00245714 a strategy for improving reliability of field-oriented controlled induction motor drives

8/13/2019 00245714 a Strategy for Improving Reliability of Field-Oriented Controlled Induction Motor Drives

1/9

910 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 5 , SEPTEMBEWOCTOBER 1993

A Strategy for Improving Reliability ofField-Oriented Controlled Induction Motor Drives

Tian-Hua Liu Member, IEEE, Jen-Ren Fu, Student Member, IEEE, and Thomas A. Lipo, Fellow, IEEE

Abstract- Although design of induction motor drives em-ploying field-oriented control has reached a relatively maturestate, relatively little effort has been expended on improvingthe reliability of these drives. In this paper a new, improvedinduction motor drive topology and control strategy is proposedthat allows for continuous, disturbance ree operation of the driveeven w i t h complete loss of one leg of the inverteror motor phase.A complete analysis and computer simulation of this new controland circuit concept is included.

I. INTRODUCTIONN the past several decades substantial progress has beenI ade in the development of ac motor drives using both

hard switched dc link and resonant-link schemes that utilizehigh speed switching devices such as fast recovery bipolarjunction transistors, insulated gate bipolar transistors (IGBTs)and GTOs. Switching losses of hard switched converters (dcvoltage link converters) have also improved dramatically dueto the substantially reduced turn off time of third generationIGBTs. Also, control strategies, particularly field-orientedcontrol strategies, have greatly improved the performance ofac motor drives. These control strategies utilize almost exclu-sively pulse width modulated (PWM) switching strategies thatexploit the low switching loss capability of these convertersand seek to produce a precisely controlled current to thewindings of the motor. In effect, the controller serves toconvert the dc/ac inverter from a voltage to a current sourcethereby overcoming many (but not all) of the inaccuraciesinvolved in induction motor torque control.

While current regulation has greatly improved the torqueresponse of ac drives, the use of this principle as a meansof avoiding problems during system faults is relatively unap-preciated. One of the most common types of faults, is theloss of a transistor in one of the legs of the inverter, oralternatively, the loss of one of the phases of the motor. Inthis case one of the motor phases is suddenly open-circuited,essentially single phasing the motor, resulting in a loss offield orientation and in high pulsating torques. Braking of themotor is then typically initiated using one of the following

Paper IPCSD 92-47, approved by the Industrial Drives Committee of theIEEE Industry App lications Society for presentation at the 1991 AnnualMeeting of the IEEE Industry Applications Society, Dear bom,MI September28-October 4.Manuscript released for publication December 28, 1992.T.-H. Liu is with the Department of Electrical Engineering of NationalTaiwan Institute of Technology, Taipei, Taiwan.J. R.u is with the Taiwan Power Company, Taipei, Taiwan.T. A. Lip0 is with the Department of Electrical and C omputer Engineering,

IEEE Log Number 9212089.University of Wisconsin-Madison, Madison WI 3706.

Fig. 1. Induction motor drive with machine neutral fed back to dc busmidpoint.

strategies: 1) friction braking, 2 ) dc current injection into thestator of the machine, 3) capacitive self-excitation braking,or 4) magnetic braking by shorting two or three stator leads[11-[3] or combinations of these braking schemes. In all cases,these strategies involve expensive mechanical devices such asfriction brakes or mechanical contactors to initiate the brakingprocess.

The cost and complications arising during emergency brak-ing can be eliminated by considering the drive circuit shownin Fig. 1. It can be noted that this topology differs fromthe standard ac drive topology only in that the neutral pointof the motor is returned to the midpoint on the dc voltagelink. The midpoint is created by simply splitting the capacitorbank into two equal sections. During normal operation thestator current is regulated in normal PWM fashion by thefield-oriented controller. In this case, the current flowing inthe neutral of the machine, connected to the midpoint ofthe dc link, is essentially zero with possibly a very smallcurrent flowing due to PWM operation of the inverter. In theevent that a transistor fails open in the inverter, however, anew current control strategy can be initiated by the converterthat preserves the torque at its original value or changes thetorque to any desired value while eliminating the negativepulsating torque usually associated with operation with anopen phase.

11. ANALYSISIn order to illustrate the concept for control during a single

phase open circuit it is useful to employ the principle of spacephasors [4]. In particular, assume that at t = 0- all three

0093-9994/93 03.00 993 IEEE


2/9

LIU et al.: A STRATEGY FOR IMPROVING RELIABILITY OF FIELD OFUENTED CONTROLLEDNDUCTION MOTOR DRIVES 91 1

as and q-axis

cs-axis d-axisFig. 2. Orientation of the current space vector just before one phase isopen-circuited.

phases are excited normally and the current space phasor islocated as shown in Fig. 2.

It is assumed that before any phase is open-circuited, thecurrents feeding the induction motor are regulated to be threebalanced positive sequence sinusoidal currents, i.e.

JI b s

Fig . 3. Phasor relationships before and after phase b is open-circuited.ICs

I,, = I cos(wt + q5I,, = I cos (w t + 5 + 27r/3).Ibs = I c o s w t + - 2 ~ / 3 ) (1)

The rotating MMF generated by the armature currents is the lJS Ibs I b ssum of the MMF's caused by each of the three phases. Basedon the three axes shown in Fig. 2 , this MMF can be expressedby the complex vector

(a) (b)Fig. 4. Phasor relationships required for a unaffected operation when (a)phase a is open -circuited and (b) phase c is open-circuited.

From these two equations, it can be determined thatMF = MMFa+ MMFb+ MMF,= N I , , + UNIb, + a 2 N I , , ( 2 )

where U = 1L120 and N is the effective number of statorturns per phase. For balanced three-phase operation

I, ,= -&I sin3 1I:, = -I(cos 0 - in 0)2 (8)- or, equivalently3

2MMF = -Fe? (3)(4)

32= -F(cosO+jsinO)

where F = N I and 0 = w t + 4).Assume that at any time t , the current in phase b suddenly

drops to zero. In this case, the rotating MMF will be the sumof MMF, and MMF, only, i.e.

MMF ' = N I : , + u2NI, ,

I:, = &I cos wt + 4 + 7r/6)I: = Icos(wt + q5 + 7r/2). 9)Hence, if phase b is open-circuited, disturbance-free con-

trol is possible if phase a current is regulated to advance by30 and phase c current regulated to be retarded 30'. Bothphase a and phase c current magnitude must also be increasedto imes their previous value. Fig. 3shows the phasor rela-tionships before and after phase is suddenly open-circuited.

The same analysis can be applied to the cases of phase a orphase b being open-circuited. If phase a is open-circuited, thencontrol of the remaining two currents after the open circuitshould be

I;, = &I cos(wt + 5 - 57r/6)I, , = Icos(wt + 4 + 5 ~ / 6 ) .I:, = &I cos(wt + 4 - ~ / 6 )

(10)(11)

where the prime indicates the new values of each variableafter one phase is open-circuited.

by setting (4) equal to 5 ) and then solving for real andimaginary parts separatelyThe same MMF is maintained after phase b is open-circuited

If phase is open-circuited(12)

( 6 ) I;, = Icos(wt + 4 - 7r/2). (13)3 1-F COS 0 = NI L, - -N I , ,2 23 Fig. 4 shows the phasor relationship before and after phases-FsinO = - -NI , , .2 2 (7) a and c are open-circuited. In any of the above three cases,


3/9

912 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 5, SEPTEMBEWOCTOBER 1993

it is noted that after the opening of one phase the remainingtwo phases should be regulated to a magnitude of fi imesof their original value and phase-shifted 60 with respect toeach other. The new phase angle is accomplished by regulatingone phase to jump forward 30 and the other phase to jumpbackward 30 with respect to their positions before the opencircuit.

Fig. 5shows an idealized example of a typical scenario inthat the motor phase c is open-circuited at t = 0 for the specialcase in which f = 0. The only torque transient to be expectedis the result of the d i / d t limit of the inverter imposed by themaximum dc link voltage.

111. COMPUTERIMULATIONA. Simulation Model

To verify the aforementioned principles, a simple computermodel for field-oriented controlled induction motor wasestablished. In this model, indirect field orientation wasimplemented based on the slip relationship for field orientationP

Rr 1 4 sLr Idssw, = .

A speed control loop with a PI controller is used to controlthe motor speed. The PI controller compares the preset speedsetting to the rotor speed signal and generates a torquecommand, i.e., I:,. The flux command I& is assumed tobe constant in the simulation. The field angle B f whichis required to transform the d-q axis commands to a-b-caxis commands, is generated by summing the rotor-positionsignal and the slip-position signal. The rotor-position sig-nal is obtained directly from the integration of the rotorspeed. Using the calculated value of 8.f. the current com-mands in the d-q axis I:, and I& are transformed to currentcommands in the a-b-c axis I&,I;, and I&. The com-manded currents I:,, I;,, and 12, are then compared to themotor stator currents I,,, Ib and ICs which are gener-ated from the dynamic model of the induction motor. Threecurrent-regulated delta modulators operated under a speci-fied sampling frequency are used to generate the switchingpattern for each phase. These switching patterns are thenused to control the switches of the current regulated PWMinverter.

The following motor parameters are used in the simulation:L, = 0.07131 h, L , = 0.07131 h, L, = 0.06931 h, R, =0.4332, R, = 0.816Q V&= 198 V, KI = 0.01, K p = 0.1,Commanded speed = 1000rpm, Switching frequency = 5 kHz.

The dynamic model of the induction motor is basedon Stanley's equations. In general, the simulation of aone-phase open-circuited condition can be achieved bytwo means. The first method employs the use of theinduced voltage in the open winding 161. The voltageappears across the open winding can be determined bythe mutual inductance, rotor leakage inductance and therotor rotating flux. The other method assumes zero currentin the open phase as a constraint. Substituting this

2

Bv Oc l

-2

2

? o--2-360 0 360

at (deepree)Fig. 5. Phase currents before and after a sudden open circuit of phase c,which results in the same instantaneous torque and notorque transient.

constraint into Stanley's equations, a second set of dynamicequations dedicated to one-phase open-circuit operation canbe derived.

B. Simulation ResultsBoth the simulation of motor starting and normal operation

with and without load have been implemented. In all cases, onephase current is assumed to drop to zero suddenly while theremaining two phases are regulated to the new magnitude andphase angles required to maintain the same M M F as discussed


4/9

LIU et al : A STRATEGY FOR IMPROVING RELIABILITY OF FIELD-ORIENTED-CONTROLLED INDUCTION MOTOR D RN ES 913

7s.w s.m i .so 6.10 6.30 K so1 10 1(c)

(c) I a s ~Fig. 6. Current variation after phase b is open-circuited (a) I,,, (b) IC-

Fig. I .

in Section 11. Figs. 6and 7 are the simulation results for thecase where phase b is suddenly open-circuited.

Note from Fig. 6that under this type of unbalanced op-eration, the neutral current I,, is no longer zero. Instead, itbecomes the sum of I,, and I,, and is three times the valueof the original operating line current before the open circuit.From Fig. 7 it can be noted that although the phase b currentdrops to zero at t = 0.6 s, the torque is almost unaffected. Fig. 8.

(b)Torque variation after phase b is open-circuited. (a) I,, . (b) Torque.

-I. q I II- -- : - - - - - -

System control block diagram incorporating open -circuit contingencycontrol.Iv. CONTROL SYSTEM BLOCKDIAGRAM

The system of Fig. 8shows one method for implementingthe proposed scheme by modifying the indirect field-orientedcontrol strategy. The basic concept is same as the simulationmodel described in Section 111-A.In addition, a device to sensethe operating condition is required in order to implement acontrol scheme change. This device should have the ability notonly to detect the open circuit but also to identify the differentphases so that the system can switch to the appropriatecontrol scheme accordingly. The a-b-c axis current commandsare obtained from the d-q to a-b-c axis transformation. Forexample, for normal operation, the current commands are

I:, = I;, cos 8 + Id, sin 8 (15)I;, = I& cos(8- 2 ~ / 3 ) I;, sin(8- 2n/3I,, = I,*,cos(8+ 2 ~ / 3 ) Ids sin(8 + 2 n / 3 ) . (16)(17)

When phase b is open, the current commands should becomeI;, = h [ I os(8 + ~ / 6 ) I;, sin(8 n/6) ] (18)

19):, = fi[-~, ~in + 2 ~ os 01It is clear that the controller of the proposed system can

readily take advantage of todays digital signal-processingtechnology. As shown in Fig. 8, continuous control can betransformed to the discrete domain by using a DSP. Thedigital controller provides not only a highly accurate but alsodrift-free control. The three control schemes required for thedifferent open-circuit operations can be pre-programmed andstored in the memory of the digital signal processor. The mainprogram is then interrupted to execute the contingency controlas soon as an open-circuit signal is detected.

For proper two-phase operation, it is apparent that thecapacitor midpoint voltage should be maintained at one half of


5/9

914 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 5, SEPTEMBEWOCTOBER 1993

Fig. 9. Use of a pair of resistors to force voltage sharing across the dc linkcapacitors.

Fig. 10. Use of a triac to block the neutral current during normal operation.the dc bus voltage. To prevent the capacitor midpoint voltagefrom drifting from this desired voltage, the following threecontrol methods can be considered:

a) A pair of small resistors can be used to help stabilize thedrift as shown in Fig. 9.In this case some extra lossesare clearly incurred. However, in most cases the addedlosses can be considered as negligible.

b) It can be noted in Fig. 8that the neutral current is notzero but carries a high frequency switching current withzero average value due to the PWM inverter. A triaccan be connected between the capacitor midpoint andthe motor neutral point to eliminate this current. Duringnormal operation, the triac is blocked, so no current willflow to unbalance the charge to the capacitors. The triacis only triggered on when the one-phase open signalis detected and a two-phase operation is required. Fig.10shows a sketch of this control method.

c) Introduce a flux or torque command regulator to manip-ulate the motor-current commands so that the capacitormidpoint voltage will be regulated and will not deviatemuch from the desired value. Fig. 11shows the controlblock diagram of this method using flux as the correctingsignal. The variation of the flux command may result inthe torque variation. However, this variation is consid-ered not to be significant. The speed change as a resultof flux command change will finally be compensatedthrough the speed feedback control loop.

V. FURTHER IMULATIONSOne of the important factors in the practical implementation

of this control strategy is the selection of the capacitor size.Under unbalanced two-phase operation, the neutral current

I; -Fig. 11. Flux command control method.

becomes three times that of the line current required fornormal operation so that the link voltage could be subjectedto severe pulsations in amplitude. To simplify the evaluation,it is assumed that stator currents are regulated exactly tobe a sinusoidal waveform so that the neutral current can beconsidered as a current source charging the capacitors. Bycomparing the impedances of the capacitor and the dampingresistor, one can conclude that most of the current will flowthrough the capacitor. Hence, the voltage variation across thecapacitors is estimated as

where Io, neutral currentw e electrical frequencyC dc link capacitance.

Hence, the capacitor selection strongly depends on the operat-ing current, frequency, and the voltage-variation tolerance. Arough estimate for the case simulated in this paper, to keep thevoltage variation less than 10% of the dc voltage, a capacitor ofabout 0.004 F is required. However, for higher load operationor more strict voltage-variation tolerance, a higher capacitorsize is required accordingly.

Fig. 12shows a modified simulation in which the effects ofthe finite link capacitor are incorporated into the computermodel. In particular, the value of the dc link capacitancewas chosen as 10000 pF. Method b), above, was used toprevent link voltage drift. That is, a triac is triggered totie the motor neutral back to the dc voltage midpoint onlywhen an open circuit is detected. Note that the dc linkvoltage now sustains a moderate ripple voltage. However,examination of the motor torque verifies that the control isnot affected.

In Figs. 13and 14, operation with the resistive dampingcircuit and with the flux command regulator are portrayedby means of simulation. Considering both the damping effectand power dissipation, values of 1000 R were used for thetwo damping resistors. Using this selection, with the capacitorand the dc voltage used in the simulation, the damping timeconstant is 10 s and power dissipation is 20 W. Again it can


6/9

LIU et al : A STRATEGY FOR IMPROVING RELIABILITY OF FIELD-ORIENTED-CONTROLLEDNDUCTION MOTOR DRIVES 915

I I J I I I I

8 I I I 88

I I I

Fig. 12. Opening of phase b using the circuit of Fig. 10.Traces from topto bottom are: Phase c voltage; phase b voltage; dc linkmidpoint voltage;electromagnetic torque.

be noted that the currents are well-regulated and that drift ofthe dc link midpoint voltage is well-damped both before andafter the open circuit occurs.

VI. EXPERIMENTALESULTSTo verify the control strategy discussed in Section IV, a

7.5 HP General Electric induction motor and Motorola DSP56000 Application Development System has been used toimplement the control block diagram shown in Fig. 8.Also,an Emerson adjustable frequency ac drive was modified toaccept the switching pattern signals generated from a digitalsignal processor (DSP). An encoder attached to the motorshaft provides pulse signals so that the motor speed can bemeasured by counting the pulses. Three current sensors, onefor each phase, provide analog current signals. By means ofan analog-to-digital card, these analog signals were trans-formed into digital signals and were utilized in a DSP. Aneutral line was connected between the motor neutral andthe dc midpoint so that the current of the remaining twophases can be individually controlled after one phase is open-circuited.

A pair of back-to-back connected SCR s were used to switchoff the motor phase current. The current was interrupted at azero crossing in the same manner as used in the simulationstudy, Section 111. The current was interrupted at a zero

Fig. 13. Opening of phase b using the circuit of Fig. 9.Traces from topto bottom are: phase c voltage, phase b voltage, dc linkmidpoint voltage,electromagnetic torque.

crossing to prevent possible high induced voltages due to thehigh d i l d t of the inductive motor current. In practice an opencircuit can occur at any current level and an electrical arcwill temporarily accompany an open circuit fault. However,this transient will not affect the long-term operation of theproposed control strategy. Simple RC snubber circuits werealso used to protect the SCR s in the neutral. A pair ofresistors are used to help stabilize the drift of dc midpointvoltage, as shown in the control scheme of Fig. 9. Thecapacitance and resistor used were 10000 pF and 750 ohms,respectively.

The DSP control-block diagram is shown in Fig. 15. TheDSP was programmed to begin execution of the main programfrom program memory address location 100 (hexadecimal).First, to prevent the motor from conducting, a disable sig-nal is sent to disable the PWM operation. The DSP thenbegins to obtain information regarding the rotor initial po-sition and calculates the zero offset of the three currentsensors. After this interval, the initial values of the statevariables are loaded. When all these tasks are completed,the main program jumps to a subroutine to obtain the ini-tial switching pattems. The initial switching pattems areobtained by assuming a constant torque command and fluxcommand. This process is repeated about 1000 times so thatthe machine speed decreases somewhat before the controlleris again put into a field-oriented control condition. Beforethe DSP begins to perform field-oriented control, the timer


7/9

916

2 PhaseSwitchingPatterns

IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 29, NO. 5 SEPTEMBEWOCTOBER 1993

3 PhaseSwitchingPanerns

85 I - 1 1 I8

j-p-&-?n

Fig. 14. Opening of phase b using the circuit of Fig. 11 . Tracesfrom topto bottom are: phase c voltage, phase b voltage, dc linkmidpoint voltage,electromagnetic torque.

in the DSP must be set to allow an interrupt occur. Thetimer was, in this case, set to limit the maximum switchingfrequency of the delta modulated PWM to be less than themaximum allowable hard switching frequency of the switchingdevices.

During field-orientation control, I& is calculated in thesynchronous frame from the speed error and a PI controlleras discussed in Section IV. During each calculating cycle,an open-circuit determination logic is performed to decidewhether or not an open-circuit fault has occurred. If it isin the normal operation mode, I and I are transformedinto I,*,, I6 and Izs. If one phase is open-circuited, I ,and I are transformed by the theory introduced in thispaper into current commands for the remaining two healthyphases. These current commands then compared to theactual current signals and generate the PWM switchingpatterns. In the open-circuit determination logic, specialconsideration must be given to prevent mistaking a currentzero crossing as an open-circuit fault. Normally, there isa minimum time required to confirm an open-circuit faultexists. This minimum time is determined by the currentmagnitude, frequency, and random noise level. The timerequired to justify an open-circuit fault is then converted toa number consistent with the program sampling frequencyso that DSP can count the open-circuit time by meansof clock cycles. A hysteresis CRPWM control scheme

Speed Setting :DisablePWM

Variables

lump into Loop forInitial Output

Set Up Timer

Measuringor

Fig. 15. DSP control block diagram.

m-t ar-92 kcrw17: 9: 5Panel

STATUS

Auxiliaryst chon12om . 5 vC h a n 2C h a n 3M m s .5Vulan 4

041 .5 v - 2om IVac? . 5 v .X 3 . 6 V .W 1 V - T/div20=

x-Y mode 2om 5 v

Return

Fig. 16. Current waveforms of actual system during loss of a phase. a) Phasecurrent a b) phase current b, c) phase current c, d) motor neutral current.

was implemented to maintain the current at the desiredvalue.

Fig. 16shows the results for the experimental system inwhich phase b is open-circuited by suddenly turning off thetransistor switches corresponding to this phase. A substantialflow of high-frequency current in the motor neutral can beobserved before the open circuit occurs. Clearly, this currentcan be reduced by either increasing the modulation frequencyor by increase in the zero sequence inductance either by motordesign or by physically inserting a small inductor in the motorneutral. In the practical system, the PWh4 frequency waslimited at about 5 kHz.


8/9

917IU er al. : A STRATEGY FOR IMPROVING RELIABILITY OF FIELD-ORIENTED-CONTROLLED INDUCTION MOTOR DRIVES

1r(.norystormR.coll

STATUS

PANL

-id

R5232Plot

rtu off T/divSOnm Ch2 VT r i g - - L I N

00245714 a strategy for improving reliability of field-oriented controlled induction motor drives

Documents