Beruflich Dokumente
Kultur Dokumente
(IJARET)
Volume 7, Issue 2, March-April 2016, pp. 5363, Article ID: IJARET_07_02_005
Available online at
http://www.iaeme.com/IJARET/issues.asp?JType=IJARET&VType=7&IType=2
Journal Impact Factor (2016): 8.8297 (Calculated by GISI) www.jifactor.com
ISSN Print: 0976-6480 and ISSN Online: 0976-6499
IAEME Publication
___________________________________________________________________________
1. INTRODUCTION
Control system should be able to help achieving desirable response for a given set of
specifications. More important aspect is stability of system with accuracy as per need
of the application. High starting torque and low inertia motors are essentials for
motion control applications. DC servo motors are useful for rapid changes in velocity
http://www.iaeme.com/IJARET/index.asp
53
editor@iaeme.com
http://www.iaeme.com/IJARET/index.asp
54
editor@iaeme.com
MRAC based on Dynamic Back Propagation and Fuzzy Emulator for Converter
application has been discussed in [10]. Output of master system has been used to
choose model for reference and T-S fuzzy model has been used to present the chaotic
and discrete-time slave nature system in [11]. With the fuzzy state estimator, and by
combination of the adaptive control backstepping technique with the decentralized
system design in[12] an adaptive decentralized output feedback control fuzzy
approach has been developed. For suggested control approach semi-globally and
uniformly and ultimately bounded probable value has been assured.
In [13] the sum of the output of conventional MRAC provide I/O of the neural /
fuzzy controller. Supervisory loop with intelligence has been incorporated with the
conventional MRAC framework design by using an online neural and fuzzy network
structure in parallel with it. Modeling with control orientation, design, structures and
simulation work with various techniques and control law strategies, result analysis
details and advanced control algorithms have been discussed in book [14]. A gradient
scheme known as the MIT rule is given by
d
e
e
where, error e y ym , is
dt
adaptation gain and is for adjustment mechanism. Fig. 1 shows basic block diagram
for proposed system. After MIT rule based MRAC [15] various modifications were
suggested by researchers. A summary and rigorous compilation of all methods,
control techniques and study of application has been carried out in [16].
ANFIS based Robot manipulator control has been proposed in [17].NN control
and ANFIS based results have been compared with and without disturbance.
Importance of soft computing technique based MRAC with past and recent
publication survey has been discussed in [18]. MRAC sections classified in survey
paper are based on applications, techniques and soft computing. Optimal criterion for
energy efficiency should be given higher importance in design [19]. Mechatronic
devices normally performs well but are intrinsically energy intensive. It affects overall
system sustainability so energy optimal motion is needed.
Adaptive and fuzzy based output tracking using backstepping control technique
has been performed in [20] and unknown nonlinear part has been identified using a
fuzzy logic system. Optimal observer design using soft computing has been applied
for HJB equation based algorithm in [21]. A neural adaptive controller for DC motor
tracking problem is discussed in [22]. Existing techniques have been integrated in
[23], such as the linearization technique for I/O, application of NN to linearize
equation of control law, the network estimation errors are compensated with the
http://www.iaeme.com/IJARET/index.asp
55
editor@iaeme.com
f1 (e, , , )
1
(01)
In layer 2 nodes are fixed with its output as the product of all their entry inputs. As
normalization is important, in next layer normalization takes place. After calculation
of consequent evaluation interference in layer 4, the overall output after layer 5 is
i wi fi
i wi
http://www.iaeme.com/IJARET/index.asp
(02)
56
editor@iaeme.com
Fig. 2 shows ANFIS structure with details for each layer. For training purpose the
least squares method as well as the backpropagation gradient descent method has been
used in combination. The difference between FIS output and training data output has
been considered as error at each epoch for least square estimation. After generation
and training part final fuzzy inference system calculates for adaptation
La
j
dia
Ra ia kb ea
dt
(03)
d
f k ia
dt
(04)
Figure 3 DC servomotor
http://www.iaeme.com/IJARET/index.asp
57
editor@iaeme.com
(05)
and plant
y Am y bu where u um
(06)
So
e Ame (b Bm )um
(07)
ef e
V (ef , )
1
1
(ef )2
(b Bm )2
2
2b
(08)
V ef ef
(b Bm )
(09)
V e( Am e (b Bm )um )
V Am e2
2(b Bm )
(10)
(b Bm )( um e)
(11)
Taking ume gives negative semi definite value for the remaining derivative
part of Lyapunov function. So adaptation parameter has been chosen accordingly for
stable system.
In DC servomotor motion control, derived can be written as
e( p m )
(12)
x Ax B(u f (x))
(13)
http://www.iaeme.com/IJARET/index.asp
58
editor@iaeme.com
(14)
lim x(t ) xm (t ) 0
t
T ( x)
u kxT x
(15)
(16)
T )T B 0
B(
m
Tracking error dynamics given by
T )T ( x)) A x B u A x A x) ( x(t ) x (t ))
e(t ) (( A BkxT ) x B((
m m
m m
m
m
m
(17)
T
T
(18)
e Ame B(kx x ( x))
The Lyapunov function is
(19)
(20)
kx x xeT PB sgn()
( x)eT PB sgn()
(21)
http://www.iaeme.com/IJARET/index.asp
(22)
59
editor@iaeme.com
Output, rad/msec
3
Model
ANFIS
Lyapunov
0
0
200
400
600
800
1000
1200
time, msec
Error for MRAC with ANFIS and Lyapunov Rule
Error, rad/msec
0.4
ANFIS
Lyapunov
0.2
0
-0.2
-0.4
0
200
400
600
800
1000
1200
time, msec
Figure 4 Output and Error plot for DC Servo System Reference Model Output
The response seems desirable in both Lyapunov based and ANFIS based MRAC
for system, butfor partly uncertain plant dynamics, Lyapunov based MRAC is not
able to approximate for uncertain part and gives average results. Use of ANFIS to
adapt and apply control effort helps for good results in such situations. Fig. 4 presents
the value of output and error plot for DC Servo System reference model output.
Model output varies based on variation in armature input voltage.
Adaptation parameter - Lyapunov based MRAC
2
1.5
0.5
200
400
600
800
1000
1200
1000
1200
time, msec
Adaptation parameter - ANFIS based MRAC
1.2
1
0.8
0.6
0.4
0.2
0
-0.2
200
400
600
800
time, msec
Figure 5 Adaptation Parameter Generated with ANFIS Trained using Data from Lyapunov
Based MRAC
Fig. 5 shows adaptation parameter generated with ANFIS and with Lyapunov
stability Based MRAC. It shows that approximate model tries to remove oscillations
and mostly approximates smooth value for new signal.
http://www.iaeme.com/IJARET/index.asp
60
editor@iaeme.com
V
216.48
216.47
216.46
216.45
0
200
400
600
800
1000
1200
800
1000
1200
time, msec
Vdot
0
Vdot
-0.2
-0.4
-0.6
-0.8
0
200
400
600
time, msec
Figure 6 Lyapunov Function and its Time Derivative for ANFIS Based MRAC
Results shown in Fig. 6 are Lyapunov Function V for MRAC based on adaptation
parameter generated by ANFIS and its Time Derivative. Observation shows that
positive definite Lyapunov function and negative semidefinite derivative term proves
stability of proposed system.
6. CONCLUSION
In this paper MRAC using Lyapunov and ANFIS applied to control motion
parameters of DC servo motor. Results are compared for both strategies. It has been
observed that ANFIS based MRAC gives better results in terms of error convergence.
Introduction, Literature survey, system fundamentals, development of classical
MRAC, soft computing and ANFIS based MRAC has been included to understand the
importance of the work, problem definition and simulation work. Data of adjustment
parameter has been saved after applying Lyapunov based rule and used to apply
ANFIS technique to generate . Lyapunov stability with suitable V and derivation of
V has been shown. Comparison with other new computing techniques and analysis
with case studies of higher order systems may give better vision on future scope,
importance and development of the topic.
ACKNOWLEDGEMENT
The present work is a part of PhD research work carried out at Nirma University.
REFERENCES
[1]
[2]
[3]
Karl Johan Astrom, Adaptive Control, 2nd Ed., Pearson Education, 2001
Brian D O Anderson, Arvin Dehghani Challenges of adaptive controlpast,
permanent and future, Annual Reviews in Control, Volume: 32, Issue: 2,
Pages: 123-135, ISSN: 13675788, Elsevier, 2008
Bo Zhao, Hongjie Hu "A new inverse controller for servosystem based on
neural network model reference adaptive control", COMPEL - The
http://www.iaeme.com/IJARET/index.asp
61
editor@iaeme.com
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
http://www.iaeme.com/IJARET/index.asp
62
editor@iaeme.com
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
http://www.iaeme.com/IJARET/index.asp
63
editor@iaeme.com