Introduction to knowledge based applications,

applications of artificial intelligence (AI) in chemical

engineering, AI principles
Introduction to Expert Systems

A. Description of Expert Systems

• Expert systems are one of the fastest growing applications of AI in the scientific and
engineering fields. Expert systems attempt to match the performance of human
experts in a given field.
• To do so, these systems rely on in-depth, expert knowledge.
• The better the knowledge, the better is the performance of the system.
• Knowledge is usually incorporated into expert systems through relationships.
• An expert system keeps track of relations and inferences invoked.
• Therefore, the knowledge used by the system is explicit and accessible to the user.
• An expert system can explain why certain information is needed, and how certain
conclusions are reached.
• Some advantages of expert systems are that they:
• (1) can assimilate large amounts of knowledge; and
• (2) never forget that knowledge
• These properties distinguish expert systems from conventional computer programs.
Ideally, an expert system can build its own knowledge base
• Expert systems use a combination of user interface and inference mechanisms,
sometimes called the expert system shell, for the expansion of the knowledge.
• An ideal expert system, shown in Figure 10.3, contains:
• (1) a knowledge base;
• (2) an inference engine; and
• (3) a user interface.
• The knowledge base contains specific, in-depth information
about the problem at hand.
• That knowledge consists of facts, rules, and heuristics,
as shown in Figure 10.4.
• To utilize the knowledge (facts, rules and heuristics), an expat system relies on its inference
engine.
• The inference engine uses inference mechanisms to process the knowledge and draw conclusions.
• The user interface provides smooth communication between the program and the user.
• As stated previously, the inference engine and the user interface are combined under the term
expert system shell, or simply the shell.
• Ideally, the shell:
• (1) answers how a conclusion was reached;
• (2) answers why certain information is needed; and
• (3) has the ability to add knowledge to the knowledge base.
• Note that in the absence of the knowledge base, the shell is just that—an empty shell that can do
nothing.
• But importantly, the shell has the ability to work with and update the knowledge base.
• Knowledge can be added
• 1) by a human expert who is not a programmer, or 2) by the expert system itself as a result of
inference.
• The ability to perform these duties makes expert systems powerful, flexible "thinking machines."
Usee of Expert Systems in Science and Engineering
 Representing Knowledge in Expert Systems
1. logic-based systems
2. frame-based systems
3. object-oriented programming.
• Logic-Based Systems
• Logic-based systems use principles and constructs of mathematical logic to represent and process
information.
• Many logic-based knowledge representations exist The two most common logic-based representations used in
science and engineering are rule-based and fuzzy-logic systems.
• 1. Rule-Based Systems
• Using rules is one of the most common ways to represent knowledge in expert systems.
• Rules are conditionally true, and can be viewed as IF-THEN statements.
• For instance, the Prolog statement

• says that X is flammable if X is a petroleum derivative.

• Rule-based expert systems use pattern-matching.
• As an expert system consults its knowledge base to answer a question, it develops an inference chain.
• The inference chain is a sequence of steps or rule-applications used by the expert system to analyze and
solve problems.
• In the above flammable statement, the inference chain says that we can infer that X is flammable if we can
prove that X is a petroleum derivative.
• Which is better, forward or backward chaining?
• The answer depends on the specific application. If an application is very data-
dependent, with less need for inferences, we may choose forward chaining.
• If we have complex relationships that require in-depth inferences with less data, we
may prefer backward chaining.
• Expert systems can use either forward or backward chaining.
• One advantage backward chaining has over forward chaining is that backward chaining
is goal-oriented.
• A forward-chaining system has a tendency to establish everything it can before
stopping, and therefore tends to be "trigger happy."
• Without careful control, rules can "fire" uselessly.
• Consequently, forward-chaining systems tend to wander in search of the result if much
inference is needed.
• 2. Fuzzy-Logic Systems
• Fuzzy logic grew out of a desire to quantify rule-based systems.
• Rule-based reasoning is grounded in qualitative knowledge representation, and fuzzy
logic allows us to mesh a quantitative approach with the qualitative representation.
• Fuzzy logic is used to quantify certain qualifiers such as approximately, often, rarely,
several, few, and very.
• Fuzzy logic is not a substitute for statistics.
• Indeed, fuzzy logic is used when statistical reasoning is inappropriate.
• Statistics is used to express the extent of knowledge (or lack thereof) about a value,
and relies on tools such as variance, standard deviation, and confidence intervals.
• Fuzzy logic, on the other hand, is used to express the absence of a sharp boundary
between sets of information.
• For example, we may write:
• Crude oil fractionation is an energy-intensive unit operation, 1.0.
• Thermal cracking is an energy-intensive unit operation, 0.9
• Catalytic reforming is an energy-intensive unit operation, 0.6
 • Catalytic cracking is an energy-intensive unit operation, 0.3.
• Open-air evaporation of brine to produce salt is an energy-intensive unit operation,
0.0.
• We use fuzzy logic to delineate the lack of a sharp boundary between clearly energy-
intensive (1.0) and notat all energy-intensive (0.0).
• Crude fractionation is very energy-intensive, while open-air evaporation of brine is not
at all energy-intensive. Thermal cracking, catalytic reforming, and catalytic cracking
cannot be considered either very energy-intensive or not at all energy-intensive.
• Thus, fuzzy logic is not used in a statistical sense to quantify the lack of knowledge.
• Instead, fuzzy logic is used to quantify the degree or extent of certain words and
boundaries between sets of information.
• To use fuzzy logic, we first need a fuzzy set.
• In a fuzzy set, the transition from membership to non membershipis not well-defined.
• We quantify the degree of membership with values between 0 (not a member) and 1
(definitely a member).
• With our energy-intensive unit-operation example, the fuzzy set is:
• Identifying Expert System Possibilities
• Determining the Appropriateness
of Expert-System Development
Justifying Expert-System
Development
• Fault diagnosis
• Process control
• Process design
• Planning and operations
• Modeling and simulation
• Product development
Uses of AI include: