Professional Documents
Culture Documents
Abstract – The present paper proposes an adaptive motor while making it less sensitive to load
fuzzy logic control (AFLC) for the position control of disturbances. In this paper, the closed loop control
a Stepper motor. The weights used for combining the is achieved by starting the motor with one pulse.
fuzzy rules are also updated, using the least mean Subsequent drive pulses are generated as a function
square algorithm. The paper also demonstrates a of the motor shaft speed which is estimated by the
Kalman filter for the estimation of motor parameters use of Extended Kalman filter.
like speed and flux vector position. The estimation is In this paper, it is proposed to estimate
ensured to be robust even in the presence of arbitrary the speed of the motor and the rotor position from
fluctuation of the input line currents. The modified
the measurements of two phase stator currents,
using an Extended Kalman filter (EKF). The
Extended Kalman Filter (EKF) rejects the outliers in
estimate also produces other quantities, like
real-time, thereby eliminating the need for manual
currents in direct and quadrature axes. The
intervention in tuning the parameters of the EKF.
proposed EKF estimation does not require the
The stability of the controller is checked by
information of the input voltage at all and it makes
computing the Lyapunov exponent from the evolution it different from similar other works [2, 3]. So the
of the state space of direct axis current, quadrature present estimation treats the whole problem as
axis current, synchronous speed; and found to be blind system identification [4, 5]. However, the
stable at all the time. The simulation results show that main problem associated with the EKF is its
the proposed control strategy operates robustly under dependence on the parameters like the initial state,
modeling uncertainty, with a good dynamic initial state co- variance, the measurement noise
performance. and the plant noise. The convergence is highly
dependent on the choice of the covariance matrices
Key words- Stepper motor, Adaptive Fuzzy logic of the plant noise and the measurement noise that
position control, Robust Kalman filter, Lyapunov appear in EKF algorithm [6]. Here the
exponent. measurements to the EKF are the output of space
vector PWM inverter, which is non sinusoidal and
I. INTRODUCTION rich in harmonics, there by degrading the
performance of EKF. To overcome this problem, a
Stepper motors are digital actuators Robust Kalman Filter is used in modifying the
which accept input pulses and respond to them by elements of the measurement covariance matrix, on
rotating the output shaft by equal increments for the run [7]. The estimated quantities show the
each input pulse. The stepping motors are mainly robustness of the technique adopted.
used in simple point to point positioning tasks in Now days stepper motors are mainly used
which they are open loop controlled [1]. The open
in closed loop control mode particularly for
loop control does not use any information
machine tool and robotics applications. According
regarding the motor shaft position or speed and in
to F Betin,D Pinchon et. al. [1], the conventional
open loop, the motor is driven by a train of pulses
with a predetermined time interval. There is no control strategies like PID control are not a good
guaranty that the motor uses all the pulses since choice because of its poor performance. In this
there is no feedback. The digital closed loop paper, the application of the fuzzy logic theory is
principle was introduced in the 1970’s in order to effectively used to control the speed of a stepper
increase the accuracy positioning of the Stepper motor. Compared to the traditional PI controller,
the FLC does not depend on precise system the coil.
and are the direct axis and quadrature
mathematical model [8]. Over the last many years, axis currents. is the differential operator. is the
the fuzzy logic technique has become very popular moment of inertia. is the viscous friction,!" is
in the control of ac drives [8, 9] because of the load torque. is the permanent magnet flux .( is
flexibility in accommodating overlapping the shaft position of the motor , ( # )( .Where
information in the definition of terms. The usual N is the number of rotor teeth,( is the electrical
approach is to compute a fuzzy function on the angle.
error between the set speed and the estimated
speed, using a set of predefined rules. All the rules
are fired with equal weightage, wm to produce the III. ESTIMATION OF STATE VECTOR WITH
ROBUST EKF
output, given by,
The Extended Kalman filter [13] is an
The mathematical model of hybrid stepper
!" # $ % & (2)
motor is mutually coupled and hence nonlinear .For
'
this non linear system, an Extended Kalman filter is
# (3) implemented. For a given sampling time Ts, both
where
, are the direct axis and quadrature axis the state estimate*P
+O+ and its covariance matrix
voltages,L is the inductance and R the resistance of
Pk/k are generated by the filter through a two step equation. The results also show that the Kalman
loop predictor corrector process. The corrector filter has become robust against input variation,
process starts with an initial value of x,QROI- and with this technique.
follows as below:
PM NS ZO PS PM PB PB PB
PB ZO PS PM PB PB PB PB Δ ia
q= (8)
Set position θe
where n is the no. of rules, p is the weighting
factor and e g is the membership function of ith rule. Fig 2 Block diagram representation of the proposed
The weighting factor in (10) is updated by the control system
gradient descent technique. The main purpose of
adjusting these parameters of the fuzzy controller is
to minimize the sum of square errors between the
rotor speed and the reference speed. The V. COMPUTATION OF THE LYAPUNOV
instantaneous cost function is defined as: EXPONENT TO ASSESS THE STABILITY
J(k+1)=;Oqr %7 ;&
P%7
" ;&s2 (9) The Lyaponov exponent is an important
indicator of the stability of a nonlinear system.
where is the reference motor speed and Pis
" Given a continuous dynamical system in an n
the estimated speed. Following [10], the change in dimensional phase space, one can monitor the long
weight ( 10) is given by term evolution of an infinitesimal sphere of initial
condition. The ith one dimensional Lyapunov
{
npt %7& u vwxy% &$z o %7& .. (10) exponent is then defined as follows [17]:
- %&
B is the motor parameter, and o %7& is the .|g # }1~ }H _ (12)
%R&
instantaneous error in motor speed.
where Pi : the ith state variable of the phase plane
Ki and Ku are the scaling factors. The current The nature of the Lyapunov exponents provide an
command iq* is obtained by the output of AFC as indication of the stability of the system studies, as
explained below:
Iq*(k)=iq*(k-1)+ ku q(k) (11) | = 0 a marginally stable orbit.
| < 0 a periodic orbit or a fixed point
~HB
HCh| > 0 Chaotic behaviour.
For the stepper motor studied, the state space 5.7249, -8.3621 , -4.193, there by confirming the
evolution of the vector [id, iq, ωe] it is collected in stability
sequence and the Lyaponov Exponents are
computed using Eqn. (12). It is seen that all the VII .CONCLUSIONS
Lyaponov exponents are negative i.e. : -5.7249 , - A robust EKF algorithm along with LMS
8.3621 , -4.193 based adaptive fuzzy, sensor less position
(taken from Sec. VI, below) , bring out the controller has been successfully implemented in
stability of the system under control. Simulink/Matlab environment. The performance
really shows that the controller is well suited for
VI. RESULTS AND DISCUSSIONS
Simulation of the given Stepper motor has 14
S t e p p o s it io n in r a d
Permanent magnet /hybrid type,, 2 phases, R=0.7ȳ, motor shaft position
8 Reference position
L=0.0014H, =0.005wb/m2, J= 1.2e-7-m2
The stator currents are taken from the 6
motor model, running in Simulink. The EKF based
approach for the estimation of speed and rotor 4
position are found to be robust, for different step
changes. The estimation needs only the two phase 2
stator current values, which are easily available
0
from the machine. One of the main distinction from
similar other works in literature is that only two -2
stator currents are required in order to estimate six 0 0.05 0.1 0.15 0.2 0.25
Time in sec
quantities at a time. Fig. 3 and Fig. 4 indicate that
the estimation converges in less than 1m S, in Fig 3step position catches up with the reference position
response to a step change. The robust EKF ensures
9
that the estimation is stable, with the digital drive
from the PWM inverter, without having to trim the 8
6
model of the motor, using embedded Matlab motor shaft position
5 Reference position
facility. Fig 4 shows the effectiveness in estimating
the rotor position very quickly even for step 4