Locking Compression Plates (LCP) are typically made of Ti6Al4V alloy. LCP is used to heal fractures of the bone. This article uses Wire Cut Electric Discharge Machining (WEDM) to machine Ti6Al4V alloy. Responses are optimized by two metaheuristic algorithms, the cuckoo search algorithm (CS) and particle swarm optimization (PSO). The cuckoo search algorithm shows that the maximum MRR is 59.4312 mg/min & the maximum cutting speed is 5.0625 mm/min. The single-objective PSO shows a maximum MRR of 59.3632 mg/min & the maximum cutting speed is 5.0155 mm/min. The GRA revealed that the optimum weight for MRR and cutting speed is 48.868% and 51.132% respectively. Scanning electron microscope analysis indicates proper biocompatibility after machining. This article compares CS and PSO with similar calculations for the t-test and found that the p-values for MRR and Cutting speed are 0.9432 and 0.9919.
Keywords: Optimization; manufacturing; biocompatibility; Ti6Al4V alloy; machining
Ti6Al4 V has outstanding qualities like high mechanical potency, corrosion limit, and a strength-to-mass ratio.[[
However, along with some advantages of Ti alloy, it also has disadvantages like advanced chemical interaction, poor thermal conductivity, and lesser modulus of elasticity, which makes the material difficult to cut and therefore limits its use.[[
Wire Cut EDM is unconventional machining.[[
WEDM also has advantages, such as non-contact machining. Thus, the difficulty of Ti alloy machining (due to the low elastic modulus and its mechanical and physical properties) is overcome using WEDM.[[
Optimization is nothing but maximizing desired responses and minimizing unwanted responses. Optimization is done after modeling the responses. While in engineering, these responses are multimodal, non-linear, and complex constraints in nature. The real-life optimization problems cannot be solved by local search algorithms or traditional optimization techniques like hill climbing or the Nelder-Mead downhill simplex method.[[
An LCP is a metal implant used for internal fixation fractures where the bone can't be repaired by an exterior method like casting. LCP consists of a combination of conventional screw holes with a locking screw hole. Before the development of LCP, a dynamic compression plate (DCP) existed, which was used as conventional plating. In DCP, the bone's fixation occurs through friction amidst the bone and the plate. The DCP plate contacts the bone and can cause disturbances in the blood supply. The limited contact type DCP was then developed to limit interaction with the bone and preserve the blood supply. But the disadvantage of LC-DCP is that the screws are held only vertically, whereas the screws are at different angles to a conventional plate.[[
This paper analyzes the effect of WEDM process parameters on Ti6Al4 V alloy by SEM images and MRR and Cutting Speed analysis. The manufacturing aspect of LCP is understood using two optimization methods (CS & PSO).
MD Al-Amin et al. explored the potential of powder mixed EDM (PMEDM) on biomaterials (like Mg-based alloys, Ti-based alloys, Co-Cr-Mo-based alloys, etc.). They found that the PMEDM coating improves the biocompatible surface and mechanical properties.[[
Ghate et al.[[
Devarasiddappa and Chandrasekaran employed preference-based TLBO on Ti6Al4 V alloy during WEDM. They found that the optimization improves the MRR by 40.51% and reduces the SR by 15.66% when comparing the highest preference and equal weight.[[
Chakala et al.[[
Phate et al. applied an adaptive neuro-fuzzy inference system (ANFIS) and response surface methodology (RSM) for modeling SR during WEDM on a newly fabricated 5% graphite-reinforced aluminum base composite.[[
Dey and Pandey selected a gray-based hybrid method to optimize surface roughness, kerf width, and cutting speed when performing WEDM on Al matrix composites. They found an improvement of 3.243% for kerf width and 7.053% for surface roughness.[[
Kumar et al.[[
Bharti optimized electric discharge machining through Taguchi's orthogonal array and then trained the data using a neural network-based approach. After obtaining 24 Pareto optimal solutions, they are ranked by the TOPSIS method.[[
Raj and Prabhu[[
Kumar and Raj analyzed the rake face texture of a tool manufactured by WEDM while turning AISI4340 steel. The turnings are then compared by the textured tool and the untextured tool. They have found that the textured tool is superior to the untextured tool because of flank wear, cutting zone temperature, chip trace area, cutting force, and surface finishes.[[
Singh et al. analyzed the improvement of EDM by powder mixed EDM with optimization. They obtained that MRR, TWR, and poor surface roughness could be improved using PMEDM compared to conventional EDM.[[
Sen et al.[[
Sharma et al[[
Sharma et al[[
The concept of the LCP plate evolved from the Dynamic Compression Plate (DCP). The AO Group developed the DCP plate during the 1960s.[[
ASTM std F-382 is the test method of the metal bone plate. While the following are the specifications for surgical implants of different materials, such as those for alloyed Ti, Ti6Al4 V extra low interstitial, and wrought, Ti-6Al-7Nb are F-67, F-136, and F-1295 respectively.[[
All surfaces of the LCP bone plate are machined by CNC milling, while the sixth surface, which comes into contact with the bone, is machined by EDM. Therefore, in this study, the LCP profiles have been cut by WEDM to see the manufacturing aspect of WEDM for LCP plates. The CAD model of the LCP is shown in Fig. 1.
Graph: Figure 1. CAD model of LCP plate.
MRR is the response that shows how much material is eliminated from the sample per unit period by WEDM. Measurement of MRR as a response is done by measuring the weight of the sample formerly and after machining. At the same time, time is computed with the help of a stopwatch. Therefore, MRR is measured by the following equation 1:
(
Graph
Here,
Graph
Graph
The cutting speed is the speed that shows the length of the sample to be cut per unit of time. It points directly at the CNC WEDM monitor. The machining setup for the WEDM is shown in Fig. 2:
PHOTO (COLOR): Figure 2. WEDM set up.
Therefore, MRR and Cutting Speed have been chosen as the responses in this study. The LCP manufacturing aspect is observed in these responses.
LCP comes under Biomedical Implant. These biomedical implants are often made from Ti6Al4 V alloy. The properties required for a biomedical implant are surface roughness, biocompatibility, bifunctionality, and corrosion resistance. However, the Ti6Al4 V alloy meets a biomedical implant's properties.[[
Different Ti alloys are now available based on the microstructure at room temperature. Ti alloys have alpha phase Ti, beta phase Ti, and alpha+beta phase Ti. Whereas in the 1970s, ELI Ti6Al4 V alloy was formed. It was suitable for dental implants. But after some time, it was found that Ti6Al4 V has some toxic effects, but it is not entirely proven that this toxicity is due to vanadium (V). Then this V in Ti6Al4 V alloy was changed using Nb. Hence Ti6Al7Nb has been developed for biomedical use.[[
Titanium is the fourth most ample material on the Earth's crust. But this is expensive because of its machining and extraction difficulty. Titanium-based minerals were first found in the late 18th century. After about 120 years, its extraction process changed significantly. Previously the Kroll process was used to obtain Ti metal by thermal reduction of TiCl4. Nowadays, there is more mass automation and continuous development. In the 21st century, Frey et al. developed an electrolytic process, but its stability and efficiency are still unavoidable. Therefore, self-propagating high-temperature synthesis (SHS) is a good option for Ti alloy extraction. However, even then, the widespread use of Ti alloys requires a cheap extraction process.[[
The Box Behnken Design (BBD) was applied in this study to design the experimental runs as the Response Surface Method (RSM). The BBD has a central point and other points at the center of the sides. This BBD is an almost rotatable design. The BBDs for the three factors and the three levels are shown in Fig. 3. This paper selects three levels of three parameters and two responses to design the experimental run.[[
(
Graph
PHOTO (COLOR): Figure 3. Box-Behnken Design for three process parameters.
Here Y, is the responses. X
Graph
Graph
Graph
This article selects Ton, SV, and WT as process parameters. Cutting Speed & MRR are chosen as two responses. The process parameters' levels are shown in Table 1. The experimental chart for MRR and Cutting Speed is exposed in Table 2 below.
Table 1. Coded level of Process Parameters.
Coded Value Factor Name Units Type Sub Type Low (−1) Mean (0) High (+1) A Pulse ON Time Micro Seconds Numeric Continuous 9 11 13 B Servo Voltage Volts Numeric Continuous 20 30 40 C Wire Tension Gram Numeric Continuous 1000 1600 2200
Table 2. Experimental Chart for MRR & CS.
Std. A: Pulse on Time B: Servo Voltage C: Wire Tension Material Removal Rate Cutting Speed GRC MRR GRC CS Micro Seconds Volts Gram mili gram/min mm/min 1 9 20 1600 15.76 2.48 0.333333 0.441077 2 13 20 1600 38.4 1.04 0.643004 0.333333 3 9 40 1600 16.26 2.39 0.336917 0.432343 4 13 40 1600 38.04 1.08 0.633643 0.335611 5 9 30 1000 25.4 4.85 0.41932 0.942446 6 13 30 1000 42.78 3.41 0.783892 0.557447 7 9 30 2200 29.74 4.97 0.474417 1 8 13 30 2200 47.1 3.77 1 0.620853 9 11 20 1000 19.46 1.58 0.36181 0.366947 10 11 40 1000 20.16 1.7 0.367754 0.375358 11 11 20 2200 27.38 2.01 0.44278 0.398985 12 11 40 2200 22.46 2.09 0.388737 0.405573 13 11 30 1600 36.24 3.12 0.590652 0.515072 14 11 30 1600 28.16 3.23 0.452759 0.530364 15 11 30 1600 34.26 2.84 0.549632 0.479853 16 11 30 1600 34.62 2.9 0.556661 0.486989 17 11 30 1600 28.26 2.88 0.454071 0.484587
In this paper, Cutting Speed and MRR are taken as responses. For manufacturing purposes, both Cutting Speed and MRR must be maximized together. In single objective optimization, only one response must be optimized. But in multi-objective optimization, both the responses are optimized by assigning different weights to these responses. The weights for these responses are given by the gray relational analysis method.[[
(
Graph
Here, 1/MRR and 1/Cutting Speed are taken as single objective functions, while MRR
Whereas w1 and w2 are gray relational weights for MRR and Cutting Speed, respectively.[[
Table 3. Weight calculation for objectives by GRA.
MRR (milligram/min.) Cutting Speed (mm/min.) level TON (micro seconds) Vs (volts) Wire Tension (gram) level TON (micro seconds) Vs (volts) Wire Tension (gram) −1 0.39099 0.44523 0.4831 −1 0.703966 0.385 0.5605 0 0.46276 0.58682 0.5056 0 0.449303 0.624 0.4488 1 0.76513 0.43176 0.5764 1 0.461810 0.387 0.6063 R 0.37413 0.1550 0.0932 R 0.25466 0.239 0.1575 ∑R 0.6224 ∑R 0.6513 Weight 0.48868 Weight 0.51132
In the cuckoo search algorithm, the concept of cuckoo breeding behavior is used to optimize the responses. However, cuckoos are known for their captivating voice. The cuckoo replaces its eggs with eggs of some other species. First, the cuckoos lay their eggs in another species of bird's nest (See Fig. 4). The cuckoo then removes the host egg. The time of laying the cuckoo is also perfect. After that, when the cuckoo's chick explodes. Then it removes the host eggs and tries to imitate the host chicks to get more food. After that, if the host bird knows about the cuckoo, they throw the cuckoo's eggs or build a new nest at other locations using levy flight.[[
PHOTO (COLOR): Figure 4. Cuckoo behavior & the nest.
The pseudocode and flow chart for the cuckoo search algorithm is shown in Figs. 5 and 6.
Graph: Figure 5. The pseudocode for the CS.
Graph: Figure 6. The cuckoo search algorithm flowchart.
PSO is a swarm intelligence technology. Kennedy and Eberhart established it. In this algorithm, a set of foraging behaviors of birds are used to optimize responses (see Fig. 7). In this, the position and velocity of each unit are first defined by Equations 4 and 5. Then the best particle solution and global solution are obtained. The velocity of each bird varies as the number of iterations increases. pbest and gbest get updated as iterations go on. The weight of velocity is updated according to equation 6.
PHOTO (COLOR): Figure 7. Particle behavior in PSO.
When the particle weight for velocity is greater, the algorithm is explored. In contrast, the weight of the velocity is less. So, it converges. The position and velocity of a particle are revised by the following equations 7 & 8.[[
(
Graph
(
Graph
(
Graph
(
Graph
(
Graph
Here,
Y (k, l) is the position of the l
The PSO's pseudocode and flow chart are shown in Figs. 8 and 9.
Graph: Figure 8. The particle swarm optimization pseudocode.
Graph: Figure 9. The flowchart for the Particle Swarm Optimization.
MRR was selected as the response in this paper. The quadratic model is best suited for MRR with an R
(
Graph
Table 4. Analysis of Variance for MRR.
Source Sum of Squares df Mean Square F-value p-value 1285.81 9 142.87 13.77 0.0011 significant A-Pulse ON Time 783.29 1 783.29 75.52 <0.0001 B-Servo Voltage 2.08 1 2.08 0.2006 0.6678 C-Wire Tension 44.56 1 44.56 4.30 0.0769 AB 0.1849 1 0.1849 0.0178 0.8975 AC 0.0001 1 0.0001 9.642E-06 0.9976 BC 7.90 1 7.90 0.7613 0.4118 A2 79.62 1 79.62 7.68 0.0277 B2 383.33 1 383.33 36.96 0.0005 C2 0.6787 1 0.6787 0.0654 0.8054 72.60 7 10.37 Lack of Fit 14.39 3 4.80 0.3297 0.8056 not significant Pure Error 58.21 4 14.55 1358.42 16
The normal plot of residuals indicates that the regression model is the best fit (Fig. 10). The residual versus predicted plot shows no pattern across experiments (Fig. 11). Therefore, no transformation is required for the regression model of MRR (Fig. 12).
Graph: Figure 10. MRR Residuals vs Normal Plot.
Graph: Figure 11. MRR Residuals vs Predicted plot.
Graph: Figure 12. MRR Box-Cox plot.
The perturbation plot (Fig. 13) for MRR shows that with a rise in Ton, MRR increases. Because of greater discharge energy at elevated Ton, higher molten material is removed, hence higher MRR. Whereas at some optimal value of servo voltage, MRR rises. Because of good flushing at optimum servo voltage, MRR is high. In comparison, wire tension has a very negligible effect on MRR.
Graph: Figure 13. MRR Perturbation plot.
Cutting Speed is a response to WEDM. It indicates how much length is cut by WEDM in a unit of time. ANOVA shows that the quadratic model is finest suited for Cutting Speed. The difference between Adj R
(
Graph
Table 5. Fit statistics for Cutting Speed.
Std. Dev. 0.1419 R2 0.9932 Predicted R2 0.9726 Adjusted R2 0.9844 C.V. % 5.20 Adeq Precision 36.7953 Mean 2.73
Table 6. Analysis of Variance for Cutting Speed.
Source Sum of Squares df Mean Square F-value p-value 20.45 9 2.27 112.88 <0.0001 significant A-Pulse ON Time 3.63 1 3.63 180.42 <0.0001 B-Servo Voltage 0.0028 1 0.0028 0.1397 0.7196 C-Wire Tension 0.2112 1 0.2112 10.50 0.0143 AB 0.0042 1 0.0042 0.2099 0.6607 BC 0.0004 1 0.0004 0.0199 0.8919 AC 0.0144 1 0.0144 0.7154 0.4256 A2 1.41 1 1.41 70.19 <0.0001 B2 14.04 1 14.04 697.30 <0.0001 C2 1.93 1 1.93 95.81 <0.0001 0.1409 7 0.0201 Lack of Fit 0.0238 3 0.0079 0.2707 0.8443 not significant Pure Error 0.1171 4 0.0293 20.59 16
From one factor at a time, it is concluded that the Cutting Speed decreases marginally as the pulse on time increases (Fig. 14 (a), (b) & (c)). Because more energy is discharged at higher Ton and more molten material is handled, the Cutting Speed is reduced due to higher Kerf Width. Whereas the Cutting Speed increases at the optimum value of V, there is a reasonably good discharge of the molten material. When the tension of the wire increases, Cutting Speed decreases. But when the tension of the wire is further increased, the Cutting Speed increases significantly due to the less loosening of the wire. The perturbation plot (Fig. 15) and 3D plot for Cutting Speed (Fig. 16(a) & (b)) are shown below:
Graph: Figure 14. Cutting Speed One factor plot with (a) Pulse on Time (b) Servo Voltage (c) Wire Tension.
Graph: Figure 15. Cutting Speed Perturbation Plot.
Graph: Figure 16. Cutting Speed surface plot for (a) Ton vs SV, (b) Ton vs WT.
Surface analysis of Ti6Al4 V alloy before WEDM and after WEDM is described in this section. The scanning electron microscope (SEM) is a high-resolution field emission scanning electron microscope (HR-FESEM).[[
Graph: Figure 17. SEM image of Ti6Al4 V alloy for (a) without machining and (b) after WEDM at maximum MRR optimized condition.
The surface morphology of WEDM machined Ti alloy is investigated by analyzing craters, voids, debris, small cracks, and globules. Whereas these surface characteristics can be changed by various machining input parameters such as tonnage, Vgap and wire tension. It is found that the EDM machined surface morphology is much worse than that without EDM machining. This is achieved at high Ton, moderate interval voltage, and high wire tension. Here a rougher biocompatible surface is obtained at 13 microsecond ton, 30 volt voltage, and 2200 g tension.
The SEM result is shown in Fig. 17(a). It shows that without machining conditions, the surface is smoother. Whereas Fig. 17(b) shows some cracks and debris vacancies, which are used to increase the biocompatibility. As it is the fact that a relatively rough surface is helped to improve biocompatibility. So, it is determined that WEDM enhances the biocompatibility of Ti alloy.
This article compares the CS and PSO on preference-based multi-objective optimization. This PBMOO is used to optimize multi-responses like MRR and Cutting Speed.[[
Here, (In equation 3) preference factors (w1) & (w2) are considered in the range of 0 to 1 in the step of 0.05, as w1+w2 = 1. The composite function was analyzed. A modern PC with a 1.6 GHz CPU, 64-bit OS, & 8 GB RAM was used for the analysis. For the CS and PSO, the tuning parameters are shown in Table 7 below.
Table 7. Tuning Parameters for CS & PSO.
Cuckoo Search Algorithm Particle Swarm Optimization Population 20 Population 20 Iteration 1000 Iteration 1000 Discovery Rate 0.25 Learning Factor Phi01 0.5 Phi02 0.5 Levy Exponent 3/2 Weightage of velocity 1 to 0.3
The optimization algorithm was executed ten times at each level of the weight combination, and the desired optimal (max) values are assumed to be the best value for that weight. Here the individual results for optimal MRR and optimal Cutting Speed are shown in the table below (Tables 8 & 9). 19 combinations of step weights, two varieties of the single objective, and GRA method-derived weights are used to optimize the responses.
Table 8. Optimum Results of PBMOO-CSA.
S.No. W1 (MRR) W2 (Cutting Speed) Parametric Settings Optimum MRR (mg/min) Optimum Cutting Speed (mm/min) Minimum Fitness value 1. 0.48868 0.51132 13 µ sec, 29.876 V, 2200 g 59.3377 3.7980 1.1710 2. 0 1 9 µ sec, 29.935 V, 2200 g 39.5428 5.0625 5.0625 3. 1 0 13 µ sec, 28.88 V, 2200 g 59.4312 3.7715 59.4312 4. 0.05 0.95 9 µ sec, 29.900 V, 2200 g 39.3047 5.0059 1.0273 5. 0.10 0.90 9 µ sec, 29.854 V, 2200 g 39.5304 5.0245 1.0486 6. 0.15 0.85 9 µ sec, 29.8132 V, 2200 g 39.4330 5.0151 1.0759 7. 0.20 0.80 9 µ sec, 29.772 V, 2200 g 39.5687 5.0186 1.0996 8. 0.25 0.75 9 µ sec, 29.731 V, 2200 g 39.5108 5.0200 1.1249 9. 0.30 0.70 9 µ sec, 29.6903 V, 2200 g 39.5839 5.0198 1.1493 10. 0.35 0.65 9 µ sec, 29.649 V, 2200 g 39.5364 5.0194 1.1750 11. 0.40 0.60 13 µ sec, 29.937 V, 2200 g 59.1344 3.7853 1.1965 12. 0.45 0.55 13 µ sec, 29.904 V, 2200 g 59.3540 3.7874 1.1784 13. 0.50 0.50 13 µ sec, 29.867 V, 2200 g 59.3839 3.7675 1.1655 14. 0.55 0.45 13 µ sec, 29.825 V, 2200 g 59.4003 3.7852 1.1459 15. 0.60 0.40 13 µ sec, 29.777 V, 2200 g 59.3500 3.7877 1.1298 16. 0.65 0.35 13 µ sec, 29.7235 V, 2200 g 59.2977 3.7905 1.1138 17. 0.70 0.30 13 µ sec, 29.660 V, 2200 g 59.3080 3.7872 1.0980 18. 0.75 0.25 13 µ sec, 29.587 V, 2200 g 59.2630 3.7897 1.0821 19 0.80 0.20 13 µ sec, 29.499 V, 2200 g 59.1576 3.7913 1.0674 20. 0.85 0.15 13 µ sec, 29.394 V, 2200 g 59.3106 3.7972 1.0489 21. 0.90 0.10 13 µ sec, 29.264 V, 2200 g 59.3682 3.7667 1.0331 22. 0.95 0.05 13 µ sec, 29.100 V, 2200 g 59.2816 3.7947 1.0174
Table 9. Optimum Results of PBMOO-PSO.
S.No. W1 (MRR) W2 (Cutting Speed) Parametric Settings Optimum MRR (mg/min) Optimum Cutting Speed (mm/min) Minimum Fitness value 1. 0.48868 0.51132 12.993 µ sec, 29.749 V, 2199.3 g 59.2890 3.7930 1.1654 2. 0 1 9.001 µ sec, 29.361 V, 2198.5 g 39.5836 5.0155 5.0155 3. 1 0 12.994 µ sec, 28.973 V, 2198.7 g 59.3632 3.7704 59.3632 4. 0.05 0.95 9.003 µ sec, 30.887 V, 2199.6 g 39.3047 5.0059 1.0273 5. 0.10 0.90 9.000 µ sec, 29.9744 V, 2199.4 g 39.5304 5.0245 1.0486 6. 0.15 0.85 9.001 µ sec, 30.320 V, 2197.0 g 39.4330 5.0151 1.0759 7. 0.20 0.80 9.004 µ sec, 29.667 V, 2199.0 g 39.5687 5.0186 1.0996 8. 0.25 0.75 9.005 µ sec, 30.123 V, 2199.7 g 39.5108 5.0200 1.1249 9. 0.30 0.70 9.000 µ sec, 29.4891 V, 2199.1 g 39.5839 5.0198 1.1493 10. 0.35 0.65 9.003 µ sec, 29.876 V, 2198.5 g 39.5364 5.0194 1.1750 11. 0.40 0.60 12.9829 µ sec, 29.9044 V, 2196.8 g 59.1344 3.7853 1.1965 12. 0.45 0.55 12.9988 µ sec, 29.5415 V, 2197.9 g 59.3540 3.7874 1.1784 13. 0.50 0.50 12.9989 µ sec, 28.9757 V, 2197.1236 g 59.3839 3.7675 1.1655 14. 0.55 0.45 12.999 µ sec, 29.3163 V, 2199.205 g 59.4003 3.7852 1.1459 15. 0.60 0.40 12.9987 µ sec, 29.5635 V, 2197.9 g 59.3500 3.7877 1.1298 16. 0.65 0.35 12.9912 µ sec, 29.5889 V, 2199.5 g 59.2977 3.7905 1.1138 17. 0.70 0.30 12.9986 µ sec, 29.7341 V, 2196.6 g 59.3080 3.7872 1.0980 18. 0.75 0.25 12.996 µ sec, 29.9148 V, 2199.3 g 59.2630 3.7897 1.0821 19 0.80 0.20 12.9983 µ sec, 30.4519 V, 2198.0 g 59.1576 3.7913 1.0674 20. 0.85 0.15 12.9986 µ sec, 29.9285 V, 2199.6 g 59.3106 3.7972 1.0489 21. 0.90 0.10 12.993 µ sec, 28.845 V, 2199.5 g 59.3682 3.7667 1.0331 22. 0.95 0.05 12.998 µ sec, 29.9954 V, 2198.6 g 59.2816 3.7947 1.0174
By adopting this PBMOO, the mean of MRR and Cutting Speed is calculated, and a t-test is performed for both CS and PSO. For this t-test, the null hypothesis was assumed that both algorithms give the same result. At the same time, the alternative hypothesis is that the two algorithms give different results.[[
Table 10. T-test comparison of CS & PSO.
Algorithms Mean optimized Results Size of samples DoF (Degree of freedom) Standard Deviation Types of Hypotheses t-score p-value Results CSA 52.16988 (µ1) 22 42 3 The null hypothesis is such that µ1=µ2 0.07158 0.9432 Null-Hypothesis cannot be rejected PSO 52.10513 (µ2) 22 The alternative Hypothesis is µ1≠µ2 CSA 4.24246 (µ1) 22 42 3 The null hypothesis represents µ1=µ2 0.010127 0.9919 Null-Hypothesis cannot be rejected PSO 4.23330 (µ2) 22 Alternative Hypothesis µ1≠µ2
From this t-test, it is obtained that for both CS and PSO, the p-value is greater than 0.05; hence, the null hypothesis cannot be excluded. This infers that MRR and Cutting Speed give almost identical results for both algorithms. But as the No Free Lunch Theorem (NFL) says, a particular algorithm produces improved results for a specific problem, and no one is perfect. Whereas for other problems, another algorithm is promising. The above table shows that CS gives somewhat better results than PSO. But the elapsed time for PSO is less than for CS. The time elapsed for the algorithm's execution depends on the algorithm's complexity and the operating system used.
The PBMOO for CS and PSO shows that when we prioritize the individual response. So, it gives a more favorable result for that response.
The convergence curve indicates how many iterations it gives the best optimal value. From the convergence curves of MO-PSO and MO-CS, it is clear that the cuckoo search algorithm converges the function much more quickly than the particle swarm optimization. Whereas the Pareto Front indicates how much the responses interacted with each other. The Pareto Front of MO-CS found that the cutting speed is lower at higher material removal rates (due to higher Kerf width at higher MRR). In comparison, the Pareto front of MO-PSO indicates that both the responses are scattered. The convergence curve and the Pareto front figure are shown in Figures 18 & 19 below.
Graph: Figure 18. Multi-objective Cuckoo Search Optimization Pareto front & convergence curve.
Graph: Figure 19. Multi-objective PSO Pareto front & convergence curve.
The MRR and cutting speed are chosen as the responses in the present work. Whereas these reactions are of better type. But comparatively cutting speed weightage is better type and MRR weightage is comparatively smaller better type, as kerf width should be minimum. MRR, kerf width, and cutting speed, are correlated as follows:
(
Graph
Because Ton is the most important factor for MRR, kerf width and cutting speed. When increasing Ton, MRR, and kerf width increases. But the width of the kerf should be minimum. So, the cutting speed should be better than the MRR.
The article compared PBMOO-based CS & PSO, while WEDM was performed on Ti6Al4 V alloy. It was found that CS gives somewhat better results than PSO. The critical conclusions are appended below:
- MRR is maximum at high Ton (13 micro-seconds), medium voltage (28.88 volts), and high tension (2200 g).
- Cutting Speed is maximum at lower Ton (9 microseconds), medium voltage (29.9359 Volts), & higher tension (2200 grams).
- PBMOO is considered for CS and PSO to optimize MRR and Cutting Speed. It is achieved that CS gives somewhat better results than PSO.
- The GRA sets a weighting of 48.869% for MRR and 51.13% for Cutting Speed.
- The multi-objective optimization by CS shows that at 13 micro-seconds, 29.876 volts, and 2200 g, the maximum MRR is 59.3377 milligram/min and Cutting Speed 3.7980 mm/min.
- Multipurpose optimization by GRA PSO shows that at 12.9932 microseconds, 29.7496 volts, and tension of 2199.3 g, the optimum MRR is 59.2890 mg/min, and Cutting Speed is 3.7930 mm/min.
- SEM images of the WEDM machined surface have more voids and craters than the unmachined surface. Therefore, WEDM can enhance biocompatibility.
The future scope of this work is to investigate other process parameters such as Toff, duty cycle, peak current, wire feed and various wire materials of WEDM on Ti6Al4 V alloy to analyze the machining, fabrication, and biomedical aspect. Furthermore, by developing more exploring metaheuristic algorithms and hybrid type metaheuristic algorithms, better and more accurate results can be obtained.
No potential conflict of interest was reported by the authors.
By Shubham Jain and Vishal Parashar
Reported by Author; Author