Elsevier

Computers in Industry

Volume 59, Issue 4, April 2008, Pages 321-337
Computers in Industry

Survey paper
A review of automated feature recognition with rule-based pattern recognition

https://doi.org/10.1016/j.compind.2007.09.001Get rights and content

Abstract

Automated feature recognition (AFR) has provided the greatest contribution to fully automated CAPP system development. The objective of this paper is to review various approaches for solving three major AFR problems: (i) extraction of geometric primitives from a CAD model; (ii) defining a suitable part representation for form feature identification; and (iii) feature pattern matching/recognition. A novel, detailed classification of developed AFR systems has been introduced. This paper also provides a thorough investigation of methods for geometric feature extraction, emphasizing STEP standard application and, finally, a review of recent research reports in the field of AFR with rule-based feature pattern recognition. We discuss potentials and limitations of these approaches and emphasize directions for further research work.

Introduction

Generative CAPP systems, in general, should provide two groups of functions: (i) efficient translation of part's geometrical information, defined by a CAD system (low-level entities – vertices, edges, etc.) in part's manufacturing information, necessary for process planning and CAM (high-level entities – holes, slots, pockets, etc.) and (ii) definition of feasible process plan (selection of manufacturing processes, selection of a work piece size, selection of auxiliary tools, set-up planning, selection, determination and grouping of elementary machining operations, selection of machining systems, operation sequencing and optimization of operation sequences, selection of cutting tools, determination of cutting parameters and conditions, tool path determination, selection of quality inspection methods, cost analysis, optimization of process plan elements and CNC code generation and verification).

Speaking of first group of functions, there are three strategies, all based on form feature concept, but differing in methods of form feature determination:

  • (1)

    design-by-feature – DBF;

  • (2)

    automated feature recognition – AFR; and

  • (3)

    interactive form feature definition.

Design-by-feature demands the existence of a form feature library, accommodated not to part function, but to part manufacturing needs. Product model is formed by using only form features from the library. Automated feature recognition comprises browsing some type of part representation aiming to find information, which characterizes singular form feature types. All approaches in this field have a unique goal: to form an algorithm capable of recognizing any possible type of form feature, without any interfering of manufacturing engineer. Interactive form feature definition is a system in which user selects a form feature set, determines recognition parameters to those features and then the system, using those instructions, performs an automated recognizing, whether directly in CAD model of the part, or in some structure developed from it.

In this paper, several AFR systems are discussed. All of them have to provide the solution for these interrelated tasks: (i) extraction of geometrical features from CAD model of the part, that are necessary for part representation suitable for form feature recognition. Boundary representation (B-rep) has most commonly been used; there is relatively rare usage of constructive solid geometry (CSG) part representation and, also, a few research approaches, based on wire frame part representation, have been reported. In B-rep based systems, these features comprise vertices, edges and faces of the part, in wire frame based systems vertices and edges and in CSG based systems geometrical primitives, such as sphere, cylinder and so on; (ii) formation of part representation suitable for form feature identification – in systems based on B-rep and wire frame representation topological linking of geometrical features is made, whereas in systems based on CSG the linking of geometrical primitives is made through Bull algebra operations. A certain number of developed AFR systems are using direct browsing of CAD model of the part for form feature extraction; (iii) matching of form features, recognized in part representation attained as a solution to previous task, with patterns in form feature library and, in case of advanced systems, based, for example, on artificial neural networks, knowledge acquisition—new patterns can be made of unrecognized form features.

In corresponding literature, following terminology has most commonly been used: feature extraction as the term for the first and the second task resolution, and feature recognition—as the term for the second and the third task resolution, and often for resolution of all three tasks. The authors have an opinion that the most suitable term for description of the first task resolution process would be geometric feature extraction, for the second task—form feature identification, and for the third task pattern recognition, that is proposed in [1], too. The authors use term pattern to describe notion of predefined manufacturing form in knowledge library, which form feature identified in part representation is matched with.

Studies analyzing developed AFR methodologies can be found in numerous literature sources, such as [2], [3], [4], [5], [6], [7]. In this paper, a novel detailed classification has been offered, which takes in account variety of approaches, particularly for each of three mentioned tasks. Table 1 illustrates this classification.

In the following text, a review of newer research results in the field of AFR with logic rule-based pattern recognition has been given. Methods based on B-rep part representation are specially emphasized, because CSG based approaches have not given effective results, from reasons thoroughly depicted by Subrahmanyam and Wozny [4].

Section snippets

Methods for extraction of geometric features

Methods for extraction of geometric features from CAD model of a part can be divided in external and internal ones [8]. Internal approaches comprehend use of API (application protocol interface) of the software by which the part was designed, in order to access topological and geometrical information relating to the part. In external approaches CAD model of the part is exported from software by which it was designed in a neutral data format (STEP, IGES, ACIS, etc.) ASCII file. That file is then

Methods for automated feature recognition with rule-based pattern recognition

The methods for automated feature recognition with rule-based pattern recognition apply a common basic principle: the structures identified in a part representation, formed using one of the methods given in Table 1, are matched with some pattern in the knowledge base, using if–then rules. It is essential that these rules provide uniqueness of form feature definition: there must not exist two forms with a unique definition or one form with more than one definition in the knowledge base. If set

Conclusion

AFR is the first and the most important step in the process of translation of CAD information into some instructions appropriate for manufacturing. To eliminate a need for human engagement in feature recognition process is essential for a fully automated CAPP system development. The advantages of AFR compared to feature based design are significant time and human resources saving, as well as insurance of desired part functionality without being limited in design creativity by the possibilities

Acknowledgement

The presented research work was financially supported by the Ministry of Science and Environmental Protection of Republic of Serbia in the framework of the 6319 project entitled “Implementation of automated design for manufacturing systems and manufacturing processes in the metalworking industry”.

Bojan R. Babic received his MSc and PhD degrees from Belgrade University, both in domain of discrete simulation and intelligent manufacturing systems. He is full professor at Faculty of Mechanical Engineering, University of Belgrade, Serbia for subjects in domain of CAPP. Currently he is vice dean of the faculty. His research interests include computer aided process planning and manufacturing, and intelligent manufacturing systems. He has authored more than 70 papers.

References (70)

  • H.S. Nagaraj et al.

    Automatic extraction of machining primitives with respect to preformed stock for process planning

    Journal of Manufacturing Systems

    (2001)
  • Y.S. Kim

    Recognition of form features using convex decomposition

    Computer-Aided Design

    (1992)
  • F. Parienté et al.

    Incremental and localized update of convex decomposition used for form feature recognition

    Computer-Aided Design

    (1996)
  • Y.S. Kim et al.

    Geometric reasoning for mill-turn machining process planning

    Computers and Industrial Engineering

    (1997)
  • H. Sakurai et al.

    Definition and recognition of volume features for process planning

  • H. Sakurai

    Volume decomposition and feature recognition. Part 1. Polyhedral objects

    Computer-Aided Design

    (1995)
  • H. Sakurai et al.

    Volume decomposition and feature recognition. Part 2. Curved objects

    Computer-Aided Design

    (1996)
  • Y. Woo et al.

    Recognition of maximal features by volume decomposition

    Computer-Aided Design

    (2002)
  • Y. Woo

    Fast cell-based decomposition and applications to solid modeling

    Computer-Aided Design

    (2003)
  • W.F. Bronsvoort et al.

    Multiple-view feature modeling for integral product development

    Computer-Aided Design

    (2004)
  • R. Bidarra et al.

    Representation and management of feature information in a cellular model

    Computer-Aided Design

    (1998)
  • R. Bidarra et al.

    Semantic feature modeling

    Computer-Aided Design

    (2000)
  • R. Bidarra et al.

    Efficiency of boundary evaluation for a cellular model

    Computer-Aided Design

    (2005)
  • W.C. Regli et al.

    Towards multiprocessor feature recognition

    Computer-Aided Design

    (1997)
  • J.H. Han et al.

    Integration of feature based design and feature recognition

    Computer-Aided Design

    (1997)
  • S. Gao et al.

    Automatic recognition of interacting machining features based on minimal condition subgraph

    Computer-Aided Design

    (1998)
  • G. Little et al.

    Delta-volume decomposition for multi-sided components

    Computer-Aided Design

    (1998)
  • R. Tuttle et al.

    Feature recognition for NC part programming

    Computers in Industry

    (1998)
  • X.G. Ye et al.

    A hybrid method for recognition of undercut features from moulded parts

    Computer-Aided Design

    (2001)
  • K. Rahmani et al.

    Boundary analysis and geometric completion for recognition of interacting machining features

    Computer-Aided Design

    (2006)
  • V. Sundararajan et al.

    Volumetric feature recognition for machining components with freeform surfaces

    Computer-Aided Design

    (2004)
  • Y.S. Kim et al.

    Recognition of machining features for cast then machined parts

    Computer-Aided Design

    (2002)
  • S. Subrahmanyam

    A method for generation of machining and fixturing features from design features

    Computers in Industry

    (2002)
  • M.C. Wu et al.

    Analysis on machined feature recognition techniques based on B-rep

    Computer-Aided Design

    (1996)
  • E. Wang et al.

    Form feature recognition using convex decomposition: results presented at the 1997 ASME CIE Feature Panel Session

    Computer-Aided Design

    (1998)
  • Cited by (0)

    Bojan R. Babic received his MSc and PhD degrees from Belgrade University, both in domain of discrete simulation and intelligent manufacturing systems. He is full professor at Faculty of Mechanical Engineering, University of Belgrade, Serbia for subjects in domain of CAPP. Currently he is vice dean of the faculty. His research interests include computer aided process planning and manufacturing, and intelligent manufacturing systems. He has authored more than 70 papers.

    Nenad Nesic received his Dipl.-Ing. degree in mechanical engineering from University of Belgrade (Serbia) in 1999. Since 2000, he has been a teaching and research assistant at the Production Engineering Department, Faculty of Mechanical Engineering, University of Belgrade. He has worked in the field of intelligent metrology and total quality management. His current areas of research interest include intelligent manufacturing systems and CAD/CAPP/CAM systems.

    Zoran Miljković received BS, MS and PhD degrees in mechanical engineering from University of Belgrade. He is an associate professor in Faculty of Mechanical Engineering in Belgrade, University of Belgrade. His research interests include intelligent manufacturing systems and processes, industrial robots and artificial neural networks. He has authored above 50 scientific papers, published totally.

    View full text