2020 XI International Conference on Electrical Power Drive Systems (ICEPDS), Saint-Petersburg, Russia, October 04-07, 2020

Digital Twin of an Electrical Motor Based on

Empirical Performance Model
Anton Rassõlkin, Viktor Rjabtšikov, Vladimir Kuts, Andriy Partyshev
Toomas Vaimann, Ants Kallaste Department of Mechanical and Industrial Engineering
Department of Electrical Power Engineering and Mechatronics Tallinn University of Technology
Tallinn University of Technology Tallinn, Estonia
Tallinn, Estonia
[email protected]

Abstract— The concept of Digital Twin is creating and implementation of different sensors and data acquisition
maintaining a digital representation of the real physical entity platforms. The service entity mainly includes regulation for
and supporting its performance utilizing simulation and both virtual and physical entities and may carry several sub-
optimization tools, which are fed with real data. Development services, such as maintenance and diagnostic, energy
and implementation of Digital Twin technology is a hot topic in optimization, path planning, etc. The connection entity
many industry-oriented research projects. However, more usually defines the interaction between other entities. More
detailed studies are needed on design methods for Digital studies are needed on design methods for DT, which allows
Twins, which allows full synchronization and connectivity full synchronization and connectivity between virtual and
between virtual and real environments. This paper presents a
real environments [3].
development case study of Digital Twin for an electric motor
based on an empirical performance model. The paper discusses A. Rasheed et al. [4] give a taste of what a state-of-the-art
the way how the data required for the Digital Twin were DT of an offshore oil platform looks like. The DT at the
obtained and present a detailed structural description of the example study is continuously updated with sensors data in
virtual entity. Unity3D is used for physics simulations and near real-time and can be supported by synthetic data
visualization of the Digital Twin, because of simple accessible generated from a virtual entity. In this case, DT does not only
and meticulously detailed documentation highly useful for give real physical entity real-time information for more
rapid prototyping in early stages.
precise control and decision making but may serve as
Keywords— Motor drives, Model-driven development,
prediction makers about how the asset will evolve or behave
Industry applications in the future. A perfect DT will be identical to the real
physical entity in terms of appearance and behavior with the
I. INTRODUCTION extra advantage of making future predictions. The DT
presented in the study also provide the possibility for
The concept of Digital Twin (DT) is creating and
operator or scholar to interact with the physical entity using
maintaining a digital representation of the real physical entity
an avatar.
and supporting its performance through simulation and
optimization tools, which are fed with real and updated data. To create an industrial avatar in Mixed Reality (XR)
DT is a part of the digitalization and simulation pillar of application with realistic physics as a part of a DT, various
Industry 4.0 paradigm [1]. The application of Industry 4.0 software packages can be used. For it, game engines such as
technologies in many sectors has been barely examined yet Unity3D, UnrealEngine, and CryEngine are popular tools
since the variation of sub-processes and manufacturing used for creating industrial simulations with real-time
technologies results in a large number of potential graphics requirements. For simulations that are more precise
combinations. DT technology may present a valuable tool for or mathematical or realistic task, other toolkits can be used as
the rapid integration of Industry 4.0 in today's technologies. an addition on separately, as MatLab, Visual Components
DT covers three components: knowledge content, effect, and Arena, Technomatrix.
functionality, and application domain [2]; for utilization of a
DT technology, the relations of the three components must There are some particular reasons for using game engines
be recognized, and the gaps remaining for exploration as visualization output tools:
identified. 1. The easiest way to implement graphical
DT consist of three main components – the physical applications and make an animation from
entity in the real world, its virtual entity (or virtual model(s)), external commands or internal algorithms.
and the data, furthermore services and connections are 2. Out-of-the-box support of existing XR hardware
presented as separate entities. Generally, a real physical and embedded connectivity methods for various
entity consists of assorted subsystems and sensory devices; protocols.
however, for the establishment of a virtual entity, it is
possible to use already collected data. The virtual entity 3. Community support and large knowledge bases
subsists of one or several models according to the DT both from consumers, experts in the fields, and
application; it can be the spatial, behavior, thermal, rule, or various field researchers.
another model. The DT data contain various sets of data 4. Flexible to work with real physical external
from the real physical entity that can be obtained by the equipment by telemetry connection means.

The research has been supported by the Estonian Research Council

under grant PSG453 "Digital twin for propulsion drive of an autonomous
electric vehicle."

978-1-7281-6538-7/20/$31.00 ©2020 IEEE

 Industrial and research applications created using those Motor Efficiency
engines are called "serious games."[5] Each of those visual 100
60
engines has its advantages and drawbacks – in the end, the 95

79
81
selection is made according to specific task requirements and

.0
.4

42
83 022 90

4
user competence level. 50
86
.76
2
.12
8 17 18 85
Except for game engines, there exist other software 88
.4
40
products used for industrial simulation applications: 88
.48
80
17
• MATLAB – accessible simulation environment and 30
75

programming language that can be used for running 70

physical simulations connected to game engine 20 65
code; 88.4817
86.1218
86.1218 60
• Computer-aided design (CAD) and 3D modeling 10
74.3
7619.92628
67 .6 0332 79.0
76.6
424
826 74.3228
64
62 .2
.8 43
83 4 71.963
software Solidworks, ANSYS, COMSOL 60
57.
55.
53 .52
.16
50.7
48.380
444
.08
46.00
43.64
41.28
38.92
36.565
34.206
31.84665
29.4865
27.1267
24.7669
3
6
38
4
42
46
248
52
54
56
574
9 5.8885
1
3
64.8836
62.5238
60.164
57.8042
55.4444
53.0846
50.7248
48.365
46.0052
43.6454
41.2856
38.9257
36.5659
34.2061
31.8463
29.4865
27.1267
24.7669
22.4071
20.0473
17.6875
15.3277
12.9679
55
22.4071
20.0473
17.6875
15.3277
12.9679
10.608 8 1 3.5287 1.1689 10.6081
8.2483 5.8885
3.5287 -1.190
1.168
Multiphysics, etc. that typically used for physical 8.2 4 -3.5507
-1.1909 9
1.1689 9 50
design and optimization; 400 600 800 1000 1200 1400 1600 1800
Speed
• Visual Components, Technomatrix – the software's Fig. 1. The empirical performance model of induction motor used for
for creating interactive factory layouts and plans; Digital Twin.
Processes visualizations;
The ratio of the motor shaft power ( mech_motor) to the total
• Industrial Robot offline programming software electrical input power ( in_total) is taken as motor efficiency:
packages – latest versions of robot programming
environments by ABB, Yaskawa, Kuka, and other motor = mech_motor/ in_total (1)
companies have simulation support; The graphical representation of an empirical performance
• Open-source software packages as a Robot model of IM is presented in Fig. 1 and rated data of the
Operating System (ROS) embedded with simulation motor is given in Table I.
and visualization tools as Gazebo, Rviz.
TABLE I. RATED DATA OF INDUCTION MOTOR
In the current study, Unity3D is used for physics
simulations and visualization of the DT, because of previous Parameter Unit Value
research stuff experience and high-quality documentation Motor frame size 132 MA
provided by software producer and community. Rated Power kW 7.5
A recent study presents a part of the project that aimed to Rated Current A 22
develop a specialized unsupervised prognosis and control Rated Speed rpm 1460
platform for electric propulsion drive systems performance cos• 0.6
estimation of an autonomous self-driving electric vehicle. Moment of inertia kgm 0.048 2

This goal requires the development of several subtasks, and The empirical performance model characterizes the loss
the related objectives, one of them is to develop physical distribution of IM on speed-torque characteristics. The
models of different energy system components (motors, copper losses are the dominant power losses in any electrical
gearboxes, power converters, etc.) and the related reduced motor [7]. The torque of IM is proportional to the electrical
models of these components (testbed), which will serve to loading and the magnetic loading. To increase the torque
construct the DT of the system [6]. produced by the motor, both (electrical and magnetic)
The paper is structured as follows: Section II reports how loadings must be increased. However, due to the saturation
the empirical performance model was obtained. Section III of the stator iron core in high magnetic flux density, the only
presents the detailed structure description of the Digital Twin constructive way to increase the motor torque is by
virtual entity ant its implementation framework. IN the increasing the electrical loading. That can be achieved by
discussion and conclusion section, the further development increasing the current density in motor windings, which in
and application of presented DT are characterized. turn higher copper losses as well as lower efficiency.
Numerical representation of an IM empirical
II. EMPIRICAL PERFORMANCE MODEL performance model may be used as one of the inputs for
In order to have a reasonable empirical performance building DT to evaluate the IM performance over a given
model, an induction motor (IM) ABB 3GAA132214-ADE speed-torque region.
was studied. The tested electrical motor is driven by an
industrial frequency converter (ABB ACS880) with Direct III. DIGITAL TWIN
Torque Control (DTC) algorithm. Additional frequency A. Structure of the Digital Twin
converter (ABB ACS800) IM setup was used as a load. To
obtain an efficiency map, the load torque is gradually The DT for the IM is set up in the following way: Unity
increased from zero to the calculated during the design stage 3D physics engine is used to simulate fundamental physics
rated load torque in a present ramp mode. All the while connected via Robot Operation System (ROS) bridge,
experiments are performed in a real-time setup, including Linux ROS nodes simulate more advanced electrical
motors and frequency converters. The process of efficiency machine-specific behaviors such as motor efficiency
map obtaining is described in previous research by authors following the efficiency map and motor controller. As shown
[7]. in Fig. 2. The primary benefit of Unity 3D, that it has a very
 Fig. 4. Detailed representation of Digital Twin Physics Engine.

Sensor) takes the angular velocity of the motor shaft,

converts it to feedback data, and sends it to the Linux ROS
Server.
Currently, there are 2 active nodes in the simulation: a
motor controller simulator node and a mechanical simulation
node. The motor controller, has a proportional–integral–
derivative (PID) controller that provides a control signal to
the motor. Mechanical simulation node works with an
efficiency map measured and recorded on a real motor and
applies it to a simulation to emulate proper torques and
Fig. 2. The operational architecture of Digital Twin. power output at certain angular velocity to make DT act just
like real entity motor.
attractive system to work with 3D objects and simulate the
effects of virtual physical forces. These upsides were a B. Graphical representation or spatial model
primaryreason for choosing Unity3D over more commonly The DT model does not undoubtedly mean a spatial or
used in research MATLAB, Matplotlib, or Mathworks on the graphical model. The main accent should be paid to process
recent stage of the project. flow and relations behind it as well as the data entity.
Linux ROS Server sends a User Datagram Protocol However, the application of augmented reality (AR) and
(UDP) command packet via ROS Bridge standard to the virtual reality (VR) tools for distance simulation via DT add
Unity 3D Visualizer. Its main task is to receive a control safety layer and more possibilities for accelerated lifecycle
message from a specified IP address in a standardized ROS tests, applications in hazardous environments, and remote
Bridge message format that is being published by the Linux work maintenance overall. Implementation of a virtual visual
ROS Server and control the 3D models to display the model can prevent plenty number of mistakes and errors
behavior of the motor as it would be in real life. It is also without any performance or economic loss. Besides that,
responsible for sending feedback to the Linux ROS Server such a model can be a valuable tool for the education of
for processing, just like a real encoder would send updates on technical staff and in universities. In the spatial model, the
the motor rotation rate to a motor controller. components of a real physical entity (e.g., windings, rotor,
shaft, encoder, etc.) are constructed as CAD geometric
Linux ROS Server Consists of multiple nodes that models to be assembled in the virtual entity. The spatial
simulate various aspects (e.g., mechanical, thermal, model used for DT visualization is shown in Fig. 4.
electrical, diagnostic, etc.) of the motor, a single motor
controller simulator node and a Simulation controller node Physics Engine has a ROS Bridge client that receives a
that combines information coming from all other aspect data package from ROS Server. The data in that package is
simulation nodes and feedback from the Unity 3D, process it
and send the data to a visualization client in Unity 3D.
More in the details physics engine of the DT is
represented in Fig. 3. Unity 3D Visualizer message is
received in the form of a UDP package in a custom ROS
Bridge format by the ROS Bridge Client, reserialized into
input variables (Empirically Estimated Velocity and Torque)
of the motor and passed on to the 3D Object Controller as
variables. Object Controller converts data from the empirical
performance model to DT model velocity at which the shaft
of the motor rotates. This data is applied to the motor's 3D
visualization. Then a simulated encoder (so-called Virtual
Fig. 3. The spatial model used for Digital Twin visualization.
extracted by the ROS Bridge client and sent to the 3D object entity, and supporting device performance using different
controller that recalculates new positions and orientations of simulation and optimization tools are a key goal of many
the simulated motor components. It then updates the 3D research works.
model with new states. This step ensures that a 3D model is
acting like a real motor would from a visual standpoint. Then This paper presents a development case of DT for an
model takes into account any other physical forces applied to electric motor based on an empirical performance model.
the motor, such as friction losses and moment of inertia of The paper consider obtained data required for the DT
the motor simulated via angular drag feature in Unity 3D's development. A detailed structural description of the virtual
Rigidbody, simulation and updates simulated encoder values entity based on Unity3D engine is presented.
which are then packed and sent to the ROS Server to act as The main application of the recent DT is a loading motor-
feedback for the speed controller. The simulated encoder is drive system for the test bench to estimate the performance
modular and can be replaced by any other necessary of the electric propulsion drive system of an autonomous
feedback simulation. Currently a PID controller is used for electric vehicle. The development and implementation of the
its simplicity and ease of use. concept of DT, will help to provide a brand-new approach
for measurement and estimation of the performance of
IV. DISCUSSION motor-drive systems.
The term Digital Siblings (DS) is also introduced in [4]
and can be considered as copy(ies) of the physical entity REFERENCES
which need not necessarily run in real-time but can be used [1] V. Kuts, M. Sarkans, T. Otto, T. Tähemaa, and Y. Bondarenko,
to test out hypothetical scenarios for maintenance, "Digital Twin: Concept of hybrid programming for industrial
diagnostics, "what if?" analysis and risk assessment. robots – Use case," in ASME International Mechanical
Engineering Congress and Exposition, Proceedings (IMECE),
Reduced models of the physical entities for DT and DS can 2019, vol. 2B-2019, doi: 10.1115/IMECE2019-10583.
be constructed using different model order reduction [2] K. Wang, T. Lee, Y. Hsu, and T. Lee, "Revolution on digital twin
methods. A variety of condition monitoring techniques are technology — a patent research approach," vol. 2, 2020.
available nowadays and may be combined with DT and DS [3] V. Kuts et al., "Synchronizing physical factory and its digital
approaches, e.g. electromagnetic field monitoring, noise and twin through an IIoT middleware: a case study," Proc. Est. Acad.
vibration monitoring, infrared recognition, temperature Sci., vol. 68, pp. 364–370, 2019, doi: 10.3176/proc.2019.4.03.
measurements, radio frequency emission monitoring, [4] A. Rasheed, O. San, and T. Kvamsdal, "Digital twin: Values,
chemical analysis, acoustic noise measurement, motor challenges and enablers from a modeling perspective," IEEE
Access, vol. 8, pp. 21980–22012, 2020, doi:
current signature analysis (MCSA) and most advanced
10.1109/ACCESS.2020.2970143.
artificial intelligence-based techniques, such as fuzzy logic
[5] T. Susi, M. Johannesson, and P. Backlund, "Serious Games – An
and neural networks etc. [8] Overview (Technical Report)," 2007.
Very different software can be combined to achieve the [6] A. Rassõlkin, T. Vaimann, A. Kallaste, and V. Kuts, "Digital
proper sibling. An example of hybrid Finite Element twin for propulsion drive of autonomous electric vehicle," in 60th
International Scientific Conference on Power and Electrical
Analysis (FEA)-Simulink modelling of permanent magnet Engineering of Riga Technical University (RTUCON), 2019.
assisted synchronous reluctance motor is presented by J. [7] A. Rassõlkin et al., "Efficiency Map Comparison of Induction
Pando-Acedo et al. in [9]. Cross-platform software that and Synchronous Reluctance Motors," in 26th International
combines multiphysics graphical applications with powerful Workshop on Electric Drives: Improvement in Efficiency of
pre-processing, solvers, and post-processing capabilities will Electric Drives (IWED), 2019, pp. 4–7, doi:
be preferable for DS and DT creation. Nowadays, some of 10.1109/IWED.2019.8664334.
the optimization framework as Artap [10] are linking [8] B. Asad, T. Vaimann, A. Belahcen, A. Kallaste, A. Rassõlkin,
and M. N. Iqbal, "Broken rotor bar fault detection of the grid and
together different software parts and interfaces for both of inverter-fed induction motor by effective attenuation of the
gradient-free and evolutionary based optimization methods, fundamental component," IET Electric Power Applications, 18-
meta-modeling techniques, surrogate models and FEM Jul-2019.
solvers. [9] J. Pando-Acedo et al., "Hybrid FEA-Simulink Modelling of
Permanent Magnet Assisted Synchronous Reluctance Motor with
That fact that developed DT ROS nodes are modular Unbalanced Magnet Flux," in Proceedings of the 2019 IEEE 12th
makes it very easy to add nodes responsible for International Symposium on Diagnostics for Electrical Machines,
electromagnetic, thermal, diagnostic or any other Power Electronics and Drives, SDEMPED 2019, 2019, pp. 174–
simulations. Different reduced models of the devices running 180, doi: 10.1109/DEMPED.2019.8864925.
parallel and in real-time can be used to assemble a DT. [10] D. Pánek, T. Orosz, and P. Karban, "Artap: Robust Design
Studies shows [11] that model order reduction is a promising Optimization Framework for Engineering Applications," 2019
3rd Int. Conf. Intell. Comput. Data Sci. ICDS 2019, Dec. 2019,
tool for controlling industrial processes, where some of the doi: 10.1109/ICDS47004.2019.8942318.
parameters cannot be measured directly. [11] D. Panek, T. Orosz, P. Kropik, P. Karban, and I. Dolezel,
Diagnostics as a promising application will be considered "Reduced-order model based temperature control of induction
brazing process," in 2019 Electric Power Quality and Supply
as foreground implementation of the DT based on the Reliability Conference and 2019 Symposium on Electrical
modified winding function-based model described in [12]. Engineering and Mechatronics, PQ and SEEM 2019, 2019, doi:
However, 10.1109/PQ.2019.8818256.
[12] B. Asad, T. Vaiman, A. Belahcen, A. Kallaste, A. Rassolkin, and
V. CONCLUSION M. N. Iqbal, "Modified Winding Function-based Model of
Squirrel Cage Induction Motor for Fault Diagnostics," IET Electr.
The concept of DT is gaining popularity nowdays in Power Appl., pp. 1–13, May 2020, doi: 10.1049/iet-
many different industry oriented fields. Creating and epa.2019.1002.
maintaining a digital representation of the real physical