Available online at www.sciencedirect.com ScienceDirect Procedia CIRP 4 (016 ) 7 18th CIRP Conference on Electro Physical and Chemical Machining (ISEM XVIII) EDM Drilling and Shaping of Cooling Holes in Inconel 718 Turbine Blades M. Kliuev a *, M. Boccadoro b, R. Perez b, W. Dal Bó b, J. Stirnimann c, F. Kuster a, K. Wegener a a Institute of Machine Tools and Manufacturing (IWF), ETH Zurich, Switzerland b GF Machining Solutions, Switzerland c inspire AG, Zurich, Switzerland * Corresponding author. Tel.: +41 44 6 51; fax: +41 44 6 11 5. E-mail address: klyuev@iwf.mavt.ethz.ch Abstract Machining capabilities of modern EDM drilling machines for drilling cooling holes and diffusers in turbine blades are described. For improving the process capability and machining efficiency of aerospace alloys, drilling of cooling holes and formation of diffusers can be combined into a single process by using the same electrode for both processes. Experimental results of material removal rate, relative tool wear and surface integrity for EDM drilling and shaping processes are shown. Particular attention is given to the process parameters analysis such as discharge current and discharge duration, their influence on the heat-affected zone and the influence of relative tool wear on the shape accuracy. Thereby the material removal rate reached 77 mm /min, relative tool wear was reduced down to 0 % and the average recast layer thickness was reduced to 8 µm. The best results were achieved separately with different sets of parameters. After diffusor shaping the roughness S a of the internal surface is less than 1 µm. Accordingly the EDM drilling and shaping processes demonstrate the possibility to combine high processing speed with relatively low recast layer thickness and good surface quality. 016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of 18th CIRP Conference on Electro Physical and Chemical Machining (ISEM Peer-review XVIII). under responsibility of the organizing committee of 18th CIRP Conference on Electro Physical and Chemical Machining (ISEM XVIII) Keywords: Electrical Discharge Machining (EDM), Drilling, Material Removal, Wear, Nickel alloy, White layer; 1. Introduction High burning temperature inside jet engines has made material selection complicated. Such materials required high strength and temperature resistance. According to [1] the Rolls-Royce Trend 800 and General Electric GE90, which was used in Boeing 777 required to resist temperature beyond 800 C. Presently superalloys, high creep resistant nickel based alloys, are the most widely used materials for turbine blades according to []. But also these superalloys materials need additional internal and external cooling to survive sufficiently long under these severe conditions.. For internal cooling, gas or liquid passes through channels inside the blade. One of the widely used methods for external cooling is film cooling. The turbine blade is cooled down by the air which is flowing through the cooling holes inside the blade and exits through diffusor openings in the blade wall to the outer surface of the blade. Each blade requires a large number of holes to evenly cover the blade with a cooling air film. Thereby the drilling of cooling holes became one of the primary processes in turbine blade manufacturing. The production of large number of cooling holes in super alloys is a complex production task. Holes of small diameter must be drilled with an inclination angle to the turbine blade surface to provide a correct cooling flow and protect the blade from overheating. For some applications conventional drilling and milling operations could be used. The application requires acute angles which would load the drill tip with high lateral forces leads to inaccurate machining and causes tool breakage due to bending according to []. Moreover nickel based alloys are difficult to cut with conventional processes. Presently in aerospace industry ECM, laser and EDM drilling are used to produce micro scale cooling holes in superalloy materials by [4-6]. Difficulties of conventional processes could be overcome by laser drilling, but this method had a high thermal effect on the machining surface and in the 1-871 016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the organizing committee of 18th CIRP Conference on Electro Physical and Chemical Machining (ISEM XVIII) doi:10.1016/j.procir.016.0.9
M. Kliuev et al. / Procedia CIRP 4 ( 016 ) 7 range of deep hole drilling the recast layer thickness varies between 80µm and 100µm according to [5] or 40µm with different drilling conditions according to [7], also laser drilling has worse geometric accuracy the deeper the hole becomes. It is possible to avoid any thermal effect to the surface by Electro Chemical Machining but cutting rate is relatively low 0.05 mm/s according to [8], which is equal to 1. mm /min. The material removal rate of Inconel 718 in EDM depends on following process parameters: electrode polarity, pulse duration, discharge current and dielectric properties. In [6, 9] it was in the range from 1. mg/s to 1.8 mg/s, which is equal to 9 mm /min to 1 mm /min. In [10] was shown that relative tool wear in EDM was varying between 6% and 44% and according to [11, 1] the recast layer thickness was between µm and 40 µm and could be reduced by additional finishing process. Therefore the recast layer thickness determination and minimization and also material removal rate (MRR) maximization became primary tasks of the current work. The recast layer thickness (RLT) is one of the most significant process parameters for the determination of quality limits for aerospace industry. In [1] it was found that a thick recast layer has a higher tendency to emit cracks. Therefore from recast layer thickness depends surface integrity and strength characteristics of turbine blades. The RTL of all drilled holes is measured. Specimen preparation including grinding, polishing and etching is performed. Polishing is performed with Al O abrasive with grain size µm. Etching solution VA is used and etching time is 0 s. The VA etching solution consists of a mixture of 100 ml water, 100 ml hydrochloric acid and 10 ml nitric acid. Etching temperature is 50 C. The RTL measurements and imaging are made by Alicona infinite focus microscope, Austria. Thereafter the images are transferred into a binary format as shown in Fig. 1, to make the recast layer clearly visible and measurable.. Experimental Method and Procedure A full factorial design of experiments is selected for exploring the parameter space. Independent variables are discharge current and pulse duration, which are specified in Table 1, as seven different levels. The dependent variables are relative tool wear, material removal rate MRR and recast layer thickness. Experiments are performed on Drill 00 from GF Machining Solutions, Switzerland. Table 1. Input parameters and selected levels Variable Unit Level Pulse duration (T on) Discharge current (I d) 1 4 5 6 7 µs 16 0 4 8 7 4 A 8.5 1 16 0 4 8 Inconel 718 as workpiece material and copper multi-hole electrodes are used for all drilling tests. The electrode diameter is 1 mm. Further EDM drilling machine parameters, are not changed during the experiment and are presented in Table. Table. EDM conditions EDM Condition Workpiece material Inconel 718 Electrode material Copper Dielectric Deionized water Holes type Blind Pause duration 15µs Ignition voltage 10V Feed of the electrode (H)* 10 mm Electrode polarity negative * Feed of the electrode is set on EDM machine and wear is not compensated Fig. 1. Recast layer image transformation, original image on the left and binary image on the right. The RLT measurement is based on line-by-line counting of the white pixels along horizontal lines after adjustment of the white layer vertically. The maximum thickness of recast layer is usually not higher than twice the average thickness R L, which is then, calculated from. R L = k i= 1 N i k K t Here N i is the number of white pixels in measured horizontal line i, while k is the total number of lines in the picture and K t is the transformation coefficient which represents the microscope resolution times the pixel width. Electrode wear measurements E Wt are performed after every erosion as reduction of the length of the electrode. From those measurements the relative tool wear E W = EWt 100% H E Wt (1) ()
4 M. Kliuev et al. / Procedia CIRP 4 ( 016 ) 7 is calculated as wear related to the eroded workpiece volume where E Wt is the total reduction of electrode length. Material removal rate MRR is calculated as ( H E Wt ) A MRR = () t where t is the machining time and A is the area of hole.. Results and discussion.1. EDM drilling model A second order polynomial regression is used as mathematical model for MRR and RLT; and a third order polynomial regression for the relative tool wear. Variables pulse duration T on and discharge current I d are taken into account by the regressions. Analysis of Variance (ANOVA) is used to investigate significance and the highest effect of variable independent parameters: pulse duration T on and discharge current I d to dependent process parameters: relative electrode wear, MMR and RLT. The p-value for statistical testing of experimental results in the model is 5 %, that evaluates the significance of the parameters. More information about p-value and F-value can be found in [1]. Table shows that the effect from individual variable T on is higher than the effect from I d and effects from higher order variables and interaction term. T on, T on, and T on have a significant contribution to the relative tool wear regression model, since p-values are less than 0.05. Table. Results from the analysis of variance for electrode wear. Source F-value p-value Regression 687.81 0.000 T on 46.1 0.000 I d 1.55 0.14 T on 46.97 0.000 I d 0.1 0.75 T on I d 0.5 0.471 T on 45.10 0.000 I d 0.08 0.78 Lack-of-fit 0.61 0.97 R =9.48% The value R is a statistically measured fitting of the regression model to the experimental data. For the relative tool wear model the value is 9.48% (Table ). It means that the model is able to explain the entire volume of data with a good reliability. The F-value of the regression is higher than the critical F-value (.04- tabulated value) with significance level of α=0.05. Also the p-value of lack-of-fit is 0.97, which indicates its insignificance. It means that all important terms such as interactions and quadratic terms are included in the model. Regression coefficients are present in Table 4, since all variables have their own units, the transformation of units is included in regression coefficients. Table 4. Regression coefficients for electrode wear E W. Source Coef SE Coef T-value p-value T on 6.85 0.01 6.79 0.000 I d -1.84 1.48-1.4 0.14 T on -0.78 0.098-6.85 0.000 I d 0.048 0.0785 0. 0.75 T on I d 0.0048 0.0066 0.7 0.471 T on 0.00 0.0005 6.7 0.000 I d 0.0004 0.001 0.8 0.78 Regression model based on coefficients from Table 4 is calculated as (%) = 6.85 1.84 0.78 E T I T + + + + W on d on 0.048 Id 0.0048 Ton Id 0.00 Ton 0.0004 Id Regression analysis of MRR in Table 5 shows that the effect from all variables T on, I d, T on, I d and T on I d has significant contribution to the relative tool wear regression model, since p-values are less than 0.05 Table 5. Results from the analysis of variance for MRR. Source F-value p-value Regression 499.6 0.000 T on 15.80 0.000 I d 65.07 0.000 T on 14. 0.000 I d 1.0 0.000 T on I d 1.4 0.000 Lack-of-fit 0.49 0.997 R =98.1% For the MRR model the value is R =98.1% (Table 5). It means that the model is able to explain the entire volume of data with excellent reliability. The F-value of the regression is higher than the critical F-value (.4- tabulated value) with significant level of α=0.05. Also the p-value of Lack-of-fit is 0.997, which indicates its insignificance. Table 6. Regression coefficients for MRR. Source Coef SE Coef T-value p-value T on 0.705 0.177.97 0.000 I d.061 0.56 8.07 0.000 T on -0.015 0.004 -.77 0.000 I d -0.0 0.006 -.6 0.000 T on I d 0.016 0.005.5 0.000 (4)
M. Kliuev et al. / Procedia CIRP 4 ( 016 ) 7 5 Regression model based on coefficients from Table 6 is calculated as mm = 0.705 on +.061 d 0.015 on MRR T I T min I + T I 0.0 d 0.016 on d The results of the regression analysis for RLT in Table 7 shows that the effect from individual variables T on and I d as well as effect from higher order variables T on and I d has significant contribution to the relative tool wear regression model, since the p-values are less than 0.05, whereas interaction between T on and I d is insignificant. Table 7. Results of the analysis of variance for RLT. Source F-Value p-value Regression 979.84 0.000 T on 104.0 0.000 I d 5.99 0.000 T on 48.18 0.000 I d 4. 0.000 T on I d 4.49 0.05 Lack-of-fit 0.74 0.889 R =97.% For the recast layer model the value is R =97.% in Table 7. It means that the model is able to explain the entire volume of data with excellent probability. The F-value of the regression is higher than the critical F-value (.4- tabulated value) with significant level of α=0.05. Also the p-value of Lack-of-fit is 0.889, which indicates its insignificance. Table 8. Regression coefficients for RTL. Source Coef SE Coef T-value p-value T on 0.5401 0.05 10.0 0.000 I d 0.878 0.076 5.10 0.000 T on -0.008 0.001-6.94 0.000 I d -0.009 0.00-4.9 0.000 T on I d 0.00 0.001.1 0.05 (5).. Electrode wear measurements Pulse duration parameter T on has a higher effect on the electrode wear than the discharge current parameter I d. The electrode wear decreases with increase of T on and I d parameters until T on = µs and I d = 0A. For the discharge current the optimum value is 0 A and for the pulse duration the optimum value is 8 µs. Further increase of discharge current leads to an increase of electrode wear, whereas the increase of pulse duration does not have a significant effect. The lowest achieved value of relative electrode wear is 0 %,. Fig.. The effect of main input parameters on relative tool wear... Material removal rate calculation Discharge current parameter has the highest effect on the MRR; its sensitivity is almost ten times higher than the one from the pulse duration. A discharge current of A maximizes the MRR, which is the highest tested value. The effect of pulse duration on MRR is similar to the effect on the electrode wear. The highest material removal rate is 77 mm /min. Regression model based on coefficients from Table 8 is calculated as ( μ ) = 0.54 + 0.89 0.008 R m T I T L on d on + 0.009 Id 0.00 Ton Id (6) Both independent variables have significant effect on the dependent variables: MRR, electrode wear E W and RLT and interactions between pulse duration and discharge current are insignificant in all three analyses. Fig.. The effect of main input parameters on MRR.
6 M. Kliuev et al. / Procedia CIRP 4 ( 016 ) 7 The MRR is proportional to the energy per pulse. This happens because the diameter of plasma channel grows simultaneously with the pulse duration and the same amount of power is spread into an increasing area. After µs the power per area becomes insufficient for maximum material vaporization and the MRR decreases. lateral part and it will change the shape condition of the electrode unpredictably..4. Recast layer analysis The lowest value of the RLT is achieved with the minimum values of the pulse duration T on and discharge current I d parameters. Minimum for discharge current is 8.5A and minimum for time duration is 16µm. The lowest average RLT achieved is 8 µm with 0.8 % standard deviation. Fig. 5. Layer-by-layer machining. A pocket made by EDM layer-by-layer shaping is shown in Fig. 6. The total depth of erosion is 0.5 mm and the shape is produced in three layers. The roughness S a of the bottom surface is less than 1 µm. The electrode path is shown by solid lines and the finishing path is shown by dashed lines. Finishing is performed to mitigate waviness. Fig. 4. Effect of main input parameters on RLT. The recast layer is formed by molten material which is not vaporized and flushed away with the dielectric and then solidifies on the surface. The recast layer increases with the volume of molten material which depends on the amount of energy transferred to the surface. Therefore minimum RLT is reached with minimum T on and minimum I d.5. EMD shaping results The uniform wear method introduced in [14] as shown in Fig. 5 is applied for shaping of In 718. A standard electrode of cylindrical shape is used to produce a complex shape as layerby-layer machining. The depth of layers influences the quality of the shape and the MRR of the machining process. Naturally the best condition for the shape precision would be to keep the thickness of the layers near the gap width between electrode and workpiece. But then the MRR and productivity will be very low. The other extreme will be maximization of the MRR, which is achieved when the thickness of the layer is in the range of the radius of the electrode. This limitation occurs because during the shaping process the top of the electrode becomes round and the roundness depends on the electrode size. If the thickness of the layer is higher than the electrode radius, then the electrode erosion will start to happen on the Fig. 6. EDM layer-by-layer shaped pocket in Inconel 718. Another more complicated case is the diffusor. It is made using the same layer-by-layer strategy as in the previous test and the workpiece is tilted by 45. The lateral part of the shape has a waviness from the chosen layer-by-layer strategy and can be improved by reduction of step size between layers.
M. Kliuev et al. / Procedia CIRP 4 ( 016 ) 7 7 The capability of EDM drilling cooling holes and shaping diffusor geometries is proven. Concerning EDM drilling the erosion speed of 1.6 mm/sec is achieved through process parameter optimization by means of nonlinear regression and response surfaces. The results reached with 1mm diameter electrode are as follows: An MRR of 77 mm /min is reached, the relative tool wear is reduced down to 0 % and the average RLT is reduced to 8 µm. The combination of EDM drilling and shaping processes demonstrates the possibility to produce cooling holes with diffusors without electrode change. After diffusor shaping the roughness S a of the internal surface is less than 1 µm. Particular attention is given to the RLT. It is found that the thickness after shaping is even lower than after drilling and significantly lower in comparison with laser drilling. Fig. 7. EDM shaped diffusor in Inconel 718 In Fig. 8, the cross-section of the diffusor is shown. The recast layer on the shaped part of the diffusor is even lower than in the drilled hole. For the drilling part the strategy optimum for recast layer is used (T on = µs, I d = 18А). 4. Summary 4.1. Results Fig. 8. Cross section of diffusor and recast layer analysis. The maximal MRR meaning and minimal Electrode wear are achieved with the value of pulse duration parameter T on = µs.. For discharge current I d optimization the maximum MRR is achieved under the maximal meaning of this parameter of I d = 0A. The minimum meaning of Electrode wear is achieved with paramiter I d = 18А. Minimum recast layer is reached with minimum T on and minimum I d. 4.. Conclusions The EDM study proved experimentally that the recast layer can be reduced to the average thickness of 8 µm. Moreover the significant drawbacks of EDM which are low MRR and expensive tool manufacturing are overcome. Acknowledgements This study has been provided with the financial support from the Commission for Technology and Innovation CTI, Switzerland. The authors appreciate the support provided by the Laboratory of Metal Physics and Technology, ETH Zürich and personally Mrs. Wegmann. References [1] Boeing. 777 Propulsion Design Review GE90 Strut Structure and Fairings. 1991. [] Reed RC. The Superalloys: Fundamentals and Applications: Cambrige universety press; 006. [] Chae J, Park SS, Freiheit T. Investigation of micro-cutting operations. International Journal of Machine Tools and Manufacture. 006;46:1-. [4] Naeem M. Advancement in laser drilling for aerospace gas turbines. Proceedings of the rd Pacific International Conference on Application of Lasers and Optics 008 008. [5] Okasha MM, Mativenga PT, Driver N, Li L. Sequential laser and mechanical micro-drilling of Ni superalloy for aerospace application. CIRP Annals - Manufacturing Technology. 010;59:199-0. [6] Kuppan P, Rajadurai A, Narayanan S. Influence of EDM process parameters in deep hole drilling of Inconel 718. Int J Adv Manuf Technol. 008;8:74-84. [7] Hasçalık A, Ay M. CO laser cut quality of Inconel 718 nickel based superalloy. Optics & Laser Technology. 01;48:554-64. [8] Sharma S, Jain VK, Shekhar R. Electrochemical Drilling of Inconel Superalloy with Acidified Sodium Chloride Electrolyte. Int J Adv Manuf Technol. 00;19:49-500. [9] Yilmaz O, Okka MA. Effect of single and multi-channel electrodes application on EDM fast hole drilling performance. Int J Adv Manuf Technol. 010;51:185-94. [10] Sánchez H, Estrems M, Faura F. Development of an inversion model for establishing EDM input parameters to satisfy material removal rate, electrode wear ratio and surface roughness. Int J Adv Manuf Technol. 011;57:189-01. [11] Cusanelli G, Hessler-Wyser A, Bobard F, Demellayer R, Perez R, Flükiger R. Microstructure at submicron scale of the white layer produced by EDM technique. Journal of Materials Processing Technology. 004;149:89-95. [1] Lee HT, Tai TY. Relationship between EDM parameters and surface crack formation. Journal of Materials Processing Technology. 00;14:676-8. [1] Dean A. Design and analysis of experiments: New York [etc.] : Springer; 1999. [14] Yu ZY, Masuzawa T, Fujino M. Micro-EDM for Three-Dimensional Cavities - Development of Uniform Wear Method. CIRP Annals - Manufacturing Technology. 1998;47:169-7.