You are on page 1of 17

December 2004 Technical paper submitted to the Indian Foundry Journal

Rapid Casting Development using Reverse Engineering, Rapid Prototyping and Process Simulation D. K. Pal1, B. Ravi2, L. S. Bhargava1 and U. Chandrasekhar3
1 2

Terminal Ballistics Research Laboratory, Sector-30, Chandigarh 160030, India Mechanical Engineering Department, Indian Institute of Technology, Powai, Mumbai 400076, India 3 Rapid Prototyping Section, Gas Turbine Research Establishment, Bangalore 560093, India. Correspondence: Dr. B Ravi, Mechanical Engineering Department, Indian Institute of Technology, Powai, Mumbai 400076. Tel: +91-22-25767510, Fax: +91-22-25726875, Email: bravi@me.iitb.ac.in

Abstract This paper delineates a synergistic approach involving reverse engineering, rapid prototyping and process simulation techniques for rapid development of intricate castings, required for replacement purposes. Reverse engineering (3D scanning) is employed to eliminate the solid modelling time, otherwise significant for intricate parts. Several widely-used rapid prototyping options are analysed for their impact on fabrication time, development cost, dimensional accuracy and surface quality of the part. Casting process simulation enables reducing the number of shop-floor trials to optimise the methoding and process parameters. The integration of the above three techniques compresses the lead-time for developing a casting from weeks to days. The entire approach is illustrated and validated through experimental studies on the development of an aluminium alloy impeller. It is superior to the traditional approach in terms of yielding more predictable and consistent results, but at present can be economically justified for only one-off intricate castings required in a short time. Keywords: Casting; Computer-Aided Design; Tooling; Reverse Engineering (RE); Rapid Prototyping (RP); Rapid Tooling (RT); Process Simulation.

1. INTRODUCTION In the fiercely competitive global markets, unlimited inflow of new products incorporating technological innovations provides a continuous challenge to manufacturing firms. Rapidity of product development along with concomitant cost control while maintaining the desired quality have become the underlying factors for success. Computer aided technologies for product design, analysis and manufacturing are becoming increasingly popular for achieving the above factors. In this context, conventional metal casting development cycle, which consumes several months for producing the first good sample, especially for replacing an existing part, presents an immediate opportunity for integration of advanced computer aided procedures with conventional practices. It is observed that the major portion of the lead-time of casting

development cycle, especially in the case of geometrically complex parts required in small order quantity, is spent on tooling development, modifications, trials and inspection. This leads to severe compromise in terms of time and cost competitiveness. This paper proposes an integrated approach involving reverse engineering, rapid prototyping, process simulation and metal casting to achieve speed and economy in developing such parts. Since numerous technology options are now available, especially for rapid prototyping, the present study focuses on a techno-economic comparison of the options based on their performance in terms of lead time, fabrication cost, dimensional integrity and surface quality. Related work in casting development using rapid prototyping techniques is reviewed next, followed by our proposed approach and experimental validation. 2. RAPID CASTING DEVELOPMENT: RELATED WORK Several researchers have explored the application of rapid prototyping (RP) techniques for foundry tooling development in the last few years. These applications can be mainly classified as direct, indirect and semi-direct. Direct tooling involves use of RP models themselves as patterns and core boxes for sand casting application. Indirect tooling makes use of RP models as intermediate masters for producing final patterns and core boxes, through processes such as epoxy mass casting, polyurethane face casting, metal spray and silicone rubber moulding [1, 2, 3]. This category of tooling, also known as soft tooling, can only be used for small quantity batches, due to more pronounced deterioration in comparison to conventional metal or wooden patterns. Semi-direct tooling involves the use of RP systems to make dies for producing wax patterns for investment casting. The RP-based tooling routes can also be classified as single step, double step and triple step based on the number of steps required for reaching the final tooling that is used for casting [4]. Many researchers have studied the application of a specific RP process for direct tooling for metal casting applications. For example, Mueller and Kochan found that castings produced by LOM patterns give good quality and acceptable accuracy with 25% cost saving [5]. In another investigation, Wang, Conley and Stoll concluded that the use of LOM tooling yielded about 50% saving in time and cost compared to conventional aluminium tooling [6]. Sushila et al [7] reported that the production of FDM tooling for a valve body results in 73% less time as compared to conventional methods. Even an indirect route, in which RP models are used to produce permanent tooling through investment casting, compresses tool fabrication time substantially [8]. Some studies on benchmarking of various RP processes have been reported. Different investigators have used different criteria for comparison. Wohlers compared various RP systems on the basis of total part cost consisting of maintenance, material, build, preprocessing and post-processing costs [9]. Shellabear carried out a detailed benchmarking study of dimensional accuracy and surface quality in RP models of 44 different RP processes [10]. Pham and Gault reviewed the strength and weaknesses of various rapid prototyping technologies in terms of common processes parameters such as layer thickness, system accuracy and speed of operation [11]. Lim et al studied the surface morphology of RP patterns produced using SLA, SLS and SGC by scanning electron microscope demonstrating the superior surface integrity of SLA parts [12].

A few researchers have made comparative studies of selected RP processes in terms of their suitability for producing casting patterns. In a pertinent study, Chua et al compared three alternative approaches: rapid pattern, rapid tooling and hybrid, for making moulds using Stereolithography process [13]. When RP processes were used to produce investment-casting patterns and subsequently cast metal parts, it was found that castings were generally less accurate than the RP model [14]. The foundry experience in producing casting from the RP pattern was found to be the most significant factor. A similar study was carried out by NASA to evaluate various RP techniques for fabricating the pattern for fuel pump housing, and it was concluded that RP techniques are effective for small and complex 3D patterns [15]. In general, study of accuracy in indirect tooling methods is more difficult than direct methods because it is necessary to take into account the accuracy of the moulds as well as intermediate parts produced in those moulds. With the growing number of direct and indirect routes for rapid tooling fabrication, for any stipulated casting application it is essential to select the most appropriate route. Indeed, Stoll, et al and Kochan, et al highlighted the need for a systematic approach for sand casting tool development using RP [16, 17]. In this context, Xu et al proposed an object-oriented knowledge based RP process selection approach to assist designers in decision making [18]. Braglia and Petroni proposed a methodology for selection and evaluation of RP systems using multi criteria decision making techniques [19]. Though rapid prototyping systems significantly reduce the tooling fabrication time, the total lead time for a good-first sample casting is still significant, since it involves the time for solid modelling the part (more for intricate shapes) before tooling fabrication as well as foundry trials after tooling fabrication. If the part already exists, the modelling time can be nearly eliminated by reverse engineering the shape using a 3D scanner. Similarly, the foundry trials can be reduced to just one by carrying out the trials on a computer using a casting process simulation software. This implies that all iterations and planning are carried out in the virtual domain, minimizing the expenditure of resources during actual manufacturing activities. Such an integrated approach, involving reverse engineering, rapid tooling fabrication and casting simulation, is described here, supported by a real-life example of an aluminium impeller casting. 3. REVERSE ENGINEERING AND TOOLING FABRICATION Reverse engineering involves capturing the geometry of a part, using contact or noncontact scanners, in the form of cloud of points. From this point cloud data, a CAD model of the part is created by surface fitting using suitable reverse engineering software. This CAD model is then used for design and fabrication of casting tooling using rapid prototyping technology. The approach is described in detail and illustrated using a case study of an impeller casting in the following sections. The impeller is used in an air blower in marine vehicles. The original impeller required replacement due to wear and tear, but any drawings or manufacturing data were not available, except a worn-out wooden pattern (Figure 1a). This part suited our study also because it had features appropriate for geometric comparisons: thin straight blades, a curved surface, a thick bottom portion and axi-symmetry.

Figure 1. 3D CAD model generation using laser scanner: (a) wooden pattern, (b) cloud of points, and (c) 3D CAD model

3.1 Reverse Engineering The original pattern was scanned using a Cyclone (Renishaw plc., UK) stand-alone scanning machine with a non-contact laser scanning probe. A step size of 0.1 mm was used to scan the part in a single set up in one orientation, yielding a cloud of points (Figure 1b). The point data file contained several thousands of points representing x, y, z coordinate values. This was imported in IGES format into reverse engineering software called Imageware Surfacer (EDS Inc., USA). The surface patches were fitted on each individual feature. Only those points that best described the feature were taken, leaving out other points located in worn out craters and depressions on the surface and edges. This helped in obtaining a clean model, devoid of imperfections and thus better representing the original part. The surfaces were then stitched into a water-tight 3D solid model (Figure 1c). This solid model was converted into STL file format for rapid prototyping. While scanning took about 5 hours to obtain the cloud of points, conversion to solid model (through surface fitting) took less than 4 hours. 3.2 Rapid Pattern Fabrication The part solid model was used for fabricating non-expendable patterns for sand casting as well as expendable patterns for investment casting using RP techniques. The nonexpendable RP patterns were fabricated using SLA, FDM and LOM RP techniques. The patterns FDM2 and FDM3 are made on the same machine, but the latter is made with a widely spaced cross-hatching for the interior region, reducing its fabrication time (given later). The expendable RP patterns were fabricated using SLA QuickCast and Thermojet (Figure 2.). Table 1 shows the summary of RP techniques and relevant parameters. Table 1. Summary of techniques used for producing RP patterns
RP Machine FDM1 FDM2 FDM3 SLA1 SLA2 SLAQ1 TJP1 LOM1 FDM Titan FDM 250 FDM 250 SLA 5000 SLA 250 SLA 5000 Thermojet LOM-2030H Helisys Technologies System manufacturer Stratasys Material Accuracy XY-plane (mm) 0.15 0.15 0.15 0.1 0.1 0.1 0.1 0.25 Accuracy Z-plane (mm) 0.13 0.13 0.13 0.10 0.10 0.10 0.10 0.30 Layer Thickness (mm) 0.25 0.25 0.25 0.10 0.10 0.10 0.10 0.20

3D Systems

Polycarbonate ABS(P400) ABS (P400) SLA5530 epoxy rein SLA5530 epoxy resin SLA5530 epoxy resin TJ88 wax Paper (LPH series)

All the RP processes discussed here require support structure for features such as overhangs and undercuts except the LOM process. The support structures are fabricated along with the part in the same or different material. These are subsequently removed by breaking, cutting or filing. Further filing and sanding is necessary in order to produce a good surface finish. The FDM process uses either a snap-away support structure made

from a second material that does not stick to the part material, or a water-soluble support structure. In Stereolithography parts, the supports are cut off and the areas of attachment are finished by sand blasting or filing by hand. The Thermojet systems use a hair-like support structure, which can be brushed away and the remaining downwardfacing surfaces cleaned up manually. The LOM process requires elaborate de-cubing of the trapped volume of the part with a hammer and chisel. 3.3 Soft Tooling Fabrication The RP pattern SLA1 made by Stereolithography (Figure 3a) is used as a master to fabricate a silicone rubber mould (Figure 3b). Silicone rubber is a versatile material that can be moulded around a master RP pattern to produce a cavity [23]. Since the impeller is flat at one end, the silicone rubber mould is made in a single piece without a parting line. For this purpose, the SLA1 part is placed in a cylindrical flask. Degassed silicone rubber slurry is slowly poured into the flask engulfing the master pattern, and then baked in a temperature controlled oven. The silicone rubber mould can be used for producing 40-50 wax patterns (Figure 3c) before the mould surface starts showing wear. These wax patterns are used for investment casting. 4. COMPARATIVE ASSESSMENT The different RP routes for producing non-expendable and expendable patterns for casting are compared in terms of fabrication time and cost, dimensional accuracy and surface quality. The details are given here. 4.1 Fabrication Time and Cost Comparison data on fabrication time and preparation cost for various RP options is included in Table 2. The per hour machine rate was calculated considering only the basic cost of the RP machine and 10% annual maintenance cost, and assuming 24 hour working over 365 days. The fabrication cost for a given part is given by the product of machine cost per hour and the fabrication time. The material cost is calculated by the product of material rate and part weight. The total cost of a RP part is given by the sum of the fabrication cost and material cost. The costs of conventional wooden and metal patterns are also given for comparison. These are estimated by a separate cost estimation program developed in our lab and verified by tooling experts. Table 2. Cost and time comparison
Machine cost Machine ($1000) rate ($/hr)* FDM1 100 12.56 FDM2 55 6.90 FDM3 55 6.90 SLA1 400 50.23 SLA2 200 25.11 SLAQ1 400 50.23 TJP1 60 7.53 LOM1 120 15.07 Conventional wooden pattern Conventional metal pattern Time taken (hr) 7 16 8 2.5 4 2.5 6 6 Material rate ($/kg) 330 300 300 250 250 250 225 20 Part weight (kg) 0.19 0.09 0.09 0.21 0.21 0.05 0.16 0.17 Total cost ($) 150.62 137.40 82.20 178.08 152.94 138.08 81.18 93.82 200.00 450.00

Figure 2. Fabrication of RP patterns: (a) FDM1, (b) FDM2, (c) FDM3, (d) SLA1, (e) SLA2, (f) SLAQ1, (g) TJP1 and (h) LOM1

Figure 3. Indirect route for wax pattern: (a) SLA1 master, (b) silicone rubber mould, and (c) wax pattern

4.2 Dimensional Accuracy Dimensional accuracy is a measure of the closeness of the RP part to the corresponding CAD model. The Figure 4 shows the dimensions l, b, w, d, D, H and t measured on both RP part and CAD model for dimensional comparison. Each dimension is measured three times and the average is recorded in Table 3. The absolute deviation is calculated with reference to the dimensions of the CAD model (in STL file format), since the same model was used for all RP processes. The STL file is itself an approximation of the original surface, since it involves faceting, but the corresponding deviation was limited to less than 0.02 mm by fine faceting while exporting the STL file (from the IGES surface). The percentage deviation for measuring the dimensional accuracy of an RP model is defined as the percentage ratio of the deviation from the corresponding dimension of the CAD model. Absolute deviation=Actual dimension-CAD dimension % deviation= (Absolute deviation / CAD dimension) x 100 Table 3. Dimensional accuracy comparison of RP processes Part Dimension Unit CAD Model Reference Actual FDM1 Deviation % Deviation FDM2 Actual Deviation % Deviation FDM3 Actual Deviation % Deviation SLA1 Actual Deviation % Deviation SLA2 Actual Deviation % Deviation SLAQ1 Actual Deviation % Deviation TJP1 actual Deviation % Deviation LOM1 actual Deviation % Deviation l mm 14.24 14.49 0.25 1.75 14.15 -0.09 -0.63 14.25 0.01 0.07 14.26 0.02 0.14 14.35 0.11 0.77 14.43 0.19 1.33 14.39 0.15 1.05 14.39 0.15 1.05 w mm 11.58 11.78 0.20 1.73 11.84 0.26 2.25 11.46 -0.12 -1.04 12.1 0.52 4.49 11.63 0.05 0.43 11.77 0.19 1.64 11.7 0.12 1.04 12.13 0.55 4.75 b mm 12.08 12.55 0.47 3.89 12.03 -0.05 -0.41 11.79 -0.29 -2.40 12.32 0.24 1.99 12.2 0.12 0.99 12.11 0.03 0.25 12.14 0.06 0.50 11.93 -0.15 -1.24 H mm 42.35 42.72 0.37 0.87 43.03 0.68 1.61 42.68 0.33 0.78 42.5 0.15 0.35 42.74 0.39 0.92 42.95 0.6 1.42 42.96 0.61 1.44 44.55 2.2 5.19 d mm 30.59 30.57 -0.02 0.065 30.42 -0.17 -0.56 30.73 0.14 0.46 30.76 0.17 0.56 31.04 0.45 1.47 31.13 0.54 1.77 30.93 0.34 1.11 30.77 0.18 0.59 D mm 93.0 93.10 0.10 0.11 92.85 -0.15 -0.16 93.22 0.22 0.24 93.49 0.49 0.53 93.57 0.57 0.61 94.29 1.29 1.39 93.39 0.39 0.42 92.33 -0.67 -0.72 t mm 2.30 2.33 0.03 1.30 2.29 -0.01 -0.43 2.43 0.13 5.65 2.33 0.03 1.30 2.40 0.10 4.35 2.49 0.19 8.26 2.61 0.31 13.48 2.64 0.34 14.78

Figure 4. Reference dimensions for comparison

Figure 5. Inverted Camera Microscope

10

4.3 Surface Quality The surface quality is evaluated in terms of surface finish and morphology. The surface roughness tester (Rotary TALY SURF-10, Rank Tylor Hobson, UK) was used over a distance of 2 mm at three locations on the vertical wall. The surface roughness values Ra and Rt measured for the various RP patterns are listed in Table 4. Table 4. Surface quality of RP parts FDM1 FDM2 FDM3 SLA1 SLA2 SLAQ1 TJP1 LOM1 Layer Thk (mm) 0.25 0.25 0.25 0.10 0.10 0.10 0.10 0.20 Condition Unfinished Unfinished Unfinished Polished Polished Polished Unfinished Polished and Varnished Ra (m) 17.307 19.847 17.540 2.578 4.001 4.469 2.578 1.374 Rt (m) 83.423 87.140 91.891 17.596 21.304 26.221 17.576 10.886

The surfaces of the RP patterns are observed using an inverted camera microscope (ZEISS ICM 405, Germany) shown in Figure 5. The microscope is fitted with an image grabber CCD camera with 1000X magnification. The captured image of the surface is enhanced using an image processing software (Image-Pro PLUS 4.5, Media Cybernetics, Inc., USA) for better visualisation. The images are shown in Figure 6. In our study, the SLA and LOM parts have the best vertical and best horizontal surfaces. The surface roughness on horizontal surface of the parts was found to be around Ra 1.0 m and was found to be independent of the parameters such as layer thickness, material and process. It is also observed that the SLA pattern has the highest level of surface integrity followed by LOM, FDM and Thermojet parts. 4.4 Stair-casing Effect As well known, the layer by layer fabrication of RP parts creates a stair-case effect, which is more pronounced at sloping surfaces. The effect can be reduced by decreasing the layer thickness, but with adverse impact on part fabrication time and preparation cost. With increased layer thickness, stair-casing effect becomes more pronounced, necessitating greater post processing effort through finishing operations. The staircasing effect also depends on the RP process employed to build the part, which is studied here. For this purpose, the curved geometry of the casting patterns generated by SLA, FDM, LOM and TJP is probed using a contour tracing machine (Contourecord 1600D from Carl Zeiss) with a touch probe. The results are given in Figure 7. The staircasing effect is greater on surfaces at a small angle with respect to the horizontal plane and reduces as the angle increases. This is an inherent drawback of layered manufacturing processes, which can be reduced by decreasing the layer thickness, or by employing adaptive slicing (thicker slices for vertical surfaces and thinner for inclined surfaces) if available in the RP system.

11

e Figure 6. Surface of RP patterns (at 1000X): (a) FDM, (b) SLA, (c) LOM, (d) SLA QuickCast, and (e) Thermojet

12

Figure 7. Stair-casing effect evaluated by contour tracing for: (a) FDM2, (b) SLA2, (c) SLAQ1, (d) TJP1, (e) LOM1, and (f) silicone rubber mould.

13

Figure 8. Casting methoding and analysis: (a) feeder design and modelling, and (b) verification by solidification simulation.

14

5. METHODING AND CASTING Casting methoding involves its orientation in mould, followed by location, design and modelling of feeders (for compensating solidification shrinkage) and gating system (for leading the molten metal to the casting cavity). The complete casting model (including feeders and gating) can be simulated for mould filling and solidification to predict related defects such as inclusions and shrinkage porosity. Based on the results of simulation, the methoding can be modified and verified (by simulation) until the desired quality is obtained while maximising yield (ratio of casting weight to total casting weight including feeders and gating system). In this project, the methoding and simulation of the impeller casting was carried out using AutoCAST software (Advanced Reasoning Technologies, India). For this purpose, the STL file of the impeller pattern was imported into the software and oriented to minimise undercuts, followed by setting of the metal and process (sand casting). The preliminary analysis showed a hot spot region in the centre close to the top of the casting, and suggested a top feeder (Figure 8). The feeder was modelled and verified by casting simulation, which showed that the hottest region (and therefore the shrinkage porosity) shifted to the feeder. The entire exercise, starting from importing the pattern model to feeder design and optimisation took less than 3 hours. In contrast, one iteration of pattern modification, foundry trial and casting inspection in conventional practice would have taken at least 3 days. Assuming 3 iterations to get the desired quality and yield, it would have taken at least one working week to get a good casting. For sand casting, the RP patterns made through LOM, FDM and SLA can be used directly. The corresponding castings are shown in Figure 9. A major problem is compensating for shrinkage of the RP pattern during the building process (especially if phase change is involved). The shrinkage of metal casting from solidification temperature to room temperature also affects the dimensions of the final casting. The extent of shrinkage however, also depends on the casting shape besides the metal and process. For example, free shapes shrink more than constrained shapes or features. This is a complex phenomenon and beyond the scope of the present project. In this project, no shrinkage allowance was added to the RP patterns, so that the extent of dimensional deviation between the RP parts with respect to the CAD model can be measured accurately and used for future studies. 6. CONCLUSION This paper showed how the bottleneck (in terms of lead-time) of pattern development for metal casting can be overcome by a combination of reverse engineering, rapid prototyping and simulation technologies. This approach has been successfully demonstrated by taking up an industrial case study of an impeller casting. The comparative assessment of various routes for pattern fabrication yielded valuable data in terms of time and cost, dimensional accuracy and surface quality. The proposed approach is especially useful for urgently replicating worn out parts that have a complex shape and are required one-off or in small numbers. In near future, innovations in RP

15

technology are expected to lead to advantages in terms of accuracy and cost as well, especially for small intricate parts. Acknowledgements The authors acknowledge the assistance of the Rapid Prototyping Section of Gas Turbine Research Laboratory, Bangalore in RP pattern fabrication. The project was partly funded by the Department of Science and Technology, New Delhi, through a project on New technology for one-off intricate castings. References 1. S. Ashley, Prototyping with Advanced Tools", Mechanical Engineering, 116, pp. 48-55, 1994. 2. P. Dvorak, Here Comes Rapid Tooling, Machine Design, 13, pp. 57-64, 1998. 3. C. K. Chua, T. H. Chew and K. H. Eu, Integrating Rapid Prototyping and Tooling with Vacuum Casting for Connectors, International Journal of Advanced Manufacturing Technology, 14(9), pp. 617-623, 1998. 4. M. M. Akarte, B. Ravi, 2000, RP/RT Route Selection for Casting Pattern Development, Manufacturing Technology, Proceedings of 19th AIMTDR Conference, pp. 699-706, 2000. 5. B. Mueller, and D. Kochan, Laminated Object Manufacturing for Rapid Prototyping and Pattern Making in Foundry Industry, Computers in Industry, 1, pp. 47-53, 1999. 6. W. Wang, J. G. Conley, and H. W. Stoll, Rapid Tooling for Sand Casting using Laminated Object Manufacturing process, Rapid Prototyping Journal, 3, pp. 134141, 1999. 7. B. Sushila, K. Karthik, and P. Radhakrishnan, Rapid Tooling for casting-A case study on application of Rapid Prototyping processes, Indian Foundry Journal, 11, pp. 213-216, 1999. 8. M. C. Warner, 1997, Metal Rapid Prototyping methods and case studies for metal casting and tooling, Rapid News, 6, pp. 1-5, 1997. 9. T. T. Wohlers, Chrysler compares rapid prototyping systems, Computer-Aided Engineering, 11(10), pp. 84-91, 1992. 10. M. Shellabear, Process chains for Rapid Technical Prototypes (RAPTEC), Benchmark Study of Accuracy and Surface Quality in RP Models, Final report for EU project, EOS GmbH, Germany, June 1999. 11. D. T. Pham and R. S. Gault, A Comparison of Rapid Prototyping Technologies, International Journal of Machine Tools & Manufacture, 38 (10-11), pp. 1257-1287, 1998. 12. C. S. Lim, P. Eng, S. C. Lin, C. K. Chua and Y.T. Lee, Rapid Prototyping and Tooling of Custom-Made Tracheobronchial Stents, International Journal of Advanced Manufacturing Technology, 20, pp. 44-49, 2002. 13. C. K. Chua, K. H. Hong, and S. L. Ho, Rapid Tooling Technology. Part 1. A Comparative Study, International Journal of Advanced Manufacturing Technology, 15 (8), pp. 604-608, 1998.

16

14. P. M. Dickens, R. Stangroom, M. Greul, B. Holmer, K. K. B. Hon, R. Hovtun, R. Neumann, S. Noeken, D. Wimpenny, et al., Conversion of RP Models to Investment Castings, Rapid Prototyping Journal, 1(4), pp. 4-11, 1995 15. A. T. Spada, Investment Casting Discuss RP, Ceramics Strength, Modern casting, 1, pp. 38-41, 2000. 16. H. W. Stoll, J. G. Conley, and W. Wang, Tool path selection for sand casting, 103rd AFS Congress and CastExpo held in St. Louis, Missouri, pp. 13-16, 1999. 17. D. Kochan, C. K. Chua, and Du Zhaohui, Rapid prototyping issues in the 21st century, Computers in Industry, 39 (1), pp. 3-10, 1999. 18. F. Xu, Y.S. Wong, and H. T. Loh, A Knowledge-based decision support system for RP&M process selection, Proceedings of the SFF Symposium held in Austin, Texas, pp. 9-18, 1999. 19. M. Braglia, A. Petroni, A management-support techniques for the selection of rapid prototyping technologies, Journal of industrial Technology, 15(4), pp. 1-6, 1999.

17

You might also like