters tab.
Important is, that the parameters are constant quantities, which do not
change during calculation, but have a significant influence on the result.
The user must know these parameters himself or observe them in reality
and specify them to the model. One determines them for example by
measuring or weighing the bodies.
For the mass representing our ball, the weight (here: 1 [kg]) and the
angle with which the ball is dropped down must be entered. In our example
the ball should fall vertically downwards with an degree of 90 . For the gap
between the ball and the ground, the height of 1 [m] as well as the contact
stiffness of the ball at impact (here: it is best to enter a very large value) as
well as the damping at impact (here: 150 [N/(m/s)]) must be entered. The
ball will not be able to penetrate the ground with this damping model, but
will bounce off the ground. In the sections Submodels and parameters we
have always summarized all parameters in a table (see: Table 3.1). This is
the input to Amesim.
Table 3.1 Parameters for the bouncing ball simulation model
Item Parameter
MAS001 Mass = 1 [kg]
Inclination = 90 [degree]
LSTP00A Gap or clearance with both displacements zero = 1000 [mm]
Contact stiffness = 1e+06 [N/m]
Contact damping = 150 [N/(m/s)]
Penetration for full damping = 0.001 [mm]
V001 Linear displacement = 0 [m]
In Table 3.1 the name of the model or submodel is also given under the
Item column. Under this name, you can also search for the corresponding
item using the search function in the parameter window. If all parameters
are correct, the simulation can be started as shown in Fig. 3.2. After clicking
with the mouse on the button to set the Run parameters, you first have to
enter how long the simulation should run. In our case you start at 0 [s] and
calculate up to the simulation time 4 [s]. Additionally, a reasonable step size
is selected (here: 0.01 [s]). This specifies when results are to be written for
each time step. The simulation times are given in Table 3.2.
Table 3.2 Simulation time for the bouncing ball simulation model

Simulation settings
Start time = 0 [s]
Final time = 4 [s]
Print interval = 0.01 [s]

When the progress bar has risen to 100% as shown in Fig. 3.3, the
simulation was completed successfully. Now the results can be examined. If
a simulation does not run successfully, or if it takes a very long time to
generate a result you have to change the simulation parameters. There is the
possibility to choose the simulation time (Final time) a bit lower and you
could also sets the Print interval to a smaller time step size.
Fig. 3.3 Successfully completed the simulation for the digital twin of a falling ball
This allows you to check which possible error could be present in the
simulation model. Mostly the error is caused by physical values that do not
make sense, e.g. the mass of the ball is much too high or the distance is too
large. If you select the time steps too large, it is possible that the simulation
cannot be carried out successfully. It is then said that the mathematical
model does not converge. But also if you choose the time steps too small it
is possible that the simulation cannot be executed successfully. This is
because the Amesim program writes results to the hard disk for each time
step. If too much data is written, problems with disk space may occur.
Fig. 3.4 Generating diagrams: visualisation of the result by dragging the variable on the work space
Next, let us take a look at the most important simulation results that
result from the simulation for our ball model. We want to analyse them.
Drag the desired result variables over the workspace to create a diagram of
the value as shown in Fig. 3.4. After that, we will give some suggestions for
further developments or further investigations on our digital twin. One thing
should be kept in mind when creating the model. If something would not
work in reality, e.g. if the mass of the ball or the distance to the ground
would be much too high, then this can also lead to an abort and an error
message. That’s why simulations should only be done with halfway
plausible parameters.
Fig. 3.5 Size of gap is a measure of the distance between the ball and the ground

3.2.3 Results and Analysis

The program stores a lot of different physical quantities in the background
during the simulation run. This could lead to a lot of data if the model is big
and time steps are low. However, in this section we will only look at the
really interesting quantities and discuss them. We will show you how what
this means.
To be able to draw and map simulation results in a diagram, one must
click on the result variable after the successful simulation, then drag the
name of the variable as shown in Fig. 3.4 from the variables window to the
workspace with the mouse held down. This will generate a diagram of this
result variable over the time course of the simulation.

Important: Drag a result variable with the mouse from the menu on the
right side to the workbench to plot a diagram in an x,y-chart.

What results are of particular importance to us in the bouncing ball? At

the moment, there are essentially two quantities in this model. We are
interested in how the height of the ball changes and how it bounces back
from the ground. We can best visualize this by looking at the gap or
clearance between the ball and the ground (see: Fig. 3.5). How should we
now read and interpret the diagram in Fig. 3.5? Suppose we drop the ball
from a height of 1.0 [m] to the ground. After about one second it hits the
ground for the first time at a height of 0.0 [m]. It then jumps up again but
does not reach the full height. After several jumps, the distance between the
ground and the ball becomes smaller and smaller. Your first simulation for
the bouncing ball was successful. After this first trial we look at more
advanced problems and work suggestions for this example.

Important: If you want to display two result variables in a single

diagram, you should simply drag the second variable onto the diagram
window in the same way as the first one. This way the second value will
also be displayed in the same diagram.

To get to know the digital twin of your bouncing ball problem a bit
better, let us work on a few more advanced work suggestions. In addition,
let’s take a look at the speed of the bouncing ball. For this we drag the
variable Velocity at port 1 from the variables field onto the workbench to
display this variable in a diagram as well. When the ball is released the
velocity is still small, towards the end it gets bigger and bigger until the ball
hits the ground and is bounced back.

Problems

3.1 Investigate after what time the ball first touches the ground if you set
the distance between the ball and the ground at the beginning from the
height 1.0, 2.0 and 4 [m].

3.2 How does the trajectory of the ball change when you increase the
weight of the ball from 1.0 to 1.5 [kg] and to 2.0 [kg]?

3.3 How does the trajectory of the ball change if you decrease the contact
damping force from 150 to 50 [N/(m/s)]?
3.3 The Mechanical Rocker
Now let’s consider another use case that we can recreate with the
mechanical library. We want to create a twin for a mechanical rocker. On
the right side, we want the bar of the rocker to be pushed down. On the left
side of the seesaw, we want a spring that is attached between the left lever
arm and the bottom. Let’s start right away and build the model of the rocker
as shown in Fig. 3.6.

3.3.1 Simulation Model

Our simulation model for the digital twin of the mechanical rocker initially
consists of a lever arm on the left side to which the spring is attached. The
model lever2 [LML012] is used as the rocker. The length of the left arm is 1
[m]. On the right side, a force pushes the rocker down by a distance of 1
[m]. The right lever arm of the seesaw should also be 1 [m] long. For the
time being, it does not matter how great the friction is at the joint of the
rocker. Moreover, this digital twin always assumes that we are on Earth and
that the gravity of our planet is acting.
Once you have created the simulation model for the mechanical rocker
with its associated spring, you can enter the specific values, such as spring
force, length of the lever arms, or the distance by which the right lever arm
is pushed down. We will make these specifications in the next section
Submodel and parameters.
Fig. 3.6 Simulation model for a mechanical rocker with lever arms of equal length

3.3.2 Submodels and Parameters

We want to try to push down the rocker arm on the right side in the model.
How can we teach this operation to the twin? To do this, we selected a ramp
function from the signal library (red library) in Fig. 3.7 [RAMP0]. For this
item, we specify the value for the slope of the ramp function as 1. This
means that within one second the right arm is pushed down by one meter.
On the left side of the seesaw we attach the spring.
Fig. 3.7 Ramp function that allows us to push the seesaw down on the right side
In the next step, we have selected an element that shows us how far
down to push the right arm of the rocker. In the element for the seesaw we
specify how long the two arms of should be. For both sides of the seesaw
we should choose lever arms that are 1 [m] long. The values from Table 3.3
must always be entered in the parameter mode.
The simulation time is set to 1 [s] according to the given Table 3.4. In
this calculation we have to take care that the simulation time is not too long.
This is for a very simple reason. If we choose the simulation time too long,
then the arms of the rocker on the right side will be pushed down too far,
which would lead to nonsensical behavior.
Table 3.3 Parameters for the mechanical rocker simulation model

Item Parameter
RAMP0 Slope = 1 []
XVLC01 Time constant for derivative of displacement = 0.0001 [s]
LML012 Distance port 1 to pivot = 1 [m]
Distance port 2 to pivot = 1 [m]
SPR000A Spring stiffness rate = 1000 [N/m]
Spring force with both displacements zero = 0 [N]
Table 3.4 Simulation time for the simulation model of the mechanical rocker

Simulation settings
Start time = 0 [s]
Final time = 1 [s]
Print interval = 0.01 [s]

3.3.3 Results and Analysis

We want to evaluate two different events. Of course, with the rocker, we
already know what needs to happen. It is always good to check the results
for plausibility and to critically question if everything is correct. First, in
Fig. 3.8 we have a look how the movement of the lever arms changes. The
left arm of the beam moves exactly one meter upwards in a time span of 1
[s], the right lever of the seesaw moves one meter downwards. This is a
reasonable behavior.

Important: Always check your models for plausibility and logic before
you make the models too complicated. Check the dimensions and the
units of your results.

The fact that the right bar is pushed down is shown by the negative
value. If the left bar had a greater length, it would move further up than the
right bar would move down. So we see that both lever arms are moved by 1
[m] in the time 1 [s].
Fig. 3.8 Height by which left lever is lifted up and the right lever arm is pushed down
Fig. 3.9 Angle in degrees by which the beam of the rocker rotates around its axis
We can also look at something else in Fig. 3.9. That is, by what angle
the beam of the rocker is rotated around its axis. This can be evaluated by
the size angular lever position. Here, the seesaw rotates by about 57 . The
example is relatively simple now, but we can do the following suggested
work to make it a little more interesting and explore our mechanical
problem a little further. In the Problems part you find suggestions for
further work.

Problems
Why don’t they do the following investigations with your rocker model?

3.4 How do the results change when we press the seesaw for more than
one second? At what point do the results become nonsensical and why?

3.5 Check how the height to which the weight is pulled changes when the
mass hanging at the bottom of the rope changes. Does the length of the rope
also change during the process?

3.6 What happens if we extend the right arm of the bar, e.g. to 3 [m]?
Interpret your new results.

3.7 What changes if we increase the spring force on the left arm of the
seesaw? Interpret these results as well.

3.4 How a Row on a Pulley Works

We will build a mechanical twin for a pulley. With the rope we will lift a
weight. The rope is to run over a pulley. We will also investigate whether
the pulley runs without friction or not. When modeling, you will notice that
the length of the lower end of the rope also has an influence on the result.
This is because this part of the rope can still stretch when it is pulled. Please
always remember this when you want to practice bungee jumping.

3.4.1 Simulation Model

Our simulation model in Fig. 3.10 for the mechanical twin of the rope on a
pulley consists initially on the left side of a force F with the unit Newton
[N]. This can be used to pull on the rope. Then the rope runs over the pulley
and at the hanging end of the rope there is the weight. A pulley never runs
completely without friction, so you have to add a model for the friction
(rotary load) to the pulley.
Again, it doesn’t matter how big the friction is, the force on the left side
or the weight hanging on the bottom of the rope is. You should first
understand the basic structure of the simulation model for the problem.
Then you assemble it and test it for function. Only after that they should
enter the exact parameters.
In the following step, one can then enter the concrete values, e.g. the
weight or length of the rope. In addition, the gravity of the earth is still
specified with the apple symbol. We discuss later, if this makes sense. If
you want, you can remove the apple and check what happens. In the
following section Submodel and parameters we enter the parameters of the
model.

Fig. 3.10 Simulation model for a rope to lift a weight running over a pulley

3.4.2 Submodels and Parameters

Let’s take a closer look at the parameters again. We are pulling on the left
side of the rope with a force of 50 [N]. Since we are pulling on the rope to
the left, we need to set the force to −50 [N], otherwise we would push the
rope, which makes no sense. So, the value is given with a negative sign. In
addition, we must now also specify the size of the pulley we want to use
and we must also choose the length of the rope. The weight of the body we
want to lift is also crucial, and we must specify that it should weigh 1 [kg].
All other values are given in Table 3.5.

Important: Gravity has a great influence on many mechanical models,

but also on other physical phenomena. Thus, the applications may be
different in space or on other planets.

The symbol with the apple [GROV0] indicates the magnitude of gravity.
On the earth the gravitational acceleration of 9,80665 [m/s ] acts almost
everywhere. If we were on another planet like Mars or Venus, e.g. we
would have to use a different value for gravity. If we were on our
neighboring planet Mars, we would not be able to use this digital twin
because the gravity there is 3.711 [m/s ]. On the moon, the gravity is 1.62
[m/s ].
We can’t change the gravity acceleration arbitrarily in the individual
items, and the apple icon doesn’t affect all models the same. Unfortunately,
however, you can always paste the icon on the desktop, and that’s what we
did, because the apple icon looks very nice. But inside the submodels,
unfortunately, the gravitational constant of the earth is still used in some
places. This should be adapted once in following versions of the program
Amesim. Then digital twins can also be created for other planets.
Here we want to do the calculation for 4.0 [s]. Please make sure that the
simulation time in Table 3.6 is not too long, otherwise there is a risk that
you will pull the whole rope through the pulley. Our model is not designed
for this and no useful results would come out.
Table 3.5 Parameters for the simple rope hoist simulation model

Item Parameter
CONS00 Constant value = −50 [N]
RSHE002A Diameter of the sheave = 500 [mm]
Roping angle = 90 [degree]
MECROPE0 Stiffness of unit length of rope = 1e+06 [N/m]
Viscous friction of unit length of rope = 1000 [N/m/s]
Initial length = 10 [m]
Item Parameter
RL00A Moment of inertia = 1 [kgm**2]
Coefficient of viscous friction = 1 [Nm/(rev/min)]
(All other values are set to 0)
MAS001 Mass = 1 [kg]
GRAV0 Constant gravity value = 9,80665 [m/s/s]

Table 3.6 Simulation time for the simulation model simple wire rope hoist and pulley

Simulation settings
Start time = 0 [s]
Final time = 4 [s]
Print interval = 0.01 [s]

3.4.3 Results and Analysis

What result can we evaluate now? It definitely makes sense to first look at
how many meters the weight at the bottom of the rope is lifted when pulling
with a force of 50 [N]. Again, please pay attention to the sign. It is negative
because the rope is pulled away from the pulley.
Take care, when specifying boundary conditions such as forces,
velocities or mass flows, it is always important in which direction the force
or the flows act. According to the direction you have to select the sign. If
you do not know which sign to use. Just test if it has to be a plus or a minus.
If we look at the displacement port variable, we see in Fig. 3.11 that the
weight can be pulled up by about one meter in 4 [s]. This fact is still
relatively easy to understand and we could have derived it with simple
physical considerations.
Just take a closer look at Fig. 3.12. This is because our digital twin can
predict even more. Since the rope is pulled up very quickly over the pulley,
there is a brief lengthening of the rope at the beginning of the pulling
process. It is first tensioned, so to speak, and thus stretched. You can see
this in the following figure from the brief change in length. As soon as the
mass is pulled up evenly, the change in length of the rope decreases again.
This process is much more complex and cannot be determined so easily
with paper and pencil without a simulation model of the technical system.
However, the additional lengthening of the rope can be quite important in
engineering or in daily life. Just think of bungee jumping.
Fig. 3.11 Height by which the weight is lifted at the bottom of the rope
Fig. 3.12 Change in length of the rope due to acceleration at the start

Problems
Now let’s look at what else we can use our digital twin of the rope running
over a pulley for. They can investigate the following work suggestions on
their own using their simulation model:

3.8 What would happen if they double the force with which they pull on
the rope. How fast will the mass be lifted then?

3.9 Check how the height to which the weight is pulled changes when the
mass hanging at the bottom of the rope changes. Does the length of the rope
also change during the process?

3.10 Try to create a simulation model where you run the cable over two
different pulleys. What does this do for you?

3.11 Just remove the apple symbol and see if the model still runs.

3.5 Damper of a Driving Car

A damper or shock absorber holds the wheel of a car and is a mechanical,
hydraulic or pneumatic device designed to absorb and damp shock impulses
from the road. It does this by converting the kinetic energy of the shock into
another form of energy which is then dissipated [1]. We want to show you
in this example how engineers work on a problem and try to solve it step by
step.

3.5.1 Simulation Model

Pneumatic and hydraulic damper or shock absorbers are used in conjunction
with cushions and springs. An modern automobile shock absorber contains
spring-loaded check valves and orifices to control the flow of oil through an
internal piston [1]. In this example we concentrate on the simpler
mechanical damper, because pneumatic and hydraulic library will be
introduced later. The simulation model in Fig. 3.13 for the damper consists
only of one wheel for the the whole car, this is done to make the model
simpler. After sketching the model of the damper, one can define the
parameters.
Fig. 3.13 Model for one wheel of a car with a mechanical damper and road profile

Table 3.7 Parameters for the shock absorber of a driving car

Item Parameter
MAS002 (of the car)
Mass = 1500 [kg]
Inclination = 0 [degree]
SPR000A Spring rate = 100000 [N/mm]
Spring force with both displacements zero = 0 [N]
DAM0000 Damping rate = 1000 [N/(m/s)]
MAS002 (of the wheel)
Mass = 35 [kg]
Inclination = 0 [degree]
Item Parameter
SD0000A Spring rate = 100000 [N/m]
Damper rating = 1000 [N/(m/s)]
XVLC01 Time constant for derivative of
Displacement = 0.0001 [s]
UD00 Constant gravity value = 9,80665 [m/s/s]
Number of stages = 2
Cyclic = no
Time at which duty cycle starts = 0 [s]
Output at start of stage 1 = 0 null
Output at end of stage 1 = 0 null
Duration of stage 1 = 2 [s]
Output at start of stage 2 = 0.08 null
Output at end of stage 2 = 0.08 null
Duration of stage 2 = 8 [s]

3.5.2 Submodels and Parameters

We have mentioned that the model of the car is simplified, we modeled only
one wheel of the car and the total mass. We have modeled the car with only
one mass element and estimate the weight of the vehicle to be 1500 [kg].
The second mass model is for the wheel. In Table 3.7 we give the full list of
parameters for the model.
We took a simulation time of 10.0 [s], which is of course rather short, so
we only want to model the point of time where the wheel hits the edge of
the sidewalk. The simulation time and time step is defined in Table 3.8.
Table 3.8 Simulation time for shock absorber of a driving car

Simulation settings
Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]

3.5.3 Results and Analysis

Before we have a deeper look at the results of the car movement and
damper behavior, we want to check if the simulation is running and the
model works. In Fig. 3.14 the signal of the submodel [UD00] is shown and
gives us an impression, that this function was defined in the right way. Of
course, you can make the road profile more sophisticated.
Now let us take a closer look at the more interesting results. The
damping of the wheel and bouncing of the damper after step is shown in
Fig. 3.15. You can see, that the pneumatic tire which is modeled by a spring
[SD0000A] reacts with an oscillation of the weight on the step of the
sidewalk.
The movement and oscillation of the whole cabin of the car is shown in
Fig. 3.16 and gives us an impression how the second damping acts on the
car. It can be seen that here the amplitude of the oscillation is even higher.
This is probably not very convenient and should be avoided in an normal
car.

Fig. 3.14 Height of the road profile, e.g. step of a sidewalk

Fig. 3.15 Damping of the wheel and bouncing of the damper after step

Fig. 3.16 Oscillation of the car cabin in the moving car after step of the sidewalk
 At the end of this section you can work on some problems that we have
defined for you. This is a normal and good approach and shows you how
engineers work. First we created a simple model that works and after this
introduction to the problem we will try to improve our car model more and
more until we end up with a good car behavior.

Problems
We have already discussed some problems. Now you can investigate the
following problems for your damper with your simulation model.

3.12 Try to make the damping of the spring system softer to reduce the
oscillation in the cars cabin.

3.13 Change the size of the step and examine, how the damping changes.

3.14 We have modeled only one wheel as an example, try to build a

digital twin of a car with four wheels. Consider how to model a correct road
profile for four wheels.

3.15 Work through the chapters on hydraulics and pneumatics and also try
to build a hydraulic or pneumatic shock absorber.

3.16 Try to generate a more complicated road profile and chose longer
simulation times.

3.6 How to Use 3D Geometries

To get a more vivid idea of the mechanical twin and to be able to have a
discourse during the innovation process, the results can also be visualized
three dimensional in Simcenter Amesim. An example, how to visualise such
an model in 3D is shown in Fig. 3.17. Here we use a model of a spring to
visualise the working behavior. The visualization can be done with
Animation dialog, which can be opened by clicking on the Animation button
in the menu bar of the main graphical user interface (GUI).
Based on every digital twin, 3D geometries can be generated for better
understanding of dynamic simulation results and test their behavior and
interaction. We refer to this also as computer aided engineering (CAE) or
CAx, which includes also computer aided design (CAD). The designer is
given the opportunity to perform real-time tests on the digital twin, thus
applying theory directly to the design and questions how well it works.
Theoretical knowledge is deepened through such virtual dry runs and
various application scenarios can be run through independently. In the
following chapter examine the digital twin of a chassis with a hydraulic
damping system when driving over an obstacle. A complete vehicle in a
landscape can also be visualized, as in Fig. 3.18. The 3D geometry of the
car can be created by the designer, e.g. this could also be done with the tool
Blender [2] or any other alternative CAD tool. How the vehicle behaves,
how the damper system works or which forces act on the chassis depends
on the spring model of the damper an is then again calculated by the
Amesim main program making it possible to test the virtual geometry of the
chassis under real conditions and to investigate in advance how it works.
With this model, e.g. driving cycles of the whole car can be simulated, as
shown in Fig. 3.18. The behavior of different bumpers can be examined.
The results are much more understandable, if the whole car maneuver can
be visualised.
Fig. 3.17 Animation screen for 3D visualization and animation of the model with the animation
dialog
Fig. 3.18 Results of a driving cycle with 3D geometry of a car and realistic driving behavior
So, simple 3D geometries can be added within the Animations dialog. If
an 1D model is already running, e.g. the spring of the cars damper system,
all the other object of the scene can be added with the Animations dialog.
Standard objects like boxes, spheres, splines or springs can be added by
pressing the Add menu in the menu bar. In addition, 3D models created with
Blender can also be imported. To connect the 1D model from the main GUI,
the result values of the simulation run can be dragged to geometric values
of the 3D model. This can be seen in Fig. 3.19 and is done simply by
dragging the result value on the animation window.
Fig. 3.19 Connecting a 1D spring model with a 3D spring model in the animations dialog
Fig. 3.20 3D animation of the spring at different simulation time steps
After this step the simulations result data can be used to move and
animate the 3D model by pressing the play button in the menu bar of the
animation dialog. As shown in Fig. 3.20 this will lead to an animation scene
of the deformed damper spring.
Furthermore, in conjunction with artificial intelligence (AI) as shown in
a later chapter of this book, the program can learn to find its way over the
obstacle as efficiently as possible. In the remainder of this book we will
refrain from using 3D geometry as this is beyond the scope of this book, but
we have further books on this topic planned.
Today there is a really good new smartphone app called poly-cam [3]. It
is a 3D capture application for high-quality 3D models from photos with
mobile device, and rapidly generate scans of spaces with the LiDAR sensor.
LiDAR is an acronym of light detection and ranging or laser imaging,
detection, and ranging [4]. This process is sometimes referred to as 3D
laser scanning, a special combination of 3D scanning and laser scanning.
Graphical models can be scanned from real objects very easily and you can
use them to create animated presentations of results.
After scanning the 3D models can be edited with the free of charge tool
Blender [2] and it is possible to export them in different file formats, e.g.
GLTF or STL. GLTF (GL Transmission Format) is an open-source and
royalty-free 3D file that supports static models, animation, and moving
scenes. GLTF is used in games, native web applications, and 3D ads. STL is
a file format native to the stereo-lithography CAD software created by 3D
Systems. STL has several acronyms such as Standard Triangle Language or
Standard Tessellation Language. This file format is supported by many
other software packages. It is commonly used for rapid prototyping, 3D
printing, and computer-aided manufacturing. STL files describe only the
surface geometry of a three-dimensional object without any representation
of color, texture, or other common CAD model attributes [5].

References
1. Bauer, H. (ed).: Automotive Handbook, 4th edn, Robert Bosch GmbH, ISBN 0-8376-0333-1, p
584 (1996)
2.
https://​www.​blender.​org
3.
https://​poly.​cam/​
4.
Travis, S.: Introduction to Laser Science and Engineering. CRC Press, Taylor (2019)
5.
https://​en.​wikipedia.​org/​wiki/​STL_​(file_​format)

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_4

4. The Thermal Twin

Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

4.1 Examination of Heat Transfer

After the mechanical library in the previous chapter, we now want to turn to
a completely different problem. For example, we don’t know how much a
seesaw heats up when exposed to sunlight, or what amount of heat is
generated by the friction between the rope and the pulley. Therefore, we
have chosen the so-called thermal library to create a thermal simulation
model. The color of the thermal library is dark brown.
There are three different heat transfer processes. First, we speak of heat
conduction inside a body or between two neighboring bodies. This process
is also called conduction. The second mechanism is called convection. In
this process, heat is transported by a gas or liquid. The third main heat
transfer process is radiation. There are further heat transfer processes, but
we will not go too deep into this topic.

Important: There are different transport mechanisms for heat. In most

cases, heat transport occurs from a higher temperature to a lower
temperature. Not in the opposite direction.

Let us consider practical examples of heat conduction in an aluminium

heat sink and a solar collector. The temperature of a body results due to heat
transfer with its environment or due to transformation processes and
chemical reactions inside it. In heat transport, one essentially considers heat
conduction, convection as well as heat transport by radiation. It is important
to what mass a body has. This allows us to calculate how much heat can be
stored and what temperature it assumes. All individual processes can be
described by different digital twins. At the end, further working suggestions
are given.
Also crucial to understanding the philosophy of Simcenter Amesim is
that the thermal library can be connected to the mechanical library or any
other library in one model. For example, one can calculate how much heat
is generated by friction from a brake disc during the braking process of a
vehicle to a couple of different complex twins. For example, you can link
certain items of the mechanical library with the thermal library. However, it
is only possible to connect items if this makes physical sense. If two items
cannot be connected, then this almost always has a reason.
Now let’s look at our thermal model for the heat sink of a generator. The
design and creation of heat sinks are very important for many technical
applications. For example, almost all electronic components such as electric
motors, generators, batteries or accumulators need to be cooled during
operation.
4.2 Heat Transfer in Electric Generators
When dealing with the temperature of bodies, we should first realize that
the temperature of a body (in [ C]) is usually the effect of a previous heat
or energy transfer in joules [J] per unit time. In the case of an electrical
generator, eddy currents cause a heat input. An attempt is usually made to
use heat sinks to keep the temperature of the material low and to dissipate
heat to the environment. The material from which the generator and heat
sink are made plays a role here. A heat transfer process is given the unit
watt [W].

4.2.1 Simulation Model

For the thermal simulation model, we need to specify for the metal bodies
what material they are made of. Here we want to model the heat conduction
between the two metals iron and aluminium. The necessary material
parameters are already stored in Simcenter Amesim. In addition, we want to
consider the so-called convection, i.e. the cooling of the metals by
surrounding air, as well as the heat radiation.

Important: If we cool a heat sink with a fluid like air or water it is

always important how fast the fluid moves past the body and if
turbulence occurs. This is also characterized by the so-called Nusselt
Number [1].

In our twin, we have to use a special element [THCD00] for the heat
conduction between the metal bodies. We composed our digital twin from a
total of four different metal bodies [THC00]. To tell the model which
material each item in Fig. 4.1 is made of, we have to use a so-called solid
type index.
Fig. 4.1 Simulation model for heat conduction, the heat is transferred between a body made of iron
and a body made of aluminium
A heat source of 50.0 [W] in [THHS0] is attached to the right side of the
iron block and to the aluminium block, respectively. At the top is the
aluminium block where heat is dissipated by convection and radiation.
Here, too, it is possible to change the air speed in order to influence the heat
transport.

4.2.2 Submodels and Parameters

For the model parameters in Table 4.1, note that in each case the two
aluminium bodies have an equal mass, and the same weight was also used
for the two iron bodies. Here, the solid type index = 1 is set for aluminium
and the solid type index = 2 for iron. Other materials would also be
possible. Be careful to use the right solid type index for the components.
Table 4.1 Parameters for the simulation model of an aluminium heat sink

Item Parameter
THHS0 Heat flow rate at port 1 = 50 [W]
Item Parameter
THSD00 Solid type index = 1 [-]
Material definition = pure aluminium (Al) [-]
THSD00 Solid type index = 2 [-]
Material definition = pure iron (Fe) [-]
THC00 Solid type index = 1 [-]
Mass of material = 5 [kg]
THCD00 Contact surface = 1000 [mm**2]
Thermal contact conductance = 1000 [W/mm**2/degC]
THGCV0 Inclination angle = 90 [degree]
Width = 100 [mm]
Length = 150 [mm]
Velocity of the fluid = 4 [m/s]
It has to be noted, that for the estimation of the contact surface for
[THCD00] the volumes of the [THC000] blocks have to be taken into
account. The simulation time in Table 4.2 is chosen to be very long for this
example, because such heat conduction processes can take quite a long time
power source is low and weight is relatively high. Therefore, we adjust the
time step size accordingly to higher time step sizes.
Table 4.2 Simulation time for the model of an aluminium heatsink

Simulation settings
Start time = 0 [s]
Final time = 3600 [s]
Print interval = 0.1 [s]

4.2.3 Results and Analysis

If we want to look at the results and estimate how heat from the source is
conducted first through the iron and then through the aluminium elements,
we look at the time period of 1 h (= 3600 [s]) in Fig. 4.2. We have applied
50.0 [W] of heating power on the right side which is due to eddy currents,
e.g. from an electric generator. The mass elements are numbered
consecutively and after 1 hour the temperature has risen up to approx. 80.0 [
C] for the elements that are directly applied to the 50.0 [W] heat source.
Fig. 4.2 Average temperatures inside the iron and the aluminium blocks
The temperatures inside the metal blocks in Fig. 4.3 result from the
corresponding heat fluxes between the heads and the surroundings. Towards
the left side, the temperatures of both the iron and aluminium blocks only
increased to about 50.0 [ C] each. The explanation for this is that on the
left side, cooling of the two bodies takes place by convection. During
convection, heat is transferred from the metal to the ambient air. The level
of air velocity plays a decisive role in this process.
Fig. 4.3 Heat conduction within the bodies and convection at the boundaries
Heat conduction between bodies is given in units of watts [W]. We see
that about 30.0 [W] of heat flows between both aluminium and the iron.
The convection that takes place on the left side is about 15 [W] after 1 hour.
Note that this is only the case when cooling air is flowing past the heat sink
at a velocity of 4 [m/s]. Fig. 4.4 is also interesting. It shows how the heat is
conducted from the iron core to the aluminium heat sink. We assume here
that the heat conduction is quite good. Let us have a look at the heat flows
in a diagram.
Fig. 4.4 Heat conduction from the iron block to the aluminium heat sink
Near the 50.0 [W] source on the right side, more heat is conducted than
on the left side, where cooling by the air takes place. This can be explained
by the fact that we are already at a lower temperature level there and there
are no longer such large amounts of heat stored inside the material. Now let
us have a look at a few more work proposals or problems that we can
investigate with this thermal twin.

Problems
Using your digital twin for the heat sink, investigate the following work
suggestions once:

4.1 Change the materials and use copper instead of aluminium, for
example. How does this change the temperatures of the bodies?

4.2 Investigate what happens if you increase the masses of the iron bodies
from 5 to 10 and 15 [kg].
4.3 Set the heat source on the right side from 50 to 100 and 150 [W].
Observe what happens to the temperatures of the metal bodies.

4.4 Try to cool the iron body further by increasing the air velocity on the
right side to 20 [m/s].

4.5 Have a look at the result output for convection. There you will find
different comparison numbers, like the Nusselt number or the Reynolds
number. What do these numbers stand for?

4.3 Site Selection for a Solar Collector

We will now look at heat transfer by radiation. In a solar or solar collector,
radiation from the sun strikes a collector body and heats it. Through
subsequent conduction and convection processes, this energy can be further
distributed. However, it is also crucial how much solar energy falls on the
collector on a given day and where it is located.

Fig. 4.5 Simulation model for a solar collector and the heat conduction in the adjacent material
4.3.1 Simulation Model
To design the solar collector in Fig. 4.5, it is important to know where it is
located. The location plays a crucial role in its function. A collector located
near Johannesburg in South Africa has much more solar energy available
than a collector located in Paris or Helsinki. Furthermore, the time of day at
which the solar radiation falls on the collector is of decisive importance. In
most places, the radiation is strongest around noon. In Simcenter Amesim, it
is possible to set these environmental conditions using a custom model.
This makes it possible to design the collector exactly for its particular
position. We could also create the collector from single items. But this is not
absolutely necessary. There is already a ready-made model that we can use.
The solar collector should still be connected with elements made of
aluminium, which should symbolize the design and structure of the
collector support and its heat exchangers. Through the contact, it can give
its energy to these elements and thus will heat them up.

4.3.2 Submodels and Parameters

In the simulation with the parameters from Table 4.3, all components are to
be made of aluminium. We also want to choose a very long period of
30.000 [s] simulation time for this simulation. During this time, the
collector heats up slowly. We select Paris as the location for our twin
number two.
Table 4.3 Parameters for the simulation model of a solar collector in Paris

Item Parameter
THSD00 Material definition = pure aluminium (Al)
THAMBCOND0 City name = Paris
Year = 2007
Month = December
Day = 16
Hour = 8
Minute = 00
THGCV0 Inclination angle = 90 [degree]
Width = 100 [mm]
Length = 150 [mm]
Velocity of the fluid = 5 [m/s]
Item Parameter
THC000 Solid type index = 1 [-]
Mass of material = 50 [kg] and 20 [kg]
THR03 Equivalent emission factor wall/gas = 1 []
Exchange area = 100 [m**2]
Temperature of the gas = 20 [degC]
THRSOL002 Solar radiation setting mode = using ambient conditions
Exchange area = 1 [m**2]
Adsorption factor = 0.7 []
Equivalent emission factor gas/surface = 0.9 []
Surface inlination = 45 []
Decisive for the absorption of the amount of heat is the area of the
collector and at what angle the collector is inclined to the horizon. We can
specify many different cities all over the world to indicate the position.
However, it is also possible to enter the GPS coordinates of a location. GPS
stands for global positioning system. A system by which signals are sent
from satellites used to show the position on the globe. This makes it
possible to calculate yields that a solar collector field can produce. The
simulation times are given in Table 4.4.
Table 4.4 Simulation time for the simulation model of a solar collector

Simulation settings
Start time = 0 [s]
Final time = 30000 [s]
Print interval = 1 [s]

4.3.3 Results and Analysis

We have already described that when simulating the solar collector, it is of
crucial importance where exactly it is located and on which day the
investigation is carried out. We can use the twin to determine the exact solar
path as shown in Fig. 4.6 for any given day over the collector.
Fig. 4.6 Angle with which the sun is above the solar collector
In our example, the morning of Dec. 16, 2007, was considered. These
are, together with 06/24/2011, the best days of the century or millennium.
On the morning of this day, the solar radiation heated up the collector and,
as shown in Fig. 4.7, we can increase the temperature in the components
with the amount of energy absorbed as a result. We can also see in Fig. 4.7
that air cooling of the left component will not have a high impact on the
temperature profile at the low air velocities. The radiant power absorbed by
the collector can now be accurately determined in Fig. 4.8. Unfortunately,
this lacks information on the extent to which the energy has been
suppressed by moisture in the atmosphere like fog, clouds, or rain. Dense
clouds and rain in the atmosphere have of course a very large influence on
the solar radiation.
Fig. 4.7 Temperature profile of the aluminium components of the solar collector

Fig. 4.8 Radiant power impinging on the surface of the solar collector
Problems
With the solar collector model is completed, you can now perform various
tests and modifications:

4.6 Once, change the weight of the left aluminium bracket to half its
value. What happens? How does the temperature change when they
increase the wind speed from 5.0 to 8.0 [m/s]?

4.7 How does the energy input to the collector change, if we move the
collector to Johannesburg or Helsinki? Select some alternative cities for
simulation.

4.8 Consider a model setup where they can use the collector to heat a
medium such as air or water.

Reference
1. Incropera, F.P., DeWitt, D.P.: Fundamentals of Heat and Mass Transfer, 4th edn. Wiley (1996)

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_5

5. The Hydraulic Twin

Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

5.1 What Is Hydraulics?

The word hydraulics comes from the ancient Greek word hydro for water
and aulos for pipe. We refer to a technology in which fluids are transported
or used to transmit power or energy. However, it is not just about water or
oil. Many other fluids in such a system can also be considered. When
transporting liquids, the so-called pressure loss plays a special role [1]. Any
diversion, bend, or fitting will cause a pressure loss within the pipeline. In
the case of liquids, this usually has to be overcome by a pump. Otherwise
no transport takes place and the liquid cannot flow. Systems for the use of
heat or kinetic energy can also be considered as hydraulic systems, e.g.
heating systems or hydroelectric power plants. We want to use the hydraulic
library to create them. The hydraulic library (see: Fig. 5.1) of Amesim is
located on the right side of the workbench in the Library Tree.

5.2 Two Fuel Oil Tanks and a Pump

In the application in Fig. 5.1 we have two fuel oil tanks. The models for
these tanks [TK10] can be found in the hydraulic library. We connect them
by a line with a pump [PU001] in between. With this pump we can increase
the pressure and transport the oil into the second tank. You can also connect
the hydraulic library with other libraries of Simcenter Amesim. For example,
you can connect the pump shaft to the mechanical library to account for
friction. Or you can couple the pump casing with the heat library to
calculate the heat dissipation at the casing [2].

5.2.1 Simulation Model

There is one major difficulty in creating networks of pipes. If one tank is on
a higher position compared to the other, you have to define the slope within
the line connecting the tanks and you will get a pressure difference between
these two tanks. However, since this is too complex for the first example,
we only want to connect the oil tanks with normal pipes without any slope.
The resulting pressures in the pipelines will then only result from the filling
level inside the tanks. We will show how to create inclined pipelines with a
slope later.
Fig. 5.1 Simulation model for two fuel oil tanks with one pump
When all the items are connected, the digital twin for our hydraulic
system is ready. Now we need to look a little closer and define the
submodels and set the parameters according to Table 5.1 for our model. In
this model, we have used the simplest model for a motor [PM001] without
any friction. Of course, we can use much more complicated models from
mechanical library here.

5.2.2 Submodels and Parameters

Both tanks [TK10] should be the same size and contain the same amount of
oil at the beginning of the simulation. We define the contained amount by
specifying the fill level of the tank. It matters in the model whether we
place the connections at the top or at the bottom of the tank. We also need to
define what material properties [FP04] of the fuel oil because the fluid
properties have a major impact on the flow behavior. Please make sure that
each tank contains the same fluid by choosing the same index of hydraulic
fluid for every component.
Table 5.1 Parameters for the simulation model for two connected heating oil tanks
Item Parameter
TK10 Index of hydraulic fluid = 0 []
Initial height of liquid tank = 0.25 [m]
Tank area = 0.5 [m**2]
Minimum height alarm level = 0.1 [m]
Maximum height alarm level = 1.0 [m]
PU001 Index of hydraulic fluid = 0 []
Pump displacement = 100 [cc/rev]
Typical speed of pump = 1000 [rev/min]
PM000 Shaft speed = 1500 [rev/min]
FP04 Index of hydraulic fluid = 0 []
Temperature = 40 [degC]
Density = 850 [kg/m**3]
Bulk modulus = 17000 [bar]
Absolute viscosity = 51 [cP]
You start the simulation again at 0 [s] according to Table 5.2 and let it
run until a calculation time of 10 [s]. When the progress bar has risen to 100
, the simulation has been completed successfully.
Table 5.2 Simulation time for the simulation model for two connected heating oil tanks

Simulation settings
Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]

If a simulation does not run successfully or it takes a long time to

generate a result, there is a possibility to make the simulation time smaller.
This allows to better check which possible errors are present in the
simulation model. Mostly it is due to non-physical parameters.

Important: Always pay attention to what liquid you expect, e.g. whether
it is oil, water, or another liquid. Change the liquid properties, density
and viscosity, with the droplet symbol [FP04]. The index for the fluid to
be used must also be set there. This index can be redefined for each
 hydraulic component. Several different fluids are possible within one
larger system.

If you look at the fluid properties of the liquid in this experiment, you
will notice that it is not water but an oil, e.g. hydraulic oil, heating oil or
diesel fuel, due to its low density. Here the index of the hydraulic fluid = 0
was selected.

5.2.3 Results and Analysis

During the simulation run all important quantities are saved and you can
evaluate the interesting and decisive quantities when it is finished [3]. To
check if the results make sense you can first have a look at a quantity that is
easy to understand. The level inside the two tanks shown in Fig. 5.2 is the
first thing we should look at and analyze. The level of the tanks is the
simplest quantity and we can estimate whether changing this value is
realistic or not.

Fig. 5.2 Filling level of the different oil tanks as a result of the analysis
 We want to show the filling level from the upper tank and the filling
level from the lower tank in the same diagram and can now compare the
results very well. Since both tanks are exactly the same size, the level in the
upper tank has increased by exactly the amount of oil that is missing in the
lower tank.

Fig. 5.3 Torque of the pump shaft and pressure behind the hydraulic oil pump

The next thing is, we want to evaluate the effect of the pressure increase
behind the pump. As the level in the upper tank in Fig. 5.2 continues to
increase, the pressure in the line between the tanks also increases. We have
to apply an increasingly higher pressure with the pump to move the fuel oil
to the second tank. However, we have specified in our model in Fig. 5.3 that
the speed of the pump should be kept constant. Therefore, an ever-
increasing torque must occur at the pump shaft during the pumping process.
This expected result is also shown in the graph in Fig. 5.3.
Problems
Let’s look at some problems we can explore with the digital twin of the two
connected tanks.

5.1 What happens to the simulation when the bottom tank is empty?
Investigate how long you can run the simulation. Discuss the result.

5.2 How do the results change when you increase the footprint of the
lower tank from 0.5 [m] to twice the value of 1.0 [m]?

5.3 Operate the experimental setup with water instead of fuel oil. What do
you have to change? Compare the results.

5.4 Investigate what happens when the speed of the pump is doubled.

5.5 Try to build a network of tanks with a slope of the pipes between
them.

5.3 How Does a Hydraulic Jack Work?

Just like in the previous example with the two tanks, we will have to deal
with liquids. Let’s take a closer look at the example of the hydraulic jack
[HJ001]. It basically consists of a hydraulic piston. A hydraulic jack can be
used to lift up cars or other vehicles to change the wheels.
Fig. 5.4 Simulation model of a hydraulic jack connected with a pressure compensation tank
Most liquids such as water or oil, unlike gases such as air, have the
property that they are incompressible. This means that they cannot be
compressed. Density changes are therefore usually not very large. This
property can be exploited very well in lifting platforms, jacks, or an
excavator bucket of a digger. Volume and mass flows are almost always
constant and can be easily balanced.

5.3.1 Simulation Model

Our model for the jack in Fig. 5.4 consists of an item for the force [FORC].
This force is coupled with the step signal from the red library [STEP0]. It
then acts on the model of a so-called hydraulic jack [HJ000]. The model is
already quite complete and complex. The equations behind the model of the
hydraulic jack can be found in the help function of Amesim. You can open
this help dialog by right-clicking on the item with the mouse. All equations
are very well documented and sketches for the technical models are offered
for better understanding.

Important: For liquid lines, the slope is very important because the
liquid can only flow downwards if there is no pressure difference. If the
 liquid has to flow upwards, we need additional pressure, e.g. from a
pump. Therefore, when creating the line, you must pay attention where
the starting point (1) and the end point (2) are located. Specify the angle
of the line between these starting and end points. Negative angles are
also possible or the pipes can also be connected in reverse.

A pressure accumulator is connected to the bottom of the piston via a

throttle [HYDORF0]. In this case, the pressure accumulator [HA001] is
relatively small at 0.5 [L]. The pressure inside the accumulator is 50 [bar].
An overflow for leakage is also connected to the hydraulic piston. This is
not just a simple connection, but a real line. This model [HL0002] is created
automatically by Amesim you will not find it in the library tree. With this
type of line, it is important whether the fluid can flow from top to bottom.
You must therefore specify the slope of the line.

5.3.2 Submodels and Parameters

Carefully enter the parameters according to Table 5.3. The values must be
re-entered in Parameter mode in the corresponding field. Also make sure
that the index of the hydraulic fluid is set correctly.
Table 5.3 Parameters for the jack simulation model

Item Parameter
STEP0 Value after step = 100 [zero]
Step time = 1 [s]
FP04 Index of hydraulic fluid = 0
Temperature = 40 [degC]
Density = 850 [kg/m**3]
Bulk modulus = 17000 [bar]
Absolute viscosity = 51 [cP]
HJ000 Pressure at port 1 = 40 [bar]
Pressure at port 2 = 0 [bar]
Rod velocity = 0 [m/s]
Rod displacement = 0.3 [m]
Index of hydraulic fluid = 0
Piston diameter = 25 [mm]
Rod diameter = 12 [mm]
Item Parameter
Length of stroke = 0.3 [m]
Dead volume at port 1 end = 50 [cm**3]
Dead volume at port 2 end = 50 [cm**3]
Total mass being moved = 250 [kg]
Angle rod makes with horizontal = 90 [degree]
Coulomb friction force = 0 [N]
Stiction force = 0 [N]
Viscous friction coefficient = 0 [N/(m/s)]
Leakage coefficient = 0 [L/min/bar]
HYDROF0 Index of hydraulic fluid = 0
Number of parallel orifices = 1 []
Orifice geometry = circular []
Diameter = 3 [mm]
HL0002 Index of hydraulic fluid = 0
Section type = circular
Diameter = 25 [mm]
Length = 1 [m]
Relative roughness = 1e-05 []
HA001 Pressure at port 1 = 40 [bar]
Index of hydraulic fluid = 0 []
Adiabatic initialization = 1 []
Gas precharge pressure = 40 [bar]
Accumulator volume = 0.5 [L]
Polytropic index = 1.4 [zero]
TK000 Tank pressure = 0 [bar]
In order to improve the jack’s function, an additional restriction should
be placed between the pressure vessel and the hydraulic piston to limit the
flow and reduce pressure pulsations. Set the simulation time back to 10 [s]
according to Table 5.4.
Table 5.4 Simulation time for the simulation model of a hydraulic jack

Simulation settings
Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]
5.3.3 Results and Analysis
The simulation runs for exactly 10 [s] according to the specification given
in Table 5.4. After the first second, the jack is pushed down from the top
with a force of 100 [N]. As can be seen in Fig. 5.5, this happens suddenly.
The signal occurs within a very short time window. Due to its viscosity, the
hydraulic oil cannot react as quickly as possible. This is noticeable in an
oscillation of the oil pressure.

Fig. 5.5 Force on the jack—after one second the force is increased from 0 to 100 [N]
Fig. 5.6 Downwards movement of the hydraulic piston in the jack
Fig. 5.7 Pressure oscillation—pressure in pressure accumulator increases from 50 to 52 [bar]
We see in Fig. 5.6 that the piston does move downwards in accordance
with the force signal. However, it responds with an oscillation at the end of
the force application due to the damping of the oil volume inside the piston.
The oscillation continues through the pipeline. The components “phone”
each other.
In the accumulator tank, we can see an increase in pressure from 50 to
over 52 [bar]. When the force decreases, the piston can then release the
pressure from the accumulator tank. However, in Fig. 5.7, we also see a
clear pressure oscillation in the accumulator despite the throttle. We can
now perform some experiments with the digital twin of the jack in Fig. 5.4.
We can investigate the following work suggestions for this hydraulic model
on our own.

Problems
What can we investigate with our model if we want to lift up a car? Have a
look at the following problems.

5.6 How do the results change when the moving mass increases or
decreases? Use masses with 150, 250, and 500 [kg].

5.7 What happens if we increase the accumulator? Change it once to 50

[L]. How does this affect the pressure oscillation?

5.8 How does the pressure oscillation change if you increase or decrease
the throttle cross section?

5.9 Change the material values of the hydraulic oil. What would happen if
you ran the jack with water? Does that make sense technically?

References
1. Idel’cik, I.E.: Handbook of Hydraulic Resistance, 3rd edn. Begell House Inc., (1996)
2.
Miller, D.S.: Internal Flow Systems, 2nd edn. Amazon Technology (1989)
3.
Bode, B.: Verfahren zur Extrapolation wichtiger Stoffeigenschaften von Flüssigkeiten unter
hohem Druck. Tribol. Schmierungstech. 37(4), 197–202 (1990)

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_6

6. The Pneumatic Twin

Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

6.1 Pneumatic, Fluid Flow, and Turbulence

Gases such as air have the property of being compressible under common
ambient conditions. This means that they can be compressed by a cylinder.
Simplified, we can say that gases obey the ideal gas law. Density changes
can be very large in gases. Also, the mathematical models and solution
methods are usually more elaborate for compressible media than for
incompressible fluids. Mass flows are also constant here, but volume flows
need not always be constant. Therefore, the same technical applications are
often not possible for gases as for liquid fluids. And solution instabilities
will be different.

Important: The numerical stability of a library is also influenced by its

physics. All fluid media must be divided into compressible or
incompressible media and it must be clear before selecting a library
whether the problem is hydraulic or pneumatic. Use the index of the
fluid to define the properties. The description of turbulence is weak in
Amesim. Use a computational fluid dynamics (CFD) software tool for
detailed description of turbulence.

The word pneumatic comes from the ancient Greek pneuma and means
breath or wind. In creating the digital twin, we examine how gases can be
transported inside pipe systems. When using the pneumatic library, you
should keep in mind that gases are usually compressible. This can lead to
higher computational effort and a worse convergence of the mathematical
equation system than in incompressible problems. It is also difficult to
calculate complicated phenomena such as vortexes or turbulence. To model
turbulence, one should use another simulation tool such as ANSYS CFX or
ANSYS Fluent [1].

6.2 Safety Valve for a Biogas Tank

Biogas consists largely of a mixture of methane and propane gas with small
amounts of air. We want to consider this biogas here simplified only as
methane gas. Via a pipeline of 10 [m] the gas is to be led into a tank with a
volume of 10 [L].
Fig. 6.1 Pneumatic simulation model for the safety valve of a biogas tank
Normally this is possible without any problems. To prevent over
pressure inside the tank and in the pipeline an additional safety valve has to
be installed. This safety valve is to prevent the wall of the tank or the
pipeline from bursting if the pressure inside rises too high. Often spring is
used to seal the valve from the environment. If the gas pressure in the
pipeline rises above a certain level, the valve opens and releases a certain
amount of the gas into the environment.

6.2.1 Simulation Model

Let us build a simulation model like the one shown in Fig. 6.1 in which
methane [PNGD00] is filled into the biogas tank through a pipe. The
pressure in the tank is supposed to rise above 3 [bar]. A safety valve
[PNCV01] is installed in front of the tank [PNCH022] to protect it. The
biogas tank should have a volume of 10 [L]. As soon as the pressure rises
above a value of 1.8 [bar], the valve opens. All other parameters are given
in Table 6.1.
6.2.2 Submodels and Parameters
The simulation times for the biogas tank with the safety valve are given in
Table 6.2. We set the simulation time to a high value to see the full reaction
of the pressure inside.
Table 6.1 Parameters for the simulation model biogas tank and safety valve

Item Parameter
PNGD00 Temperature at port 1 = 293.15 [K]
Fluid definition = air
PNCS001 Index of hydraulic fluid = 0 []
Pressure at port 1 = 1 [barA]
PNGD00 Gas type index = 2 []
Fluid definition = methane (CH4)
PNCS001 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 3 [barA]
PNCV001 Gas type index = 1 []
Orifice area = 5 [mm**2]
Check valve cracking pressure = 1.8 [bar]
Hysteresis for opening/closing = 0 [bar]
PNCH022 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
Gas type index = 2 []
Volume = 10 [L]
Thermal exchange coefficient = 500 [J/m**2/K/s]
Thermal exchange area = 0.1 [m**2]
Exchange temperature = 293.15 [K]
PNL000R Gas type index = 2 []
Diameter of pipe = 10 [mm]
Pipe length = 10 [m]
Relative roughness = 1e-05 [zero]

Table 6.2 Simulation time for the simulation model biogas tank and safety valve
Simulation settings
Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.1 [s]

6.2.3 Results and Analysis

In the simulation results in Fig. 6.2, two curves are now shown in the
diagram. On the one hand, the pressure in front of the tank and, on the other
hand, the mass flow of the methane gas can escape through the safety valve.

Fig. 6.2 Pressure in the biogas tank and mass flow of the gas through the safety valve

After a successful simulation, click on the result variable to draw and

map the simulation results in a diagram in Fig. 6.2. It can be seen that the
pressure increases up to the limit of the gas valve. Then the valve opens and
the methane is released into the environment for safety reasons. Thereupon,
the curve for the pressure flattens out.
Problems
Have a look at a few work proposals that we can explore with our models:
To do this, you should find the appropriate model parameters in the
simulation model and modify them.

6.1 Lower the pressure of the methane gas at the inlet. Observe what
happens.

6.2 Increase and decrease the volume of the tank and observe how this
affects the pressure curve in the tank.

6.3 Change the opening behavior of the safety valve by decreasing the
pressure at which the valve opens. What happens to the pressure in the
tank?

6.4 How do the results change if you use propane instead of methane as
biogas? What changes if you simulate with air?

6.3 Ventilation System of a Building

The design of the ventilation system for a building is not as trivial as it first
seems. However, compared to a system of water pipes, we encounter them
relatively often. In large department stores or furniture stores, you can see
them on the ceiling. Even though air flows have only small pressure
differences compared to liquid flows, differences of even a few pascals [Pa]
can lead to high velocity gradients.
Anyone who has observed ventilation pipes with an expert’s eye may
have noticed that the pipes still have a larger diameter at the beginning. At
the end, where the air exits the vents, the pipes become thinner and thinner.
This is because at the beginning you still have very high volume flows, but
these decrease with each additional branching of the pipe.
In addition, another difference compared to liquids should be pointed
out. Gas flows are usually compressible flows. Gases can be compressed
during movement or at rest. They thereby reduce their volume but increase
their density. This relationship must be taken into account during the
calculation.

6.3.1 Simulation Model

We want to build a simulation model for a building where we introduce air
from a main duct into a ventilation system and then split it to other ducts.
The pipe branches and different lengths result in different pressure losses.
In turn, the volume flow in the other parts of the pipe network depends
on the pressure drop in a pipe. The whole model is shown in Fig. 6.3. The
parameters and submodel are listed in Table 6.3 (see Table 6.4). The
simulation time is given in Table 6.5. To make this example a bit more
interesting and challenging, the reader should try to find the ramp for the
step signal [UD00] by himself. In Fig. 6.4, the mass flow of the compressor
can be seen and the duration of the steps of the signal [UD00] can be
calculated from this diagram.

Fig. 6.3 Pneumatic simulation model of a ventilation system for buildings

Table 6.3 Parameters for the simulation model of a ventilation system (part 1)

Item Parameter
PNGD00 Gas type index = 1 []
Fluid definition = air
PNCS001 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
PNCP00 Gas type index = 1 []
Item Parameter
Compressor displacement = 1000 [cc/rev]
Polytropic constant = 1.35 [null]
PNPC1 Gas type index = 1 []
Diameter of pipe = 8 [cm] or 15 [cm]
Pipe length = 5 [m]
Relative roughness = 1e-05 [null]
PNBP001 Gas type index = 1 []
Hydraulic diameter = 8 [cm]
Curvature radius = 10 [cm]
Center angle = 60 [degree]
Relative roughness = 1e-05 [zero]

6.3.2 Submodels and Parameters

Table 6.4 Parameters for the simulation model of a ventilation system (part 2)

Item Parameter
TPTE001 Gas type index = 1 []
Diameter at port 1 = 15 [cm]
Diameter at ports 2 and 3 = 20 [cm]
Friction factor in the main branch = 0.1 [null]
Friction factor side branch = 1.2 [null]
PN3P000 Gas type index = 1 []
Pressure drop coefficients = Idelchik
Side branch diameter (port 1) = 15 [cm]
Straight passage diameter (port 2 and 3) = 15 [cm]
Angle between side branch and straight passage = 90 [degree]
Critical Reynolds number = 5000 [zero]
Time constant = 1e-06 [s]
Transition accuracy = 0.9 [null]
PNL0000 Gas type index = 1 []
Model = polytropic
Diameter of pipe = 10 [cm], 15 [cm] and 20 [cm]
Pipe length = 1 [m], 2 [m], 3 [m] and 20 [m]
Polytropic constant = 1.35 [zero]
Table 6.5 Simulation time for the simulation model of a ventilation system

Simulation settings
Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.1 [s]

6.3.3 Results and Analysis

To generate the airflow, we use a compressor. Actually, a blower should be
used for this purpose. However, modeling a blower in Simcenter Amesim is
quite complex and only feasible with additional efficiency tables.
Therefore, we want to do without it in the model in Fig. 6.4. The
compressor is driven by a motor. Different speeds are to be investigated.

Fig. 6.4 Mass flow rate of the air at the compressor at different speeds

In Fig. 6.4, we look at the mass flows of air at the end of the three
branches of the ventilation system. There is a different mass flow in each
pipe. Depending on the length and diameter of the pipes, but also due to
diversions, there are different pressure drops. This further complicates the
behavior of the pipe system.
Fig. 6.5 Mass flow of the air at all three outlets of the ventilation system
You can see the air distribution at the three outlets of the duct system in
Fig. 6.5. Interpret what this means for the ventilation system. Is it desirable
to have a strong uneven distribution of air flows? How does this affect the
indoor climate inside the building? Also work on the following suggestions.

Problems
We want to look at following problems and solve them with the model of
the ventilation system:

6.5 Try changing the speed distribution on the compressor. How does this
affect the airflow?

6.6 Does the air distribution stay the same if they keep increasing the
speed of the compressor?

6.7 Investigate how the air distribution changes when they make
individual pipes thinner.
6.8 Each of the ventilation pipes should have a slider. You can open and
close it. Sliders result in changed pressure drops at the outlets, on which in
turn the volume flow depends. How can they create a uniform air
distribution?

6.9 Create additional flow channels and try to adjust them so that the air is
distributed as evenly as possible at the outlets.

Reference
1. https://​www.​ansys.​com/​

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_7

7. The Electric Twin

Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

7.1 Permanent Electric Motor with Load

The electrical motor model shown in Fig. 7.1 is a permanent magnet motor.
It can be found in larger fluid energy machines, such as for driving ship
propellers or compressors. The model can work either as a motor or a
generator. The model of the machine or generator is [EMDPMDC01] which
can be seen as reversible.

Important: The electric current in the rotor is needed to produce torque.

It is obtained by electromagnetic induction from the magnetic field of
the stator winding [1]. This produces waste heat inside the armature. An
induction motor can be made without electrical connections to the rotor.
An induction rotor can be either wound or squirrel-cage type [1].

7.1.1 Simulation Model

Let us build the simulation model as shown in Fig. 7.1, in which the motor
is introduced and connected to the electric current. The current is given by a
function from the signal library. The signal step model [UD00] is a duty
cycle submodel with a dimensionless output. The user may specify three
stages giving a start value, an end value, and the duration for each stage.
Linear interpolation is used to determine the output. Thus constant sections,
ramps, and steps may be constructed with the signal library. For the signal
library, a designer dialog is given to define the specific application to create
the piecewise linear function. It is available by double clicking on the icon
in parameter mode and in simulation mode. The item is used to generate
piecewise linear signals like ramps, steps, squares, saw tooth, or trapezoidal
signals.
Fig. 7.1 Permanent electric DC motor with voltage source and mechanical load
Using the electrical motor convention, the rotor relative speed can be
measured by sensors for torque T and rotary speed as an output speed at
the mechanical port and the electromagnetic torque can be measured with T
as an output torque at the mechanical port. When and T are of same sign,
the machine is working as a motor and the output mechanical power P is
positive. When and T are of opposite sign, the machine is working as a
generator and the output mechanical power P is negative. This formula can
be used for further examinations.
Fig. 7.2 Stepwise voltage increase as an input to the permanent electric motor
The diagram in Fig. 7.2 shows a stepwise voltage increase and the
resulting rotor speed of this example. You can choose whether the machine
is acting as a motor or generator by looking at the signs. If the signs of the
speed (n) and the torque (M) are different, the result of the mechanical
power is negative (P) and thus the machine is operated as a generator. As
can be seen, the signs are different in this example and the machine is
regarded as a generator. The following formula gives the ratio of speed and
torque:
(7.1)
With the variables P for power in [W], M torque [Nm], and n speed [1/s] .
The electromotive force and torque coefficient and the armature winding
resistance are corrected with the temperature. The variables to solve this
equation could be taken from the sensors in the model (see Table 7.1).

7.1.2 Submodels and Parameters

Table 7.1 Parameters for the simulation model of the permanent motor with load
Item Parameter
EMDPMDC01 Reference temperature = 25 [degC]
Armature winding resistance = 0.6 [Ohm]
Corrective coefficient armature winding = 0 [1/K]
Armature winding inductance = 0.012 [H]
Electromotive force and torque constant = 1.8 [V*s/rad]
Corrective coefficient electromotive force = 0.0 [1/K]
UD00 Number of stages = 3 []
Cyclic = no
Time at which duty cycle starts = 0 [s]
Output at start of stage 1 = 48 []
Output at end of stage 1 = 48 []
Duration of stage 1 = 3 [s]
Output at start of stage 2 = 100 []
Output at end of stage 2 = 100 []
Duration of stage 2 = 3 [s]
Output at start of stage 3 = 200 []
Output at end of stage 3 = 200 []
Duration of stage 3 = 3 [s]
THTS1 Temperature at port 1 = 20 [degC]
RL02 Moment of inertia = 0.005 [kg m**2]
To improve the function of the electric model, set the simulation time to
9.0 [s] according to Table 7.2. The armature current can then be computed
inverting the generalized Ohm’s law equation for the rotor phase. For
details concerning the calculation of power, energy, and activity in this
submodel, please refer to the definitions specific to this library in the help
menu.
Table 7.2 Simulation time for the simulation model for permanent motor with load

Simulation settings
Start time = 0 [s]
Final time = 9 [s]
Print interval = 0.001 [s]
 The simulation times for the twin of the permanent electrical motor with
load are given in Table 7.2.

7.1.3 Results and Analysis

When looking at the simulation results of speed and torque in Fig. 7.3, two
curves are now shown in the diagram. This is always helpful if you want to
explain how the values influence each other. After successful simulation,
you can click on the result variable and drag it into an existing diagram like
in Fig. 7.3.

Fig. 7.3 The torque [rev/min] and speed [Nm] of the electric permanent DC motor

On one hand, the relative motor speed due to the different currents is
shown. After each step, an oscillation of the rotor speed can be seen, even if
the electromagnetic torque remains more or less constant. It reacts to the
oscillations in the generator’s magnetic field.

Problems
For our model of the electric motor we can make several working
suggestions. Take a look at a few work suggestions that we can explore with
our twin: To do this, they should find and modify the appropriate model
parameters in the simulation model.

7.1 Change the armature winding resistance to 1.0 [Ohm] and to 2.4
[Ohm]. Explain what happens to the results.

7.2 Change the temperature of the [THTS1] from 20 [degC] to 40 [degC].

What happens?

7.3 Couple the motor’s armature with an additional convective cooling

element from the thermal library. Test different materials, such as iron, cast
iron, aluminium, etc.

7.4 Change the load’s moment of inertia. Explain what is happening.

Fig. 7.4 Asynchronous squirrel-cage induction machine with three phase and mechanical load

7.2 How an Asynchronous Motor Works?

A three-phase, asynchronous motor consists of two basic components, stator
and rotor. The stator is the fixed part of the motor. It has a stator core and
field windings often made of metal wires. Inside the stators of three-phase
asynchronous motors, there is a separate winding for each phase. It is
operating with three-phase alternating current. An asynchronous motor
model setup in Fig. 7.4 shows a three-phase motor operated with all three
phases of alternating current. These asynchronous motors are often used in
industry for pumps, tools, or even electric vehicles. Friction at mechanical
load lowers the efficiency of the motor. Heat conduction plays an important
role for its efficiency.

7.2.1 Simulation Model

Let us build a simulation model of the electric motor where we can change
the three phases of the electric current stepwise by a sinus signal function.
The sinus functions can be modeled by [SIN0]. Figure 7.4 shows the input
voltage of the asynchronous motor. The phases are each shifted by 120
[degree]. The motor is a squirrel-cage induction machine in a rotor frame.
The simulation model [EMDSCIM01] is used to model linear Squirrel-
Cage Induction Machine (SCIM) and it is used to rotate a load.
In the [EMDSCIM01], the reference angle used for transformation is the
rotor position. In the rotor frame, the model quantities vary with the slip
frequency at steady-state operating points. The reference angle used for
transformation is the rotor flux linkage position. In the rotor flux linkage
frame, the model quantities are constant at steady-state operating points.

7.2.2 Submodels and Parameters

The important parameters for the submodels of the three-phase motor setup
are compiled in Table 7.3. Two submodels [EMDSCIM01] and
[EMDSCIM02] return the same results. The model [EMDSCIM02]
produces usually faster simulations but it may raise numerical problem
when the currents tend to zero.
To improve the function, an additional constriction should be placed, set
the simulation time to 0.04 [s] according to Table 7.4. Use [EMDSCIM01]
when the best robustness is required. Note also that [EMDSCIM01] internal
variables in the frame are not consistent with the internal variables of the
control elements.
Table 7.3 Parameters for the simulation for the electric three-phase motor
Item Parameter
EMDSCIM01 Winding connection = star [connectType]
Number of pole pairs = 1 []
Reference temperature = 25 [degC]
Stator winding resistance = 1 [Ohm]
Corrective coefficient on stator resistance = 0 [1/K]
Stator cyclic inductance = 0.25 [H]
Rotor time constant = 0.153 [s]
Corrective coefficient on rotor time = 0 [1/K]
Dispersion coefficient = 0.066 []
RL02 Moment of inertia = 0.005 [kgm**2]
THTS1 Temperature at port 1 = 20 [degC]
W000 Fixed angle = 0 [degree]
SIN0-1 Sine wave frequency = 50 [Hz]
Mean level = 0 []
Sine wave amplitude = 230 [s]
Phase shift = 0 [degree]
SIN0-2 Sine wave frequency = 50 [Hz]
Mean level = 0 []
Sine wave amplitude = 230 [s]
Phase shift = −120 [degree]
SIN0-3 Sine wave frequency = 50 [Hz]
Mean level = 0 []
Sine wave amplitude = 230 [s]
Phase shift = +120 [degree]

Table 7.4 Simulation time for the simulation model of the electric three-phase motor

Simulation settings
Start time = 0 [s]
Final time = 0.04 [s]
Print interval = 0.0001 [s]

7.2.3 Results and Analysis

In Fig. 7.5, you can see the phase shift of the three different currents. The
machine windings are balanced, and two different connections are possible.
The enumeration winding connection enables to choose between star and
delta connection for the equivalent windings A, B, and C.

Fig. 7.5 Phase-shifted voltage curves of the asynchronous motor with three phase and mechanical
load

The three-phase electrical ports are associated on the icon with the
notations I, II, and III. The star and delta connections for the windings A, B,
and C as well as the winding voltages and currents are defined within this
convention.

Problems
Let’s look at some problems that we can now investigate with our electric
motor twin. To do this, you should modify the corresponding model
parameters in the simulation model.
7.5 Change frequency of the three sinusoidal signal functions up to 100
[Hz].

7.6 Try a different stator winding resistance like 2 or 3 [Ohm].

7.7 Change the rotor time constant to 0.165 [s]. What happens?

7.8 Change the load of the model and change the limit temperature of the
armature of the motor. Give an interpretation of your results.

7.3 Electric Generator with Resistor and Heat

Transfer
The permanent magnet motor or generator can be defined as a motor which
includes a permanent magnet pole, and therefore it is called Permanent
Magnet DC Motor (PMDC). The motor includes an armature core,
commutator, and armature winding. There are usually two different types of
windings in a conventional DC motor. A solenoid can be used to make the
flux working within the air gap in its place of the field winding. This type of
machine is always reversible, so it can work either as a motor or as a
generator. The rotor structure is similar to the straight DC motor.

7.3.1 Simulation Model

Let us build a motor simulation model in combination with the thermal
library that was already described before and combine it with a heat transfer
problem. Figure 7.6 shows the input voltage permanent magnet DC motor.
Fig. 7.6 Electric permanent magnet DC motor (PMDC) in combination with the thermal library
The model setup in Fig. 7.4 shows a generator operated by an
mechanical drive [OMEGC0]. The main function of field winding is to
produce the functioning magnetic flux within the air gap as well as wound
on the stator of the motor while armature winding can be wound on the
rotor. Inactive carbon brushes are pushed on the commutator like in
conventional DC motor. Using the motor convention, the rotor relative
speed can be measured with W as an output speed at the mechanical port
and the electromagnetic torque can be measured with T as an output torque
at the mechanical port. When W and T are of same sign, the machine is
working as a motor and the output mechanical power P is positive. When W
and T are of opposite sign, the machine is working as a generator and the
output mechanical power P is negative.

7.3.2 Submodels and Parameters

The important parameters for the submodels of the generator setup are
compiled in Table 7.5. There are two submodels, namely, [EMDPMDC01]
and [THC000]. The model [EMDPMDC01] produces a heat source to heat
up the mass model [THC000].
Table 7.5 Parameters for the simulation for the electric three-phase motor
Item Parameter
EMDPMDC01 Reference temperature = 25 [degC]
Armature winding resistance at reference temperature = 0.6 [Ohm]
Corrective coefficient on armature winding resistance = 0 [1/K]
Armature winding inductance = 0.012 [H]
Electromotive force and torque constant at reference temperature = 1.8 [V*s/rad]
Corrective coefficient on electromotive force and torque constant = 0 [1/K]
EBR02 Type of reference resistor value = resistor value directly entered by user
Resistance = 1 [Ohm]
THSD00 Solid type index = 1
Material definition = pure aluminium (Al)
THC000 Solid type index = 1
Mass of material = 0.5 [kg]
UD00 Number of stages = 3 []
Cyclic = no
Time at which duty cycle starts = 0 [s]
Output at start of stage 1 = 48 []
Output at end of stage 1 = 48 []
Duration of stage 1 = 3 [s]
Output at start of stage 2 = 100 []
Output at end of stage 2 = 100 []
Duration of stage 2 = 3 [s]
Output at start of stage 3 = 200 []
Output at end of stage 3 = 200 []
Duration of stage 3 = 3 [s]
To improve the function, an additional constriction should be placed, set
the simulation time to 0.04 [s] according to Table 7.6. Use [EMDSCIM01]
when the best robustness is required. Note also that [EMDSCIM01] internal
variables in the frame are not consistent with the internal variables of the
control elements.
Table 7.6 Simulation time for the simulation model of the electric three-phase motor
Simulation settings
Start time = 0 [s]
Final time = 10.0 [s]
Print interval = 0.01 [s]

7.3.3 Results and Analysis

To examine the simulation model of the generator, we check the torque at
the shaft of the generator firstly to see if mechanical load is provided. In
Fig. 7.7, we can see that all boundary conditions are set correctly.

Fig. 7.7 Checking the boundary conditions of the generator’s torque based on mechanical drive

In Fig. 7.8, the electric power is presented which can be calculated by

the P formula. The mechanical load will be converted into electric capacity
in the generator and the direct current machine will not only generate
electric energy, but also heat.
(7.2)
The thermal losses are identified with Joule’s losses in the machine which
can be estimated by the changing temperature in the generator’s housing, as
can be seen in Fig. 7.9. Joule heating, also known as resistive, resistance, or
Ohmic heating, is the process by which the passage of an electric current
through a conductor produces heat.

Fig. 7.8 Changes of the current and voltage steps based of the production of the generator
Fig. 7.9 Joule’s losses produce a heat source in the housing of the generator
The temperature change in the housing is influenced by the mass of the
housing [THC000] and the material. Here in Fig. 7.9 we used aluminium,
but alternative materials like iron or cats iron can also be used. We can see
how the temperature changes inside the generator housing.
Cooling of housing for electric generators as well as the construction
and design of fins for heat conductors is an important topic. It is widely
used for different applications like wind turbines, power plant, and other
fluid energy machines. Here we see a good example of the connection
between the thermal library and the electric library.

Problems
We want to have a look at some problems that we now can explore our
generator. To do this, you should modify the simulation model.

7.9 Change the thermal mass that is connected to the housing to 5 [kg]
and to 10 [kg] and compare the results. Use cast iron as alternative housing
material. Discuss the results.

7.10 Change the armature winding resistance at reference temperature to

2 and 6 [Ohm]. What happens?
7.11 How long does it take to heat up the housing up to 350 [K]. Do you
see any influence on the current?

7.12 Try to build a convective cooling unit with the thermal library to
cool down the electric housing.

Reference
1. IEC 60050 Rotation Machinery – General, IEV ref. 411-31-10: Induction Machine – an
asynchronous machine of which only one winding is energized. Publication date: 1990-10.
Section 411-31

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_8

8. Analysis of Complex Technical Systems

Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

8.1 How Does the Liquid Piston Compressor

Work?
The storage of energy, e.g. from wind or water turbines, is one of the main
objectives in the future development of stable power grids [1, 2]. In this
chapter, a model for a liquid piston compressor (LPC) is presented. This
LPC can be used to store mechanical energy produced by a wind or water
turbine [3]. Based on a short explanation of the structure and function of an
LPC, it is shown how such a system can be modeled from existing
hydraulic, pneumatic, and mechanical components. In sections that build on
one another, a basic model is first developed and parameterized.
Subsequently, the model is successively extended, e.g. second piston,
control system, etc. and simulation results are presented and discussed in
diagrams after each section. Finally, a parameter study on the overall model
is presented.

Important: In Amesim all elements of the different libraries can be

connected and combined to create more complex models. They can be
combined with each other or with the signal library. In the student
version only some basic libraries are included.

8.2 Design and Function of a Liquid Piston

Compressor
In one of the simplest designs of a compressor with a liquid piston, the
liquid piston acts as a direct hydraulic-pneumatic transformer. This
possibility of compressing air in the application of an energy storage system
was presented, among others, in [4]. Figure 8.1 schematically shows the
circuit diagram of such a system with a hydraulic-pneumatic transformer.
Fig. 8.1 System with technical components of a liquid piston compressor (LPC) [3]
In this process, a cylinder filled with gas is filled with a liquid, e.g.
hydraulic oil or water, by a hydraulic pump (the hydraulic valve was
controlled for this purpose—switching position 2). The liquid flowing into
the cylinder compresses the gas in the cylinder and the pressure in the
system increases. When the pressure prevailing in the accumulator is
reached, the check valve “2” opens and releases the gas flow to the
accumulator. When the cylinder is almost completely filled with the liquid,
the hydraulic valve switches again (switching position 1) and the liquid in
the cylinder can run back into the liquid tank. The vacuum thus created
ensures that the cylinder is filled with gas from the environment via check
valve “1”. The process can then run again. We want to present three variants
of the LPC in detail.

8.3 Liquid Piston Compressor with One Cylinder

for One Stroke
8.3.1 Simulation Model
For modeling purposes, the existing components from the areas of
hydraulics, pneumatic, and mechanics are to be used first. For this purpose,
the real system is replaced by the combination of a spring-loaded single-
acting hydraulic cylinder in combination with a spring-loaded single-acting
pneumatic cylinder. To couple the two cylinders, a mass is interposed, since
coupling of both cylinders is not possible without an inertial mass (natural
frequency of the system). The model for a liquid piston Compressor with
One Cylinder for One Stroke can be seen in Fig. 8.2.

Fig. 8.2 Digital twin for liquid piston compressor with one cylinder for multiple strokes
The role of the relief valve [RV010] is to limit the upstream pressure
within a hydraulic circuit and thus protect hydraulic components from over
pressure. This component is also known as pressure limiting valve,
maximum-pressure valve, or safety valve [5].

8.3.2 Submodels and Parameters

To improve the function of the model, set the simulation time to 10 [s]
according to Tables 8.1 and 8.2 in this submodel, refer to the definitions
specific to this library in the help menu. The fluid [FP04] inside the
hydraulic cylinder is an hydraulic oil.
Table 8.1 Parameters for simulation model of the liquid piston compressor with one cylinder for one
stroke (part 1)

Item Parameter
PNCH022 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
Gas type index = 1
Volume = 0.1 [L]
Thermal exchange coefficient = 0 [J/m**2/K/s]
Thermal exchange area = 0.1 [m**2]
External temperature = 293.15 [K]
PNCV010 Gas type index = 1
Check valve cracking pressure = 0.2 [bar]
Check valve mass flow rate pressure gradient = 10 [g/s/bar]
Valve hysteresis = 0 [bar]
PNGD00 Gas type index = 1
Fluid definition = air
Properties definition = semi perfect
PNAS001 Temperature at port 1 = 293.15 [K]
Item Parameter
PNJ002 Temperature at port 1 = 293.15 [K]
Pressure at port 1 = 1.013 [barA]
Gas type index = 1
Model = with thermal exchange
Use initial displacement = yes []
Displacement of piston = 0.5 [m]
Piston diameter = 25 [mm]
Rod diameter = 12 [mm]
Length of stroke = 0.5 [m]
Dead volume at port 1 end = 50 [cm**3]
Thermal exchange coefficient = 0 [J/m**2/K/s]
External temperature = 293.15 [K]
Viscous friction coefficient = 0 [N/(m/s)]
Leakage coefficient = 0 [g/s/barA]
Spring preload = 0 [N]
Spring rate = 1 [N/m]
Pressure in rod chamber = 1.013 [barA]
Spring rate at endstops = 100000 [N/mm]
Damping coefficient at endstops = 100000 [N/(m/s)]
Deformation on endstops at which damping rate = 0.001 [mm]

Table 8.2 Parameters for simulation model of the liquid piston compressor with one cylinder for one
stroke (part 2)

Item Parameter
RV010 Index of hydraulic fluid = 0
Relief valve cracking pressure = 5 [bar]
Relief valve flow rate pressure gradient = 500 [L/min/bar]
Valve hysteresis = 0 [bar]
Item Parameter
HJ0023 Pressure at port 1 = 1.013 [bar]
Index of hydraulic fluid = 0
Use initial displacement = yes
Displacement of piston = 0 [m]
Piston diameter = 25 [mm]
Rod diameter = 12 [mm]
Length of stroke = 0.5 [m]
Dead volume at port 1 end = 50 [cm**3]
Viscous friction coefficient = 0 [N/(m/s)]
Leakage coefficient = 0 [L/min/bar]
Spring preload = 0 [N]
Spring rate = 1 [N/m]
Pressure in rod chamber = 1 [bar]
Spring rate at endstops = 100000 [N/m]
Damping coefficient on endstops = 10 [N/(m/s)]
Deformation on endstops at which damping rate is fully effective = 0.001 [mm]
PU001 Index of hydraulic fluid = 0
Pump displacement = 2 [cc/rev]
Typical pump speed = 1500 [rev/min]
PM000 Shaft speed = 1500 [rev/min]
FP04 Type of fluid properties = elementary
Index of hydraulic fluid = 0
Temperature = 40 [degC]
Name of fluid = unnamed fluid
Density = 850 [kg/m**3]
Bulk modulus = 17000 [bar]
Absolute viscosity = 51 [cP]
Item Parameter
MECMAS21 Velocity at port 1 = 0 [m/s]
Displacement at port 1 = 0 [m]
Use friction = no
Endstop type = none
Mass = 0.1 [kg]
Inclination (+90 port 1 lowest, −90 port 1 highest) = 0 [degree]
TK000 Tank pressure = 0 [bar]
The existing model must now be parameterized accordingly. In the
following only those parameters are shown, which deviate from the basic
settings. The simulation times for the simulation model are given in Table
8.3.
Table 8.3 Simulation time for the simulation model of the liquid piston compressor

Simulation settings
Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]

8.3.3 Results and Analysis

To check, if the twin of the first liquid piston compressor is working, we
plot the movement of the mass element between the pneumatic and the
hydraulic chamber. The result is presented in Fig. 8.3, and we can see that
the hydraulic pressure-relief valve [RV010] is working and switches and the
mass reaches its maximum position at about 4.5 [s].
Fig. 8.3 Displacement of the mass element between hydraulic and pneumatic pistons
From the results, the pressure curves inside the hydraulic cylinder, the
pneumatic cylinder, and the pressure accumulator can be evaluated, among
other things. The pressure curves in the hydraulic cylinder and the
pneumatic cylinder are shown in Fig. 8.4.
Fig. 8.4 Diagram of pressure curve in the hydraulic, pneumatic cylinder
It can be clearly seen that the compression process takes place exactly
once. When the mass reaches its maximum position, the cylinders are not
able to move further on. So the hydraulic pump increases the pressure in the
hydraulic system very fast. At a pressure of 5 [bar] the pressure-relief valve
opens and the hydraulic oil flows through this valve directly back to the
tank. The pressure curves of the pressure accumulator are shown in Fig. 8.5.
First, the pressure in the pneumatic cylinder is increased by the extension of
the hydraulic cylinder, which corresponds to a retraction of the pneumatic
cylinder, until the accumulator pressure is reached. A slight pressure
fluctuation in the fluid [FP04] of the hydraulic cylinder can be seen.
The valve [RV010] is normally closed. When the pressure drop across
the valve exceeds the relief valve cracking pressure (typically a spring
force), the valve opens and let the fluid flow across so that the pressure drop
gets regulated to the cracking pressure [5]. Subsequently, the pressure in
Fig. 8.5 remains almost constant and the air present is forced into the
pressure accumulator. In the process, the pressure increases both in the
pressure accumulator and in the pneumatic and hydraulic cylinders. When
the hydraulic cylinder is completely extended (end position), the hydraulic
pressure increases like mentioned before until the opening pressure of the
pressure-relief valve is reached and the process is finished.

Fig. 8.5 Diagram of pressure curve in the pressure accumulator chamber

8.4 Liquid Piston Compressor with One Cylinder

and Multiple Strokes
8.4.1 Simulation Model
In order to be able to run through the compression process several times, a
control system for the movement of the piston must be implemented. For
this purpose, a 2/3-way hydraulic valve [HSV23 02], a displacement
sensor [MECDS0B], and a “THRESHOLD” are to be added to the model,
as can be seen in Fig. 8.6. The sensor gives a trigger signal for the lift and
controls the valve.
Fig. 8.6 Digital twin for liquid piston compressor with one cylinder for one stroke
In Fig. 8.6, we can see the digital twin for liquid piston compressor with
one cylinder for multiple strokes with components from the Amesim model.
We improved the model from the further section and added a two-position
three-port hydraulic servo-valve [HSV23 02] to the model. The existing
model must now be parameterized accordingly. In the following only the
parameters of the added components are shown, which differ from the basic
settings to avoid repetition. Only additional parameters will be given in
Table 8.4.

8.4.2 Submodels and Parameters

Most parameters for the digital twin of the liquid piston compressor with
one cylinder and multiple strokes stay the same as in the previous model in
Fig. 8.6 for the system with one stroke. Therefore, we will make Table 8.4
for the parameters a bit shorter, and add only the parameters for the
additional submodels and deviating parameters.
Table 8.4 Additional parameters for liquid piston compressor with one cylinder for multiple strokes

Item Parameter
MECDSS0B Sign convention: positive from = 0
Offset to be subtracted from displacement = 0 [m]
Gain for signal output = 1 [1/m]
TRIG0 Initial output value = low
High input threshold value = 0.499 null
Low input threshold value = 0.001 null
High output value = 0 null
Low output value = 1 null
HSV23 02 Index of hydraulic fluid = 0
Valve rated current = 1 [mA]
Valve natural frequency = 80 [Hz]
Valve damping ratio = 0.8 null
Ports P to A characteristic flow rate at maximum opening = 100 [L/min]
Ports P to A corresponding pressure drop = 0.01 [bar]
Ports P to A critical flow number (laminar/turbulent) = 100 null
Ports A to T flow rate at maximum opening = 100 [L/min]
Ports A to T corresponding pressure drop = 0.01 [bar]
Ports A to T critical flow number (laminar/turbulent) = 1000 null
Time steps and simulation time for the liquid piston compressor with
multiple stroke are given in Table 8.5.
Table 8.5 Simulation time for the liquid piston compressor with multiple strokes

Simulation settings
Start time = 0 [s]
Final time = 10 [s]
Print interval = 0.01 [s]

8.4.3 Results and Analysis

The travels of the hydraulic and pneumatic cylinders show that they both
work in opposite directions. The signal from the displacement sensor is
used to switch over the hydraulic valve via the “THRESHOLD” when the
hydraulic cylinder is almost completely extended. In Fig. 8.7, we can see
that movement of the mass occurs two times according to the trigger of the
signal [TRIG0] and the displacement of the mass element between
hydraulic and pneumatic pistons with multiple strokes.

Fig. 8.7 Displacement of the mass element between hydraulic and pneumatic pistons with multiple
strokes
In Fig. 8.9, the pressure in the pneumatic and hydraulic cylinder is
shown. The hydraulic cylinder after the first stroke retracts again due to the
residual pressure in the pneumatic cylinder and the spring force. The
pneumatic cylinder draws in new air.
Fig. 8.8 Digital twin for liquid piston compressor with two cylinder and multiple strokes
When the hydraulic cylinder is almost completely retracted, the process
is repeated. With each conveying process, the pressure in the pressure
accumulator increases step by step. It can be seen that pressure fluctuations
inside the hydraulic fluid are rather high. In technical applications, this
behavior should normally be avoided (see Fig. 8.10).

8.5 Liquid Piston Compressor with Two Cylinder

and Multiple Strokes
8.5.1 Simulation Model
In order to prevent the system from not compressing air during the
emptying of the hydraulic cylinder, i.e. not working, so to speak, the system
can be extended by another fluid piston. This configuration is shown in
Fig. 8.8. The two fluid pistons then work alternately, each using the
hydraulic oil of the cylinder that is just emptying to compress the air. For
this purpose, the model must be appropriately planed with two cylinders.

Fig. 8.9 Pressure curves in hydraulic, pneumatic cylinder with multiple stroke and high oscillations
Fig. 8.10 Pressure curves in the pressure accumulator chamber with multiple strokes
Now, we added a two-position four-port hydraulic valve [HSV24 02]
to the model to switch between the two different hydraulic cylinders. This
valve is controlled also by the trigger signal [TRIG0].

8.5.2 Submodels and Parameters

Additional parameters and data for the new two-position four-port hydraulic
servo-valve [HSV24 02] were sampled in Table 8.6. All other parameters
are similar to the model in the previous section.
Table 8.6 Additional parameters for liquid piston compressor with one cylinder for multiple strokes
Item Parameter
HSV24 02 Index of hydraulic fluid = 0
Valve rated current = 1 [mA]
Valve natural frequency = 80 [Hz]
Valve damping ratio = 0.8 null
Ports P to A characteristic flow rate at maximum opening = 100 [L/min]
Ports P to A corresponding pressure drop = 0.1 [bar]
Ports P to A critical flow number (laminar/turbulent) = 1000 null
Ports B to T characteristic flow rate at maximum opening = 100 [L/min]
Ports B to T corresponding pressure drop = 0.1 [bar]
Ports B to T critical flow number (laminar/turbulent) = 1000 null
Ports P to B characteristic flow rate at maximum opening = 100 [L/min]
Ports P to B corresponding pressure drop = 0.1 [bar]
Ports P to B critical flow number (laminar/turbulent) = 1000 null
Ports A to T characteristic flow rate at max. opening = 100 [L/min]
Ports A to T corresponding pressure drop = 0.1 [bar]
Ports A to T critical flow number (laminar/turbulent) = 1000 null
Fig. 8.11 Three displacements of the mass elements of the two pistons of the LPC

8.5.3 Results and Analysis

Finally, this model is to be extended according to the basic scheme in
Fig. 8.11. For this purpose, it was recorded how the cylinders move in both
pistons. It can be seen clearly in the diagram that the movement of the right
piston is inverse to the movement of the left piston. When the right piston
moves up, the left piston moves down. This is completely plausible. Also,
the motion is in the same range because both pistons are of the same size. It
could be investigated how the movement of the pistons changes if we
choose different volumes.

Important: Comparing different variants of the same technical

application is an effective method for selecting the best solution. Even
though the absolute results may be somewhat different when compared
with measurements. The decision for a favored variant might be
possible.
Fig. 8.12 Pressure curve in hydraulic, pneumatic cylinder with multiple strokes

Figure 8.12 shows the pressure curves of the pneumatic actuator on the
left and the pressure curve of the pneumatic actuator on the right. It should
be mentioned that we have not visualized the pressure variations in the
hydraulic chamber as in the previous section, but they are comparable. The
pressure in the chamber of the pressure accumulator is shown in Fig. 8.13.
It can be seen that there are many more stages forming the stepwise
pressure rise in the pressure accumulator chamber.
Fig. 8.13 Pressure curve in the pressure accumulator chamber with multiple strokes
We want to analyze our complex technical system and have a look at
some problems to improve the model. You can modify the appropriate
parameters in the model and try out what’s happening.

Problems

8.1 Change the size of the pressure accumulator from 0.1 to 10 [L]. What
happens?

8.2 Change the trigger signal of the sensor and try to reduce the
fluctuations inside the hydraulic chamber.

8.3 Add an additional third hydraulic and pneumatic actuator and repeat
the simulations.

8.4 Examine the temperature of the gas inside the pneumatic accumulator.
Is it adiabatic? How can you change the model to reduce the temperature
inside the accumulator?
8.5 Think about a new component to store the heat energy of the gas in
the accumulator.

8.6 Add a hydraulic chamber in the back flow area between the 2/4-way
hydraulic valve and the pump.

References
1. Staudacher, T., von Roon, S., Vogler, G.: Energy storage - status, perspectives and economic
viability; study summary report; client boarfeleven; performed by Forschungsstelle für
Energiewirtschaft e.V., 03.2009
2.
German Federal Ministry for Economic Affairs and Energy (The energy of the future, 2021): 8th
Monitoring Report on the Energy Transition – Reporting years 2018 and 2019, Access: 2021-05-
09. https://​www.​bmwi.​de/​Redaktion/​DE/​Publikationen/​Energie/​achter-monitoring-bericht-
energie-der-zukunft.​html
3.
Grün, S., Hübner, D.H., Ortwig, H.H., Rückert, F.U.: Investigation and optimization of a three
stage inter-cooled piston compressor for an energy storage system with pressurized air. In:
Proceedings of the Interdisciplinary Conference on Mechanics, Computers and Electrics
(ICMECE 2022), 6–7 October 2022, Barcelona, Spain
4.
Grün, S., Hübner, D.H., D., Molter, J.: Investigation and simulation based optimization of an
energy storage system with pressurized air. Arch. Thermodyn. 42(4), 183–200 (2021). https://​doi.​
org/​10.​24425/​ather.​2021.​139658. https://​journals.​pan.​pl/​Content/​122213/​PDF/​art11_​corr.​pdf
5.
Siemens Digital Industries Software: Simcenter Amesim Student Edition (2020)

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_9

9. Digital Twins and Artificial Intelligence

Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

9.1 Neural Networks in Nature

We want to give an example how one can use an artificial neural network
for the physical libraries described before. Neural networks can be used
widely in different fields of business administration, economics, science,
and technology. They were built similar to biological neural networks of
animals. In the following, we will see that our brain or the brain of any
other animal in a certain form performs its services with the help of matrix
calculation. For this some basics have been formulated by Spitzer [1].

Important: Artificial intelligence (AI) can be used to reproduce results

from digital twins much faster, but the quality of the results could be
weaker. We can also use AI to control the twins or make results of the
calculations available faster and with less effort. AI can be used in many
different areas of science.

9.2 Neural Networks and Digital Twins

Sensory cells in the eye, ear, nose, skin, etc. react specifically to physical or
chemical stimuli and produce output by discharging a binary electrical
signal. The so-called action potential. We can describe this in mathematics
with the coded 0 and 1. This signal is transmitted via the axon of the
sensory cell and usually reaches a large number of other neurons. The nerve
cells, which further process the corresponding input. In our brain, each
neuron receives impulses from about 10.000 other neurons in this way. An
input cell is connected to another nerve cell and, given sufficient stimulus,
sends an electrical impulse along the connecting nerve fiber.
Fig. 9.1 Artificial neural networks are used to collaborate with digital twins and predict results
The AI can also collaborate with complex networks of digital twins and
predict new results. Often the predictions are much faster than with a
complex system of digital twins, but the quality can be weaker. To produce
data for such neural networks, you also need, for example, the pneumatic or
thermal library as shown in Fig. 9.1. However, the current does not flow
directly into a nerve cell, but into a synapse, which usually docks with so-
called dendrites or with the cell body of N. The synapse then generates
more or less neurotransmitter molecules, e.g. glutamate, depending on its
own strength. The synapse then generates more or less neurotransmitter
molecules like glutamate depending on its own strength [1].
Fig. 9.2 Creation of an artificial neural network with Simcenter Amesim based on signal library
The synapse then generates more or less neurotransmitter molecules,
depending on its own strength, which pass over to the nerve cell. The
synapse then generates an impulse, which passes over to the nerve cell,
depending on its own strength. The synapse then generates an impulse when
nerve cell receives an impulse. The nerve cell in turn generates an impulse
when a sufficient number of transmitter molecules arrive. An important role
is played by the fact that very many nerve fibers end and therefore receive
transmitter molecules from many synapses simultaneously.
This behavior has to be transferred to a mathematical formulation. It
depends on the synapse whether it emits inhibitory or excitation transmitter
molecules. The synapse thus multiplies the incoming signal (0 or 1) by its
own strength, which can be modeled as a number in the interval [−1, 1].
The neuron sums all the values obtained in this way and fires when the
summed value exceeds a critical threshold characteristic of the neuron.
Creation of an artificial neural network (ANN) is simple in Simcenter
Amesim, as shown in Fig. 9.2. A ready-to-use model is already included.
The [DYNNNFF01] submodel allows evaluating a restricted feed-forward
neural network model having one hidden layer.
Replacing some Simcenter Amesim physical components by a feed-
forward artificial neural network might be a source of undesired algebraic
loops since all the dynamics of the original model have been removed.
Indeed, such a neural network establishes a direct algebraic relationship
between each output and all the inputs. If necessary, algebraic loops can be
broken by placing some first-order lags or some discrete delays on some of
the neural network’s outputs. Beware that time constants need to be
carefully set, depending on the dynamics of the system from which a
surrogate model is being created.
The icon of [DYNNNFF01] submodel is a dynamic icon. When placing
it on the sketch, the user must indicate the number of inputs and the number
of the outputs of the neural network model. In addition, the number of
neurons used for the training process under MATLAB or an external
program which could be provided on our homepage needs to be provided
[4]. A backpropagating neural network is a network of neurons, the simplest
unit of this technique. A neuron receives inputs, changes its internal state
(activation state), and produces outputs in accordance with inputs and
activation states. The connection of these elementary units forms the
network, connecting the outputs of certain neurons with the inputs of other
ones [3].
Fig. 9.3 Adding data elements to the artificial neural network based on signal library
To start a first test with the neural network data for learning is needed.
In Fig. 9.3, we connect our first neural network to input data and output
data. We have to make sure that we have two columns of input data for the
network and one column of output data as defined in Fig. 9.2. We will come
back to this setup later. Firstly, we want to explain how a neural network
works and how learning can train the brain.

9.3 The Artificial Frog Model

Ever since Alessandro Volta used a frog to demonstrate the influence of
electric current, frogs have been widely used for scientific experiments.
How a frog processes what it sees? We will discuss a simple frog example
for a neural network. Think about a frog whose eye consists of three photo
receptors which is shown in Fig. 9.4. Thus, inputs of the network are
propagated to the network’s output through the neurons, following weighted
connections. When the direction of propagated signals goes always from
inputs to outputs, i.e. signals are never propagated backwards, the network
is called a feed-forward neural network. The weights and activation
functions are modified during the training process in order to get values that
provide the best outputs. To train a neural network, a learning algorithm is
used to minimize a related cost function. For example, mean least squares
can be used between the desired output and the predicted one [3].

Fig. 9.4 Photoreceptors of a frog are now connected to three neurons that initiate an action when
they are sufficiently excited [2]
There are three events that are important for the frog. The frog sees a
stork, a fly, or just the blue sky. Each time, the photoreceptors are excited in
a characteristic way. The photoreceptors are now connected to three
neurons that initiate an action when they are sufficiently excited.

9.3.1 Submodels and Parameters

Let us consider a simple test case for the frog model [2]. We imagine a
simple living creature, e.g. like a frog in a pond. How can a frog processes
what it sees? If the stork is detected like excitation in pattern A, flight is
initiated excitation of nerve cell 1. For the frog the fly is the prey. If the fly
is sighted (B), the frog sticks out its tongue to catch the fly. This means
excitation of nerve cell 3. If the frog sees the blue sky (C), it activates
digestion. Excitation of nerve cell 2. For this event, data can be generated.
In Fig. 9.5, we give you an example data set with 10 hidden neurons for this
case.

Fig. 9.5 Definition file frog example.net for the frog with training and test data
Fig. 9.6 Enter the input data for the experience of the frog’s life and hunting behavior
We now graphically represent the connection between the three visual
cells and the three nerve cells, the connecting nerve fibers and the synapses.
In each of the synapses, their strength is plotted as a value between -1 and
1. Each of the three nerve cells fires when the sum of the transmitters
arriving via the synapses exceeds the measure 0.8 in an interval from 0 to 1
(see Fig. 9.6).
The model [DYNNNFF01] can be compared with the [DYNPSM02].
For now, this process needs to be done using the MathWorks Neural
Network toolbox for MATLAB [5]. It is also a submodel that allows creating
global approximations, and it is especially suitable in presence of non-
smooth or highly non-linear models. The input file listed in Table 9.1 has to
be generated by MATLAB or by our in-house programmed optiflow neural
network tool.
The only drawback of this submodel is that it currently has a
dependency to the commercial code MATLAB which is rather slow for
creating the artificial neural network, but not for evaluating it. Note that,
just as [DYNPSM02], [DYNNNFF01] only uses standard Amesim XY
tables, which is particularly interesting to create a surrogate model that can
be efficiently exported as a Simulink blackbox, as a Functional Mock-up
Unit (FMU), or executed on a real-time target. Examples for the XY-tables
needed are given in Table 9.1.

Important: Artificial intelligence and artificial neural networks are

currently a hot topic and under very fast development. Therefore, the
information in this book might be outdated already. To prevent this we
developed a standalone tool to calculate the weights of the neural
network.

Table 9.1 Data files written by optiflow neural network tool for Simcenter Amesim

Data file Description

B1.data Bias file for the connection of the input layer with hidden layer
B2.data Bias file for the connection of the hidden layer with output layer
W1.data Weighting file for the connections between input and hidden layers
W2.data Weighting file for the connections between hidden and output layers
Scaling in.data File with the input scaling matrix
Scaling out.data File with the output scaling matrix

The icon of [DYNNNFF01] submodel is a dynamic icon. When placing

it on the sketch, the user must indicate the number of inputs and the number
of the outputs of the neural network model. In addition, the number of
neurons used for the training process under MATLAB needs to be provided
or you can use our in-house programmed code [3]. For our frog model, we
should define three inputs and outputs, and 10 hidden neurons. The item
component comprises two ports that are listed in Table 9.2.
Table 9.2 Ports for the neural network model

Port Description
Port 1 Used to supply inputs to the neural network model
Port 2 Provides the responses estimated by the neural network model

9.3.2 Optiflow Neural Network (ONN)

To try out simple neural network on your own and to predict test data
results, we offer you the optiflow neural network (ONN) tool. The tool can
be downloaded under [4]. It is easy to use and free of charge. The user
interface to the tool can be seen in Figs. 9.7 and 9.8. After starting the tool
one can open a net-file to define the neural network. Doing this is quite
simple. An example file is given in Fig. 9.5. For our frog we have to define
the number of input cells ( ), the number of hidden neurons (
), and the number of output neurons ( ).

Fig. 9.7 Software tool optiflow neural network (ONN) can be used to create weights
Fig. 9.8 Testing your neural network of the frog’s example
Under the topic Connectcalls you define how to connect the sensor cells
and neurons against each other. The line (1 3 4 13) means that all three
input cells have one connection to the hidden layer (4 to 13). And all hidden
neurons (4 to 13) have a connection to the output cells (14 to 16). After
loading the example file frog example.net you can start the training of
your frog brain.

Important: Note that some basic knowledge of artificial neural

networks is beneficial before using this component [3]. We provide our
own in-house programmed tool optiflow neural network (ONN).
The tool and an example parameter set with data files for the frog
model can be downloaded at https://​www.​optiflow.​htwsaar.​de/​.
As can be seen in Fig. 9.7 you just have to press the button Start Training.
The tool will start with the training for a certain number of epochs. The
number of epochs can be defined in the GUI together with the random seed
of our random number generator. This seed is used to ensure a reproducible
simulation run.

9.3.3 Results and Analysis

After finishing the training of your frog brain you can test the neural
network. As can be seen in Fig. 9.8, the neural network gives quite good
results for prediction of the frog’s behavior. So we can learn that 10 hidden
neurons might be pretty well to reproduce the frog’s behavior. The data files
that are described in Fig. 9.5 can be written with the menu topic File/Save
Weights. Weights from further simulation runs can be loaded and used for
prediction.

Problems
We want to have a look at some problems that we now can explore with an
artificial neural network. To do this, you try the following things with our
froggy.

9.1 Create a data set for the learning behavior of the frog. It should
contain 100 entries of training data. Verify the results with your own test
data.

9.2 Train your neural network with the frog information data set. And try
to predict the behavior.

9.3 Visualize the test cases in a diagram and compare input and output
data. Discuss the result.

9.4 Try to create a mesh with 20 hidden layers and compare the results of
both meshes.

References
1. Spitzer, M.: Lernen – Gehirnforschung und die Schule des Lebens. Elsevier (2007)
2.
Mathematik für Wirtschaftsingenieure; Skript, Hochschule für Wirtschaft und Technik des
Saarlandes (htw saar) (2020)
3.
LMS Imagine.Lab Amesim Users Guide, Siemens Industry Software S.A.S (2016)
4.
https://​www.​optiflow.​htwsaar.​de/​
5.
https://​www.​mathworks.​com/​help/​matlab/​ref/​rand.​html

OceanofPDF.com
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023
F. U. Rückert et al., Digital Twin Development
https://doi.org/10.1007/978-3-031-25692-9_10

10. Conclusions
Frank U. Rückert1 , Michael Sauer2 , Tuomo Liimatainen3 and
Dirk Hübner4
(1) Fluid Energy Machines, University of Applied Sciences Saarbrücken,
Saarbrücken, Saarland, Germany
(2) Fluid Machinery and Measurement, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany
(3) Mechanical Engineering, LAB University of Applied Sciences,
Lappeenranta, Etelä-Suomi, Finland
(4) Lightweight Construction, University of Applied Sciences
Saarbrücken, Saarbrücken, Saarland, Germany

Frank U. Rückert (Corresponding author)

Email: [email protected]

Michael Sauer
Email: [email protected]

Tuomo Liimatainen
Email: [email protected]

Dirk Hübner
Email: [email protected]

10.1 Overview of This Textbook

At the beginning of this book, we introduce the handling of the program
Simcenter Amesim. It was chosen because it is easy to learn and no deep
mathematical skills or programming knowledge is necessary. It can be used
by students free of charge. We started with the creation of a simple pocket
calculator, with which they can perform first mathematical calculations. In
the individual chapters, digital twins from different, technical fields are
presented and developed. The authors have chosen to divide them into
mechanical, thermal, hydraulic, pneumatic, or electric twins. For example,
simple models of ventilation and tank systems or a solar collector are
modeled. The combination of digital twins with artificial intelligence is also
presented. The combination of digital twins with artificial intelligence is
also presented. In the last chapter, it is shown how a simple artificial neural
network of the behavior of a frog is modeled and how this can be used to
control the digital twins. In the last chapter, it is shown how a simple
artificial neural network of the behavior of a frog is modeled and how this
can be used to control the digital twins.
A complex technical twin in which all the individual libraries have been
combined into one overall model of a new innovation, the liquid piston
compressor (LPC). This invention could be used, after further development,
to store mechanical energy generated by wind or water turbines. The book
closes with a description how to connect the digital twins with artificial
intelligence, e.g. neural networks to control the twins and learn from the
data.
To carry out the exercises and solve the problems, the appropriate
equipment is very important. Each student must have the opportunity to
work independently on a computer. A desire was expressed to provide a
second screen for each student. This will allow one screen to be used to
work with the program and the other to follow the lecture. Switching back
and forth between the program and the online meeting on only one screen
makes it difficult to follow the lecture, the second screen would make this
easier. Also it has to be mentioned that the program has a really good help
dialog with many examples from industry.
So far, exercises are given in the form of tasks from lesson to lesson,
without handing out a proper task template. What is desired is a written
accompaniment to the respective tasks, which have a real reference for
better understanding, as well as a textbook in which some processing steps
can be looked up. Another suggestion for improvement is to set aside time
at the beginning of each lecture hour to discuss final papers and clarify
difficulties.
10.2 What You Can Take Away
The first run-through of the lecture with the program Simcenter Amesim
proved to be still very expandable, since both lecturers and students were
new to using it. Suggestions for improvement are to provide better
equipment for students, a textbook, as well as written task descriptions and
to discuss the tasks at the beginning of each lecture.
Through student evaluation, the critical points can be revised and
eliminated for future lectures. Thus there is a continuous learning process to
improve the handling of the system. If the program Amesim can be taught in
an understandable way, this serves very well as a basis for similar modeling
and simulation programs. Furthermore, the students are introduced to the
practical experience and overlapping topics with other lectures through the
handling. They thus gain a deeper understanding in topics that are otherwise
discussed purely theoretically. But what outcome can we take from this
book?
You will have noticed that in this essential we have tried to introduce the
program at the beginning and only briefly present the examples later on.
We don’t want to try to document all examples completely, but rather to
encourage you to create your own simulation model.
Our goal is to make the reader want to create digital twins of technical
devices and machines himself. The aim is to train the eye for the
essentials. You start with simple models and roughly estimated volumes,
distances, forces, or weights. In the course of the work one goes then
ever further into the detail.
The interested reader, student, engineer, or designer should be able to
create and test first detailed, quantitative estimations and calculations for
his new product before he starts with the complex work of designing and
constructing with common CAD tools. Further, the simulation engineer
can use the twin to search for boundary conditions for more complex
simulations.
Digital twins may be a new phenomenon. However, drawings, sketches,
and mathematical and physical equations have been created for
engineering products in advance of a design for a much longer time.
The combination of digital twins together with artificial intelligence is an
important step. You can learn how to train an artificial neural network
 with a data set and how to use this trained network to predict further
results and compare them with data for testing.

10.3 Teaching Methods and Gamification

The book was written during winter semester 2020 and summer semester
2022 at the University of Applied Sciences Saarbrücken (Germany). During
this time, all lectures had to be held as online classes with Microsoft Teams
due to Corona crisis regulations. Students could install the program Amesim
at home and use it for their homework.
In addition, all exercises were tested in face-to-face summer school
together with students from LAB University in Lappeenranta (Finland). We
tested our textbook in two semesters with students from LAB University
(Finland) in such a way that we first formed teams of two who sketched the
examples one by one and solved the given problems. During these exercises
they learned how to improve the models and at the end of the seminar they
had to create their own complex digital twin together with a 3D
visualization.
The summer school lasts 1 week, during which time not all examples
can be worked on. Students had to create each twin from the illustrations in
the book by themselves using their own computers. During lecture, it was
helpful to form collaborating teams of two or three to help each other.
Initial uncertainties and difficulties hopefully subsided during the review
sessions. At the end of each seminar, all team members created their own
digital twins to a new, previously unknown problem and had to present
them to the audience to pass the exam.
Discussing and presenting the twin in front of the other ones always was
a helpful and interesting situation and often resulted in new ideas and
further improvements of the models.
Showing animations of the digital twin in 3D environment gave a good
overview and was always fun at the end of each presentation, e.g. they had
to build a robot and its 3D model was moving between cones to an exit.
Here all the different libraries of Amesim could be tested. To improve the
teaching method, students were asked for their feedback.
Further improvements of the teaching method with gamification were
done in a 1-week workshop at Universidade de Aveiro (Portugal) in
November 2022. Within the framework of the teaching project e-CLOSE
financed by the European Union (EU), we have developed this form of
seminar into a more self-motivating method [1]. We combined the tasks
from the problem section of each twin with the innovative teaching method
of role-playing. A role-play is a form of experiential learning described by
Russell and Shepherd in 2010 [2]. Students take on assigned roles of a
fictive character and act out those role through a scripted play. The method
should be carried out by two teams with four persons in each team. The two
teams have to work against each other. Each member of the group should
take on a specific role, like an actor on stage.

Important: Communication and new teaching methods are important.

When working with the models, users and engineers should learn how to
create the real applications. But always question how realistic a result is.
Rely on experience, never decide based on math alone, or you will end
up failing.

Then the teams create a technical scenario with the digital twins and
changed the parameters and constraints to produce different results for the
same model. Based on the results, roles and rules need to be clearly defined
in two scripts. As with the hunting scene from the petroglyphs in Bryce
Canyon, the moderator with the antlers handed the two alternative scripts
with the different results to the two teams and gave them the task of
discussing them and trying to convince the other side. Teams could act out a
specific discussion in this safe environment.
Our idea was that this could be a dynamic learning experience for actors
to gain knowledge about physical behavior by immersing them in simulated
real-world problems. But also in industry, such an artificial dispute in a
safe, moderated environment can help teams learn very quickly and focus
on a technical problem. Adrenalin is provided free of charge and no
additional coffee is needed.

10.4 Our Outlook for the Future

For further work, it would be very interesting to combine the digital twins
on one hand with 3D visualization to show the behavior in a much more
practical way and to develop more 3D models for the physical twins. The
second most important step has to be the combination of the digital twins
with artificial intelligence or to depict the physical models to neural
networks. However, it must also be said that artificial neural networks are
definitely not the end of the line in artificial intelligence. Further work will
focus much more on reduced-order models (ROM) that lead to more
sophisticated predictions and better represent the digital twins. We also
want to improve the gamification teaching method role-play and create
scripts with controversy, technical tasks.

10.5 Important: Disclaimer for Our Work

The information contained in this book has been obtained solely from
personal experience. The information provided may not be correct,
complete, or accurate. The authors assume no responsibility for losses or
other related liabilities and make no claim as to the accuracy of the
information contained in this book resulting from its use in any manner, or
as to the infringement of any patent rights which may result.
Likewise, the authors and the publisher do not warrant that the
procedures and software tools described are free of third-party intellectual
property rights. Simcenter Amesim is a registered trademark of Siemens
Industry Software NV. The tools, Blender, ANSYS CFX, ANSYS Chemkin,
ANSYS Fluent, openFOAM, and MATLAB, are also registered trademarks.
The reproduction of common names, trade names, product designations,
etc. in this work does not entitle the user to assume that such names are or
would be considered free within the meaning of trademark protection
legislation, even without special identification. Further information can be
found on the Siemens PLM Software homepage [3].

References
1. https://​www.​ua.​pt/​en/​news/​9/​69847
2.
Russell, C., Shepherd, J.: Online role-play environments for higher education. Br. J. Educ.
Technol., Special Issue: Learning Objects in Progress 41, 6 (2010)
3.
https://​www.​plm.​automation.​siemens.​com/​global/​en/​

OceanofPDF.com