HIGH PERFORMANCE DRIVES 100 ---------------------------------------------------------------------------------------------------------------------------------------

6. SPEED ESTIMATION FOR SENSORLESS HIGH PERFORMANCE VECTOR CONTROL

6.1.

INTRODUCTION

It has been assumed so far in all the considerations of induction machine and SPMSM vector control that a speed (or position) sensor is available and that it provides the required feedback signal for closed loop speed control (and the speed/position information for co-ordinate transformation, where required). Sensors mounted on the machine shaft are in general not desirable for a number of reasons. First of all, their cost is substantial. Secondly, their mounting requires a machine with two shaft ends available - one for the sensor, and the other one for the load coupling. Thirdly, electrical signals from the shaft sensor have to be taken to the controller and the sensor needs a power supply, and these require additional cabling. Finally, presence of a shaft sensor reduces mechanical robustness of the machine and decreases its reliability. It is for all these reasons that substantial efforts have been put in recent past into possibilities of eliminating the shaft mounted sensor. However, the information regarding actual speed and/or position of the rotor shaft remains to be necessary for closed loop speed (and/or) position control (and co-ordinate transformation, if applicable) even if the shaft sensor is not installed. Hence the speed (and/or position) has to be estimated somehow, from easily measurable electrical quantities (in general, stator voltages and currents). All the schemes of vector control, introduced in Chapters 3 and 4, are valid regardless of whether the machine speed (and/or position) is measured or estimated. This Chapter therefore discusses only issues related to speed estimation. An induction machine is under consideration at all times, although the same approaches as those explained in what follows are in general applicable (with some modifications) to permanent magnet synchronous machines as well. A drive in which the speed/position sensor is absent is usually called sensorless drive, where sensorless symbolises absence of the shaft sensor. However, the sensors required for stator current measurement (and, in many cases, stator voltage measurement as well) remain to be present so that the term sensorless is somewhat misleading. Sensorless vector control of an induction machine has attracted wide attention in resent years. Many attempts have been made in the past to extract the speed signal of the induction machine from measured stator currents and voltages. The first attempts have been restricted to techniques which are only valid in the steady-state and can only be used in low cost drive applications, not requiring high dynamic performance. Different, more sophisticated techniques are required for high performance applications in vector controlled drives. In a sensorless drive, speed information and control should be provided with an accuracy of 0.5% or better, from zero to the highest speed, for all operating conditions and independent of saturation levels and parameter variations. In order to achieve good performance of sensorless vector control, different speed estimation schemes have been proposed, so that a variety of speed estimators exist nowadays. In general, all the existing speed estimation algorithms belong to one of the following three groups: 1. speed estimation form the stator current spectrum; 2. speed estimation based on the application of an induction machine model; 3. speed estimation by means of artificial intelligence techniques (artificial neural networks and fuzzy logic). In this Chapter, basic principles of some of the simplest speed estimation schemes will be reviewed and discussed. Only methods belonging to the first two groups are covered. Speed estimation from stator current spectrum, by utilising various parasitic effects (rotor slot harmonics, saliency etc.), is elaborated in Section 6.2. Majority of existing speed estimation schemes are based on the induction machine model.
E Levi, 2001

HIGH PERFORMANCE DRIVES 101 ---------------------------------------------------------------------------------------------------------------------------------------

Hence detailed discussion of induction machine model based speed estimation occupies the remainder of the chapter. In Section 6.3, speed estimation schemes based purely on an induction machine model are reviewed. All the schemes discussed in Section 6.3 belong to so-called open-loop speed estimation group of methods (the meaning of this definition will be addressed later). Closed loop speed estimation is reviewed in Section 6.4. Model reference adaptive control (MRAC) based method is the one discussed in detail, since it is the simplest one. Finally, Section 6.5 illustrates the performance of the MRAC based speed estimator within a sensorless vector controlled induction motor drive.

6.2.

SPEED ESTIMATION FROM STATOR CURRENT SPECTRUM

Good feature of all the methods that belong to this group is that speed estimation is not affected by variation of the machines electrical parameters. On the other hand, the most serious shortcoming of these methods is the need for extensive signal processing, since in general Fast Fourier Transform (FFT) has to be executed on-line. This at the same time reduces the accuracy of speed estimation during rapid transients. Stator current harmonics arising from rotor mechanical and magnetic saliencies, such as rotor slotting, rotor eccentricity and magnetic saturation, are speed dependent. They are independent of time-varying machine parameters and exist at any non-zero speed. Digital signal processing and spectral estimation techniques have made it possible to extract these harmonics with minimal data collection and processing time. Therefore, the slip frequency and rotor speed of the machine can be estimated by utilising, for example the rotor slot harmonics, monitored from the stator currents. Many schemes which employ certain harmonics for slip frequency and rotor speed estimation have been reported. These methods can be divided into two major groups:

1) those that use the fundamental excitation to estimate the rotor speed, and 2) those that use a separate excitation signal from the fundamental excitation to estimate the
rotor speed. Only one method that belongs to the first group, speed estimation on the basis of rotor slot harmonics, is discussed in what follows. The rotor slot harmonics can be detected by using monitored stator currents or stator voltages. As monitoring of stator currents is always required anyway, detection of rotor slot harmonics by using stator currents is preferred, so that voltage monitoring can be avoided. A speed estimation scheme that uses FFT technique to extract information regarding rotor slot harmonics has been proposed for the first time in 1992. Presence of rotor slotting causes existence of certain harmonics in the stator current spectrum. By performing the FFT of the stator current, one is capable of extracting the frequency information related to the slot harmonics, so that estimation of the rotor speed becomes possible. Speed is estimated by decomposition of stator current signal into its harmonic components to determine the speed dependent slot harmonic frequency, fsh, and the machine fundamental frequency fe. The basic principle of this scheme is as follows. For a given supply frequency fe, the speed of the fundamental rotating field is constant and is the synchronous speed. It is given in revolutions per minute (rpm) by the well-known equation:

ne =

60 f e P

(6.1)

where ne is the synchronous speed and P is the number of pole pairs. The slip, s, is the difference between the actual speed n and the synchronous speed ne. It is often expressed as a fraction of the synchronous speed as:

s=
E Levi, 2001

ne n ne

(6.2)

HIGH PERFORMANCE DRIVES 102 ---------------------------------------------------------------------------------------------------------------------------------------

It follows that n=(1-s)ne and f=(1-s)fe Angular frequency of slot harmonics is given with: (6.4) (6.3)

sh =

Z e P

(6.5)

where Z is the number of rotor slots. Note that the number of rotor slots is normally not known and this is one of the serious drawbacks of this method. In order to implement the method, it is necessary to somehow at first acquire the information regarding the number of rotor slots. Using (6.1) to (6.5), the rotor speed n in rpm is expressed and calculated by the following expression:

n=

60 ( f fe ) Z sh

(6.6)

where fundamental harmonic frequency is obtained from the FFT analysis as well. The scheme can operate quite satisfactorily in a wide speed range, down to 2 Hz. The rotor slot harmonic signal is however nowadays regarded as insufficient in bandwidth as speed feedback signal in a high performance drive. Apart from rotor slot harmonics, numerous other parasitic effects can be utilised for extraction of the speed estimate from the stator current spectrum. These include rotor mechanical eccentricities, existing magnetic saliencies, and even deliberately introduced rotor asymmetries. These will not be discussed in more detail, since their industrial relevance at present is limited.

6.3.

OPEN LOOP SPEED ESTIMATION USING INDUCTION MACHINE MODEL ONLY

In the context of speed estimation based on an induction machine model, the term open loop speed estimation means that the speed estimation purely relies on the equations of an induction machine model. In other words, a corrective action within the speed estimator is not present. If there is certain corrective action within the model based speed estimator, such an estimator is termed closed loop speed estimator (see next Section). Note that the meaning of open loop and closed loop in this context is not in any way related to the speed control loop of the drive - this loop is always closed and that is precisely the reason why the speed is estimated in the first place. The first attempt to operate the induction machine with closed loop speed control but without using a speed sensor was based on an analogue slip calculator that computed the slip frequency and dates back to 1975. The slip frequency is the difference between the stator frequency and the electrical frequency corresponding to rotor speed. By calculation of the slip frequency, the speed of the rotor can be determined. The slip information is obtained by measuring the electrical quantities applied to the machine. By performing simple signal processing operations on the measured quantities, an analogue signal proportional to the slip level is derived and used to control the machine. This scheme is applicable only in steady-state, in a limited speed range, and is therefore inappropriate for high performance vector control. During the last couple of years, several open-loop rotor speed estimation methods were developed for sensorless vector control of induction machine. Calculation of the rotor speed is based on the induction machine dynamic model. Rotor speed is calculated as the difference between the machines synchronous electrical angular speed and the angular slip frequency. In other words,

= r sl
E Levi, 2001

(6.7)

HIGH PERFORMANCE DRIVES 103 ---------------------------------------------------------------------------------------------------------------------------------------

where is the rotor speed, r is the speed of rotor flux and sl is the angular slip frequency. In a rotor flux oriented controlled induction machine, it is possible to obtain the angular slip frequency by using the rotor voltage equation of the machine in the rotor flux oriented reference frame. The angular slip frequency can be calculated from:

sl =

Lm iqs Tr r

(6.8)

where iqs can be obtained from the torque equation as:

iqs =

2Te Lr 3 P Lm r

(6.9)

Substitution of equation (6.9) into (6.8), considering that r= dr in the rotor flux oriented reference frame and that

Te =
yields

3 Lm ( i r is ) P 2 Lr r s

(6.10)

sl =

Lm 1 i r is Tr r2 r s

(6.11)

The electrical angle r of the rotor flux vector is defined as:

r = tan 1

r r

(6.12)

The derivative of the angle equation (6.12) can be used to obtain the electrical angular speed of the rotor flux. Therefore,

r =

dr = dt

d r

dt 2r + 2r

d r dt

(6.13)

If the rotor flux components are known, the electrical angular speed of rotor flux can be calculated by using equation (6.13). It is convenient to estimate the rotor flux components from the stator voltage equations. Derivatives of the rotor flux components can be then given as:

d r L di = r vs Rs is Ls s dt Lm dt d r dt
Hence

di s L = r v s Rs i s Ls Lm dt
Lr Lm Lr Lm

(6.14)

r = r =

[ (v

Rs is ) dt Ls is

[ (v

Rs i s dt Ls i s

(6.15)

2 r = 2r + r

E Levi, 2001

HIGH PERFORMANCE DRIVES 104 ---------------------------------------------------------------------------------------------------------------------------------------

Electrical angular speed of rotor flux, given with (6.13), and angular slip frequency (6.11) are thus calculated using (6.14)-(6.15) and measured stator voltages and currents. Finally, the rotor speed is estimated from:

= r sl =

d r

dt 2 2r + r

d r dt

Lm 1 i r is Tr r2 r s

(6.16)

Figure 6.1 shows the diagram of implementation of the speed estimation scheme based on the equations given above. The inputs are stator currents and stator voltages in stationary reference frame. Stator currents can be measured from the machine terminals. Stator voltages can be measured from the machine terminals or reconstructed from the inverter switching states and measured DC link voltage. Summarising the described procedure, the following equations are utilised:

d r L di = r vs Rs is Ls s dt Lm dt d r dt = di s Lr v s Rs i s Ls Lm dt

r = r =

Lr Lm Lr Lm

[ (v

Rs is ) dt Ls is

[ (v

Rs i s dt Ls i s

]
Lm 1 i r is Tr r2 r s

(6.17)

2 r = 2r + r

= r sl =

d r

dt 2 2r + r

d r dt

Note that all the quantities in (6.17) are in stationary reference frame (that is, alfa-beta components). Stator current and voltage alfa-beta components are obtained directly from measured phase currents and voltages, using constant parameter 3/2 transformation (that does not require any angle) of (3.19) and (3.24). The problems encountered in the implementation of this scheme are two-fold. Firstly, since it is model based, accuracy of speed estimation is affected by parameter variation effects. Secondly, the scheme involves pure integration that fails at very low and zero frequency due to offset and drift problems. This kind of speed estimator works without failure above 10% of rated synchronous speed. An alternative speed estimation method is again based on the induction machine voltage equations and the flux equations in stationary reference frame. The rotor speed can be calculated directly using these equations. From the machine voltage equations and flux equations the rotor current components in stationary reference can be expressed as a function of the stator flux:

ir = i r

1 ( Lsis ) Lm s 1 = Ls i s Lm s

(6.18)

E Levi, 2001

HIGH PERFORMANCE DRIVES 105 --------------------------------------------------------------------------------------------------------------------------------------Lm Tr

r
vs is is
p Rs Rs p Lr Lm

sl r
-

Ls

Ls

vs

Lr Lm

r2 r2

Fig. 6.1 - Block-diagram of an open-loop speed calculation method.

From rotor voltage equation, eliminating the rotor resistance Rr, it is possible to obtain the rotor speed as follows:

ir

d r dt dt ir r + i r r i r Ls is ) d r dt s Ls is d r dt

d r

(6.19)

Substitution of equations (6.18) into rotor speed equation (6.19) enables rotor speed to be expressed as:

Ls is

( ) + (
r

Ls i s

) )

(6.20)

where stator flux components and rotor flux components can be estimated by means of the following equations:

s = s =
r = r

(v

(v

s s

Rs is )dt Rs i s dt

(6.21)

Lr L L s s r is Lm Lm

L L L = r s s r is Lm Lm

(6.22)

Problems related to practical application of this method are the same as those stated in conjunction with the previous method. The open-loop speed estimation methods, reviewed here, are simple to implement. However, it should be noted once more that the accuracy of open-loop speed estimators depends greatly on the accuracy of the machine parameters used. In general, at low speed the accuracy of open-loop speed estimators is

E Levi, 2001

HIGH PERFORMANCE DRIVES 106 ---------------------------------------------------------------------------------------------------------------------------------------

reduced. Furthermore, the integration problem makes application of these schemes at zero and very low speed impossible.

6.4.

CLOSED LOOP SPEED ESTIMATION BASED ON INDUCTION MACHINE MODEL

In Section 6.3 open-loop speed estimation methods have been elaborated. These have utilised the stator and rotor voltage equations of the induction machine. However, the accuracy of these open-loop schemes depends strongly on the machine parameters. In closed-loop speed estimators the accuracy of estimation can be improved. This is achieved by introducing certain corrective action, based on an error between two conveniently chosen quantities, within the speed estimator. There are three basic types of closed-loop machine model based speed estimators. The first one is the model reference adaptive control (MRAC) based speed estimator. The second one is the speed estimator based on an observer. The third type of speed estimation relies on extended Kalman filter (EKF) technique. MRAC based speed estimation is the simplest approach and is therefore the one analysed in detail in what follows. The measured signals that are used in MRAC based speed estimators (and in all other model based approaches) are stator phase voltages and currents. Alternatively, stator voltages are reconstructed from measured DC link voltage and inverter switching functions. The model reference approach makes use of two independent machine models of different structure to estimate the same state variable on the basis of different sets of inputs variables. The estimator that does not involve the quantity to be estimated (in this case, the rotor speed) is considered as a reference model. The other estimator, which involves the estimated quantity, is regarded as an adjustable model. The error between the outputs of the two estimators is used to drive a suitable adaptive mechanism that generates the estimated rotor speed for the adjustable model. When the estimated rotor speed in the adjustable model attains the correct value, the difference between the output of the reference model and the output of the adjustable model becomes zero. The estimated rotor speed is then equal to the actual rotor speed, under ideal conditions. The idea of a MRAC based speed estimator is illustrated in Fig. 6.2.
vs is Reference model

A(1)

Error calculation Adjustable model A(2)

PI contr.

est

Fig. 6.2 - Conceptual block diagram of MRAC estimator.

The outputs of the reference and the adjustable model in Figure 6.2 are denoted with A and they depend on which quantity is used for speed estimation. The most frequently used scheme has rotor flux space vectors at the output of the reference and the adjustable model and this is the scheme that is discussed further on. However, other solutions are possible as well. The outputs of the two models may be back emf, reactive power or air-gap power. As already noted, MRAC based speed estimation method makes use of two independent models which are constructed to estimate the same quantity, using measured stator voltages and currents. The two models can be obtained from the induction machine model equations. One of them does not involve the quantity that is to be estimated (in this case rotor speed) and is regarded as the reference model of the
E Levi, 2001

HIGH PERFORMANCE DRIVES 107 ---------------------------------------------------------------------------------------------------------------------------------------

speed estimator. The second one does involve the quantity that is to be estimated, and it is regarded as the adjustable model within the speed estimator. The error between the outputs of the reference model and the adjustable model is then used as an input into a suitable adaptation mechanism, that turns out to be a simple PI controller. The estimated rotor speed is obtained as the output of the PI controller. Rotor flux based MRAC method was proposed in 1992. In this scheme the outputs of the reference and the adjustable models are two estimates of the rotor flux space vector, that are obtained from the machine model in the stationary reference frame. From (2.36) one has by letting a = 0 the following two space vector equations:
v s = Rs i s + d 0 = Rr i r + d
r s

dt dt j

(6.23)
r

Elimination of the stator flux vector and rotor current vector enables rotor flux vector to be expressed in the form of:
d dt d dt
r r

Lr Lm 1 Tr

v s R s i s L s + j +
r

di s dt

(6.24)
Lm Lr is

The first equation of (6.24) can be used to calculate rotor flux space vector on the basis of the measured stator voltages and currents. The equation is independent of rotor speed and it therefore represents the reference model of Fig. 6.2. On the other hand, calculation of rotor flux from the second equation of (6.24) requires stator currents only and is dependent on the rotor speed. Hence the second equation of (6.24) represents the adjustable model of Fig. 6.2. By resolving equations (6.24) into two-axis components, the rotor flux components in the stationary reference frame are obtained as:

r Lr = r Lm

vs v s

( R s + Ls p ) 0 0 ( R s + L s p)

is i s

(6.25)

r 1 / Tr = r

1 / Tr

r L is + m r Tr i s

(6.26)

where p is the differential operator. Equation (6.25) is the reference model in developed form, while (6.26) is the adjustable model in developed form. The angular difference between the two rotor flux space vector positions is used as the speed tuning signal (error signal). The speed tuning signal actuates the rotor speed estimation algorithm, which makes the error signal converge to zero. The adaptation mechanism of MRAC based speed estimation method is a simple PI controller algorithm

est = K p + K i dt
where the input of the PI controller is
(1) (2) ( 1) (2) = r r r r

(6.27)

(6.28)

and Kp and Ki are arbitrary positive constants (parameters of the PI controller), while superscripts (1) and (2) identify outputs of the reference and adjustable model in accordance with Fig. 6.2, respectively. Figure 6.3 shows the complete scheme of the MRAC based speed estimation using rotor flux for adaptation purposes. It should be noted that the factors Lr/Lm in (6.25) and Lm/Tr in (6.26) have been incorporated into the adaptation mechanism gain constants Kp and Ki. The outputs of the two models thus only represent the rotor flux space vector in angle, but not in amplitude.

E Levi, 2001

HIGH PERFORMANCE DRIVES 108 ---------------------------------------------------------------------------------------------------------------------------------------

Reference model

i s vs vs is

Rs + Ls p

1 p

1) (r
( 1) r

Rs + Ls p
1 Tr

1 p

+ -

1 p

Kp +

(r2 )
(2) r

Ki p

est
+ +
1 p
1 Tr

Adjustable Model
Figure 6.3. - Block diagram of MRAC based speed estimator using rotor flux for the speed tuning signal creation.

In practice, it is very difficult to implement the pure integrator in the reference model due to the problems of initial conditions and drift. In order to eliminate the pure integrator, it is possible to implement the reference model by using a low-pass filter, with the transfer function 1/(p+1/T), instead of the pure integrator. However, since 1/(p+1/T) = (1/p)[p/(p+1/T)], the pure integrator can be eliminated by inserting a high pass filter [p/(p+1/T)] in the output side of the reference model (which contains 1/p). Since the output of the reference model gives the modified rotor flux, the adjustable model needs to be modified as well. Therefore, the identical high pass filter [p/(p+1/T)] is placed in the output channel of the adjustable model as well. The modified rotor flux based MRAC speed estimator using auxiliary variables is shown in Figure 6.4.

vs Reference is model

r(1)

p p+1/T

r(1)mod
Eq. (6.28) PI contr.

est

Adjustable model

r(2)

p p+1/T

r(2)mod

Fig. 6.4 - Modified rotor flux based MRAC speed estimator.

E Levi, 2001

HIGH PERFORMANCE DRIVES 109 ---------------------------------------------------------------------------------------------------------------------------------------

6.5.

PERFORMANCE OF THE ROTOR FLUX BASED MRAC SPEED ESTIMATOR

Performance of the MRAC based speed estimator, with rotor flux space vector selected as the output of the reference and the adjustable model, is illustrated by means of experimental results. A commercially available indirect rotor flux oriented induction motor drive, with speed sensor, is used in the experimental study. The drive has the 1,500 rpm base speed. It is equipped with the standard indirect vector controller, so that the drive configuration fully corresponds to the one of Fig. 3.11. A schematic diagram, showing the major parts of the experimental rig, is shown in Fig. 6.5. The induction motor is a 2.3 kW, 50 Hz, 4-pole motor (1500 rpm = 1 p.u. speed) and it is supplied by the commercially available DBS 04 type vector controller. A DC machine is used for loading purposes. The drive is equipped with resolver feedback and is in all the experiments operated as a drive with measured position and speed feedback. Actual speed of the drive is measured for display purposes using a dynamic signal analyser, using analogue speed signal output available in the drive. The analysed speed estimator is shown in its basic form in Fig. 6.3. The integration problem is overcome using the scheme of Fig. 6.4. Stator voltages and currents are measured and filtered using analogue circuitry, prior to being sampled using LabView software. Speed estimator is implemented in a PC using Simulink/Matlab and is at all times operated in parallel to the rig (i.e. estimated speed is not used either for speed feedback or for orientation angle calculation). Such an approach enables fine tuning of all the necessary filters within the speed estimation algorithm, as well as the PI controller of Fig. 6.4. Tuning was performed using experimental recording of actual speed response to application of a step 500 rpm speed command from standstill and the estimate of the same speed response (detailed description of the tuning procedure is beyond the scope here). The experimental rig is shown in Fig. 6.5, while the final structure of the speed estimator, with all the filters and their parameters included, is given in Fig. 6.6.
Current and voltage measurement ia ib Data cable

PC reference speed setting

PC with data acquisition card, Labview and Matlab for speed estimator evaluation

Serial link va vb vc speed feedback

3 phase
Single phase

Rectifier DC generator

Induction motor DBS 04 controller and PWM inverter Actual speed measurement HP35665A analyser Resistor bank

Fig. 6.5 - The experimental rig.

E Levi, 2001

HIGH PERFORMANCE DRIVES 110 ---------------------------------------------------------------------------------------------------------------------------------------

c=500 rad/s
vs v s is i s
Adjustable model, (6.26) Reference model, (6.25)

1/T=800 s-1

Butterworth 2nd order filter

p p +1 T
_

Kp=150 Ki=800

PI
1/T=800 s-1

1 + T1 p

est

p p +1 T

T1=0.03 s

Fig. 6.6 - Implementation of the constant parameter MRAC speed estimator. Testing consists of a series of acceleration transients, obtained as a response to the applied speed command, followed by appropriate steady-state operation at the set speed. Tests are performed for both no-load case and loaded operation (load torque varies approximately linearly with speed and is approximately rated at 1500 rpm). As an example, Figs. 6.7 and 6.8 illustrate comparison of actual speed and estimated speed for the no-load case with 900 rpm set speed and loaded case with 1500 rpm set speed. Excellent agreement can be observed from these figures, in both transient and steady-state operation.
1100 1000

920 915

Actual and estimated speed (rpm)

Actual and estimated speed (rpm

Actual

900 800 700 600 500 400 300 200 100 0 0 0.5 1 1.5 2 Time (s) 2.5 3 3.5 4 Estimated

910 905 900 895 890 885 880 2 2.05 2.1 Time (s) 2.15 2.2 Estimated Actual

Fig. 6.7 - Actual speed and estimated speed for the acceleration transient with 900 rpm set speed under no-load conditions: the complete transient and a zoomed extract in vicinity of steady-state operation.
1600 1400 Actual and estimated speed (rpm) 1200 1000 800 600 400 200 0 0 0.5 1 1.5 2 Time (s) 2.5 3 3.5 4 Actual Estimated

1520 1515 Actual and estimated speed (rpm) 1510 1505 1500 1495 1490 1485 1480 Estimated 1475 1470 2.3 2.35 2.4 Time (s) 2.45 2.5 Actual

Fig. 6.8 - Actual speed and estimated speed for the acceleration transient with 1500 rpm set speed in loaded operation: the complete transient and a zoomed extract in vicinity of steady-state.
E Levi, 2001

HIGH PERFORMANCE DRIVES 111 ---------------------------------------------------------------------------------------------------------------------------------------

Finally, a set of acceleration transients for low values of the set speed (75, 100, 125, 150 175 and 200 rpm), recorded under no-load conditions, is shown in Fig. 6.9, where a comparison of actual and estimated speeds is provided. The two are in excellent agreement in all the case.

250

Actual and estimated speed (rpm)

200

150

100

50 Estimated Actual 0 0 0.5 1 1.5 2 Time (s) 2.5 3 3.5 4

Fig. 6.9 Comparison of actual and estimated speeds for no-load acceleration transients (75, 100, 125, 150, 175, 200 rpm set speeds).

E Levi, 2001