SCEA International Conference and Educational Workshop

1
Knowledge Based Estimating Models

• SCEA International Conference and Educational
  Workshop
• June, 2004

Dr. Johnny Gilliland
Methods and Models 1 Track
C 2004, Vought Aircraft Industries, Inc.
2
We Are Vought
1917
1922
1935
1939
1961
1965
1976
1983
1986
1992
1994
A new company with deep roots in the aerospace
marketplace
2000
3
Vought History

• The Legacy of the Vought Name Dates Back to 1917
  When Chance Vought Co-founded the Lewis Vought
  Corp (LTV).
• 1992, LTV Sold Its Aircraft Division Assets to
  The Carlyle Group and Northrop Corp, Creating a
  New Company Named Vought Aircraft Company.
• 1994, Northrop, Concurrently With Its Purchase of
  Grumman Aerospace, Acquired Vought Aircraft
  Company From Carlyle
• 2000, The Carlyle Group Purchased Northrop
  Grumman's Aerostructures Business Group, Creating
  a New Company.
• The New Business Uses the Heritage Name Vought
  Aircraft Industries, Inc. And Remains Based in
  Dallas.
• 2003, Vought Acquired The Aerostructures
  Corporation Currently in Transition to Integrate
  Operations
• Today, Vought Aircraft Industries Is a Major
  Subcontractor on Many Commercial and Military
  Aircraft Programs.

4
Company Overview Facilities
Wings Tail Sections
Fuselage Skins, Spars, Stringers, Pylons
Machined Components

Nashville, TN
Facilities - 2.17M ft2
Everett, WA
Facilities - 150K ft2
Bonding, Subassembly Door MRO



Wing Skins, Spars, Stringers, Chords Ribs




Milledgeville, GA
Facilities - 622K ft2
Military Other Commercial Wings, Empennages
Fab
Brea, CA
Fuselages
Doors, Wing Structures, Nacelles Nacelle MRO
Facilities - 90K ft2
Boeing Empennages
Stuart, FL
Dallas, TX
Grand Prairie, TX
Hawthorne, CA
Facilities - 361K ft2
Facilities - 4.96M ft2
Facilities - 1.16M ft2
Facilities - 2.65M ft2
5
Commercial Products
777
767
737
747
757
A340-300/-500/-600
A320
A330
Citation X
ERJ-170/190
Integrated Aerostructures
Subassemblies/Kits
CFM56
Components/Parts
Flexibility/Agility to Produce Aerostructures for
Business Jets to Dual-Aisle Transports
G300/400
G500/550
CF6
HAWKER 800
6
Military Products
P-3
C-17
E-8C/JSTARS
C-5
S-3
F-14
E-2C
V-22
EA-6B
RQ-4A/Global Hawk
C-130J
F-15
Participation on Key Aircraft Acquisition
and Modification Programs
T-38
F/A-18E/F
F-35/JSF
F/A-22
7
Systems Engineering Cost Estimating

• Systems Engineering Cost Estimating
• Apply on Expert Systems and Math Models
• Provide ROM Estimates When Detail Design Is Not
  Available
• Establish Design to Cost Targets
• Independent Estimating Based on Parameters
• Decision Support
• Aid in Bid/No Bid Decision-Making Process
• Analyze Sensitivity of Program/Design Factors
• Evaluate Technical/Cost Relationships to
  Determine Key Cost Drivers
• Analyze Competitive Cost Data
• Develop Market Price Benchmarks
• Optimize Design for System Engineering Trade
  Studies

8
Trade Study Methodology
Bid and Departmental Targets
Market Price
Expert System
Recurring Cost Model
Build to Print Design/Mfg
Cost Optimization
New Technology And Innovative Ideas
Investigate a Better Way
9
Creating Your Expert System

• Define the System
• Use a Commercial Expert System Shell
• Build Your Own Model
• Establish the Relationships Between the Key
  Characteristics and Historical Data.
• Logic Engineer Builds the System Using
  Historical Data and the Expert From Each
  Department
• Fuzzy Logic Is a Useful Tool When the
  Historical Data Base Is Too Small to Get Good
  Statistical Correlation.
• Multi-Attribute Utility Theory (Hierarchical
  Analytical Process) Is a Useful Technique for
  Creating This Type of System
• Spend the Time and Resources Needed
• Not a Part Time Assignment

10
Some Characteristics of Intelligence
Ability to Communicate
Ability to Learn
Creativity
Internal Knowledge
World Knowledge
Goal-Directed Behavior
11
Components

• A Rule-Based System (Production System) Includes
  a Knowledge Base, a Set of Rules Specifying How
  to Apply the Knowledge, and a Control Scheme to
  Mediate Rule Application

Knowledge
Rules
Controls
12
Artificial Intelligence Tools
INTELLIGENT COMPUTER AIDED INSTRUCTION
FUZZY LOGIC
ALGORITHMS
KNOWLEDGE AND UNDERSTANDING
KNOWLEDGE BASED SYSTEM
EXPERT SYSTEMS
HEURISTICS
HISTORICAL DATA
MACHINE LEARNING
13
Why Use an Expert System?
Application Areas
Problem Solving Diagnosis Advising Control Int
erpretation Prediction Design Instruction Trad
e Studies
14
When to Use an Expert System?

• Ultimate Users Agree That Payoff Will Be High
• Application Is Knowledge Intensive
• Task Is Repeatable - Defined Process on Best
  Approach
• Automation of Process Reduces Risk of Error or
  Cost
• A Wide Range of Test Cases Are Available
• Optimal Results Are Not Required - Early Design
  Stage
• A Human Expert Exists
• Transfer of Knowledge Is Hard, but Not Too
  Difficult

.
15
Building an Expert System

• Eight Steps Identified to Build an Expert System
• 1. Expert Systems Product Definition
• 2. Available History
• 3. Experts Identification
• 4. The Knowledge Engineer
• 5. Key Characteristics Determination
• 6. Fuzzy Logic
• 7. Decision Analysis Methodology
• 8. Innovative Concepts Incorporation

16
Building an Expert System
Step 1. Expert Systems Product Definition

• Determine What Is to Be Estimated
• Total Airplanes Major Subcomponents
• Detail Parts Avionics
• Define the Output of the Expert System
• Hours Dollars
• Weight Material Mix
• Initial Settings for Other Cost Models
• Determine the Approach
• Use Commercial Expert Systems Shell
• Build Your Own Using Fuzzy Logic
• Historical Data
• Expert Opinion

17

Building an Expert System
Step 2. Available History

• Determine What Historical Data Is Available
• In-house Programs
• Standards
• Industry Data
• Government Data
• Normalize Historical Data
• Hours Per Pound
• Dollars Per Pound
• Component Per Square Foot
• Support Ratios
• Economic Adjustments
• Derive Company Cost Estimating Relationships
  (CERs)

18

Building an Expert System
Step 3. Experts Identified

• Who Is an Expert?
• Senior Estimator
• Manufacturing Lead Person
• Shop Manager
• Senior Manufacturing Engineer
• Senior Design Engineer

Anyone Who Understands the Processes to Be
Estimated
19
Natural Language Processing

• Interfaces With Knowledge and Reasoning
• Problem
• Use of Natural Language Presumes Understanding by
  the Listener, Not Simple Decoding
• Examples
• British Left Waffles on Falklands
• Kicking Babies Considered Healthy
• Translation
• The Spirit Is Willing, but the Flesh Is Weak
• The Vodka Is Fine, but the Meat Is Rotten

Care Must Be Exercised When Building an Expert
System
20
Building an Expert System
Step 4. The Knowledge Engineer

• What Is a Knowledge Engineer?
• A Person Who Designs the Logic Paths in an Expert
  System
• May Not Be an Expert in the Subject Matter of
  That Particular Expert System
• A Person Who Understands the Decision/logic
  Process of Reaching a Conclusion
• A Person That Can Interpret the Logic Process
  Used by the Experts

21
Building an Expert System
Step 5. Key Characteristics Determination

• What Are Key Characteristics
• Key Characteristics Are the Properties of an Item
  That an Expert Uses to Estimate
• Weight Fastener Count Schedule
• Size Materials Power
• Speed Surface Contour Part Count
• Manufacturing Technology Environment
• Who Identifies the Key Characteristics Used?
• Identified by the Experts in Each Functional
  Department
• How Are Key Characteristics Used?
• Knowledge Engineers Use the Key Characteristics
  and Fuzzy Logic to Construct a Relationship
  Between the Key Characteristics and the Relevant
  Historical Data

22
Building an Expert System
Step 6. Fuzzy Logic

• What is Fuzzy Logic?
• Fuzzy Logic Is a Calculus of Compatibility.
  Unlike Probability, Which Is Based on Frequency
  Distribution in a Random Population, Fuzzy Logic
  Deals With Describing the Characteristics of
  Properties.
• Fuzzy Logic Describes Properties That Have
  Continuously Varying Values by Associating
  Partitions of These Values With Semantic Label
• Bill Is Tall
• Tom Is Short
• A One Pound Part Is Average
• A Small Part Has a Value of Less Than 0.8 Lb.
• A Heavy Part Has a Value of over 5.5 Lb.
• Fuzziness Is a Measure of How Well an Instance
  (Value) Conforms to a Semantic Ideal or Concept

23
Boolean Logic
Boolean Logic Is Binary One Law A or Not A
Black and White World
24
Why Fuzzy?

• Most Modes of Human Reasoning and Common Sense
  Reasoning Are Approximate in Nature
• Approximation of Data
• Incompleteness of Data
• Uncertainty of Knowledge
• Is a Statement Absolutely True and Auditable?
• Imprecision of Knowledge
• Inflation 3.8 Versus Low Rate
• Fuzzy Logic Handles Partial Truth Value
• Between Completely True and Completely False

25
Building an Expert System
Step 7. Decision Analysis Methodology

• Incorporates Objective and Subjective Selection
  Criteria in a Structured Approach
• Considers Relative Importance of Criteria in
  Determining the Worth of the Alternatives
• Overall Performance Is Summation of Weighted
  Utility Value of Each Criteria
• Output Is a Single Value That Represents the
  Relative Worth of an Alternative

26
Multi-Attribute Utility Analysis

• Management Science and Systems Engineering Tool
• Systematic
• Repeatable
• Accountable/Traceable
• Flexible
• Fast
• Concept Originally Applied by Economists and
  Market Researchers
• Also known as Hierarchical Analytical Process

27
The Hierarchical Analytical Process
Multi-Attribute Utility Theory
28
Weighting Factors

• Establish Weighting Factors for Each Step of the
  Hierarchy
• Determine the Value (or Utility) of Each
  Contributor to Cost
• The Sum of the Utilities Must Equal 1
• Compute a Relative Adjustment Factor

29
Some Multi-Attributes Contained in an Expert
System
T1
STEALTH
SYSTEMS
MFG. TECH.
COMPLEXITY
COMM/MIL/MOD
1.1
1.0
1.2
1.0
0.3
0.4
MIL
MOD
COM
PRE-80S
CUR
FUTURE
1.5
1.0
1.0
0.6
0.4
0.0
NONE
CUR
ADV
NONE
PARTIAL
FULL
3.0
1.2
3.0
1.2
1.4
1.0
1.0
1.5
1.5
0.7
0.8
1.0
0.4
0.6
0.4
1.0
SUB
SUP
HYP
LOW
MED
HIGH
10
50
100
150
LOW
AVG
HIGH
FLAT
CURVE
CPLX
135 Cost/Utility Functions Identified
30
Building an Expert System
Step 8. Innovative Concepts Incorporation

• Value/Producibility Engineering
• Design/Build Teams
• Design to Cost
• 2-D and 3-D Computerized Modeling
• Advanced Tooling Philosophy
• Determinate Assembly
• Automated Factory
• Integrated Design/Manufacturing Data Base
• New Materials
• Advanced Processes
• Can Do Attitude

31
Benefits From Establishing an Expert System

• Gives a Consistent Starting Point for All
  Estimates
• Captures the Thought Process of Senior Estimators
• Helps Train New Personnel
• Helps Retain the Experiences of the Expert When
  They Retire or Leave the Company
• Captures
• Rules to Be Applied
• Questions to Be Asked
• Default Values
• Puts Discipline in the Estimating System.
• Helps Support Estimator's Opinion for Proposal
  Justification

32
Top Down Parametric Model

• Used to Estimate Total Recurring Production Costs
  by Function
• Cost Drivers Include
• Weight and Material Mix
• T1 Man-hours by Material Type
• Multiple Slope Improvement Curves for Labor and
  Material
• Programmatics
• Economics
• T1 Cost and Improvements Curves for
• Avionics
• Equipment
• Raw Materials

33
Role of Top-Down Recurring Cost Model
1. Establish a Target 2. Build a Reasonable
Estimate Using Top Down Model (TOPS)
3. Assess Alternative/New Ways to
Meet the Target (SPOT)
Given Descriptions Weights Rates
Run TOPS
Research Historical Data
Analyze Estimate
Expert System Settings
Accept Target
Run SPOT
Analyze Result
Propose Alternatives
Determine What to Hold Constant
Comment TOPS and SPOT are Vought Developed
Proprietary Software Programs
34
Expert Systems and Process Modeling

• Apply Knowledge Based Expert System Techniques to
  Design for Manufacturing and Assembly (DFMA)
  Models
• DFMA Models Are Based on Rule Building
• Process Models Help Retain Expert Knowledge

DFMA Reflects Cost of Each Activity in the
Manufacturing Process
35
Process Based Cost Prediction Tool

• Design for Manufacturing and Assembly DFMA

• Implement Boothroyd-Dewhurst Design For
  Manufacture and Assembly
• Structured Multi-Functional Team Trade Studies
• Quantifies Product Designs Impact on Assembly
• CATIA Interface for Part Geometry
• Addresses Tooling and Material Handling
• Consistent Metric for Ranking Lean Design
  Concepts
• Provides Weight Versus Cost Visibility
• Lower Cost
• Apply to Advanced Systems Engineering Trade
  Studies
• Vought Tailors DFMA Database to Its Processes

36
Process Based Cost Model

• Vought Applies the Boothroyd Dewhurst Inc. Design
  for Manufacturing and Assembly (DFMA) Software
  Suite of Design for Assembly (DFA) and Design for
  Manufacturing (DFM) Concurrent Costing Models
  From For Use in Performing Cost Trade Studies
• Design for Assembly (DFA) Involves Step-by-step
  Assessment of Manufacturability Issues
• Design for Assembly Is a Systematic Procedure
  Used to Reduce Overall Product Cost Through
  Design Simplification.
• User Can Develop Relationships Based on
  Historical Data

37
DFMA Application to Trade Studies

• DFMA Provides Real-time Support for Trade Studies
  Dealing With Alternative Product Configurations
  and Manufacturing Processes
• Rank Concepts With Consistent Set of Costing
  Assumptions
• Considers Material and Labor Costs
• Not an Exhaustive Model of the Total Cost
• Can Include Non-Recurring, Burdens, Support,
  Tooling, Capital, Systems, etc.

38
Metallic Processes

• Metallic Product Analysis Is Generated With
  Current Levels of Process Information
• Evaluate Differences in Configuration Concepts
• Metallic Component Assembly Is Mature and Well
  Understood

39
Composite Processes

• Vought Aircraft Has a Wide Range of Programs That
  Employ Composite Material Application
• Composite Product Analysis Is Generated With
  Current Levels of Process Information
• Vought Derived Equations are proprietary
• Models Voughts Current Composite Manufacturing
  Processes
• Data Concerning New Processes Developed From
  Synthesized From Simulations

40
Trade Study Approach Using DFMA

• Define/Plan Operations Required
• Assign Labor Hours to Each Operation Based
  Manufacturing Engineering Experts
• Determine Characteristics of High Speed Machinery
• Speed, Feed, Cutter Width
• Calculate Times to Machine Contact Surface Area
  for Manufacturing Equipment
• Derive Equations for Each Operation and Program
  Formulas in DFMA
• Select Each Operation in a Stepwise Manner for
  Each Alternative
• Analyze Trade Study Results From Application of
  DFMA

41
In Conclusion
ANALYSIS EXPERIENCE
EXPERT SYSTEM
PROCESS
ESTIMATE