Toward Hybrid VariantGenerative Process Planning

1
Toward Hybrid Variant/Generative Process Planning

• i-Design Lab.
• ? ? ?

2
Contents

• Introduction
• Backgrounds
• Our Approach
• Summary and Conclusion

3
Introduction
Variant Process Planning
Hybrid Process Planning
Generative Process Planning
Variant Process Planning
Generative Process Planning

• Tools of choice
• Commercially available
• Some drawbacks

• Realistic Process Plan
• Only in restricted Domain
• Not achieved industrial use

4

• Backgrounds

5
Group Technology (GT)

• Classifying similar products into groups
• To capture critical design and manufacturing
  attributes of a part
• GT Code
• Scheme for describing products
• Consists of two types of positions
• Global property of design material, size, etc.
• Relevant only for certain types of designs
• Human-oriented
• ? Difficulty in automating the generation of
  GT codes

6
Geometric Approaches

• To use geometric properties of solid and CAD
  models
• CSG Constructive Solid Geometry
• Not unique and Not robust method
• Not correspond to the manufacturing operations
• Sun et al. properties of boundary
  representations
• Polyhedral approximation of a solid using a graph
• Relation between solid faces
• Difficulty
• Only with polyhedral objects
• Not symmetrical ( A B ? B A )
• Not incorporate manufacturing considerations

7
Variant Process Planning

• Based on the use of the Group Technology coding
  schemes
• When given a new design D
• Determines a GT code for D
• Retrieve a process plan P for a design D
• Modify retrieved process plan P for a design D
  manually
• If the part mix varies over time
• ? Some drawbacks

8
Generative Process Planning

• When given a new design D
• Synthesize a process plan directly for D
• Retrieved candidate processes
• Select feasible processes
• Difficulties
• Only in restricted domains
• Not really achieved significant industrial use
• If Generative Process Planning gives feedback to
  designer
• ? Improve its manufacturability

9
Hybrid Approaches

• No comprehensive solutions

10
Hybrid Approaches

• An approach for acquiring knowledge
• Use inference rules to find explanations behind
  the plan
• Very Simple ? No other manufacturing information
• Capture plan knowledge that specifies necessary
  process
• Search the old subplans ? Modify using process
  capability rules
• Each feature is independent ? No group features
• Robust Hybrid Approach
• Consider feature interaction
• Consider store, classify, retrieve

11

• Our Approach

12
Our Approach

• Preprocessing
• Given a database of existing designs
• Build an indexing and classification structure
  for search and retrieval
• Design signatures, Design similarity, Finding
  similar designs, Computational issues, Validation
• Planning for new designs
• Use the classification structure
• Synthesize a plan
• Slicing Designs, Retrieving plans and extracting
  plan slices, Combining plan fragments

13
Design Signatures

• Instead of GT Code, more detailed structure
• Basic signature
• Represents all properties of the design
• If basic signatures are isomorphic, two designs
  are identical
• Higher level of abstraction
• Simplify ? Remove some of the less important
  details

14
Design Signatures
F-Rex Feature Extractor
15
Design Signatures
lt Basic Signature gt
16
Design Similarity
lt Basic Signature gt
lt Simplified Basic Signature gt
17
Design Similarity
lt Simplified Basic Signature gt

• More similar ? If simplified, Isomorphic
• How much simplification is needed ??

18
Design Similarity

• D a set of designs
• P a set of all design properties
• Sn(d) basic signature Pn1, ... , Pn
• Rn(d,d) Relation Sn(d) Sn(d) ?
  Rn(d,d)
• Single equivalence class

• Classification Tree - hierarchically

19
Finding Similar Designs

• v vertex of the classification tree
• d member of equivalence class represented by v
• ? v matches d

lt Search Algorithm gt
20
Preprocessing

• Computational Issues
• To determine whether a vertex v of depth k
• Algorithm compares the design signature
• NP (Nondeterministic Polynomial time problem)
• ? Exponential time
• How to speed up ??
• Use special properties
• Unless d matches vs parent ? Not Check whether d
  matches v
• Validation
• Obstacle
• Lack of a generally available data set of CAD
  model
• Lack of an agreed-upon standard from
  manufacturing point of view

21
Planning for New Designs

• Traditional approach
• Retrieve a process plan from existing design
• ? Other portions may not match so well
• Our approach

22
Slicing Designs

• Manufacturing dependencies
• If two features intersect ? precedence constraint
• If feature has tolerance that is defined relative
  to the other
• If there is a thin section between them
• If the same tool is used
• If both have the same approach direction

• If both have the same corner radii

23
Planning for New Design

• Retrieving plans and extracting plan slices
• Design d ? d, Slice s ? s ? Plan ? p
• Slice s ? s ? Plan slice ? q
• ? Operation m
• Combining Plan Fragments

X

• Tool, approach direction ? Same slice
  Future Works

24

• Summary and Conclusion

25
Summary and Conclusion

• Anticipated Benefits
• Accelerating the product development process
• Utilize the strengths of both process planning in
  ad innovative way
• Provide the designer feedback

26

• Thanks!!!
• Q A