KTH Engineering Sciences
Modelling and simulation of
electromagnetic audible noise generated
by traction motors
Fredrik Botling
Licentiate Thesis
Stockholm, Sweden
Academic thesis with permission by KTH Royal Institute of Technology, Stockholm, to be submitted for public examination for the degree of Li-centiate in Vehicle and Maritime Engineering, Wednesday the 14th of
December, 2016 at 10.00, in E2, Lindstedtsvägen 3 (floor 3), KTH - Royal Institute of Technology, Stockholm, Sweden.
TRITA-AVE 2016:84 ISSN 1651-7660
ISBN 978-91-7729-200-5 c
Fredrik Botling, 2016
Postal address: Visiting address: Contact:
KTH, AVE Teknikringen 8 botling@kth.se SE-100 44 Stockholm
Abstract
An annoying tonal noise is produced by modern electrical trains during acceleration and deceleration. This noise is caused by electromagnetic forces generating structural vibrations, especially from the traction mo-tors. The electromagnetic noise is dominant at low train speeds and af-fects both the passengers on the train and on platforms, as well as people living near the track. The focus on this issue has increased the last years, both regarding legislation, contractual requirements and also because of expectations from citizens and travelers. To be able to design low noise electric drive systems, a thorough understanding of the cause and the possibility to predict the electromagnetic noise is needed. This thesis describes the modelling and simulation of an complete multi-physics real-time environment for prediction and analysis of the electromagnetic noise. The simulation results are validated against measurements of the structural vibration and acoustic response of a real traction motor fed by a power converter running in the entire operational range.
Keywords: Vibro-acoustics, Electromagnetic noise, traction motor, modal analysis, multi-physics
Sammanfattning
Ett irriterande tonalt ljud skapas av moderna elektriska tåg under accel-eration och retardation. Detta ljud orsakas av elektromagnetiska krafter som genererar strukturella vibrationer, speciellt från traktionsmotorerna. Detta elektromagnetiska ljud är dominant vid låga tåghastigheter och påverkar både passagerare på tåget och på perronger, samt för personer som bor i närheten av spåret. Fokus på det här området har ökat de senaste åren, både när det gäller lagstiftning, avtalskrav och på grund av förväntningar från medborgare och resenärer. För att kunna utforma elektriska drivsystem med låg ljudnivå, krävs en grundlig förståelse av orsaken av, och möjligheten att prediktera, de elektromagnetiska lju-det. Denna avhandling beskriver modellering och simulering av en komplett multifysik realtidsmiljö för prediktion och analys av elektro-magnetiskt ljud. Simuleringsresultaten har validerats mot mätningar av strukturella vibrationer samt akustisk respons från en riktig trak-tionsmotor som matats från en strömriktare som körts i dess hela arbet-sområde.
Nyckelord: Vibroakustik, Elektromagnetiskt ljud, traktionsmotor, mod-alanalys, multifysik
Acknowledgements
This work is supported by the KTH Railway group at the Royal Institute of Technology in Stockholm, Sweden, and by Propulsion and Controls, at Bombardier Transportation AB in Västerås, Sweden.
I would like to thank my PhD supervisors, Ines Lopez Arteaga and Siv Leth, at the Marcus Wallenberg Laboratory for Sound and Vibra-tion Research (MWL), for all their support and help. I also would like to thank everyone at the Centre of Competence Acoustics and Vibration at Bombardier Transportation AB in Västerås, for help and support with general topics related to acoustics and especially the art of performing acoustic measurements.
Last but not least, I thank my family for all supporting during this time. Thanks Oscar and Ulrika, I love you!
Fredrik Botling
Dissertation
This thesis consists of two parts: The first part gives an overview of the research area and work performed. The second part contains the following research papers (A-B):
Paper A
F. Botling, I. Lopez Arteaga and S. Leth, Combined Experimental and Ana-lytical Vibro-Acoustic Model of an Electrical Motor, Submitted to the Journal of Experimental Mechanics, (2016).
Paper B
F. Botling, I. Lopez Arteaga and S. Leth, Modelling framework for electro-magnetic noise generation from traction motors, Submitted to the Journal of Notes on Numerical Fluid Mechanics and Multidisciplinary Design, (2016).
Division of work between authors
All work has been performed by F. Botling, with the main supervision performed by Prof. I. Lopez Arteaga and by deputy advisor Adj. Prof. S. Leth.
Publications not included in this thesis
Conference papers:
F. Botling, I. Lopez Arteaga and S. Leth, Vibro-Acoustic Modal Model of a Traction Motor for Railway Applications, Conference Proceedings of the Society for Experimental Mechanics Series, Springer, 2016.
F. Botling, I. Lopez Arteaga and S. Leth, Modelling framework for electro-magnetic noise generation from traction motors, Conference Proceedings of the 12th International Workshop on Railway Noise, IWRN12, 2016.
Contents
I
OVERVIEW
1
1 Introduction 3
1.1 Background . . . 3
1.2 Electro-magnetic noise generation . . . 4
1.3 Modeling of electromagnetic noise generation . . . 8
1.4 Objectives . . . 9 2 Methodology 11 2.1 Converter control . . . 12 2.2 Electrical model . . . 13 2.3 Electromagnetic model . . . 13 2.4 Structural model . . . 16
2.4.1 Experimental modal analysis . . . 16
2.4.2 Finite element model . . . 17
2.4.3 Reduced order model . . . 19
2.5 Acoustic model . . . 20
2.6 Simulation framework . . . 21
3 Simulation and validation results 23 3.1 Simulation and validation based on measured motor cur-rents . . . 24
3.2 Complete real time simulation and validation results . . . 26
3.2.1 Tonal components . . . 26
CONTENTS
4 Conclusions and recommendations 31
4.1 Conclusions . . . 31 4.2 Recomendations and future work . . . 32
5 Summary of appended papers 33
Bibliography 35
II
APPENDED PAPERS A-B
37
Part I
1
Introduction
1.1
Background
The first known electrical locomotive was built in 1837, and the first electric passenger train was presented in 1879 by Siemens (figure 1.1).
Figure 1.1: First electric passenger train, 1879 [siemens.com]
Electromagnetic noise and vibrations from electrical motors has been a topic of interest for a long time. Heubach, studied electrical motors and the electromagnetic effects on noise and vibration as early as 1903 [1]. The increasing numbers of power converters and electric motors in hy-brid and pure electrical cars has raised the awareness and focus on elec-tromagnetic noise and its influence of the sound quality of the vehicles [2]. The same holds true for electrified trains but due to higher elec-trical power ratings, the problems associated with electromagnetic noise is even higher for trains compared to electrical cars.
CHAPTER 1. INTRODUCTION
An annoying tonal noise is produced by modern electrical trains dur-ing acceleration and deceleration. This noise is dominant at low train speeds and affects both the passengers on the train and on platforms as well as people living near the track [3].
In Europe mandatory legislation [4] states the maximum overall levels of permitted noise during acceleration from platforms. In these situations the electromagnetic noise is the dominant noise source. These require-ments have become more and more demanding the last years reflecting the expectations from both citizens and travelers.
1.2
Electro-magnetic noise generation
There are three different types of noise sources on a train [5]: mechan-ical, aerodynamic and electro-magnetic. The two first sources are more important at higher vehicle speeds. At low vehicle speeds the electro-magnetic noise is often the dominant noise source [6].
Figure 1.2: Propulsion equipment on a train
Several components on a train (figure 1.2) can generate electromagnetic noise as the main transformers, auxiliary 3-phase transformers, power converters and traction motors. The most critical component for the gen-eration of electromagnetic noise at low vehicle speeds is the traction mo-tor.
1.2. ELECTRO-MAGNETIC NOISE GENERATION
Electromagnetic noise from traction motors originates from different types of forces. Two of the most important ones [7] are the slotting induced magnetic forces and the converter control induced magnetic forces. Slotting harmonics can however be neglected if the combination of rotor and stator slot numbers are selected according to well known rules [8]. The converter control of an electrical motor should realize a voltage source with varying amplitude and frequency. This is done with the so called pulse width modulation technique (PWM). The converter PWM control is creating a wanted waveform by pulsing a constant DC-link voltage.
The power converter uses a constant DC-link voltage Vdc for the
pro-duction of the 3-phase AC voltages (Va, Vb and Vc) fed to the traction
motors as seen in figure 1.3
Figure 1.3: Power converter with switches S1 to S6 for generation of the 3-phase PWM
voltage connected to the traction motor windings
The 3-phase output voltage is created by controlling the power con-verter switches S1 to S6 for generation of pulses with variable
duty-cycle. In figure 1.4 the principle of PWM generation is shown where the blue curve m is the wanted fundamental voltage, and the red PWM pattern is the actual voltage applied. This technique of producing AC voltage is very energy efficient but will cause harmonic components due to the PWM pattern.
CHAPTER 1. INTRODUCTION
Figure 1.4: Principle of pulse with modulation
In figure 1.5 the wanted voltage is depicted with the yellow curve and the output voltage from the converter is shown in green. This voltage will produce the 3-phase motor currents that is needed to produce the wanted torque of the motor. The wanted fundamental motor current is shown by the magenta curve but the actual motor current is shown in blue. It can be seen that the actual motor current consists of the wanted fundamental motor current plus unwanted harmonic compon-ents. These harmonics are the cause for the unwanted PWM generated electromagnetic noise from traction motors.
Figure 1.5: Motor voltage and currents caused by PWM modulation
1.2. ELECTRO-MAGNETIC NOISE GENERATION The PWM induced electromagnetic noise in traction motors is charac-terised by many narrow banded harmonic components spread in a wide frequency range. Figure 1.6 is showing the spectral behavior during an acceleration from stand still upp to 45 km/h. The spectral content vary-ing with the operational conditions as e.g. train speed and load condi-tions.
Figure 1.6: Electromagnetic noise characteristics during a normal acceleration, with many narrow band components in a wide frequency range
CHAPTER 1. INTRODUCTION
1.3
Modeling of electromagnetic noise
generation
The generation of electromagnetic audible noise from traction motors is influenced by effects in many different domains depicted in figure 1.7.
Figure 1.7: Model domains relevant for the generation of electromagnetic noise
To be able to simulate the final electromagnetic acoustic noise, all these domains needs to be modelled. The physical relations between the do-mains also needs to be calculated by the model.
• Converter control: Controls the power converter electronics to gen-erate a pulse with modulated 3-phase converter voltage fed to the electrical windings of the traction motor.
• Electrical model: The converter voltage will cause electrical cur-rents to flow in the 3-phase windings of the traction motor, indu-cing spatially distributed electromagnetic flux waves in the air-gap of the motor.
• Electromagnetic model: The induced rotating flux density waves will cause both tangential and radial force components generating the wanted torque but also unwanted radial force waves acting on the motor.
• Structural model: The rotating radial force waves will excite struc-tural vibrations on the surface of the motor. The response of the structure depends on both space and frequency harmonics of the excited electromagnetic force, and on the corresponding structural behavior of the motor.
• Acoustic model: The radial vibrations on the surface of the mo-tor will cause pressure variations in the surrounding air causing radiated acoustic noise.
1.4. OBJECTIVES Different methods have been used for simulating electromagnetic noise generation as; numerical finite element methods [9, 10], analytical mod-els based on physical relations [11] and experimental techniques based on measurements [12, 13]. Some researchers have used a combination of the mentioned methods to utilize the best parts from the different meth-ods [14, 15, 16].
Previous research on electromagnetic noise generation have often been focused in one or a few of the domains related to the generation of the electromagnetic noise. These models often have the limitation to only be able to simulate one working condition at a time.
If all domains must be simulated during various operational conditions (as the normal condition for a train during acceleration), most of these models can not handle this at all, and the majority of the other methods will take a considerable effort and time to achieve such a test case.
1.4
Objectives
• The first objective with this research is to build a thorough under-standing of the cause of the electromagnetic noise including the effects in all of the domains related to electromagnetic noise gen-eration.
• The second objective is to implement an efficient modeling frame-work for simulation of a complete propulsion system including all the physical domains related to the generation of the electromag-netic noise. The model shall run i real time and shall cover all operational conditions.
• Final objective is to validate the simulation model of the complete propulsion control system against measurement of a real propul-sion system.
2
Methodology
Two different methods for modelling and analysis of the electromag-netic response from a traction motor have been used. The first method, presented in paper A, is based on prediction of the electromagnetic forces in the traction motor by using measured motor currents. The structural response from the motor is created from a reduced order model based on the results from an experimental modal analysis. The coverage of paper A is shown in figure 2.1
Figure 2.1: Model domains covered in paper A
The second method, presented in paper B, is an extension of the work described in Paper A to include simulations of all domains related to the electromagnetic noise generation. This simulation model is implemen-ted in a real time simulation environment making it possible to simulate the vibro-acoustic response of the motor in any operational condition in real time.
Figure 2.2: Model domains covered in paper B
CHAPTER 2. METHODOLOGY
motor fed by a PWM controlled power converter during various op-erational conditions, all the domains involved in the generation of the electromagnetic noise needs to be modeled at the same time.
In this work all the domains except for the converter control is imple-mented in Matlab/Simulink in a real-time simulation environment. The converter control used in the simulator is exactly the same controller (both software and hardware wise) as used on real trains, and the con-troller communicates in real time with the simulation environment. This modeling approach has a number of advantages compared to more traditional ways of simulation of magnetic noise. First, the models run in real time, making it very time efficient. Secondly, the simulation can be done with any operational condition similar to normal running of a train, in contrast to traditional models where only one working condi-tion is modeled during the simulacondi-tion.
During a normal acceleration of a train, most operational conditions rel-evant for the magnetic noise generation are changing, as: speed, torque, magnetic flux, motor currents and PWM switching frequency. These changing conditions directly affects the generation of the electromag-netic forces, structural vibrations and the radiated acoustic sound. The simulation environment will continuously update all parameters in real-time to adapt for the changing operational conditions.
2.1
Converter control
To be able to produce the PWM generated voltage to the traction motor, a control system is needed for controlling the power switches in the con-verter. See explanation i section 1.2.
It is very hard to create a simplified model of the converter control due to its high complexity. A normal traction motor control software con-sists of typically million lines of code.
To be able to simulate the acoustic response during converter control op-eration, the actual converter control computer (as used on trains) can be used in the real-time simulator (RTS) environment. All converter control input and output signals (as analog, digital, PWM pulses etc.) are con-12
2.2. ELECTRICAL MODEL nected to a dSPACE computer running the compiled Matlab/Simulink code in real time.
2.2
Electrical model
The converter control generates a PWM voltage to the traction motor via the power converter. The electrical behavior of the asynchronous trac-tion motor can be modeled as an equivalent electrical circuit diagram as shown in figure 2.3. All important electrical characteristics of the motor can be modeled by this simple and compact representation of the motor.
Figure 2.3: Equivalent circuit diagram of the asynchronous traction motor
The model simulates the conversion of the PWM input voltage into 3-phase motor currents (both in the stator and the rotor) and the corres-ponding rotor and stator fluxes. A more detailed explanation of the model can be read in paper B.
The acoustically most important output signal from the electrical model is the air-gap magnetic flux density that is divided in both the tangential and radial direction. As later explained in section 2.3 these flux com-ponents are generating different force comcom-ponents important for torque production and radial vibrations.
2.3
Electromagnetic model
The electromagnetic forces in the motor are created by the magnetic flux in the motor. This flux is generated by the 3-phase motor currents. To generate the torque of the motor, the magnetic flux needs to rotate in
CHAPTER 2. METHODOLOGY
relation to the speed of the motor. This rotating flux density wave can be expressed as a function in time and space as
b(x, t) =
∞
∑
i=0
Bicos(vix−ωit−ψi), (2.1)
where b is the air gap flux expressed in time t and space x, Biis the flux
amplitude, viis the wave number and ωiis the corresponding frequency
of one particular wave.
The force has both a tangential component generating the wanted torque and a radial unwanted force component causing structural vibrations and noise. The Maxwell stress tensor can be used to calculate the radial magnetic force density in the air gap [6] as
Pr = 1
2µ0Br
2, (2.2)
where Pr is the magnetic force density in radial direction, µ0i the
mag-netic permeability of vacuum and Br is the flux density in radial
direc-tion.
The radial force density in the air-gap can now be calculated by com-bining equation 2.1 and 2.2 as
b(x, t) = 1 2µ0 ∞
∑
i=0 Bicos(vix−ωit−ψi) !2 . (2.3)The rotating force density wave, expressed in equation 2.3, will contain both frequency (ωi) and spacial (vi) harmonic components that will
ex-cite the structure of the motor causing rotating vibration waves on the surface of the motor.
Figure 2.4 shows a force density wave derived from a electromagnetic finite element model. The frequency harmonics corresponds to the char-acteristics in the time domain, and the spatial harmonics corresponds to the distribution of the force wave in the circumferential direction.
2.3. ELECTROMAGNETIC MODEL
Figure 2.4: 3D plot of force density wave
A spatial FFT in the circumferential direction can be calculated for the force density wave in figure 2.4. This gives valuable information of dom-inant space harmonics (viin equation 2.3) of the force density wave.
Figure 2.5: Spatial FFT of force density wave in the circumferential direction
Figure 2.5 shows the spatial FFT in circumferential direction for the force density wave. Dominant spatial harmonic components can be seen for the orders 0, 4 and 8. This information is very useful for creation of a reduced order modal model that will be explained in section 2.4.3.
CHAPTER 2. METHODOLOGY
2.4
Structural model
A model of the structural behavior can be made in different ways. In this work an experimental modal analysis is used. A comparison with two different finite element models is also performed.
The model shall predict the surface vibration levels of the outer shield of the motor. The tested motor is shown in figure 2.6.
Figure 2.6: CAD drawing of traction motor with endplates
2.4.1
Experimental modal analysis
An experimental modal analysis was performed on a traction motor for a metro application. The motor was tested in the Bombardier Power lab in Västerås. The excitation of the motor was done with an impact ham-mer that measured the impulse force to the structure. 72 accelerometers where mounted on the surface of the motor measuring the radial vibra-tions. The frequency response function (FRF) was calculated for all 72 accelerometer positions. The impact force excitation was made at four different positions on the motor surface to be able to resolve coupled coupled modes and a high modal density.
2.4. STRUCTURAL MODEL
Figure 2.7 shows a few of the modes and their corresponding resonance frequency from the experimental modal analysis.
Figure 2.7: Structural modes from experimental modal analysis
2.4.2
Finite element model
The results from the experimental modal analysis were compared to the results from a simple finite element model of a thick wall cylinder (fig-ure 2.8), that is often used by acoustic engineers to simulate the struc-tural behavior of a traction motor.
This simple finite element model can predict the correct mode shapes, but the eigen-frequencies are not correct compared to the experimental results.
CHAPTER 2. METHODOLOGY
Figure 2.8: Structural modes from finite element model without end plates
To investigate the structural influence of end shields on the motor, a second finite element was made, based on the thick walled cylinder but with the extension with end plates (fig 2.9).
Figure 2.9 shows a finite element model with end plates.
Figure 2.9: Structural modes from finite element model without end plates
The results from the finite element model of the thick walled cylinder with end plates corresponds better to the experimental results, espe-cially for the two most important modes m = 0 and 4. This indicates that the end plates of the motor is important for the structural behavior 18
2.4. STRUCTURAL MODEL of the surface of the motor.
2.4.3
Reduced order model
The results from the experimental modal analysis give several hundred eigen-modes within the frequency range of interest. Only a fraction of these modes are relevant for the generation of electromagnetic noise during converter fed operation. Hence it is important to understand which modes are excited by the electromagnetic force. This can be done with knowledge of the spatial characteristics of the electromagnetic force. The equation of motion for a mechanical system can be described by
M¨x(t) + C ˙x(t) + Kx(t) = f(t). (2.4) This describes a coupled system with N degrees of freedom, where M,
Cand K are known as the mass matrix, viscous damping matrix and stiffness matrix. The system can be uncoupled by performing a modal coordinate transformation,
x(t) =Ψξ(t), (2.5)
whereΨ corresponds to the mode shape matrix and ξ is the vector of modal coordinates. Substituting 2.5 in 2.4 and pre-multiplying with the transpose of the mode shape matrixΨTgives
ΨTMΨ ¨ξ(t) + ΨTCΨ ˙ξ(t) + ΨTKΨξ(t) = ΨTf(t). (2.6)
This equation can be written as a system of N uncoupled equations where the right hand side becomes
n(t) =ΨTf(t). (2.7)
If the components of the modal force vector n(t) become zero (or close to zero) for some of the modes, this implies that these modes will not be ex-cited by the applied force and can therefore be excluded from the model. The analysis of the electromagnetic force made in section 2.3 leads to the conclusion that only two modes are excited by the PWM generated electromagnetic force.
CHAPTER 2. METHODOLOGY
Figure 2.10: Transfer function for mode 0 and 4
Figure 2.10 shows the transfer functions from input force density to out-put structural vibration amplitude for the two structural modes relevant for electromagnetic noise generation.
The resonance peaks for the two modes, modelled as the transfer func-tion in figure 2.10, corresponds to the results from the experimental modal analysis shown in figure 2.7. The structural behavior can be con-sidered as broad banded (in comparison to the electromagnetic forces components) with high response even far from the resonance peaks. Even at zero hertz the structural response is not negligible. This means that narrow band force harmonic components, with frequencies far from the structural resonances, will still excite structural vibrations of the mo-tor and contribute to the acoustic noise radiation.
2.5
Acoustic model
The structural vibrations from the surface of the motor will cause acous-tic noise. The radiated acousacous-tic sound power can be calculated based on the vibration velocities vnon the stator using equation 2.8
W = σρ0cShv2ni, (2.8)
where σ is the radiation efficiency, ρ0is the density of air, c is the speed
of sound, S is the area of the vibrating body and v2nis the squared spatial
average of the vibration velocity. The radiation efficiency is assumed to 20
2.6. SIMULATION FRAMEWORK be one for this type of application. The equivalent sound pressure levels (SPL) at some distance from the motor can then be estimated from the calculated acoustic power.
The A-weighted sound pressure level can be calculated in real-time, based on equation 2.8, where the surface velocities are calculated by the structural model domain.
2.6
Simulation framework
The modelling principle is to calculate the vibration response from the motor as the the transfer function of the structure of the motor multi-plied by the electromagnetic force excitation. It is important to note that the exciting electromagnetic force consists of many narrow band har-monic components, that will excite a broad band structural response, resulting in mainly narrow band vibration components, as seen in fig-ure 2.11.
Figure 2.11: Principle of the modeling framework with the spatial harmonics shown in red and blue
The spatially distributed electromagnetic force can be calculated for the two modes 0 and 4 corresponding to the two structural modes shown as red = mode 0, and blue = mode 4.
3
Simulation and
validation results
In this chapter the results from two different simulation methods are presented. The fist method is also presented in paper A, and the second one in paper B.
The first method uses measured motor currents for estimation of the electro magnetic forces created in the air gap in the traction motor. These forces are used to calculate the vibration response based on a structural model of the motor.
This method does not involve the modeling and/or simulation of the converter control, electrical and electromagnetic domains. The method can not simulate other working conditions other than the ones present while the the motor current was measured. However, this method is valuable for investigation of the actual structural response for projects during commissioning or at early prototype testing.
The second method is a complete simulation environment for all do-mains related to electromagnetic noise generation. The model runs in real time and can be used for any operational condition without adapt-ation.
Both methods are evaluated for the six most dominating harmonic com-ponents seen in table 3.1. The dominant frequency and spatial harmonic components can be calculated analytically, as seen in the last column of the table, where fswcorresponds to the applied PWM switching
fre-quency, and fmto the motor rotational frequency. The spatial mode also
CHAPTER 3. SIMULATION AND VALIDATION RESULTS a positive and negative sign.
Table 3.1: Frequencies for the six dominating components
Component Frequency [Hz] Mode (sign = rotation) Analytical exp.
f1 1275 -4 2 fsw−2 fm f2 1340 0 2 fsw f3 1415 4 2 fsw+ 2 fm f4 2615 -4 4 fsw−2 fm f5 2680 0 4 fsw f6 2745 4 4 fsw+ 2 fm
3.1
Simulation and validation based on
measured motor currents
This section describes the simulation and validation results of the simu-lation model using estimated electromagnetic forces based on measured motor currents (Paper A).
Fig. 3.1 shows the simulated (red) and measured (blue) structural vi-bration response on the stator shield in the frequency range from 1260 to 1430 Hz. This frequency range includes the three first dominating fre-quency components f1, f2 and f3 listed in Table 3.1. The peak vibration amplitudes of these three components are shown by blue and red circles.
Figure 3.2 shows the simulated and measured average structural vibra-tion response in the frequency range from 2590 to 2770 Hz. This fre-quency range corresponds to the last three dominating frefre-quency com-ponents f4, f5 and f6 shown in table 3.1.
The simulation results are close to the measurements and validates the models ability to simulate the narrow band harmonic components.
3.1. SIMULATION AND VALIDATION BASED ON MEASURED MOTOR CURRENTS
Figure 3.1: Tonal components 2*fs based on measured motor currents
CHAPTER 3. SIMULATION AND VALIDATION RESULTS
3.2
Complete real time simulation and
validation results
This section describes the simulation and validation results of the com-plete real-time simulation model (Paper B).
3.2.1
Tonal components
In figure 3.3 and 3.4 the simulated vibration responses for mode 0 and 4 for the six frequency components according to table 3.1 is compared with the measured average displacements and 2 times the standard de-viation (STD) of the measured values. The simulated vibration response for mode 0 and 4 are shown by the blue and red curves. The measured average vibration levels for each of the six frequency components are plotted as circles, and 2 times the standard deviation for each compon-ent are marked with diamonds.
Figure 3.3: Tonal components 2*fs
3.2. COMPLETE REAL TIME SIMULATION AND VALIDATION RESULTS
Figure 3.4: Tonal components 4*fs
3.2.2
Total sound pressure
Sound pressure level as function of torque
To investigate the influence of the applied motor torque in relation to the acoustic noise radiation, measurements and simulations were per-formed with constant flux and variable motor torque.
The results shows that the acoustic noise radiation is independent on the applied motor torque, that is in contrast to what is stated in e.g. [17]. However, the measurement results is in line with the simulation results and by theoretical calculations.
Sound pressure level as function of flux
Another investigation was performed with constant low torque (5%) and by variable flux levels. As expected the radiated noise is highly influenced by the motor flux levels despite almost zero applied motor torque. These results are also in line with both the simulation results and by theoretical calculations.
CHAPTER 3. SIMULATION AND VALIDATION RESULTS
Sound pressure level as function of motor speed
The electromagnetic noise behavior during constant acceleration is seen in figure 1.6, and the corresponding test case is one of the most import-ant ones for evaluating electromagnetic noise generation. The fact that the harmonic force components are changing with motor speed means that the different force components will change in respect to the reson-ances for the structure and consequently changing the acoustic response for the different harmonic components.
The best way of evaluating the human annoyance for the electromag-netic noise generation during an acceleration is of course to listen to the actual sound. This is possible to do in the real time simulator making it possible to perform clinical tests of e.g. acceptance and annoyance of different characters of the electromagnetic noise. The second best way to characterize the electromagnetic noise is to plot a narrow band spectral plot see figure 3.5. The normally used 1/3 octave band plot is too rough to be able to catch the characteristics of the sound, and is therefore not recommended for narrow band electromagnetic noise sources.
Figure 3.5: Simulation results of narrow band spectral plot of acoustic response during acceleration
3.2. COMPLETE REAL TIME SIMULATION AND VALIDATION RESULTS One of the most used criteria for noise requirements is the A-weightening total sound pressure level. The simulation model can calculate the A-weightening sound pressure level during stationary or dynamic situ-ations. Figure 3.6 shows the simulated sound pressure level during an acceleration. In figure 3.7 a corresponding measurement is shown.
Figure 3.6: Simulated total sound pressure during acceleration from stand still to 45 km/h
4
Conclusions and
recommendations
4.1
Conclusions
The cause of the electromagnetic noise from traction motors is influ-enced by effects in many different physical domains. To be able to pre-dict and analyze the generated sound, all these domains needs to be simulated. Furthermore, the acoustic response is highly dependent on the operational point of the motor (e.g motor speed, PWM switching frequency, applied torque and magnetic flux level), that implies that the simulation model needs to have the capability to simulate varying oper-ational conditions, as acceleration and deceleration of the motor at dif-ferent levels of motor torque.
If all physical domains in the whole operational range shall be simu-lated with finite element methods, the calculation time for a normal ac-celeration will be in the region of days or weeks. If analytical models in combination with a reduced order experimental model of the structure, implemented in a real-time environment is used, the same test case is done in a couple of seconds.
This thesis describes the modelling and simulation of such a model. The simulation model is capable of calculating the 3-phase motor currents, electromagnetic forces, structural vibrations and the total sound pres-sure level in a real time environment during varying operational condi-tions. The model has been validated against measurement of a traction motor running in various operational conditions. The simulation results were compared to measurements of the same motor that was used for
CHAPTER 4. CONCLUSIONS AND RECOMMENDATIONS the experimentally obtained structural model.
4.2
Recomendations and future work
All parts of the model, except for the experimentally determined struc-tural model can be adapted for new different vehicle projects.
To be able to simulate new motor designs, a finite element model (or de-tailed analytical model) can be used to derive the reduced order modal parameters that can be implemented in the real-time system.
Recommendations for future work is to implement and evaluate the simulation model against other vehicle project with different traction motors.
5
Summary of
appended papers
Paper A - Combined Experimental and Analytical
Vibro-Acoustic Model of an Electrical Motor
F. Botling, I. Lopez Arteaga and S. Leth, Submitted to the Journal of Experi-mental Mechanics, (2016).
This paper describes the implementation and simulation of a model for prediction of the electromagnetic noise generated by traction motors, based on prediction of the electromagnetic forces in the traction motor by using measured motor currents. The structural response from the motor is created from a reduced order model based on the results from an experimental modal analysis.
CHAPTER 5. SUMMARY OF APPENDED PAPERS
Paper B - Modelling framework for electromagnetic noise
generation from traction motors
F. Botling, I. Lopez Arteaga and S. Leth, Submitted to the Journal of Notes on Numerical Fluid Mechanics and Multidisciplinary Design, (2016).
This paper is an extension of Paper A to include simulations of all do-mains related to the electromagnetic noise generation. The motor cur-rents are simulated with electrical models of the traction motor, and by the real converter control unit. This simulation model is implemented in a real time simulation environment making it possible to simulate the vibro-acoustic response of the motor in any operational condition in real time.
Bibliography
[1] Heubach, Der Drehstrommotor, 1903.
[2] D. Lennström, T. Lindbom, A. Nykänen, Prominence of tones in electric vehicle interior noise, in: Proceedings of Internoise, 2013. [3] D. Thompson, Railway Noise and Vibration, Elesvier, 2008.
[4] Technical specification for interoperability relating to the subsys-tem rolling stock noise, Official Journal of the European Union. [5] I. P. Tsoumas, H. Tischmacher, Influence of the inverter’s
modula-tion technique on the audible noise of electric motors, IEEE Trans. Ind. Appl. 50 (1) (2014) 269–278.
[6] J. L. Besnerais, V. Lanfranchi, M. Hecquet, P. Brochet, Characteriz-ation and reduction of audible magnetic noise due to pwm supply in induction machines, IEEE Trans. on Ind. Elec. 57 (4) (2010) 1288– 1295.
[7] O. Bíró David A. Lowther, P. Alotto, M. van der Giet, R. Rothe, K. Hameyer, Asymptotic fourier decomposition of tooth forces in terms of convolved air gap field harmonics for noise diagnosis of electrical machines, COMPEL-The international journal for com-putation and mathematics in electrical and electronic engineering 28 (4) (2009) 804–818.
[8] J. L. Besnerais, V. Lanfranchi, M. Hecquet, P. Brochet, Optimal slot numbers for magnetic noise reduction in variable-speed induction motors, IEEE Transactions on Magnetics 45 (8) (2009) 3131–3136. [9] S. Wiak, B. Weilharter, O. Bíró, S. Rainer, Numerical investigation of
BIBLIOGRAPHY
rotating force waves, COMPEL-The international journal for com-putation and mathematics in electrical and electronic engineering 31 (5) (2012) 1503–1512.
[10] A. Jacobus, W. Muller, Numerical solutions of forces and torques, IEEE Transaction on Magnetics 19 (6) (1983) 2589–2592.
[11] J. L. Besnerais, V. Lanfranchi, M. Hecquet, P. Brochet, G. Friedrich, Prediction of audible magnetic noise radiated by adjustable-speed drive induction machines, IEEE transactions on Industry Applica-tions 46 (4) (2010) 11367–1373.
[12] I. C. Rosca, M. Filip, E. Helerea, Three-phase squirrel-cage induc-tion motor modal analyses. theoretical and experimental aspects, in: International Conference on Optimization of Electrical and Elec-tronic Equipment, 2012.
[13] C. Wang, J. C. S. Lai, Vibration analysis of an induction motor, Journal of Sound and Vibration 224 (4) (1999) 733–756.
[14] G. Qi, J. T. Chen, Z. Q. Zhu, D. Howe, L. B. Zhou, C. L. Gu, Influ-ence of skew and cross-coupling on flux-weakening performance of permanent-magnet brushless ac machines, IEEE Transaction on Magnetics 45 (5) (2009) 2110–2117.
[15] D. Braunisch, B. Ponick, G. Bramerdorfer, Combined analytic-alâ ˘A¸Snumerical noise calculation of electrical machines consider-ing nonsinusoidal mode shapes, IEEE Transaction on Magnetics 49 (4) (2013) 1407(9).
[16] A. Saito, H. Suzuki, M. Kuroishi, H. Nakai, Efficient forced vibra-tion reanalysis method for rotating electric machines, Journal of Sound and Vibration 334 (2015) 388(16).
[17] A. Ruiz-Gonzalez, M. J. Meco-Gutierrez, F. Perez-Hidalgo, F. Vargas-Merino, J. R. Heredia-Larrubia, Reducing acoustic noise radiated by inverter-fed induction motors controlled by a new pwm strategy, IEEE Transactions on Industrial Electronics 57 (1) (2010) 228–236.
