Micro-ECDM performance analysis and hybrid-optimization during micro-channel generation on silica using HGAPSO-ANN

Abstract

Hybridizing non-traditional machining (NTM) processes, such as micro-ECDM, is a great challenge in measurement and microfabrication. To find out the solution to this problem, in the present research, the overall model is demonstrated in three steps, during the first stage, ANN is used to construct the linear model of width of cut (WOC), metal removal rate (MRR), and surface roughness (SR) from experimental data consisting of process parameters, that is, voltage (V) pulse frequency (PF), electrolyte concentration (EC), and duty ratio (DR). In the second phase, to get the best-fitted model, we applied both Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and a hybrid of these two algorithms for cross-validation and validation on the train and test datasets. Based upon root mean square error, accuracy, and computational time, the proposed algorithm is more efficient than PSO and GA. Furthermore, in the final phase, the ANN model was optimized using hybrid GAPSO, which helped to determine the optimal process parameters responsible for maximum MRR, minimum WOC, and SR formation. The result shows that maximum MRR, minimum WOC, and SR were formed for the optimized value of voltage 45 volts, electrolytic concentration 30 wt%, DR 0.45, and PF 75 Hz. Moreover, hybrid GAPSO-ANN shows better convergence, accuracy, and computational time (seconds) for micro-machining characteristics analysis.

Keywords

Micro-ECDM hybrid-optimization ANN HGAPSO micro-channel

Introduction

Hybrid machining processes, such as the ECDM process, accrue better performances to cut brittle hard materials like silica glass, wafers, and advanced ceramics. Chak et al.¹ reported that the amount of material detached from the job specimen was augmented with an increase in electrolyte conductivity, voltage (V), and duty factor. Kulkarni et al.² designed micro-channels on glass using the electrochemical spark micro-machining (ECSMM) process with the square pulsating waveform as a machining power source to investigate the duty cycle effects on the micro-machining performances. Biswas et al.³ presented an optimization technique to optimize experimental data by utilizing multi-objective optimization by ratio analysis (MOORA). To enhance the efficiency of milling particulate-reinforced metal matrix composites (MMCs), a grinding-aided electrochemical discharge machining (GECDM) method has been suggested. To enhance the efficiency of milling particulate-strengthened MMCs, Liu et al.⁴ proposed a GECDM method. Cao et al.⁵ Research was done into a combined ECDM/micro-grinding process using polycrystalline diamond (PCD) tools to speed up milling and enhance the surface quality produced by the ECDM process. A random model based on experiments was presented by Jiang et al.⁶ for electrochemical discharge machining spark energy estimation. The electrolyte used was 30-weight percent NaOH, and the tungsten instrument had a width of 250 m. Different types of electrolytes, as well as various shapes of micro-tools, are used to consider their effects during the ECDM process.⁷ Mallick et al.^8–11 applied Genetic Algorithm (GA), RSM, and ANOVA for multi-criteria optimization during the ECDM process. Machining performances, as well as machining depth and surface quality, have been improved using mixed electrolytes in the ECDM process by Mallick et al.^{12, 13} Bellubbi et al.¹⁴ applied the GRA-Taguchi technique for multi-objective optimization during micro-channeling on glass. A square heat source is applied for the simulation of the metal removal rate (MRR).¹⁵ Temperature distribution for the Gaussian heat source predicted MRR of micro-ECDM.¹⁶ MRR increases by escalating energy distribution and reduces with enlarging spark radius.¹⁷ Experimental discharge energy is anticipated, and spark energy ensures the mixed log-normal distribution.⁶ Heat is dissipated during the machining channel to the job specimen.¹⁸ At present, Gaussian heat sources are mainly considered during micro-machining performances.¹⁹ Finite element approach modeling propounded by Paul for the prediction of MRR and width of cut (WOC) during the ECDM process.²⁰ Due to ineffective mathematical modeling of input and output process parameters of micro-ECDM, selecting effective optimization algorithms to achieve the optimal process parameters is still a challenging task for the researcher.

Population-based optimization techniques like Particle Swarm Optimization (PSO) and GA each have advantages and disadvantages. Although the GA technique is highly resilient and better suited for addressing numerous objective problems, the existence of evolutionary operators like selection, crossover, and mutation in the solution generation process causes the method’s convergence to occur slowly.^{21, 22} On the other hand, the PSO offers quick convergence as it generates solutions using mathematical rather than evolutionary operators.^{23, 24} Due to a lack of diversity, PSO generates premature convergence, which is a disadvantage. Thus, a better algorithm would incorporate the strengths of these two algorithms to achieve fast convergence and high diversity.²⁵ A hybrid algorithm, HGAPSO, combines GA and PSO for their feasible approach, similar working methodology, and finding non-dominant solutions.²⁶ The main procedural steps of the hybrid GAPSO algorithm are as follows: In the first stage. The initial population is generated. Binary codes 1 and 0 are used to encode the hidden nodes. In Step 2, the training dataset was applied, and the fitness of each particle was updated to the best position $P_{b e s t}$ (t) and the global best position $P_{g b e s t}$ (t). If any updated fitness value does not satisfy termination, then it automatically proceeds to the next step or produces the final solution. In Stage 3, the GA process is performed by applying a roulette–wheel selection algorithm to determine population members.²⁵ When the value of recent fitness is more efficient than the preceding value, then the overall process is terminated. Once the termination is satisfied, final solutions are obtained, and the appropriate network architecture and connection weights are otherwise used to update the parameters of PSO.²⁶ In the final step, the output of the testing dataset was obtained by an evolutionary ANN with a defined set of iteration numbers.

Earlier several authors have proposed different hybridized evolutionary algorithms, GAPSO is a hybrid methodology combining PSO, and GA for cloud process optimization, enhancing task assignment efficiency, reducing costs, and improving resource utilization,²⁷ robust analysis in breast cancer diagnosis,²⁸ optimizing chemo virotherapy treatment schedules,²⁹ optimizing electric logistics vehicle routing,³⁰ working path of an AUV manipulator,³¹ optimizing circular antenna arrays,³² optimizing micro-tools ambient temperature,³³ and Taguchi process design in drilling.³⁴ Different tools were designed to increase the discharge rate for the enhancement of machining accuracy.³⁵ Lower HAZ achieved during groove cut on titanium alloy using laser power of 90 kHz.³⁶ Hybrid GAPSO optimization played a vital role with ANN during the manufacturing process for finding the optimal results.³⁷ L₉ Taguchi design proposed during micro-machining of borosilicate glass by WECSMM.³⁸ Micro-features, as well as micro-holes on the composite, were created by the ECDM process precisely.³⁹

Motivated by the aforementioned discussion, we propose a novel hybrid approach that considers the GA and PSO algorithms for the parametric optimization of micro-ECDM.

The present research is organized into the following parts: Section “Experimental setup and methodology,” describes the experimental setup and methodology, followed by the result analysis in section “Results analysis for hybrid-optimization.” This section describes linear model implementation with a neural network, and the mathematical model is optimized by three algorithms: GA, PSO, and hybrid GAPSO. In the last section, explain the key points observed in this research along with the future scope in the conclusion section.

Experimental setup and methodology

Figure 1 depicts this experimental setup, which consists of an electric power supply with an oscilloscope, main chamber, graphite auxiliary electrode, micro-tool holder unit, and spring feed. The template acts as a CAM, and the guide pin acts as a follower. Silica glass is the job specimen that has been repeatedly used. Figure 2 shows the neural network configuration with input, hidden and output layers. In the hybrid GAPSO algorithm, the flexible operator of GA merges into the PSO algorithm to construct the appropriate network with optimum connection weights, as shown in Figure 3. Hybrid GAPSO with ANN has an effective convergence to find optimized outputs with proper input combinations. Figure 4 represents the performance characterization based on ANN models.

Figure 1.

Micro-ECDM experimental setup.

Figure 2.

Neural network architecture of the proposed model.

Figure 3.

Flowchart of the hybrid GAPSO algorithm.

Figure 4.

Proposed methodology in the present research.

ANN models are highly used in AI and are stimulated by the neuron topology to solve critical nonlinearities and complex problems. An NN model can create a complex structure make a computer learn from given datasets and predict better results in the domain. A conventional ANN model consists of layers: input, hidden, and output layers with a feed-forward configuration of an architect. The input layer is connected to the hidden layer with the output layer response.

Input vectors $x = x_{1}, x_{2}, \dots, x_{n}$ and $w_{1,} w_{2}, w_{3}, \dots, w_{n}$ are weights to determine the associated neuron connection weight $x^{T} w$ is shown in Equation (1).

x = \sum_{i = 1}^{n} x_{i} w_{i}

(1)

The node’s threshold is b, which indicates offset values are shown in Equation (2).

y = f (x) = f \{\sum_{i = 1}^{n} x_{i} w_{i} - b\}

(2)

Earlier, ANN initialized the random weight values, but after training, these weight values are updated according to the input and hidden layer structure. In such cases, we used either backpropagation learning or other updated learning algorithms. These updated weighted values are adjusted in such a manner that we get optimized calculated responses for an unknown set of inputs.

Experimental planning

Every experiment is conducted in the laboratory three times, and the average value of the results is used for analysis.

For the three sets of experimental results, only average values are taken, and the deviation for MRR, WOC and surface roughness (SR) 5%–10%.

MRR is measured by the Toledo weighing machine by calculating the weight of the job specimen before and after machining for 20 minutes, and WOC is measured by an optical microscope, and a Taly surf stylus of 5 µm is used to find out SR.

NaOH and KOH are mixed in equal proportions for preparing electrolytes, and a tungsten carbide tool of 250-µm cylindrical micro-tool has been used for micro-channeling on silica glass. Face-centered design (FCD) of experiments of 27 experiments are conducted, ranges of variables are shown in Table 1, which consists of V, electrolyte concentration (EC), pulse frequency (PF), and duty ratio (DR) and parametric combinations as well as results of experimentation have been disclosed in Table 2. From Table 2, it is seen that MRR, WOC, as well as SR are increased due to high voltage, EC and DR and decrease due to higher PF because of the sparking rate falling.

Table 1.

Variables range.

Process parameters (variables)	Range
	–1	0	+1
Voltage (V)	45	50	55
Electrolyte con. (wt%)	20	25	30
Duty ratio	0.45	0.50	0.55
Pulse frequency (Hz)	50	75	100

Table 2.

Experimental results and parametric combinations.

SL. no.	Voltage (V)	Electrolyte con. (wt%)	Duty ratio	Pulse frequency (Hz)	MRR (mg/hr)	WOC (µm)	SR (µm)
1	45	20	0.45	50	11.15	471.13	1.13
2	45	20	0.45	50	12.16	492.17	2.14
3	45	20	0.45	50	11.46	480.49	1.42
4	45	25	0.50	75	12.61	473.60	2.41
5	45	25	0.50	75	11.60	462.56	1.40
6	45	25	0.50	75	12.06	462.10	2.01
7	45	30	0.55	100	11.15	456.10	1.12
8	45	30	0.55	100	11.86	472.90	1.56
9	45	30	0.55	100	11.09	462.08	1.05
10	50	20	0.50	100	11.85	462.76	1.65
11	50	20	0.50	100	11.01	483.23	1.11
12	50	20	0.50	100	12.70	472.78	2.60
13	50	25	0.55	50	11.14	472.15	1.11
14	50	25	0.55	50	11.87	472.98	1.77
15	50	25	0.55	50	11.85	462.90	1.55
16	50	30	0.45	75	12.97	472.98	2.77
17	50	30	0.45	75	11.06	471.11	1.02
18	50	30	0.45	75	12.15	483.24	2.11
19	55	20	0.55	75	11.98	483.99	1.78
20	55	20	0.55	75	12.35	482.35	2.25
21	55	20	0.55	75	11.30	481.35	1.29
22	55	25	0.45	100	11.68	492.78	1.78
23	55	25	0.45	100	12.08	492.11	2.05
24	55	25	0.45	100	12.29	488.34	2.27
25	55	30	0.50	50	12.04	482.12	2.05
26	55	30	0.50	50	12.07	492.23	2.08
27	55	30	0.50	50	12.02	488.34	2.04

Results analysis for hybrid-optimization

The overall optimization is done by 27 experimental data of micro-electrochemical discharge machining (Micro-ECDM) process parameters, including input parameters V, EC, PF, and DR and output parameters WOC, material removing rate (MRR), and SR. Overall, research was performed in three sections. In the first phase, cross-validation was performed on 90% (24) of the datasets. Table 7 represents the RMSE and accuracy of all the applied optimization techniques. In the second phase, the rest of the 10% (three datasets) data was used for validation of the proposed model, Table 8 representing the accuracy and RMSE of GA-ANN, PSO-ANN, and the hybrid algorithm. In the final stage, the optimal micro-ECDM process parameters were obtained for maximum MRR, minimum WOC and SR. The overall research is represented in Figure 4. These extensive simulations have been performed on MATLAB 2015a, with a system configuration of Processor 16GB RAM, 11th Gen Intel (R) Core (TM) i5 @ 2.50GHz.

Finding a neural network model for MRR, WOC, and SR

In this work, three output variables, MRR, WOC, and SR, are considered a linear function of input variables V, EC, DR, and PF, which can be represented by Equations (3–5).

M R R = f_{1} (V, E C, D R, P F)

(3)

W O C = f_{2} (V, E C, D R, P F)

(4)

S R = f_{3} (V, E C, D R, P F)

(5)

Four weights, $w_{1,} w_{2}, w_{3}$ , $w_{4}$ associated with four inputs, and the output node associated with $β$ .

M R R = D \times w_{1, M} + E C \times w_{2, M} + D R \times w_{3, M} + P F \times w_{4, M} + β_{M}

(6)

W O C = D \times w_{1, W} + E C \times w_{2, W} + D R \times w_{3, W} + P F \times w_{4, W} + β_{W}

(7)

S R = D \times w_{1, S} + E C \times w_{2, S} + D R \times w_{3, S} + P F \times w_{4, S} + β_{S}

(8)

Now PSO, GA, and HGAPSO evolutionary algorithms are used to generate the optimum values of neural network weights and bias values from Equations (6) to (8). From 27 experiment datasets, 90% of the data has been taken for training and the rest of the data for testing. ANN-based linear mathematical model designed using a training dataset. Proposed three optimization techniques applied to random populations $w_{1,} w_{2}, w_{3}, w_{4}, a n d β$ can be obtained, which provides the least squares error for the training dataset using Equations (9–11).

E = \sum_{i = 1}^{m} {(t_{i} - c_{i})}^{2}

(9)

Objective function E is a function of t and c (target and calculated output), which should be minimized by optimization techniques. From optimization technique, an optimal solution is obtained when the error is minimized. The process variables of PSO, GA, and HGAPSO are shown in Table 3. Tables 4–6 represent the coefficients of the ANN-based linear model of MRR, WOC, and SR. Statistical results and computational time of each algorithm are represented in Table 7, and the best-fitted model. The mathematical equation of RMSE can be defined as follows

R M S E = \sqrt{\frac{1}{m} \sum_{i = 1}^{m} {(\frac{X_{e x p} - X_{C a l}}{X_{e x p}})}^{2}} \times 100 %

(10)

Accuracy = (100 - R M S E) %

(11)

Table 3.

Process variables of GA, PSO, and HGAPSO for NN modeling.

PSO variables	Value	GA variables	Value	HGAPSO variables	Value
Population ( $n_{p o p}$ )	50	Population ( $n_{p o p}$ )	50	Population ( $n_{p o p}$ )	50
$ρ_{m a x}$	0.9	$P_{c}$	0.7	$W_{d a m p}$	0.99
$ρ_{m i n}$	0.4	$γ$	0.4	$C_{1}$	1.5
		$P_{m}$	0.2	$C_{2}$	2.0
		$m_{u}$	0.05	$P_{c}$	0.7

Table 4.

Coefficient of MRR after optimization.

	$w_{1, M}$	$w_{2, M}$	$w_{3, M}$	$w_{4, M}$	$β_{M}$
PSO-ANN	–14.346	14.4951	15	6.6531	15
GA-ANN	7.020	3.1709	15.000	0.2876	15.000
HGAPSO-ANN	–2.597	–0.529	5.887	3.010	5.250

Where, $X_{e x p}$ and $X_{c a l}$ is the target as well as optimized value, respectively.

Table 7 represents a comparative study on cross-validation, where it has been seen that the proposed ANN-optimized hybrid GAPSO obtained an accuracy of 97.146% and 97.748% with the least computational time of 21.4162 sec and 32.145 sec for MRR and WOC, while the ANN-optimized GA achieved maximum accuracy for SR.

Equations (12–14) representing the nonlinear mathematical model of ANN-based hybrid GAPSO of output variables are as follows:

MRR = V \times (- 2.59) + E C \times (- 0.525) + D T \times (5.8102) + P F \times 3.0124 + 5.25

(12)

W O C = V \times (10) + E C \times (9.99) + D T \times (- 1.10) + P F \times (- 3.92) - 0.63

(13)

S R = V \times (- 3.30) + E C \times (3.89) + D T \times (3.457) + P F \times (0.862) - 0.3515

(14)

The remaining 10% of the data is used to check the accuracy of testing the dataset or forecasting the model. The entire model can accurately forecast MRR, WOC, and SR for new conditions. Table 8 represents the efficacy of proposed model efficacy of the proposed model. Convergence characteristics of the proposed algorithm decreased abruptly for all the response variables shown in Figures 5, 6 and 7, from where, by testing convergence, it is obviously found that hybrid GAPSO is the best method for optimization. From Tables 7 and 8, it is clear that hybrid GAPSO with ANN provides less error of root mean square (RMSE) with higher accuracy and minimum computational time as well as minimum cost. Figure 8 represents SEM of the micro-channel with HAZ on Silica, and validated the test results of the micro-channel fabricated at 45 volts, 30 (wt%), 0.45, and 75 Hz.

Table 5.

Coefficient of WOC after optimization.

	$w_{1, W}$	$w_{2, W}$	$w_{3, W}$	$w_{4, W}$	$β_{W}$
PSO-ANN	14.9604	–13.828	13.7697	8.14937	4.6155
GA-ANN	7.2628	2.6811	15.00	0.2937	15.000
HGAPSO-ANN	–3.301	3.892	3.475	0.862	–3.5.15

Table 6.

Coefficient of SR after optimization.

	$w_{1, S}$	$w_{2, S}$	$w_{3, S}$	$w_{4, S}$	$β_{S}$
PSO-ANN	–6.26	–12.3854	15	7.5930	15
GA-ANN	7.342	2.6008	15.000	0.2678	15.000
HGAPSO-ANN	10	9.999	–1.108	–3.9222	–0.6377

Table 7.

Outcomes using PSO-ANN, GA-ANN, and HGAPSO-ANN by cross-validation.

Parameters		ANN-PSO	ANN-GA	ANN-HGAPSO
MRR	Error of root mean square (RMSE)	7.4627	5.9099	2.8543
	Accurateness	92.5373	94.091	97.146
	Computational time	36.4667 sec	139.650 sec	21.4162 sec
	Minimum cost	0.00362	0.00219	0.00134
WOC	Error of root mean square (RMSE)	6.691	2.3064	2.252
	Accurateness	93.309	97.6936	97.748
	Computational time	63.4632 sec	105.07 sec	32.145 sec
	Minimum cost	0.0063675	0.00748	0.003185
SR	RMSE	4.521	3.767	3.965
	Accuracy	95.479	96.233	96.035
	Computational time	58.632 sec	74.523 sec	29.213 sec
	Minimum cost	0.00652	0.00654	0.00324

Table 8.

Comparative study for testing the data set.

Variables		Calculated value			RMSE (%)			Accuracy (%)
	Exp. Value	ANN-PSO	ANN-GA	ANN-HGAPSO	ANN-PSO	ANN-GA	ANN-HGAPSO	ANN-PSO	ANN-GA	ANN-HGAPSO
MRR	12.04	11.92	11.83	12.13	11.24	7.65	4.08	88.76	92.35	95.92
	12.07	13.45	12.98	11.85
	12.02	10.65	12.96	11.56
WOC	482.12	471.63	473.65	476.35	12.91	8.84	6.82	87.09	91.16	93.18
	492.23	501.63	496.32	484.36
	488.34	470.96	476.24	481.68
SR	2.05	1.95	1.97	2.01	15.85	12.12	6.21	84.15	87.88	93.79
	2.08	2.21	2.19	2.14
	2.04	2.28	2.2	2.12

Figure 5.

Characteristics graph between performances with the number of epochs of the ANN model.

Figure 6.

Convergence curve for MRR using different algorithms.

Figure 7.

Convergence curve for WOC using different algorithms.

Figure 8.

Convergence curve for SR using different algorithms.

Optimal conditions for model

Optimum input variables are obtained when the objective function contains maximum MRR, and minimum WOC and SR represented in Equation (15).

F (M R R, W O C, S R) = M R R + \frac{1}{W O C} + \frac{1}{S R}

(15)

To obtain the optimized objective function for a given range of input, it has been discovered that the best combination of V, E, D, and f is 45 volts, 30 (wt%), 0.45, and 75 Hz, respectively, and it takes 56.433 seconds by the hybrid GAPSO-ANN model. This result indicates that the best drilling condition was obtained for maximum electrolytic concentration and the least value of V and DR. Figure 9 shows the SEM of micro-channel cutting at 45 volt, 3owt%, 0.45 duty ratio and 75Hz.

Figure 9.

SEM of micro-channel with HAZ at 45 volt, 30 (wt%), 0.45 duty ratio, and 75 Hz.

Conclusions

Finally, an account of the present research on micro-ECDM performances

ANN is the best fitness linear model, while hybrid GAPSO is the most relevant optimization technique.

The best combination of input parameters is 45 V, 30 wt% electrolytes, 0.45 DR, and 75 Hz PF when WOC and SR are minimized and MRR is maximized during multi-objective soft computational hybrid GAPSO is used in the ECDM process.

From cross-validation, it is seen that the ANN-hybrid GAPSO optimization technique provides maximum accuracy rather than ANN-GA and ANN-PSO.

During the testing dataset, it has been observed that the proposed hybrid algorithm achieved maximum accuracy for MRR, WOC, and SR at about 95.92%, 93.18%, and 93.79%, respectively.

A heat-affected area was also observed at the micro-channel after proper SEM investigations.

Test results are cross-validated for micro-channeling on (silicon dioxide + sodium silicate) silica.

Footnotes

Acknowledgments

The authors are thankful to Shobhit Institute of Engineering and Technology (Deemed to be University), Meerut, UP, India and GIMT, Nadia, WB, for permitting us to conduct the experimentation work.

Authors’ contributions

All the authors have equal contributions in experimentation, results analysis and writing of the manuscript.

Availability of data and material

The datasets generated during and/or analyzed during the current study are available from the authors and would be provided to the journal if required.

Compliance with ethical standards

This article does not contain any studies with human participants or animals performed by any of the authors.

Consent to participate

The authors are free to contact any of the people involved in the research to seek further clarification and information.

Consent for publication

The authors give their consent for the publication of identifiable details, which can include photograph(s) and/or details within the manuscript to be published in the esteemed journal.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Bijan Mallick

Azzam Sabah Hameed

References

Chak

and Rao

PV.

Machining of SiC by ECDM process using different electrode configurations under the effect of pulsed DC. All India Manufacturing Technology, Design and Research Conference 2012; 1: 513–519.

Kulkarni

, Jain

and Misra

KA.

Performance of micro machining using ECSMM with square pulsating power source. All India Manufacturing Technology, Design and Research Conference 2012; 2: 995–1000.

Biswas

, Sarkar

, Doloi

, . Parametric optimization of µ-ECDMing of silicon nitride ceramics. All India Manufacturing Technology, Design and Research Conference 2012; 2: 1079–1084.

Liu

, Yue

and Guo

ZN.

Grinding-aided electrochemical discharge machining of particulate reinforced metal matrix composites. Int J Adv Manuf Technol 2013; 8: 4846.

Cao

, Kim

and Chu

CN.

Hybrid micromachining of glass using ECDM and micro grinding. Int J Precis Eng Manuf 2013; 14(1): 5–10.

Jiang

, Lan

, Ni

, . Experimental investigation of spark generation in electrochemical discharge machining of non-conducting materials. J Mater Process Technol 2014; 214: 892–898. DOI: 10.1016/j.jmatprotec.2013.12.005

Mallick

, Biswas

, Sarkar

, . On performances of electrochemical discharge micro-machining process using different electrolytes and tool shapes. Int J Manuf Mater Mech Eng 2019; 10(2). DOI: 10.4018/IJMMME.2020040103

Mallick

, Sarkar

, Doloi

, . Analysis on the effect of ECDM process parameters during micro-machining of glass using genetic algorithm. J Mech Eng Sci 2018; 12(3): 3942–3960.

Mallick

, Sarkar

, Doloi

, . Analysis of electrochemical discharge machining during micro-channel cutting on glass. Int J Precis Technol 2017; 7(1): 32–50.

10.

Mallick

, Sarkar

, Doloi

, . Modelling and analysis on performance of ECDM process for the fabrication of µ-channels on glass through response surface methodology. Manuf Technol Today 2019; 18: 9.

11.

Mallick

, Sarkar

, Doloi

, . Multi criteria optimization of electrochemical discharge micro-machining process during micro-channel generation on glass. Appl Mech Mater 2014; 592–594: 525–529.

12.

Mallick

, Hameed

, Sarkar

, . Experimental investigation for improvement of micro-machining performances of µ-ECDM process. Mater Today 2020; 27(Part 1): 620–626.

13.

Mallick

, Sarkar

, Doloi

, . Improvement of surface quality and machining depth of µ-ECDM performances using mixed electrolyte at different polarity. Silicon 2021. DOI: 10.1007/s12633-021-01587-2

14.

Bellubbi

, NS and Mallick

Multi response optimization of ECDM process parameters for machining of microchannel in silica glass using Taguchi–GRA technique. Silicon 2021. DOI: 10.1007/s12633-021-01167-4

15.

Jain

, Dixit

and Pandey

PM.

On the analysis of the electrochemical spark machining process. Int J Mach Tools Manuf 1999; 39: 165–186. DOI: 10.1016/S0890-6955(98)00010-8

16.

Goud

and Sharma

AK.

A three-dimensional finite element simulation approach to analyze material removal in electrochemical discharge machining. Proc Inst Mech Eng Part C J Mech Eng Sci 2017; 231: 2417–2428. DOI: 10.1177/0954406216636167

17.

Panda

and Yadava

Finite element prediction of material removal rate due to traveling wire electrochemical spark machining. Int J Adv Manuf Technol 2009; 45: 506–520. DOI: 10.1007/s00170-009-1992-0

18.

Elhami

and Razfar

MR.

Numerical and experimental study of discharge mechanism in the electro chemical discharge machining process. J Manuf Process 2020; 50: 192–203. DOI: 10.1016/j.jmapro.2019.12.040

19.

Rajput

, Goud

, Suri

, . Surface characteristics study of Ti–6Al–4V alloy for biomedical applications using finite element modeling. J Mater Res 2022; 37: 2710–2721. DOI: 10.1557/s43578-022-00677-0

20.

Paul

and Hiremath

SS.

Model prediction and experimental study of material removal rate in micro ECDM process on borosilicate glass. Silicon 2022; 14: 1497–1510. DOI: 10.1007/s12633-021-00948-1

21.

Dutta

and Kumar

Modelling of liquid flow control system using optimized genetic algorithm. Stat Optim Inf Comput 2020; 8(2): 565–582.

22.

Dutta

and Kumar

Design an intelligent calibration technique using optimized GA-ANN for liquid flow control system. J Eur Syst Autom 2017; 50(4–6): 449.

23.

Dutta

, Cengiz

and Kumar

Study of bio-inspired neural networks for the prediction of liquid flow in a process control system. In: Cognitive Big Data Intelligence with a Metaheuristic Approach. Academic Press, 2022, pp.173–191.

24.

Shami

, El-Saleh

, Alswaitti

, . Particle swarm optimization: A comprehensive survey. IEEE Access 2022; 10: 10031–10061.

25.

Sun

, Shen

and Vogel-Heuser

A hybrid genetic algorithm for distributed hybrid blocking flowshop scheduling problem. J Manuf Syst 2023; 71: 390–405.

26.

El-Shafiey

, Hagag

, El-Dahshan

ESA

, . A hybrid GA and PSO optimized approach for heart-disease prediction based on random forest. Multimed Tools Appl 2022; 81(13): 18155–18179.

27.

Rani

, Savitha

and Naval

Cloud process optimization: A hybrid GA-PSO methodology. In: 2023 International Conference on Advances in Computation, Communication and Information Technology (ICAICCIT). IEEE, 2023, pp.938–943.

28.

Lin

, Lee

, Chen

, . Automated marker-free longitudinal infrared breast image registration by GA-PSO. Phys Med Biol 2023; 68(24): 245026.

29.

Amini

, Niroomandfard

, Khalili

, . The improvement of chemovirotherapy effectiveness utilizing hybrid GA-PSO. In: 2023 30th National and 8th International Iranian Conference on Biomedical Engineering (ICBME). IEEE, 2023, pp.205–212.

30.

Wang

and Xie

Logistics pure electric vehicle routing based on GA-PSO algorithm. Mechanics 2023; 29(3): 235–242.

31.

Cheng

Working path optimization of AUV manipulator based on PSO-GA algorithm. In: 2023 4th International Conference for Emerging Technology (INCET). IEEE, 2023, pp.1–6.

32.

Kumar

, Ram

, Kar

, . Radiation pattern comparison of circular antenna arrays using GA and PSO. In: 2022 IEEE International Symposium on Smart Electronic Systems (iSES). IEEE, 2022, pp.164–168.

33.

Adin

MŞ

, İşcan

and Baday

Ş.

Machining fiber-reinforced glass-epoxy composites with cryo-treated and untreated HSS cutting tools of varying geometries. Mater Today Commun 2023; 37: 107301. DOI: 10.1016/j.mtcomm.2023.107301

34.

Adin

MŞ.

Machining aerospace aluminium alloy with cryo-treated and untreated HSS cutting tools. Adv Mater Process Technol 2023; 10(3): 2664–2689. DOI: 10.1080/2374068X.2023.2273035

35.

Walia

and Singh

A comprehensive review on electrochemical discharge machining (ECDM): Process configurations, optimization techniques, simulations, and challenges. J Micromanuf 2025; 0(0). DOI: 10.1177/25165984251369298

36.

Doloi

, Pandey

and Bhattacharyya

Analysis of fiber laser microgrooving characteristics of Ti6Al4V. J Micromanuf 2025; 0(0). DOI: 10.1177/25165984251348179

37.

Mallick

, Hameed

, Chakraborty

, . Hybrid evolutionary algorithms (hybrid-GAPSO) based optimization of MIG welding process. J Micromanuf 2024; 0(0). DOI: 10.1177/25165984241290900

38.

Bhuyan

, Bhuyan

and Porwal

RK.

Hybrid approach for parametric optimization of wire-electrochemical spark micromachining of borosilicate glass. J Micromanuf 2025; 8(2): 132–143. DOI: 10.1177/25165984241308644

39.

Bhuyan

, Bhuyan

, Porwal

, . Prospects for electrochemical discharge micro-drilling of sustainable materials: An extensive review. J Micromanuf 2025; 8(2): 120–131. DOI: 10.1177/25165984251345304