Embed Size (px)
Citation preview
1922 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 48, NO. 6, NOVEMBER/DECEMBER 2012
Encoderless Servo Drive With Adequately DesignedIPMSM for Pulse-Voltage-Injection-Based
Position DetectionSohji Murakami, Takayuki Shiota, Motomichi Ohto, Kozo Ide, Member, IEEE, and Masaki Hisatsune, Member, IEEE
Abstract—This paper presents a technology fusion of encoder-less control technique and encoderless-specialized motor appliedfor servo drive applications. First, a fast encoderless control tech-nique which employs a high-frequency pulsating signal injectionmethod with square-wave pulse voltage at d-axis is proposed. Theestimation principle is almost the same as that of a resolver, andthe control technique includes a fast initial position estimationalgorithm based on magnetic hysteresis phenomenon. Next, a pro-totype interior permanent-magnet synchronous motor is designedto act like a resolver. Then, evaluation results of the prototypeencoderless drive system are reported. Finally, the prototype sys-tem is confirmed to fulfill major servo specifications of about 10-bnet resolution of the position detection at standstill and morethan 60-Hz speed control bandwidth in operation speed from 0up to 1000 min−1. In conclusion, the proposed encoderless drivesystem is verified to achieve major control performances almostcomparable to those of encoder-assisted servo drive systems.
Index Terms—Encoderless, high frequency, hysteresis, interiorpermanent magnet (PM) (IPM), polarity detection, saliency, sen-sorless, signal injection, synchronous motor.
I. INTRODUCTION
T ECHNICAL troubles of an encoder attached to an ac ser-vomotor often become significant issues when the motor
is operated with huge impact or vibration forces or it is driven inhigh-temperature and high-humidity environmental conditions.
In order to avoid those encoder troubles and to improvedurability and reliability of the servomotor, encoderless tech-nique, or position sensorless technique in other words, has beenattracting attentions, and various studies have been reportedover the past two decades or so.
The encoderless techniques are classified into two cate-gories by its principle of detecting rotor position: the backelectromotive force (EMF) voltage detection method [1], [2]and the high-frequency signal injection method [3]–[10]. Thelatter utilizes magnitude of magnetic saliency which varies with
Manuscript received December 27, 2011; revised April 26, 2012; acceptedMay 30, 2012. Date of publication October 23, 2012; date of current versionDecember 31, 2012. Paper 2011-IDC-667.R1, approved for publication in theIEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the Industrial DrivesCommittee of the IEEE Industry Applications Society.
S. Murakami, T. Shiota, M. Ohto, and K. Ide are with the CorporateResearch and Development Center, Yaskawa Electric Corporation, Kitakyushu803-8530, Japan (e-mail: [email protected]; [email protected];[email protected]; [email protected]).
M. Hisatsune is with the Drives and Motion Division, Yaskawa America,Inc., Waukegan, IL 60085 USA (e-mail: [email protected]).
Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIA.2012.2226132
rotor position and not with rotor speed, so that it can operatemotor steadily at standstill and low speed. This control stabilityand feasibility at standstill and low speed are indispensableparticularly in motion and position controls required for servodrive applications.
More closely, there are two different methods in the high-frequency signal injection method category in terms of injectedsignal pattern: the rotating signal vector method [3]–[5] and thepulsating signal method. Furthermore, the latter is classifiedinto two subcategories by injection position of the injectedsignal: the d-axis injection method [3], [6]–[8] and the q-axisinjection method [9].
A comparison of the rotating signal vector method and thepulsating signal method is reported in [10], which implies thatthe pulsating signal method can be potentially less sensitive tovarious factors involving position estimation error.
In terms of industrial servo drive applications, evaluationresults are reported in [11]. Those results are promising andprospective; however, a comprehensive design scheme of totalencoderless drive system, which includes not only encoderlesscontrol technique but also encoderless-specialized motor, is notthoroughly discussed.
Regarding motor itself, a crucial problem is recognized tooccur when the motor is operated at large load: The high-frequency inductance distribution, which is a distribution ofrotor saliency as a response to injected high-frequency signal,distorts heavily when load on the motor increases [12]–[17].This phenomenon is caused by armature reaction, and it makesrotor position estimation extremely difficult at large load con-ditions and deteriorates position estimation accuracy hugely.
To solve this problem, several investigations on various mo-tor configurations and their influences on encoderless controlperformances are conducted in [9], [12]–[16], and [18]. Amongthem, motor design methods, which improve rotor positionestimation accuracy but do not degrade basic motor perfor-mances, i.e., output torque, cogging torque, efficiency, etc., arepublished in [9], [13], and [18]. However, those reports focus onmotor rotation speed range under around 200 min−1, which isnot a typical operating speed range in servo drive applications.
This paper discusses a total designing method of an encoder-less drive system which considers both position control perfor-mances and basic servomotor performances. Section II presentsfast and accurate position estimation and initial polarity de-tection principles. Section III shows a motor design conceptand how to realize it by specializing the motor configuration.
0093-9994/$31.00 © 2012 IEEE
MURAKAMI et al.: ENCODERLESS SERVO DRIVE WITH DESIGNED IPMSM FOR POSITION DETECTION 1923
Fig. 1. Control block diagram of position estimation.
Section IV and V illustrate evaluation results of our prototypeencoderless system by experiments.
II. ENCODERLESS CONTROL TECHNIQUE
A fast encoderless control technique for servo drive applica-tions is proposed in this section.
A. Position Estimation Technique
Fig. 1 shows the control block diagram of position estima-tion, where square-wave pulse voltage is injected at d-axis tosearch for rotor position. In comparison with sine-wave signalinjection, the speed control bandwidth is expected to be higherbecause of square-wave pulse injection, which is suitable forservo applications. Switching period of the injected voltage4ΔTs is twice as long as that of the pulsewidth modulation(PWM) 2ΔTs as shown in Fig. 2, where ΔTs is the samplingtime of position detection and current control. The injectionangle Δθh is adjusted to inject the pulse voltage right on theestimated d-axis. Response currents of U -phase and V -phase,whose waveforms are shown in Fig. 3(a), are monitored bycurrent sensors; then, alpha-phase and beta-phase current wave-forms are obtained through three-phase-to-two-phase transfor-mation of the stationary reference frame. Fig. 3(b) shows thewaveforms of those two phase currents. The envelope curves,which are continuously connecting adjacent peak points of eachcurrent waveform, indicate the rotor position. The rotor positionis calculated directly as an arc-tangent value on the assumptionthat the alpha-phase envelope is a cosine curve and the beta-phase envelope is a sine curve. Obviously, this assumptionis reasonable only when rotor saliency and high-frequencyinductance vary almost sinusoidally with rotor rotation. Thissinusoidal high-frequency inductance distribution is realized bya motor design method presented in Section III. The positionestimation principle mentioned earlier is identical to that of aresolver, and it ensures sufficiently high estimation accuracyand high speed required for servo drive applications.
Fig. 2. Waveforms of PWM voltage and injected voltage.
Mathematical explanation of this position estimation princi-ple is as follows.
First, the voltage equation in the synchronous referenceframe (rotor d−q coordinate system) is expressed as
[vdvq
]=
[Ra + pLd −ωLq
ωLd Ra + pLq
] [idiq
]+
[0
ωΨa
](1)
where vd and vq are d-axis and q-axis voltages, respectively,id and iq are d-axis and q-axis currents, respectively, Ra iselectric resistance of one phase, Ld and Lq are d-axis and q-axisinductances, respectively, Ψa is d-axis linkage magnetic flux,p = d/dt is a differential operator of time, and ω is rotationangular speed in electrical radian [1].
If the frequencies of injected voltage and current are muchhigher than the synchronous frequency of the motor, productterms of inductance and differentiated current are much moresignificant than other terms in (1). Thus, (1) can be approxi-mated as
[vdhvqh
]=
[Ldh 00 Lqh
]p
[idhiqh
](2)
1924 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 48, NO. 6, NOVEMBER/DECEMBER 2012
Fig. 3. High-frequency current waveforms. (a) Magnified waveform ofU-phase current. (b) Overall waveform in the two-phase stationary referenceframe.
where subscript “h” means that the value is concerned withhigh-frequency signal.
High-frequency voltage is set to be injected into the esti-mated d-axis named gamma axis. With the definition that angleθ is the gamma-axis position from the alpha-phase axis andangle θreal is the real d-axis position from the alpha-phaseaxis, phase advance of the gamma axis from the real d-axisis equal to θ − θreal. Therefore, the left-hand side of (2) isexpressed as[
vdhvqh
]=
[cos(θ − θreal) − sin(θ − θreal)sin(θ − θreal) cos(θ − θreal)
] [vγhvδh
](3)
where delta axis is defined as an axis positioned forward fromthe gamma axis by electrical 90◦.
Meanwhile, the right-hand side of (2) is expressed as[idhiqh
]=
[cos θreal sin θreal− sin θreal cos θreal
] [iαhiβh
]. (4)
With (3) and (4), (2) can be transformed into[Ldh 00 Lqh
]p
([cos θreal sin θreal− sin θreal cos θreal
] [iαhiβh
])
=
[cos(θ − θreal) − sin(θ − θreal)sin(θ − θreal) cos(θ − θreal)
] [vγhvδh
]. (5)
On the assumption that high-frequency currents iαh andiβh are approximately sine waves, the left-hand side of (5) isequal to
ωh
[Ldh 00 Lqh
] [cos θreal sin θreal− sin θreal cos θreal
] [iαhiβh
](6)
where ωh is the angular frequency of the high-frequency cur-rent. Thus, high-frequency currents are given by[iαhiβh
]=
1
ωh
[cos θreal sin θreal− sin θreal cos θreal
]−1 [Ldh 00 Lqh
]−1
×[cos(θ − θreal) − sin(θ − θreal)sin(θ − θreal) cos(θ − θreal)
] [vγhvδh
]. (7)
After expanding the right-hand side of (7), considering thatthe gamma-axis voltage vγh is substituted for ΔVh and thedelta-axis voltage vδh is zero, then, the following equation isobtained:[IcosIsin
]=
ΔVh
ωh (L20h − L2
1h)
[L0h cos θ − L1h cos(2θreal − θ)
L0h sin θ − L1h sin(2θreal − θ)
]
(8)
where the envelope curves of Icos and Isin are substituted forthe alpha-phase current iαh and the beta-phase currents iβh, re-spectively, L0h = (Ldh + Lqh)/2, and L1h = (Ldh − Lqh)/2.This substitution operation corresponds to envelope detectionprocess mentioned later in this section.
As can be deduced from (8), the envelope curves will notcontain any rotor position information of θreal if the d-axisand the q-axis high-frequency inductances are identical to eachother, which means that magnetic saliency disappears com-pletely. In order to avoid this crucial problem, a specializedmotor design method is explained in Section III.
If the estimated d-axis position θ converges to the real d-axisposition θreal, then (8) becomes
[IcosIsin
]=
ΔVh
ωh (L20h − L2
1h)
[(L0h − L1h) cos θ
(L0h − L1h) sin θ
]. (9)
The convergence process is executed by a phase-locked loopafter the first position detection process where the first d-axisposition θ0 is calculated as an arc-tangent value by
θ0 = tan−1
(IsinIcos
). (10)
Identification of the envelope curves is executed by an en-velope detector shown in Fig. 4. Difference between the formerand the present values of each phase current is multiplied by thesign of the injected pulse voltage. Then, the envelope signalsare obtained as resultant output from low-pass filters (LPFs).This process is similar to that of a peak-hold circuit. Cutofffrequency of those LPFs is fixed at a low value sufficient forreducing only noise components caused by the PWM, but it ishigher than the maximum synchronous frequency of the interiorpermanent-magnet (PM) (IPM) synchronous motor (IPMSM)rotation in order to avoid a large phase delay and deteriorationof control dynamics.
MURAKAMI et al.: ENCODERLESS SERVO DRIVE WITH DESIGNED IPMSM FOR POSITION DETECTION 1925
Fig. 4. Block diagram of envelope detector.
Fig. 5. Hysteresis phenomenon in stator and rotor cores. (a) Main loop;N pole. (b) Minor loop; N pole. (c) Main loop; S pole. (d) Minor loop; S pole.
B. Initial Position Estimation Technique
The rotor position estimation algorithm explained in theprevious section is not able to detect the polarity, i.e., N poleor S pole, at initial start-up conditions. Therefore, it occursfrequently that a result of the estimation at start-up differs fromthe real position by 180 in electrical degrees. This is a com-mon problem in encoderless control technique which utilizesmagnetic saliency, and many researches have been reported[19], [20].
In terms of estimation processing speed, this proposedmethod injects a square-wave pulse signal and utilizes magnetichysteresis phenomenon in stator and rotor cores. When a pulsevoltage is injected into the tentative d-axis, which is not yetconfirmed to be whether on N pole or on S pole, responsecurrent occurs and magnetic condition in rotor and stator coresvaries and circulates along magnetic hysteresis minor loopwhich is profiled by the curves of a, b, c, and d in Fig. 5.The circulation period is the same as that of the injected pulsevoltage, and the circulation direction whether clockwise (CW)or counterclockwise (CCW) is determined by the initial polarityon the tentative d-axis. If the initial polarity on the tentatived-axis is N pole, then the circulation direction is CCW asshown in Fig. 5(b). Meanwhile, if the initial polarity is S pole,then the direction is CW as shown in Fig. 5(d). Obviously,
Fig. 6. Response current waveforms at initial polarity detection. (a) N pole.(b) S pole.
Fig. 7. Block diagram of initial position estimation.
the relation between the initial polarity and the circulationdirection depends on the winding direction of armature coilsin IPMSM. However, the relation can be defined in advance atmotor designing.
Response current waveforms are different according towhether the initial polarity is N pole or S pole because of thesehysteresis minor loop characteristics. Fig. 6 shows responsecurrent waveforms induced by injected high-frequency voltage.Note that these waveforms are exaggerated for a clear under-standing. Fig. 6(a) shows the waveform when the initial polarityis N pole. On the other hand, Fig. 6(b) shows the one when itis S pole. These response current waveforms are determinedby inductance characteristics which are approximately definedby the gradient value of hysteresis minor loop in Fig. 5(b)and (d). This means that the second-order harmonic componentof this current waveform contains information about the initialpolarity.
Fig. 7 shows a block diagram of the proposed initial positionestimation. The bandpass filter (BPF) extracts only a high-frequency component from the monitored current waveformwithout switching noises and dc components. Output signalfrom the BPF is multiplied by pulse signal whose frequencyis twice as fast as that of the injected pulse voltage, so thatthe second-order harmonic component is extracted from theBPF output signal. The second harmonic contains informationabout the initial polarity, and the information is extracted bycalculating the integral of the harmonic value during a certaintime interval. The initial polarity is determined by whetherthe integral value is positive or negative according to N poleor S pole, and a compensational value is set at 0 or 180 inelectrical degrees, depending on the obtained initial polarity.
1926 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 48, NO. 6, NOVEMBER/DECEMBER 2012
Fig. 8. IPM motor configurations. (a) Conventional configuration. (b) Pro-posed configuration.
Meanwhile, the position estimation process explained in theprevious section is temporarily executed in advance of the ini-tial position estimation process. Eventually, the final resultantvalue of initial position estimation θinit is determined afteradding the compensational value to the output value from thetemporary position estimation process θ.
III. MOTOR DESIGN METHOD
In this section, an encoderless-specialized motor designmethod for the encoderless control technique in the previoussection is proposed.
A. Basic Design Concept
When load on the motor increases, the high-frequency in-ductance distribution is distorted by armature reaction. Thearmature reaction is a phenomenon where field magnetic fluxproduced by PM is disturbed by main current in armaturewindings which is necessary to generate load torque.
For the purpose of reducing the armature reaction, intensify-ing magnetic loading and eccentric rotor shape is effective.
First, if the magnetic loading is increased, the main currentnecessary for load torque decreases; then, the armature reactionweakens.
Next, on the conditions that the motor is operated by zerod-axis current control and only q-axis current is applied, thearmature reaction is reduced by broadening the gap distanceat q-axis. However, in designing the eccentric rotor shape,careful consideration is essential for a sinusoidal magnetic fluxdistribution in the gap to avoid increase of high-order harmoniccomponents of back-EMF voltage.
B. Motor Configuration
Fig. 8 shows a conventional IPM configuration and a pro-posed IPM motor configuration.
The conventional configuration in Fig. 8(a) holds only onePM block per one rotor magnetic pole. On the other hand,the proposed configuration holds two PM blocks per one rotormagnetic pole in a V-letter array. With this configuration, thetotal surface area of PM is larger than that in the conventionalconfiguration. Therefore, the total amount of PM flux is largeras well, so that the magnetic loading is intensified.
Fig. 9. Simplified magnetic circuit in IPM motor.
The other approach of eccentric rotor shape is also shown inFig. 8(b). The gap-distance designing formula is theoreticallyderived as mentioned as follows.
Fig. 9 shows a simplified magnetic circuit of field magneticflux in an IPM motor, where rotor and stator are linearlyunrolled for a plain explanation. Field magnetic flux producedby PM is flowing out of the rotor through the gap into thestator and returning to the same magnet. To make the discussioneasier, there is no slot for windings in the stator core whosesurface is drawn as a straight line.
In the following deduction, variable x is circumferentialdistance on the imaginary circle in the middle of the gap. Theorigin of coordinate x represents the position where the gapdistance is the smallest, i.e., the d-axis position.
In the circulating magnetic flux path shown in Fig. 9, mag-netic flux in the gap φg , magnetic resistance in the gap Rg ,and gap distance lg are functions of variable x; then, φg isexpressed as
φg(x) =Fm
Rg(x)=
FmΔSg
ρg
1
lg(x)(11)
where Fm is magnetomotive force, ρg is magnetic permeabilityin the gap, and ΔSg is cross-sectional area of magnetic fluxpath in the gap.
Hence, magnetic flux density in the gap Bg is given by
Bg(x) =φg(x)
ΔSg=
Fm
ρg
1
lg(x). (12)
In order to achieve a sinusoidal magnetic flux density distri-bution in the gap, which varies periodically along to pole pitchτp, it is essential for the gap distance lg to be expressed as
1
lg(x)= A cos
(2π
τpx
)+B (13)
where A and B are constants of integral, which are determinedby boundary conditions that the gap distance lg is the smallestvalue lgmin at x = 0 and the largest value lgmax at x = ±τp/2.
MURAKAMI et al.: ENCODERLESS SERVO DRIVE WITH DESIGNED IPMSM FOR POSITION DETECTION 1927
Fig. 10. Back-EMF voltage by calculation.
TABLE IBACK-EMF VOLTAGE AND PM VOLUME
Therefore, the design formula of the gap distance is ex-pressed as
lg(x) =2lgmaxlgmin
(lgmax − lgmin) cos(
2πτpx)+ (lgmax + lgmin)
.
(14)
The gap distance lg is the smallest at the d-axis and thelargest at the q-axis. The profile between the d-axis and theq-axis is formed by a smooth curve.
Fig. 10 shows back-EMF voltage waveforms of the conven-tional and the proposed configurations at the same rotor speedand with the same winding turn number. These waveformsare obtained by finite-element-method magnetic field analysis.The waveform of the proposed configuration is more sinusoidalowing to the eccentric rotor shape mentioned earlier, and itspeak value is larger than that of the conventional one owing tothe intensified magnetic loading.
Table I shows volumes of PM used in these two configura-tions and ratios of the fundamental component amplitude to thePM volume. This proposed configuration efficiently produceslarger EMF voltage than the conventional configuration.
IV. EVALUATION RESULTS OF MOTOR PERFORMANCES
Basic performance evaluation results of the prototype en-coderless motor designed in the previous section are mentionedin this section.
A. Basic Motor Performance
Fig. 11 shows waveforms of back-EMF voltage obtained bycalculation and experiment. The measured waveform is almostidentical to the calculated waveform, and also, its waveform is
Fig. 11. Back-EMF voltage.
TABLE IISPECIFICATIONS OF THE PROTOTYPE MOTOR
Fig. 12. High-frequency inductance distribution.
almost sinusoidal. It is confirmed that both magnetic loading isintensified and sinusoidal magnetic flux density distribution isrealized as expected in the previous section.
Owing to this back-EMF characteristic, which is the mostimportant PM motor characteristic, it is confirmed that theprototype motor satisfies adequate basic motor specificationsshown in Table II.
B. High-Frequency Inductance Distribution
Fig. 12 shows high-frequency inductance distributions ob-tained by calculation and experiment. Although magneticsaliency, which means difference between the maximum valueand the minimum value of the inductance, deteriorated un-expectedly, the measured waveforms are almost sinusoidalas expected. In addition, positions where the high-frequencyinductance is the largest and the smallest are almost steadywithout severe position shifting during load increase from noload up to 100% rated load.
Owing to these robust characteristics against load increase,encoderless control is expected to be feasible in operating
1928 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 48, NO. 6, NOVEMBER/DECEMBER 2012
Fig. 13. Initial position estimation. (a) Initial polarity is N pole. (b) Initialpolarity is S pole.
region from no load to 100% rated load, which is verified inthe next section.
V. EVALUATION RESULTS OF CONTROL PERFORMANCES
Control performance evaluation results of the prototype en-coderless drive system are reported in this section.
In the position estimation process during operation afterstart-up, the sampling time ΔTs is set at 31.25 μs and thefrequency of the injected pulse voltage is 8 kHz which ishalf of the PWM frequency of 16 kHz. Meanwhile, in theinitial position estimation process at start-up, the frequencyof the injected pulse voltage is set at 1 kHz with the PWMfrequency set at 8 kHz. The reason why these two frequenciesare decreased is that sampled data points of the phase currentincrease, and accordingly, the polarity detection result becomesmore accurate.
A. Initial Position Estimation
Fig. 13 shows transient processes of the initial positionestimation at standstill by experiment. Position signal outputfrom encoder, which is used only for the standard referenceand not for the estimation process, is also illustrated as “realposition.”
The estimation process explained in Section II-A is executedas the first procedure. Resultant values from that estimationprocess are temporarily stored, and difference between theprevious value and the present value is calculated at every timestep. If the difference converges within 0.5 in electrical degreesduring 20 ms, then the estimation process is terminated. Asthe second procedure, the initial polarity detection explainedin Section II-B is executed during a time period of 10 ms.As the third procedure, a compensational value, i.e., 0 or 180
Fig. 14. Estimation accuracy during speed change.
in electrical degrees, is determined by the initial polarity andis added to the result of the estimation process. As the finalprocedure, additional waiting time of 10 ms is provided forstart-up safety.
When the initial polarity is N pole, the total estimation time is32 ms as shown in Fig. 13(a). On the other hand, the estimationtime of 62 ms is necessary when the initial polarity is S pole asshown in Fig. 13(b). This case costs the longest estimation timeof all the cases where the rotor is positioned at from 0 up to 360in mechanical degrees. Furthermore, by another experiment,this high-accuracy initial position estimation is confirmed tobe repeated without any failure regardless of the initial rotorposition.
B. Position Control
Fig. 14 shows an experimental result of positioning accuracyduring speed change. Position command is determined based onthe profile curve of the speed command. As shown in Fig. 14,position estimation error, which is calculated from resultantvalue from the position estimation process and output valuefrom the standard encoder output signal, decreases down towithin three in mechanical degrees when the speed commanddrops from 500 down to 0 min−1. The settling time from thespeed command drop to the estimation convergence within the3◦ is about 82 ms, which is almost the same as that of encoder-assisted servo drive systems. Steady-state position error isconfirmed to be less than one mechanical degree by anotherexperiment.
Fig. 15 shows an experiment result of positioning accuracyduring load change. In this experiment, the prototype motorwhich is operated still at zero speed is coupled to anotherloading motor which produces the step load change from 0%to 100% of the rated load. In Fig. 15, the load torque on theprototype motor is displayed as 78% of the rated torque of theloading motor. Recovering time of the prototype encoderlessservo drive system is about 57 ms, which is almost comparableto that of encoder-assisted servo drive systems. This resultverifies that the proposed encoderless drive system is able to beoperated even when the load increases up to 100% rated load.
MURAKAMI et al.: ENCODERLESS SERVO DRIVE WITH DESIGNED IPMSM FOR POSITION DETECTION 1929
Fig. 15. Estimation accuracy during load change.
Fig. 16. Speed control dynamics.
C. Speed Control Dynamics
Fig. 16 shows an experiment result of speed control dynam-ics. When the speed command jumps up from 0 to 1000 min−1,settling time tset is about 7.4 ms. The settling time tset isdefined as time interval between the time when the real speedstarts to rise up and the time when the real speed reaches 95%of the speed command value.
On the simple assumption that the control system is ap-proximately the first-order system, the typical speed controlbandwidth ftyp is defined as
ftyp =1
2π
1tset3
=3
2πtset. (15)
With this definition, the typical speed control bandwidthftyp of this drive system is 64 Hz. This result indicates thatthe prototype encoderless drive system can be employed formost of the servo applications except for those which requireultraprecision motion and position controls.
VI. CONCLUSION
This paper has presented a trial result of technology fusionof encoderless control technique and encoderless-specializedmotor applied for servo drive applications.
The proposed encoderless control technique employs a high-frequency pulsating signal injection method with square-wavepulse voltage at d-axis. The frequency of injected voltage isset at half of that of the main current control loop frequencyand PWM frequency. The position estimation principle is al-most the same as that of a resolver, and envelope curves ofhigh-frequency current waveforms are extracted to obtain rotorposition information. Because of this resolver-like algorithm,good dynamic performances of control technique are achieved.The control technique includes a fast initial position estimationusing magnetic hysteresis phenomenon.
In order to make the most of these advantages, anencoderless-specialized motor is designed to act like a resolverbased on a proposed motor design method. The basic designconcept is the reduction of armature reaction. For realizingthis concept, V-shaped PM array and eccentric rotor shape areadopted after deriving a gap-distance designing formula basedon a magnetic circuit model of IPMSM.
Evaluation results of the prototype encoderless motor in-dicate that the prototype motor satisfies basic servomotorperformances and, also, that the high-frequency inductancedistribution of the prototype motor is robust and nearly zerophase shift is observed in spite of load increase from no load upto 100% rated load.
Concerning position control performances, the prototypeencoderless drive system is confirmed to start up immediatelyand successfully without any failure and to fulfill major servospecifications, such as about 10-b net resolution of the positiondetection at standstill and more than 60-Hz speed controlbandwidth in operation speed range from 0 up to 1000 min−1.
These evaluation results verify that the proposed encoderlessdrive system demonstrates control performances almost com-parable to those of encoder-assisted servo drive systems.
REFERENCES
[1] S. Morimoto, K. Kawamoto, M. Sanada, and Y. Takeda, “Sensorlesscontrol strategy for salient-pole PMSM based on extended EMF in ro-tating reference frame,” in Conf. Rec. IEEE IAS Annu. Meeting, 2001,pp. 2637–2644.
[2] N. Matsui and M. Shigyo, “Brushless DC motor control withoutposition and speed sensor,” in Conf. Rec. IEEE IAS Annu. Meeting, 1990,pp. 448–453.
[3] H. Kim and R. D. Lorenz, “Carrier signal injection based sensorlesscontrol methods for IPM synchronous machine drives,” in Conf. Rec.IEEE IAS Annu. Meeting, 2004, pp. 977–984.
[4] L. M. Gong and Z. Q. Zhu, “Improved rotating carrier signal injectionmethod for sensorless control of PM brushless AC motors, account-ing for cross-saturation effect,” in Proc. IEEE ICPE–ECCE Asia, 2011,pp. 1132–1139.
[5] S. Kim, Y.-C. Kwon, S.-K. Sul, J. Park, and S.-M. Kim, “Position sen-sorless operation of IPMSM with near PWM switching frequency signalinjection,” in Proc. IEEE ICPE–ECCE Asia, 2011, pp. 1660–1665.
[6] J.-H. Jang, S.-K. Sul, J.-I. Ha, M. Ohto, and K. Ide, “Analysis of per-manent magnet machine for sensorless control based on high frequencysignal injection,” IEEE Trans. Ind. Appl., vol. 40, no. 6, pp. 1595–1604,Nov./Dec. 2004.
[7] Y. Li, Z. Q. Zhu, D. Howe, C. M. Bingham, and D. Stone, “Improved rotorposition estimation by signal injection in brushless AC motors, accountingfor cross-coupling magnetic saturation,” IEEE Trans. Ind. Appl., vol. 45,no. 5, pp. 1843–1850, Sep./Oct. 2009.
[8] Y.-D. Yoon, S.-K. Sul, S. Morimoto, and K. Ide, “High bandwidth sen-sorless algorithm for AC machines based on square-wave type voltageinjection,” in Proc. IEEE ECCE, 2009, pp. 2123–2130.
[9] S. Yang, T. Suzuki, R. D. Lorenz, and T. M. Jahns, “Surface permanentmagnet synchronous machine design for saliency-tracking self-sensing
1930 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 48, NO. 6, NOVEMBER/DECEMBER 2012
position estimation at zero and low speeds,” in Proc. IEEE ECCE, 2010,pp. 3493–3500.
[10] D. Raca, P. Garcia, D. Reigosa, F. Briz, and R. Lorenz, “A comparativeanalysis of pulsating vs. rotating vector carrier signal injection-basedsensorless control,” in Proc. IEEE APEC, 2008, pp. 879–885.
[11] O. C. Ferreira and R. Kennel, “Encoderless control of industrial servodrives,” in Proc. EPE–PEMC, 2006, pp. 1962–1967.
[12] N. Bianchi and S. Bolognani, “Influence of rotor geometry of an interiorPM motor on sensorless control feasibility,” in Conf. Rec. IEEE IAS Annu.Meeting, 2005, pp. 2553–2560.
[13] Y. Kano, T. Kosaka, N. Matsui, and T. Nakanishi, “Design of saliency-based sensorless drive IPM motors for general industrial applications,” inProc. IEEE Ind. Appl. Soc. Annu. Meet., 2008, pp. 1–6.
[14] N. Bianchi, S. Bolognani, J. H. Jang, and S. K. Sul, “Advantages of insetPM machines for zero-speed sensorless position detection,” IEEE Trans.Ind. Appl., vol. 44, no. 4, pp. 1190–1198, Jul./Aug. 2008.
[15] N. Bianchi, S. Bolognani, and M. Zigliotto, “Design hints of an IPMsynchronous motor for an effective position sensorless control,” in Proc.IEEE PESC, 2005, pp. 1560–1566.
[16] N. Bianchi, S. Bolognani, J. H. Jang, and S. K. Sul, “Comparison of PMmotor structures and sensorless control techniques for zero-speed rotorposition detection,” in Proc. IEEE PESC, 2006, pp. 1–7.
[17] J.-I. Ha, M. Ohto, J.-H. Jang, and S.-K. Sul, “Design and selection of ACmachines for saliency-based sensorless control,” in Conf. Rec. IEEE IASAnnu. Meeting, 2002, pp. 1155–1162.
[18] N. Limsuwan, Y. Shibukawa, D. Reigosa, and R. D. Lorenz, “Novel de-sign of flux-intensifying interior permanent magnet synchronous machinesuitable for power conversion and self-sensing control at very low speed,”in Proc. IEEE ECCE, 2010, pp. 555–562.
[19] J.-I. Ha, K. Ide, T. Sawa, and S.-K. Sul, “Sensorless position controland initial position estimation of an interior permanent magnet motor,”in Conf. Rec. IEEE IAS Annu. Meeting, pp. 2607–2613.
[20] Y. Jeong, R. D. Lorenz, T. M. Jahns, and S.-K. Sul, “Initial rotor po-sition estimation of an interior permanent-magnet synchronous machineusing carrier-frequency injection methods,” in Proc. IEEE IEMDC, 2003,pp. 1218–1223.
Sohji Murakami received the B.S. degree in sys-tems engineering from Ibaraki University, Hitachi,Japan, in 2005.
He joined an electric machines manufacturingcompany as a Motor Engineer. Since 2008, he hasbeen with the Corporate Research and DevelopmentCenter, Yaskawa Electric Corporation, Kitakyushu,Japan, where he has been engaged in research anddevelopment of high-power ac servomotors, en-coderless motors, etc.
Takayuki Shiota received the B.S. and M.S. degreesin mechanical engineering from Kyushu University,Fukuoka, Japan, in 2007 and 2009, respectively.
Since 2009, he has been with the Corporate Re-search and Development Center, Yaskawa ElectricCorporation, Kitakyushu, Japan. His research inter-ests include the control of ac machines.
Motomichi Ohto received the B.Eng., M.Eng., andD.Eng. degrees from Oita University, Oita, Japan, in1984, 1986 and 2005, respectively.
Since 1986, he has been with the Corporate Re-search and Development Center, Yaskawa ElectricCorporation, Kitakyushu, Japan. His main fieldsof interest include technologies on electromagneticfield analysis and motors.
Dr. Ohto was the recipient of the Third Prize PaperAward of the IEEE Industry Applications SocietyElectric Machines Committee in 2011.
Kozo Ide (S’92–M’96) received the B.S., M.S., andPh.D. degrees in electrical engineering from KyushuInstitute of Technology, Kitakyushu, Japan, in 1991,1993, and 1996, respectively.
From 1991 to 1992, he was a Visiting Researcherwith L’Aquila University, L’Aquila, Italy, supportedby the Italian government. From 2002 to July 2003,he was a Visiting Researcher with Siemens AG,Munich, Germany. Since 1996, he has been withthe Corporate Research and Development Center,Yaskawa Electric Corporation, Kitakyushu, where he
is currently a Manager for research and development of power electronics. Hiscurrent research interests are control technology for ac machines and energyconversion systems.
Dr. Ide is a Senior Member of the Institute of Electrical Engineers of Japanand a member of the European Power Electronics and Devices Association(EPE).
Masaki Hisatsune (M’11) received the B.S. andM.S. degrees in electrical engineering from KyushuInstitute of Technology, Kitakyushu, Japan, in 1992and 1994, respectively.
In 1994, he joined Yaskawa Electric Corporation,Kitakyushu, where he was engaged in the devel-opment and design of ac motors, optical encoders,and ac motor drives. Since 2011, he has been withthe Drives and Motion Division, Yaskawa America,Inc., Waukegan, IL, as a Research and DevelopmentEngineer. His research interests include motor drives
and power electronics.
本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。
学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，
提供一站式文献检索和下载服务”的24 小时在线不限IP
图书馆。
图书馆致力于便利、促进学习与科研，提供最强文献下载服务。
图书馆导航：
图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具
