You are on page 1of 8

Electrical Engineering (2006) 88: 395402 DOI 10.

1007/s00202-005-0301-7

O R I GI N A L P A P E R

zdemir mu s Bilal Gu Mehmet O

Sensorless vector control of a Permanent magnet synchronuous motor with fuzzy logic observer

Received: 24 December 2004 / Accepted: 12 March 2005 / Published online: 15 June 2005 Springer-Verlag 2005

Abstract In this paper, a new method for sensorless vector control of a permanent magnet synchronous motor (PMSM) using a fuzzy logic observer is developed. This method is based on determination of rotor position and thereby speeds by estimating back-emf components which result from fuzzy logic observers. The rotor position angle and rotating speed are estimated by evaluating the instantaneous values of stator voltages and currents. The estimators are two fuzzy logic observers. They have two inputs: the estimated stator currents and the dierence between the measured and estimated stator currents. In addition, the outputs of the fuzzy logic observers are the back-emf components in an ab reference frame. The proposed method was implemented using a MATLAB/Simulink software package program. The obtained results are within acceptable error limits for a wide speed range, from 40 rad/s up to 500 rad/s. Keywords Permanent magnet synchronous motor Sensorless control Fuzzy logic observer

1 Introduction
Permanent magnet synchronous motors (PMSM) are frequently used in industrial applications. Especially their compact design, high eciency, high power/weight and torque/inertia ratios can be shown as the most important advantages of PMSMs. On the other hand, the high cost and their time-varying magnetic characteristics are the disadvantages of PMSMs [18]. PMSMs
B. Gu (&) mu s Department of Electrical & Electronics Engineering, Dicle University, Engineering and Architecture Faculty, 21280 Diyarbakr, E-mail: bilgumus@dicle.edu.tr zdemir M. O Department of Electrical & Electronics Engineering, Engineering Faculty, Frat University, Elazg E-mail: mozdemir@rat.edu.tr

can be divided into two main groups with respect to the placement of their magnets on the rotor. PMSMs are classied as surface- mounted PMSM (SPMSM) and interior PMSM (IPMSM) if their magnets are mounted on the surface of the rotor and inside the rotor, respectively. In this work, SPMSM has been used. The vector control technique is one of the most important closed loop techniques for AC machines in variable speed applications. In vector control applications, speed and rotor position can be obtained by using conventional electromechanical sensors such as tachogenerators and encoders mounted on the rotor. However, this increases the dimensions and cost of the driver system. [14, 6, 79]. The need for using sensors can be eliminated if the speed and position of the rotor are estimated. This is called sensorless control technique. The main aim of sensorless control technique is to estimate the rotor position and hence the speed using machine parameters and measured voltages and currents. In PMSMs employing a back-emf-based sensorless scheme, the information on the rotor position is extracted from the estimated back-emf waveforms by means of inverse trigonometric functions [4, 7]. The estimated position and speed of the motor are used in close loop control applications. There are various developed estimating methods such as sliding estimators [2] or voltage injection through the machine [9]. Fuzzy systems have been used extensively and successfully in control systems over the past few decades [10]. Fuzzy logic is an intelligent control method which has proved to be convenient in some applications and has successful results when compared with classic control methods [10]. Fuzzy logic processing can be suggested to achieve a better approximation of the absolute rotor position when measurement uncertainties are present and there is a lack of a precise law of variation for the current dierences in the PMSM control. In this study, a new method using a fuzzy logic observer for sensorless vector control of the PMSM is presented for estimating the position with acceptable error values in a wide speed range.

396

the electrical speed. Supposing that the machine is sinusoidal, the induced voltage has the following form: 2 3 sin he ~ m 4 sin he 2p 5 xe W ~ m K he : ~ E xe W 2 3 4p sin he 3 In the PMSM case, the torque is expressed by: ~ m K he ; ~ Is t W Te p ~

where Is is stator currents matrix and p is the number of pole pairs. The torque equation can be rewritten as
Fig. 1 Phasor diagram of SPMSM

~ m ~ ~ T e pW I a Ka h ~ Ib Kb h ~ Ic Kc h

2 Mathematical model of PMSM


A one-phase electrical equation of the machine can be given as [5]: dW d v Zi Ri Ri Li Wm h; 1 dt dt where Wm corresponds to the amplitude of the natural magnetic ux of the permanent magnets. The term d the back-emf (induced voltage) dt Wm h corresponds to m h and can also be written dW dh xe ; where xe corresponds to

If the sinusoidal stator currents are replaced in Eq. 4 then the torque expression becomes [4]: 3 ~m ~ ~ Te p W Is 2 If the mechanical equation of the machine is: j dxm Kd xm Tl Te dt 6 5

Here, Kd is the friction coecient, Tl represents load torque and Te the induced torque.

Fig. 2 Sensorless vector control scheme (of PMSM)

397

Fig. 3 Input, output functions of the proposed fuzzy logic observer

Fig. 5 Scaled membership functions of ia ^ ia input function

Fig. 4 Scaled Leave single space scaled membership functions of ia input function

The mathematical model of a PMSM can be dened in the dq frame by the following equations [4]: ! ! ! ! vd Rs pLd id xLq 0 7 vq xe Ld R pLq iq KE xe ; where td ;tq and id, iq determine dq axis voltages and currents, respectively. Rs is the stator resistance, xe is the electrical angular speed, KEis the back-emf constant, p is the dierential operator and Ld and Lqcorrespond to the dq axis inductances. In a SPMSM, Ld Lq L: 8

Fig. 6 Scaled membership functions of the ea output function

! ! ! ! va ia 0 sin he Rs pL KE x e ; vb 0 Rs pL ib cos he

The general representation of Eq. 7 constitutes the mathematical model of an interior mounted PMSM. Consequently Eq. 7 can be used for a surface mounted PMSM by replacing L with Ld and Lq. The equation in the ab reference frame for SPMSM is obtained by considering Eq. 8 as:
Table 1 Rule base of proposed fuzzy logic observer INPUT 1 INPUT 2 NL NL NM NKS Z PS PM PL X NL NM NM NM NL NL

where, ta and tb represent ab axis voltages, respectively. This equation can be rewritten as ! ! ! ! ia va ia sin he pL K E xe ; 10 R vb ib ib cos he |{z} |{z} |{z}
A B C

where, the portion C of the equation represents the back emf [4]. The phasor diagram for SPMSM is illustrated in Fig. 1.

NM NL NM NM NS NM NM NL

NS NM NS NS NS NS NS NM

Z Z Z Z Z Z Z Z

PS PM PS PS PS PS PS PM

PM PL PM PM PS PM PM PL

PL PL PL PM PM PM PL X

NL Negative large, NM Negative medium, NS Negative small, Z Zero, PS Positive small, PM Positive medium, PL Positive large

398 Fig. 7 The general structure of the proposed position estimator

Fig. 8 Simulation results of sensorless vector control. a Rotor speed, b Stator currents, c Induced torque, d id and iq currents

399

3 Sensorless vector control of SPMSM


The block representation of a sensorless vector control structure for PMSM is shown in Fig. 2. Here, the main dierence between sensorless control and the classical vector control scheme [8] is that in the sensorless control, the speed sensor is removed and replaced by a position estimator. The vector control method shown in Fig. 2 imposes a rotor ux vector angle he which is aligned to the d-axis. In the case of PMSM, he is equal to the instantaneous ux vector angle since there is no slip frequency and reference isd is taken to be zero. A current-controlled voltage source inverter was used as a driver, since the power value of the motor is low. Here, a hysteresis band current controller with xed switching frequency is chosen and the ideal switches are used in the simulation. The switches frequency is taken to be 10 kHz with a hysteresis current band of 0.02 A.

3.1 Fuzzy logic observer Two fuzzy logic observers are proposed for this operation. Each obtains one of the components of back emfs. The fuzzy logic observer has two inputs and one output. One of the inputs is the estimated current component and the other is the dierence between the estimated and measured currents. The output function is induced back emf. Seven membership functions were used for the two inputs and output (Fig. 3). Limits of the membership functions were determined by considering the parameters of PMSM. The general structure of the fuzzy logic observer is given in Fig. 4. Membership functions for ^ ia current is shown in Fig. 4 where triangular membership functions have been chosen and scaled in the range of 3.6 A. These are the maximum values for PMSM currents. Similarly, membership functions for the other input ia ^ ia are given in Fig 5. Here, the membership function was scaled in the

Fig. 9 Simulation results for 40 rad/s reference speed with fuzzy logic observer: a Estimated and measured positions, b error between measured and estimated positions, c a component of back-emf output from fuzzy logic observer, d b component of back-emf output from fuzzy logic observer

400

Fig. 10 Simulation results for 300 rad/s reference speed with fuzzy logic observer: a Estimated and measured positions, b error between measured and estimated positions, c a component of back-emf output from fuzzy logic observer, d b component of back-emf output from fuzzy logic observer

range of 7.2 A for the worst case. Functions for the b-axis currents were formed in the same way. In addition, ^ ea and ^ eb output functions of the fuzzy logic observer were formed from seven triangular membership functions scaled in the range of 18 V as shown in Fig. 6. The rule base for the fuzzy logic observer was constructed by evaluating the truth degrees of the estimated and measured currents. The rule base for both observers is the same and is shown in Table 1. Rules were extracted using the Mamdani method. In the table, two rules are ignored because they make no sense. In the observer, the centroid method is used as the defuzzication method. The position estimator shows similarities with the sliding mode estimations. But in the proposed method, back emfs are obtained from the fuzzy logic observer instead of the machine model.

3.2 General structure of the position estimator It is known that back-emf terms contain the rotor position information as given in Eq. 9. Hence, it is essential to obtain these back emfs in the ab reference frame to calculate the rotor position. The block diagram of the proposed method is given in Fig. 7. The relationships between va, vb and the estimated currents, ^ ia ; ^ ib ; and voltages, ^ ea ; ^ eb ;can be given as d^ ia va Rs ^ ^ ea ia ; dt L L L 11 ^ d^ ib vb Rs ^ eb ib : dt L L L eb are obtained from Here, back-emf components, ^ ea ; ^ the output of fuzzy logic observers, then estimated currents ^ ia ; and ^ ib are calculated using Eq. 10. In the closeloop operation, the currents ^ ia ; ^ ib and voltages ^ ea ; ^ eb are

401

Fig. 11 Simulation results for 500 rad/s reference speed with fuzzy logic observer: a estimated and measured positions, b error between measured and estimated positions, c a component of back-emf output from fuzzy logic observer, d b component of back-emf output from fuzzy logic observer

obtained for every sampling period. The estimated position information, he can be calculated as   ^ ea 1 ^ 12 he tan ^ eb So the determination of position was done by the fuzzy logic observer that works to decrease the error with feedback. The estimated position information varies between p/2 and p/2. A Matlab function was written for summing the position values by considering the starting point as reference in order to avoid instability. The same function is used for reversal speed operation by modifying the sign of position value. Velocity estimation is performed by the help of variation in position information obtained by this function. A longer sampling period can be used in speed estimation because of the fact that speed variation is much slower than position

variation. The proposed system eliminates analytic solution of the higher order system, error determination and correction methods. Error correction methods are quite complex [14, 6, 79]. The purpose of this study is to overcome such diculties by using the proposed method.

4 Simulation results
The general structure of PMSM, including the sensorless vector control, is implemented in the Matlab/ Simulink software package program. Therefore, the proposed sensorless control technique has been simulated for various speed references to demonstrate the performance of the system. Figure 8 shows the simulation results obtained for sensorless vector control using the fuzzy logic observer. A sinusoidal reference speed of 16.6 Hz with a

402

magnitude of 500 rad/s was given to demonstrate the performance of the system in four-quadrant regions. The initial position estimation in Fig. 8 was taken as zero. The motor speed follows the reference speed as shown in Fig. 8a. Three-phase stator currents are illustrated in Fig. 8b. The reference torque shown in Fig. 8c is not saturated due to the sinusoidal change in the reference speed. Figure 8d shows dq currents where the torque component current follows the reference torque. The back-emf components, ea, eb, which are estimated by the fuzzy logic observer, the position information calculated from these back emfs, and the error between the measured and estimated position values for various speeds are illustrated in Fig. 9 and Fig. 10. Figure 9 shows the measured and estimated positions for 40 rad/s which can be accepted as a low rotating speed for PMSMs. It can be seen that the maximum error between the measured and estimated positions is not more than 20. The proposed control algorithm makes position estimation acceptable at low speed ranges where sensitive position control is not required. In Fig. 10, it can be seen that the maximum electrical position error is approximately 10. The proposed fuzzy logic observer makes the estimation at high accuracy at medium and high-speed ranges. Therefore, this error results in negligible speed variations. Similar results are given in Fig. 11 for a speed range of 500 rad/s. As discussed above, the error rate of the position information determined by the proposed method does not exceed 20 for a wide speed bandwidth [40500 rad/s]. Mechanical rotor position error in [9] is more than 20. In this work, the maximum electrical position error was approximately 20 which corresponds to 10 of mechanical rotor position error. Hence, these error values are acceptable as bad speed conditions for most of the PMSM applications.

parameters. So the error that occurs in the system is decreased by this feedback. In this work, the position estimation has been done with acceptable error in the wide speed range from 40 rad/s to 500 rad/s. In this speed interval, the maximum electrical position error was approximately 20 and the error did not exceed 5 for most values of the position. If the number of the poles of the machine is taken into consideration, it would be seen that the mechanical position error becomes smaller since the electrical position is twice that of the mechanical position. It can be observed that the proposed method can make estimations with a high accuracy ratio for a wide speed interval.

Appendix
Table 2 Motor parameters
P W RS (W) (rpm) (X) Ld (H) J (kgm2) Kd p Un (Nms) (V) Tm Ke (Rated (Wb). torque) (Nm) 0.0246

75

8,000

5.25 0.46103 9.9107 0

2 19.1 0.029

References
1. Batzel TD, Lee KY (2000) Slotless permanent magnet synchronous motor operation without high resolution rotor angle sensor. IEEE Trans Indust App 15(4):366376 2. Consoli A, Scarcella G, Testa A (2001) Industry application of zero-speed sensorless control techniques for PM synchronous motors. IEEE Trans Indust App 37(2):513519 3. Corley MJ, Lorenz RD (1998) Rotor position and velocity estimation for a salient- pole permanent magnet synchronous machine at standstill and high speeds. IEEE Trans Indust App 37(4):784789 4. Chen Z, Tomita MD, Doki S, Okuma S (2003) An extended electromotive force model for sensorless control of interior permanent-magnet synchronous motors. IEEE Trans Indust App 50(2):288288 5. Simon T (1999) Implementation of a speed eld oriented control of 3-Phase PMSM motor using TMS320F240. TI Application report SPRA588, Texas Instruments 6. Tursini M, Petrella R, Parasitili F (2003) Initial rotor position estimation method for PM motors. IEEE Trans Indust App 39(6):16301640 7. Kojabadi HM, Ahrabian G (2000) Similation, analysis of the interior permanent magnet synchronous motor as a brushless AC drive. Simulation Practice and Theory 7:691707 8. Su nter S, Altun H (2005) Control of a permanent magnet synchronous motor fed by a direct ACAC converter. Elect Eng 87(2):8392 9. Tursini M, Petrella R, Parasitili F (2003) Initial rotor position estimation method for pm motors. IEEE Trans Indust App 39(6):16301640 10. Simon D (2000) Design and rule base reduction of a fuzzy lter for the estimation of motor currents. Int J Approx Reasoning 25

5 Conclusions
Methods depending on the complex and some constant parameters are used in order to decrease errors that occur in the position estimation method used in the sensorless drive systems. It is necessary to determine the correction coecients in these methods for each system. It was seen that improper choosing of the coecients leads to some errors in the system. However, they were eliminated without the necessity for correction coecients by the proposed method based on fuzzy logic. Moreover, the disadvantages of using dierent estimation methods for lower and higher speeds are eliminated. In the proposed method, the dierences between the estimated currents, which are determined by means of back emfs, and the measured currents are used as input

You might also like