INTEGRATING DESIGN AND PRODUCTION OF PRECISION TRADITIONAL CERAMICS T. Giannakakis1, G.-C.Vosniakos1, D.

Pantelis2
1. School of Mechanical Engineering, Manufacturing Technology Division 2. School of Naval Architecture and Marine Engineering, Shipbuilding Technology Laboratory National Technical University of Athens, 9 Heroon Polytechniou Ave.,GR-15773, Zografos, GREECE

ABSTRACT This paper advocates using state-of-the-art engineering tools in design and production of ceramic parts and the corresponding pressing and casting dies. A feature library is developed for the parametric design of parts with simple geometry, focusing on concave pottery parts. A user-friendly interface is developed, so that parametric design can be achieved by technicians with no proficiency in CAD In addition, special routines were developed for enhancing commercially available functionality in conversion of point clouds, taken from a laser scanner, into solid models. Last, a fuzzy system is developed for the selection of suitable machining operations, machining strategy, cutting tools and parameters, respecting geometry complexity, total material removal volume and a coupled time-accuracy criterion depending on the user. KEYWORDS: Design with features, CAD-CAM, Fuzzy system, Laser scanning

1.

INTRODUCTION

In the area of design and production of traditional ceramic parts, new technologies are barely used, since most of the work is still performed mostly without use of CAD-CAM tools. Artistic methodologies are usually applied, which rely on the skill of traditional artists. When it comes to reproduction of duplicates of existing ceramic parts, the main problem is accuracy, which for artistic reasons can be dropped to the fairly acceptable levels. This problem is particularly significant in relation to free-form surfaces, which are often met in traditional pottery. This work advocates using state-of-the-art engineering tools in design and production of ceramic parts and the corresponding press-forming and casting dies. Feature-based design, point-cloud based modelling scanning of artefacts and fuzzy systems as approximate decision making tools in the area of process planning are the main technologies employed. The scope of this work is to introduce these technologies to traditional ceramics technicians, in order to advance their technical level step-wise. The use of features as a tool for parametric design not only of industrial components /1/, but also of free-form solids in an artistic and aesthetic manner is widely established /2-3/. Design with features is widely accepted particularly because they can refer directly to machining operations /4/. Features can be even produced from point clouds /5/, and customized feature libraries can be automatically generated /6/. A feature library is developed in this work for the parametric design of parts with simple geometry, focusing on concave pottery parts. A user-friendly interface is developed in Visual Basic, so that parametric design can be achieved by technicians with no proficiency in CAD.
Proceedings of the 3rd International Conference on Manufacturing Engineering (ICMEN), 1-3 October 2008, Chalkidiki, Greece Edited by Prof. K.-D. Bouzakis, Director of the Laboratory for Machine Tools and Manufacturing Engineering (), Aristoteles University of Thessaloniki and of the Fraunhofer Project Center Coatings in Manufacturing (PCCM), a joint initiative by Fraunhofer-Gesellschaft and Centre for Research and Technology Hellas, Published by: and PCCM

747

As far as copying and reconstruction of existing artifacts is concerned, many methods are introduced based on registration of point clouds /7-8/, while the large number of points received from scanners always have to be filtered /9/. In this work, special routines were developed for enhancing commercially available functionality in conversion of point clouds, taken from a laser scanner, into solid models, with or without surface approximation techniques. Solid models resulting from both methods lead to the appropriate die geometry. Traditional ceramists are by no means experienced machinists, therefore for the actual production of the die, they need help in deciding how to machine to die. Fuzzy sets have also proved a reliable tool for the evaluation and estimation of machining parameters /10-12/. In this work an easy to use fuzzy system is developed for the selection of suitable machining operations (roughing, semi-finishing, finishing), machining strategy, cutting tools and cutting parameters (scallop height, overlap, stepdown etc.), respecting geometry complexity, total material removal volume and a coupled time-accuracy criterion depending on the user. The system is developed so as to gradually introduce the appropriate parameters to the user, to allow comprehension of their impact on the process and their mutual interaction and to recommend appropriate values for the relevant parameters. Each of the three areas of development mentioned above will be presented in some detail in the respective sections 2, 3 and 4 followed by a draw-up of conclusions.

2.

PARAMETRIC DESIGN USING FEATURES

There are many cases where a part has similar geometry to other, already modeled, parts. Many companies offer products, some parts of which are variations of a standard geometric family. In other cases, the designer does not know from start the final dimensions of the model, or he/she just needs to experiment with the parts shape. In any case, the modern CAD systems provide the necessary tools to accomplish these goals. This is when parametric design proves indispensable. This approach involves different design techniques which allow the designer to describe the dimensional parameters of the model not only through numbers, but also through mathematical equations, which can relate them with variables and other parameters. This allows the model to be defined first in terms of general topology, leaving the definition of the final geometry to a subsequent designing stage. This results in faster design and quicker geometry changes. This design philosophy is accomplished in CAD systems through features. 2.1. Feature Library A feature library was developed, with the use of which the ceramics designer can easily model many common geometries, without having to start from scratch, when it comes to new parts. The library comprises of geometric families often met in the area of ceramics. The library features fall into certain categories, according to their general geometry and the part of the model they belong to. The following general categories are distinguished: Bases Concave walls Handles Solid bodies Tiles

The bases category includes features pertaining to the design of the lowest part of any ceramic product, e.g. the base of a ceramic plate, an amphora or a statuette. Regardless of the final ceramic object as a whole, it is certain that many of them have common bases. As a result, the user can pick from this general category the desired base feature and import it into the model. Some of the library features included in this category are presented in Figure 1.

748

3rd ICMEN 2008

(a) Figure 1: Library features in the bases category.

(b)

The second category, namely walls includes features that represent the main geometry of a usually hollow item, e.g. pots, bottles, vials, amphora, but also plates, etc., while the handles category is named after those parts of a ceramic item that are used to lift, hold and move them. It includes many kinds of solids that can be applied usually on a wall feature (Figure 2).

(a) Figure 2: Library features included in the walls category.

(b)

The last two general categories contain features with solid geometry. The former includes any geometry that is not hollow, while the latter those that are square or rectangular with a certain pattern only on the upper surface. In practice, the user can combine features from different categories to build a new model. Since there is the possibility of editing a feature before and during its import to the model, there are infinite combinations that can be acceptable. Feature libraries are useful tools supported from almost all modern CAD systems. Its use can dramatically reduce the design time, especially when the desired object can be modeled entirely from features within the library. Use of the feature library is very simple, even to the non-familiar user, because the transition from features to the final product is achieved through few steps. Every feature has its characteristic name and it is accompanied by a distinguishing thumbnail. Features are inserted in the model by simply dragging & dropping from the library to the design space. They can be left loose until the final design stage or they can be bound to the rest of the model from start with specific dimensions or parametric relations and constrains.

CAD, CAM, CMM, Vitrual Manufacturing

749

2.2. Expanding the feature library One of the most important parameters of the library is its possibility of expansion. It is not necessary to import each new model in the library, but when it becomes obvious that the specific geometry should prove useful in the future, it is better to save it as a new feature for later use. Consequently, it should be named appropriately and located in the right category. Of course, new categories can be created, when needed. The design of features does not differ from any other kind of modelling, except that they should be saved with the respective extension, according to the CAD platform used. Special attention should be also given to feature internal relations and constrains, for optimal control of its behavior during its import to new models. Usually, features are designed by expert users and designers. In order to make this technology available even to non-familiar users, special software was developed in Visual Basic, adopting a very practical paradigm. According to this, the user should sketch the necessary views of the model in plain paper and locate the XY coordinates of each views characteristic points. These coordinates should be then entered in a text or spreadsheet file, following the X,Y number format. Then, the software developed takes on, by importing these points automatically into the CAD environment. At all times, guidance by the user is absolutely necessary, on how these points are to be connected (using lines, arcs or splines). The user-system interface is friendly, as presented in Figure 3.

(a) Figure 3:

(b)

User-system interaction forms (a) for connecting points of a sketch and (b) for confirmation.

Groups of points can be separately imported, forming separate sketches in the same model. These sketches are then used, to form the final feature geometry, with the automatic execution of the appropriate commands (extrude, revolve, extruded cut etc.), as specified by the user. He should always follow the instructions given by the system, to successfully complete the model. All answers are critical to the procedure and should be precise and logical.

3. SCANNING AND MODELLING OF PARTS WITH FREE-FORM SURFACES Before one decides to scan an object, one must be sure that it cannot be designed by any means in a CAD system. Scanning is a time-consuming operation, therefore it should be used only when it is absolutely necessary. This usually happens when dealing with free-form or sculptured surfaces, often met in traditional ceramic objects. All 3D scanners actually digitize the scanned surfaces and usually have two exporting options. They either export a text file containing a point cloud, or an STL model of the scanned object, which results after the system has applied a triangulation technique. The latter is useful only if one needs an exact copy of the object. STL files can be opened by nearly all commercial CAD platforms. However, if editing of the file is needed, e.g. a copy of the original object in a different scale, the first option should be preferred. The point cloud contained in the exported text file,

750

3rd ICMEN 2008

can be easily read by CAD systems. If not, it is still possible to develop a tool in this direction, as shown in this work. In any case, a large number of points is necessary during the scan, in order to acquire the surface precision needed. On the other hand, a large point cloud is hard to manipulate, thus highend computers are necessary. To overcome this problem, the user has to achieve different local point densities, according to the demanded accuracy and to the shape of the scanned surface. For example, areas with intense changes in the radius of curvature need a far more dense point cloud, than areas that are almost flat. One solution is to apply different resolutions in different areas of the part during a scan, which can prove difficult and time-consuming. Some CAD systems provide clever algorithms for local point cloud filtering, which can be used after scan. Which of the two methods is to be used, relies upon the types of surfaces found on the object and on the experience of the user. Some of the point clouds scanned in this work appear in figure 4. Both objects have actually two surfaces: one flat and one free-form. Scanning was performed with uniform resolution and cloud filtering was applied aftewrwards.

(a)

(b)

Figure 4: Point clouds achieved with a laser scanner, using uniform scanning resolution: (a) 246.000 points and (b) 763.000 points.

(a) Figure 5: Solid models achieved after surface reconstruction.

(b)

Surface fitting tools provided by the CAD system were applied. The point clouds presented in figure 4 were fitted with a 2nd grade spline surface, resulting in maximum deviation of 0.35mm and average deviation of far less than 0.1mm, which was deemed acceptable. After thorough checking of the fitted surfaces in the CAD system, it was found that deviations over 0.1 mm were obtained in areas where lack of points was observed, i.e. in close neighborhood of edges

CAD, CAM, CMM, Vitrual Manufacturing

751

of the object, as clearly presented in both point clouds of figure 4. The surfaces reconstructed by these algorithms appear in figure 5. The surface in figure 5b is more complex than the one presented in figure 5a, consequently a comparatively less aggressive point cloud filtering algorithm was applied. Furthermore, other algorithms, provided by the CAD system, were also used. The steps followed for every object scanned were the following: Point cloud filtering Triangulation and mesh creation Hole filling of the mesh Surface fitting Solid model creation

Special routines were developed for this purpose, to help users handle the creation and hole filling methods of the mesh. These routines are actually scripts that, when activated by the user inside the CAD environment, start an interactive user-system procedure. The system triggers sequentially the necessary commands and the user has to select the appropriate geometries in the design area, or enter the appropriate values asked for. Explanatory messages pop-up in every step to inform the user of the procedure he is engaged in and its particulars. The first script has a point cloud as an input, and performs all necessary actions to create a proper, closed mesh. However, in some cases the object is divided into areas each of which is scanned separately with possible overlaps of multiple point clouds, which have to be initially fitted together. The fitting is performed by the second script, which actually joins the point clouds. The newly created point cloud can then be processed by the first script.

4. PRODUCTION OF THE CASTING DIE After creating the solid model of an object, it is easy to create the corresponding die, using Boolean commands on solids. The problem that arises is the clay shrinkage during casting, which has to be taken into account on the solid model before the die design. Clay shrinkage ranges between 5% and 20%, depending on the material and the objects cross-section. On this level of progress of the work, a uniform shrinkage was applied (10%). This is not realistic, of course, thus research is focused on an algorithm that can predict local shrinkage, taking into account all relevant parameters. A uniform shrinkage can be achieved in the corresponding surfaces of the model with the surface offset command. Using this methodology, the dies resulting from the models shown in figure 5 are presented in figure 6.

(a) Figure 6: Respective dies of the models presented in figure 5.

(b)

752

3rd ICMEN 2008

As far as the actual production of the die is concerned, it comes to a multi-parametric process planning problem. Usually, in these kind of dies, three sequential machining operations are applied: roughing, semi-finishing and finishing. The second one is not always necessary, depending mostly on the surface complexity of the die, the tools used in the other two processes and the precision demanded. At the same time, in every process different cutting strategies can be applied, and for every strategy, different parameters can be used. Table 1 presents the parameters involved in every cutting strategy, according to the operation used. Furthermore, independently of all other parameters, the surface offset, the tool diameter and the cutting conditions (e.g. feed per tooth and surface speed) should be taken into account. Table 1: Parameters involved in every cutting strategy and machining operation used. Cutting Strategies Roughing Hatch Contour Semi-Finishing Linear Circular pocket Constant Z Finishing Linear Circular pocket Constant Z Paremeters involved Overlap, stepdown Overlap, stepdown Stepdown, stepover, scallop, max.stepover Stepover Stepdown, scallop, max.stepover Stepdown, stepover, scallop, max.stepover Stepover Stepdown, scallop, max.stepover

In this work, all machining operations were studied, while every strategy was simulated with different combinations of parameters, for two different dies shown in figure 6. This led to 188 scenarios in total, simulated in a commercial CAM platform. This way, valuable trends were discovered on how all the above mentioned parameters affect the time needed for the die production and the quality achieved. The evaluated results were used for the development of a fuzzy system for the selection of suitable machining operations, machining strategy, cutting tools and cutting parameters. Fuzzy system The developed fuzzy system is addressed to those users that are unfamiliar with machining operations. It constitutes a good tool for preliminary definition of the machining processes for the construction of the die, recommending the tool diameter, the cutting strategy needed and the appropriate cutting parameters. It comprises of 6 sub-systems which are executed sequentially: Roughing Tool Finishing Tool Semi-finishing Rough Strategy SemiFinish Strategy Finish Strategy

The name of each sub-system evidently explains its purpose. Generally, for every parameter involved in the system, a fuzzy membership function was created, and parameter values were associated according to the results of the cases studied experimented with in CAM. The output of each rule is a fuzzy set; fuzzy sets are aggregated into a single output fuzzy set. Finally the resulting set is defuzzified, or resolved to a single number. The first two sub-systems are independent, but have to be both executed before going on to the

CAD, CAM, CMM, Vitrual Manufacturing

753

third. The decision over the roughing tool diameter depends on two parameters: the total volume of the material to be removed, and the initial to final surface ratio. This ratio, representing the ratio of corresponding areas, is an index of the complexity of the final surface of the die. The relationship of these two parameters is presented in figure 8. The decision over the finishing tool diameter accordingly depends on the minimum radius of curvature of the final die surface and on the Time-Quality criterion, set by the user. The latter has to be chosen by the user, as an integer between one and five, with five meaning that the target is die quality, without any concern about time (figure 7). Both these first sub-systems give a result that is not necessarily an integer. It is obvious that the user should finally choose the closest available tool diameter. Time Quality

Figure 7: The Time-Quality membership function used in the fuzzy system. When both roughing and finishing tool diameters are chosen, the semi-finishing sub-system should be activated, for a conclusion to be drawn over the necessity of such an operation. This relies on three fuzzy membership functions, which describe the initial to final surface ratio, the time-quality criterion, as well as the roughing tool diameter, as already described. Combination of these three parameters leads to conclusions on the necessity of a semi-finishing operation (figure 8).

RoughToolDia

RoughToolDia

T_V (a)

In-FSR

T_V (b)

In-FSR

SemiFinishing In-FSR (c)

TTQ

Figure 8: Surfaces expressing relationships between input and output parameters of the fuzzy sub-systems, where Tot_Vol the total volume to be removed, In-FSR the initial to final surface ratio and TTQ the target time-quality parameter. (a) Roughing Tool (b & c) Semi-finishing for different pairs of input parameters.

After all decisions concerning the machining operations and the corresponding tool diameters are made, the last three sub-systems activate sequentially. The cutting strategies depend on the tool diameter and the values of the respective cutting parameters, as shown in Table 1. However, in this case, instead of having the cutting parameters as an output, the time-quality criterion is used. Treating the tool diameter as a constant, since it is already chosen, the user has to properly adjust the cutting parameters so as to achieve the initially set value of the timequality criterion, see figure 9. The reason for this inversion is simple: instead of executing the system as a black-box, the user can now visualize the effect of each parameter, being gradually introduced to the involved parameters and comprehending both their impact on the process and

754

3rd ICMEN 2008

FinishToolDia Stepover Scallop MaxStepover FinishStrategy Figure 9: Structure of the fuzzy sub-system for the selection of the finishing strategy. FinishStrategy Selection

Target Time-Quality

Figure 10: The appropriate value for the time-quality criterion in the last column can be approached by adjusting the values of all input parameters.

their mutual interaction. Adjustment of the parameters is achieved by the user-friendly interface of the fuzzy system in the Matlab environment, see figure 10.

5. CONCLUSIONS In this work, product engineering technologies such as laser surface scanning, feature-based CAD modelling and CAM techniques are integrated for next generation design and production of precision traditional ceramics and the corresponding press-forming and casting dies. Since ceramic technicians or artists are not familiar with these technologies, all tools were developed in such a manner, so as to provide a comfortable, explanatory and user-friendly environment. The feature library was developed for particular classes of shapes and their function on the ceramic artefact, but it can equally well be expanded towards other geometries. In addition, a lowlevel, practical methodology was adopted for guided step-by-step conversion of 2D sketches on paper to 3D model in CAD systems. In laser scanning of artefacts with a view to transferring their models into CAD environments, it was realised that some steps of the process needed enhancement, so special scripts were also written to sew different point clouds together and to repair unwanted holes in fitted meshes. Solid model offsetting is used to create an expanded die cavity to account for shrinkage, is automatically offset, and subtracted to result in the model of the die, taking into account the clay shrinkage. Help in deciding machining processes, tool diameters and cutting parameters is certainly necessary for a ceramics technician to enter the machinists world, therefore a fuzzy system prototype was developed.

CAD, CAM, CMM, Vitrual Manufacturing

755

Improvements can be achieved by expanding the fuzzy system with more parameters, in order to increase results reliability. Furthermore, the clay shrinkage is to be studied thoroughly, so as to provide a local prediction algorithm and incorporate it into a variable offset methodology.

ACKNOWLEDGEMENT This work has been funded by ELKEA SA.

6. REFERENCES 1. Kim C., OGrady P.J., A representation formalism for feature-based design, ComputerAided Design, 28 (1996) 451-460. 2. Stamati V., Fudos I., A parametric feature-based CAD system for reproducing traditional pierced jewellery, Computer Aided Design, 37 (2005) 431-449. 3. Pernot J.-P. et al., Incorporated free-form features in aesthetic and engineering product design: State-of-the-art report, Computers in Industry, 59 (2008) 626-637. 4. Sun G. et al., Operation decomposition for free-form surface features in process planning, Computer-Aided Design, 33 (2001) 621-636. 5. Ke Y., Fan S. et al., Feature-based reverse modeling strategies, Computer-Aided Design, 38 (2006) 485-506. 6. Qamhiyah A.Z., A strategy for the construction of customized design libraries for CAD, Computer-Aided Design, 30 (1998) 897-904. 7. Benko P., Martin R.R., Varady T., Algorithms for reverse engineering boundary representation models, Computer Aided Design, 33 (2001) 839-851. 8. Zhang L.Y., Zhou R.R., Model reconstruction from cloud data, J. Mat. Processing Technology, 138 (2003) 494-498. 9. Huang J., Menq C.H., Combinatorial manifold mesh reconstruction and optimization from unorganized clouds with arbitrary topology, Computer Aided Design, 34 (2002) 149-165. 10. Ip R.W.L. et al., An economical sculptured surface machining approach using fuzzy models and ball-nosed cutters, J. Mat. Processing Technology, 138 (2003) 579-585. 11. Hashmi K., El Baradie M.A., Ryan M., Fuzzy-logic based intelligent selection of machining parameters, J. Mat. Processing Technology, 94 (1999) 94-111. 12. Wong S.V., Hamouda A.M.S., El Baradie M. A., Generalized fuzzy model for metal cutting data selection, J. Mat. Processing Technology, 89-90 (1999) 310-317.

756

3rd ICMEN 2008