This paper reports a system for automatic extraction of geometric and non-geometric part information from ‘engineering drawings’ created using computer-aided design and drafting (CADD) tools. A heuristic search is employed to interpret the characteristic attributes of dimension sets that denote linear, diametrical, radial and angular dimensions. Textual callouts are processed using natural language processing (NLP) techniques to interpret information related to part/feature function and related processes. The part information so recognized is represented using object oriented paradigm (OOP) suitable for linking to the downline CAD/CAM activities of the product cycle. The system, thus provides an effective alternative for design automation using CADD models.
The past decade has witnessed significant research on a variety of 3D part modeling techniques like constructive solid geometry (CSG), boundary representation (B-Rep), features technology, constraint-based modeling and of late the concept modeling. The primary objective has been to provide an informationally rich part representation scheme to support, if not all, a majority of business and manufacturing activities of the product cycle in an integrated manner [1] and [2].
Solid modeling schemes employing B-Rep and CSG form a core technology for part modeling as they provide a robust means of object construction, validation, manipulation and visualization. However, they lack explicit information in terms of functional features and non-geometric part attributes, which are very important in the down-line applications tasks [3]. Nevertheless, the current generations of CAD/CAM systems are based on these schemes.
To overcome these limitations, researchers developed feature supported representation schemes following two different strategies viz. feature extraction from solid models and feature-based modeling techniques. Several algorithms having a variety of capabilities have been proposed to extract and classify design and manufacturable features from B-Rep-based solid models. However, they lack ability to extract and represent non-geometric part attributes such as dimensions and tolerances, geometrical tolerances and specific information related to processes, tooling, etc. [4], [5], [6], [7] and [8].
An alternative strategy of feature-based modeling envisages a ‘design by features’ approach of model creation using functional feature libraries [9]. Here the features provide a means of representation of both geometric and non-geometric part data. Though this strategy is effective for specific part domains and related down-line activities like CAPP, DFM, DFA, etc. it lacks generality in terms of application [10] and [11].
On the other hand, industries typically use a large amount of part and assembly drawings for various engineering (design/manufacturing) and business activities [12]. To switch over to the CAD/CAM technology in the current scenario, industries will be required to model a large number of existing and new parts using solid/feature-based modelers. The effort involved in this activity will be labor intensive leading to larger lead times and costs.
In contrast, low cost computer-aided design and drafting (CADD) systems which provide 3D drafting (wireframe modeling) capabilities are quite popular in industries. These systems employ standard symbology for representation of diverse geometric as well as non-geometric part data relating to material, process and commercial functions. Though low cost and elegant, the CADD drawing models have a major limitation, in that they are suitable for human interpretation only.
A need therefore exists to develop algorithms for automatic interpretation of these CADD drawing models to extract and represent the relevant part/product information, for achieving the perceived design automation and CAD/CAM integration. This would establish an alternative route to the creation of informationally complete part model from CADD drawings.
This paper reports an intelligent interpreter for the extraction and representation of part information using a part model paradigm, from CADD models. Algorithms for extraction of part geometry and topology have been discussed at length in our earlier paper [13]. The present paper focuses on the recognition of non-geometric information for CADD models using NLP techniques.
AUTOFEAT has been extensively tested using standard engineering drawings stored in CADD model format of commercial packages. It has been found to be consistent in the recognition and classification of the geometry and non-geometric part/feature information, including the information on machining and/or functional requirement. The OOP-based design allows the system to be easily customized and extended by incorporating cogent classes with relevant data members and methods to support user defined features as well as the interface to down-line application tasks.
AUTOFEAT, thus demonstrates a comprehensive methodology of extracting part/feature information to provide an effective interface with down-line application tasks for design automation and integration.
