of wax patterns created by RTV silicone rubber molding using the Taguchi approach S Rahmati Imam Hossein University, Tehran, Iran J Akbari Department of Mechanical Engineering, Sharif University of Technology, Tehran, Iran, and E Barati Sharif University of Technology, Tehran, Iran Abstract Purpose Dimensional accuracy analysis of wax model created by room temperature vulcanization (RTV) silicone rubber molding to be used in investment casting is presented The purpose of this paper is to discuss the effective parameters of dimensional accuracy in RTV tooling technique Design/methodology/approach After producing a pattern by stereolithography (SLA) and then creating the RTV silicone rubber mold by the SLA pattern, dimensional accuracy of wax models produced by RTV tool are analyzed Design of experiments (DOE) using the Taguchi approach is used for analysis and determination of optimum condition Findings Experiments show that the dimensional accuracy in RTV technique is as accurate as in traditional molding Hence, RTV tooling technique can be used in investment casting Using Taguchi approach based on DOE, it was realized that the optimum condition to achieve acceptable accuracy is 358C for mold temperature, 858C for wax temperature, and 205 barg for vacuum pressure Practical implications RTV silicone rubber mold is a useful alternative of metallic mold to produce wax patterns for investment casting It has benefits such as reduction in production lead-time and cost, compared with traditional metallic mold Originality/value A case study for research of dimensional accuracy of wax patterns created by RTV silicone rubber mold had not been attempted as such In addition to reduction in production lead-time and cost, the dimensional accuracy of wax patterns using RTV tooling technique are as accurate as in traditional technique Keywords Oils, fats and waxes technology, Waxes, Rapid prototypes, Taguchi methods Paper type Research paper 1 Introduction Investment casting also known as the lost wax process is one of the oldest manufacturing processes It can be used to make parts that cannot be produced by normal manufacturing techniques, such as turbine blades that have complex shapes, or airplane parts that have to withstand high temperatures This process starts by creating a mold to the desired specifications This mold will be used to inject wax to create the patterns needed for investment casting Traditionally, a metallic mold is used that involved significant time and expense Rapid prototyping (RP) and rapid tooling (RT) technologies can reduce these factors RP techniques are capable of producing prototypes of very complex part geometry directly from three-dimensional CAD software The current issue and full text archive of this journal is available at wwwemeraldinsightcom/1355-2546htm 13/2 (2007) 115 122 q Emerald Group Publishing Limited [ISSN 1355-2546] [DOI 101108/13552540710736803] (Kochan et al, 1999) RT is a technology that transforms prototype into a functional part (Zhou and He, 1999) Producing a pattern by stereolithography (SLA) and then creating the room temperature vulcanization (RTV) silicone rubber mold by the SLA pattern is a useful method for producing wax pattern This method has advantages such as faster and cheaper than traditional tooling Thus, metallic mold can be replaced by silicone rubber mold However, silicone rubber mold must have good dimensional accuracy in order to be used in investment casting There are many researches about variety methods of RP and RT technologies, but less material about silicone rubber molding technique or dimensional accuracy of wax pattern have been presented Horacek and Lubos (1996) studied the influence of injection parameters on the dimensional stability of wax patterns produced by injection molding process In their work, they found the interrelationship between various The authors would like to thank Mr Ali Reza Soraya, the Leader of RP and tooling team of Sapco Co for his help during the experimentation for this study Received: 25 April 2006 Reviewed: 27 May 2006 Accepted: 2 August 2006 115
S Rahmati, J Akbari and E Barati injection parameters and their dependency on other parameters Yarlagadda and Hock (2003) compared accuracy of wax patterns created by hard tool (polyurethane mold) and soft tool (RTV mold) They found that for the soft tool, the injection pressure and holding time have very little effect on the accuracy of wax patterns and the injection temperature is the only parameter that has an effect on the dimensional accuracy Figure 1 The pattern shape 2 Methodology In this work, first, a pattern was produced by SLA and then, silicone rubber mold was generated by SLA pattern Next, vacuum casting machine was used for creating wax patterns Experiments were focused on optimization of casting parameters such as wax temperature, vacuum pressure, and mold temperature to achieve better dimensional accuracy of the product The Taguchi approach for design of experiments (DOE) was applied to this research After the determination of optimum condition by means of Taguchi approach, other experiments were done using those optimum parameters in order to verify the results 21 Design of pattern The first step in this research was designing the specific shape of the product Several issues were considered for this purpose Some of the key issues considered for the design of shape of the product are as follows: various lengths for analysis of length effect on dimensional accuracy; various thicknesses for analysis of thickness effect; constrained and unconstrained sections for comparing of these sections and their differences to influence on dimensional accuracy; and complexity of the shape for easy removal of the wax pattern from the mold By considering these items, a pattern with F-shape as shown in Figure 1 was considered The dimensions which are shown in Figure 1 are the nominal dimensions which have been decided arbitrarily On the other hand, the ideal dimensions are the nominal dimensions plus the shrinkage factors Finally, the actual dimensions are the dimensions of the actual casting wax 22 Determination of shrinkages In traditional molding, the shrinkages of casting metal and wax must be considered to create the cavity of mold Thus, the cavity dimensions of mold should be bigger than nominal data in order to compensate the shrinkages of wax and casting metal In silicone rubber mold, in addition to these shrinkages, the shrinkage of silicone rubber should be applied to nominal dimensions, too In this research, for obtaining the average shrinkage of silicone rubber, an experimental mold was produced The production of this mold was done by a metallic pattern, under the same conditions of main mold The difference between dimensions in metallic pattern and mold cavity in each section indicates the shrinkage of silicone rubber Although the average shrinkage of silicone is claimed to be 01 percent using silicone commercial catalogue, our research indicates to be 015 percent It should be noted that the shrinkage of wax produced by silicone rubber mold is supposed to differ from what was produced by metallic mold, because the thermal conductivity of these two molds are different from each other and, therefore, the rate of solidification of wax are different In order to apply a reasonable amount for wax shrinkage produced by silicone rubber, some experiments were done at various conditions It was found that the wax has an average shrinkage of 065 percent for constrained sections, and 285 percent for unconstrained sections The shrinkage of casting metal was also determined from casting design handbook 23 Silicone rubber mold production For manufacturing of a physical master model by SLA, first 3D-CAD model was created by Solidworks software The shrinkage of wax, silicone rubber, and metal casting were applied to the nominal dimensions Then, this model was converted to STL format by 3D Lightyear software STL is a standard format in RP industries which approximates 3Dmodel surfaces with several triangles After implementation of some complementary actions on the STL model, like model review, defining supports and build orientation, final file was sent to RP apparatus In this project, master model was produced by photo curable Cibatool SL 5195e resin with a 3D systems SLA- 5000e machine Part layer thickness used in this process was 01 mm After producing SLA master, some post-processing activities such as washing excessive resin, removing supports and it was finally post cured in an UV oven To make the silicone rubber mold, first SLA master model was located into the paste while the half of the master was inside it and the other half was outside it Then, the master and the paste were placed in a frame V-750 silicone rubber with CAT 750 catalyst (10:1 mixing ratio) 116
S Rahmati, J Akbari and E Barati was mixed by a mixer machine for 20 min and was poured into the frame Next, the whole assembly degassed by a vacuum machine for 20 min Next day, the assembly was upside down and the paste was removed After rubbing silicone barrier on the surface of silicone mold, the next half was built similar to the first half Finally, the half of silicone rubber mold was separated from the other half, and SLA master was removed from the mold Figure 2 shows the steps of this procedure 3 Experiments 31 The Taguchi approach Assume an engineering experiment requires n control factors and m control levels per control factor to understand the influence and interaction of its input data on the output results By using a traditional experimental process, usually all the possibilities (m n tests) need to be carefully conducted before an optimal performance can be concluded The Figure 2 The subsequent steps of silicone rubber tooling 117
S Rahmati, J Akbari and E Barati number of tests can rapidly get very large To reduce the number of such tedious and costly tests while still be able to maintain an insight into the overall effects of the input factors on the output, Taguchi approach based on the DOE should be considered (Roy, 2001) In this technique, only a few numbers of tests that systematically choose certain combinations of values of those control factors are required, but it is possible to separate their individual effects This approach not only saves considerable time and cost but also leads to a more fully developed process by providing systematic, simple and efficient methodology for the optimization of the near optimum design parameters with only a few well defined experimental sets (Prasad et al, 2005) Figure 3 A sample of wax pattern 32 Experimental program This study presents the optimization of dimensional accuracy of wax patterns produced by RTV silicone rubber mold Three factors with four levels for each factor were considered These factors and their levels are shown in Table I The Taguchi method reduced the mandatory experiment runs from 4 3 ¼ 64 to only 16, by using modified L-16 orthogonal array The other factors such as room temperature, room humidity, injection rate and so on were assumed constant in all experiments First, the mold and sprue were cleaned and placed in an oven with specified temperature in order to stabilize the mold temperature at the desired temperature (for example, 40 min to an hour) Then, the mold was placed in the vacuum casting machine (MCP 006) and vacuumed to the desired pressure while cast After an hour, the mold was opened and the wax pattern was removed from the mold The pattern was measured by a digital caliper three times for each experiment, after removal from the mold, subsequently after 1 h, 2 h, and 5 h, to reduce the measuring errors Then, the average of these measurements was reported It should be noted that all measurements were carried out at the same condition by the same measuring apparatus and operator A sample of wax pattern is shown in Figure 3 It was intended to determine the effects of production parameters on dimensional accuracy of wax patterns Some parameters may have insignificant effects on accuracy such as ambient temperature, room humidity, or sprue diameter On the other hand, some parameters could not be controlled because of equipment disadvantages For example, the injection speed may have some significant effects on dimensional accuracy, but the MCP machine could not control this parameter So it was not possible to analyze the effect of this parameter By considering these problems and elimination of parameters which may no effect or insignificant effect on output parameter (dimensional accuracy), it seems that those three parameters were good choice It is clear that if other significant and controllable parameters are suggested, they could be used as future work There are ten sections on the pattern named A, B, Jas shown in Figure 4 For example, in Section D, it is important to specify the location of measurement in order to pressure consistency Thus, all measurements of all patterns and mold cavity were done at the same location each time It should be reported only one data for each experiment Thus, it should be determined a characteristic number for each experiment that indicates dimensional errors For this purpose, first the ideal dimensions were calculated These dimensions are the nominal dimensions, which the metal casting shrinkage has been applied on them Then, the ideal and actual dimensions for each feature were compared, and the percent differences between them were calculated for each section The root mean squares (RMS) of these data is a useful value as the characteristic number for each experiment Based upon the RMS results, it would be evident that as RMS Figure 4 Parametric shape Table I Factors and their levels Levels Serial Factors 1 2 3 4 A Wax temperature ( o C) 70 75 80 85 B Vacuum pressure (barg) 0 2025 205 2075 C Mold temperature ( o C) 25 35 45 55 118
S Rahmati, J Akbari and E Barati value gets smaller, ie less dimensional error and more accuracy For the state of repeatability, each trial was repeated twice The modified L-16 (M-16) orthogonal array and results are shown in Table II The term of Result in Tables V and IX is the characteristic value Since, each trial was repeated twice in M-16 orthogonal array, the Result 1 indicates the characteristic number of first experiment, and Result 2 indicates that value for second experiment mean squares (or variance): F-ratio: pure sum of squares: V A ¼ S A f A F A ¼ V A V e ð3þ ð4þ 33 Analysis of variance (ANOVA) Analysis in DOE refers to the things that are done with results after experiments are carried out and test samples are evaluated (Roy, 2001) The total effects of factors are obtained by adding the results containing the effects of factors at the desired level The average effects (or the main effects) are obtained by dividing the total effects by the number of results The main effects of factors and their plots are shown in Table III and Figure 5, respectively For a set data (results) Y 1 ; Y 2 ; Y N the total variation can be calculated by adding deviations of the individual data from the mean value, hence: S T ¼ Xn i¼1 ðy i 2 YÞ 2 Following a similar approach, the variation caused by an individual factor, for example, A, is obtained by an expression called the factor sum of squares, as: S A ¼ A2 1 þ A2 2 þ 2 T 2 N A1 N A2 N Where T ¼ total of results; N A1 ¼ the total number of experiments in which level 1 of factor A is present; and A 1 ¼ the total results that include factor A 1 The other related terms are: ð1þ ð2þ percent influence: S 0 A ¼ S A 2 ðv e f A Þ P A ¼ S0 A S T ð6þ where V e ¼ the variance for the error term (obtained by calculating error sum of squares and dividing by error degrees of freedom); and f A ¼ the degrees of freedom of factor A (Roy, 2001) The obtained experimental data was processed with smaller is better quality characteristics for the determination of the optimum dimensional accuracy of wax pattern, in order to identify individual factors influence on the accuracy and to estimate the performance of optimum conditions The ANOVA of these experiments is shown in Table IV From Table IV, it is clear that the influence of other/error term is 736 percent This significant number includes these terms: the interaction influence between factors; the experimental errors; and the error caused by variations of shrinkage percent of wax in each section of pattern Since, this value is bigger than other percents, it should be determined that how much of this number is caused by the interaction influences between factors ð5þ Table II Experimental design of the Taguchi analysis with modified M16 orthogonal array and results of experiments A B C Trial Test sequence Column 1 Column 2 Column 3 Column 4 Column 5 Result 1 Result 2 1 14 1 1 1 1 1 07719 07402 2 9 1 2 2 2 2 07587 07181 3 13 1 3 3 3 3 05387 05583 4 6 1 4 4 4 4 06855 06791 5 10 2 1 2 3 4 08746 08216 6 7 2 2 1 4 3 06530 05555 7 1 2 3 4 1 2 07272 07113 8 8 2 4 3 2 1 05738 05347 9 15 3 1 3 4 2 06487 06532 10 11 3 2 4 3 1 05437 05311 11 5 3 3 1 2 4 04494 04825 12 2 3 4 2 1 3 09838 09599 13 12 4 1 4 2 3 06655 06845 14 3 4 2 3 1 4 05886 05027 15 16 4 3 2 4 1 06016 06099 16 4 4 4 1 3 2 05123 05918 119
S Rahmati, J Akbari and E Barati Table III Main effects of selected factors Serial Factor Level 1 Level 2 Level 3 Level 4 A Wax temperature 0681 0681 0657 0595 B Vacuum pressure 0733 0606 0585 0690 C Mold temperature 0748 0608 0622 0636 Figure 5 The main effects of factors The main effects of factors and ANOVA for these experiments are shown in Tables VI and VII, respectively 35 Determination of optimum condition The contribution of individual factors is the key for obtaining optimum point In Taguchi approach, ANOVA is used to analyze the results of experiments and to determine how much variation each factor has contributed By studying the main effects of each factor, the general trends of the influence of the factors towards the process can be characterized The characteristics can be controlled in such away that the lowest value in a particular influencing factor produces the preferred result Thus, the levels of factors which produce the best results could be predicted The contribution of each factor and optimum point are shown in Table VIII 36 Verification In order to verify the methodology, three trials under the optimum condition were done The results are shown in Table IX The results have shown that under this condition, the patterns are more accurate than the other conditions Thus, the results of experiments are in good agreement with Taguchi approach Table IV ANOVA Factors DOF Sum of squares Variance F-ratio Pure sum Percent A 3 00403 00134 1058 00022 04 B 3 01161 00387 3052 00781 146 C 3 00987 00329 2594 00607 114 Other/error 22 02790 00127 736 Total 31 100 34 Interaction Understanding the interaction between two factors gives a better insight into the overall process analysis Any individual factor may interact with any or all of the other factors creating the possibility of presence of a large number of interactions (Prasad et al, 2005) For obtaining influences of interactions between factors, by using the L-8 orthogonal array, eight experiments had to be done In these experiments only levels of 1 and 4 (two extremes) for each factor were considered The L-8 orthogonal array and the results are shown in Table V 4 Discussion This F-shape pattern is a case study for determination of input parameters (wax temperature, vacuum pressure, and mold temperature) effects on output parameter (dimensional accuracy) and determination of optimum conditions This work has some suggestions for future work The results of experiments which were carried out at the same condition indicate that the results are close to each other Thus, it could be concluded that the results have good reliability The individual factor influences in Table I, clearly indicates that its results are in good conformance with the results of Table IV In meanwhile, the insignificant differences of the results are due to the fact that two separate set of experiments have been executed which would obviously be not exactly identical It could also conclude that the wax temperature which is only 1 percent has minimum effect on the dimensional accuracy However, the factor of vacuum pressure has the maximum influence on the dimensional accuracy of wax patterns However, the interactions between factors have significant effect on accuracy of wax patterns By considering these effects (Table VII) it is found that the interaction between wax temperature and vacuum pressure has the highest effect on dimensional accuracy Table V Experimental design with L8 orthogonal array and results of experiments A B A 3 B C A 3 C B 3 C Trial Test sequence Column 1 Column 2 Column 3 Column 4 Column 5 Column 6 Column 7 Result 1 3 1 1 1 1 1 1 1 07361 2 8 1 1 1 2 2 2 2 06533 3 6 1 2 2 1 1 2 2 06354 4 1 1 2 2 2 2 1 1 05333 5 4 2 1 2 1 1 1 2 06159 6 7 2 1 2 2 2 2 1 05963 7 2 2 2 1 1 1 2 1 06017 8 5 2 2 1 2 2 1 2 06265 120
Table VII ANOVA Factors Dimensional accuracy analysis S Rahmati, J Akbari and E Barati Table VI Main effects of selected factors Factors Level 1 Level 2 L1 2 L2 A 0639 0610 0029 B 0650 0599 0051 A 3 B 0654 0595 0059 C 0647 0602 0045 A 3 C 0649 0601 0048 B 3 C 0628 0622 0006 DOF Sum of squares Variance F-ratio Pure sum Percent A 1 00017 00017 3398 00002 10 B 1 00052 00052 10294 00037 162 A 3 B 1 00070 00070 13776 00055 239 C 1 00040 00040 7933 00025 110 A 3 C 1 00045 00045 8869 00030 131 B 3 C 1 00001 00001 0155 00014 61 Other/error 1 00005 00005 287 Total 7 00231 1000 The interaction effects play an important role in analysis and conclusion Referring to Table IV, it is observed that the wax temperature has 04 percent of influence, while vacuum pressure and mold temperature have 146 and 114 percent of influences, respectively But these data are not accurate enough for calculating overall influence of these factors In other word, the total effects of these factors on dimensional accuracy aren t equal to sum of influences of individual factors (For example, 04 þ 164 þ 114 from Table IV) Therefore, it is necessary to include the interaction influences in addition to individual effects for calculating total effects of these factors The overall effects of these three factors can be calculated by sum of individual factors effects and interaction effects between factors from Table VII Thus, the total effects of wax temperature, vacuum pressure, and mold temperature are 100 2 287 ¼ 713 percent Although the wax temperature has very little effect on accuracy of patterns (04 percent from Table IV or 10 percent from Table VII), its interaction with other factors have significant effect on accuracy (162 percent for its interaction with vacuum pressure and 110 percent for its interaction with mold temperature) Therefore, it could not be excluded from other factors The interaction between mold temperature and vacuum pressure has little effect (61 percent) Wax pattern shrinkages of were assumed to be constant for constrained and unconstrained features (285 percent for unconstrained sections and 065 percent for constrained sections) But experiments have shown for any individual trial, the ten different features have different shrinkage values Therefore, it is impossible to apply a fixed shrinkage value for a mold This is reflected from the other/error value is 287 percent from Table VII However, if the wax casting is reformed at optimum condition, the above mentioned shrinkage errors would fell into permissible range In general, the dimensional deviations of wax patterns are accumulated by SLA process, silicon rubber molding and wax injection The dimensional deviations of silicone rubber mold are insignificant and could be neglected Because of the shrinkage of silicone rubber was 015 percent and comparing with wax shrinkage (285 percent), this value is very small (1:19) In addition, the shrinkage of silicone rubber was measured by creating an experimental mold with similar shape This shrinkage percent then was applied to nominal dimensions to compensate silicone rubber dimensional deviations The dimensional deviations of SL pattern were accumulated on deviations caused by wax injection So, the dimensional errors are because of wax injection and SLA process deviations However, these errors in wax pattern which is produced under optimum condition are in the range of permissible tolerance of investment casting Thus, these deviations caused either by wax injection or by SLA process, are insignificant For other shapes and sizes, the results may not be exactly identical to the present results However, at optimum conditions, the dimensional errors are shown to be insignificant, and due to the fact that this model is arbitrary chosen, it was concluded that by applying the same methodology and correct determination of shrinkage at optimum conditions, the dimensional errors are not worrying In other word, if other shapes and sizes are designed applying the same methodology, first the shrinkages of the new pattern must be examined (which may differ from the values of the F-shape pattern), and then the optimum conditions to be determined 5 Conclusions The wax injection parameters such as wax temperature, vacuum pressure, and mold temperature have significant effect on the dimensional accuracy of wax patterns created by RTV tooling The importance of each factor was investigated and it was concluded that the vacuum pressure and its interaction with wax temperature are the most effective factors on the overall accuracy Using Taguchi approach based on DOE it was realized that the optimum condition to achieve acceptable accuracy is 358C Table VIII Optimum point of factors and their contribution Serial Factors Value Level Contribution A Wax temperature 85 4 200589 B Vacuum pressure 205 3 200686 C Mold temperature 35 2 200451 Total contribution from all factors 201726 Current grand average performance 06535 Expected result at optimum condition 04809 121
for mold temperature, 858C for wax temperature, and 205 barg for vacuum pressure The Taguchi results were confirmed by the results of experiments which were carried out under optimum conditions The errors resulted at optimum conditions are in the range of permissible for investment casting process The wax pattern created by RTV tooling is a successful alternative for traditional tooling References Dimensional accuracy analysis S Rahmati, J Akbari and E Barati Table IX Experiments under optimum condition Trial Wax temperature Vacuum pressure Mold temperature Result 1 85 205 35 04701 2 85 205 35 04824 3 85 205 35 04513 Horacek, M and Lubos, S (1996), Influence of injection parameters to the dimensional stability of wax patterns, paper presented at Ninth World Conference on Investment Casting, San Francisco, CA, pp 1-20 Kochan, D, Kai, CC and Zhaohui, D (1999), Rapid prototyping issues in the 21th century, Journal of Computers in Industry, Vol 39, pp 3-10 Prasad, KK, Mohan, SV, Rao, RS, Pati, BR and Sarma, PN (2005), Laccase production by Pleurotus ostreatus 1804: optimization of submerged culture conditions by Taguchi DOE methodology, Biochemical Engineering Journal, Vol 24, pp 17-26 Roy, RK (2001), Design of Experiments Using the Taguchi Approach, Wiley, Mississauga, ISBN 0-471-36101-1 Yarlagadda, P and Hock, TS (2003), Statistical analysis on accuracy of wax patterns used in investment casting, Journal of Materials Technology, Vol 138, pp 75-81 Zhou, J and He, Z (1999), A new rapid tooling technique and its special binder study, Journal of Rapid Prototyping, Vol 5 No 2, pp 82-8 Further reading King, D and Tansey, T (2002), Alternative materials for rapid tooling, Journal of Material Processing Technology, Vol 121, pp 313-7 Song, Y, Yan, Y, Zhang, R and Lu, Q (2001), Three dimensional non-linear coupled thermo-mechanical FEM analysis of the dimensional accuracy for casting dies in rapid tooling, Finite Element in Analysis and Design, Vol 38, pp 79-91 Corresponding author S Rahmati can be contacted at: rahmati@rapidtoolpartcom To purchase reprints of this article please e-mail: reprints@emeraldinsightcom Or visit our web site for further details: wwwemeraldinsightcom/reprints 122