2. Introduction
In the past four decades considerable
research has been performed in the field of
artificial intelligence (fuzzy logic, neural
networks, fuzzy–neural networks, genetic
algorithms, etc.), which has resulted in many
industrial applications.
3. Cont.
This covers the details of AI-based machine
design, control, parameter and state
estimation (e.g. speed, position, flux, torque,
etc., estimation), condition monitoring and
diagnosis, simulation (steady state and
transient), various firing schemes, etc.
4. Artificial Neural Networks (ANNs)
ANNs are universal function approximations
and are capable of closely approximating
complex mappings, which can be extended to
include the modelling of complex, nonlinear
systems.
5. Cont.
Much research has been done to improve the
learning speed of the backpropagation
algorithm, but some of these increase the
complexity of the neural network. It is also
possible to enhance the backpropagation
learning by using fuzzy concepts.
6. Fuzzy Logic Systems (FLSs)
Fuzzy logic systems (FLSs) are also
universal function approximators. However,
a fuzzy logic system is an expert system,
where the heart of a fuzzy logic system (FLS)
is a linguistic rule-base, which can be
interpreted as the rules of a single ‘‘overall’’
expert, or as the rules of ‘‘sub experts’’
7. Cont.
and there is a mechanism (inference
mechanism) where all the rules are
considered in an appropriate manner to
generate the output(s).
8. Cont.
These rules may directly originate from
experts, but if the experts are not available,
they can also be obtained by the appropriate
processing (e.g. clustering) of available
input=output data. Again it is important to
note that the system can be nonlinear, and
nonlinearity is incorporated into the fuzzy
logic system.
9. Genetic Algorithms
(GAs) Genetic algorithms (GAs), are not
function approximation techniques, but they
are simple and powerful general purpose
methods (learning mechanisms), which have
been inspired by the Darwinian evolution of
a population subject to reproduction,
crossover and mutations in a selective
environment where the fittest survive.
10. Cont.
GA combines the artificial survival of the
fittest with genetic operators abstracted from
nature to form a very robust mechanism that
is suitable for a variety of optimization
problems.
11. AI Applications in Electrical
Machines and Drives
In the literature most publications on the
application of artificial intelligence (AI) in
electrical drives discuss fuzzy logic-based, or
seldom neural-network-based, speed or
position controller applications, where an
existing PI or PID controller is simply
replaced by an AI-based controller.
12. Cont.
Although this is an important application,
and the AI-based controllers can lead to
improved performance, enhanced tuning and
adaptive capabilities, there are further
possibilities for a much wider range of AI-
based applications in variable speed ac and
dc drives.