永磁同步电机无位置传感器

Performance Comparison of Permanent Magnet Synchronous Motors and Controlled Induction Motors in Washing Machine Applications using Sensorless Field Oriented ControlAengus Murray, Marco Palma and Ali HusainEnergy Saving Products DivisionInternational RectifierEl Segundo, CA 90245Abstract—This paper describes two alternative variable speed motor drive systems for washing machine applications. Three phase induction motors with tachometer feedback and direct drive permanent magnet synchronous motors with hall sensor feedback are two drive systems commonly used in North American washers today. Appliance manufacturers are now evaluating sensorless drive systems because of the low reliability and high cost of the speed and position feedback sensors. A Field Oriented Control Algorithm with an embedded rotor flux and position estimation algorithm enables sensorless control of both permanent magnet synchronous motors and induction motors. The estimator derives rotor shaft position and speed from rotor flux estimates obtained from measured stator currents and the applied voltages. Sampling of currents in the dc link shunt simplifies stator current measurement and minimizes cost. Field oriented control algorithm allows good dynamic control of torque and enables an extended speed range through field weakening. The digital control algorithm runs on a unique hardware engine that allows algorithms to be designed using graphical tools. A common hardware platform can run either the PMSM or IM using sensorless field oriented control in a front loading washer application. Test results are presented for both drives in standard wash cycles.Keywords-component; Advanced Control; Field Oriented Control Algorithm;, Appliance control architecture;I.I NTRODUCTIONAccurate control of drum speed is required in both horizontal and vertical axis washer machines [1]. In front loading horizontal axis washers, the drum speed determines the washing action. There is a critical drum RPM, depending on the drum radius, above which the clothes stick to the inside edge of the drum. At this speed, the centrifugal force due to rotation balances the weight of the wet clothes. At speeds below this, the clothes will stick to the side of the drum until the component of the weight acting along the radius is greater than the centrifugal force. Once this angle is reached, the clothes fall back down into the base of the drum. The speed of the drum determines how vigorously the clothes are washed and allows a gentle wash cycle to be selected for delicate items. In the spin mode, the water is drained and the drum speed is increased well beyond the critical speed and the water forced out of the cloths by the centrifugal force. In traditional top loading vertical axis machines, the agitation action is produced mechanically using a gearbox and clutch. However, the introduction of speed control systems not only simplifies the mechanical system but also allows for wash cycle control. The control of the speed and angle of stroke allows the system designer to better manage the washing action and so develop wash cycles that use less water.European front-loading washers have used variable speed control for many years and typically use a universal ‘brush type’ motor. However, the American washer uses a larger drum size, which requires a motor with a power range beyond that of the universal motor solution. The front-loading drive solutions on the market today include direct drive permanent magnet synchronous motor drives or a belt drive using an induction motor. Appliance manufacturers are now evaluating these two drive types in top-loading machine to reduce cost and improve performance. However, both these drive systems use shaft feedbacks sensors. The direct drive PMSM typically uses a Hall Effect sensor for position feedback while the induction motor drive typically uses an analog or digital tachometer for speed feedback. The ideal universal drive can run either a PMSM or an induction motor without shaft feedback sensors. However, a single hardware platform can efficiently run either a PMSM or an induction motor using sensorless field oriented control algorithm. In both cases, speed and position estimates derive from motor terminal voltages and currents.Induction motors were initially preferred for washing machine drives because of the ease of running in high speed field weakening mode even with simple scalar control methods. However, the PMSM is now becoming a viable solution because field oriented control approach enables high speed field weakening. In an induction motor, the torque producing current flows in both the rotor and stator windings while the air gap field generation needs additional field current. Therefore, in washing mode, the total copper losses are more than doublethe PM motor losses since all the PM motor stator current can generate torque. However, in spin mode the PM motor has a disadvantage since field weakening consumes a large fraction of the stator current. In the induction motor, the magnetizing current is almost zero in field weakening mode. This paper compares the performance of two similarly rated motors operating in both parts of the wash cycle. It also examines specific advantages for each of these drives in the washer application.II.F IELD ORIENTED CONTROL OF PERMANENT MAGNET SYNCHRONOUS MOTORS AND INDUCTION MOTORSField oriented control (FOC) in three phase motors is a well know technique that decouples the ac stator currents into quasi dc field and torque components to simplify the control system design [2]. The control architecture, shown in Fig. 1, is based around rotating reference frame models that were developed many years ago to analyze the operation of ac induction and synchronous machines. In the control model, stationary windings carrying ac currents are represented by equivalent rotating windings carrying dc current. The well known Clarke and Park vector transformations translate three phase stator currents into d-axis and q-axis current components in a reference frame synchronized to the angle of the rotor flux. In the rotating frame reference system, we define the d-axis as being aligned with the rotor field and the q-axis as being in quadrature with the rotor field. In both the synchronous motor and the induction motor, the rotor flux is a function of the d-axis current and the motor torque is proportional to the product of the rotor flux and the q-axis current. Thus, the d-axis current is equivalent to the field current in a dc motor while the q-axis current is equivalent to the dc motor armature current. Two independent current loops vary the applied stator winding voltages in order to deliver the desired motor torque and flux. The use of the above transforms reduces the problem of controlling the magnitude and frequency of three ac currents to a problem of controlling the magnitude of two dc currents. A further advantage is that the stator winding impedance, which is time varying in the stationary ac reference frame, becomes a fixed inductance in the rotating reference frame. This means that the current control loop dynamics become independent of the stator frequency improving the robustness of the system robustness. The input to the q-axis current controller is the torque reference, which is derived from the velocity or position control loop. The input to the d-axis controller is derived froma field current control function depending on the motor type.Figure 1. Field Oriented Control ArchitectureSince the rotor flux in a PMSM is generated by a permanent magnet, the field reference is normally set to zero. In the case of the induction motor, d-axis current is required to magnetize the core just like in a field wound dc motor. Since there is cross coupling between the q-axis and d-axis windings, the d-axis current controller must apply voltage even to maintain zero d-axis current. At some speed, known as the base speed, the required q- axis and d-axis voltage exceeds the inverter voltage capacity. Above base speed, the controller operates in field weakening mode and reduces the rotor flux to lower q-axis voltage generated by the motor windings. In the induction motor, the rotor flux is reduced by lowering the d-axis current. However, in the case of the PMSM, a negative d-axis current is required to oppose the permanent magnet field in order to weaken the rotor flux. The PMSM is more efficient than an induction motor at low speeds because it does not require magnetizing current but the induction motor draws lower current at high speeds.III. R OTOR F LUX ANGLE ESTIMATIONPermanent magnet synchronous motors with rotor position sensors have been used in high end industrial drives for many years. In sensorless drive systems, a motor circuit model estimates the rotor flux λr from the measured stator currents and applied voltages. The two-phase equivalent circuit model for a PMSM yields two equations (1) that are sine and cosine functions of the rotor flux angle θr . Integration of the voltage equations yields flux estimates (2) that are sine and cosine functions of the rotor angle. The flux estimates drive a phase locked loop that extracts the rotor angle and velocity. The rotor angle feedback enables field oriented control of stator currents while the rotor velocity feedback enables closure of the outer velocity loop.()()()()inductanceand resistance stator the are currents and tages stator vol phase two the are ,,,where,sin ...cos ...s s r r d di S S r r dt d dt di S S ,L r i i v v L i r v L i r v βαβαββααθλθλβα++=++= (1)()()()()βββαααθλθλi L dt i r v i L dt i r v ssrr ssr r ...sin ....cos .∫∫−−=−−= (2) Field oriented control of a PMSM is easily implemented because the rotor flux angle is locked to the rotor position. The rotor flux angle is not easily measured in an induction motor because it is not synchronized to the rotor shaft angle or speed. However, the above rotor flux angle estimation technique applies equally well to an induction machine [3]. In this case, the winding inductance term L s ’, in the motor circuit model also includes contributions from the rotor circuit (3).inductanceself rotor the is inductance g magnetizin the is inductance self stator the is where,2'r m S rms sL L L L L L L −= (3) This flux angle estimate enables direct field oriented control of an induction motor where the motor torque is proportional to the product of the rotor flux magnitude λr and the q-axis stator current i qs . The velocity estimate cannot be directly used to close the outer velocity loop because of rotor slip is required for the motor to deliver torque. However, the motor shaft velocity can be calculated under field orientation since there is a linear relationship (4) between the motor slip s and the q-axis stator current i qs .resistancerotor the is where,...r rqsm r r e r i L L r s λω=(4)The control schematic of the rotor flux estimator and the phase locked loop (PLL) is shown in Fig. 2. A vector rotation function calculates the error between the flux angle θr and the estimated angle θest . The integrator in the PLL feedback loop ensures that it also calculates both the rotor angle and velocity. Digital implementations support precision integration without drift however, special care is still required to avoid flux integrator saturation due to dc offsets in the current sampling circuits. The actual implementation includes low frequency cut off in the flux integrator functions with a matching low frequency correction function in the PLL. Good performance can be achieved down to about 5% of motor rated speed but theangle and velocity estimates are not reliable at lower speeds.Figure 2. Flux estimator and angle and frequency PLLIV. PMSM AND I NDUCTION MOTOR STARTINGStarting the PMSM is a challenge with sensorless control because the flux estimator does not provide information at zero speed. Starting the induction motor is easier but starting torque is low with open loop volts per hertz control. The approach tooptimized starting of both motors is similar and so a common starting controller is used. There is a three stage starting sequence: dc injection, open loop velocity ramp and closed loop operation. The startup controller, shown in Fig. 3, manages the operation of the PLL switching it from open loopto closed loop modes based on the flux magnitude output |λr | of the vector rotator. It also selects the reference inputs d-axis andq-axis current loops as a function of the operating mode.Figure 3. PLL starting controlThe dc injection phase, also know as parking, aligns the PMSM rotor to a known angle. The starting controller sets the initial value of the angle integrator to the parking angle θinit and the d-axis reference current i d * to the parking current i park . The flux integrator is also primed with initial rotor flux values λinit based on the parking angle. When the controller applies the parking current, the PM rotor aligns with the d-axis stator flux. The magnitude and duration of the parking current pulse is adjusted to match the shaft load characteristics. The parking phase maximizes the starting torque after the controller switches to open loop velocity ramp control and so improves starting reliability. In the case of the induction motor, the dc injection phase magnetizes the rotor core before starting. The parking current value is set to the motor magnetizing current while the parking time is matched to the rotor time constant. In the second phase, the q-axis reference current i q * is set at a fixed value i start in order to deliver a constant accelerating torque. The starting controller drives the integrator in the PI compensator with a fixed value matching the acceleration rate αinit set by the starting torque. The starting controller switches the PLL into closed loop mode when the motor reaches the minimum speed at which the flux integrator can reliably operate. The controller tests for start failure by comparing the rotor flux magnitude |λr | with predicted values.In ideal conditions with zero static torque, the dc injection current will park the PM rotor at an angle where the motor torque is zero. However, when there is non-zero static load, there will be a parking angle error. Fig. 4 helps demonstrate that the drive can tolerate relatively high static starting torque. The parking torque T park is a sine function of the angular distance from the parking position while the starting torque T park is a cosine function of the error in the parking angle. In the case where the static torque is equal to 50% of the maximum available parking torque the parking error εpark will be 30o (=sin -1(0.5)) which only reduces starting torque to 87% (=cos(30o )) of its maximum value. For most initial rotor angles,the parking torque will move the rotor towards the parking angle until the parking torque matches the static load torque. However, there is a dead zone around ±180o for initial rotor angles where the static torque will prevent any movement towards the parking position causing the initial starting torque to be in the reverse direction. A two stage parking scheme with a second parking angle advanced by 60o from the first positionwill avoid this problem.Figure 4. Parking error and the initial starting torqueIn a perfectly tuned system, the rotor angle will track the PLL angle in the open loop velocity ramp phase. The open loop starting acceleration is predicted from the starting current, the motor torque constant and the load inertia. However, the load inertia is difficult to characterize and real application loads have static friction. The starting process is tolerant to errors in the starting acceleration prediction as long as it is lower than the correct value. In this case, the initial acceleration will cause the rotor angle to move ahead of the angle predicted by the PLL resulting in a reduction in the generated torque. The system reaches equilibrium when the angle advances to the point at which the resultant acceleration falls to the predicted value. When the PLL switches to closed loop mode it will correct any angle offset angle generated during the open loop velocity ramp phase. V.S ENSORLESS AC MOTOR C ONTROL USING THE M OTIONC ONTROL E NGINEThe sensor-less field oriented control algorithms for both motor types are implemented on new digital control architecture described in Fig. 5 [4]. The ICs Motion Control Engine (MCE) includes digital control ASIC modules to implement the algorithm control calculations. MCE sequencer instructions stored in RAM enable flexibility while dedicated ASIC implementations of motor control functions significantly speed up algorithm execution. The MCE control modules are connected to hardware interface blocks that include a three-phase space vector PWM unit and an analog subsystem to measure the motor currents. The PWM timing unit generates the power inverter gating signals and the A/D sample timing signals required to reconstruct the motor winding current from the dc link current. The flux angle estimator, as described previously, calculates the rotor flux angle from the applied v α and v β voltages and measured i α and i β currents. This angledrives the vector rotation blocks that allow the current loop tobe closed in the rotating reference frame.Figure 5. Digital Control ICThe control schematics for the PMSM controller and the induction motor controller are shown in Fig. 6 and Fig. 7. The field oriented current control loops are identical in structure in both systems. The outer velocity loop is also identical for both drives; but for the induction motor, the slip frequency is subtracted from the rotor flux frequency to calculate the motor speed. The PMSM does not require field current so the d-axis current reference can be set to zero. However, a negative d-axis current enables extended speed range operation under so called “field weakening control”. The opposite is the case for the induction motor. The d-axis reference current is set to the magnetizing current at low speeds but is reduced at higher speeds to enable field weakening control. The field weakening controller determines the d-axis current that will optimize the use of the dc bus based on the outputs of the currentcompensators.Figure 6. Sensorless PMSM drive algorithmFigure 7. Sesorless IM drive algorithmGraphical design tools supporting the IRMCK341 digital control IC enables easy modification of the outer velocity loop and field control functions. These tools have also been used by washing machine manufacturers to add special functions to support load out of balance detection before entering the spin cycle. A hardware reference design demonstrates recommended circuit layout for the dc link current signal. Thisis the most critical part of the circuit design because this signalis the only source of information on motor current, rotor flux position and velocity.VI.WASHING MACHINE M OTOR P ERFORMANCECOMPARISONThis section describes results for a PMSM and an induction motor evaluated for use in a front loader washer. Both motors met the washer application requirements including torque speed range and speed accuracy using sensorless control. Fig. 8 shows an example of the variation in the PMSM motor current during a wash cycle. The current varies significantly during the cycle as the drum lifts the cloths then lets them fall to the bottom of the drum. The current variation is quite unpredictable during the cycle depending on how the cloths become lumped together. The test data presented here compares starting performance and efficiency during washing. Both motors run the washer drum via a 15:1 belt drive where the typical washing speed is 47 RPM corresponding to a motor speed of approximately 700 RPM. The maximum spin speed of 1100 RPM requires a motor shaft speed of 16,500 RPM which iswell over 3 times base speed.Figure 8. Motor currents during wash cycleDynamometer test data is presented to allow a comparison in drive performance during washing. Fig. 9 and Fig. 10 show the starting performance of the PMSM and IM with a target speed of 470 RPM. The PMSM achieves a staring torque of 2.7 Nm, which exceeds the application requirements. The maximum motor starting current is 6 Arms and the total starting time was 2 seconds. The induction motor achieves a higher starting torque of 3.4 Nm with a maximum starting current of 5.6 Arms and a starting time of less than 1 second. Both motors met the application specification but the inductionmotor performed better.Figure 9. PMSM startup test with 2.7 Nm loadFigure 10. Induction motor starting with 3.4 Nm loadFig. 11 and Fig. 12 show the motor efficiencies at washing speed over the application torque range. The induction motor efficiency drops to less than 40% at the maximum torque range while the PMSM efficiency is still above 70% at this point. The inverter efficiency is almost 90% at the maximum torque range for both motors. This is expected since the rated currents for both motors are almost same.Figure 11. PMSM efficiency at 700 RPMFigure 12. Induction motor efficiency at 700 RPM The motor efficiency is a minor contributor to washing machine energy efficiency rating since the energy consumed by the motor is still significantly less than the energy consumed in heating the hot water used. However, since the PMSM is more efficient it can be made smaller and therefore may cost less to build than the induction motor. The PMSM efficiency is a little lower at spin speeds (16500 RPM) because field weakening consumes most of the current but this is still higher than the induction motor. It was not possible to take dynamometer efficiency measurements for the induction motor at spin speeds but the drive operates in this mode for a relatively short time.VII.W ASHING M ACHINE C ONTROL S YSTEM While the motor control function is a key part of the washing control system, the washing machine application software adds most the appliance product features. The IRMCF341 washing machine control IC integrates an independent 8-bit microcontroller for the application layer functions. This solution preserves all the advantages of having an independent application layer processor that can control the relatively slow washing process without having to compete for system resources required by the faster motor algorithm. The 8-bit microcontroller loads the control loop parameters into the dual port RAM and special control registers in the ASIC control modules. It also loads sequencing code into the MCE program memory to define the motor control algorithm. This makes is possible switch between induction motor control and PMSM control algorithms without changing the drive hardware. The 8-bit microcontroller can also access the internal algorithm variables such as speed and torque current required to implement washing machine functions such as wash load detection or out of balance measurements. In the case of the PMSM drive, it is possible to optimize starting parameters to match different wash load conditions and so minimize starting currents.VIII.C ONCLUSIONThe paper presented two implementations of field oriented control using a common control platform. Both the induction motor and the PMSM meet the washer application requirements. The induction motor has a higher starting torque than the PMSM but has a much lower efficiency in washing mode. The use of sensorless field oriented control using dc link current feedback enables the use of a single drive platform to control either motor.R EFERENCES[1]Bianchi, A. and Buti, L. “Three-Phase A.C. Motor Drive and Controllerfor Clothes Washer” Appliance Magazine, June 2003[2]Krause, P.K., Wasynczuk, O. and Sudhoff, S.D. “Analysis of ElectricMachinary and Drive Systems” 2nd Edition, IEEE Press 2002[3]Novotny, D.W. and Lipo, T. A. “Vector Control and Dynamics of ACdrives” Oxford University Press 1996 pp 257-282[4]Murray, A. and Ho, E. “New Motion Control Architecture SimplifiesWashing Machine Control System Development” IAS 2006 Conferencer Record pp 1229-1234。