Form Feature and Tolerance Transfer from a 3D Model to a Setup

Transcription

1 Int J Adv Manuf Techno! (2002) 2002 Springer-Verlag London Limited The Intemationa! Journal of Advanced manufacturing Jechnology Form Feature and Tolerance Transfer from a 3D Model to a Setup Planning System F. Zhoul, T.-C. KUO,2 S. H. Huang3 and H.-C. Zhang4 1 Reltec Corporate. Bedford, Texas, USA; 2Department of Industrial Engineering and Management, Shu- Teh Institute of Technology and Commerce, Taichung, Taiwan; 3Department of Mechanical, Industrial and Nuclear Engineering, University of Cincinnati, USA; and 4Department of Industrial Engineering, Texas Tech University, Lubbock, Texas, USA Currently, most CAD systems, even the feature-based design systems which were developed for CAPP, cannot provide exact information about an object (e.g. dimensions and tolerances). Some feature-based design systems can provide product data directly or indirectly; however, most CAPP systems still do not have an interface with CAD systems. The product data required by CAPP systems usually has a specific format which is unique to the system. In a CAPP system, it is essential for set-up planning to ensure the precision of the machining processes. Therefore, it is necessary to develop an interface with CAD models, so that the part data file can be obtained directly from the CAD representation. This paper proposes an approach for integrating the set-up planning system with a feature-based CAD system. By using an object-oriented approach -Product Data Translator (PDT), the computerautomated extraction of geometry and complete tolerance information was achieved; and the automated generation of the tool approach direction was developed. Keywords: CAD; CAPP; Dimensions; Set-up planning; Toler. ance 1. Introduction Process planning is the sequence of technical planning and decision-making steps related to the manufacturing of a product [1]. It not only includes the identification of the set-ups, fixtures, processes, machine tools, cutting tools, and operations necessary to produce the desired part [2], but also includes the geometric information about the product. To be more specific, process planning requires additional data on the shapes of thepart (e.g. their dimensions, locations, tolerances, and surface Correspondence and offprint requests to: H.-C. Zhang, Department of Industrial Engineering, Texas Tech University, Lubbock, TX , USA. hzhang@coe.ttu.edu finish condition). This infonnation should consider more specialised factors including tool accessibility, fixturing possibilities, and inspectability. Therefore, product data transfer is important and necessary in automated process planning. Currently, many CAD systems have been developed; however, most CAD systems, even the feature-based design systems which were developed for the requirements of CAPP, cannot provide exact infonnation about an object (e.g. dimensions and tolerances). Until now, most CAD models considered only the geometry of a part, rather than other product infonnation; however, product data, such as tolerance infonnation, is absolutely crucial to CAPPo To realise product data transfer, featurebased design can be used, although it has not yet fulfilled its promise. Some feature-based design CAD systems can provide product data directly or indirectly; however, most CAPP systems still do not have an interface with those CAD systems. The product data required by these CAPP systems usually has a specific fonnat, which is unique to the system. Among them is an automated set-up planning system developed by the Advanced Manufacturing Laboratory in the Department of Industrial Engineering of Texas Tech University. In a CAPP system, setup planning is essential to ensure the precision of the machining processes. An effective set-up planning program was developed by using a systematic approach [2]. The disadvantage of this program is that it lacks an interface with the CAD models. The part data file has to be manually edited and input, which is a very tedious and time-consuming task. Therefore, it is necessary to develop an interface with CAD models so that the part data file can be obtained directly from the CAD representation. This paper proposes an approach far integrating the set-up planning system with a feature-based CAD system. By using an object-oriented approach -Product Data Translator (PDT), the computer-automated extraction of geometry and complete tolerance infonnation was achieved; and the automated generation of tool approach direction was developed. Since the CAD export file used in this work does not include all the product infonnation clearly, it was modified to be a tree search method.

2 2. Form Feature and Tolerance Transfer 89 Background 2.1 Computer-Aided Process Planning (CAPP) and Set-up Planning Computer-aided process planning (CAPP) plays a key role in the integration of CAD/CAM functions, because it is the function responsible for converting design specifications into manufacturing instructions. Considerable effort has been devoted to modelling and automating process planning activities. More than 200 CAPP systems have been developed [3]; however, owing to the complexity of the problems involved, a truly generative process planning solution still does not exist [4]. Set-up planning and fixture design are two closely related tasks in process planning. To set up a workpiece requires the location of the workpiece in a desired position on the machine table. A fixture provides a clamping mechanism to maintain the workpiece in that position and to resist the effects of gravity and/or operational forces [5]. The purpose of set-up and fixturing is to ensure the stability and, more importantly, the precision of the machining processes. The important guidelines for set-up planning and fixture design are design tolerance requirements; however, tolerance analysis for set-up planning and fixture design are relatively unexplored issues. Some rules for set-up planning based on tolerance analysis have been proposed [6,7], but there is no systematic approach available. Recently, a graph-matrix approach to set-up planning, which is the first step towards building a scientifically rigorous base for CAPP research, was developed by Huang [2]. 2.2 Feature-Based CAD Systems Feature-based CAD systems, as a new development of CAD, emerged from a desire to integrate CAPP with CAD in the mid-1970s. They demonstrated a clear potential for creating attractive design environments [1]. Feature-based design links geometry information between design and manufacturing and is regarded as an important factor in CAD/CAPP integration. From a design point of view, feature-based design offers the possibilities for supporting the design process better than current CAD systems [8]. Three types of feature are defined by Shah and Rogers [9,10], i.e. form features, precision features, and material features. Form features are intended to achieve a given function or to modify the appearance of a part [11,12]. They described portions of the nominal (or idealised) geometry of a part [1], and can be attached to several features. Form features also have several different definitions. In this work, the definition of a form feature in ISO standard [13, p. 78] was adopted: A form feature is to be a shape aspect that conforms to some preconceived pattern or stereotype, while a shape aspect is defined as a share or a distinguishable portion of a shape. The shape aspects mentioned in the definition are specified as plane and cylindrical surfaces in this work. Precision features, including tolerances and surface finish, describe the additional geometric characteristics of the existing geometric model and the form features of the part. Material features deal with material composition, treatment, condition, etc. Form features and precision features are all closely related to the geometry of parts, and are called collectively geometric features [1]. Current feature-based CAD systems mainly address geometric features, in particular, form features. Geometric features can also be classified, according to the intended application, into groups such as design features, manufacturing features, and inspection features; etc. Feature recognition in a CAD model is a crucial task for a CAPP system. It is useful to recognise when plans can be adapted from previous parts, especially for reducing computation during replanning. Also, it is always worthwhile for the designer, when contemplating a new design, to consult previous designs with similar attributes. Secondly, features are sometimes created as a side effect of other operations. 2.3 Feature-Based Design in CAPP Process planning converts design infornlation into process documentation using a generative approach based on individual features [14]. A great deal of work has focused on CAPP; and all process planners, explicitly or implicitly, use the features of parts [15]. In the early process planning systems, group technology (GT) codes were used to describe designs in ternls of geometric and manufacturing features; however, such codes tended to emphasise either geometric or manufacturing features, with the result that they were good for design retrieval or for manufacturing planning but not for both [16]. Other drawbacks of conventional GT codes also exist (e.g. tedious data entry, difficulties in matching partial codes, ambiguity leading to finding unrelated parts when the code is short, or overspecialisation leading to finding nothing when the code is long) [15]. As a result, GT codes have been replaced in general process planners by feature descriptions of the part [14,17-20]. Thus, systems such as GAR! [17] or AIFlX [20] include features (e.g. "socket", and "profile") and relational statements such as "s-aside-of: () or "(opens-into &X &y)" were developed [14]. 3. The Graph-Matrix Approach to Set-up Planning Set-up planning is a complex problem. For a simple example part, as shown in Fig. 1, there are 2(i)(1) = 12 alternative setup plans (Table 1). Given the design tolerance requirements, some of the set-up plans are appropriate, and one of them should be selected to ensure the resultant part dimensions; therefore, the purpose of set-up planning is to select a set-up plan to facilitate tolerance control. To solve the set-up planning problem effectively, a graph-matrix approach was developed by Huang [2]. Set-up planning is divided into three steps: 1. Set-up formation. 2. Patum selection. 3. Set-up sequencing.

3 90 F. Zhou et unilmm Table 1. Set-up methods for the part shown in Fig. 1 (modified from [2]. Set-up plan Method A Machined features: A, B, C, D, and E Possible setup plans: 2(;)(:)=12 Fig. 1. The complex problem of set-up planning (redrawn from [2]). 3.1 Set-up Formation 6 The first step in set-up planning is to group features of a part into set-ups, namely, set-up formation. An important factor in determining set-ups is the tool approach direction of a feature, which is an unobstructed path that a tool can take to access the feature [4]. Features with the same tool approach direction can be grouped into one set-up. A feature may have more than one tool approach direction; therefore, it can be grouped into different set-ups. The feature might have tolerance relationships with another feature which has only one tool approach direction. If these two features share a common tool approach direction, then they should be grouped into the same set-up so that set-up method I is applied to assure the tolerance. This is the purpose of set-up formation [2]. The problem of set-up formation can be represented using the language of graphs [2]. The features within a part and their relationships (in terms of tool approach direction) can be represented as a feature relationship graph Go = (V, E). The feature relationship graph consists of a number of complete subgraphs. Each complete subgraph represents a set-up. The vertices within a complete subgraph represent the features to be machined in the set-up. The purpose of set-up formation is to find these subgraphs. After the set-up has been formed, s must be selected and the sequence of the set-ups must be decided. The concerns at this stage are those dimensions that cannot be obtained using set-up I. The primary purpose is to make sure that the dimension with the tightest tolerance is obtained using set-up II. The secondary purpose is to sequence the set-ups so that set-up II is applied to obtain as many dimensions with specified tolerances as possible. Datum selection and set-up sequencing are both based on the datum graph [2]. The set-up planning program has been implemented using C/C+ + under the Microsoft Windows environment. It can generate set-up plans for both rotational and prismatic parts based on tolerance analysis. To obtain a set-up plan for a part, a part data file must be read. The file format for rotational parts is as follows: n E\E2 A\A2 SlS2 t11t12 t21t22 En An 8n tin t2n feature A as a and then E as a Machine features A and C using feature E as a and then machine features E, G, and I using feature A as a feature A as a and then G as a Machine features A and C using feature G as a and then machine features E, G, and I using feature A as a Machine features A and C using feature I as a and then machine features E, G, and I using feature A as a feature C as a and then E as a Machine features A and C using feature E as a and then machine features E, G, and I using feature C as a feature C as a and then G as a Machine features A and C using feature G as a and then machine features E, G, and I using feature C as a feature C as a and then G as a Machine features A and C using feature I as a and then machine features E, G, and I using feature C as a (number of feature) (feature type of each feature) (tool approach direction of each feature) (existence of each feature or the stock) (tolerance matrix) SH-PART-R (header) tnltn: tnn

4 -E Datum Name Reference Datum Datum Type Form Feature and Tolerance Transfer 91 The results include the user interface, part drawing, and input and output files. The graph-matrix approach is a mathematical approach and is the first step towards building a scientifically rigorous base for CAPP research. The set-up planning algorithm based on it is both efficient and effective. However, the setup planning program still needs to be improved since the program does not have an interface with CAD models. The part data file has to be manually edited, which is very tedious, especially when the part is complex. It would be helpful to develop an interface with CAD models so that the part data file required by the set-up planning program can be obtained directly from CAD representation. 4. Feature Recognition and Product Files Transfer 4.1 CAD Export File Preparation of product data for CAPP is an important yet difficult task. A wide variety of methods and procedures have been proposed and many of them were unsuccessful because of the lack of consistency or generality [21]. As CAD systems gained popularity and acceptability, people began to pay attention to extracting product data directly or indirectly from a CAD database. In order to obtain the product data file required by the set-up planning program [2], Pro/ENGINEER was used to model 3D parts in this research. The product data information can be extracted from the neutral file, an export file of Pro/ENGINEER. A neutral file is a formatted text file containing information about parts and assemblies created using the Pro/ENGINEER system. The feature-based design of the part is retained by providing topology, geometry, symbolic parameters and their corresponding values [22]. The structure of a neutral file of a part is shown in Fig. 2. Neutral file uses a high-level, featurebased and object-oriented approach to specify a product which can represent boundary, conditions, tolerances, and so on. It includes all the product data needed by CAPP and is easier to read than other CAD files both for humans and computers, although the tolerance information is not exact. A neutral file includes the attributes of dimensional tolerances and geometric tolerances, but the related surfaces of the tolerance are not described. Therefore, some confusing cases will occur when the model is created. This is also a weakness of a neutral file which will be modified in this research. 4.2 Feature Recognition The automation of process planning requires that product data is extractable from the product model automatically. However, CAD product representations in product-models usually differ from the type of information required in CAPP [8]. Feature recognition has now been considered as the most common approach to extract product data for CAPP from CAD models. Feature recognition is typically thought of as a process that is performed on a geometric model of a finished part. For different applications, the methods of feature recognition may PARl Revision Number '-Volume Mass Properties '-Density -Center of Gravity "--Moments of Inertia Dimensions { Value Tolerance Name Plus Tolerance Minus Name of the Feature (Comment Line) F \1J H d ---rfeature Type Feature&--ea re ea e~id -[ DimensionS-Dim Ids LFeature Data AssocIated Entily'-Surface_,ds 1 =;v:~~;:s 'Id Trimmed Surfaces ~uvvalues,-xyz Values -Orient -Loops-Edge_ids -Surface Type Surface Structure L -[ un~vectors 0 f L I C rd at Syst I -Id j { ;0.. ","Ids of,.- Surfaces ~- Conntected.~ to the Edge Ed eo Direction g Boundary Points Curve Type Curve Structure Tolerance Class t- Tolerance Type f-feature Id -Reference Type Geometric Tolerances -Reference Id -Value '- Material Condition ngln 0 oca 00 In e em Datum Id Fig. 2. The structure of the neutral file. be different. For example, applications such as fixture planning do most of their reasoning on the basis of surfaces; also setup planning for turning and face milling generally requires surface features [I]. In this work, surface features are recognised from the part model of Pro/ENGINEER, since the extracted product data here should have the same format as the input data file in the set-up planning program. An objectoriented approach -Product Data Translator (PDT) is developed to extract the product data. Although it is convenient for humans to understand the neutral file, to realise the integration of CAD/CAPP, a program is still required to recognise features and extract the product data from the neutral file automatically. The PDT is implemented by using Borland C under the MS-DOS environment. PDT can read the neutral file for rotational parts and output the product data which is formatted as the one required in the set-up planning program. The tasks of PDT can be divided into two parts: 1. Form feature recognition from geometric data, which can obtain the tool approach directions of the models. 2. Precision feature recognition, which can extract the tolerance information and obtained a tolerance matrix. The structure of PDT is shown in Fig Form Feature Recognition by PDT Usually, the output data files of CAD systems include some information that is unavailable for the design process and downstream applications because of non-standardisation. There-

5 I 92 F. Zhou et at. Pro/ENGINEER Model i"-a (0) Class surfaces Attributes surface_ids fore, extracting the available information based on the application protocol is a goal of the integration structure for product data exchange. The geometry and topology of a part in a neutral file is represented clearly and completely. As shown in Table 2, the entities -features, surfaces, and edges -describe the geometry and topology of a part. In the set-up planning approach developed by Huang, only the surfaces are considered as an important fact. Therefore, form features are specified as surfaces in this work which includes planar surfaces and cylindrical surfaces. The geometric shape of a part can be recognised from the surface entity of a neutral file. Feature entity and edge entity are not important here, thus they are not analysed. Table 2 shows part of a neutral file describing two kinds Table 2. Part of a neutral file describing surfaces. Planar surface 2 surfaces 3 id 12 3 uv_rnin (2) 4 uv~n 2* UV-Inax (2) 4 uv_max 2*15. 3 xyz--rnin (3) 4 xyz--rnin 0.,2*-15 3 XYZ-Inax (3) 4 xyz-inax 0.,2*15. 3 orient-i 3 loops (1) 4 loops 5 edge~ds (2) 6 edge~ds 63,64 3 surface_type 34 3 surface (plane)-> 4 el (3) 5 el 0.,-1.,0. 4 e2 (3) 5 e2 2*0.,-1. 4 e3(3) 5 e3 1.,2*0. 4 origin (3) 5 origin 3*0. Cylindrical surface 2 surfaces 3 id 18 3 uv-min (2) 4 uv-inin 0., uv-max (2) 4 uv-max , xyz-inin (3) 4 xyz--lnin 15.,2*20. 3 xyz-max (3) 4 xyz-max ,20.,0. 3 orient-l 3 loops (1) 4 loops 5 edge-1ds (2) 6 edge-1ds 112, 115,65, surface_type 36 3 surface (cylinder)-> 4 el (3) 5 el 0.,-1.,0. 4 e2 (3) 5 e2 2*0.,-1. 4 e3 (3) 5 e3 1.,2*0. 4 origin (3) 5 origin 3*0. of surface: plane and cylindrical surfaces. All the geometric infonnation can be extracted from this part of a neutral file. In order to recognise a fonn feature, an object-oriented class -surfaces class is developed. The surfaces class provides all the required geometric and topological infonnation which is extracted from the neutral file. Another class, named two-d surfaces, is also constructed because rotational parts are axial symmetric and they can be simplified to 2D variables. The descriptions of these two classes are shown in Fig. 4. In a neutral file, the shape and size of a surface and its positional relationship with other surfaces are decided by its absolute Cartesian coordinate. The surfaces are all the trimmed surfaces which are called faces in topology. Therefore, by considering the values of minimum and maximum x, y, and z and the surface type, the geometric and topological infonnation can be obtained. The flowchart of the fonn feature recognition process is given in Fig. 5. When a neutral file is read, surfaces ills, and minimum and maximum values of X, Y, and Z, and surface types are found by geometric recognition. Then, these data are converted and filtered by coordinate transfer, cylindrical feature recognition, and 3D to 2D transferring to obtain the 2D data of a part. Using these filtered data, a 2D drawing can be made by 2D graphics, the external and internal surfaces of the part can be decided by internal and external feature recognition. Finally, from the recognised shape of the part, the tool approach direction can be detennined by the tool approach determination Precision Feature Recognition by PDT Precision features include tolerances and surface finish, etc., which describe additional geometric characteristics of the form features of a part. In this work, only tolerance is recognised

6 Form Feature and Tolerance Transfer 93 Table 3. Part of a neutral file describing dimensions and geometric tol. erances. - Dimensional tolerance 1 Dimensions (7) 2 Dimensions 3 name d2 3 value tolplus tolminus surface~ds [2] 4 surface~ds 18,20 2 dimensions -- Geometric tolerance I geometric_tolerance (2) 2 geometric_tolerance 3 tolclass Runout 3 toltype Circular Runout 3 fealid 9 3 ref_type Surface 3 ref~d 20 3 value surface~ds (4) 4 surface~ds 18,20,135, because only tolerance infonnation is required for the set-up planning program. Tolerance can be divided into two fonds: dimensional tolerance and geometric tolerance. In the neutral file, they are described in different parts. Hence, in the precision feature recognition, they are considered separately. Table 3 gives the parts of dimensional tolerance and geometric tolerance in a neutral file. In Table 3, it is obvious that the neutral file has independent parts to describe dimensional tolerance and geometric tolerance and both of them are not integral. For example, a dimension and tolerance of a part is given in a neutral file, but the surfaces related to this dimension are not known. Users cannot recognise the tolerance relationship according to this file. In Fig. 6, parts with the same geometry may have the same neutral file, but the dimensional notations can be different. Tolerance is important in precision manufacturing. The different dimensional notation means different precision requirements of a part. The machining steps are also different according to the different precision requirements. In Fig. 6(a), the larger cylinder must ensure the required precision when being machined, whereas in Fig. 6(b), the smaller cylinder requires a higher precision. Therefore, the product infonnation is not clear if there is no dimension notation in a neutral file. For geometric tolerance, the same problem exists. In Fig. 6, the reference datum of the geometric tolerance is surface A. Actually, surface A is the cylindrical surface which has a smaller radius, but users cannot read this infonnation from a neutral file. In order to obtain complete tolerance infonnation, the neutral file must be modified first. The modification of the neutral file is divided into two parts: 1. Modifying dimensional tolerance information. 2. Modifying geometric tolerance information. Manufacturing of parts with identical dimensions is known from experience to be impossible [2]. Thus, in industry, exact dimensions are not usually specified, but always continuous regions for dimensions. The continuous region is called a tolerance zone. Therefore, for a manufactured part, every dimension may have a tolerance which is related to one or two surfaces. Usually, all the axial dimensions of a rotational part are related to two surfaces -a dimensioned surface and a reference surface; all the dimensions of diameter are related only to one surface -a cylindrical surface. Since in Pro/ENGINEER, a whole cylindrical surface is represented by two half cylindrical surfaces, all the diameter dimensions are related to two half cylindrical surfaces in the neutral file. The dimensions of a part always consist of an open loop, because a closed dimensions chain is not allowed in manufacturing. That means, for rotational parts, if there are n planar surfaces, the axial dimensions must have n -1 surfaces. In Fig. 7, the nodes in the related surfaces tree (RST) represent all the planar surfaces in the rotational part, and the lines connecting them represent dimensions. In the example drawing in Fig. 7 the surfaces which a dimension is related to, can be obtained. For example, surfaces A and G are a dimensioned surface and a reference surface of dimension 50, respectively. The related surfaces tree can also be used to represent n planar surfaces. According to the RST, the dimension entity in a neutral file can be modified in this way. For every axial dimension, insert the two surfaces -a line with two ending nodes in the tree into the neutral file. For a diameter dimension, since it is not decided which of the two symmetrical surfaces would be filtered, both of them should be added to define the diameter dimension. The bold lines are added to the original lines. Geometric tolerance is divided into five classes and in each class there are several tolerance types [23]. The related surface numbers of these tolerance types are different, and some of them are not related to any other surfaces. The modification of the geometric tolerance in a neutral file has to refer to the tolerance types. For all the geometric tolerance types in the form class, only the surface itself which has a geometric tolerance, should be added into the neutral file. For other geometric tolerance types, not only the surface itself, but also the reference surfaces should be added into the neutral file. The modified neutral file is shown in Table 3. (a) (b)

7 94 F. 7lIou et at. After modifying the tolerance entities of a neutral file, tolerance values and the related surfaces can be recognised by precision feature recognition. Dimensional tolerance recognition includes: 1. Extracting tolerance values and related surfaces from a neutral file. 2. Examining the cylindrical surfaces to skip those which are filtered in form feature recognition. Since in the set-up planning program, tolerance was input as a tolerance matrix, the recognised tolerance results should be expressed in a matrix. Geometric tolerance recognition also needs to extract the required data and examine those related surfaces. The difference is that in dimensional tolerance, only two related surfaces exist, but in geometric tolerance, the related surfaces may be one, two, or more, depending on the geometric tolerance types and the numbers of the reference surfaces. Therefore, the algorithm of geometric tolerance recognition is considered as a similar process to the algorithm of dimensional tolerance recognition. 5. Illustration Example An example, which can be recognised by PDT, is given in this section. PDT is a tool to extract product data of rotational parts from the neutral files of Pro/ENGINEER. Therefore, the example parts should be created in Pro/ENGINEER first, and then the neutral files of these parts are exported. Mter being modified, these neutral files can be read by PDT to recognise both form features and precision features. Product data are saved as a text file. This file is sent to the set-up planning program to realise the integration of CAD/CAPP. Suppose all the tolerances of the diametric dimensions of three parts are equal to This part has a through hole. Two cylindrical surfaces have a geometric tolerance reference from datum F. The drawing of the part and the tool approach directions are shown in Fig. 8. The modified neutral file is read by PDT and the necessary product data are extracted. The final output file includes: header, number of features, feature type, tool approach direction of each feature, existence of each feature on the stock, and tolerance matrix. This part output file is given in Fig O.OIIFI 0 B Fig. 9. The output file of example After the product data of the models is extracted by PDT, it should be sent to the set-up planning program to realise the integration of CAD/CAPP. The example is checked to see whether the output file (Fig. 9) can be accepted by the set-up planning program and produce a set-up plan. The set-up planning program interface is shown in Fig Conclusions Set-up planning is the function of preparing detailed work instructions for setting up a part. It is the first and the most important step in process planning. However, it is also one of the least studied areas in automated process planning. Recently, workers are beginning to consider the issue of automated setup planning. Huang [2] developed a systematic approach -a graph -matrix approach to generate practical set-up plans based or tolerance analysis. Based on this approach, an efficient and effective set-up planning algorithm was developed. However, to realise the real automation of process planning, the set-up planning program still lacked an interface with CAD models. The work described in this paper was to establish an interface with CAD models in order to obtain the product data required by the set-up planning program [2]. Recently, feature-based modelling has become a wellaccepted approach to geometric modelling in integrated CAD/CAPP systems. Features are considered as an effective representational means for providing a more abstract product model than geometry alone. Feature recognition also becomes an important way for providing, without any human intervention, the product data input needed by applications such as process planning, NC part programming, etc. Therefore, a feature-based CAD system -Pro/ENGINEER is chosen to create product models in this work. It was found that neutral files are suitable for extracting product data, but neutral files loose some important information, so they must be modified first. Therefore, another objective in this work is to modify a neutral file in order to obtain an exact product data file. To extract the product data from the exported file of Pro/ENGINEER, an object-oriented approach -Product Data Translator (PDT) was developed. The program was developed under the MS-DOS environment using C+ +. With this interface, the part data file in the set-up planning program can be directly transferred from the CAD model without human intervention. An integration of a CAD/CAPP system was thus truly realised.

8 Form Feature and Tolerance Transfer 95 ~ Fig. 10. Part drawing, data file, and set-up plan fot the example part shown in Fig. 8. The tool approach direction is an essential part of the product data file of the set-up planning program. In this research, a systematic method is developed, using computers instead of the human experience to check the tool approach direction. By using form feature recognition in PDT, the tool approach direction of all the surfaces of the models can be determined automatically. Most CAD systems cannot usually provide exact product information. Although the neutral file has several advantages over some other CAD product data files, it is not perfect. The product information in the neutral file is insufficient if the tolerance information needs to be extracted. Thus, modification should be added to the neutral file before extracting tolerance information. The modification of a neutral file is significant because with this modification, a neutral file can be considered as an exact file and all the product data can be extracted from a modified neutral file. References 1. J. Shah and M. Mantyla, Parametric and Feature-Based CAD/CAM, John Wiley, S. H. Huang, "Graph-Matrix approach to setup planning in computer-aided process planning (CAPP)", PhD dissertation, Texas Tech University, H.-C. Zhang and L. AIting, Computerized Manufacturing Process Planning Systems, Chapman & Hall, T.-C. Chang, Editorial, Journal of Design and Manufacturing, 2(4), pp. i-ii, J. I. Karash, "Principles of locating and positioning", in Handbook of Fixture Design, McGraw-Hill, pp , X. Huang and P. Gu, "Tolerance analysis in setup and fixture planning for precision machining", Proceedings of the Fourth International Conference on Computer Integrated Manufacturing and Automation Technology, Rensselaer Polytechnic Institute, Troy, New York, pp , October, J. Mei and H.-C. Zhang, "Tolerance analysis for automated setup selection in CAPP", Concurrent Engineering, PED-59, ASME, pp , O. W. Salomons, F. J. A. M. van Houten and H. J. J. Kals, "Review of research in feature-based design", Journal of Manufacturing Systems, 12(2), J. J. Shah and M. T. Rogers, "Functional requirements and conceptual design of the feature-based modeling system", Computer Aided Engineering Journal, 5(1), pp. 9-15, J. J. Shah and M. T. Rogers, "Expert form feature modeling shell", Computer-Aided Design, 20(9), pp , II. N. Wang and T. M. Ozsoy, "The presentation of assemblies for automatic tolerance chain generation", Engineering with Computers, 6, pp , N. Wang and T. M. Ozsoy, "Scheme to represent features, dimension and tolerances in geometric modeling", Journal of Manufacturing Systems, 10(3), pp , M. Dune (ed.), Industrial Automation Systems and Integration - Product Data Representation and Exchange -Part 48: Integrated Generic Resources: Form Features, 2nd edn, ISO/WD , T.-C. Chang and R. A. Wysk, An Introduction to Automated Process Planning Systems, Prentice-Hall, M. R. Cutkosky, J. M. Tenenbaum and D. Muller, "Features in process based design", in Proceedings of the International Computers in Engineering Conference, San Francisco, CA, American Society of Mechanical Engineers, 1988.

9 96 F. Zhou et al. 16. R. L. Eckert, "Codes and classification systems", in Group Technology at Work, San Francisco, CA, Society of Manufacturing Engineers, pp , Y. Descotte and J.-C. Latombe, "Taking compromises among antagonist constraints in a planner", Artificial Intelligence, 27, pp , H. R. Berenji and B. Khoshnevis, "Artificial intelligence in automated process planning", Computer in Mechanical Engineering, pp , September B. S. Lim and J. A. C. Knight, "Foundation for a knowledge-based computer integrated manufacturing system", International Journal for Artificial Intelligence in Engineering, 2(1), pp , January P. N. Ferreira, B. Kochar, C. R. Liu and V. Chandru, "IFIX: an expert system approach to fixture design", Second International Conference on Design Theory and Methodology DTM '90, Chicago, IL, L.-H. Qiao, C. Zhang, T.-H. Liu, H.-P. B. Wang and G. W. Fischer, "PDES/STEP-based product data preparation procedure for computer-aided process planning", Computers in Industry, 21, pp , PTC1, Pro/ENGINEER Interface Guide, Waltham, MA, Parametric Technology, PTC2, Pro/ENGINEER Drawing User Guide, Waltham, MA, Parametric Technology, 1994.