Expert Systems - Digital Signal Processing in Power System Protection and Control - page 293

Chapter 14
Expert Systems
Expert System (ES) is a kind of software that simulates the problem-solving
behavior of a human expert of given domain. ES can be used to solve a complex
problem or give an advice, mainly in cases when the amount of data to be
processed is very high. There is no sense to apply ESs for simple decision prob-
lems, they should be used wherever a conventional data processing (DP) solution
is not possible or may appear troublesome. Referring to Fig. 14.1 one can say that
instead of traditional DP by algorithmic, well-defined methods, expert systems
perform processing of knowledge and make intelligent searches.
Why use Expert Systems? When a situation or data to handle is very complex
and the solution of the problem is neither visible at the first sight nor can be easily
found with available explicit formulae, an expert of given domain is usually called
to perform some reasoning and give an advice. Since human experts are not always
available, and even if they are, their advice is costly, may not always be reliable or
consistent, not to mention problems with organizing meetings with them every
time when it is needed. Therefore, building an artificial ES that would replace a
human expert becomes a reasonable and cost-effective alternative for many
technical problems.
It is obvious that an ES is as good as the experts that shared their domain-
specific knowledge and the manner they solve problems. Important is also the way
the knowledge and reasoning mechanisms are stored and organized in a formalized
way by a knowledge engineer and programmers. The structure of an ES, reasoning
mechanisms and other aspects related to their application are all touched in the
following sections of this chapter.
It is good to remember that expert systems are not panacea for all conceivable
decision problems. Their usage is limited to the domain and problems they have
been designed for. The ESs are not always up to date, they do not learn, they also
do not offer any ''common sense'' (general view), and are in most cases inflexible.
And, of course, human experts are still needed to setup and maintain the system.