Elements Of Expert System And Knowledge Representation
Expert systems are not concerned along with understanding language, or other aspect of intelligence, but are only concerned along with solving problems. As distinct from other modules of artificial intelligence, expert systems are not based upon any common theory of intelligence or universal science of cognition. Expert systems are based upon the engineering premise that intelligence lies in knowledge; which is, people who know more exhibit more intelligence than those who know less. Thus, it follows from this premise that, if sufficient knowledge can be represented in a computer, the computer will display intelligent as like behaviour.
One of the first expert systems was the program MYCIN, developed in the early 1970s at Stanford University to assist medical doctors in the diagnosis of bacterial infections in human patients. From this beginning, many diagnostic expert systems have been developed in manufacturing and engineering, like it turns out that the reasoning included in the cause of problems is suitable to represent by expert systems. A report by National Academy of Sciences defines expert systems to having two parts: a body of knowledge and a mechanism of interpreting this knowledge or National Academy of Sciences. The body of knowledge in turn has two parts: rules or heuristics that represent the knowledge to solve problems in a exact domain, and facts about the specific problem to be solved. The objective of expert systems development is to capture the knowledge of an expert in an exact area, to computerize this knowledge, and to transfer it to other experts.
An expert system then contains all the given elements as:
Ø An inference engine
Ø A knowledge base separate from inference engine
Ø A data base
Ø An explanatory facility
The inference engine draws logical inferences or conclusions, from the expert knowledge contained in the knowledge base and the exact problem conditions contained in the database. One of the discoveries that made expert systems helpful was the realization that by separating the inference engine from the knowledge base, the inference engine may be utilized in any number of expert systems.
The knowledge base is a computer represent of the knowledge of the domain expert, expressed in one of several forms, for illustration, as rules. A knowledge base is exact to each expert application.
The database is a set of input data, pertinent to exacting problem, which is utilized by the inference engine to reach a conclusion. These data may be acquired from the user of the expert system, from a computer database, from a CAD system, or from the output of other computer program.
Expert systems also involve explanatory facilities, which explain to the user how or why the expert system has arrived at a particular conclusion. Which is, if the user asks the expert system to explain its results, the explanatory facility determines the exact knowledge, from the knowledge base, combined along with the user's input data, which lead to this conclusion and displays these to the user. The explanatory facility makes the expert system easier to employ, easier to understand, and easier to accept. The existence of an explanatory facility is one of the features that differentiate an expert system from any other computer program.
A rule base is the most general kind of knowledge base. In a rule base, a domain expert's knowledge is expressed in the form of production rules, all of which consists of a set of conditions and set of consequents. These production rules are generally expressed in a form as close as possible to natural language for example English. A typical rule is of the form
IF <Condition A is true> AND <Condition B is true>
AND <Condition N is true>
THEN <Conclusion X can be drawn> ELSE <Conclusion Y can be drawn>
If the conditional statements are each evaluated to be true, then the rule is said to fire, and the consequent statements are asserted to be true too. The consequents of one rule may be utilized as the conditions of other rules. It produces a chain reaction effect, as the firing of one rule leads to the firing of others; such chain reaction is carried out by the inference engine.
A particular expert system doesn't consist of a few rules, each along with a very large number of conditions and consequents, however rather a large number of rules, each with a few consequents and conditions.
Rule based system's chain reaction behaviour is of three types: backward chaining, forward chaining, and mixed. In forward chaining, or data driven systems, the inference engine employs the given data as from the user or from the data base and determines whether any rules can be fire. If so, then chaining process continues until one or more conclusions or terminal consequents are reached or no extra rules can fire. Before a conclusion is reached if rule firing stops, the inference engine asks the user for more information, which may reason new rules to fire.
In backward chaining, and goal driven systems, the inference engine posits a possible conclusion and then chains backward through the rules from this conclusion to the data. If the conditions required sustaining the hypothesis is incompatible along with the actual data, then the conclusion is rejected and a new conclusion is hypothesized. It continues until one or more conclusions have been validated by the data or all possible hypotheses have been refused. Mixed chaining is a composite of backward chaining and forward chaining.
Not all expert systems are based upon rules. Some are based on another knowledge representational scheme, as like frames or objects. Though, rules are the easiest for the designer or manufacturing engineer to know and are the basis for most available expert system shells.
The domain expert, who possess the knowledge or expertise, interact along with a knowledge engineer, who is experienced in asking questions and converting the expert's answers into rules. The knowledge engineer is experienced also in choosing which inference method to apply and in structuring the rule base efficiently. From this collaboration the rules resulting of domain expert and knowledge engineer then from the knowledge base.
A user interacting along with a completed expert system so the knowledge base in the computer has been created earlier via a collaboration of the domain expert and the knowledge engineer. The user supplies the input data exact to this problem, or data problem may be extracted from a computer data base or CAD system. The inference engine compares the user's input along with the rules in the knowledge base to verify if it can arrive at any conclusion. If not, then inference engine will ask the user for more information. The user may ask or inquire the inference engine why it requires this information, or why it has arrived at an exact conclusion, through the explanatory facility.
Subsequent to the knowledge base is first developed; this should be tested by the user, in order to refine the knowledge base and to make the system reliable and easy to employ. An iterative process is preferred. This is generally more efficient to start along with a few rules, test them in practice or in a prototype system, and after that to add rules to increase the depth of knowledge and to handle conditions not originally contemplated.
Because the rules in the rule base are expressed in English, or something as like English, rules may be aided over a period of time by the original domain expert or by others. Such evolutionary improvement is characteristic of expert systems, like compared to other types of computer programs. If a lot of domain experts contribute their knowledge to a rule base, the expert system over time may become extra knowledgeable and capable of problem resolving than any one of the experts who contributed to its development. It is particularly true of expert systems that manufacturing knowledge available to designers throughout CAD systems as: the rule base may become a synthesis of manufacturing know- how beyond the capabilities of several single persons.
The demonstrated gains of expert systems include making manufacturing and design expertise more wildly available in the organization, enhanced design decision-making, and enhanced performance of design personal and manufacturing equipment. Manufacturing and Design engineering are fertile fields for the application of the expert systems since much expertise is heuristic - and is thus generally suitable for representation by rules. The English as like nature of the rules lends itself to knowledge base improvement by practicing industrial engineers and designers, and experience has implied that practicing engineers can readily express their expertise in rule form.
A lot of early expert systems were stand alone applications, however the full potential of expert systems in engineering and design can merely be achieved by integration of the knowledge base along with the manufacturing database and the computer aided design system. Stand - alone expert systems need excessive manual data input to be truly helpful in practical design situations. By integrating an expert system shell along with the database and the CAD system, the expert system can automatically interrogate the design database for any necessary information to satisfy their rules. The results of the expert system session can be automatically inserted also into design model.
With such integration, designers can execute expert systems interactively when in the process of design utilizing the CAD system. The expert system can retrieve any essential information from the CAD model or from the database. Advice given by the expert system is displayed on the workstation screen during the computer aided design and engineering session. The output of the expert system can be geometric information also, which is inserted directly into the computer aided design and engineering models. In such way, expert system can advice the designer, make suggestions to enhance the design for manufacturability, change the dimensions of objects in the design model, or even produce entire designs automatically. As the familiar engineering analysis programs, expert systems free engineers from tedious details in order to provide them more time to deal along with the significant issues of design and engineering.