In Fig. 19, a screen of assembly procedure for intake valve is given. When the assembly of the intake valves is completed, the program asks to the user whether or not there is any part to be assembled. If there is a part to be assembled, the program turns to the first process in the methodology. If there is no part to be assembled, the program shows the assembly model of the parts assembled. This method is implemented to all parts of the diesel automobile engine.

Part recognition based computer aided assembly system developed in this study is also applied to sub-assemblies of diesel engine. For example, exploded view of the crankshaft and connecting rod mechanism assembled by the program is shown in Fig. 20. Assembled view of it is shown in Fig. 21.

7. Implementation

In the development of the part recognition based computer aided assembly system, solid modeler of AutoCAD is used as the geometric modeler. Standard machine parts are obtained from standard part library of Mechanical Desktop a set of program for part recognition and assembly algorithms was completely written in Visual BASIC. An expert system for part recognition approach is developed by the authors for this study. All the parts in the system are recognized by the inference engine of the expert system through forward chaining reasoning method. Windows based a text editor (NotePAD) has been used as knowledge base of the expert system. For recognizing of all the parts, total 184 rules were written into the knowledge base. The approach is implemented and tested to parts of a diesel automobile engine and standard machine parts commonly used in industry namely, bolts, nuts, gears, washers, pins, rivets, springs, etc. To demonstrate the effectiveness of the methodol-ogy, the most complex parts in the diesel automobile engine have been chosen. The results are quite successful. For example, cylinder block which has the most surfaces has been handled within the system (Fig. 22). After performing the STEP translation for the cylinder block, a STEP file with 20,287 rows generated and saved. When the faces of the cylinder block are represented and numbered in basic format, 662 faces are determined on the cylinder block by the program. The program decreases the number of faces into 479 faces by stitching of the

183 equivalent faces and constructs FORM belonging to the cylinder block as 479 479 square matrix. Cylinder block is recognized at 56 min 16 s by reasoning of FORM generated in the system and rule represented in the knowledge base of it due to computational complexity.

For assembly operations, an assembly file within AutoCAD is generated. The part recognition and assembly procedures have been implemented to a diesel automobile engine having 370 parts (Fig. 23). However, the system developed can also be applied to other mechanical systems. The implementation results show that the approach is very effective in part recognition. Especially, rule writing process has been facilitated with the optional rule writing module and in this way all the parts in manufacturing area are recognized by the recognition procedure. The components recognized in this study have regular and/or free form surfaces.

8. Discussion and concluding remarks

In this paper, a new method is proposed to recognize the parts using their STEP AP 203 files and to assemble the recognized parts in a CAD platform. Major contributions of the study presented in this paper can be viewed from three aspects: FORM, part recognition, computer aided assembly. A new representation scheme named FORM is introduced to represent a 3D CAD model in terms of topological and geometrical data. FORM is applicable both simple and complex parts and it can serve as communication tool between CAD systems and different manufacturing activities. FORM has been used as input for part recognition algorithm along with the rule of the part. This part description scheme can be improved with attributes such as angles, concavity and convexity between faces. When this scheme is improved, it can be used as input for CAD/CAM applications such as especially feature recognition. Since FORM is a square matrix corresponding to computer format, it can be decomposed easier than graph-based representation. Because part recognition algorithm is based on an expert system embedded within the developed system,