Paper Review - 1

@Article{Kim/O'Grady:1996,
   author      = { Chongsu Kim and Peter J O'Grady},
   year        = { 1996},
   keywords    = { DESIGN FEATURE BASED DESIGN CONCURRENT ENGINEERING},
   institution = { Samsung Data systems/University of Iowa},
   title       = { A representation formalism for feature based design},
   journal     = { Computer Aided-Design},
   volume      = { 28},
   number      = { 6/7},
   pages       = { 451-460},
   annote      = {

Feature based design lacks a formal methodology for system development and operation . One repercussion of this is that feature based design is , at present, implemented in a relatively ad hoc manner .This paper aims to address this deficiency by establishing a domain independent representation formalism for feature-based design.

This proposed representation formalism aims to make it possible to develop a feature based design system for a specific design domain in a structured way . In this representation formalism , a formal representation scheme is established so that the design representation can be based on an appropriate logic and semantic and syntactic consistency of design can be maintained through out the design process.

The feature algebra formally defines what should be represented . The feature based design algorithm describes how design can be performed. This design algorithm serves as a computational model of representation.

	         },
}



 
Paper Review - 2
 

@Article{Kim/O'Grady:1996,
@Article{Nezis/Vosniakos:1997,
  author       = { Nezis, Konstantinos and Vosniakos, George},
  year         = { 1997},
  keywords     = { Feature Recognition Neural Networks Connectedness Boundary
		   Models Heuristic Algorithms},
  institution  = { Elsevier Science Ltd.},
  title        = { Recognizing 2(1/2)-D Shape Features using a Neural Network
		   and Heuristics},
  journal      = { Computer Aided Designing :1997},
  volume       = { 29 },
  number       = { 7 },
  pages        = { 523-539 },
  annote       = {

	ABSTRACT:  The article presents a "Feature Recognition System" using a
	previously trained Neural Network. Feature Recognition is a function
	that takes information in some form of part representation and converts
	it into information about the features existing in the part. A feature
	is defined as a physical constituent of a part having predictable
	properties and is mappable to a generic shape, besides having
	engineering significance.


	A set of Heuristics is used for breaking down a compound feature, made
	up of faces in the part and represented in the form of "Attributed
	Adjacency Graph" (B-rep's Solid Modeller's Database), into sub-graphs
	that correspond to simple features. These simple features are used to
	construct "Special Representation Patters", which are the presented to
	a Neural Network which classifies them into "Feature Classes" : Pockets,
	Slots, Passages, Protrusions, Steps, Blind Slots, Corner Pockets and
	Holes. The scope of instances/variations of these features that can be
	recognized becomes very wide as the Neural Net is trained.


	COMMENTS:  The paper is well written and uses fairly simple language,
	making it an interesting read. The advantage of this approach is the
	high performance of the recognizer in terms of speed due to Neural
	Nets. The basic limitation is that of the Heuristics used to break down
	compound features into simple ones.
		},
}