Control of Large Salient-Pole Synchronous Machines

3372 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 6, JUNE 2015

Control of Large Salient-Pole SynchronousMachines Using Synchronous Optimal

Pulsewidth ModulationJoachim Holtz, Life Fellow, IEEE , Gilberto da Cunha, Student Member, IEEE ,

Norton Petry, and Paulo José Torri

Abstract—High-power grinding mills are used in thecement and mining industries to crush clinker or copperore and grind these materials to fine powder. The multi-megawatt speed-controlled mill drives operate at a very lowangular speed. Synchronous motors with a high number ofpole pairs are used as the prime movers. They are tradition-ally fed by load-commutated thyristorized cycloconverters.These are prone to failure modes that can lead to excessivetorque pulsations and high overcurrents. The huge stator,which was built as a separate ring-shaped structure aroundthe tubular mill, may then get mechanically displaced, andthe operation of the plant is interrupted. A novel andreliable direct drive uses a voltage source inverter thatoperates at the unity power factor for increased efficiency.Synchronous optimal pulsewidth modulation ensures a lowharmonic current distortion and reduced switching lossesat a very low switching frequency. The optimization of thepulse patterns takes the anisotropic magnetic properties ofa separately excited synchronous motor into account. Theimplementation in a 23-MW semiautonomous grinding millinstalled in a Zambian copper mine is intended.

Index Terms—Alternators, ball milling, Fourier transform,gradient methods, pulse width modulation inverters, soft-ware algorithms, variable speed drives.

I. INTRODUCTION

THE production of cement requires machinery for crush-ing the cement clinker produced by a rotary kiln and

to subsequently grind it to powder. The same procedure isfollowed in copper mining [1]. Crushed copper ore is groundto powder, from which the copper content is extracted in achemical process.

The production of fine powder is efficiently done in semiau-tonomous grinding (SAG) mills. Fig. 1 shows the constructionof a SAG mill [2]. It consists of a rotating hollow cylinder ofabout 12 m in diameter. Two sets of bearings in ring-shapedstructures support the cylinder to let it rotate. State-of-the-artgearless drives have about 76 salient poles attached aroundthe mill cylinder. They form the rotor of a separately excited

Manuscript received May 22, 2014; revised July 22, 2014, August 29,2014, and October 19, 2014; accepted November 7, 2014. Date ofpublication December 9, 2014; date of current version May 8, 2015.

J. Holtz is with Wuppertal University, 42119 Wuppertal, Germany(e-mail: [email protected]).

G. da Cunha, N. Petry, and P. J. Torri are with WEG Automaçao,89256-900 Jaraguà do Sul, Brazil.

Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TIE.2014.2378732

Fig. 1. Construction of a SAG mill with the stator and the rotorseparately mounted.

Fig. 2. Cross section of a SAG mill.

synchronous ring motor. The stator is separately mounted to thebasement. It is fed by a load-commutated cycloconverter with apower rating of typically 20 MW at a maximum fundamentalfrequency of 6 Hz [3], [4]. The converter is equipped withthyristors, which are the most powerful semiconductor devicesavailable.

The interior of a SAG mill is shown as a cross section inFig. 2 [5]. The shaded area in the lower right portion indicates

0278-0046 © 2014 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

HOLTZ et al.: CONTROL OF SALIENT-POLE SYNCHRONOUS MACHINES USING SYNCHRONOUS OPTIMAL PWM 3373

where the grinding material and a number of steel balls areencountered. The cylinder rotates in the anticlockwise direc-tion, which elevates the steel balls and the major-sized lumpsof ungrounded material up to a certain level. These objectswill eventually fall down into the lower portion of the cylinder,smashing the ungrounded material there to smaller pieces and,finally, to powder with a particle size of 180 μm. This typeof mill is called semiautonomous as it is the inertial forcesof falling steel balls and ungrounded material that provide thegrinding effect.

Ring-motor-driven SAG mills in copper mines have shownfrequent malfunctions in the past [3], [6], [7]. The problemis owed to the characteristics of the cycloconverter that feedsthe synchronous motor. Cycloconverters are composed of twoparallel-connected six-phase bridges per phase, with one bridgeconducting the positive current and the other conducting thenegative current. Zero-current intervals are inserted wheneverthe direction of a phase current changes. This allows thethyristor of the outgoing bridge to regain its voltage-blockingcapability. Zero-current intervals produce zero torque. Elec-tromechanical oscillations are the consequences [3], accom-panied by extreme mechanical forces that tend to break thefixtures of the stator.

Even more critical are commutation failures that occur whena sudden voltage sag appears in the feeding mains while theconverter operates in the regeneration mode. A commutationfailure leads to a short circuit between the terminals of the statorwinding.

In any of these failure modes may the stator get displaced onits basement, and the minimum airgap may not be maintained.The production of the mill must be interrupted for extensiverepair work. The standstill of production entails considerablecost.

To overcome the aforementioned problems, a new drivesystem was developed for Cleantech Company Ltd., Finland. Itwas introduced as the world’s largest SAG mill to be installed atFirst Quantum’s Kansanshi Copper Mine in Zambia in2014 [8].

II. ALTERNATIVE DRIVE SYSTEM FOR SAG MILLS

A. Medium-Voltage Three-Level Inverter

The failures that have occurred during the operation of SAGmills were caused by using a mechanically sensitive ring motorand a cycloconverter for its control. Improvement is sought byinstalling a classical drivetrain, with the drive motor and theSAG mill having their own bearings. A 23-MW salient-polesynchronous motor, as shown in Fig. 3, was developed for thatpurpose. The huge dimensions of this motor are demonstratedby a person standing in the foreground. Fig. 4 shows the rotorpoles and the armature winding.

The unreliable cycloconverter is replaced by a pulsewidth-modulated (PWM) voltage source inverter for energy con-version and control [9]. Pairs of parallel-connected 6.5-kV600-A insulated-gate bipolar transistors (IGBTs) serve as theswitching elements in a three-level topology [10], [11]. Usingoptimal PWM permits operating at an extremely low switching

Fig. 3. Two synchronous motors coupled for testing in a factory inBrazil.

Fig. 4. Rotor poles and armature winding.

frequency without sacrificing on harmonic current distortion[12]. Switching losses are thus reduced, and the nominal powerrating of the inverter is increased [13].

The minimization of the harmonic currents by synchronousoptimal PWM has become an established technology formedium-voltage induction motor drives. For SAG mill drivesoperating at a very low mechanical speed, synchronous ma-chines are a better choice. The absence of slip reduces machinelosses and leads to a better tradeoff between the installation costand efficiency. A large airgap increases the reliability.

Different from induction motor drives, salient-pole syn-chronous machines exhibit anisotropic magnetic properties.The reactions of this machine to transients caused by therepeated voltage step changes of a PWM waveform is analyzednext.

B. Dynamics of Synchronous Machines

The anisotropic magnetic properties of separately excitedsynchronous machines require using a rotor-fixed dq referenceframe for dynamic analyses. The d-axis is aligned with the


Fig. 5. Winding structure of a synchronous motor.

direction of the rotor pole of an equivalent two-pole machine.The complex-plane topology of a synchronous machine isshown in Fig. 5.

The following differential equations hold:

ud =Raid+Lddiddt

+ωLqiq + Lfdifdt

+ LDddiDdt

+ ωLQqiQ

uq =Raiq + Lqdiqdt

−ωLdid − ωLf if − ωLDdid + LQqdiQdt

uD =RDiD + LdDdiddt

+ LfDdifdt

+ LDdiDdt

uQ =RQiQ + LqQdiqdt

+ LQdiQdt

uf =Rf if + Ldfdiddt

+ Lfdifdt

+ LDfdiDdt

(1)

where ud and uq are the components of the armature (stator)voltage, uf and if are the field winding’s excitation voltageand respective current, id and iq are the components of thestator current, and iD and iQ are the components of the damperwinding current in the rotor. The respective winding resistancesRi are marked by subscripts. Ld and Lq are the inductancesof the armature winding, and Lf is the inductance of the fieldwinding. Mutual inductances are marked by pairs of subscripts,e.g., LDq describes the magnetic coupling between dampercurrent iD and the q-axis armature winding.

A second set of equations describes the flux linkages asfunctions of the currents. We have

ψd =Ldid + Lfdif + LDdid

ψq =Lqiq + LQqiQ

ψD =LDiD + LdDid + LfDif

ψQ =LQiQ + LqQiq

ψf =Lf if + Ldf id + LDf iD. (2)

C. Calculation of Harmonic Currents

The objective of this paper is to optimize the PWM pulse se-quence of the feeding inverter for minimum harmonic distortionof the armature current. The structure of (1) indicates that thisalso minimizes the harmonic currents in the damper windingand in the field winding.

State equations (1) will be used to compute the harmoniccurrents. These are excited by the harmonic content of theinverter pulse sequences and represented by harmonic spectrain the frequency domain. Accordingly, the solution of (1) and(2) is obtained in the frequency domain, for which the algorithmin [14] is followed. The approach relies on the definition of thecomplex inductance operators Λd(s) and Λq(s), each of whichcompletely describes the magnetic properties in the respectivereference axis, i.e., d or q. The inductance operators are

Λd(s)

=

(LfLD−L2

fD

)Ld−L2

dDLf−L2fdLD+2LfdLdDLfD(

LfLd−L2fD

)s2+(RfLD+RDLf )s+RfRD

s2

+

[(RfLD+RDLf )Ld−L2

fDRD−L2dDRf

]s+LdRfRD(

LfLd−L2fD

)s2+(RfLD+RDLf )s+RfRD

(3)

Λq(s) =Lq −L2qQs

RQ + LQsmmm (4)

where s denotes the complex frequency.Using the inductance operators eliminates currents of minor

interest from state equations (1). This is done here for dampercurrents iD and iQ, and for field current if . Only two equationsthen remain. They describe how currents id and iq depend onarmature voltages ud and uq as follows:

ud(s) = (Ra + sΛd(s)) id(s) + ω1Λq(s)iq(s) (5a)

uq(s) = − ω1Λd(s)id(s) + (Ra + sΛq(s)) iq(s). (5b)

Moreover, the flux linkage components are solely expressedby id and iq as follows:

ψd(s) =Λd(s) · id(s) (6a)

ψq(s) =Λq(s) · id(s). (6b)

Equations (5) and (6) serve for computing the harmonic cur-rents generated by the three-level inverter voltage waveforms.

III. SYNCHRONOUS OPTIMAL PWM

A. Modulation Index

Modulation index m is the normalized fundamental voltage,which is defined as

m =u1

u1 six-step(7)

where u1 is the fundamental voltage component of a given pulsepattern, and

u1 six-step =4ud

π(8)

is the fundamental voltage at the six-step operation of the in-verter. The modulation index ranges within 0 ≤ m ≤ 1, wherethe unity value corresponds to the six-step mode.


B. Current Harmonics

The machine currents are composed of a fundamental currentcomponent i1 and the harmonic currents ih generated by theinverter switching. Harmonic currents substantially contributeto the machine losses. Minimizing the RMS harmonic currentis therefore an objective of synchronous optimal PWM [15].

The RMS harmonic current, i.e.,

Ih rms =

√√√√ 1

T

∫(T )

[i(t)− i1(t)]2 dt (9)

does not only depend on the respective pulse pattern but alsoon the internal impedance of the machine. This influence iseliminated when the distortion factor is used as a figure ofmerit. The distortion factor is derived from the normalized RMSharmonic content, i.e.,

Ih rms

I1=

1

I1

√√√√ ∞∑k=2

Ik2 =

ω1lσU1

√√√√ ∞∑k=2

(Uk

kω1lσ

)2

(10)

of a periodic current waveform, where I1 is the fundamentalRMS current, Ik and Uk are the respective Fourier componentsof the RMS machine current and the inverter output potentialper phase, ω1 = 2πf1 is the fundamental angular frequency,and Iσ is the effective leakage impedance of the machine.

A new variable Ih rms six-step/I1 is obtained by insertingin (10) the Fourier components Uk of the rectangular six-stepvoltage waveform. This serves to define the distortion fac-tor, i.e.,

d =1h rms

Ih rms six-step(11)

which is a quantity that represents the harmonic content of aPWM waveform independently of the properties of the ma-chine. We have d = 1 at the six-step operation by definition.

C. Optimum Pulse Patterns

The optimization of the PWM pulse patterns is done underthe restriction that steady-state conditions exist. The approachimplies that switching frequency fs and fundamental frequencyf1 = ω1/2 are synchronized. Their ratio, which is the pulsenumber as follows:

N =fsf1

(12)

is then an integer number.Constructing the synchronous optimal pulse pattern for a

three-level inverter starts with establishing the half- and quarter-wave symmetries of the phase voltages using the followingconditions:

ua(ω1t) =ua(π − ω1t), 0 ≤ ω1t ≤ π/2 (13a)

ua(ω1t) = − ua(ω1t− π), 0 ≤ ω1t ≤ π. (13b)

A pattern thus defined does not contain harmonics of even orderor subharmonics. It is characterized by N switching angles ai,

Fig. 6. Definition of the optimal switching angles αi in phase a. Theangles reappear in phases a and b at phase displacements of 2π/3 and4π/3, respectively.

i ∈ 1, . . . , N , as illustrated in Fig. 6 for N = 5, where ud isthe dc-link voltage. The three waveforms show the standarddisplacement angles 2/3 between the respective phases.

The optimum switching angles ai are determined such thatthe distortion factor is minimum and a predefined low valuefsmax of the switching frequency is not exceeded. Operatingat the synchronous optimal modulation instead of the conven-tional space vector modulation, both at an identical distortionfactor, can almost double the power rating of a given inverter[16]. Skin effect losses are not considered in this comparison.

Satisfying the objective function, i.e.,

d → min (14)

for every steady-state operating point 〈N, m〉 defines the re-spective sets of N switching angles ai per fundamental periodT1 = 2π/ω1. The optimal sets are precalculated offline andstored in a memory table of the controlling microprocessor. Theswitching angles are retrieved during the operation in real timeand used for inverter control.

A pulse pattern is now considered the superposition of al-ternating positive and negative voltage steps at the respectiveswitching angles αi. The waveform of phase a is examined inthe following. Waveforms ub and uc, as shown in Fig. 6 forN = 5, are identical except for their respective phase displace-ments. The fundamental voltage component of the pattern is

ua1(αi) =4ud

π

N∑i=1

(−1)i+1 cos(ai). (15)

Using (8) to normalize (15) yields the modulation index ofpattern P (m, N) as follows:

m =

N∑i=1

(−1)i+1 cos(ai) (16)

which enters as one condition to determine the sets of optimumswitching angles that define pulse patterns P (m, N). A secondcondition is the objective function given by (11) and (14).


The harmonic voltage components of the phase a waveformin Fig. 6 are written as a Fourier series expansion as follows:

uah(S)(νω1t) =

4ud

π

∞∑n=1

1

6n∓ 1·

(N∑i=1

(−1)i+1 cos ((6n∓ 1)ai)

)sin ((6n∓ 1)ω1t) . (17)

Operator ∓ addresses both the harmonic voltage components inthe clockwise rotation, which was represented by the negativesign of the operator, and those in the anticlockwise rotation,which was represented by the positive sign of the operator.Superscript (S) denotes stationary coordinates.

The three harmonic phase voltages uah, ubh, and uch definethe stator voltage space vector in terms of its n harmonic vectorcomponents, where n ∈ 1, . . . ,∞. Each value of n stands fora pair of spectral vector components that rotate in opposeddirections. Harmonic components having positive phase se-quences rotate in the positive direction at angular velocities(6n− 1)ω1. The harmonic components with negative phasesequences rotate in the negative direction at angular velocities(6n+ 1)ω1. The resulting harmonic orders are

1∓ 6n, n ∈ 1, . . . ,∞. (18)

Existing half-wave and quarter-wave symmetries (13) eliminatethe harmonics of even and triplen order.

D. Transformation to Synchronous Coordinates

As the synchronous machine is modeled in synchronouscoordinates, (17) and harmonic phase voltages ubh and uch

in their respective displacements are transformed to obtain thevoltage space vectors of the harmonics in the synchronouscoordinates [superscript (SY )]. The transformation yields

u(SY )h (ω1t) =

4ud

π

∞∑n=1

1

6n∓ 1

·(

N∑i=1

(−1)i+1 cos ((6n∓ 1)αi)

)

exp((−1)

12 (3∓1) · j6nw1t

)(19)

where expressions 6n∓ 1 and exp[(−1)1/2(3∓1) create twosumming terms each, with the first term using 6n− 1 towhich exp((−1)1/2(3−1)) associates and with the second termusing 6n+ 1 to which exp[(−1)1/2(3+1)] associates. Term(−1)1/2(3∓1) defines the direction of the rotation of the respec-tive harmonic.

These components rotate at angular velocities ∓6nω1, eitherin a positive or in a negative direction, going past the d-axisat a high frequency. Hence, the load angle of the synchronousmachine does not influence the space vector of the harmonicvoltages.

The orthogonal dq components of space vectors (19) are

udk=4ud

π

∞∑n=1

1

(6n∓ 1)·(

N∑i=1

(−1)i+1 cos(6n∓ 1)αi

)

× (−1)12 (3∓1) · cos ((6n∓ 1)ω1t) (20a)

uqk =4ud

π

∞∑n=1

1

(6n∓ 1)·(

N∑i=1

(−1)i+1 cos(6n∓ 1)αi

)

× (−1)12 (3∓1) · sin ((6n∓ 1)ω1t) . (20b)

Pulse pattern optimization is done offline assuming steady-state conditions. The harmonic currents are represented by dis-crete spectra ik, k = 1∓ 6n, n ∈ 1, . . . ,∞. They are computedby replacing Laplace operator s → jkω1 in (3) and –(5). Theharmonic components of the machine current are then withgood approximation obtained by solving (5) for the respectivecurrents as follows:

idk =uqk − jk udk

ω1Λd(k2 + 1)(21a)

iqk =udk − j k uqk

ω1Λq(k2 + 1). (21b)

They define the distortion factor, i.e.,

d =√idk

2 + iqk2. (22)

E. Optimization Procedure

The optimization of pulse patterns is done in the followingsteps:

1) For a constant stator flux, the fundamental voltage u1 ∝f1; hence, m ∝ f1, and from (12), we have

fs = m ·N. (23)

2) A maximum permitted value fsmax of the switchingfrequency is chosen. Every pattern P (m, N) is then validin the following range:

fs max

N + 1≤ m ≤ fs max

N. (24)

3) Switching angles αi characterize harmonic voltages (20)on which harmonic currents (21) depend. Objective func-tion d(αi, m), i ∈ 1, . . . , N , defines an N -dimensionalhyperplane. To find the optimum pattern P (m, N) fora given N , the hyperplane is searched for minimumvalues, varying the N different switching angles αi andmodulation index m within its range (24).

4) A large number of local minimums is found from whichthe global minimum is selected.

5) An optimum pattern P (m, N) only extends over itspertaining range of m, as in (24). The full range of mis built from all computed patterns.


Fig. 7. Signal flow graph of the drive control system.

6) The resulting function αi(m) will show various discon-tinuities. These will cause undesired transients duringthe operation of the drive. They are eliminated by apostoptimization procedure.

Pulse pattern optimization is described in detail in [16].

IV. SYNCHRONOUS MOTOR DRIVE CONTROL

The synchronous optimal PWM is embedded in the drivecontrol system, as shown in Fig. 7. The superimposed currentcontroller receives the fundamental stator current vector froman estimator [17]. In addition, the stator flux linkage vectoris estimated. Its argument δ is used for field-oriented control,whereas its magnitude acts through a field-weakening con-troller and a flux controller on the excitation of the synchronousmotor. Superimposed is a controller for speed and the powerfactor.

V. RESULTS

As an example, the full set of optimal pulse patterns for N =6 is shown in Fig. 8(a). The range of m selected for fsmax =400 Hz according to (24) is indicated in Fig. 8(b).

Fig. 8(b) shows that the discontinuities of the switchingangles exist at the changes in pulse number N . These abruptchanges may generate current transients, which are avoidedby employing the stator flux trajectory control [12]. Fig. 8(c)shows that the optimization results in a low distortion factor.The method permits a smooth transition to the overmodulationmode at m > 0.9.

The waveforms in Fig. 9 were recorded from a downscaled30-kW laboratory setup. The operation at a 200-Hz space vectormodulation produces a highly distorted stator current, as shownin Fig. 9(a). An increase in the switching frequency to 1 kHzreduces the current distortion, as shown in Fig. 9(b). The samedistortion is achieved with the synchronous optimal modulationat a switching frequency of only 200 Hz, as shown in Fig. 9(c).

Fig. 8. Optimal switching angles αi, i =∈ 1, . . . , N , versus modulationindex m. (a) Optimal switching angles αi for N = 6. (b) Complete set ofoptimal switching angles. (c) Resulting distortion factor.

Fig. 9. Measured waveforms of the phase a stator current and inverteroutput potential at a fundamental frequency of 33.5 Hz. (a) Spacevector modulation at a switching frequency of 200 Hz. (b) Space vectormodulation at a switching frequency of 1 kHz. (c) 200-Hz synchronousoptimal modulation at a switching frequency of 200 Hz.


Fig. 10. Commutation at t1 = 40 ms of the stator current from theinverter to the ac mains.

Fig. 11. SAG mill installation with a diameter of 10 m at First Quantum’sKansanshi Copper Mine in Zambia.

After acceleration from standstill, the SAG mill operates ata constant speed. It is then decoupled from the inverter andcommutated to the ac mains to increase the efficiency. Thecommutation is shown in Fig. 10. A photograph of that SAGmill installation at First Quantum’s Kansanshi Copper Mine inZambia is shown in Fig. 11. The diameter of the mill cylinderis 10 m.

VI. SUMMARY

SAG mills are widely used in the cement and copper miningindustries for mineral grinding. Separately excited synchronousmachines with a high number of pole pairs are used as drivemotors. They operate at a very low rotational speed.

State-of-the-art SAG mills are driven by ring motors with thestator poles mounted on the circumference of the mill cylinder.The stator and the rotor thus form mechanically separated units.The stator windings are fed by a thyristorized load-commutatedcycloconverter. Present drive systems have shown undesiredoutages owed to frequent malfunctions of the cycloconvertersrequired for extensive repair work.

High reliability is achieved with the synchronous motor con-structed as a separate unit, coupled to the mill cylinder througha common shaft. An IGBT voltage source inverter is used for

energy conversion and control. This permits the unity powerfactor operation that increases the efficiency of the drive system.

Extreme low-frequency switching reduces the dynamiclosses of the medium-voltage devices. The harmonic currentdistortion is minimized using synchronous optimal PWM.

The optimization of the switching angles makes an allowancefor the anisotropic magnetic properties of the synchronousmotor. The machine is modeled by a fifth-order set of stateequations. The methodology of inductance operators reducesthe order of the state equations. The relationship between theharmonic voltages and currents is then represented by only twostate equations. This simplifies the optimization algorithm.

Experimental results are obtained from a downscaled labora-tory drive and from a 23-MV drive system in the field.

ACKNOWLEDGMENT

The authors would like to thank Dr. N. Oikonomou andC. Erbach for their contributions to this paper.

REFERENCES

[1] J. Rodríguez et al., “Technical evolution and practical experience of grind-ing mill drives in mining applications,” IEEE Trans. Ind. Appl., vol. 41,no. 3, pp. 866–674, May/Jun. 2005.

[2] J. Pontt, J. Rodríguez, and G. Sepulveda, Accionamientos GearlessPara Molinos de Alta Potencia. Santiago, Chile: LOM Ediciónes Ltda.,Nov. 2003.

[3] V. Guerrero and J. Pontt, “Oscillatory torque caused by dead time in thecurrent control of high power gearless mills,” in Proc. IEEE Ind. Electron.Conf., Melbourne, Australia, 2011, pp. 1966–1970.

[4] J. Rojas, G. Urrea, L. Gutiérrez, and J. Navarrete, “A review of SAG millperformance at Chuquicamata,” in Proc. SAG 96, Vancouver, Canada,Oct. 1996.

[5] J. Pontt et al., “Current issues on high-power cycloconverter-fed gearlessmotor drives for grinding mills,” in Proc. IEEE Int. Symp. Ind. Electron.,Budapest, Hungary, 2003, pp. 369–374.

[6] J. O. Pontt, J. P. Rodríguez, J. C. Rebolledo, K. Tischler, and N. Becker,“Operation of high-power gearless drives under abnormal conditions,”IEEE Trans. Ind. Appl., vol. 43, no. 3, pp. 814–820, May/Jun. 2007.

[7] J. O. Pontt, “Current topics on reliability of high-power electrical ma-chines employed in minery,” in Conf. Rec. IEEE IAS Annu. Meeting,Houston, TX, USA, Sep. 2009, pp. 1–5.

[8] [Online]. Available: http://www.cleantechfinland.com/content/outotec-presents-worlds-largest-sag-mill

[9] A. S. Dias, D. B. Cundido, G. Cunha, M. Sari, and P. J. Torri, “A NPCbased medium voltage inverter applied to salient pole wound rotor syn-chronous machine,” in Proc. 12th Brazilean COBEP, Gramado, Brazil,Oct. 2013, pp. 852–857.

[10] J. Holtz, “Self-controlled inverter,” (three-level and multi-level), GermanPatent DE 23 39 034 C2 (1975), Patent in Japan (1983).

[11] A. Nabae, I. Takahashi, and H. Akagi, “A new neutral point clampedPWM inverter,” IEEE Trans. Ind. Appl., vol. 17, no. 5, pp. 518–523,Sep./Oct. 1981.

[12] P. J. Torri, G. da Cunha, T. Boller, A. K. Rathore, and J. Holtz, “Optimalpulsewidth modulation for multilevel inverter systems,” Eur. Patent EP 2312 739 A1, Aug. 25, 2013, filed Sept. 29, 2009.

[13] J. Holtz and N. Oikonomou, “Synchronous optimal pulsewidth modula-tion and stator flux trajectory control for medium voltage drives,” IEEETrans. Ind. Appl., vol. 43, no. 2, pp. 600–608, Mar./Apr. 2007.

[14] K. P. Kovács and I. Rácz, Transient Processes in AC Machines.Budapest, Hungary: Hungarian Academy of Sciences, 1959.

[15] G. S. Buja, “Optimum output waveforms in PWM inverters,” IEEE Trans.Ind. Appl., vol. 16, no. 6, pp. 830–836, Nov./Dec. 1980.

[16] A. Rathore, J. Holtz, and T. Boller, “Synchronous optimal pulsewidthmodulation for low switching frequency control of medium-voltage multi-level inverters,” IEEE Trans. Ind. Electron., vol. 47, no. 7, pp. 2374–2381,Jul. 2010.

[17] J. Holtz and N. Oikonomou, “Estimation of the fundamental current in lowswitching frequency high-dynamic medium voltage drives,” IEEE Trans.Ind. Appl., vol. 44, no. 5, pp. 1597–1605, Sep./Oct. 2008.


Joachim Holtz (M’87–SM’88–F’93) receivedthe Dipl.-Ing. and Ph.D. degrees in electri-cal engineering from the Technical UniversityBraunschweig, Braunschweig, Germany, in1967 and 1969, respectively.

In 1969, he was an Associate Professor withthe Indian Institute of Technology, Madras, India,where he became a Full Professor with and theHead of the Control Engineering Laboratory in1971. In 1972, he joined the Siemens ResearchLaboratories, Erlangen, Germany. From 1976 to

1998, he was a Professor with and the Head of the Electrical Machinesand Drives Laboratory, Faculty of Electrical, Information and MediaEngineering, Wuppertal University, Wuppertal, Germany. He is currentlya Professor Emeritus and a Consultant with Wuppertal University. He isthe author of two invited papers in the PROCEEDINGS OF THE IEEE and12 invited papers in other journals, and he is a coauthor of four books.He is the holder of 32 patents.

Dr. Holtz was the Editor-in-Chief of the IEEE TRANSACTIONS ON IN-DUSTRIAL ELECTRONICS, and he is a Distinguished Lecturer of the IEEEIndustry Applications Society and of the IEEE Industrial ElectronicsSociety, where he is also a Senior Administrative Committee Member.He was a recipient of 16 Prize Paper Awards, the IEEE IndustrialElectronics Society’s Dr. Eugene Mittelmann Achievement Award, theIEEE Industry Applications Society’s Outstanding Achievement Award,the IEEE Power Electronics Society’s William E. Newell Award, the IEEEThird Millenium Medal, the Anthony J. Hornfeck Service Award, and theIEEE Lamme Gold Medal.

Gilberto da Cunha (S’13) was born in Ibirama,Brazil, in 1972. He received the B.E. and M.E.degrees in control engineering from the Fed-eral University of Santa Catarina, Florianopolis,Brazil, in 1996 and 2003, respectively.

Since 1997, he has been with the Re-search and Development Department, WEGAutomaçao, Jaraguà do Sul, Brazil. His currentresearch interests include medium-voltage in-verters.

Norton Petry was born in Porto Alegre, Brazil.He received the Diploma in electrical engineer-ing from the Federal University of Rio Grandedo Sul, Porto Alegre, Brazil, in 1981 and thePostgraduate Diploma in control and powerelectronics from the Federal University of SantaCatarina, Florianopolis, Brazil, in 2003.

Since 1981, he has been with WEGAutomaçao, Jaraguà do Sul, Brazil, leadingresearch and development activities and newproduct development in the field of variable-

speed drives.

Paulo José Torri received the Diploma in elec-trical engineering from the Federal Universityof Santa Maria, Santa Maria, Brazil, in 1984and the M.Sc. degree in electrical engineeringfrom the Federal University of Santa Catarina,Florianopolis, Brazil, in 1986.

In 1987, he joined WEG Automaçao, Jaraguàdo Sul, Brazil, where he worked on the re-search and development (R&D) of products forpower electronics systems, drives, and electricmachines. He is currently the Manager of the

R&D Department, WEG Automaçao, where he is responsible for thedevelopment of products in the field of low-voltage and medium-voltagepower converters, high power, and renewables.

Control of Large Salient-Pole Synchronous Machines

Documents