Scribd
Upload
28 views

Literature Review

Uploaded by

Mati Ur Rehman
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
28 views

Literature Review

Uploaded by

Mati Ur Rehman
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

1.

Literature Review / State of the Art (Relevant Work Section)

Since start, it was understood to system developers or researchers that nearly all
dynamical systems are nonlinear in real-world. However, linearization of these type of
system over the stability positions generate linear models, which are precisely compliant.
Suwandi [4], in his paper described the adaptive proportional-Integral-derivative (PID)
control approach to control the speed of DC motor. The controller is developed to
compute the control factors which are adjusted adaptively to provide required control
functioning even if limitations of DC Motor are altered. The limits of controller are set
online once the motor is operating by employing a Recursive Least Squares (RLS)
method. The controller can adjust the values of its factors to retain functioning of motor
as it is required once factors of DC motor are altered. Firstly, a signal for 0.07sec of
Pseudo Random Binary Sequence is provided to the system to obtain the expected
transfer function of DC motor utilizing the RLS algorithm. By coefficients of the
estimated transfer function of system, poles of required characteristic equation can be
found for the system that possesses the suitable output. As a result, the proportional,
integral, and derivative constants of controller can be found by employing online pole
placement approach. An online detection technique is applied to decide the new control
parameters.
The most control system processes, in which the system constraints are expected to be
unidentified are more attractive in applied situations since maximum nonlinear systems
can be estimated reasonably by linear replicas in their ordinary operation ranges. It is tis
collective consequence of simplicity of investigation and real-world application that
accounts for the countless accomplishment of linear models and has ended them the issue
of thorough research for many years. Since latest times, a fast-developing technology and
a competitive market needed procedures to function in several incidents in zones in the
state space wherever linear estimates are no more acceptable. To deal with such nonlinear
challenges, investigation has remained in progress on their recognition and control
utilizing artificial neural networks founded totally on measured outputs and inputs.
The phrase artificial neural networks were established to imply any architecture that has
mainly parallel interconnection of uncomplicated processors. As of a theoretic viewpoint,
a neural network can be deemed as suitably a parameterized classification of nonlinear
plans. For the Duration of 80’s and early 90’s convincing verifications were provided by
several writers, that multi-layer feedforward networks can demonstrate any nonlinear
continuous functions to any level of demanded precision given that the networks are
adequately sizeable and appropriately equipped. This trend in ANN has earned broad
consideration in control functions.
This fundamental parallel and distributed architecture of ANN can be effectively utilized
for control of DC motor. Quite a few valuable studies on the DC motor speed control
techniques utilizing ANN based speed controllers have been described in [13], [14], [2],
[5], [4]. Weerasooriya and Sharkawi in [13-14] described a DC motor control system
employing an ANN, having a capacity of attaining the unspecified, time invariant,
nonlinear operating attributes of the DC motor. However, their compositions are mainly
centered on an off-line trained ANN with indirect MRAC (model reference adaptive
technique). Because of the lack of on-line training of the ANN, the speed control is not
entirely acceptable. As under undetermined operational situations, that are not deemed
through the off-line ANN training procedure, the ANN controller does not work
appropriately. El-Khouly and others [2] incorporate on-line updating of ANN centered
adaptive controller for a DC motor. They discovered that whilst they were capable to
achieve decent control operation in their work, at some point the on-line updating of the
weights turn out to be volatile causing the DC motor running away. The technique they
applied, have being changed as of the inverse dynamic model, the reference speed was
randomly brought as one of the inputs of the ANN structure, causing in the control
system suffering from the problem of instability.
In [4], [5] Hoque, Zaman and, Rahman have described study on a real-time use of an
ANN based controller of a DC motor drive system. A permanent magnet DC motor
control system with ANN speed controller is designed by them. To attain an adaptive
speed control over a large operating range with load and parameter variations, A multi-
layer ANN structure with one feedback loop is implemented. This strategy includes
equally on-line and off-line weight and bias updating for the ANN utilizing the back-
propagation algorithm. Stability across a large array of operational positions was
achieved by utilizing an ANN structure with feedback loop. Even Though the control
system stability has been enhanced, the estimated system reactions have significant
amounts of speed exceeding in certain working situations. Because the learning speed is
non-adaptive throughout the on-line weights and biases updating.
In [12] Rubaai and Kotaru deal with the challenge in an extra conventional way. Effort
was not rendered to linearize the changing aspects of electric motor, maintaining the
trustworthiness of the model entirely. Dynamics of motor are modeled online and
managed employing a dynamic backpropagation neural network. Two control topologies
were discussed. previous experience of the load dynamics is not presumed in each
topology, whilst the subsequent topology as well believes no knowledge of the motor
boundaries. An adaptive learning algorithm that utilizes an adaptive learning rate for
training the neural network is introduced.

You might also like