Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Design of Robust H-infinity Speed Controller for

High Performance BLDC Servo Drive


Dhaneesh Krishnan T V Krishnan C M C K Panduranga Vittal
Research Scholar Assistant Professor Professor
Electrical and Electronics Engineering Electrical and Electronics Engineering Electrical and Electronics Engineering
NIT Karnataka, India NIT Karnataka, India NIT Karnataka, India
Email: dhaneeshktv@gmail.com Email: cmckrishnan@nitk.edu.in Email: vittal@nitk.ac.in

Abstract—In Brushless DC (BLDC) motors, commutation is external disturbances are some of the significant problems and
done externally by using a three phase inverter and therefore need researchers have proposed different robust control techniques
more complex control algorithm as compared to brushed DC like Sliding Mode Control (SMC), H-innity control, back
motors. BLDC motors are found to be the most suitable actuator
for high performance servo drive used in robotics, industrial au- stepping algorithm, model predictive control, fuzzy and neural
tomation, automobiles, aerospace etc. The motor controller must network based control etc.
adapt to the time varying motor characteristics and maintain the Sliding Mode control (SMC) is the most popular and
performance in its long service life. The controller also should commonly used robust control technique used in permanent
be robust against external disturbances like load disturbances, magnet motor drive [4],[5]. SMC is comparatively easier to
measurement noise etc. This paper presents the design of robust
H∞ speed controller for BLDC servo drive to address the design, especially for non-linear systems since it does not
problems mentioned above. The H∞ speed controller is designed demand the linearization of the model. In SMC, because of
and simulations are carried out in MATLAB/SIMULINK and its using a discontinuous switching function, the performance
performance is compared with a PI controller. is affected by a phenomenon called chattering. Chattering
causes high frequency variation in the control signal and the
Keywords – BLDC motor drive; H∞ control; Robust magnitude of this variation in the control signal varies with the
control; Speed control; PI control amount of uncertainty present in the system. Different control
algorithms have been proposed by the researchers to reduce the
I. I NTRODUCTION
effect of chattering. Some of them are, boundary layer control
BLDC motor has gained popularity in recent years due to [6], quasi-SMC [7], higher order SMC [8], adaptive SMC
the significant developments in permanent magnet motors and using fuzzy and neural networks [9] etc. An observer based
it is extensively used in precision speed and position control disturbance rejection algorithm is used in adaptive control
systems. As the name indicates, it does not have brushes and and feed forward compensation is used to compensate the
the commutation is done externally by electronic means [1]. effect. A nonlinear model predictive controller for the speed
BLDC motors are found to be more suitable for applications control of permanent magnet motor has been proposed [10],
like robotic arm, CNC machines, industrial automations, au- which is basically the optimization of a cost function related
tomobiles, aerospace etc., due to the advantages over other to the tracking error and it is not suitable for the systems
motors. Some of them are, better speed-torque characteristics, having faster dynamics like drive control. Adaptive back-
high efficiency, high dynamic performance, high torque to stepping control is another proposed robust control which
weight ratio, long operating life, noiseless operation etc. is robust to the dynamic uncertainty and mechanical and
Conventional methods use PI or PID controller for speed electro-mechanical parameter variation of the system and more
and position control of BLDC motor drive. It is easy to suitable for nonlinear dynamical systems [11].
implement and has good control performance in and around H∞ based robust control has been proposed for position and
a particular operating point. Many researchers have proposed speed control applications [12]. H∞ controller shows good ro-
PI controller for speed control applications in BLDC motor bust performance against the disturbances even at a low sensor
drive [2],[3]. The parameters of the controller are tuned resolution. The excellent performance and robustness makes
for particular operating conditions under the assumption that H∞ control an attractive alternative to the conventionally used
the operating conditions will not vary significantly. But in controllers. Neural network based adaptive H∞ control has
practical applications, the operating conditions as well as been presented in [13], which is based on the disturbance
the system parameters may vary and hence the PID based estimation and feed forward algorithm. This increases the
control scheme sometimes gives erroneous results. Also, it robustness, but reduce the reliability due to the increased
does not have enough robustness to the internal and external complexity.
disturbances. In a motion servo system, the deterioration of In this paper, the design of an efficient H∞ controller is
performance due to parameter variation and other internal and presented for the speed control of high performance BLDC

978-1-5090-6348-2/17/$31.00 2017
c IEEE 37
Fig. 1. Block diagram of field oriented controlled BLDC motor drive

servo drive. The controller is designed offline which does Ignoring mutual inductance between the phase windings and
not demand online computations and therefore gives increased assuming symmetry in inductances [15], Ld = Lq = Ls
reliability. Simulation study has been carried out to validate By assuming that the control is good enough to keep id = 0,
the performance of the controller and compared with the
diq
performance of the conventionally used PI controller. Vq = R s i q + L q + ω e λm (2)
The paper is organized into five sections. In Section II, the dt
modelling of BLDC motor drive is described; in Section III, The torque equation is,
  
the design of H∞ speed controller is presented; in Section IV, 3 P
the simulation results are discussed and finally, the conclusion Te = λm iq = K t iq (3)
2 2
is presented in Section V.
where Kt is the torque constant and P is the number of pole
II. M ODELLING OF BLDC MOTOR DRIVE pairs. Considering zero load torque, the mechanical equation
Figure 1 shows the block diagram of field oriented con- is given by,
dωm
trolled (FOC) BLDC motor drive. FOC is used to decouple Te = J + Bωm (4)
the stator current into magnetizing and torque producing dt
ωe
components and control them separately to obtain DC motor where, ωm = , Te is the electromagnetic torque, J
like performance. The harmonics and torque ripples are less in (P/2)
is the inertia of the rotor and B is the frictional coefficient.
FOC based control when compared to trapezoidal control and Taking the transfer function of equations (2), (3) and (4)
the dynamic performance could be improved significantly. The Iq (s)
drive has inner current control loop and outer speed control and rearranging for the transfer function , the following
Vq (s)
loop. expression is obtained.
A. Modelling of BLDC motor in d-q refrence frame   
1 B
The dynamic equation of stator circuit in d-q reference s+
Iq (s) Ls J
frame is given by [15], =    
Vq (s) 2
R s J + L s B Rs B + Ke Kt (P/2)
diq s + s+
Vq = Rs iq + Lq + ω e L d i d + ωe λm (1) Ls J Ls J
dt (5)
where, Vq = q-axis stator voltage where Ke = λm = Back emf constant
Ld and Lq = d and q axis winding inductances The parameters of BLDC motor used in the simulation are
id and iq = d and q axis currents given in Appendix. The transfer function after substituting the
Rs = Stator resistance per phase parameters and rearranging is,
ωe = Angular velocity in electrical radian per second Iq (s) K1 (1 + sTm )
λm = Rotor flux linkage = (6)
Vq (s) (1 + sT1 )(1 + sT2 )

38 International Conference on Smart Grids, Power and Advanced Control Engineering (ICSPACE2017)
where, K1 = 7.4128×10−4 , Tm = 3.5386, T1 = 2.097×10−3
and T2 = 2.3766 × 10−4

B. Design of PI controller in q-axis current loop


The inner q-axis current control loop with PI control is
modelled as shown in figure 2. i∗q is the reference q-axis
current coming from outer loop and iq is the actual current.
The inverter is modelled as a first order transfer function as Fig. 3. Closed loop speed control system
[14],
Kin
Gin (s) = (7) the outer speed control loop, the first order approximation is
1 + sTin
well justified [14].
For a DC link voltage of 24V and a PWM frequency of
10kHz, the value of Kin and Tin are 0.65 and 0.5 × 10−4 sec. iq Ki
respectively. = (11)
i∗q 1 + sTi

where, Ki = 0.9545 and Ti = 9.77 × 10−5

A. Modelling of plant with parameter uncertainty


Ki , Ti , Kt , J and B are the five plant parameters which may
be subjected to the variations. Assume that their values are
within known intervals. For example, Ki = Ki (1 + PKi δKi )
Fig. 2. Inner q-axis current control loop with PI controller
where Ki is the nominal value Ki , PKi represents the maxi-
mum possible perturbation in percentage of nominal value and
Assuming 1 + sTm ∼
= sTm , the loop transfer function can be −1 < δKi < 1. A similar definition can be given for other
written as, parameters.
K(1 + sTc )
GH(s) = (8)
(1 + sT1 )(1 + sT2 )(1 + sTin )
where,
K1 Kc Kin Tm
K= (9)
Tc
The controller parameters are calculated to get the damping
ratio of 0.707 for getting good damping performance of closed
loop system. The parameters are calculated as, Kc = 8 and
Tc = 2.3766 × 10−4 respectively. The values of proportional
and integral constants are KP = 3 and KI = 12623
respectively.

III. D ESIGN OF SPEED CONTROLLER Fig. 4. LFT representation of open-loop plant with uncertainty
The figure 3 shows the closed loop speed control system.
ωref is the reference speed and ω is the actual speed measured The Linear Fraction Transformation (LFT) representation of
using a sensor. The measured speed consists of measurement open-loop plant with parameter uncertainty is shown in figure
noise which is considered as an external disturbance repre- 4, where, G denotes input-output dynamics of the open-loop
sented by the term n. K(s) is the robust H∞ controller, which plant with parameter uncertainty. G has six inputs, six outputs
is to be designed. Kt is the torque constant and is assumed and two states (x1, x2). The state space representation of the
as constant. Td represents the combination of load torque generalized plant G is written as,
and disturbance torque on the shaft. The transfer function of ⎡ ⎤ ⎡ ⎤
x˙1 x1
mechanical part is derived from the torque equation, ⎢ x˙2 ⎥ ⎢ x2 ⎥
⎢ ⎥ ⎢ ⎥
dω ⎢ yK i ⎥ ⎡ ⎤ ⎢ uK i ⎥
Te − T d = J + Bω (10) ⎢ ⎥ A B1 B2 ⎢ ⎥
⎢ yT i ⎥ ⎢ ⎥
dt ⎢ ⎥ = ⎣ C1 D11 D12 ⎦ ⎢ uTi ⎥ (12)
⎢ yK t ⎥ ⎢ uK t ⎥
where, J is the moment of inertia and B is the frictional ⎢ ⎥ C2 D21 D22 ⎢ ⎥
⎢ yJ ⎥ ⎢ uJ ⎥
coefficient. The current control loop is approximated by a first ⎢ ⎥ ⎢ ⎥
⎣ yB ⎦ ⎣ uB ⎦
order transfer function for the simplicity in speed controller
design. As the dynamics of current loop is much faster than y u

International Conference on Smart Grids, Power and Advanced Control Engineering (ICSPACE2017) 39
−B/J Kt /J The performance weight We is selected to reflects the
where, A = ,
0 −1/Ti desired performance characteristics. A low-pass weight is used
0 0 PKt /J −PJ −PB /J so that We−1 reflect the desired shape of sensitivity function
B1 = ,
PKi /Ti −PTi⎡ 0 0 ⎤ 0 and to get minimum overshoot and steady state tracking error.
0 0 A simple gain is selected as Wu to limit the controller output.
⎢ 0 −1/Ti ⎥
0 ⎢ ⎥ The purpose of this weight is to prevent actuator saturation
B2 = , C1 = ⎢ ⎢ 0 Kt ⎥⎥, C2 = 1 0 , and limit the amplification of sensor noise signal on the
Ki /Ti ⎣−B/J Kt /J ⎦
controller input signal [16]. The performance weights used
⎡ B 0 ⎤ in this simulation are,
0 0 0 0 0
⎢PKi /Ti −PTi 0.284(s + 13.34)
⎢ 0 0 0 ⎥
⎥ We = and Wu = 0.1
D11 = ⎢ 0 0 0 0 0 ⎥ s + 0.012
⎢ ⎥,
⎣ 0 0 PKt /J −PJ −PB /J ⎦ A third order controller is obtained with infinity norm value
0 0 0 0 0 less than 0.49 for all possible uncertain transfer matrix Δ. The
D12 = K K /T 0 0 0
T
and D21 = D22 = 0 controller transfer function is,
i i i
3.783 × 107 s2 + 4.737 × 1011 s + 6.301 × 1012
B. Design of H∞ speed controller K(s) =
s3 + 1.87 × 107 s2 + 8.038 × 1011 s + 9.646 × 109
The block diagram of closed-loop system is given in figure C. Design of PI speed controller
5. We and Wu are the performance weight functions used The PI speed controller is designed using Symmetric Opti-
to tune the controller’s performance and robustness charac- mum method [14]. A filter transfer function of 1/(1+0.0016s)
teristics by shaping the loop transfer functions and it can be is considered in the feedback path in figure 3. The controller
frequency dependent or a simple gain. The maximum param- parameters are obtained as KP = 0.0318 and KI = 4.683.
eter variation in the system is taken as 50% in all parameters.
For robustness, the closed loop system must internally remain IV. S IMULATION R ESULTS
stable for any variation within this limit. In order to get The speed controlled BLDC motor drive is simulated in
good robust performance, the sensitivity functions from the MATLAB/SIMULINK. A comparative study has been done
disturbance to the performance variables ee and eu should be on the performance of H∞ controller and PI controller.
minimum to get desirable command following and disturbance Various operating cases are considered to validate the dynamic
rejection. Hence the performance criterion is that the transfer performance and robustness. The simulation results in each
functions from the disturbance (ref , Td and n) to ee and eu case are discussed in the following subsections.
should be small in the sense of infinity norm for all possible
uncertain transfer matrices Δ [16]. A. Case-1: With nominal parameters and constant load
Figure 6 and 7 shows the speed tracking performance with
H∞ controller and PI controller respectively. A constant load
with nominal motor parameters as shown in table is considered
for the simulation. The graph of speed, error, controller output
and electro-magnetic torque are shown separately. A constant
load torque of 0.1 N-m is applied at 0.1 sec. The reference
speed is varied with a constant acceleration. It is very clear
that the transient and steady state response of H∞ controller is
better than that of PI controller. When considering the transient
response, with PI controller, there is an overshoot of 5% during
Fig. 5. Closed loop system structure with H-infinity controller
the speed variation and a speed drop of 400 rpm (40%) when
the load is applied, whereas, the H∞ controller does not make
⎡ ⎤ any overshoot or considerable speed drop during transient. The
ref
ee G11 G12 G13 ⎣
= Td ⎦ (13) steady state error is zero in both the cases, but, there is a
eu G21 G22 G23 high frequency oscillation with small amplitude in the speed
n
at steady state with PI controller which is undesirable as it
where, G11 = We (I + GK)−1 , G12 = −We G(I + GK)−1 , reduces the life of the machine.
G13 = −We (I + GK)−1 , G21 = Wu K(I + GK)−1 ,
B. Case-2: With nominal parameters and load torque distur-
G22 = −Wu GK(I + GK)−1 , G23 = −Wu K(I + GK)−1
bance
Therefore for robust performance, the following criterion In servo system, the effect of load disturbance is a signif-
has to be fulfilled. icant problem and therefore the simulation is carried out to
 
 G11 G12 G13  validate the performance of the controller in such situation.
 
 G21 G22 G23  < 1 (14) The speed tracking performance of the drive for load torque

40 International Conference on Smart Grids, Power and Advanced Control Engineering (ICSPACE2017)
at 0.3 sec and 0.5 sec as shown in figure 8. The load torque
and load disturbances are applied with a slope to resemble the
practical situation. The H∞ controller has good disturbance
rejection capability and the speed variation during transient is
negligible. The maximum speed variation is around 2%. The
torque disturbance causes significant speed variation with PI
controller and the amount of speed variation is 16% for each
0.1 Nm change. This will become more significant when the
drive is operating at very low speed and the amount of speed
variation increases with the increase in slope of disturbances.
Therefore, it is evident that, the H∞ controller gives good
transient and steady state response with better load torque
disturbance rejection as compared to the PI controller.

Fig. 6. Speed tracking performance with H∞ controller

Fig. 8. Speed performance with load torque disturbances

C. Case-3: With variation in flux


The BLDC motor uses permanent magnet. When subjected
to external magnetic fields or temperature changes, the mag-
netic properties of permanent magnet may change, leading
to demagnetization, which may affect the performance of the
motor. It is therefore very important to take this phenomenon
into account when designing the controller. The simulation is
done with nominal and 70% of the nominal flux. The reference
Fig. 7. Speed tracking performance with PI controller speed is kept constant at 1000 rpm and a load torque of 0.1 Nm
is applied at 0.2 sec. The total simulation time is 0.5 sec. Apart
from the negligible speed variation, no change in the speed
disturbance is depicted in figure 8. The reference speed is kept tracking performance is observed with H∞ controller. With the
constant at 1000 rpm and a load torque of 0.1 Nm is applied at PI controller, the drop in speed during load variation increases
0.1 Sec. Two torque pulses of magnitude 0.1 Nm are applied with the variation in the flux. Therefore, it can be concluded

International Conference on Smart Grids, Power and Advanced Control Engineering (ICSPACE2017) 41
that, the H∞ controller gives better robustness against flux A PPENDIX
variation as compared to the PI controller. BLDC motor parameters

D. Case-4: With variation in the parameters Rs and B Parameters Value


Rated Voltage 24V
The simulation is carried out with nominal and 150% of Stator phase resistance, Rs 0.89Ω
nominal value of Rs and B. The speed and torque profile Stator phase inductance, Ls 0.19mH
applied are as same as case-3. The significant variation in the Flux linkage established by magnets 0.0196V.s
performance is not observed. Rotor inertia, J 6.05 × 10−6 kgm2
Viscous damping, B 1.71 × 10−6 N.m.s
For the effective comparison between the two controllers, Number of poles 4
the three performance measurements are considered. They are, No-load speed 5860rpm
Integral of the Absolute Error (IAE), Integral of the Time- Peak current (stall) 26.97A
Max. continuous torque 0.135N m
weighted Absolute Error (ITAE) and Peak error during tran-
sients. Table 1 shows the comparison of these error functions
between the two controllers for the speed and torque profile as R EFERENCES
discussed in each cases. It is clear from the table that, all the [1] Dhaneesh Krishnan T V, Samskriti and K.P.vittal, “Review of devel-
error function values with H∞ controller is much less than that opments in BLDC motor controllers along with study of four-quadrant
with PI controller, which indicates the increased robustness operation and active power factor correction”, IEEE 10th International
Conference on Industrial and Information Systems, ICIIS 2015, Dec. 18-
and speed tracking capability of H∞ controller. 20, 2015, Sri Lanka.
[2] U. Vinatha, S. Pola, and K. P. Vittal, “Simulation of four quadrant
operation and speed control of bldc motor on matlab / simulink,” in
TENCON 2008 - 2008 IEEE Region 10 Conference, pp. 16, Nov 2008.
[3] C. S. Joice, S. R. Paranjothi, and V. J. S. Kumar, “Digital control
strategy for four quadrant operation of three phase BLDC motor with
load variations,” IEEE Transactions on Industrial Informatics, vol. 9, pp.
974-982, May 2013.
[4] Ilyas Eker, “Sliding mode control with PID sliding surface and experi-
mental application to an electromechanical plant,” ISA transactions, vol.
45, no. 1, pp. 109-118, 2006.
[5] A. Sabanovic, “Variable structure systems with sliding modes in motion
control- A survey,” IEEE Transactions on Industrial Informatics, vol. 7,
no. 2, pp. 212-223, 2011.
[6] K. D. Young, V. I. Utkin, and U. Ozguner, “A control engineers guide to
sliding mode control,” IEEE Transactions on Control Systems Technology,
vol. 7, pp. 328-342, May 1999.
[7] M. L. Corradini, G. Ippoliti, S. Longhi, and G. Orlando, “A quasi-
sliding mode approach for robust control and speed estimation of PM
synchronous motors,” IEEE Transactions on Industrial Electronics, vol.
59, pp. 1096-1104, Feb 2012.
[8] Y. Wang, X. Zhang, X. Yuan, and G. Liu, “Position-sensorless hybrid
sliding mode control of electric vehicles with brushless DC motor,” IEEE
Transactions on Vehicular Technology, vol. 60, no. 2, pp. 421-432, 2011.
[9] J. B. Cao and B. G. Cao, “Fuzzy-logic-based sliding-mode controller
design for position-sensorless electric vehicle,” IEEE Transactions on
Power Electronics, vol. 24, pp. 2368-2378, Oct 2009.
V. C ONCLUSION [10] R. Errouissi, M. Ouhrouche, W. H. Chen, and A. M. Trzynadlowski,
“Robust cascaded nonlinear predictive control of a permanentmagnet
In this paper, a simple, robust H∞ speed controller is synchronous motor with antiwindup compensator,” IEEE Transactions
on Industrial Electronics, vol. 59, pp. 3078-3088, Aug 2012.
designed for BLDC motor used in servo applications. The field [11] M. Morawiec, “The adaptive backstepping control of permanent magnet
oriented controlled BLDC motor is modelled in d-q reference synchronous motor supplied by current source inverter,” IEEE Transac-
frame. PI current controller is designed for inner current tions on Industrial Informatics, vol. 9, pp. 1047-1055, May 2013.
[12] R. Bautista-Quintero and M. J. Pont, “Implementation of H-innity
control loop and approximated the loop by a first order transfer control algorithms for sensor-constrained mechatronic systems using low-
function for the simplification of speed controller design. The cost micro controllers,” IEEE Transactions on Industrial Informatics, vol.
H∞ controller is designed for accurate speed control. Sim- 4, pp. 175-184, Aug 2008.
[13] F. F. M. El-Sousy, “Hybrid H ∞ -based wavelet-neural-network tracking
ulation study has been carried out in MATLAB/SIMULINK. control for permanent-magnet synchronous motor servo drives,” IEEE
The performance and robustness of H∞ speed controller are Transactions on Industrial Electronics, vol. 57, pp. 3157-3166, Sept 2010.
validated for different operating cases and compared with a PI [14] R. Krishnan, Electric Motor Drives Modelling, Analysis and Control,
Pearson India Education Services Pvt. Ltd, 2015.
speed controller, which is tuned using the symmetric optimum [15] M. Lazor and M. Stulrajter, “Modified field oriented control for smooth
method. The performances are compared for different operat- torque operation of a bldc motor,” in ELEKTRO, 2014, pp. 180-185, IEEE,
ing conditions. For the numerical comparison of performances, 2014.
[16] D.W.Gu, P. Petkov, and M. M. Konstantinov, Robust control design with
different error functions are considered. In all the cases, the MATLAB. Springer Science and Business Media, 2005.
performance and robustness of H∞ controller are found to be
better as compared to the PI controller.

42 International Conference on Smart Grids, Power and Advanced Control Engineering (ICSPACE2017)

You might also like