Feature Recognition

Feature Recognition

Xun Xu
DOI: 10.4018/978-1-59904-714-0.ch005
OnDemand:
(Individual Chapters)
Available
$33.75
List Price: $37.50
10% Discount:-$3.75
TOTAL SAVINGS: $3.75

Abstract

Conventional CAD models only provide pure geometry and topology for mechanical designs such as vertices, edges, faces, simple primitives, and the relationship among them. Feature recognition is then required to interpret this low-level part information into high-level and domain-specific features such as machining features. Over the years, CAD has been undergoing fundamental changes toward the direction of feature-based design or design by features. Commercial implementations of FBD technique became available in the late 1980’s. One of the main benefits of adopting feature- based approach is the fact that features can convey and encapsulate designers’ intents in a natural way. In other words, the initial design can be synthesized quickly from the high-level entities and their relations, which a conventional CAD modeller is incapable of doing. However, such a feature-based design system, though capable of generating feature models as its end result, lacks the necessary link to a CAPP system, simply because the design features do not always carry the manufacturing information which is essential for process planning activities. This type of domain-dependent nature has been elaborated on in the previous chapter. In essence, feature recognition has become the first task of a CAPP system. It serves as an automatic and intelligent interpreter to link CAD with CAM, regardless of the CAD output being a pure geometric model or a feature model from a FBD system. To be specific, the goal of feature recognition systems is to bridge the gap between a CAD database and a CAPP system by automatically recognizing features of a part from the data stored in the CAD system, and based on the recognized features, to drive the CAPP system which produces process plans for manufacturing the part. Human interpretation of translating CAD data into technological information required by a CAPP system is thus minimized if not eliminated.
Chapter Preview
Top

Basic Concepts Of Feature Recognition

The input of an automatic feature recognition system is often a conventional description of the geometry of a part, i.e. a geometric model. The output is a list of recognised instances of features with necessary technological parameters and their proper organisation. A distinction may be made between several phases: feature identification, parameter determination, feature extraction, and feature organisation (Figure 1) (Xu, 2001). In the identification phase, features are identified from/within a geometric model. In the parameter determination phase, the parameters of the features, i.e. geometrical parameters such as the diameter and depth of a hole, and technical parameters such as tool approaching directions, are computed from the geometric model. In the extraction phase, the features are removed from the geometric model to form a stand-alone feature entity. And finally, in the organisation phase, the features are named and arranged in a particular structure, e.g. a graph tree. The concept of feature recognition is sometimes narrowed to only mean feature identification -- the first phase mentioned above. The reason for this is that feature identification is easily the most difficult task among the four.

Figure 1.

Feature recognition process

978-1-59904-714-0.ch005.f01
Top

Classification Of Feature Recognition Systems

Many attempts have been made to classify feature recognition systems. One of the most common ways is to classify them based upon the actual techniques used by the systems. Some of these techniques are known as, syntactic pattern recognition approach, geometric decomposition approach, expert system rule/logic approach and graph-based approach.

In the manufacturing domain, feature recognition approaches can also be classified based on the machining natures of the features (Singh & Qi, 1992), for example, approaches for recognising rotational (turning) features and approaches for recognising non-rotational (milling) features.

Shah and Mäntylä (1995) classified feature recognition techniques into four groups, those recognising features (a) from boundary models, (b) by volume decomposition, (c) from CSG models and (d) from 2D drawings. In a nutshell, some systems recognise features by detecting them from the geometric model with help of a completely pre-defined feature library. The capability of such a feature recogniser is therefore limited to the robustness of the feature database that has been pre-defined. Other systems recognise features by generating them using various techniques. Rules are normally built into systems to assist feature identification, but they are by no means complete feature libraries as mentioned above. For this reason, the author distinguishes feature recognition techniques between feature detection as in the former case and feature generation as in the later case, and discussions below are organised accordingly.

Complete Chapter List

Search this Book:
Reset