27
Expert system-based optimization in wrought copper alloy production Dr.-Ing. George Power, Universidad de Lima [email protected]

Expert Systems Presentation - Dr. George Power

  • Upload
    gfpp58

  • View
    20

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Expert Systems Presentation - Dr. George Power

Expert system-based optimization in wrought copper alloy productionDr.-Ing. George Power, Universidad de [email protected]

Page 2: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 2

Contents

Introduction to expert systemsRule-based expert system for job preparation in wrought copper alloy productionProgramming componentsResults obtainedSummary and conclusions

Page 3: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 3

What is an expert system?

Software for specific problem areas involving expert knowledgeDeveloped after AI’s failure of “General Problem Solver”Commercial XPS can be:

Case-basedRule-basedDecision trees

Page 4: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 4

Case-based expert systems

Databases of specific problems and their solutionsFor a given problem, the expert system tries to find a similar case and its solutionExample: Medical diagnosis –appropriate treatment for a patient with concrete symptoms

Page 5: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 5

Rule-based expert systems

Have a large set of “if-then-else”rules, algorithms, formulas, databases, etc.Business Rule Management Systems (BRMS) are specific for business aplicationsRules represent general cases from which solutions are drawnRules must be entered and maintained by experts

Page 6: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 6

Decision trees

Expert systems for classification tasksInductive learning based on a set of examplesObject attributes are knots of a search path or tree, characteristics represent linksGoal is the leaf or end knot gives the object class to be assigned

Page 7: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 7

Individual roles in expert systems

End-userInteracts with the XPS and can also supply some expertise

Problem domain expertBuilds and supplies the knowledge base, can also be the programmer

Knowledge (software) engineerAssists the expert in representing his knowledge

Page 8: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 8

Semis of wrought copper alloys

Source: www.wieland.de

Page 9: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 9

Manufacturing of flat products

Source: www.wieland.de

Page 10: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 10

Some fabrication parameters for Cu alloy sheet and strip

Material description (alloy type, temper, standards, …)Other technological properties (flexural, die wear, soldering tests) Hardening and softening behaviour (cold-work and annealing curves)Surface finish, protectionFabrication steps: hot rolling, cold rolling, annealing, pickling, stretching, slitting, traverse winding, etc.

Page 11: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 11

Why an expert system for job preparation?

Time-consuming consults to expertsHuge amount of data in production and customer specific proceduresDelays in job preparation increase lead times in PPSOrder changes and cancellations (storno)Reduce overhead costs

Page 12: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 12

Team structure

Material properties, machine specifications, databases, formulas

Quality Engineer

Machine setup, production, maintenance and down time, cost factors

Time and cost calculation

Object structure, production steps, resources (machines, materials), input and output forms

Job preparation

Programming tasksDepartment

Page 13: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 13

enginObject: An object-oriented programming tool for XPS

MaterialsMachines

Tools

tn = …

Preparation

Set up

SolderingDecision tables Databases

Formulas

Texty blocks

Development system

Planingdocuments Knowledge base Planing

algorithms

Shop specific planingknow-how

Job plans

CAD

CAM

PPS

SQLCreate plan

Edit / actualize job plans

SML

User system

Source: camos GmbH, 1994

Page 14: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 14

enginObject structureComponents:

Decision tablesProceduresDatabasesForms2D tablesFormulasMethodsObjectsVariables

Development system:

Class browser“Compiler”Interpreter / Debugger

User system:Class browserRun-time Interpreter

Page 15: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 15

Decision tablesAllow precise modelling of complex logic structures in a compact formatClear table structure replaces several nested “if-then-else” clausesConditions, rules and actions are coded as enteredStandardized in DIN 66241 (1979): “Decision tables for information exchange”A special case of decision tables are two-dimensional (2D) tables

Page 16: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 16

Decision tables

Conditions

Rules

Actions

Page 17: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 17

More examples

Decision table with priorization of machine selection

2D table for rolling cost factor assignment.

Page 18: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 18

Procedures

Allow semi-structured text programming of sentences, loops, arrays, etc.Syntax similar to Basic or Pascal without strict variable declarationCalls to other components (decision tables, formulas, other procedures) within the same object

Page 19: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 19

Procedures

Logical decision

Call decision table

Call formula

Comment line

String constant

Numeric constant

Reserved word

Variable (variant type)

Page 20: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 20

Databases

Allow access to structured data (material properties, standard tempers, etc.)Simple query structuresAdvanced queries with SQL server for accessing external databases

Page 21: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 21

Databases (example: material properties)

Page 22: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 22

Forms (input/output)

Page 23: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 23

Class browser (object tree)

Colapsed treestructure

Expanded treestructure

Page 24: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 24

Interpreter and debugger

Useful for test runs during program developmentOnce tested, a run-time version of the expert system can be “compiled”File extension is “.knb” (knowledge base)Users access expert system from a local or remote server through the run-time interpreter.

Page 25: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 25

Interpreter

Page 26: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 26

ResultsSystem programmed and maintaned by same experts. Realization time ca. 8 months10+ years in use, extended to other production sites running remotely on same Citrix serverReduction of personnel in job preparation, employes can be reoriented to other activitiesReduced lead timesIncreased product qualityMore precision in cost calculationXPS never forgets a single rule or procedure less quality claimsSuccesful experience in job preparation applied for another expert system: technical assistance of sales department.

Page 27: Expert Systems Presentation - Dr. George Power

02/11/2006 Copyright © George Power 27

ConclusionsExpert systems allow efficient management of one of the company’s main assets: technical know-howProgramming tools are easy to learn. Expert and programmer tasks can be performed by the same personnel within the company, freeing IT resourcesKnowledge base can be accesed remotely and must be maintained regularly by the same programming teamLimitations in structured programming aren’t a drawback: they simplify documentation and maintenance.