Intended learning outcomes: Explain the organization of an expert system (or knowledge-based information system). Describe a production rule.
It is hard to find a precise definition of the term expert system in the literature (see [Apel85]). One practical definition relates, in particular, to the way in which an expert system works:
Expert systems are knowledge-based information systems. Such systems firstly attempt to represent large amounts of knowledge concerning a limited application in a form that is suitable for the particular problem; secondly, help to acquire and modify this knowledge, and thirdly, at the user’s request, draw conclusions from the knowledge and make the result available to the user.
Here, the term knowledge incorporates all the stored information that is needed to answer queries. Most expert systems differentiate among facts, rules (i.e., knowledge about the facts), and metarules, i.e., knowledge about the rules.
The term fact base is used to describe the rules as a whole. The term rule base designates the rules as a whole.
The inference engine is a programming logic that applies rules to facts in order to derive new facts so as to answer questions.
Figure 17.3.1.1 illustrates the interaction between the various components of an expert system and its users for the purposes of design and operation.
Fig. 17.3.1.1 Organization of an expert system (or knowledge-based information system).
- A programmer is responsible for designing the system.
- An expert drafts and maintains the rules and any metarules.
- The user records and maintains the facts.
- The user starts the inference engine in order to make a query.
It must be possible to operate an expert system without the help of programmers. It must be possible to make queries on the expert system without the help of experts. In practice, however, there is periodic contact between the users and experts in order to supplement or modify the rule base. The inference engine is independent of knowledge and facts. If the knowledge changes, the inference engine must not change in any way.
The rules of a knowledge base can be presented in different ways. The simplest and the most intuitive form is the production rule.
A production rule is a statement of the “if (condition), then (action)” type, i.e.,
· If a certain situation is true (a number of facts), then conclude (infer) various actions (a certain number of facts).
The structure of positions in a bill of material and routing sheet, which is made conditional though the use of IF clauses (see the example in Figure 7.3.2.1), precisely corresponds to the structure of production rules in an expert system expressed regressively (from effect to cause): Here, a production rule in the true sense of the word, that is, of a product to be manufactured, corresponds to a production rule within the expert system in the applied sense.
The facts of the expert system are formed by the item, production equipment, and work center logistical objects and by the values assigned to the query parameters (e.g., for an existing order). The experts are the designers and process planners within the company. The users are the people who issue, monitor, and produce the orders. See Section 7.3.2.
The inference engine works on the chaining principle: The inferred facts can in turn occur in rules (e.g., in the IF clause of a production rule). Further facts can be inferred if the engine is then applied iteratively, particularly to this type of rule. In this case, the inference engine is generally only needed for forward chaining. By analyzing production rules containing IF clauses with the relevant parameters, it is able to return the order bill of material and order routing sheet applicable to the specified parameter values.
A more complex expert system also contains a declaration component which makes the rules transparent to the user. In practice, however, most bill-of-material positions and operations are self-explanatory. Some expert systems also suggest methods for handling incomplete knowledge or knowledge arising from conclusions by analogy. They belong to the methods of artificial intelligence.
Artificial intelligence comprises computer programs that can learn and reason in a manner similar to humans (cf. [ASCM22]).
Course section 17.3: Subsections and their intended learning outcomes
17.3 Extensions Arising from the Variant-Oriented Concept
Intended learning outcomes: Produce an overview on expert systems and knowledge-based systems. Explain the implementation of production rules. Present a data model for parameterized representation of a product family.
17.3.1 Expert Systems and Knowledge-Based Systems
Intended learning outcomes: Explain the organization of an expert system (or knowledge-based information system). Describe a production rule.
17.3.2 Implementation of Production Rules in a Knowledge-Based System
Intended learning outcomes: Differentiate between the representation of the bill of material or routing sheet for a product with options and the standard version without variants. Explain the extended primary key for a bill of material with options.
17.3.3 A Data Model for Parameterized Representation of a Product Family (*)
Intended learning outcomes: Produce an overview on parameter and parameter class. Differentiate between primary and secondary parameters. Explain plausibility or compatibility test. Identify various types of formulas.
