OPSEARCH

https://doi.org/10.1007/s12597-019-00410-2

APPLICATION ARTICLE

Parametric analysis of a green electrical discharge

machining process using DEMATEL and SIR methods

Partha Protim Das1 · Shankar Chakraborty2 

Accepted: 6 September 2019

© Operational Research Society of India 2019

Abstract
To achieve safer machining environment, and minimize emission of harmful and
toxic substances during electrical discharge machining (EDM) process along with
improvement in its performance, this paper emphasizes on identifying the best para-
metric combination of a green EDM process using superiority and inferiority rank-
ing (SIR) method. Decision making trial and evaluation laboratory method is also
employed to visualize the interrelationships between the responses of the said pro-
cess while splitting them into cause and effect groups. In this process, peak cur-
rent, pulse duration, dielectric level and flushing pressure are the input parameters,
while process time, relative tool wear ratio, process energy, concentration of aerosol
and dielectric consumption are considered as the responses. The optimal paramet-
ric combination as derived employing the SIR method is validated with the help
of developed regression equations for each of the responses, which show that the
adopted approach outperforms the other popular optimization techniques in obtain-
ing the best mix of the green EDM process parameters for having improved machin-
ing performance and less hazardous effects on the environment.

Keywords  Green EDM · DEMATEL · SIR method · Process parameter · Response ·

Optimization

1 Introduction

Most of the industrially-accepted manufacturing processes give rise to consider-

able amount liquid, solid and gaseous wastes, contaminating the earth and pollut-
ing the environment which results in various health-related issues. Besides forma-
tion of hazardous wastes, these manufacturing processes are considered to have

* Shankar Chakraborty
[email protected]
1
Department of Mechanical Engineering, Sikkim Manipal Institute of Technology, Sikkim
Manipal University, Majhitar, Sikkim, India
2
Department of Production Engineering, Jadavpur University, Kolkata, West Bengal, India

13
Vol.:(0123456789)
 OPSEARCH

energy-intensive activities, thereby affecting the environment indirectly. Strict

government rules and regulations as well as rising public awareness related to
environmental issues have made these processes as one of the most vital areas for
deployment of strategic manufacturing decisions [34, 48]. In view of these, green
manufacturing, an advanced mode of manufacturing, is gaining interest among the
researchers that aims not only in reduction of the environmental effect and resource
utilization, but also improves efficiency of the manufacturing processes [25, 40].
Die-sinking electrical discharge machining (EDM) is a popular non-traditional
machining process, extensively used by the mould, tool and die making, aerospace,
automobile and electronics industries for generating prototype components, where
volume of production is comparatively low. It has become popular due to its abil-
ity to machine various intricate shape features with higher dimensional accuracy on
numerous advanced, hard and difficult-to-machine materials [16, 23]. In spite of its
several advantages, it is considered to be a hazardous process, because during the
machining operation, it releases a large amount of harmful solid and liquid wastes
along with expulsion of toxic gases, thereby polluting the environment [44]. During
its operation, material is eroded from the workpiece as well as from the tool by a
series of rapid discharges of current, causing formation of inorganic substances, like
tungsten carbide, titanium carbide, barium and chromium, which contaminate the
nearby environment. A high discharge of energy during the EDM process may also
result in generation of reaction products from the dielectric being used, which may
escape as aerosols and toxic gases. In addition to emission of these gases, hazardous
substances, like eroded workpiece or tool material, may concentrate in the slurry as
well as in the dielectric itself. These emitted harmful and toxic substances may enter
into the operators’ body through inward breath, ingestion or skin contact. It has been
observed that emission of the toxic gases and harmful wastes along with various
performance measures of an EDM process are significantly influenced by its differ-
ent input/controllable parameters. Hence, to achieve green environment for an EDM
process, it has become very much necessary to run the machining set-up at the opti-
mal combination of the considered process parameters [24, 35, 36, 42, 49]. Various
mathematical tools, such as analytic hierarchy process (AHP), grey relational analy-
sis (GRA), technique for order of preference by similarity to ideal solution (TOP-
SIS), Vlse Kriterijumska Optimizacija Kompromisno Resenje (VIKOR) etc. can be
systematically deployed to resolve this problem. These techniques have already been
recognized as effective decision making aids in solving problems arising in different
dimensions of engineering and management [5, 14, 17].
Selection of the optimal combinations of input parameters for EDM processes
using different mathematical approaches has already gained much interest among the
researchers. Singh et  al. [37] applied GRA technique for multi-objective optimiza-
tion of various performance measures of an EDM process, such as material removal
rate (MRR), tool wear rate (TWR), radial overcut, taper and surface roughness (Ra)
while machining Al-10%SiCP metal matrix composites. El-Taweel [10] adopted
response surface methodology for achievement of maximum MRR and minimum
TWR, while considering pulse-on time, peak current and flushing pressure as the
EDM process parameters. Joshi and Pande [22] applied an intelligent approach com-
prising artificial neural networks and genetic algorithm for process modeling and

13
OPSEARCH

parametric optimization of an EDM process resulting in improved MRR and TWR

values. Mukherjee and Chakraborty [27] employed biogeography-based optimiza-
tion approach for choosing the appropriate EDM process parameters to enhance vari-
ous quality characteristics, i.e. Ra, surface crack density (SCD), MRR, TWR, white
layer thickness (WLT) and gap size. Considering discharge current, lifting height,
gap voltage, duty factor and polarity as the EDM process parameters, Tang and Du
[41] combined GRA technique with Taguchi method to derive the optimal perfor-
mance characteristics, i.e. TWR, MRR and Ra for the considered process. Tang and
Du [42] proposed the integrated application of GRA and Taguchi methods to decide
the optimal operating levels of peak current, lifting height, pulse width, pulse inter-
mission and open circuit voltage in a green EDM process while using tap water as
the working fluid. Gopalakannan and Senthilvelan [13] performed EDM operation on
metal matrix composites with copper as an electrode. Pulse current, pulse-on time,
gap voltage and pulse-off time were the control parameters, and desirability func-
tion approach was later applied for simultaneous optimization of three responses,
i.e. MRR, Ra and electrode wear rate. Taking into account pulse-on time, discharge
current,  tool lift time and tool work time as the EDM process parameters, Dewan-
gan et al. [9] proposed a proficient methodology combining GRA with fuzzy logic
to obtain the optimal parametric mix while machining AISI P20 tool steel material.
Various surface characteristics, such as WLT, SCD and Ra were considered as the
responses. While treating surfactant concentration, peak current and graphite pow-
der concentration as the three input parameters, Reddy et al. [30] examined the per-
formance of an EDM process in machining of PH17-4 stainless steel material using
surfactant-mixed and graphite powder-mixed dielectric fluids. An integrated Taguchi-
data envelopment analysis-based multi-response optimization technique was applied
for achieving superior values of responses, such as MRR, Ra and TWR. Bhuyan and
Routara [2] applied VIKOR method for selection of the optimal parametric combina-
tion of peak current, pulse-on time and flushing pressure in an EDM process. At that
parametric mix, better values of MRR, TWR, Ra and radial overcut were attained.
Chakraborty and Das [4] presented an effective multivariate quality loss function
approach for concurrent optimization of multiple responses in an EDM process. Roy
and Dutta [33] employed fuzzy-AHP and fuzzy-TOPSIS methodologies for paramet-
ric optimization of an EDM process while treating pulse-on time, duty cycle, dis-
charge current and gap voltage as the input parameters, and MRR, tool wear ratio and
tool overcut as the responses. Das and Chakraborty [8] combined Taguchi method
and super ranking concept as an optimization tool for obtaining better responses in an
EDM process. Wang et al. [45] applied analytic network process (ANP) and decision
making trial and evaluation laboratory (DEMATEL) methods to study the principles
of an EDM process to validate the concept of green manufacturing theory.
Based on the above-cited literature, it becomes quite evident that the optimal
combination of EDM process parameters is very much essential so as to achieve
the target response values. But, there is a scarcity of published research works that
have dealt with the applications of different multi-criteria decision making (MCDM)
tools for parametric optimization of green manufacturing processes. It is also
noticed that some conventional MCDM methods, mainly in the form of AHP, TOP-
SIS, GRA, VIKOR etc., have only been employed for parametric optimization of

13
 OPSEARCH

EDM processes [19, 20, 37]. However, no concrete literature is available dealing
with parametric analysis of green EDM processes, which can be considered as vital
to attain the concept of sensible manufacturing. The literature survey thus identi-
fies the following research gaps: (a) lack of parametric analyses to discover the sig-
nificant effect of each of the responses on the others, (b) dearth of a mathematically
sound MCDM tool for parametric optimization of any of the non-traditional machin-
ing processes, (c) assignment of precise weights to each of the responses from infor-
mation complexion is missing in the traditional approaches, and (d) absence of cor-
relation between the attributes and uniformity in the judgments.
In this paper, DEMATAL method is first applied for parametric analysis of a
green EDM process so as to investigate the significant effect of each of the responses
on the others. Additionally, superiority and inferiority ranking (SIR) method, a
novel MCDM approach, is adopted for obtaining the optimal combination of process
parameters for the said green EDM process. Presently, no literature is accessible that
has studied the application of the SIR method for parametric optimization of any
of the manufacturing processes. Thus, it would be interesting to examine the effec-
tiveness of DEMATEL and SIR methods in analyzing and solving a multi-response
optimization problem for having improved machining performance of a green EDM
process within the perimeter of sustainable manufacturing.

2 Green EDM process

In a die-sinking EDM process, material is removed from the workpiece by a

sequence of sparks generated between the electrode and the workpiece. This gen-
eration of sparks during the machining operation yields high temperature causing
melting and vaporization of the work material. The evaporated metal and some
portion of the molten material are then rinsed out from the machining area by
the dielectric fluid [1]. A die-sinking EDM process with minimum detrimental
effects on the environment is often known as a green EDM process. The interre-
lationships between various green EDM process parameters and responses can be
presented employing an input–output model, as depicted in Fig. 1 [7]. The inputs
to this process include several machining parameters, workpiece material, tool
material and electrical energy. On the other hand, the outputs of this process con-
sist of MRR, TWR, toxic wastes, emission of harmful gases, heat, noise etc. To

Fig. 1  Input-output model for a green EDM process

13
OPSEARCH

Fig. 2  Manufacturing and environmental aspectsofa green EDM process

attain green manufacturing for the EDM process, organizations should take into
account two aspects of machining outputs, i.e. manufacturing aspect and environ-
mental aspect. Each of these aspects is related to the process outputs. The deci-
sion making model for a green EDM process thus consists of two manufacturing
aspects, i.e. process time and relative tool wear ratio, and three environmental
aspects, i.e. process energy, aerosol exposure and consumption of dielectric. For
a green EDM process, these environmental aspects need to be substantially min-
imized, causing less harmful effects to the surroundings. A schematic diagram
considering these two aspects of a green EDM process is provided in Fig. 2.
Process time is the time needed to remove unit volume of material from a given
workpiece sample during the machining operation. It is also directly related to
machining cost. Relative tool wear ratio is measured as the fraction of the amount
of tool material abraded to the amount of workpiece material removed. The amount
of process energy absorbed is determined by estimating the total energy consumed
during discharges and time duration of the current flow. Consumption of more elec-
trical energy indirectly impacts the environment as more waste would be produced
while generating the same. Formation of toxic aerosols during EDM operation is
mainly due to the use of hydrocarbons as the dielectric fluid. Exposure to mass con-
centration of aerosols is very much harmful to the operators as it may consist of
metallic particles and toxic gases formed as the reaction products from the dielectric
fluid. Dielectric consumption is the amount of dielectric consumed per unit volume
of material removal. Dielectric is mainly consumed due to three reasons, e.g. depo-
sition of dielectric fluid on the workpiece, deposition of dielectric fluid on the tool
material as well as on the material removed from both the tool and the workpiece,
and vaporization of the dielectric into the surroundings. Inhalation of such toxic
gases (vaporized dielectric) may cause serious health menaces to the operators.

3 Mathematical models

3.1 DEMATEL method

The DEMATEL method is principally utilized to perceive the interrelations between

various attributes (responses) and identify the key elements to illustrate their

13
 OPSEARCH

effectiveness [11, 12]. The attributes are evaluated based on pair-wise comparisons,
where they are contrasted with deference to relative influence of each on the other. This
method is best suited for dealing with complicated and entwined problems, and primar-
ily compares the interaction relationships between different factors of a complex system
to decide upon the immediate and circuitous causal relationships, and corresponding
impact levels. The results of DEMATEL method are visual portrayals (a visual struc-
tural matrix and a causal diagram) that indicate the causal relationships and impact lev-
els between the considered attributes, and help with settling on suitable choices. The
evaluation procedure usually involves several complicated facets, counting economical
and non-economical aspects, along with developing qualitative viewpoints, which are
either directly or indirectly connected. The application of this method depends on the
idea of generation of digraphs, which can isolate the assumed decision criteria/attrib-
utes into groups of cause and effect so as to visibly portray the inside of a complicated
problem. It presumes a system comprising a set of components C = {C1,C2,…,Cn},
which can be evaluated with respect to their pair-wise comparisons. The implementa-
tion of DEMATEL method involves the following steps [6, 15, 46]:

Step 1 Initiation of the direct-relation matrix (A)

In this stage, the decision maker first assesses the correlation between the groups of
paired criteria to identify the direct influence that each ith criterion (response) exercises
on each jth criterion, being evaluated over a scale (score) spanning from 0 to 4, where
0, 1, 2, 3 and 4 respectively signify no influence, low influence, medium influence, high
influence and very high influence. A direct-relation matrix (A) of n × n matrix is thus
formed as the result of this assessment. Every distinct element (aij) signifies the level by
which ith criterion influences jth criterion, with n being the number of the considered
criteria.
⎡ 0 a12 ⋯ a1j ⋯ a1n ⎤
⎢ a21 0 ⋯ a2j ⋯ a2n ⎥
⎢ ⎥
A=⎢
⋯ ⋯ ⋯ ⋯ ⋯ ⋯⎥
⎢ ai1 ai2 ⋯ aij ⋯ ain ⎥
⎢⋯ ⋯ ⋯ ⋯ ⋯ ⋯⎥
⎢ ⎥
⎣ an1 an2 ⋯ anj ⋯ 0 ⎦

Step 2 Normalization of the direct-relation matrix

The above matrix is now normalized using Eq. (1), where the elements of matrix X
have values between 0 and 1.
Y = k.A (1)
where
1
k= � � ( i, j = 1, 2, … , n)
∑n
max aij
1≤i≤n j=1

13
OPSEARCH

Step 3 Estimation of the total-relation matrix (T)

The total-relation matrix (T) is calculated using Eq. (2), where I denotes the iden-
tity matrix. Each element tij in matrix T represents the indirect effect of ith criterion
on jth criterion and the matrix T denotes the total relationship within each pair of
criteria.
[ ]
T = tij n×n ( i, j = 1, 2, … , n) (2)

( )[ ]
T = Y + Y 2 + Y 3 + ⋯ + Y k = Y I + Y + Y 2 + ⋯ + Y k−1 (I − Y)(I − Y)−1
( )
= Y I − Y k (I − Y)−1

Then
T = Y(I − Y)−1 , when k → 𝛼, Y k = [0]n×n

T = Y(I − Y)−1
Step 4 Compute the sums of rows and columns of matrix T
The sum of row values and sum of column values of the total-relation matrix T
are represented by vectors R and C, and are calculated applying Eqs.  (3) and (4)
respectively.
[ n ]
∑ [ ]
Ri = tij = ti n×1 , i = 1, 2, … , n (3)
j=1 n×1

[ ]
∑
n
[ ]
Cj = tij = tj n×1 , j = 1, 2, … , n (4)
i=1 1×n

Step 5 Define a threshold value (α)

The threshold value (α) is derived from the average of the elements in the total-
relation matrix T, as calculated using Eq. (5).
n n � �
∑ ∑
tij
i=1 j=1 (5)
𝛼=
N

where N represents the total number of elements in matrix T. It basically enables the
decision maker in removing some minor effect elements from matrix T.

Step 6 Formation of the casual diagram

The vector (Rk+ Ck) in horizontal axis termed as ‘prominence’, expresses the
importance of a criterion and is calculated by adding R with C. Likewise, the vertical
axis vector (Rk − Ck), styled as ‘relation’, segregates various criteria into two groups,
i.e. a cause group and an effect group, and is computed by subtracting C from R. If
Ri is the sum of all the elements of ith row in matrix T, then Ri summarizes both the

13
 OPSEARCH

direct effect as well as the indirect effect exerted by ith criterion over the other crite-
ria. On the contrary, Cj indicates the sum of all the elements of jth column in matrix
T, which represents both the direct and indirect effects exerted by jth criterion over
the other criteria. Hence, (Rk+ Ck) signifies the level of significance that ith criterion
exerts in the entire system. On the other hand, (Rk − Ck) defines the net effect that ith
criterion imparts on the total system.
Usually, when (Rk − Ck) becomes positive, it designates a criterion to belong to
the cause group, and if (Rk − Ck) is negative, the corresponding criterion goes to the
effect group. Thus, a casual diagram, also known as the cause and effect diagram,
can be developed while aligning all the coordinate sets of (Rk+ Ck, Rk − Ck) to visual-
ize the complex interrelationships in a decision making problem. It provides infor-
mation to identify the most important criteria and their influence levels in order to
recognize the driving variables of the considered problem in a complex system, and
design for reasonable measures to take care of the issues based on the attribute type
and influence level.

3.2 SIR method

Superiority and inferiority scores were first introduced to MCDM models by

Rebai [28, 29]. These scores are estimated by contrasting different criteria values
(responses). For instance, let the ordinal scores for two alternatives A and Aʹ are
determined corresponding to a criterion g (maximization type) utilizing the follow-
ing preference structure {P, I}:
( ) ( )
APA� A is preferred to A� iff g(A) > g A�
( ) ( )
AIA� A is indifferent to A� iff g(A) > g A�
where g(A) and g(Aʹ) represent the values of criterion g for alternatives A and Aʹ
respectively. The differentiation of cardinal data is more convoluted than ordinal
data, which can be attributed to the effects of incorrectness, indetermination and
lack of certainty in the evaluation process of criteria values [47]. Various meth-
ods have been proposed to address and resolve such complicated problems, either
by introducing limits [32] or by calculating differences between criteria values by
means of generalized criteria [3]. The SIR method is a conjecture of the representa-
tion of superiority scores and inferiority scores as defined by Rebai [28, 29], while
taking into account distances between criteria values and varying types of general-
ized criterion [3]. These generalized criterion types are utilized to identify the char-
acteristics of functions signifying the specified criterion. The steps involved in the
SIR method, when all the criteria need to be maximized, are presented as below [26,
39, 43]:

Step 1 Estimation of intensity of preference

Let A1, A2,…, Am be m number of alternatives, and g1, g2,…, gn be n number of
cardinal criteria with gj(Ai) be the performance value for ith alternative against jth crite-
rion. Thus, the following decision matrix can be formed.

13
OPSEARCH

⎡ g1 (A1 ) ⋯ gj (A1 ) ⋯ gn (A1 ) ⎤

⎢ ⋯ ⋯ ⋯ ⋯ ⋯ ⎥
⎢ ⎥
G = ⎢ g1 (Ai ) ⋯ gj (Ai ) ⋯ gn (Ai ) ⎥
⎢ ⋯ ⋯ ⋯ ⋯ ⋯ ⎥
⎢ g (A ) ⋯ gj (Am ) ⋯ gn (Am ) ⎥⎦
⎣ 1 m

In this step, the values of each criterion are compared to estimate the intensity of
preference. Consider two alternatives A and Aʹ, and a criterion g, with g(A) and g(Aʹ) be
the values of the criteria for alternative A and Aʹ respectively. The intensity of prefer-
ence for A over Aʹ is given by the difference between their criteria values, as shown in
Eq. (6):
( ) ( ( ))
P A, A� = f (r) = f g(A)−g A� (6)
Brans et al. [3] introduced six generalized criteria, as shown in Table 1, which can
be used to identify characteristics of the functions that illustrate different attitudes
towards the preference structure and intensity of preference. The type of the general-
ized criterion is to be selected by the decision maker based on the attitude and intensity
of preference for the specified decision making problem.

Step 2 Calculation of superiority and inferiority indexes

The superiority index Sj(Ai) and inferiority index Ij(Ai) for alternative Ai with respect
to jth criterion can be respectively computed using Eqs. (7) and (8).

( ) ∑ m
( ) ∑m
( ( ) ( ))
Sj Ai = P Ai , Ak = fj gj Ai − gj Ak (7)
k=1 k=1

( ) ∑ m
( ) ∑m
( ( ) ( ))
Ij Ai = P Ak , Ai = fj gj Ak − gj Ai (8)
k=1 k=1

The calculated indexes are then translated in the form of superiority matrix
(S-matrix) and inferiority matrix (I-matrix). The S-matrix shows the intensity of superi-
ority that each alternative has on each criterion, while I-matrix informs about the inten-
sity of inferiority that each alternative has on each criterion.

⎡ S1 (A1 ) S2 (A1 ) ⋯ Sn (A1 ) ⎤

⎢ S (A ) S (A ) ⋯ Sn (A2 ) ⎥
S=⎢ 1 2 2 2
⋯ ⋯ ⎥
⎢ ⎥
⋯ ⋯
⎣ S1 (Am ) S2 (Am ) ⋯ Sn (Am ) ⎦

⎡ I1 (A1 ) I2 (A1 ) ⋯ In (A1 ) ⎤

⎢ I (A ) I (A ) ⋯ In (A2 ) ⎥
I=⎢ 1 2 2 2
⋯ ⋯ ⎥
⎢ ⎥
⋯ ⋯
⎣ I1 (Am ) I2 (Am ) ⋯ In (Am ) ⎦
13
 OPSEARCH

Table 1  Six types of generalized Criterion Shape

criterion [3]
Type 1: true criterion
{
1 if r > 0
f (r) =
0 if r ≤ 0

Type 2: quasi criterion

{
1 if r > a
f (r) =
0 if r ≤ a

Type 3: criterion with linear preference

⎧ 1 if r>b
⎪
f (r) = ⎨ r∕b if 0 < r ≤ b
⎪ 0 if r≤0
⎩
Type 4: level criterion

⎧ 1 if r>b
⎪
f (r) = ⎨ 1∕2 if a < r ≤ b
⎪ 0 if r≤a
⎩
Type 5: criterion with linear preference and
indifference area

⎧ 1 if r>b
⎪
f (r) = ⎨ (r − a)∕(b − a) if a < r ≤ b
⎪ 0 if r≤a
⎩

Type 6: Gaussian criterion

{
1 − exp(−r2 ∕2𝜎 2 ) if r > 0
f (r) =
0 if r ≤ 0

Step 3 Estimation of flows

The superiority and inferiority indexes, as represented in the form of S-matrix
and I-matrix are then aggregated using the standard MCDM aggregation procedures
to form two global preference indexes, i.e. superiority flow (𝜙> (.))(S-flow) and infe-
riority flow (𝜙< (.))(I-flow). The S-flow and I-flow respectively denote the global
intensity of superiority and inferiority of each alternative. The S-flow implies how
an alternative outranks all other alternatives, whereas, I-flow signifies how an alter-
native is outperformed by all other alternatives. The aggregation procedures, such as
simple additive weighting (SAW), TOPSIS etc. are usually employed to obtain the
corresponding S-flows and I-flows. Between these two, SAW is considered to be the
simplest procedure and is typically deployed as a benchmark to compare solutions
derived from other MCDM methods. On the contrary, TOPSIS method is considered
to follow a logical way in solving discrete MCDM problems and is computationally

13
OPSEARCH

more complicated than SAW method [21]. The evaluation procedures of SAW and
TOPSIS methods while calculating S-flow and I-flow are presented as below:
SIR-SAW method The S-flow and I-flow, in this method, are calculated using
Eqs. (9) and (10) respectively, while considering weights of different criteria.

∑
n
𝜙> (Ai ) = wj Sj (Ai ) (9)
j=1

∑
n
𝜙< (Ai ) = wj Ij (Ai ) (10)
j=1

where wj is the weight of jth criterion.

SIR-TOPSIS method Using ideal solution, A+S and negative-ideal solution, A−S  , the
S-flow is calculated from the superiority matrix (S-matrix).
S− (Ai )
𝜙> (Ai ) = (11)
(S− (Ai ) + S+ (Ai ))
where

{ }1∕
( ) ∑n
| ( ( ) )|𝜆 𝜆
+
S Ai = |wj Sj Ai − S+ | ; 0≤𝜆≤∞
| j |
j=1 | |

{ }1∕
( ) ∑n
| ( ( ) )|𝜆 𝜆
S− Ai = |wj Sj Ai − S− | ; 0≤𝜆≤∞
| j |
j=1 | |

( ( ) ( ))
A+S = (S1+ , … , Sn+ ) = max S1 Ai , … , max Sn Ai
i i

( ( ) ( ))
A−S = (S1− , … , Sn− ) = min S1 Ai , … , min Sn Ai
i i

The evaluation procedure is based on varying values of λ (a positive constant),

i.e. λ = 1 (the city block distance), λ = 2 (the Euclidean distance) and λ = 10 (the
distance with a significantly large λ). Similarly, depending on ideal solution (A+I )
and negative-ideal solution (A−I ) , the I-flow is calculated from the inferiority matrix
(I-matrix).
( )
( ) I + Ai
𝜙 Ai = ( ( )
<
( )) (12)
I + Ai + I − Ai

13
 OPSEARCH

where

{ }1∕
( ) ∑n
| ( ( ) )|𝜆 𝜆
+
I Ai = |wj Ij Ai − I + | ; 0≤𝜆≤∞
| j |
j=1 | |

{ }1∕
( ) ∑n
| ( ( ) )|𝜆 𝜆
−
I Ai = |wj Ij Ai − I − | ; 0≤𝜆≤∞
| j |
j=1 | |

( )
A+I = (I1+ , … , In+ ) = min I1 (Ai ), … , min In (Ai )
i i

( ) ( ( ) ( ))
A−I = I1− , … , In− = max I1 Ai , … , max In Ai
i i

The net flow (n-flow) 𝜙n (.) and relative flow (r-flow)𝜙r (.) are now estimated based
on S-flow and I-flow.
( ) ( ) ( )
𝜙n Ai = 𝜙> Ai − 𝜙< Ai (13)
( )
𝜙> Ai
𝜙r (Ai ) = ( ) (14)
𝜙> (Ai ) + 𝜙< Ai

Step 4 Complete and partial rankings

Complete ranking Four complete rankings can( be derived
) from( S-, )I-, n- and r-flows,
( )
as shown in Fig. 3,
( ) and are named as S-ranking ℜ
( > {  , I-ranking})ℜ<  , n-ranking ℜn
and r-ranking ℜr respectively. The S-ranking ℜ> = P> , I> is obtained based on
the declining order of S-flows (𝜙> (Ai )), as shown below:
( )
Ai P> Ak iff 𝜙> (Ai ) > 𝜙> Ak
( )
Ai I> Ak iff 𝜙> (Ai ) = 𝜙> Ak
( { })
The I-ranking ℜ< = P< , I< is attained depending on the increasing order of
I-flows (𝜙< (Ai )).
( )
Ai P< Ak iff 𝜙< (Ai ) < 𝜙< Ak
( )
Ai I< Ak iff 𝜙< (Ai ) = 𝜙< Ak
Similarly, based on the decreasing order of n-flows and r-flows, the n-ranking and
r-ranking are also derived. ( )
Partial( ranking
) The partial ranking takes into account both S-ranking ℜ> and
I-ranking ℜ<  , as shown in Fig. 4, and can be denoted using the following expression:
ℜ = {P, I, R} = ℜ> ∩ ℜ<

13
OPSEARCH

Fig. 3  Complete ranking procedure for the SIR method

For partial ranking, in order to compare any two alternatives, the intersection princi-
ple, as proposed by Brans et al. [3] and Roy et al. [31], can be employed.
The preference relation, P is given as:
( ) ( ) ( )
APA′ iff AP> A′ and AP< A′ or AP> A′ and AI< A′ or AI> A′ and AP< A′
The indifference relation, I is denoted as:
AIA′ iff AI> A′ and AI< A′
The incomparability relation, R is expressed as:
( ) ( )
ARA′ iff AP> A′ and A′ P< A or A′ P> A and AP< A′
To attain green environment during EDM operation, it is always advised to oper-
ate the machining set-up at the most suitable parametric combination in order to
reduce emission of harmful and toxic substances into the environment which may
adversely affect the concerned operators. It has also been noticed that emission of
the individual substances is correlated with the others, which compels to analyze
the effect of each of the responses on the remaining ones. Identification of the best
parametric combination for a green EDM process is considered to be a demanding

Fig. 4  Partial ranking procedure for the SIR method

13
 OPSEARCH

task due to the involvement of a huge number of input machining parameters and
responses. Thus, it corresponds to a multi-response optimization problem where all
the considered responses need to be optimized concurrently. Typically, in manufac-
turing industries, the machining data handbooks as provided by the manufacturers
are consulted for setting of various EDM process parameters, which usually do not
always ensure attainment of a really global parametric mix. To overcome this prob-
lem, the SIR method, being an efficient MCDM tool, is applied here to attain the
optimal parametric combination of a green EDM process. To examine the effects
of each of the responses on the others, DEMATEL technique is also employed. It
is observed that the parametric combination derived while using the SIR method
is superior to that as obtained by the past researchers. The combined application of
DEMATEL and SIR methods for parametric analysis of a green EDM process is
demonstrated through Fig. 5. It is worthwhile to mention here that different group
decision making techniques, like brainstorming, Delphi method, nominal method,
fish bowling and didactic interaction method can also be augmented to subjectively
analyse the effects of the green EDM process parameters on the responses.

4 Parametric analysis of a green EDM process

Availability of a large number of process parameters along with a set of conflicting

responses turns it as one of the complex problems to find out the most significant
process parameters and responses in a green EDM process. It is also inappropriate

Fig. 5  Flowchart for parametric analysis of a green EDM process

13
OPSEARCH

to consider the elements within the EDM operation to be independent of each other.
Thus, it becomes essential to evaluate the relationships between the considered
responses and identify those significant responses affecting the others. In order to
accomplish this task, DEMATEL method is initially employed to study the critical
relationships between different responses of a green EDM process casually and visu-
ally. Moreover, to select the optimal combination of the process parameters, the SIR
method is later adopted for having improved machining quality of the said process to
achieve the concept of sustainable manufacturing.
Based on Taguchi’s L9 orthogonal array, Sivapirakasam et al. [38] performed nine
experiments on a high carbon high chromium tool steel plate of size 4 × 4 × 1.5 cm
using a green EDM process while considering copper rod of diameter 25 mm as an
electrode. Fuzzy-TOPSIS methodology was applied as a multi-response optimiza-
tion technique to derive the optimal parametric mix with peak current (PC) (in A),
pulse duration (PD) (in μs), dielectric level (DL) (in mm) and flushing pressure (FP)
(in kg/cm2) as the input parameters, with three levels each. Five responses, i.e. pro-
cess time (PT) (in s), relative tool wear ratio (RTWR), process energy (PE) (in W),
concentration of aerosol (CA) (in mg/m3) and dielectric consumption (DC) (in ­cm3)
were simultaneously minimized. The considered process parameters along with
their operating levels are presented in Table 2. All the five responses have lower-the-
better characteristics (non-beneficial criteria). The design plan for experiments in
addition to the measured responses is summarized in Table 3.
The application of DEMATEL method starts with the development of the cor-
responding direct-relation matrix. The interrelationships between the considered
responses are estimated using an integer scale ranging between 0 and 4, as discussed
earlier. Thus, the initial 5 × 5 direct-relation matrix (A) is developed by pair-wise
comparisons between the five responses with respect to their influences and direc-
tions, as shown in Table  4. The developed matrix (A) is then normalized while
employing Eq.  (1) so as to bring all of its elements within a comparable range of
0–1, as given in Table  5. The total-relation matrix (T) is subsequently developed
using Eq.  (2), as provided in Table  6. The sum of rows and sum of columns of
matrix (T), as denoted by vectors R and C respectively, are calculated using Eqs. (4)
and (5), and the corresponding results are provided in Table  7. The total and net
effects for each response are then computed, where the total effect (R + C) is the sum
of vectors R and C, and the net effect (R − C) is calculated by subtracting C from R.
The estimated total and net effects (influence levels) for each response are shown
in Table 8. A positive value of the net effect (R − C) for a response indicates that it

Table 2  Green EDM machining Machining parameter Symbol Unit Levels

parameters with their levels [38]
1 2 3

Peak current PC A 2.0 4.5 7.0

Pulse duration PD μs 2 261 520
Dielectric level DL mm 40 60 80
Flushing pressure FP kg/cm2 0.3 0.5 0.7

13
 OPSEARCH

Table 3  Experimental details [38]

Exp. no. Process parameter Response
PC PD DL FP PT RTWR​ PE CA DC

1 1 1 1 1 0.7258 0.3899 54.433 0.82 0.0665

2 1 2 2 2 1.5357 0.0055 115.178 0.77 0.0981
3 1 3 3 3 1.6393 0.0051 122.951 0.64 0.0865
4 2 1 2 3 0.4705 0.3496 79.389 1.22 0.0510
5 2 2 3 1 0.3415 0.0041 57.620 2.13 0.0332
6 2 3 1 2 0.3942 0.0049 66.516 1.98 0.0394
7 3 1 3 2 0.4062 0.3452 106.632 2.40 0.0497
8 3 2 1 3 0.2381 0.0065 62.4884 4.12 0.0351
9 3 3 2 1 0.2646 0.0076 69.469 5.05 0.0434

Table 4  Initial direct-relation Response PT RTWR​ PE CA DC

matrix for the responses
PT 0 2 4 3 3
RTWR​ 2 0 1 2 1
PE 4 3 0 3 3
CA 2 1 3 0 3
DC 1 2 3 2 0

Table 5  Normalized direct- Response PT RTWR​ PE CA DC

relation matrix for the responses
PT 0 0.1538 0.3077 0.2308 0.2308
RTWR​ 0.1538 0 0.0769 0.1538 0.0769
PE 0.3077 0.2308 0 0.2308 0.2308
CA 0.1538 0.0769 0.2308 0 0.2308
DC 0.0769 0.1538 0.2308 0.1538 0

Table 6  Total-relation matrix Response PT RTWR​ PE CA DC

for the responses
PT 0.5501 0.6250* 0.8876* 0.7804* 0.7907*
RTWR​ 0.4228 0.2577 0.4217 0.4492 0.3953
PE 0.8103* 0.6990* 0.6771* 0.8068* 0.8140*
*
CA 0.5637 0.4636 0.7009 0.4656 0.6657*
*
DC 0.4580 0.4742 0.6280 0.5408 0.4119

*tij> 0.5904

13
OPSEARCH

Table 7  Computation of vectors Response Rk Ck

R and C 
PT 3.6338 2.8049
RTWR​ 1.9467 2.5195
PE 3.8071 3.3154
CA 2.8596 3.0427
DC 2.5129 3.0776

influences all other responses more than the other influencing it, whereas, a negative
value signifies that the response is significantly influenced by the other responses.
A pictorial view, as depicted in Fig. 6, also known as the casual diagram, is devel-
oped by delineating the dataset of Table 8. Thus, it can be propounded from Table 8
that PT has the highest net influence level, followed by PE for the considered green
EDM process.
It is quite expected that PT with its highest level of influence would directly dom-
inate all the other responses. The longer the duration of machining, the larger would
be the consumption of process energy along with an increase in the amount of die-
lectric consumption. Moreover, as the machining time elapses, amount of tool and
work material erosion also increases, leading to an increase in RTWR and CA. On
the other hand, an increase in PE enhances the amount of melting and vaporization
of materials, thereby increasing RTWR. An increase in erosion of materials from
the tool and the workpiece into the dielectric fluid, along with excess emission of
toxic gases, also results in an increase in CA. Besides, higher DC is attributed to the
generation of significant amount of heat that intensifies causing an increase in PE.
Generation of high discharge energy also leads to initiation of large amount of heat,
which consequently increases the amount of evaporation of the dielectric fluid from
the machining zone.
The considered responses can also be split into cause and effect groups, as
depicted in Fig.  6, where PT and PE with positive influence levels constitute the
cause group, and RTWR, DC and CA having negative influence levels are added to
the effect group. It is quite apparent that PT and PE are the key driving parameters
for the remaining responses. The response PE with the highest (R + C) value is iden-
tified as the most significant one as it has the utmost intensity of relativeness as com-
pared to the other responses. However, PT with the highest (R − C) value is identi-
fied as the most influencing response that affects all the other responses. Thus, PT

Table 8  Total and net effects of Response R + C R − C Criteria group

the responses
PT 6.4387 0.8289 Cause
RTWR​ 4.4662 − 0.5728 Effect
PE 7.1225 0.4917 Cause
CA 5.9022 − 0.1831 Effect
DC 5.5904 − 0.5647 Effect

13
 OPSEARCH

Fig. 6  DEMATEL casual diagram for the responses

and PE play significant roles in this performance assessment problem of the green
EDM process as they have the greatest influences on the other responses. On the
contrary, RTWR with the maximum negative value of (R − C) is greatly influenced
by other responses. The threshold value (α) is now calculated by averaging the ele-
ments in the total-relation matrix (T) as 0.5904. The values of tij in matrix T are then
compared with that threshold value. If a value in the matrix T is greater than 0.5904,
the element is then marked as t*ij, which signifies that there is a significant interac-
*
tion present between the two responses. For instance, as t12 (0.6250) > α (0.5904), an
arrow in the diagraph is drawn from PT to RTWR, as shown in Fig. 7. The arrows in
this digraph depict the contextual interactions among different elements of the green
EDM process. There are bi-directional interactions between PT and PE, PE and CA,
and PE and DC. On the other hand, the remaining interactions are unidirectional.
Further, in order to achieve the optimal parametric combination for the green
EDM process, the SIR method is now employed. As discussed earlier, the type of
the generalized criterion for each of the responses along with their relevant param-
eter values is selected, as shown in Table 9. Two separate scenarios are considered
here, where scenario 1 (S1) assumes equal weights to all the responses and scenario
2 (S2) adopts weights estimated using the entropy method [18]. The normalized
weights for the five responses derived while employing entropy method are also pro-
vided in Table 9. It is interesting noticed that RTWR has the maximum importance,
followed by PE. Now, applying Eqs. (7) and (8), the S-matrix and I-matrix are sub-
sequently developed, and the results are tabulated in Table 10.
SIR-SAW: The S-flows and I-flows are calculated for both the scenarios (S1 and
S2), based on Eqs. (9) and (10) respectively. Using Eqs. (13) and (14), the n-flow

13
OPSEARCH

Fig. 7  DEMATEL diagraph for

the green EDM process

Table 9  Characteristics of the Response PT RTWR​ PE CA DC

selected responses
Type of criterion Type 4 Type 1 Type 4 Type 3 Type 1
Parameter a = 0 a = 0 b = 2
b = 1 b = 40
Weight (S1) 0.20 0.20 0.20 0.20 0.20
Entropy weight (S2) 0.1912 0.2555 0.2199 0.1619 0.1715

Table 10  S-matrix and I-matrix

Exp. no. S-matrix I-matrix
PT RTWR​ PE CA DC PT RTWR​ PE CA DC

1 1 0 5.5 4.225 2 3 8 0 0.115 6

2 0.5 5 0.5 4.35 0 6.5 3 6 0.065 8
3 0 6 0 4.74 1 7 2 7 0 7
4 2.5 1 2 3.425 3 2.5 7 2.5 0.715 5
5 4 8 5 2.13 8 1 0 0.5 2.61 0
6 3.5 7 4 2.285 6 1.5 1 1.5 2.235 2
7 3 2 1 1.86 4 2 6 5 3.42 4
8 5 4 4.5 0.465 7 0 4 1 6.855 1
9 4.5 3 3 0 5 0.5 5 2 7.465 3

and r-flow values for each experimental run are also calculated. These values of
S-, I-, n- and r-flows for both the scenarios are presented in Table  11. The partial
ranking relationships for the considered scenarios along with the partial ranking dia-
grams are provided in Table 12 and Fig. 8 respectively. It can be clearly observed
from Fig. 8 that for both the scenarios with different weight assignments, experiment
trial number 5 with a parametric combination of A ­ 2B2C3D1 is identified as the best
experimental run among the nine conducted trials, whereas, experimental trial num-
ber 2 (­ A1B2C2D2) is the least preferred choice. For both the scenarios, experimental

13
 OPSEARCH

Table 11  SIR flows using SAW procedure

Exp. no. S-flow I-flow n-flow r-flow
S1 S2 S1 S2 S1 S2 S1 S2

1 2.545 2.4277 3.423 3.6652 − 0.878 − 1.2375 0.4264 0.3984

2 2.07 2.1873 4.713 4.7112 − 2.643 − 2.5239 0.3052 0.3171
3 2.348 2.4719 4.6 4.5892 − 2.252 − 2.1173 0.3379 0.3501
4 2.385 2.2423 3.543 3.7895 − 1.158 − 1.5472 0.4023 0.3717
5 5.426 5.6251 0.822 0.7237 4.604 4.9014 0.8684 0.886
6 4.557 4.7362 1.647 1.577 2.91 3.1592 0.7345 0.7502
7 2.372 2.2916 4.084 4.2546 − 1.712 − 1.963 0.3674 0.3501
8 4.193 4.2433 2.571 2.5232 1.622 1.7201 0.6199 0.6271
9 3.1 3.1441 3.593 3.536 − 0.493 − 0.3919 0.4632 0.4707

Table 12  Partial ranking relationships between the experimental trials using SIR-SAW procedure

For scenario 1 5 > 1 6 > 1 8 > 1 9 ~ 1 1 > 2 4 > 2 7 > 2 3 > 2
5 > 2 6 > 2 8 > 2 9 > 2 1 > 3 4 > 3 7 > 3
5 > 3 6 > 3 8 > 3 9 > 3 1 > 4 4 > 7
5 > 4 6 > 4 8 > 4 9 ~ 4 1 > 7
5 > 6 6 > 7 8 > 7 9 > 7
5 > 7 6 > 8 8 > 9
5 > 8 6 > 9
5 > 9
For scenario 2 5 > 1 6 > 1 8 > 1 9 > 1 3 ~ 1 1 > 2 7 > 2 4 > 2
5 > 2 6 > 2 8 > 2 9 > 2 3 > 2 1 > 4 7 ~ 4
5 > 3 6 > 3 8 > 3 9 > 3 3 ~ 4 1 > 7
5 > 4 6 > 4 8 > 4 9 > 4 3 ~ 7
5 > 6 6 > 7 8 > 7 9 > 7
5 > 7 6 > 8 8 > 9
5 > 8 6 > 9
5 > 9

trial number 5 is followed by trial numbers 6 and 8 as the best candidate parametric
mixes. The average values of the calculated r-flows at different levels of operation of
the green EDM process parameters are calculated for both the scenarios, as shown
in Table 13. These average r-flow values are now treated as the responses to search
out the possibility of any better parametric mix for the considered process. From
the response tables for both the weight scenarios, the best combination of the green
EDM process parameters (shown in bold faced) is identified as ­A2B2C1D1 which is
different from experiment number 5 only with respect to dielectric level. In experi-
ment number 5, the dielectric level is set at 80 mm, whereas, in the proposed para-
metric combination, it is required to be fixed at 40 mm. This is because of the fact

13
OPSEARCH

Fig. 8  Partial ranking diagrams using SIR-SAW procedure

Table 13  Response table Process Level 1 Level 2 Level 3 Max–Min Rank

for r-flows using SIR-SAW parameter
procedure
(a) Scenario (S1)
PC 0.3565 0.6684 0.4835 0.3119 1
PD 0.3987 0.5978 0.5119 0.1991 3
DL 0.5936 0.3902 0.5246 0.2034 2
FP 0.5860 0.4690 0.4534 0.1326 4
(b) Scenario (S2)
PC 0.3552 0.6693 0.4826 0.3141 1
PD 0.3734 0.6101 0.5237 0.2367 2
DL 0.5919 0.3865 0.5287 0.2054 3
FP 0.5850 0.4725 0.4496 0.1354 4

that the average of r-flows at level 1 of dielectric level is more than that of level
3. Thus, the chance of having higher value of r-flow is always with level 1 than
with level 3. Lower value of dielectric level leads to its more controlled use caus-
ing less emission of the toxic gases from the machining zone. Hence, in order to
attain the best machining performance, the parametric combination is to be set as
peak current = 4.5 A, pulse duration = 261 μs, dielectric level = 40 mm and flushing
pressure = 0.3 kg/cm2. The max–min column in Table 13 with a maximum value for
peak current identifies it as the most influential input parameter for the considered
green EDM process.
Note i > j denotes alternative i outperforms alternative j; and i ~ j signifies alterna-
tive i is not comparable to alternative j

13
 OPSEARCH

SIR-TOPSIS The S-, I-, n- and r-flows are also computed using TOPSIS method
for both the scenarios, while taking λ = 1, 2 and 10. These values are provided in
Table 14. The corresponding partial ranking diagrams are shown in Fig. 9. Similar
to SIR-SAW procedure, it can be observed from Fig. 9 that for both the weight sce-
narios, experiment number 5 with a parametric combination of A ­ 2B2C3D1 is identi-
fied to be the best choice. The average values of the estimated r-flows using SIR-
TOPSIS procedure at different λ values are now calculated. From those response
values, it can be revealed that with λ as 1 and 10 in scenario S1, the optimal para-
metric combination is derived as ­A2B2C1D1 for concurrent optimization of all the

Table 14  SIR flows using TOPSIS procedure

Exp. no. S-flow I-flow n-flow r-flow
S1 S2 S1 S2 S1 S2 S1 S2

For city block distance (λ = 1)

1 0.4073 0.3824 0.5518 0.5677 − 0.1445 − 0.1853 0.4247 0.4025
2 0.3313 0.3445 0.6196 0.6179 − 0.2883 − 0.2734 0.3484 0.358
3 0.3758 0.3893 0.538 0.5528 − 0.1622 − 0.1634 0.4112 0.4133
4 0.3817 0.3532 0.3469 0.3586 0.0348 − 0.0054 0.5239 0.4962
5 0.8684 0.886 0.1233 0.1088 0.7452 0.7772 0.8757 0.8906
6 0.7294 0.746 0.3258 0.3269 0.4036 0.4191 0.6913 0.6953
7 0.3796 0.361 0.4535 0.4608 − 0.0738 − 0.0998 0.4557 0.4393
8 0.6711 0.6684 0.3973 0.3888 0.2738 0.2795 0.6281 0.6322
9 0.4962 0.4952 0.3241 0.3203 0.1721 0.1749 0.6049 0.6072
For Euclidean distance (λ = 2)
1 0.4033 0.3742 0.537 0.5717 − 0.1338 − 0.1974 0.4289 0.3956
2 0.3799 0.4108 0.5609 0.5504 − 0.181 − 0.1397 0.4038 0.4274
3 0.4257 0.4604 0.5563 0.5735 − 0.1306 − 0.1131 0.4335 0.4453
4 0.3681 0.3196 0.3809 0.4005 − 0.0128 − 0.0808 0.4915 0.4439
5 0.8227 0.8556 0.1776 0.1479 0.6451 0.7077 0.8224 0.8526
6 0.7342 0.7647 0.3202 0.3206 0.414 0.4441 0.6964 0.7046
7 0.3862 0.3493 0.4398 0.463 − 0.0537 − 0.1137 0.4675 0.43
8 0.6357 0.6233 0.4448 0.4318 0.1909 0.1915 0.5883 0.5908
9 0.5006 0.4859 0.3661 0.3557 0.1345 0.1302 0.5776 0.5774
For distance with a sufficiently large λ (λ = 10)
1 0.4077 0.3718 0.5204 0.6054 − 0.1126 − 0.2336 0.4393 0.3805
2 0.3896 0.4794 0.5188 0.488 − 0.1292 − 0.0086 0.4289 0.4955
3 0.4609 0.5416 0.5356 0.5825 − 0.0748 − 0.0409 0.4625 0.4818
4 0.3338 0.2474 0.4481 0.4655 − 0.1143 − 0.2181 0.4269 0.3471
5 0.7667 0.829 0.2643 0.1907 0.5024 0.6383 0.7436 0.813
6 0.7416 0.8128 0.3112 0.2885 0.4304 0.5243 0.7044 0.7381
7 0.3996 0.3131 0.446 0.5143 − 0.0463 − 0.2012 0.4726 0.3784
8 0.6122 0.5488 0.51 0.4738 0.1022 0.0751 0.5455 0.5367
9 0.4961 0.423 0.4556 0.3932 0.0405 0.0298 0.5213 0.5183

13
OPSEARCH

Fig. 9  Partial ranking diagrams using SIR-TOPSIS procedure

13
 OPSEARCH

five responses. However, from the response values evaluated forscenario S1 at λ = 2,
and for scenario S2 at λ = 1, 2 and 10, the optimal parametric combination is derived
as ­A2B2C3D1, which is quite similar to that of experiment number 5 for the con-
sidered green EDM process. On the contrary, Sivapirakasam et  al. [38] observed
the optimal parametric combination as ­A2B2C1D2 while employing fuzzy-TOPSIS
method.
For fully justifying the superiority of the derived optimal combination of the
process parameters over that as obtained by Sivapirakasam et  al. [38], the follow-
ing regression equations are developed for the five responses while considering
the main and interaction effects for the said green EDM process. Based on these
regression equations, the estimated response values at different parametric com-
binations are compared with those as predicted by Sivapirakasam et  al. [38], as
exhibited in Table 15. From this table, it can be clearly concluded that the response
values derived at the optimal parametric combination are significantly better than
those obtained by the past researchers. It is also interestingly noticed that grey-AHP
method also provides the same parametric combination as that of SIR method.
PT =1.67−0.41 PC−0.00032 × PD−0.0198 × DL + 1.24 × FP−0.00010
(15)
× PC × PD + 0.0044 × PC × DL + 0.000027 × PD × DL

REWR = 1.679−0.363 × PC−0.001720 × PD−0.0327 × DL + 0.833 × FP + 0.0000645

× PC × PD + 0.00612 × PC × DL + 0.000022 × PD × DL
(16)
PE = 120−37.8 × PC + 0.040 × PD−1.88 × DL + 141 × FP−0.0004
(17)
× PC × PD + 0.580 × PC × DL + 0.00050 × PD × DL

CA = −0.14 + 0.965 × PC−0.00595 × PD + 0.0027 × DL + 1.63 × FP + 0.00054

× PC × PD−0.0067 × PC × DL−0.000082 × PD × DL
(18)
DC = 0.147−0.0333 × PC−0.000078 × PD−0.00174 × DL + 0.090 × FP + 0.000009
× PC × PD + 0.000400 × PC × DL + 0.000001 × PD × DL
(19)
In order to highlight the contribution and novelty of this paper, Table 16 com-
pares its contribution with that of the past researchers towards parametric opti-
mization of green EDM processes. The earlier works have mainly emphasized on

Table 15  Comparison of the derived response values

Parametric combination PT RTWR​ PE CA DC

A2B2C1D1 (SIR method) 0.2779 0.0032 36.5902 1.8187 0.0272

A2B2C1D1 (grey-AHP method) 0.2779 0.0032 36.5902 1.8187 0.0272
A2B2C3D1 (experiment trial 5) 0.3415 0.0041 57.62 2.13 0.0332
A2B2C1D2 [38] 0.5259 0.1698 64.7902 2.1447 0.0452

13
 OPSEARCH

Table 16  Optimization techniques applied for green EDM processes

Sl. no. Authors Optimization techniques Process parameters Responses Outcome

1. Jagadish and Ray [19] Entropy-GRA approach Discharge current, PD, DL Energy consumed, PT, RTWR, Optimal parametric combination
and FP CA, DC
2. Jagadish and Ray [20] PCA-GRA approach Discharge current, PD, DL Energy consumed, PT, RTWR, Optimal parametric mix
and FP CA, DC
3. Sivapirakasam et al. [38] Fuzzy-TOPSIS approach PC, PD, DL and FP PT, RTWR, PE, CA, DC Optimal parametric combination
4. Tang and Du [41] GRA-Taguchi method Discharge current, gap voltage, Electrode wear ratio, MRR, SR Optimal parametric mix
lifting height, negative polar-
ity, pulse duty factor
5. Tang and Du [42] GRA-Taguchi method PC, open circuit voltage, lifting Tool electrode wear rate, MRR, Optimal parametric mix
height, pulse width, pulse SR
intermission
6. Present paper DEMATEL, entropy and SIR PC, PD, DL and FP PT, RTWR, PE, CA, DC Optimal parametric combina-
method tion along with effects of each
response over the others

13
 OPSEARCH

the applications of different combinations of mathematical tools solely for deter-

mination of the best parametric mixes of the considered processes. In this paper,
entropy method is applied for estimation of weights for the five responses, the
interrelationships among those responses are portrayed using DEMATEL method
in the form of causal diagram and finally, the optimal parametric mix is identified
based on the SIR method. Besides parametric optimization, it thus attempts to
divide the considered responses into cause and effect groups, and envisage how a
particular response is being influenced by the others.

5 Conclusions

In this paper, a novel technique combining DEMATEL and SIR methods is

applied to investigate and describe the interaction relationships and influence
levels between different responses as well as to achieve the optimal parametric
combination for a green EDM process. The DEMATEL method assists in devel-
opment of the casual relationship diagram, thereby partitioning the considered
responses into the corresponding cause and effect groups. The results obtained
using DEMATEL method show that process time and process energy play signifi-
cant roles in this performance evaluation problem having the greatest influences
on the other responses. Although relative tool wear ratio is the most important
response, it is maximally influenced by the other responses. On the other hand,
the SIR method is adopted to determine the optimal parametric combination of
the green EDM process for having increased performance with minimum haz-
ardous effect on the environment. It is found that the adopted approach provides
better parametric combination for the said green EDM process, which is also
confirmed with the results estimated utilizing the developed regression equations
for each of the responses. For the considered process, moderate values of peak
current and pulse duration, and low values of dielectric level and flushing pres-
sure lead to simultaneous attainment of all the target responses. This approach
thus outperforms other popular methods while proving its competency as a multi-
objective optimization tool. It is simple to apply, easy to apprehend and free from
any complex mathematical calculations. As the entire calculation is based on the
data obtained from the past literature, there is no scope here for any confirmatory
experiment for real time validation of the derived results. It is also noticed that
the performance of SIR method in finding out the optimal parametric mix of the
considered green EDM process is quite similar to that of grey-AHP approach.
This approach can also be augmented to other conventional as well as non-con-
ventional machining processes for their parametric analyses to derive the optimal
settings of their input parameters leading to improved machining performance.

References
1. Abbas, N.M., Solomon, D.G., Bahari, M.F.: A review on current research trends in electrical discharge
machining (EDM). Int. J. Mach. Tools Manuf. 47(7–8), 1214–1228 (2007)

13
OPSEARCH

2. Bhuyan, R., Routara, B.: Optimization the machining parameters by using VIKOR and entropy weight
method during EDM process of Al-18% SiCp metal matrix composite. Decis. Sci. Lett. 5(2), 269–282
(2016)
3. Brans, J.P., Vincke, P., Mareschal, B.: How to select and how to rank projects: the PROMETHEE method.
Eur. J. Oper. Res. 24(2), 228–238 (1986)
4. Chakraborty, S., Das, P.P.: A multivariate quality loss function approach for parametric optimization of
non-traditional machining processes. Manag. Sci. Lett. 8(8), 873–884 (2018)
5. Chakraborty, S., Das, P.P., Kumar, V.: Application of grey-fuzzy logic technique for parametric optimiza-
tion of non-traditional machining processes. Grey Syst.: Theory Appl. 8(1), 46–68 (2018)
6. Chen, Y.C., Lien, H.P., Tzeng, G.H.: Measures and evaluation for environment watershed plans using a
novel hybrid MCDM model. Expert Syst. Appl. 37(2), 926–938 (2010)
7. Choi, A.C.K., Kaebernick, H., Lai, W.H.: Manufacturing processes modelling for environmental impact
assessment. J. Mater. Process. Technol. 70(1–3), 231–238 (1997)
8. Das, P.P., Chakraborty, S.: Parametric optimization of non-traditional machining processes using Taguchi
method and super ranking concept. Yugosl. J. Oper. Res. (2018). https​://doi.org/10.2298/YJOR1​80821​
033D
9. Dewangan, S., Gangopadhyay, S., Biswas, C.K.: Multi-response optimization of surface integrity char-
acteristics of EDM process using grey-fuzzy logic-based hybrid approach. Eng. Sci. Technol., Int. J.
18(3), 361–368 (2015)
10. El-Taweel, T.A.: Multi-response optimization of EDM with Al–Cu–Si–TiC P/M composite electrode.
Int. J. Adv. Manuf. Technol. 44(1–2), 100–113 (2009)
11. Fontela, E., Gabus, A.: The DEMATEL Observer, DEMATEL 1976 Report. Battelle Geneva Research
Center, Geneva (1976)
12. Gabus, A., Fontela, E.: Perceptions of the world problematique: communication procedure, communi-
cating with those bearing collective responsibility (No. 1). DEMATEL Report (1973)
13. Gopalakannan, S., Senthilvelan, T.: Optimization of machining parameters for EDM operations based
on central composite design and desirability approach. J. Mech. Sci. Technol. 28(3), 1045–1053 (2014)
14. Govindan, K., Rajendran, S., Sarkis, J., Murugesan, P.: Multi criteria decision making approaches for
green supplier evaluation and selection: a literature review. J. Clean. Prod. 98, 66–83 (2015)
15. Hamidi, N., Yousefi, P., Rahimi, A., Jabari, F.: A hybrid of Borda and DEMATEL for productivity
improvement. Manag. Sci. Lett. 2(8), 2757–2764 (2012)
16. Ho, K.H., Newman, S.T.: State of the art electrical discharge machining (EDM). Int. J. Mach. Tools
Manuf. 43(13), 1287–1300 (2003)
17. Ho, W., Xu, X., Dey, P.K.: Multi-criteria decision making approaches for supplier evaluation and selec-
tion: a literature review. Eur. J. Oper. Res. 202(1), 16–24 (2010)
18. Hwang, C.L., Yoon, K.: Multiple Attribute Decision Making Methods and Applications. Springer, Ber-
lin (1981)
19. Jagadish, Ray, A.: Multi-objective optimization of green EDM: an integrated theory. J. Inst. Eng.
(India): Ser. C 96(1), 41–47 (2015)
20. Jagadish, Ray, A.: Optimization of process parameters of green electrical discharge machining using
principal component analysis (PCA). Int. J. Adv. Manuf. Technol. 87(5–8), 1299–1311 (2016)
21. Janic, M., Reggiani, A.: An application of the multiple criteria decision making (MCDM) analysis to
the selection of a new hub airport. Eur. J. Transp. Infrastruct. Res. 2(2), 113–141 (2002)
22. Joshi, S.N., Pande, S.S.: Intelligent process modeling and optimization of die-sinking electric discharge
machining. Appl. Soft Comput. 11(2), 2743–2755 (2011)
23. Kung, K.Y., Horng, J.T., Chiang, K.T.: Material removal rate and electrode wear ratio study on the
powder mixed electrical discharge machining of cobalt-bonded tungsten carbide. Int. J. Adv. Manuf.
Technol. 40(1–2), 95–104 (2009)
24. Kuo, T.C., Chang, S.H., Huang, S.H.: Environmentally conscious design by using fuzzy multi-attribute
decision-making. Int. J. Adv. Manuf. Technol. 29(3–4), 209–215 (2006)
25. Liu, F., Zhang, H.: A decision-making framework model of green manufacturing. Chin. J. Mech. Eng.
35, 11–15 (1999)
26. Marzouk, M.: A superiority and inferiority ranking model for contractor selection. Constr. Innov. 8(4),
250–268 (2008)
27. Mukherjee, R., Chakraborty, S.: Selection of EDM process parameters using biogeography-based opti-
mization algorithm. Mater. Manuf. Process. 27(9), 954–962 (2012)
28. Rebai, A.: BBTOPSIS: a bag based technique for order preference by similarity to ideal solution. Fuzzy
Sets Syst. 60(2), 143–162 (1993)

13
 OPSEARCH

29. Rebai, A.: Canonical fuzzy bags and bag fuzzy measures as a basis for MADM with mixed non cardi-
nal data. Eur. J. Oper. Res. 78(1), 34–48 (1994)
30. Reddy, V.V., Valli, P.M., Kumar, A., Reddy, C.S.: Multi-objective optimization of electrical discharge
machining of PH17-4 stainless steel with surfactant-mixed and graphite powder-mixed dielectric using
Taguchi-data envelopment analysis-based ranking method. Proc. Inst. Mech. Eng., Part B: J. Eng.
Manuf. 229(3), 487–494 (2015)
31. Roy, B., Slowinski, R., Treichel, W.: Multicriteria programming of water supply systems for rural areas.
J. Am. Water Resour. Assoc. 28(1), 13–31 (1992)
32. Roy, B.: Multicriteria Methodology for Decision Aiding, vol. 12. Springer, Berlin (2013)
33. Roy, T., Dutta, R.K.: Integrated fuzzy AHP and fuzzy TOPSIS methods for multi-objective optimi-
zation of electro discharge machining process. Soft Comput. (2018). https​://doi.org/10.1007/s0050​
0-018-3173-2
34. Sheng, P., Srinivasan, M., Kobayashi, S.: Multi-objective process planning in environmentally con-
scious manufacturing: a feature-based approach. CIRP Ann.—Manuf. Technol. 44(1), 433–437 (1995)
35. Singh, J., Sharma, R.K.: Green EDM strategies to minimize environmental impact and improve process
efficiency. J. Manuf. Sci. Prod. 16(4), 273–290 (2016)
36. Singh, N.K., Pandey, P.M., Singh, K.K., Sharma, M.K.: Steps towards green manufacturing through
EDM process: A review. Cogent Eng. 3(1), 13 (2016). https​://doi.org/10.1080/23311​916.2016.12726​62
37. Singh, P.N., Raghukandan, K., Pai, B.C.: Optimization by grey relational analysis of EDM parameters
on machining Al-10% S ­ iCP composites. J. Mater. Process. Technol. 155, 1658–1661 (2004)
38. Sivapirakasam, S.P., Mathew, J., Surianarayanan, M.: Multi-attribute decision making for green electri-
cal discharge machining. Expert Syst. Appl. 38(7), 8370–8374 (2011)
39. Tam, C.M., Tong, T.K., Wong, Y.W.: Selection of concrete pump using the superiority and inferiority
ranking method. J. Constr. Eng. Manag. 130(6), 827–834 (2004)
40. Tan, X.C., Liu, F., Cao, H.J., Zhang, H.: A decision-making framework model of cutting fluid selection
for green manufacturing and a case study. J. Mater. Process. Technol. 129(1–3), 467–470 (2002)
41. Tang, L., Du, Y.T.: Experimental study on green electrical discharge machining in tap water of Ti–6Al–
4V and parameters optimization. Int. J. Adv. Manuf. Technol. 70(1–4), 469–475 (2014)
42. Tang, L., Du, Y.T.: Multi-objective optimization of green electrical discharge machining Ti–6Al–4V in
tap water via grey-Taguchi method. Mater. Manuf. Process. 29(5), 507–513 (2014)
43. Tavana, M., Zareinejad, M., Santos-Arteaga, F.J.: An intuitionistic fuzzy-grey superiority and inferior-
ity ranking method for third-party reverse logistics provider selection. Int. J. Syst. Sci.: Oper. Logist.
5(2), 175–194 (2018)
44. Tönshoff, H.K., Egger, R., Klocke, F.: Environmental and safety aspects of electrophysical and electro-
chemical processes. CIRP Ann. 45(2), 553–568 (1996)
45. Wang, X., Chen, L., Dan, B., Wang, F.: Evaluation of EDM process for green manufacturing. Int. J.
Adv. Manuf. Technol. 94(1–4), 633–641 (2018)
46. Wu, X., Zhang, S., Qiu, S., Sun, L.: Decision making method of process parameter selection for green
manufacturing based on a DEMATEL-VIKOR algorithm. J. Mech. Eng. 49(7), 91–100 (2013)
47. Xu, X.: The SIR method: a superiority and inferiority ranking method for multiple criteria decision
making. Eur. J. Oper. Res. 131(3), 587–602 (2001)
48. Yeo, S.H., Neo, K.G., Tan, H.C.: Assessment of health hazards in production of printed paper packages.
Int. J. Adv. Manuf. Technol. 14(5), 376–384 (1998)
49. Yeo, S.H., New, A.K.: A method for green process planning in electric discharge machining. Int. J. Adv.
Manuf. Technol. 15(4), 287–291 (1999)

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps
and institutional affiliations.

13