+ All Categories
Home > Documents > 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31...

2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31...

Date post: 27-Aug-2019
Category:
Upload: dinhkhuong
View: 214 times
Download: 0 times
Share this document with a friend
26
2. Lecture Expert Systems Soft Control (AT 3, RMA)
Transcript
Page 1: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

2. Lecture

Expert Systems

Soft Control

(AT 3, RMA)

Page 2: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

24 WS 17/18 Georg Frey

2. Outline of the Lecture

1. Introduction to Soft Control: Definition, Limitations, Basics of

“SMART” Systems

2. Knowledge representation and knowledge processing

(Symbolic AI) Application: Expert Systems

3. Fuzzy-Systems: Dealing with Fuzzy Knowledge

Application: Fuzzy-Control

4. Connective Systems: Neural Networks

Usage: Identification and Neural Control

5. Genetic algorithms: Stochastic Optimization

Application: Optimization

6. Summary & Literature

Page 3: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

25 WS 17/18 Georg Frey

Contents of the 2nd Lecture

1. Expert Systems

1. Idea

2. Areas of applications

3. Compared with conventional programs

2. Basic Architecture of Expert Systems

Explanation of the components

Forms of knowledge base

Inference mechanism

3. Case Based Reasoning

4. Summary

Page 4: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

26 WS 17/18 Georg Frey

Expert Systems

• Core idea (Natural Model)

Human-like abstract thinking

• History

First expert systems began in 1970's (though faced the problem of high

computing expenses)

• Application in Automation Engineering

Today: Manifold industrial use at higher levels of automation

• Examples

Expert systems to support process control

Expert systems for fault diagnosis

Training Systems (simulators)

Page 5: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

27 WS 17/18 Georg Frey

Objectives of Expert Systems

Page 6: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

28 WS 17/18 Georg Frey

Comparison: Conventional Programs vs. Expert Systems

In expert systems the knowledge & the problem-solving strategy

are separated, while in conventional programs knowledge and

problem-solving strategy are implicitly embedded in algorithms.

Algorithms

Data

Knowledge

Data

Troubleshooting

-strategy

Conventional Programs Expert Systems

Page 7: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

29 WS 17/18 Georg Frey

Architecture of Expert Systems

Quelle: Lunze

Page 8: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

30 WS 17/18 Georg Frey

Knowledge Acquisition Components

• The knowledge of the system is changed or reviewed by introducing Interface to the expert systems

• Ideally, the interface is designed such that no programming knowledge is required

• Very often, there is a system developer (knowledge engineer) between the knowledge acquisition component and the expert system

• The knowledge engineer supports the experts in the command input or he usually asks the expert’s knowledge in interviews and prepares the input according to it.

• Most importantly, knowledge acquisition creates a bottleneck in the expert system creation

Page 9: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

31 WS 17/18 Georg Frey

Knowledge Base

• Here the existing knowledge in the system is suitable saved

• There are three different components of the knowledge base

1. Array based knowledge: The actual knowledge base, is the knowledge, that is

entered by knowledge engineer or expert

2. Specific case knowledge: knowledge about the current system problem, which

is entered by the user or automatically,

3. Interim results: Results that have been produced by individual rules of the

problem-solving components can serve in other rules and further processing

Page 10: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

32 WS 17/18 Georg Frey

Problem Solving Components

• Application neutral (application independent) part of the knowledge

Processing (also inference machine)

• Program system, the means of knowledge basis generation for

problem-solving

• Construction of the inference machine depends on the type of

knowledge base

Chain-rule (forward, backward) for production control systems

Derivation (resolution) for logic based systems

Comparison with case-based reasoning system

Page 11: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

33 WS 17/18 Georg Frey

Dialog Components

• Interface for users of the expert system

• Ask the problem from suitable

• Presentation and explanation of the solutions

• Explanation of the problem

System can be transparent, as it was concluded. (Useful for troubleshooting and

necessary confidence-building)

Deep explanations for "why" and “for which reason" a solution is possible or not

possible.

The XPS approaches in the deep structure of problems; that are still in their

infancy ( deep structure = the problem underlying technical, physical or

chemical relationships)

Page 12: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

34 WS 17/18 Georg Frey

Forms the Knowledge Base

• Control Based Systems

Production Control Systems

Logic Oriented Systems

• Case Based Reasoning Systems

• Object-Oriented Systems (are not discussed here)

Page 13: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

35 WS 17/18 Georg Frey

Production Control Systems

• Production rules as the smallest building blocks of knowledge base

• Construction:

IF certain conditions are met,

THEN will be closed on the following facts

• Example: failure diagnosis (syntax of the system Babylon)

Page 14: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

36 WS 17/18 Georg Frey

Example: Failure Diagnosis (2)

Page 15: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

37 WS 17/18 Georg Frey

The General Approach to the Inference

Quelle: Lunze

• MATCH: It examines

which rules are

applicable in the

current problem state

conflict quantity =

quantity in all

applicable rules

• SELECT: selecting a

rule from the conflict

quantity

• ACT: application of

the selected rule to

the current problem

state new problem

state

Page 16: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

38 WS 17/18 Georg Frey

Inference: Forward Chaining vs. Backward Chaining

• Forward Chaining:

Carried out on the basis of known facts

Assumptions for these are new facts

Interruption, if no rule can fire more

Undirected (Untargeted) search

• Backward Chaining

Carried out on a hypothesis that would find the rule for re-check

A rule (to be found); that is conclusion to hypothesis

Then, with the assumption of this rule; proceed to the already known rules

• Mixed Strategy

Initially Hypothesis formation, (forward) = selection of the rule that is valid for

most of the assumptions

Then Hypothesis check (backward) = attempt the missing assumptions to verify

• For all chaining methods search may still exist between depth and

the breadth

Page 17: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

39 WS 17/18 Georg Frey

Example: Forward Chaining

Page 18: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

40 WS 17/18 Georg Frey

Example: Forward Chaining

• Rules are given in the form

of a decision tree

• Problem: Find reason for

low pressure

Page 19: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

41 WS 17/18 Georg Frey

Logic Based Systems

• Based on the statements or first order logic

• Use of variables is allowed

• Application e.g. Logic oriented programming PROLOGUE

• Knowledge base is developed from facts (statements) and rules

(implications)

• Example:

Statements:

A1 = stirrer runs

A2 = inert atmosphere is concerned

A3 = footway is off

A4 = dosage runs

A5 = dosage does not work

Implications

(A1 & A2 & A3) A4

A3 A4

Instructions

Be A1, A2, A3 TRUE then the value of A4 will also be derived to TRUE

Page 20: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

42 WS 17/18 Georg Frey

Logic Based Systems: Extended

• Default logic: For unknown assumptions are default values are (generally) TRUE

• Multi-logics: Z.B. tetravalent (TRUE, probably TRUE, FALSE Probably, FALSE)

• Modal logic: "It is possible that ... applies "

• Auto-epistemic logic: "I believe that ... applies "

• Temporal logic: the light temporal relations: "A, applies after B“

• Inclusion of probability statements in the logic

• Fuzzy Logic: details in the section of fuzzy systems

Page 21: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

43 WS 17/18 Georg Frey

Problems Rule-Based Systems

• Basic assumption that technical expertise expressed in rules can be

at least questionable

• Rules are often encroached, control logic in the system to encode

• In rules the context (scope) is often encoded

• In rules structural relations are often encrypted, for example,

specializations of some rules that other rules

• Rules can not be structured and organized

Mixing of knowledge (facts, context, control) means that expert

systems are not able to create their own system behavior that is

sufficient to explain .

Lack of structure leads to the technical applicability limits (generally

sensible systems have thousands of rules)

Way out: Case based reasoning and object-oriented approaches

Page 22: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

44 WS 17/18 Georg Frey

Case Based Reasoning Systems

Case Based Reasoning, CBR

• Case is some kind of experience in solving a problem .

• Use of the experience, or a case, the solution for a sufficiently similar problem to new current problem can be applied.

• Special development of knowledge-based systems

• Cases from other analogous or same area (different domains: the solar system and atomic model) Case Retrieval (solution unchanged over)

Close Case (Case adapt)

• Renunciation of truth, instead usefulness (optimization problem)

• Approach is based on dynamic memory (basic assumptions) Remembering and adjust (adaptation) are key in understanding mental

processes

Indexation is important for remembering

Understanding leads to the reorganization of memory, which is why this dynamic

The memory structure for the knowledge processing are the same as for the storage of knowledge

Page 23: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

45 WS 17/18 Georg Frey

CBR: Application Requirements

1. There must be sufficiently many experiences available

2. It is must be easier to use this experience as the problems may be

solved directly

3. The use of the solutions on the case by case basis should not

conflict with safety requirements

4. The available information is incomplete or vague and uncertain

5. A modeling within the meanings of traditional knowledge systems is

not easily available

• Frequently used in the area of diagnostics and electronic sales but

also configuring and planning

Page 24: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

46 WS 17/18 Georg Frey

CBR: Approach= CBR-Cycle

Page 25: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

47 WS 17/18 Georg Frey

CBR: Example

• Appropriate selection of similarities among cases

• How does one define similarity

1. Definition of similar dimensions for the individual attributes (local similarity

degree)

2. Summary of an overall dimensions (e.g. weighted sum)

Page 26: 2. Lecture Expert Systems - uni-saarland.de · expert system creation . SC WS 17/18 Georg Frey31 Knowledge Base • Here the existing knowledge in the system is suitable saved ...

SC

48 WS 17/18 Georg Frey

Summary and Learning of the 2nd Lecture

To know what Experts Systems are

Application areas of Expert Systems

To know basic architecture of Expert Systems and its components

To know basic types of Expert Systems and their functional

principle:

Production Control Systems

Logic Based Systems

Case based systems


Recommended