Abstract In this paper, a new approach has been developed to recognize the CAD models through their face adjacency relations and attributes and to automatically assemble the recognized parts in a CAD environment. Adjacent faces and the face attributes belonging to each face of parts designed in a CAD platform are derived by using of standard for the exchange of product data (STEP) file. They are represented in a square matrix format named face oriented relation matrix (FORM). An expert system was developed and embedded within the system. A knowledge base of the expert system is generated using a text editor. Reasoning the face adjacency relations and the face attributes in the knowledge base and FORM, the parts are recognized. Then, a reference face belonging to recognized parts is determined and they are positioned and assembled in assembly file through their reference faces. Part recognition approach developed in this study is applied to a computer aided assembly system. But it may be useful and practical for different CAD/CAM applications such as process planning and group technology too. The algorithm has been applied to a diesel automobile engine which has complex parts to demonstrate its efficiency and capability.63208

Keywords: Step; Form; Part recognition; Computer aided assembly; Expert system

 1. Introduction

An important issue in CAD/CAM integration is the automatic extraction of manufacturing and assembly informa-tion from 3D CAD models. Many of current CAD systems contain all geometric information of the part. However, they define mechanical components by low-level information such as points, curves, surfaces, and primitive solids [1,2]. But, in CAD/CAM form features such as slot, hole and pocket, etc., and technological information, namely tolerance, surface roughness, hardness, material, etc., are required. This means that CAD databases are not suitable for use directly in CAM systems, thus, a common database, which can be used by CAD and CAM, is necessary [3–5]. In order to interface CAD and CAM, one approach may be part recognition system which analyses a CAD model and identifies it. The parts analyzed in the part recognition algorithm can be then used as input to CAD/ CAM applications such as process planning, group technology (GT) and computer aided assembly. Part recognition can play an

important role in grouping of the parts and extracting of process plans of them. In the last decade, a lot of research has been done in the scope of extracting of manufacturing information from CAD models. The majority of these research activities have concentrated on feature recognition [6–11]. Therefore, part recognition based CAD/CAM applications may be a new work field.

Kruth et al. [12] described a method for defining and extracting user definable manufacturing feature information using wire frame and feature based CAD systems. The approach allows to extract geometric as well as technological feature information by analyzing related dimensions, form and location tolerances. Prabhu and Pande [13], and Prabhu et al. [14] developed a system for automatic extraction of geometric and non-geometric part information from engineering draw-ings. 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. Intelligent drawing interpreter called AUTOFEAT extracts both geometric and non-geometric part data. Part information recognized is represented using

 object-oriented paradigm (OOP) suitable for linking to the down line CAD/CAM activities of the product cycle. Joshi and Chang [15] presented the development of a concept called attributed adjacency graph (AAG) for the recognition of machined features from a 3D boundary representation of a solid. The features in the part are sub graphs of the complete AAG and recognition of the features involves in identification of the sub graphs that correspond to the features. A heuristic method is proposed to identify components of the graph that could form a feature. Each graph component is analyzed to determine the feature type. Qamhiyah et al. [16] presented a boundary-based procedure for the sequential extraction of form features from CAD models of objects with planar faces. Form features are first classified based on their effect in changing a basic shape. Geometric reasoning is then used to obtain generalized properties of the form features classes. Finally, form features classes are sequentially extracted based on the obtained properties. Lockett and Guenov [17] presented a novel CAD feature recognition approach for thin-walled injection molded and cast parts in which molding features are recognized from a mid-surface abstraction of the part geometry. The main contribution of the research has been the development of an attributed mid-surface adjacency graph to represent the mid-surface topology and geometry, and a feature recognition methodology for molding features. The conclusion of the research is that the mid-surface representa-tion provides a better basis for feature recognition for molded parts than a B-REP solid model. Mehalawi and Miller [18,19] proposed a representation scheme of the CAD model in a database. Components are represented using attributed graphs in which the nodes correspond to the faces of the component and the links correspond to the edges of the component. The graph is based on the STEP physical file of the component. The graph and its attributes describe the topology of the component completely together with some geometric data that are not dependent on any coordinate system such as face and curve types. Lee and Kim [20,21] proposed a new approach for extracting machining features from a feature based design model, based on an integrated geometric modeling system that supports both feature based modeling and feature recognition. Feature recognition is achieved through an incremental feature converter. The incremental feature converter not only keeps the design model consistent, but also incrementally extracts machining features from design features as a design evolves. Kao et al. [22] developed super relation graph method for extracting prismatic features from the CAD boundary representation of a machined part. Using the definition super relations and validity of a feature volume, this method recognizes features with all three types of interactions: face splitting, face merging, and edge truncation. Woo and Sakurai [23–25], and Sakurai and Dave [26] presented a volume decomposition method to decompose a solid model fast into maximal volumes. The maximal volume decomposition has been considered as an effective method for solid modeling such as recognition of intersecting machining features. The maximal volume decomposition comprises of three steps: cellular decomposition, cell collection and volume generation.