International Journal of Performability Engineering Vol. 10, No. 2, March 2014, pp. 155-162. © RAMS Consultants Printed in India
Use of Robust Design Technique in Job Shop Manufacturing: A Case Study of Die-Sinking Electro Discharge Machining R.M. CHANDIMA RATNAYAKE1* and I. VALBO2 1
Department of Mechanical and Structural Engineering and Materials Science, Faculty of Science and Technology, University of Stavanger, N-4036 Stavanger, NORWAY 2 Salte AS, Lalandsvegen 227, N-4353 Klepp Stasjon., NORWAY (Received on April 05, 2013, revised on November 07, 2013)
Abstract: Job shops typically run in small manufacturing businesses handling job production. In general, they move on to different jobs when each job is completed. The nature of the job shop type of manufacturing operation means that it usually requires different skills, expert knowledge, machine settings, materials and processes. In this context, when the die-sinking electro discharge machining (EDM) process has been used for job shops, as it involves several parameters, their settings have to be pre-determined to achieve optimized manufacturing and quality performance. It is possible to accomplish this via a ‘parameter design’ approach suggested in the robust design technique (RDT). The ‘parameter design’ focuses on designing a process to make the performance minimally sensitive to the various causes of variation. This manuscript illustrates the use of RDT in optimizing the performance of die-sinking EDM. It also verifies the reliability of the approach using a verification experiment. Keywords: Job shop, robust design technique, die-sinking EDM, parameter design 1.
Introduction
To date super alloys are used in the following industries: oil and gas (O&G) (offshore and onshore), marine, space, and nuclear. They are also used in other applications due to their strength capacity at high temperatures and high corrosion resistance ability [1]. The high level mechanical performance of these alloys significantly hinders their machinability [2, 3]. For instance, titanium grade 5, which is also known as Ti-6Al-4V, is one of the most commonly used alloys that has been used for a range of applications in the offshore O&G, aerospace, marine and power generation industries. This grade has an excellent combination of strength, corrosion resistance and weldability. In general, titanium is classified as difficult to machine by traditional methods due to its physical, chemical, and mechanical properties [4]. For example, as titanium is a poor conductor of heat, the heat generated by the cutting action does not dissipate quickly; this results in most of the heat being concentrated on the cutting edge and the tool face, leading to premature cutting tool failure. The chemical reactivity with materials at cutting temperatures causes rapid destruction of cutting tools, due to galling, welding and smearing. Moreover, low modulus of elasticity causes chatter, tool rubbing and tolerance-related complications [5]. Hence, it is almost impossible to use conventional techniques and cutting materials for machining titanium. Hence, in industry, the machining of titanium is carried out by wire electrical discharge machining (WEDM) or die-sinking EDM, depending on the client’s requirements.
___________________________________ ∗
Corresponding author’s email:
[email protected]
155
156
R.M. Chandima Ratnayake and I. Valbo
In this context, the selection of optimum machining parameters in the EDM is an important step before starting machining. Current research reveals that improperly selected parameters can significantly degrade manufacturing and quality performance [6, 7]. For instance, material removal rate (MRR), surface roughness (Ra) and die erosion rate (ER) determines production and quality performance [8,9]. Although a particular computer numerically controlled (CNC) EDM machine automatically generates machining parameters (e.g., current, voltage, on-time, off-time, etc.) and output values (MRR, surface finish (SF), ER, etc.), the selection of machining parameters is not fully dependent on machine control. Rather, it is dependent on the material which is going to be machined. The aforementioned circumstances are further exacerbated in the job shops, which are typically referred to as small manufacturing businesses that handle job production. They comprise of tailored or semi-tailored manufacturing processes to cater for small to medium-size customer orders or batch jobs. Generally, job shops move on to different jobs when each job is completed [10]. When EDM processes are utilized in job shops, it is vital to establish a technique selecting machining parameters effectively and efficiently for optimizing manufacturing (e.g., ER and MRR) and quality performance (e.g., SF). This manuscript illustrates the use of the RDT to to tackle/overcome the aforementioned challenges and to establish it as a formal mechanism for ‘parameter design’ for carrying out machining operations effectively and efficiently in the job shops type of manufacturing firms. The technique is illustrated using a case study; a verification experiment has been performed and results are presented. 2.
Background and Industrial Challenge
The small manufacturing firm in the case study carries out machining on a job shop basis. The machining operations are usually performed with super alloys to make expensive items required for the offshore O&G industry. Conventionally, copper has been used as electrode material for EDM. However, when the copper electrode used for die-sinking EDM (with die-sinking EDM machine Agie_FORM_20) of titanium grade 5, more erosion of the electrode than the work piece is observed (i.e., even after changing the polarity). Once the problem was reported to the electrode material manufacturing experts, the recommendation was to utilize a copper-tungsten electrode (i.e., Cu25% and W75%). The operational manual of the Agie_FORM_20 die-sinking EDM machine suggests providing the data about electrode material, work piece material, surface area of the electrode, depth of the erosion in the work piece, and the expected surface roughness of the final product. Then, the machine control system estimates the values for the input parameters and values for the output parameters. An experiment has been performed with the copper-tungsten electrode of titanium grade 5. The results revealed that actual output values indicate a significant difference, when they were compared to the control system of Agie_FORM_20 die-sinking EDM generated input parameters and estimated output values (see Table 1).
Use of Robust Design Technique in Job Shop Manufacturing: A Case Study of Die-Sinking Electro Discharge Machining
157
Table 1: Automatically Generated Values from Aige_Form_20 and Observed Actual Values Agie_FORM_20 control system generated values Input parameter Current
Value
Observed actual values
Output Value parameter 39 A Machining 44.6 min 102.5 min time Voltage 250 V Electrode 0.0103 g 0.1963 g erosion On-time 11.5 μs Workpiece 2.1697 g 2.0812 g erosion Off-time 154 μs *Surface *4.0 μm 4.87 μm finish, Ra *Surface finish has to be specified by the machine operator
The time taken to carry out the machining of 1mm depth was quite high with the diesinking EDM Agie_FORM_20 control system generated values. The MRR, electrode ER and surface finish (selected and actually observed) are summarized in Table 2. Table 2: MRR, Electrode Erosion and Surface Finish (selected and actually observed)
Output parameter
Die-sinking EDM Agie_FORM_20 control system generated values
Observed actual values
Electrode ER 0.0002309 0.001915 (g/min) Workpiece ER 0.0486479 0.020304 (MRR) (g/min) * SF (μm), Ra *4 4.87 *Surface finish has to be specified by the machine operator
The experiment (see Table 1) reveals that machining time (102.5 > 44.6 min) and electrode erosion rate (0.001915 > 0.0002309 g/min) are very high compared to the machine control system calculated values [see columns (4) and (5)]. However, the workpiece erosion rate (0.020304 < 0.0486479 g/min) and surface finish (4.87 > 4.0 μm) were poorer than the control system estimated values. In order to overcome the aforementioned challenge, it is necessary to carry out ‘parameter design’ for the particular machine and material combination (i.e., recognizing parameter settings for the machining operation based on the selected material). The RDT provides a mechanism to establish a simple methodology to carry out ‘parameter design’ and alternatively to optimize manufacturing (e.g., MRR and electrode ER) and quality (e.g., SF) requirements. 3. Methodology The RDT is structured to recognize the optimum combinations of the input parameters, based on the statistical results generated from test matrix experiments [11, 12]. The standard orthogonal array L 9 (see Table 3) suggested in RDT has been selected to perform the matrix experiment. The manufacturing and quality performance measures have been selected as ER of the die (i.e., electrode ER), MRR (i.e., work piece ER) and SF respectively [1, 12].
158
R.M. Chandima Ratnayake and I. Valbo
Table 3: The Standard Orthogonal Array L 9 Exp. No. 1 2 3 4 5 6 7 8 9
Current level (A) 1 1 1 2 2 2 3 3 3
Voltage level (B) 1 2 3 1 2 3 1 2 3
On-time level (C) 1 2 3 2 3 1 3 1 2
Off-time level (D) 1 2 3 3 1 2 2 3 1
As suggested in the RDT, the-smaller-the-better case has been selected for SF and ER. The given in (1) has been utilized for calculating signal to noise ratio [12]. The mean values of the parameter values of each experiment are transformed into S/N ratio values (η i ; i = 1, 2...9). 1 2 𝑆/𝑁 𝑟𝑎𝑡𝑖𝑜 (𝜂𝑖 ) = −10 𝑙𝑜𝑔10 �� � ∑𝑛𝑖=1 𝑦𝑖𝑗 � (1) 𝑛 where, n = number of replications , and y ij = performance indicator value; (i = 1,2…n and j = 1,2…9). As suggested in the RDT, the-larger-the-better-case has been selected for MRR. The relationship given in (2) has been utilized for calculating signal to noise ratio [12]. The mean values of the parameter values of each experiment are transformed into S/N ratio values (η i ; i = 1, 2...9). 1
𝑆/𝑁 𝑟𝑎𝑡𝑖𝑜 (𝜂𝑖 ) = −10 𝑙𝑜𝑔10 �� � ∑𝑛𝑖=1 𝑛
1
2 𝑦𝑖𝑗
�
(2)
where, n = number of replications, and y ij = performance indicator value; (i = 1,2…n and j = 1,2…9). The relationship in (3) calculates the signal to noise ratio under the optimum levels of parameter design (ηopt) [12]. (3) S/N ratio (η ) = α + α iA + α Bj + α Ck + α lD + ε ijklmn where, α = overall mean S/N ratio over all the possible combinations, i,j,k,l = particular levels of each of the factors which were selected (so in this model i,j,k,and l must all take on one of the values 1, 2 or 3) 𝛼𝑖𝐴 = deviation from α caused by setting factor A at level i (similarly, other terms can be defined) 𝜀𝑖𝑗𝑘𝑙 = error term. Each experiment was performed with three replicates (i.e., thee electrodes and three workpieces) in order to mitigate the bias. Then, using (3), signal to ratio at optimum settings was calculated. The relationship in (4) has been utilized to calculate corresponding output parameter values [12]. −η opt
y = 10
10
(4) A verification experiment has been carried out using recognized optimum settings, and results were compared to verify the validity and establish the RDT methodology for recognizing machining parameter values to meet optimum manufacturing and quality requirements in future job shops
Use of Robust Design Technique in Job Shop Manufacturing: A Case Study of Die-Sinking Electro Discharge Machining
159
4. Results Twenty-seven experiments (i.e., 9*3 replicates) were performed with 27 copper tungsten electrodes and 27 tungsten grade-5 work pieces. The results were recorded and the necessary calculations were performed. 4.1 Results of the Matrix Experiment Table 4 illustrates levels of input factors and S/N ratios of outputs (i.e., MRR, SF and ER) for each of nine experiments. The factor levels were recognized based on experts’ knowledge. Table 4: Levels of Input Factors and S/N Ratios of Outputs (i.e., MRR, SF and ER) Type Size (pts)
A (Amp) 29
Input parameter levels B C D (Volt) (μs) (μs) 180 6.5 133.4
S/N ratio (dB) SF
MRR
ER
-35.8504036
109.082651
53.5485526
29 29
220 250
11.5 31.6
154.0 177.8
-33.9681927 -48.1823334
108.306538 108.091708
48.5125909 50.3577989
39 39
180 220
11.5 31.6
177.8 133.4
-32.4324146 -41.5331586
106.194957 107.083278
48.6386954 44.7012090
6
39
250
6.5
154.0
-34.8838750
108.323322
51.8398729
7
52
180
31.6
154.0
-30.4416609
103.983897
41.8310365
8 9
52
220
6.5
177.8
-35.0081256
106,801621
51.4024483
52
250
11.5
133.4
-27.6439430
105.763473
44.2884385
1 2 3 4 5
Plot of factor effects for MRR, SF and ER are presented in Figure 1, Figure 2 and Figure 3 respectively. -31
-31
-31
-31
-33
-33
-33
-33
-35
-35
-35
-35
-37
-37
-37
-37
-39
-39
-39
-39
-41
-41 A1
A2
A3
-41 B1
B2
B3
-41 C1
C2
C3
D1 D2 D3
Overall mean MRR Figure 1: Plots of Factor Effects: MRR (Factor Levels at the best MRR are A3-B1-C2-D2)
160
R.M. Chandima Ratnayake and I. Valbo
53 52 51 50 49 48 47 46 45
53 52 51 50 49 48 47 46 45
53 52 51 50 49 48 47 46 45 B1
A1 A2 A3
B2
B3
C1 C2 C3
53 52 51 50 49 48 47 46 45 D1 D2 D3
Overall mean ER Figure 2: Plots of Factor Effects: ER (Factor Levels at the minimum ER are A1-B3C1-D3) 109
109
109
109
108
108
108
108
107
107
107
107
106
106
106
106
105
105 A1 A2 A3
105 B1 B2 B3
105 C1 C2 C3
D1 D2 D3
Overall mean SF Figure 3: Plots of Factor Effects: SF (Factor Levels at the best SF are A1-B2-C1-D1)
Best settings and the corresponding theoretically calculated optimum values [using (3)] are summarized in Table 5. Table 5: Theoretically calculated optimum values Output parameter
MRR (g/min) SF (μm) ER (g/min)
Best 𝜂𝑜𝑝𝑡 (dB) parameter [Equation (3)] levels combination A3-B1-C2-D2 -26.830081 A1-B2-C1-D1 109.492669 A1-B3-C1-D3 54.723201
Theoretically calculated optimum value [Equation (4)] 0.0455000 3.1407600 0.0018358
Using the optimum values of the output parameter levels estimated (i.e., using the designed parameters) by the matrix experiment, a verification experiment was performed. This is to verify whether the theoretically calculated output parameter values (see column (4) of Table 5) can be obtained when the machining operation is repeated with optimum values of the control factor levels. 4.2 Results of the Verification Experiment A verification experiment was performed to verify the theoretically calculated output values. The results are summarized in Table 6.
Use of Robust Design Technique in Job Shop Manufacturing: A Case Study of Die-Sinking Electro Discharge Machining
161
Table 6: Results of Verification Experiment Output parameter MRR (g/min) SF (μm) ER (g/min)
Best parameter levels combination A3-B1-C2-D2 A1-B2-C1-D1 A1-B3-C1-D3
Observed values 0.041156 3.18 0.00152526
5. Discussion The summary of the estimated values of the Agie_FORM_20 die-sinking EDM, theoretically calculated optimum values [i.e., using (4)] and actually observed values from the verification experiment are summarized in Table 7. Table 7: Results of Verification Experiment Output parameter
Estimated values of the Agie_FORM_20 die sinking EDM
Theoretically calculated optimum value [Equation (4)]
MRR (g/min) SF (μm)
0.048647 *4
0.0455079 3.1407600
ER (g/min)
Observed values from the verification experiment 0.041156 3.18
0.000231 0.0018358 0.0015252 *Surface finish has been specified by the machine operator
Table 7 reveals that although the Agie_FORM_20 die-sinking EDM control system’s estimated value of electrode ER (0.000231) has not been achieved, the actual vale of the electrode ER observed from the verification experiment has been reduced significantly after the ‘parameter design’. On the other hand, MRR and SF have been significantly improved after the ‘parameter design’ compared to the Agie_FORM_20 diesinking EDM control system’s estimated values. 4.
Conclusion
The verification experiment revealed that the ‘parameter design’ methodology suggested in the manuscript provided significantly improved values than the estimated values of the Agie_FORM_20 die-sinking EDM. Hence, this approach can be utilized as a mechanism for ‘parameter design’ and alternatively improve manufacturing and quality demands. Inherently, job shops need to move on to different jobs when each job is completed. Hence, the mechanism (i.e., ‘parameter design’ using RDT) suggested in this manuscript provides the means to cater for this when different machines and materials have to be utilized frequently, depending on the clients’ requirements; i.e., the mechanism suggested a simple approach to recognize machine settings depending on a particular job’s requirement (i.e., based on the selected material and machine). The reliability of the approach has been verified with a verification experiment. Hence, the suggested approach can be adapted for future job-shop manufacturing requirements. Future research should be carried out to establish a formal mechanism to recognize machine factor levels using the experts’ knowledge. Acknowledgment: The authors convey their sincere thanks to Laila Salte Gausel, Rolf Undheim and Oddvar Salte in the Salte AS (http://salteas.no), Norway, for the extended support in accomplishing this research work.
162
R.M. Chandima Ratnayake and I. Valbo
References [1]
[2]
[3] [4] [5] [6]
[7] [8]
[9]
[10]
[11] [12]
Muthu Kumar, V., A. Suresh Babu, R. Venkatasamy, and M. Raajenthiren. Optimization of the WEDM Parameters on Machining Incoloy800 Super Alloy with Multiple Quality Characteristics. International Journal of Engineering Science and Technology, 2010; 2(6): 1538-1547. Ezugwu, E. O. Key Improvements in the Machining of Difficult-to-Cut Aerospace Superalloys. International Journal of Machine Tools and Manufacture, 2005; 45 (12-13): 1353–1367. Choudhury, I. A., and M. A. El-Baradie. Machinability of Nckel-base Super Alloys: A General Review. Journal of Materials Processing Technology, 1998; 77 (1-2): 278–284. Azom. Titanium Alloys - Ti6Al4V Grade 5. 2002. http://www.azom.com/article.aspx? ArticleID=1547 (accessed on 27.02.2013). Supraalloys. Technical Information: Characteristics Influencing Machinability. 2013. http://www.supraalloys.com/ technical-machine.php. Secilmis, A., A. M. Olmez, M. Dilmec, H. S. Halkaci, and O. Inan. Determination of Optimal EDM Machining Parameters for Machined Pure Titanium-Porcelain Adhesion. International Journal of Advanced Manufaturing Technoogy, 2009; 45 (1-2): 55-61. Joshi, S. N., and S. S. Pande. Intelligent Process Modeling and Optimization of Die-Sinking Electric Discharge Machining. Applied Soft Computing, 2011; 11 (2): 2743-2755,. Patel, B. A., and D. S. Patel. A Review: Influence of Electrode Geometry and Process Parameters on Surface Quality and MRR in EDM using Artificial Neural Network. International Journal of Engineering Research and Applications, 2013; 3 (1); 1645-1654. Bharti, P. S., S. Maheshwari, and C. Sharma. Multi-Objective Optimization of Die-Sinking Electric Discharge Machining. Applied Mechanics and Materials, 2011. 110-116: 18171824. Chen, L.H., and Y.H. Chen. A Design Procedure for a Robust Job Shop Manufacturing System under a Constraint using Computer Simulation Experiments. Computers & Industrial Engineering, 1996; 30 (1): 1-12. Daniel, C. Applications of Statistics to Industrial Experimentation. John Wiley and Sons, New York, 1976. Phadke, M. S. Quality Engineering using Robust Design. Englewood Cliffs, New Jersey: PTR Prentice-Hall, Inc., 1989.
R.M. Chandima Ratnayake is an Associate Professor in Mechanical Engineering in the University of Stavanger and maintenance specialist in the Applysorco AS, Norway. He received a Ph.D. degree in Offshore Engineering from University of Stavanger – Norway and M.Sc. degree in Manufacturing Engineering and B.Sc. degree in Mechanical (specialized in Production) Engineering from the University of Peradeniya - Sri Lanka. He served as a Senior Maintenance Engineer in the AkerSolutions Offshore Partner from 2009 to 2011 in Stavanger, Norway. He also served as a visiting Associate Professor and Assistant Professor from August 2007 to July 2010 in the University of Stavanger, Norway. Dr. Ratnayake has published number of book chapters as well as journal and conference papers in international journals and conferences. I. Valbo is a Mechanical Engineer working at the Salte AS, Norway. She graduated from Department of Mechanical and Structural Engineering and Materials Science, University of Stavanger, Norway.