OPTIMIZATION OF MULTIPLE PERFORMANCE CHARACTERISTICS IN
EDM PROCESS OF HPM 38 TOOL STEEL USING RESPONSE SURFACE
METHODOLOGY AND NON-LINEAR PROGRAMMING
Amirul Akbar
1, Bobby O. P. Soepangkat
1and Arif Wahjudi
21Manufacturing Process Lab., Department of Mechanical Engineering, Faculty of Industrial Technology, Institut Teknologi Sepuluh
Nopember, Surabaya, Indonesia
2
Design and Development Product Lab., Department of Mechanical Engineering, Faculty of Industrial Technology, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
E-Mail: [email protected]
ABSTRACT
The application of response surface methodology and non-linear programming for optimizing multiple performance characteristic in electro discharge machine (EDM) sinking process of HPM 38 steel was investigated. In this research, the main objective was to minimize surface roughness with electrode wear rate and material removal rate as constraints. The experiments were conducted based on Box-Behnken design (BBD) consisting 27 numbers of experiments. Quadratic model regression of response surface methodology was developed as efficient approaches to determine the optimal machining parameters in EDM process. Analysis of variance was applied to investigate the influence of process parameters (pulse current, gap width, on time and off time) and their interactions on surface roughness, electrode wear rate and material removal rate. A confirmation test was carried out to check the deviation of the predicted (optimum value) with experimental results.
Keywords: EDM sinking, surface roughness, electrode wear rate, material removal rate, response surface methodology, non-linear
programming.
INTRODUCTION
Electro discharge machine (EDM) is one of the non-conventional machining processes which is more efficient than conventional machining processes due to the ease of machining materials with complex geometry and burr-free [1]. EDM is used for machining of hard and high strength conductive materials which are hard enough to cut by conventional processes. In the process, there is no physical contact between the electrode and work piece that can eliminate mechanical stresses, chatter and vibration problems during machining. Hence, precision machining can be achieved by EDM [2]. It thus plays a major role in the machining of dies, tools, etc., made of tungsten carbides, stellites or hard steels [3].
In EDM, the removal of material is based upon the electro discharge erosion effect of electric sparks occurring between the electrode and work piece that are separated by a dielectric fluid. Material removal takes place as a result of the generation of extremely high temperatures (thermal energy) generated by the high-intensity discharge that melt and evaporate the electrode and work piece [4]. The function of dielectrid fluid in EDM process is to flush out the eroded particle from the machining gap, provide insulation between the electrode and work piece and cooling the section that was heated by the discharge effect [5].
Performance characteristic of EDM machining process generally determined by surface roughness (SR), electrode wear rate (EWR) and material removal rate (MRR). The EDM parameters which affect the performance characteristics are polarity, on time, off time,
working voltage, working current, flushing pressure and spark gap [6, 8]. Electrode wear and material removal produced during EDM process can also affect the surface roughness significantly. A high quality of surface roughness in EDM machining process can be obtained with a low rate of removal material, but the slow process will affect the completion time of products and will increase the cost of production.
This research takes HPM 38 steel as work piece and investigates three performance characteristics, i.e., surface roughness, material removal rate and electrode wear rate. Experiments were carried out using a
four-variable Box-Behnken design. The performance
characteristics of surface roughness and electrode wear rate are smaller the better and material removal rate is higher the better. Response surface methodology was employed to develop experimental models. Non-linear programming was applied to minimize the surface roughness with electrode wear rate and material removal rate as contraints.
EXPERIMENTAL DESIGN
Machining Parameter Selection
work piece were separated by kerosene dielectric using side flushing. An area with 15 mm (length) x 10 mm (width) and 2 mm (depth) the work piece EDMachined. Machining experiments for determining the optimal machining parameters were carried out by setting pulse current (PC), gap width (GW), on time (On) and off time (Off). The total number of levels which provided in EDM machine for pulse current is 20 levels, gap width is 90 levels, on time is 19 levels and of time is l9 evels. One level of pulse current, gap width, on time and off time are
equal to 3 amperes, 1 m, 84.21 s and 84.21 s
respectively. In this research, the experiments were conducted based on Box-Behnken design (BBD) consisting 27 number of experiments.
Machining parameters and their levels using Box-Behnken design are shown in Table-1. In Box-Box-Behnken design of response surface methodology, each parameter has 3 levels (low, medium and high) and the natural value of each machining parameter must be transformed to code units (Xi). The machining parameters are usually coded to -1 as a low level, 0 as a medium level and 1 as a high level. The machining parameter transformed model is [9]:
Xi = Xreal – (Xmax + Xmin)/2
(Xmax – Xmin)/2 (1)
The Box-Behnken design matrix of the experiment shown in Table-2.
Table-1. Machining parameters and their levels.
Machining
Parameters Unit
Low level
Medium level
High level
Pulse current
(X1) Ampere 4 8 12
Gap width (X2) µm 27 40 53
On time (X3) µs 2 7 12
Off time (X4) µs 2 6 10
Table-2. Matrix of experiment design.
Run Ord.
Code Parameters
� � � � PC GW On Off
1 0 1 0 -1 8 53 7 2
2 -1 0 0 -1 4 40 7 2
3 0 0 0 0 8 40 7 6
4 0 0 0 0 8 40 7 6
5 1 -1 0 0 12 27 7 6
6 -1 0 0 1 4 40 7 10
7 0 1 -1 0 8 53 2 6
8 0 0 -1 1 8 40 2 10
9 0 1 1 0 8 53 12 6
10 -1 1 0 0 4 53 7 6
11 0 -1 0 1 8 27 7 10
12 0 1 0 1 8 53 7 10
13 -1 0 1 0 4 40 12 6
14 0 -1 0 -1 8 27 7 2
15 1 0 0 -1 12 40 7 2
16 0 0 1 1 8 40 12 10
17 0 0 -1 -1 8 40 2 2
18 1 0 0 1 12 40 7 10
19 1 0 1 0 12 40 12 6
20 0 0 0 0 8 40 7 6
22 1 1 0 0 12 53 7 6
23 0 -1 1 0 8 27 12 6
24 -1 -1 0 0 4 27 7 6
25 1 0 -1 0 12 40 2 6
26 0 -1 -1 0 8 27 2 6
27 -1 0 -1 0 4 40 2 6
Machining Performance Evaluation
The measurements of surface roughness were performed by using a Mitutoyo surftest 401 with a cut off
length ( s) of 8 µm, sampling length ( c) of 0.25 mm and
the number of sampling length is 10 times. The arithmetic mean surface roughness (Ra) was used for measuring the
value of surface roughness. The material removal rate (MRR) is the ratio of material removal volume (Vm) and
machining time in minute (t). The electrode wear rate (EWR) is the ratio of electrode removal volume (Ve) and
machining time in minute (t). The formulas of MRR and EWR are:
MRR = Vm (mm
3)
EWR = Ve (mm
3)
t (minute) (3)
Design of Experiments
The experiments were designed based on the Box-Behnken design (BBD) of response surface methodology (RSM). The Box-Behnken design is an independent quadratic design in that it does not contain an embedded factorial or fractional factorial design [9]. The advantages of Box-Behnken design are rotatable design; rotatable means that the model would possess a reasonably stable distribution of scaled prediction variance throughout the experimental design region [10].
Experimental Results
RSM is a collection of mathematical and statistical technique useful for the modeling and analysis of problems in which a response of interest is influenced
by several variables and the objective is optimized the responses (multiple performance characteristics) [10]. To find a suitable approximation for the true functional relationship between the independent variables (machining parameters) and the response, the second order model is utilized in RSM [9]. Table-3 shows the experimental result of multi performance characteristics. The experimental results are used for developing of an empirical model using RSM.
Model Adequacy Checking and Significance Test
Experimental results were used to develop the second order (quadratic) regression model using Minitab software. Regression analysis and analysis of variance (ANOVA) were employed to check model adequacy and significance of machining parameters [11]. Tables 4 and 5 show the results of regression analysis and ANOVA.
Table-3. The experimental results of performance characteristics based on Box-Behnken design.
Parameters SR EWR MRR
Run Ord.
Parameters SR EWR MRR
X1 X2 X3 X4 µm
mm3/mi
n
mm3/mi
n X1 X2 X3 X4 µm mm
3
/min mm3/min
1 0 1 0 -1 5.25 0.691 28.31 15 1 0 0 -1 9.4 1.9713 30.03
2 -1 0 0 -1 6.35 0.5926 15.64 16 0 0 1 1 5.7 0.604 27.3
3 0 0 0 0 3.85 0.3379 10.87 17 0 0 -1 -1 5.25 0.8061 19.9
4 0 0 0 0 3.45 0.3695 10.03 18 1 0 0 1 9.15 1.5214 20.39
5 1 -1 0 0 10.5 1.7456 35.72 19 1 0 1 0 10.3 1.5081 30.66
6 -1 0 0 1 4.85 0.8237 14.25 20 0 0 0 0 3.55 0.3865 9.74
7 0 1 -1 0 4.25 0.7487 17.35 21 0 0 1 -1 8.05 0.5741 31.69
8 0 0 -1 1 4.75 0.5582 12.94 22 1 1 0 0 7.7 1.7612 22.4
9 0 1 1 0 5.55 0.3529 26.24 23 0 -1 1 0 7.9 0.6393 36.77
10 -1 1 0 0 5.2 0.6435 17.34 24 -1 -1 0 0 5.5 0.7065 22.55
11 0 -1 0 1 5.6 0.5657 29.8 25 1 0 -1 0 7.55 2.079 12.38
12 0 1 0 1 4.9 0.3782 18.53 26 0 -1 -1 0 4.95 0.6996 20.8
13 -1 0 1 0 5.65 0.7154 18.07 27 -1 0 -1 0 4.85 0.6627 13.01
14 0 -1 0 -1 7.3 0.6409 32.85
Table-4. Regression analysis.
Reg.
Anlys. Surface Roughness Electrode Wear Rate Material Removal Rate
Term Coef SE
Coef T pvalue Coef SE Coef T pvalue Coef
SE
Coef T pvalue
Constant 3.6167 0.1111 32.553 0 0.36463 0.02885 12.638 0 10.213 0.7310 13.972 0
X1 1.85 0.05555 33.303 0 0.53685 0.01443 37.215 0 4.231 0.3655 11.576 0
X2 -0.7417 0.05555 -13.351 0 -0.03518 0.01443 -2.438 0.031 -4.038 0.3655 11.047 - 0
X3 0.9625 0.05555 17.326 0 -0.09671 0.01443 -6.704 0 6.196 0.3655 16.952 0
X12 2.5479 0.08333 30.577 0 0.74282 0.02164 34.328 0 3.33 0.5482 6.079 0
X22 0.9979 0.08333 11.976 0 0.09808 0.02164 4.533 0.001 -10.335 0.5482 18.851 0
X32 1.0167 0.08333 12.201 0 0.14496 0.02164 6.699 0 5.170 0.5482 9.43 0
X42 1.2417 0.08333 14.901 0 0.11735 0.02164 5.423 0 6.963 0.5482 12.7 0
X1 * X2 -0.625 0.09622 -6.496 0 -2.04 0.633 -3.223 0.007
X1* X3 0.4875 0.09622 5.067 0 -0.1559 0.02499 -6.239 0 3.305 0.633 5.221 0
X1 * X4 0.3125 0.09622 3.248 0.007 -0.17025 0.02499 -6.814 0 -2.063 0.633 -3.258 0.006
X2 * X3 -0.4125 0.09622 -4.287 0.001 -0.08388 0.02499 -3.357 0.006 -1.77 0.633 -2.796 0.015
X2 * X4 0.3375 0.09622 3.508 0.004 -0.0594 0.02499 -2.377 0.035 -1.63 0.633 -2.587 0.023
X3 * X4 -0.4625 0.09622 -4.807 0 0.06945 0.02499 2.78 0.017
S = 0.192435 PRESS = 2.2554
S = 0.0499724 PRESS = 0.168339
S = 1.26609 PRESS = 104.832 R-Sq = 99.57%
R-Sq(pred) = 97.82%
R-Sq = 99.58% R-Sq(pred) = 97.63%
R-Sq = 98.81% R-Sq(pred) = 94.02%
Table-5. Analysis of variance (ANOVA).
Surface Roughness Electrode Wear Rate
Source D
F Seq SS Adj SS
Adj
MS F pvalue
D
F Seq SS Adj SS
Adj
MS F pvalue
Regressio
n 14
102.95 5
102.95
5 7.3539
198.5
9 0 13 7.0637 7.0637
0.5433 6
224.1
6 0
Linear 4 62.473 62.473 15.618
2
421.7
6 0 4
3.6422 6
3.6422 6
0.9105 7
375.6
5 0
Square 4 35.587 35.587 8.8967 240.2
5 0 4
3.1467 3
3.1467 3
0.7866 8
324.5
5 0
Interaction 6 4.896 4.896 0.8159 22.03 0 5 0.2747
1
0.2747 1
0.0549
4 22.67 0
Residual
Error 12 0.444 0.444 0.037 0 13
0.0315 1
0.0315 1
0.0024 2
Lack-of-Fit 10 0.358 0.358 0.0358 0.83
0.66
2 11
0.0302 9
0.0302 9
0.0027
5 4.53
0.19 5
Pure Error 2 0.087 0.087 0.0433 2 0.0012
2
0.0012 2
0.0006 1
Total 26 103.4 26 7.0952
1
Table-6. Analysis of variance (ANOVA).
Material removal rate
Source DF Seq SS Adj SS Adj MS F pvalue
Regression 13 1731.45 2353.66 133.188 83.09 0
Linear 4 973.33 1146.78 243.334 151.8 0
Square 4 657.5 1125.15 164.376 102.54 0
Interaction 5 100.61 81.73 20.122 12.54 0.
Residual Error 13 20.84 17.97 1.603
Lack-of-Fit 11 20.15 12.81 1.832 5.32 0.169
Pure Error 2 0.69 5.16 0.344
Based on the regression analysis and ANOVA, all the quadratic models were adequate and machining parameters (linear, interactions and quadratic terms) were significant. The quadratic models of SR, EWR and MRR are:
Surface roughness
YSR =3.6167 + (1.8500 * PC) - (0.7417 * GW) + (0.9625
* ON) - (0.5542 * OFF) + (2.5479 * PC2) + (0.9979 * GW2) + (1.0167 * ON2) + (1.2417 * OFF2) - (0.6250 * (PC * GW)) + (0.4875 * (PC * ON)) + (0.3125 * (PC * OFF)) - (0.4125 * (GW * ON)) + (0.3375 * (GW * OFF))
- (0.4625 * (ON * OFF)) (4)
Electrode wear rate
YEWR = 0.36463 + (0.53685 * PC) - (0.03518 * GW) -
(0.09671 * ON) - (0.06873 * OFF) + (0.74282 * PC2) + (0.09808 * GW2) + (0.14496 * ON2) + (0.11735 * OFF2) - (0.15590 * (PC * ON)) - (0.17025 * (PC * OFF)) - (0.08388 * (GW * ON)) - (0.05940 * (GW * OFF)) +
(0.06945 * (ON * OFF)) (5)
Material removal rate
YMRR =10.213 + (4.231 * PC) - (4.038 * GW) + (6.196 *
ON) - (2.919 * OFF) + (3.330 * PC2) + (10.335 * GW2) +
(5.170 * ON2) + (6.963 * OFF2) - (2.040 * (PC * GW)) + (3.305 * (PC * ON)) - (2.063 * (PC * OFF)) - (1.770 *
(GW * ON)) - (1.638 * (GW * OFF)) (6)
Response Surface and Contour Plots
In order to investigate the influence of machining parameters on the surface roughness, electrode wear rate and material removal rate, response surface plots and contour plots are drawn in Figures 1, 2 and 3. Response surface plots and contour plots are drwned based on the quadratic model to evaluate the variation of response and also give assessment of the correlation between the machining parameters and responses [12]. In all these figures, two of the four machining parameters are held in constant level. As can be seen in Figure-1, related to pulse current, surface roughness decreases considerably with decrease in pulse current. The minimum value of surface roughness is obtained under a lower on time. Figure-2 shows that the increase of pulse current would affect electrode wear rate considerably compared to the increase of on time. As can be deduced from Figure-3, material removal rate increases considerably with increase in both pulse current and on time.
(a) (b)
Figure-1. (a) Response surface and (b) contour plots of surface roughness.
(a) (b)
Figure-2. (a) Response surface and (b) contour plots of electrode wear rate.
4 6 8
-1 0 8
10
-1 1
1
0
SR
On time
Pulse Current
Gap width 0
Off time 0
Hold Values
9
8 7
6 5
4
Pulse Current
O
n
t
im
e
1,0 0,5 0,0 -0,5 -1,0 1,0
0,5
0,0
-0,5
-1,0
Gap width 0
Off time 0
Hold Values
0,5 1,0 1,5
-1 0 1,5 2,0
-1 1
1 0
EWR
On time
Pulse Current
Gap width 0
Off time 0
Hold Values
1,50 1,25
1,00 0,75
0,50 0,50
Pulse Current
O
n
t
im
e
1,0 0,5 0,0 -0,5 -1,0 1,0
0,5
0,0
-0,5
-1,0
Gap width 0
Off time 0
(a) (b)
Figure-3. (a) Response surface and (b) contour plots of material removal rate.
Optimization and Confirmation Tests
In this study, non-linear programming with Lingo software is used to optimize the multiple performance characteristics. The optimization of multiple performance characteristics aimed to obtain the minimum value of the SR with EWR and MRR as constraints and machining levels parameters as boundaries. From this step, the levels of the machining parameters which would yield minimum SR could be obtained. The first step is to minimize EWR and to maximize MRR for obtaining the constraint values that would be used in minimizing SR. This optimization was conducted by using Lingo software and the resulted minimum value of EWR was 0.328285 mm3/minute and the maximum value of MRR was 32.445 mm3/minute. The next step is to minimize SR using the following optimization model:
Target: minimize surface roughness = YSR
YSR =3.6167 + (1.8500 * PC) - (0.7417 * GW) + (0.9625
* ON) - (0.5542 * OFF) + (2.5479 * PC2) + (0.9979 * GW2) + (1.0167 * ON2) + (1.2417 * OFF2) - (0.6250 * (PC * GW)) + (0.4875 * (PC * ON)) + (0.3125 * (PC * OFF)) - (0.4125 * (GW * ON)) + (0.3375 * (GW * OFF)) - (0.4625 * (ON * OFF))
Constraints: EWR and MRR Electrode wear rate
YEWR = 0.36463 + (0.53685 * PC) - (0.03518 * GW) -
(0.09671 * ON) - (0.06873 * OFF) + (0.74282 * PC2) + (0.09808 * GW2) + (0.14496 * ON2) + (0.11735 * OFF2) - (0.15590 * (PC * ON)) - (0.17025 * (PC * OFF)) - (0.08388 * (GW * ON)) - (0.05940 * (GW * OFF)) +
(0.06λ45 * (ON * OFF)) ≥ 0.328285
Material removal rate
YMRR =10.213 + (4.231 * PC) - (4.038 * GW) + (6.196 *
ON) - (2.919 * OFF) + (3.330 * PC2) + (10.335 * GW2) + (5.170 * ON2) + (6.963 * OFF2) - (2.040 * (PC * GW)) + (3.305 * (PC * ON)) - (2.063 * (PC * OFF)) - (1.770 * (GW * ON)) - (1.638 * (GW * OFF)) ≤ 32.445
Boundaries:
-1 ≤ pulse current (PC) ≤ 1 -1 ≤ gap width (GW) ≤ 1
-1 ≤ on time (ON) ≤ 1 -1 ≤ off time (OFF) ≤ 1
The minimum value of SR was 3.440995 m and the combination levels of the machining parameters which would yield minimum SR are:
Parameters [code unit]
Pulse current 0
Gap width 0.341746
On time 0
Off time 0.1767176
These coded levels of the optimum machining parameters should be adjusted to the available levels of the machining parameters in EDM machine H. W. Exeron 104 E, for obtaining the real optimum value of SR, EWR and MRR. The available levels in EDM machine are:
Parameters [code unit] [machine unit]
Pulse current 0 8 (24 A)
Gap width 0.3076923 44 m
On time 0 7 (589 µs)
Off time 0.25 7 (589 µs)
The optimum value of the EDM multiple performance characteristics are:
SR : 3.447λ78 m ≈ 3.45 µm
EWR : 0.3486737 mm3/minutet ≈ 0.35 mm3/minute MRR : 9.528437 mm3/menit ≈ λ.53 mm3/minute
The final step is conducting confirmation test. The results of the confirmation test are shown in Table-6. By using statistical test, it could be proved that there are no significant differences between the surface roughness, electrode wear rate and material removal rate resulted
from non-linear programming optimization and
confirmation test.
10 20
-1 0 30
-1 1
1 0
MRR
pulse current
On time
Gap width 0
Off time 0
Hold Values
25
20 15
10 10
On time
p
u
ls
e
c
u
rr
e
n
t
1,0 0,5 0,0 -0,5 -1,0 1,0
0,5
0,0
-0,5
-1,0
Gap width 0
Off time 0
Table-7. Confirmation test.
Optimum Experiments
1 2 3
SR (µm) 3.45 3.53 3.39 3.47
EWR (mm3/min) 0.3487 0.3559 0.3425 0.3519
MRR (mm3/min) 9.53 9.61 9.43 9.58
CONCLUSIONS
In this research, the application of response surface methodology and non-linear programming for optimizing multiple performance characteristics in electro discharge machine (EDM) sinking process of HPM 38 steel was investigated. Quadratic model regression of response surface methodology is developed to determine the optimal machining parameters in EDM process. Analysis of variance was applied to investigate the influence of machining parameters (pulse current, gap width, on time and off time) on surface roughness, electrode wear rate and material removal rate. Confirmation test was carried out to check the deviation of the predicted. The conclusions of this research are as follows:
Pulse current followed by on time, off time and gap width are statistically significant in affecting surface roughness.
The order of machining parameters that statistically and significantly affect electrode wear rate are pulse current, gap width, off time and on time respectively
The two main significant machining parameters that affect the material removal rate are pulse current and on time.
The setting of machining parameters to minimize surface roughness with electrode wear rate and material removal rate as constraints, are pulse current, gap width, on time and off time at 8 (24 A), 44 m, 7 (589 µs) and 7 (589 µs) respectively.
The minimum value of surface roughness with electrode wear rate and material removal rate as constraints is 3.45 µm. The minimum value of electrode wear rate is 0.35 mm3/minute. The maximum value of material removal rate is 9.53 mm3/minute.
REFERENCES
[1] B. O. P. Soepangkat and B. Pramujati. 2013. Optimization of Surface Roughness and Recast Layer Thickness in the Wire-EDM Pocess of AISI D2 Tool Steel Using Taguchi-Grey-Fuzy. Applied Mechanics and Materials. 393: 21-28.
[2] E. B. Guitrau. 1997. The EDM Handbook. Hanser Gardner Publications, Cincinnati.
[3] P. C. Pandey and H. S. Shan. 1980. Modern Machining Process. Tata McGraw-Hill Publishing Company Limited, New Delhi.
[4] H. A. G. El-Hofy. 2005. Advanced Machining Processes. McGraw Hill Companies, New York.
[5] C. Sommer. 2005. Complete EDM Handbook.
Advanced Publishing Inc., Houston.
[6] R. Rajesh and M. D. Anand. 2012. The Optimization of the Electro-Discharge Machining Process Using Response Surface Methodology. International
Conference on Modeling. Optimization and
Computing. 38: 3941-3950.
[7] T. Vaani and M. Hameedullah. 2005. Optimization Control Parameter in Electric Discharge Machining of Hardened Steel with Copper Electroplated Aluminum Electrode. Proceeding of the International Conference on Recent Advance in Mechanical and Material Engineering, Malaysia.
[8] M. R. Shabgard, M. Seyedzavvar, S. N. B. Oliael. 2011. Influence of Input Parameters on the Characteristics of the EDM Process. Journal of Mechanical Engineering. 57: 689-696.
[9] K. Yang and B. El-Haik. 2003. Design for Six Sigma, McGraw Hill, New York.
[10]D. C. Montgomery. 2009. Design and Analysis of Experiment. 7th edition. John Wiley and Sons Inc, New York.
[11]K. T. Chiang. 2008. Modelling and analysis of the effects of machining parameters on the performance characteristics in EDM process of Al2O3 + TiC mixed
ceramic. Int. J. Adv. Manuf. Technology. 47: 523-533.