Introduction to Expert Systems (2 of 2)

download Introduction to Expert Systems (2 of 2)

of 28

Transcript of Introduction to Expert Systems (2 of 2)

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    1/28

    3/16/2011 Total Slides 27 1

    INTRODUCTION TOEXPERT SYSTEMS

    Lecture-2/2

    By

    Dr. M. Tahir Khaleeq

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    2/28

    2

    Benefits of Expert Systems

    1. Increased output and productivity

    Increased outputfever workers and reduced cost

    2. Increased quality

    3. Reduced downtime

    4. Capture of scarce expertise

    5. Flexibility

    6. Easier equipment operation

    7. Elimination of the need for expensive equipment.

    8. Operation in hazardous environments.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    3/28

    3

    9. Accessibility to knowledge and help desks

    10. Reliability

    11. Increased capabilities of other computerized systems.

    12. Integration of several experts opinion.

    13. Ability to work with incomplete or uncertain

    information.

    14. Provision of training.

    15. Enhancement of problem solving.

    16. Ability to solve complex problems.

    17. Knowledge transfer to remote location.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    4/28

    4

    Problems of Expert Systems

    1. Knowledge is not always readily available.

    2. Expertise is hard to extract from humans.

    3. It is hard, even for a highly skilled expert, to abstract

    good situational assessment when he or she is under time

    pressure.

    4. Users of expert systems have natural cognitive limits.

    5. Expert systems work well only in a narrow domain.

    6. Most experts have no independent means of checking

    whether their conclusions are reasonable.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    5/28

    5

    7. The vocabulary, or jargon, that experts use for

    expressing facts and relations is frequently limited and

    not understood by others.

    8. Help is frequently required from knowledge engineers

    who are rare and expensive.

    9. Lack of trust by end-users may be a barrier to expert

    system use.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    6/28

    6

    Types of Expert Systems

    Expert Systems may be classified as following: Method of knowledge representation

    Nature of system

    Requirement of the system

    Nature of the application.Method of Knowledge Representation

    1. Rule-Based Expert Systems:

    The knowledge is represented as a series of

    production rules based on human expertise. Because the technology of rule-based systems is

    relatively well developed, most expert systems

    currently being produced are rule-based.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    7/28

    7

    Ex: Consider the likely behavior of an engineer with

    a great deal of repair experience.

    He looks briefly at the console. Noting the pattern of lights and error message.

    Goes over to one of the cabinets, open it,

    Pulls out the faulty circuit board

    Insert the healthy one Restart the machine.

    2. Model-Based Expert Systems:

    The knowledge is represented using a model of a

    system that simulates the structure and function of thesystem.

    Model-based expert systems are especially useful in

    diagnosing equipment problems or troubleshooting.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    8/28

    8

    Ex: Consider the new engineer fresh from training.

    a great deal of repair experience.

    He carefully notes the sysmptoms Gets out a thick book of schematics and spends the

    next half an hour over them

    At last he goes over to one of the cabinets, opens

    it, Replace the faulty circuit board with the healthy

    one and restart the machine.

    3. Frame-Based Expert Systems:

    The knowledge is represented as frames.

    A representation of the object-oriented programming

    approach.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    9/28

    9

    4. Hybrid Expert Systems:

    These systems include several knowledge

    representation approaches. Ex: Frames and Rules.

    Systems Classified by their Nature:1. Evidence gathering:

    Gather the evidences, which lead to the goal2. Stepwise Refinement:

    It deals with the large numbers of possible outcomes bymeans of successive levels of detail.

    3. Stepwise Assembly: The subject domain can have an extremely large

    number of possible outcomes. Special type is called a catalog selection.

    Deals with problems like choosing the thingsfrom a catalog choices.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    10/28

    10

    Requirement of the System: Expert systems can be developed to meet the particular

    needs of a user, which are called Custom-made ExpertSystems.

    Expert systems can be purchased as ready-made

    packages for any use. The systems are called Ready-made (Turnkey) Expert Systems.

    Ready-made systems are less expensive than the

    customized systems.

    Ready-made systems are general in nature so may not beuseful in complex situations.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    11/28

    11

    Nature of Application:

    Real-time Expert Systems are systems in which there is a

    strict time limit on the systems response time, which must

    be fast enough for use to control the process being

    computerized.

    Real-time Systems always produces response by the time

    it is needed.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    12/28

    12

    Rule-Based Expert Systems (RBES) In such systems the knowledge is represented as a

    series of production rules based on human expertise.

    RBES useful for certain classes of problems, which do

    not have direct algorithmic solutions.

    Rules:

    Rule [rule-label]

    IF logical condition statement

    [AND/OR] logical condition statementsTHEN Statement conclusion

    [Else] Alternate statement conclusion

    Comments, may be inserted any where.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    13/28

    13

    Example: LYOPHILIZER DIAGNOSIS

    Lyophilizer Fault Tree

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    14/28

    14

    Rules SectionRULE 1

    IF Shelf_heating_on = YESAND Faulty_heat_control = YES

    THEN High_shelf_temp = TRUE;

    RULE 2

    IF Air_leakage = YES

    OR Vacuum_pump_failure = YES

    OR Faulty_pressure_control = YES

    THEN Inadequate_vacuum = TRUE;

    RULE 3

    IF Compressor_failure = YESOR Inadequate_coolant = YES

    OR Faulty_temp_control = YES

    THEN `High_condenser_temp = TRUE;

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    15/28

    15

    RULE 4

    IF High_shelf_temp = TRUEAND Inadequate_vacuum = TRUE

    THEN Excessive_heat = TRUE;

    RULE 5

    IF High_condenser_temp = TRUEOR Vapor_flow_throttling = YES

    THEN Low_drying_rate = TRUE;

    RULE 6

    IF Excessive_heat = TRUE

    OR Low_drying_rate = TRUE

    THEN Product_melt = TRUE;

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    16/28

    16

    Can be difficult to create. Can be difficult to maintain.

    In large rule-bases, adding a rule can cause many

    unforeseen interactions and effects => difficult to

    debug. Many types of knowledge are not easily represented by

    rules.

    Uncertain knowledge: if it is cold it will probably

    rain Information which changes over time

    Procedural information (e.g. a sequence of tests to

    diagnose a disease)

    Limitations of Rule-Based Representations

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    17/28

    17

    Where do the rules come from?

    Manual Knowledge Acquisition: Interview experts

    Expert describes the rules

    Process of translating expert knowledge into a

    formal representation.

    Automated Knowledge Acquisition:

    derive rules automatically from formal

    specifications e.g., by automatic analysis of solution manuals

    machine learning

    learn rules from data

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    18/28

    18

    Types of Rule-Based Systems

    There are two broad kinds of rule-based systemsaccording to the search strategies:

    1. Forward Chaining Systems

    2. Backward chaining Systems

    Forward Chaining Systems

    The systems use data-driven search strategy so the

    system is also called data-driven rule-based expertsystem.

    The systems draw new conclusions from existing

    data.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    19/28

    19

    Forward Chaining Systems (continue)

    In data-driven search the problem solver begins with

    the given facts of the problem and a set of rules. Searchproceeds by applying rules to facts toproduce new facts.

    The process continues until it generates a path that

    satisfies the goal condition.

    In the system the facts are represented in a working

    memory, which is continually updated. Rules in the

    system represent possible actions to take when

    specified conditions hold on items in the working

    memory. The conditions a usually patterns that must

    match items in the working memory while the actions

    usually involve adding or deleting items from the

    working memory.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    20/28

    20

    Backward Chaining Systems

    The systems use goal-driven search strategy, so thesystem is also called goal-driven rule-based expert

    system.

    The expert system starts from the goal, searches

    backward through successive sub-goals until it worksback to the facts of the problem.

    The search is the chain of moves or rules leading from

    a goal to the data.

    The process is recursive, that is, a condition in a rule

    may be a conclusion in another rule (other rules).

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    21/28

    21

    EXAMPLE(Monitoring of a Car)

    1. Two rules for the car monitoring:

    Rule 1 (Forward chaining inference)

    IF There is an overheatingOR The brakes respond slowly when pressed

    THEN Give a message to the driver to stop the car.

    Rule 2 (Backward chaining inference)

    IF The temperature meter works properly

    AND The temperature is over 120

    THEN There is an overheating.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    22/28

    22

    Facts

    Goal

    Fact: Temperature = 130 Fact: Meter-OK

    Inferred:overheating

    Brakes respond

    slowly

    Inferred: stop the car

    Forward Chaining Inference

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    23/28

    23

    Stop the car

    There is over heating

    The brakesrespond slowly

    The gaugeworks properly The temperatureis over 120

    Goal

    Yes Yes (128 )

    Sub-goal

    Goal

    Facts

    Backward Chaining Inference

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    24/28

    24

    Forwards vs Backwards Chaining

    The use of forwards or backwards reasoning to solve a

    problem depends on the properties of the rule set and

    initial facts.

    If we have some particular goal ( to test some hypothesis)

    then backward chaining will be much more efficient.

    If we have many possible ways of trying to prove

    something and we may have to try almost all of them

    before we find one of that works. In this case forward

    chaining is useful.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    25/28

    25

    Forwards vs Backwards Chaining (continue)

    Forward chaining may be batter if we have lots of thingsto prove (or if we just want to find out in general what

    new facts are true); when we have a small set of initial

    facts; and when there tends to be lots of different rules

    which allow us to draw the same conclusion.

    Backward chaining may be better; if we are trying to

    prove a single fact, given a large set of initial facts, and

    where, if we used forward chaining, lots of rules would be

    eligible to fire in any cycle.

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    26/28

    26

    An Application

    A Product of Digital Equipment Corporation XCON = rule-based expert system for computer

    configuration

    decides how peripherals are configured on new orders

    has 10,000 rules

    developed in early 1980s

    based on reactive rule-based systems

    if antecedents then action forward chaining in style

    estimated to have saved several hundred million $s

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    27/28

    27

    END

  • 8/7/2019 Introduction to Expert Systems (2 of 2)

    28/28

    28

    All lectures are available on

    www.geocities.com/mtkhaleeq/AI.htm