Reduction of Audible Noise of a TractionMotor at PWM Operation

Hanna Amlinger

Licentiate ThesisStockholm, Sweden

2018

Academic thesis with permission by KTH Royal Institute of Techno-logy, Stockholm, to be submitted for public examination for the degreeof Licentiate in Vehicle and Maritime Engineering, Tuesday the 6th ofFebruary, 2018 at 13.00, in E2, Lindstedsvägen 3 (floor 3), KTH - RoyalInstitute of Technology, Stockholm, Sweden.

ISBN 978-91-7729-649-2TRITA-AVE 2017:93ISSN 1651-7660

c© Hanna Amlinger, 2018

Postal address: Visiting address: Contact:KTH, AVE Teknikringen 8 amlinger@kth.seSE-100 44 Stockholm Stockholm

ii

Abstract

A dominating source for the radiated acoustic noise from a train at lowspeeds is the traction motor. This noise originates from electromagneticforces acting on the structure resulting in vibrations on the surface andthus radiated noise. It is often perceived as annoying due to its tonalnature. To achieve a desirable acoustic behavior, and also to meet legalrequirements, it is of great importance to thoroughly understand thegeneration of noise of electromagnetic origin in the motor and also tobe able to control it to a low level.

In this work, experimental tests have been performed on a traction mo-tor operated from pulse width modulated (PWM) converter. A PWMconverter outputs a quasi-sinusoidal voltage created from switched volt-age pulses of different widths. The resulting main vibrations at PWMoperation and their causes have been analyzed. It is concluded that anappropriate selection of the PWM switching frequency, that is the rate atwhich the voltage is switched, is a powerful tool to influence the noise ofelectromagnetic origin. Changing the switching frequency shifts the fre-quencies of the exciting electromagnetic forces. Further experimental in-vestigations show that the trend is that the resulting sound power leveldecreases with increasing switching frequency and eventually the soundpower level reaches an almost constant level. The underlying physicalphenomena for the reduced sound power level is different for differentfrequency ranges. It is proposed that the traction motor, similar to a thinwalled cylindrical structure, shows a constant vibration over force re-sponse above a certain frequency. This is investigated using numericalsimulations of simplified models. Above this certain frequency, wherethe area of high modal density is dominating, the noise reducing effectof further increasing the switching frequency is limited.

Keywords: electromagnetic noise, pulse width modulation (PWM),switching frequency, traction motor

iii

Sammanfattning

Traktionsmotorn är en starkt bidragande källa till ljudet från ett tåg vidlåga hastigheter. Elektromagnetisk krafter i motorn verkar på dess ytaså att denna vibrerar vilket resulterar i utstrålat ljud. Detta ljud uppfat-tas ofta som irriterande eftersom att det är väldigt tonalt. För att minskadet utstrålade ljudet från motorn samt för att uppfylla lagkrav, är detväldigt viktigt att förstå de bakomliggandet orsakerna till hur ljud frånelektromagnetiska krafter uppkommer och även att kunna kontrolleradetta ljud till en låg nivå.

I detta arbete har experiment utförts på en traktionsmotor matad frånen pulsbreddsmodulerad (PWM) frekvensomriktare. En PWM omrikta-re genererar en sinusliknanade spänning utifrån switchade spännings-pulser av olika bredd. De dominerande vibrationerna vid PWM mat-ning och dess bakomliggande orsaker har analyserats. Ett lämpligt valav PWM switchfrekvens, dvs. hur snabbt spänningen switchas, är ett ef-fektivt sätt att påverka ljudet som uppkommer från elektromagnetiskakrafter. När switchfrekevensen ändras, ändras de exciterande elektro-magentiska krafternas frekvens. Ytterligare experimentella mätningarvisar att trenden är att ljudtrycket minskar med ökad switchfrekvensoch till slut når ljudtrycket en nästan konstant nivå. De bakomliggan-de fysikaliska orsakerna till minskningen av ljudtrycket är olika för oli-ka frekvensområden. En hypotes är att traktionsmotorn, likt en tunn-väggig cylinder, har ett konstant förhållande mellan resulterande vibra-tion och exciterande kraft över en viss frekvens. Detta studeras närma-re genom numeriska simuleringar av förenklade modeller. Detta skullebetyda att över denna frekvens, i området som domineras av hög mo-daltäthet, är effekten av att öka switchfrekvensen ytterligare med avsiktatt reducera ljudet begränsad.

Nyckelord: elektromagnetiskt ljud, pulsbreddsmodulering, switchfrek-vens, traktionsmotor

iv

Acknowledgements

First of all, I would like to that The KTH Railway Group and BombardierTransportation for the financial support of this research.

To my supervisors Ines Lopes Arteaga and Siv Leth - thank you both forall your support, encouragement, feedback and discussions during thiswork!

I would also like to thank all present and former colleagues at Bom-bardier: my colleagues at the Control Dynamics department always an-swering all my questions related to converter control and all acousti-cians within the Center of Competence Acoustics helping out with ques-tions related to acoustics and acoustic measurements. A special thanksto Fredrik Botling with whom I shared a large part of this journey forthe good collaboration, to Florenece Meier for all the interesting discus-sions and all your valuable input and feedback; and to Daniel Jansson,Mattias Hill and Tommy Sigemo for all the help with measurements.

Thank you to all colleagues at KTH, even though I have not been at KTHvery often you have always made me feel very welcome!

Last but not least, to all my loving friends and family, without you thiswould never have been possible. Karin, Fredrik and especially Lina:it has been so valuable to share my thoughts with someone who hasalready been on a similar journey, thank you for always being there andlistening and cheering! Christoffer, thank you for your endless love andsupport; and Linnea, there is nobody or nothing that reminds me whatit is really important in life the way you do - I love you!

Hanna AmlingerVästerås, January 2018

v

Dissertation

This thesis consists of two parts: The first part gives an overview ofthe research area and work performed. The second part contains thefollowing research papers (A-B):

Paper A

H. Amlinger, F. Botling, I. L. Arteaga, and S. Leth, “Operational Deflec-tion Shapes of a PWM-fed Traction Motor,” in Rotating Machinery, HybridTest Methods, Vibro-Acoustics & Laser Vibrometry, Volume 8: Proceedings ofthe 34th IMAC, A Conference and Exposition on Structural Dynamics 2016,J. De Clerck and D. S. Epp, Eds. Springer International Publishing, 2017,pp. 209-217.

Paper B

H. Amlinger, I. L. Aretaga, and S. Leth, “Reduction of Radiated AcousticNoise of a Traction Motor at PWM Converter Operation”, submitted toIEEE Transactions on Industry Applications (January 2018).

Division of work between authors

H. Amlinger initiated the direction of and performed the studies, madethe analysis, and produced the papers. I. Lopez Arteaga and S. Lethsupervised the work, discussed ideas and reviewed the work.

vii

Publications not included in this thesis

Conference papers:

H. Amlinger, I. L. Aretaga, and S. Leth, “Impact of PWM SwitchingFrequency on the Radiated Acoustic Noise from a Traction Motor,” inProceedings of the 20th International Conference on Electrical Machines andSystems (ICEMS), 2017, Sydney, Australia. Available: IEEE Xplore.

F. Botling, H. Amlinger, I. L. Arteaga, and S. Leth, “Vibro-Acoustic ModalModel of a Traction Motor for Railway Applications,” in Rotating Ma-chinery, Hybrid Test Methods, Vibro-Acoustics & Laser Vibrometry, Volume8: Proceedings of the 34th IMAC, A Conference and Exposition on StructuralDynamics 2016, J. De Clerck and D. S. Epp, Eds. Springer InternationalPublishing, 2017, pp. 197-208.

viii

Contents

I OVERVIEW 1

1 Introduction 31.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.3 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Acoustic noise of electromagnetic origin in traction motors 92.1 Pulse width modulation . . . . . . . . . . . . . . . . . . . . 92.2 Electromagnetic forces . . . . . . . . . . . . . . . . . . . . . 12

3 Summary of appended papers 193.1 Paper A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.2 Paper B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

4 Conclusions 23

Bibliography 25

II APPENDED PAPERS A-B 29

Part I

OVERVIEW

1 Introduction

1.1 Background

The electrification of the traction systems used for rolling stock emergedat the end of the nineteenth century. However, experiments in elec-tric rail have been traced back to the mid-nineteenth century. The firstknown electric locomotive was built by Robert Davidson and was pow-ered by galvanic cells, i.e. batteries. The first electric passenger trainwas presented by Werner von Siemens in Berlin in 1879. It was run-ning on a 300 meter long circular track and the electricity was suppliedthrough a third, insulated rail situated between the tracks. Much of theearly development of electric locomotion was driven by the increasinguse of tunnels, particularly in urban areas. Smoke from steam locomot-ives was noxious and municipalities tended to prohibit their use withintheir limits. Much has occurred since the first electric locomotive in theend of the nineteenth century and in 2006, 240,000 km (25% by length) ofthe world rail network was electrified and 50% of all rail transport wascarried by electric traction. [1]

Nowadays, the pollution of smoke from steam locomotive is not of-ten on the agenda but one topic that is instead more commonly dis-cussed is the pollution of noise. According to the World Health Organiz-ation (WHO), noise is the second largest environmental cause of healthproblems, after the impact of air quality (particulate matter). Road trafficis the most dominant source of environmental noise in Europe. Expos-ure to road traffic is followed by rail traffic noise, aircraft noise and in-dustrial noise. In 2012, it was reported that nearly 7 million people inEurope was exposed to rail traffic noise above 55 dBlen. Complementingthe reported data with estimations, it was estimated that the total num-

3

CHAPTER 1. INTRODUCTION

ber of people exposed to railway noise was nearly 14 million. 55 dBlenis the EU threshold for excess exposure, indicating a weighted averageduring the day, evening and night [2]. It is reported that exposure tonoise in Europe contributes to about 910 thousand additional prevalentcases of hypertension, 43 thousand hospital admissions per year, and atleast 10 thousand premature deaths per year related to coronary heartdisease and stroke [3].

The attention for noise related to trains and railways has increasedduring the last years. This is due to the enhanced awareness of noise pol-lution from trains and the railway sector and the resulting consequenceson health and wellness. This is also recognized in legalization, in con-tractual requirements from customers on suppliers in the railway in-dustry as well as in the expectations from citizens and travelers. Since2002 there are technical specifications for interoperability (TSI) [4] in-cluding limits for noise emission of new vehicles operating in Europe.The latest revision is from January 2015 with reduced permitted noiselevels compared to the previous versions.

Figure 1.1: Example of typical speed dependency for different noise sources, figure from[5].

There are many types of acoustic noise on a train, e.g. rolling noise,aerodynamic noise and noise related to the traction and auxiliary sys-tems (compressors, transformers, traction motor etc.). The train speedhas a large impact on the radiated noise, and the importance of thedifferent sources adding up to the total noise level differs for different

4

1.1. BACKGROUND

Figure 1.2: Electromagnetic noise characteristics during acceleration.

speeds of the train as is illustrated in Figure 1.1 [5].As can be seen in Figure 1.1, the noise related to the traction system

is dominating at low speeds when the rolling noise is very low. Thisapplies especially to electromagnetic noise from the drive system causedby magnetic forces acting on structures. An important contributor tothe electromagnetic noise on trains is the traction motor. This noise ispassengers on platforms as well as in the compartments exposed to. Thenoise of the traction motor is often perceived as quite annoying due toits tonal nature. A typical frequency spectra during acceleration for amotor is shown in Figure 1.2 and as can be seen it is characterized bynarrow-band harmonics in a wide frequency range.

To be able to reduce the noise of a train and also to meet the TSIrequirements it is of great importance to understand and predict theacoustic noise of electromagnetic origin, for example from the tractionmotor. It is desirable to control the acoustic noise of the traction motorto a low level, preferably already on the design stage. The generationof acoustic noise of electromagnetic origin from the traction motor is aprocess involving several different physical domains. Looking at it froma modeling point of view this process could be divided into five differ-ent sub-domains as shown in Figure 1.3. The converter control controlsthe power converter using so called pulse width modulation (PWM) tooutput a pulse width modulated 3-phase voltage to the traction motor.The voltage results in electrical currents in the 3-phase windings of themotor which is described in the electrical domain. These motor cur-rents induce a magnetic field that generates magnetic forces in the air-

5

CHAPTER 1. INTRODUCTION

Figure 1.3: Generation of acoustic noise of electromagnetic origin.

gap of the motor. The magnetic field is rotating and has both tangentialand radial components. The tangential components result in the desiredtorque of the motor. The radial components are unwanted and act radi-ally on the stator and cause vibrations. The resulting deflections of thestator are described in the structural domain. The response is depend-ent of a match in both frequency and spatial order between the excitingforces and the structural natural modes. The vibrations on the surface ofthe motor generate pressure variations in the air, thus radiated acousticnoise of the motor.

The resulting radiated noise of a traction motor can be controlled inmultiple ways. One possibility is to change the physical structure of themotor and thereby affect the structural response of the motor. Anotherpossibility is to change the excitation of the structure by changing theforces resulting from the converter control and the used PWM method.This thesis focuses on the latter.

1.2 Objectives

The overall objective of this thesis is to gain a thorough understandingon the generation of acoustic noise of electromagnetic origin in a trac-tion motor to be able to reduce the resulting noise to a low level bycontrolling the excitation forces. More specific, the following researchquestions are studied:

• What are the causes of the main vibrations (and hence also radi-ated noise) of a traction motor at PWM operation?

• How is the radiated noise of a traction motor affected by the PWMswitching frequency?

• Is it possible, from simplified modeling of the motor, to determ-ine a engineering optimum value of the switching frequency thatresult in a good acoustic behavior?

6

1.3. THESIS OUTLINE

1.3 Thesis outline

This thesis is organized in two parts. Part I gives some backgroundto the problem studied and a motivation for the research area. Thenthe problem studied and the objectives of this thesis is presented. Inchapter 2, an introduction to pulse width modulation and the generationof acoustic noise of electromagnetic origin in traction motors is given.The contributions of this thesis are briefly presented in Chapter 3 whereeach of the appended papers is summarized. The main conclusions ofthe work are listed in Chapter 4. Part II contains the full contribution ofthis thesis in the form of appended papers.

7

2Acoustic noise ofelectromagneticorigin in tractionmotors

The acoustic characteristics of induction motors and the correspondingsources have been subject to research for a very long time. One of theearliest works on the subject was presented already in 1921, investigat-ing different sources for acoustic noise of electrical machines and howthis noise could be reduced by an improved mechanical design [6]. Thecharacterization of acoustic noise generated by Maxwell forces, i.e. noiseof electromagnetic origin, was later addressed in 1954 [7]. Since theseearly works, the research has continued and also included new areasand aspects as the development of the related electrical machines andtechnology has evolved [8–10].

2.1 Pulse width modulation

The traction motor 3-phase voltage is generated from a voltage sourceconverter having a constant DC-link voltage Udc as input. The outputfrom this converter cannot be varied constantly, it can only be switchedbetween discrete levels. The aim of the converter control is to appropri-ately choose the switching instants. This is done using a pulse widthmodulation (PWM) method with the purpose to select the switchinginstants so that the voltage-time area over time of the switched pulsepattern coincides with that of the desired reference waveform.

9

CHAPTER 2. ACOUSTIC NOISE OF ELECTROMAGNETIC ORIGININ TRACTION MOTORS

A very basic PWM method is the carrier-based method where theswitching instances are determined from intersections between the de-sired reference and a high-frequency carrier. Typically the carrier is atriangular or a sawtooth wave. When the reference is higher than thecarrier, the output is in its high state. Otherwise it is in the low state.This is illustrated in Figure 2.1 where the red line is the reference, theblue line the carrier and the green line the switched output voltage.

0 0.2 0.4 0.6 0.8 1

time [s]

-1

0

1

vo

lta

ge

[V]

0 0.2 0.4 0.6 0.8 1

time [s]

-1

0

1

vo

lta

ge

[V]

Figure 2.1: Principle of pulse width modulation.

The ratio between the switching frequency fc and the fundamentalfrequency of the reference f0 is defined as the pulse number n:

n =fc

f0. (2.1)

The effect of an increased switching frequency, i.e. increased pulse num-ber, is illustrated evaluating the frequency spectra of the pulse widthmodulated signal for two different pulse numbers. The result is plottedin Figure 2.2 and as can be seen, an increased switching frequency shiftsthe harmonic content to higher frequencies. A change of the harmoniccontent of the PWM output voltage also affects the motor currents. Asan example, the motor current for a certain application is plotted forthe different pulse numbers in Figure 2.3. It is evident that the desiredcurrent wave form is less disturbed for the higher pulse number.

An important measure for evaluating the performance of a specificPWM method is the harmonic distortion. This factor quantifies to what

10

2.1. PULSE WIDTH MODULATION

0 20 40 60 80 100

Harmonic order

10-2

10-1

100

Ha

rmo

nic

ma

gn

itu

de

(p.u

.)

(a)

0 20 40 60 80 100

Harmonic order

10-2

10-1

100

Ha

rmo

nic

ma

gn

itu

de

(p.u

.)

(b)

Figure 2.2: Harmonic components for pulse number (a) 9 and (b) 24.

degree undesirable harmonics are created by the used PWM method.The consequences of the harmonics are for example: additional lossesin equipment connected to the converter, torque pulsations in electricalmachines fed by the converter and also acoustic noise. A commonlyused measure for the harmonic distortion is the weighted total harmonicdistortion (WTHD) which is calculated as

WTHD =1

Urms,1

√

√

√

√

∞

∑k=2

(

Urms,k

k

)2

, (2.2)

where Urms,k is the RMS voltage of the kth harmonic and k = 1 is the fun-damental frequency. If the two examples above are again considered, theWTHD is reduced from 10.6% to 3.9% as the pulse number is increasedfrom 9 to 24.

11

CHAPTER 2. ACOUSTIC NOISE OF ELECTROMAGNETIC ORIGININ TRACTION MOTORS

0 0.005 0.01 0.015 0.02

time [s]

-1

-0.5

0

0.5

1

no

rm.

mo

tor

ph

ase

cu

rren

t [-

]

(a)

0 0.005 0.01 0.015 0.02

time [s]

-1

-0.5

0

0.5

1

no

rm.

mo

tor

ph

ase

cu

rren

t [-

]

(b)

Figure 2.3: Example of one motor phase current for pulse number (a) 9 and (b) 24.

There are numerous different PWM methods, except different carrier-based methods there are for example also various programmed methodsand direct methods. For programmed methods, the switching instancesare computed off-line with the objective to achieve a certain perform-ance. For direct methods, the switching instances are determined dir-ectly, typically with the aim to keep a control variable within a specifictolerance band. Despite the used PWM method, the main objective isstill to determine the switching instances to output the desired funda-mental waveform and at the same time preferably minimize unwantedharmonic distortion [11].

2.2 Electromagnetic forces

The electromagnetically generated acoustic spectrum can be determinedfrom the radial component of the air gap Maxwell forces. The electro-magnetic forces result from the interaction of two different flux densitywaves, i.e. a combination of two permeance harmonics and two mag-netomotive (mmf) harmonics. The electromagnetic forces can be classi-fied into three different groups [12]:

1. forces resulting of the interaction of the stator and rotor magneto-motive force (mmf) and permeance slot field harmonics,

2. forces resulting of the interaction of the fundamental stator mmfand a harmonic mmf caused by a PWM time harmonic of the statorcurrent,

12

2.2. ELECTROMAGNETIC FORCES

mode 0 mode 1 mode 2

Figure 2.4: Examples of spatial modes of order 0, 1 and 2.

3. forces resulting of the interaction of the permeance slot field har-monics and and a harmonic mmf caused by a PWM time harmonicof the stator current.

The three different groups of electromagnetic forces have been fur-ther studied, and analytical expressions for the the force harmonics withhighest magnitude have also been derived [13–16].

All electromagnetic forces can be characterized by their frequencyand their mode order, that is the circumferential spatial shape. Examplesof mode order 0, 1 and 2 are shown in Fig. 2.4. The resulting vibrationsare dependent of a match in both frequency and spatial order betweenthe exciting forces and the structure.

The main electromagnetic forces in the first group, the so called slot-ting vibrations, are listed in Table 2.1. Variables kr and ks are positiveintegers from the Fourier series of the permeance distribution, Zr andZs are the number of rotor and stator slots respectively, s is the slip,and p the number of pole pairs. The higher kr and ks are, the lower arethe permeance harmonics, and the lower is the corresponding force har-monic. Therefore, the force harmonics with highest magnitude are givenby kr = ks = 1.

Table 2.1: Frequency and spatial order of the main slotting vibrations

Name Frequency f Spatial order m

F−slot fs(krZr(1 − s)/p − 2) krZr − ksZs − 2p

F0slot fs(krZr(1 − s)/p) krZr − ksZs

F+slot fs(krZr(1 − s)/p + 2) krZr − ksZs + 2p

The characteristics of the forces in the second group, the so calledPWM vibrations are listed in Table 2.2. The fundamental stator fluxdensity is of order p and frequency fs where p is the number of pole-pairs and fs the fundamental stator frequency. The harmonic flux dens-

13

CHAPTER 2. ACOUSTIC NOISE OF ELECTROMAGNETIC ORIGININ TRACTION MOTORS

Table 2.2: Frequency and spatial order of pure PWM force lines

Name Frequency f Spatial order mF−

pwm fs − ηs f sn p − p = 0

F+pwm fs + ηs f s

n p + p = 2p

ity caused by a PWM time harmonic of the stator current is of order pand frequency f s

n where f sn is the frequency of the PWM harmonic. De-

pending on the propagation direction of the harmonic mmf, the interac-tion of these flux density waves results in two groups of force harmon-ics, F−

pwm and F+pwm. As can be seen in Table 2.2, all pure PWM vibrations

have a spatial order of 0 or 2p

Figure 2.5: Relationship between amplitude of harmonics and modulation index, f I is thefundamental stator frequency (figure from [17]).

For PWM with a triangular carrier, the harmonics f sn can be written

as

f sn = n1 fs ± n2 fc, (2.3)

where fc is the PWM switching frequency and n1 and n2 have an oppos-ite parity. The amplitude of the harmonics of the motor current dependson both the carrier type of the PWM as well as the modulation index.The relationship between the amplitude of the current harmonics andthe modulation index for PWM with a triangular carrier is shown in Fig.2.5. As can be seen, the amplitude of 2 fc ± fs is dominating for lowermodulation indexes but at high modulation indexes the amplitude offc ± 2 fs is instead largest [17].

Substituting these frequencies in Table 2.2 gives the frequencies cor-responding to the main peaks in the PWM related vibration response fora triangular carrier signal and their spatial orders as listed in Table 2.3.

14

2.2. ELECTROMAGNETIC FORCES

Table 2.3: Frequencies and spatial orders of main PWM-related vibration peaks for a tri-angular carrier

( f sn = 2 fc ± fs) ( f s

n = 4 fc ± fs) ( f sn = fc ± 2 fs) ( f s

n = 3 fc ± 2 fs) SpatialFrequency f Frequency f Frequency f Frequency f order m

2 fc − 2 fs 4 fc − 2 fs fc + fs 3 fc + fs −2p2 fc 4 fc fc + 3 fs 3 fc + 3 fs 02 fc 4 fc fc − 3 fs 3 fc − 3 fs 0

2 fc + 2 fs 4 fc + 2 fs fc − fs 3 fc − fs 2p

The main electromagnetic forces in the third and last group, the socalled slotting vibrations, are listed in Table 2.4. They have the same spa-tial order as the slotting vibrations but they occur at a higher frequency.

Table 2.4: Frequency and spatial order of slotting PWM vibrations

Frequency f Spatial order m

F−slotpwm fs(krZr(1 − s)/p − 1) − ηs f s

n krZr − ksZs − 2p

Fslotpwm fs(krZr(1 − s)/p − 1) − ηs f sn krZr − ksZs

Fslotpwm fs(krZr(1 − s)/p + 1) + ηs f sn krZr − ksZs

F+slotpwm fs(krZr(1 − s)/p + 1) + ηs f s

n krZr − ksZs + 2p

The first group of forces is present also at sinusoidal supply but theother two groups are added at PWM operation and therefore a motorthat is quiet at sinusoidal supply can still be noisy at PWM supply.

The number of force harmonics is infinite, but the amplitude of thecorresponding harmonics in the vibration response is inversely propor-tional to m4 where m is the spatial order. Therefore, only the lowestspatial order forces result in vibration levels (and thereby noise) of sig-nificance. For traction motors, the spatial orders of interest are those oforder zero to four.

The power spectral density of an accelerometer measurement dur-ing an acceleration is plotted in Figure 2.6. The acceleration is constant,hence time is corresponding to motor speed. In the obtained spectrum,vibrations resulting from all thre groups of electromagnetic forces can beobserved. The slotting vibrations are depending on motor speed withthe corresponding force lines starting in the origin. Both PWM vibra-tions and slotting PWM vibrations are located around multiples of theswitching frequency (1050 Hz in this case) but with different depend-ence on the motor speed.

15

CHAPTER 2. ACOUSTIC NOISE OF ELECTROMAGNETIC ORIGININ TRACTION MOTORS

Figure 2.6: Spectrogram of one accelerometer during constant acceleration

With an appropriate selection of number of stator and rotor slots [18],the influence of the forces related to permeance slot field harmonics canbe reduced and the main vibration forces will be related to PWM vibra-tions, i.e. the second group of electromagnetic forces. The character-istics of these vibrations are related to the PWM switching frequency,the stator frequency and the pole pair number. Therefore, the PWMswitching frequency can be used as a design parameter to to influencethe acoustic performance of the traction motor.

The impact of PWM switching frequency on the radiated acousticnoise is addressed in several publications investigating different aspectsof the problem. It has been concluded that it is of great importance toavoid coincidence with the natural frequency of the stator. If the PWMforces acting as excitation matches the natural modes and frequenciesof the structure, higher levels of vibration and acoustic noise are cre-ated [19, 20]. The recommendation is to select a switching frequency be-low or above the frequency region of the stator natural modes of order0 and 2p since these are the spatial orders of the pure PWM forces [16].Another aspect of increasing the switching frequency is that the electro-magnetic force excitation is reduced due to an reduced harmonic content

16

2.2. ELECTROMAGNETIC FORCES

in the motor currents. However, since the structural resonances are ofgreat importance for the resulting noise level, there is not necessarily aproportional relationship between the reduced harmonic current amp-litude and the resulting acoustic noise [21].

When designing a system for high power applications, as a tractionconverter is, there is always a tradeoff between performance, size, costand operating switching frequency of the switching devices used. Theprogress in silicon carbide (SiC) material opens for replacing the conven-tionally used silicon power semiconductors with silicon carbide powersemiconductors. This enables the possibility for much higher switchingfrequencies for high power applications with the efficiency and perform-ance still retained [22,23]. Except improved acoustic performance, otheradvantages are reduced losses, reduced temperature and reduced size.Since these new devices enables a new range of feasible PWM switchingfrequencies for traction converters and train applications, it is of greatinterest to fully understand the behavior in this frequency range and theimplications for the system in total. This also include the effect on theresulting radiated acoustic noise of the traction motor.

The evaluation of different PWM methods with respect to acous-tic performance has also been subject of research. There are numerouscommonly used PWM methods [24] and on-line methods have in com-mon that strong current harmonic components are present in specificfrequency bands. These frequency bands in combination with the mech-anical design, i.e. structural behavior, determine the resulting radiatednoise [25, 26]. There are also possibilities to implement off-line optim-ized PWM methods. With these, there are openings to for example min-imizing specific problematic harmonic components [25, 27]. However,off-line methods might be difficult to implement and not always pos-sible to use in practice.

A common approach to reduce the acoustic noise at PWM opera-tion is various randomization techniques [28,29]. However, it should benoted that at randomization the total energy is not reduced but spreadover a wider range of frequencies. Thereby, tonality is reduced but of-ten the sound pressure level does not decrease and at times it even in-creases [25, 30].

17

3Summary ofappended papers

The contribution of the appended papers includes:

• Experimental measurements and evaluation of operational deflec-tion shapes of a traction motor at PWM operation (Paper A)

• Experimental investigation and evaluation of the impact of an in-creased PWM switching frequency on the radiated acoustic noiseof a traction motor. Initial validation of the proposal that the trac-tion motor shows a constant mobility behavior at higher frequen-cies (Paper B)

19

CHAPTER 3. SUMMARY OF APPENDED PAPERS

3.1 Paper A

Operational Deflection Shapes of a PWM-fed Traction MotorThe operational deflection shapes of a three-phase induction motor fedby a PWM frequency converter are studied. At operation, the radial de-flections of the stator are measured with a mesh of accelerometers in alab environment. The tests are performed for different motor speeds anddifferent PWM switching frequencies. The frequency and spatial modesof the deflections with largest amplitude are determined. The resultingvibrations are dependent of match both in frequency and space betweenthe exciting forces and responding structure.The dominating deflections all have spatial mode 0 or 4. They can allbe related to electromagnetic forces resulting from the PWM that are ofspatial mode 0 and 4. These PWM forces are a result of the combina-tion of the fundamental flux density and its time harmonics. The fun-damental flux density is given by the stator frequency, and its harmon-ics by the stator frequency and the switching frequency of the PWM.The frequency of the PWM forces can therefore be determined from thestator frequency and the switching frequency.There is a significant difference in amplitude of the largest deflections,and radiated noise, for the two different tested switching frequencies.Changing the switching frequency shifts the frequencies of the excit-ation forces and thereby influences the match between the structuralmode natural frequency of the stator and the PWM excitation. An ap-propriate selection of the PWM switching frequency is therefore a power-ful tool to influence the acoustic radiation of the motor originating fromthe electromagnetic excitation.

Figure 3.1: The operational deflection shape at 2036 Hz. The switching frequency is 1050Hz and motor speed 940 rpm.

20

3.2. PAPER B

3.2 Paper B

Reduction of radiated acoustic noise of a traction motor at PWM con-verter operationThe impact of the PWM switching frequency on the radiated acousticnoise from a traction motor is studied. A four pole 150 kW traction mo-tor is tested at PWM converter operation in a lab environment. The testis repeated for four different motor speeds and the PWM switching fre-quency is stepwise increased. The resulting sound pressure level as anaverage of three different microphone positions is quantified and ana-lyzed. It is found that the sound pressure level decreases as the switch-ing frequency increases. The trend for all four tested motor speeds issimilar and for higher switching frequencies, the sound pressure levelseems to asymptotically reach a constant level.It is proposed that the traction motor shows similar behavior as a cyl-indrical structure approaching a plate with a constant mobility abovea certain frequency. If the excitation from the PWM forces is in theconstant mobility region there are no structural advantages to increasethe switching frequency even higher. For lower frequencies, the modaldensity is lower and coincidence between the exciting forces resultingfrom the PWM supply with structural natural modes of the motor res-ult in high noise levels. The frequency where the constant mobility re-gion is entered could be used as an engineering optimum with respectto acoustic noise for the switching frequency. Noise and motor losses aredecreasing with increasing switching frequency, converter losses on theother hand increase with switching frequency. As a trade-off on systemlevel it is desirable to select a high switching frequency resulting in lownoise, but not too high to have acceptable converter losses.

0 5 10 15

Norm. switching frequency (-)

-35

-30

-25

-20

-15

-10

-5

0

Norm

. so

un

d p

ress

ure

lev

el (

-)

140 rpm

552 rpm

970 rpm

1380 rpm

Figure 3.2: Normalized average sound pressure level for increasing switching frequencyfor the four motor speeds tested.

21

4Conclusions

The main conclusions of this work are:

• An appropriate selection of the PWM switching frequency is apowerful tool to influence the acoustic radiation of the motor. Achanged switching frequency shifts the frequencies of the excita-tion forces and thereby influences the match between structuralmode natural frequencies and the PWM excitation.

• The sound pressure level decreases as the switching frequency in-creases and for higher switching frequencies, the sound pressurelevel seems to asymptotically reach a constant level.

• It is proposed that the traction motor shows constant mobility abovea certain frequency and there are no structural advantages to in-crease the switching frequency even higher than moving the PWMexcitation forces into this constant mobility region. This could beused as an engineering optimum for the switching frequency withrespect to acoustic noise.

23

Bibliography

[1] S. Frey, Railway Electrification Systems & Engineering, 1st ed. WhiteWord Publications, 2012.

[2] European Environment Agency, Noise in Europe 2014. PublicationsOffice of the European Union, 2014, no. 10.

[3] D. J. M. Houthuijs, A. J. van Beek, W. J. R. Swart, and E. E. M. M.van Kempen, Health implication of road, railway and aircraft noise inthe European Union. National Institute for Public Health and theEnvironment, 2014.

[4] TSI NOI Technical specification for interoperability relating to the sub-system ‘rolling stock –noise’, EU Std. 1304/2014.

[5] WG Railway Noise of the European Commission, Position Paperon the European strategies and priorities for railway noise abatement.European Communities, 2003, version 19403.

[6] H. Fritze, “Über die geräuschbildung bei elektrischen maschinen,”Archiv für Elektrotechnik, vol. 10, no. 3-4, pp. 73–95, 1921.

[7] P. L. Alger, “The magnetic noise of polyphase induction motors,”Electrical Engineering, vol. 73, no. 6, pp. 497–497, June 1954.

[8] P. Timar, Noise and vibration of electrical machines, ser. Studies in elec-trical and electronic engineering 34. Amsterdam: Elsevier, 1989.

[9] A. Bonnett, L. Wittman, and J. Miller, “Motor vibes: noise and vi-bration bibliographies and abstracts,” in Pulp and Paper IndustryTechnical Conference, 1993., Conference Record of 1993 Annual. IEEEPublishing, 1993, pp. 184–205.

25

BIBLIOGRAPHY

[10] P. Vijayraghavan and R. Krishnan, “Noise in electric machines: areview,” IEEE Transactions on Industry Applications, vol. 35, no. 5,pp. 1007–1013, Sep. 1999.

[11] D. G. Holmes, Pulse width modulation for power converters : principlesand practice. Hoboken, N.J.: John Wiley, 2003.

[12] W. C. Lo, C. C. Chan, Z. Q. Zhu, L. Xu, D. Howe, and K. T.Chau, “Acoustic noise radiated by pwm-controllel induction ma-chine drives,” IEEE Transactions on Industrial Electronics, vol. 47,no. 4, pp. 880–889, Aug 2000.

[13] J. L. Besnerais, “Reduction of magnetic noise in pwm-supplied in-duction machines - low-noise design rules and multi-objective op-timisation,” Ph.D. dissertation, Ecole Centrale de Lille, 2008.

[14] J. L. Besnerais, V. Lanfranchi, M. Hecquet, P. Brochet, andG. Friedrich, “Acoustic noise of electromagnetic origin in afractional-slot induction machine,” COMPEL - The internationaljournal for computation and mathematics in electrical and electronic en-gineering, vol. 27, no. 5, pp. 1033–1052, 2008.

[15] J. L. Besnerais, V. Lanfranchi, M. Hecquet, R. Romary, andP. Brochet, “Optimal slot opening width for magnetic noise reduc-tion in induction motors,” IEEE Transactions on Energy Conversion,vol. 24, no. 4, pp. 869–874, 2009.

[16] J. L. Besnerais, V. Lanfranchi, M. Hecquet, and P. Brochet, “Char-acterization and reduction of audible magnetic noise due to pwmsupply in induction machines,” IEEE Transactions on Industrial Elec-tronics, vol. 57, no. 4, pp. 1288–1295, Apr 2010.

[17] S. Ueda, K. Honda, T. Ikimi, M. Hombu, and A. Ueda, “Magneticnoise reduction technique for an ac motor driven by a pwm in-verter,” IEEE Transactions on Power Electronics, vol. 6, no. 3, pp. 470–475, Jul 1991.

[18] J. L. Besnerais, V. Lanfranchi, M. Hecquet, and P. Brochet, “OptimalSlot Numbers for Magnetic Noise Reduction in Variable-Speed In-duction Motors,” IEEE Transactions on Magnetics, vol. 45, no. 8, pp.3131–3136, Aug. 2009.

26

BIBLIOGRAPHY

[19] A. Malfait, R. Reekmans, and R. Belmans, “Audible noise andlosses in variable speed induction motor drives with igbt inverters-influence of the squirrel cage design and the switching frequency,”in Proceedings of 1994 IEEE Industry Applications Society Annual Meet-ing, Oct 1994, pp. 693–700 vol. 1.

[20] C. Wang, A. Astfalck, and J. C. S. Lai, “Sound power radiated froman inverter-driven induction motor: experimental investigation,”IEE Proceedings - Electric Power Applications, vol. 149, no. 1, pp. 46–52, Jan 2002.

[21] A. C. Binojkumar, B. Saritha, and G. Narayanan, “Acousticnoise characterization of space-vector modulated induction motordrives: An experimental approach,” IEEE Transactions on IndustrialElectronics, vol. 62, no. 6, pp. 3362–3371, June 2015.

[22] A. Elasser, M. H. Kheraluwala, M. Ghezzo, R. L. Steigerwald, N. A.Evers, J. Kretchmer, and T. P. Chow, “A comparative evaluation ofnew silicon carbide diodes and state-of-the-art silicon diodes forpower electronic applications,” IEEE Transactions on Industry Ap-plications, vol. 39, no. 4, pp. 915–921, Jul. 2003.

[23] J. Mookken, “Sic mosfets enable high frequency in high power con-version systems,” 2 2016, available: www.wolfspeed.com.

[24] J. Holtz, “Pulsewidth modulation-a survey,” in , 23rd Annual IEEEPower Electronics Specialists Conference, 1992. PESC ’92 Record, Jun.1992, pp. 11–18 vol.1.

[25] I. P. Tsoumas and H. Tischmacher, “Influence of the inverter’s mod-ulation technique on the audible noise of electric motors,” IEEETransactions on Industry Applications, vol. 50, no. 1, pp. 269–278, Jan2014.

[26] V. Radha and Z. M. Ripin, “Correlation between pwm modulationof an induction motor inverter and radiated electromagnetic acous-tic noise,” in 2014 IEEE International Conference on Control System,Computing and Engineering (ICCSCE 2014), Nov 2014, pp. 267–272.

[27] D. Franck, M. v. d. Giet, and K. Hameyer, “Active reduction ofaudible noise exciting radial force-density waves in induction mo-tors,” in 2011 IEEE International Electric Machines Drives Conference(IEMDC), May 2011, pp. 1213–1218.

27

BIBLIOGRAPHY

[28] A. M. Trzynadlowski, F. Blaabjerg, J. K. Pedersen, R. L. Kirlin,and S. Legowski, “Random pulse width modulation techniquesfor converter-fed drive systems-a review,” IEEE Transactions on In-dustry Applications, vol. 30, no. 5, pp. 1166–1175, Sep 1994.

[29] A. C. B. Kumar and G. Narayanan, “Variable-switching frequencypwm technique for induction motor drive to spread acoustic noisespectrum with reduced current ripple,” IEEE Transactions on In-dustry Applications, vol. 52, no. 5, pp. 3927–3938, Sept 2016.

[30] S. Laurentiu Capitaneanu, B. de Fornel, M. Fadel, and F. Jadot, “Onthe Acoustic Noise Radiated by PWM AC Motor Drives,” Auto-matika : journal for control, measurement, electronics, computing andcommunications, vol. 44., no. 3-4, pp. 137–145, Dec. 2003.

28

Part II

APPENDED PAPERS A-B