Archives of Computational Methods in Engineering (2022) 29:2285–2308

https://doi.org/10.1007/s11831-021-09663-x

REVIEW ARTICLE

A Review of Multi‑objective Optimization: Methods and Algorithms

in Mechanical Engineering Problems
João Luiz Junho Pereira1 · Guilherme Antônio Oliver1 · Matheus Brendon Francisco1 · Sebastião Simões Cunha Jr1 ·
Guilherme Ferreira Gomes1

Received: 30 December 2020 / Accepted: 1 October 2021 / Published online: 21 October 2021
© CIMNE, Barcelona, Spain 2021

Abstract
The optimization problems that must meet more than one objective are called multi-objective optimization problems and
may present several optimal solutions. This manuscript brings the most important concepts of multi-objective optimization
and a systematic review of the most cited articles in the last years in mechanical engineering, giving details about the main
applied multi-objective optimization algorithms and methods in this field. Some of the applications that can be found in
this study are: (i) problems in design optimization, (ii) problems in manufacturing: welding, machining and molding and
(iii) problems in structural health monitoring. It can be seen that classic optimization methods had their importance in the
past, but lost space for new algorithms that emerged with the advancement of computing, better able to deal with a greater
number of variables, objectives and nonlinearities. These powerful algorithms, still little used in Mechanical Engineering,
showed significant improvement where they were applied. Meta-heuristics with a posteriori decision-making techniques
proved to be a modern trend in solving multi-objective problems, although it is not limited due to the constant battle of new
algorithms more adapted to specific problems.

Abbreviations SPEA (I and II) Strength Pareto Evolutionary

MOP Multi-objective Problems Algorithm
PF Pareto Front MOPSO Multi-objective Particle Swarm
ZDT Zitzler-Deb-Thiele test function Optimization
DM Decision Maker MOGWO Multi-objective Grey Wolf Optimizer
NBI Normal Boundary Intersection MOALO Multi-objective Ant Lion Optimizer
GA Genetic Algorithm OM Orthogonal Method
MOEA Multi-objective Evolutionary FG Functionally Graded
Algorithm TOPSIS Technique to Order of Preference by
VEGA Vector Evaluation GA Similarity to Ideal Solution
MOGA Multi-objective GA ORC Organic Rankine Cycles
NSGA (I and II) Non-dominated Sorting GA DMS Direct Multi-Search
SAW Submerge Arc Welding
* João Luiz Junho Pereira FCAW​ Flux Cored Arc Welding
joaoluizjp@gmail.com GMAW Gas Metal Arc Welding
Guilherme Antônio Oliver FEM Finite Element Method
gaoliver@unifei.edu.br DOE Design of Experiments
Matheus Brendon Francisco ANOVA Analysis of Variance
matheus_brendon@unifei.edu.br FSW Friction Stir Welding
Sebastião Simões Cunha Jr ANN Artificial Neural Network
sebas@unifei.edu.br NSTLBO Non-dominated Teaching–Learning
Guilherme Ferreira Gomes Based Algorithm
guilhermefergom@unifei.edu.br TLBO Teaching–Learning Based Optimization
HMOGWO Hybrid grey wolf optimizer
1
Mechanical Engineering Institute, Federal University SHM Structural Health Monitoring
of Itajubá (UNIFEI), Itajubá, Brazil

13
Vol.:(0123456789)
2286 J. L. J. Pereira et al.

1 Introduction This manuscript summarizes the most cited applications

in the last five years and some others which, although not so
Optimization can be described as a process of searching for recent, have great relevance in mechanical engineering. The
the best solution within a set of possible solutions [7]. As in main focus is to highlight the most modern and efficient trends
practical engineering problems most applications are non- on algorithms and decision-making techniques used, indicat-
linear, have complicated or nonexistent analytical solutions ing precisely the decision variables and objective functions
and must often serve more than one objective or function in real multi-objective optimization problems in mechanical
that may even be in conflict with each other, they require engineering.
sophisticated optimization tools to be determined [65, 179]. Some of the applications that can be found in this work
In optimizing a single objective, it is possible to deter- are: (i) problems in design optimization: meta-materials, func-
mine given one set of solutions, which is better than oth- tionally graded plates, airfoils, wind turbines, fan and pumps,
ers. As result, a single solution is usually obtained. How- security and support structures, suspensions, exchangers and
ever, in multi-objective optimization there is no direct expanders, beams, composites, etc. (ii) problems in process
method to determine if one solution is better than another engineering: welding, machining and molding and (iii) prob-
because the answer is a set of solutions that involve multi- lems in structural health monitoring.
ple conflicting objectives considered simultaneously [66]. So this paper review has two purposes: (i) to synthesize
There are several techniques, between algorithms and the main concepts of multi-objective optimization: multi-
decision-making techniques, for dealing with a multi- objective problem; Pareto front and its types; decision-making
objective optimization problem and even more recently. techniques—no preference, a priori, interactive and a poste-
The methods can be divided into optimization processes in riori techniques and algorithms—enumerative, deterministic
which an operator can participate at any time: not partici- (gradient-based) and especially stochastic (evolutionary and
pating, participating at the beginning of the optimization others meta-heuristics) with examples and a critical discus-
process, during or only at the end. The solutions found sion of its drawbacks and (ii) situate which techniques and
depend heavily on when the problem operator acts [31]. algorithms within this vast area have been applied to problems
Regardless of how the operator works, there are several in mechanical engineering, pointing out modern and more effi-
optimization algorithms that can work with him according cient trends.
to the methodology adopted. Understanding the basic con- This manuscript is organized as follows: Sect. 2 presents a
cepts of multi-objective optimization that are available in general theoretical background review about multi-objective
the literature is extremely important to situate the operator optimization. Section 3 presents a systematic review of the
of the problem and provide support for him to choose the literature on the main applications of multi-objective optimi-
best possible approach to his problem. zation in Mechanical Engineering detailing which algorithms,
Mechanical engineering is a vast and complex area with techniques, decision variables and objectives were used in each
numerous possible applications for multi-objective optimiza- of the problems and Sect. 4 brings conclusions.
tion. Nowadays, according to the author’s best knowledge,
there is no work in the literature that compiles the main 2 Backgrounds
applications and allows the researcher an overview of the
subject in this area. The review works found, and some are The optimization problems that must meet more than one
even out of date, usually focus on the (i) multi-objective objective are called Multi-objective Optimization Problems
optimization itself: Long et al. [102], Gunantara [68], Wang (MOPs) and present several optimal solutions [28]. The solu-
et al. [175], Marler and Arora [108] focusing on engineering tion is the determination of a vector of decision variables
in general, (ii) in algorithms and where they have already X = {x1, x2, …, xn} (variable decision space) that optimizes
been applied: Liu et al. [101] in meta-heuristics for discrete the vector of objective functions F(X) = {f1(x), f2(x), …, fn(x)}
optimization, Mane and Rao [107] in Evolutionary Algo- (objective function space) within a feasible region of solutions
rithms, Tamaki et al. [165] in Genetic Algorithms, Song subject to equality hi(x) or inequalities gi(x) constraints where
and Gu [110] in Particle Swarm Optimization, Leguizamón xmin and xmax are the limits that determine the search space for
and Coello [93] in Ant Colony Optimization,or (iii) spe- each of the variables, or vector of variables [17]. As described
cific cases: Ridha et al. [148] in photovoltaic system, Kumar in Eq. (1) [67].
et al. [90] in machining, Ojstersek et al. [123] in production
scheduling, Liu et al. [100, 101] in wind energy, Rangaiah
et al. [140] in chemical process engineering, Afshari et al.
[2] in concrete structures, Cui et al. [34] in energy saving,
Fadaee et al. [46] in renewable energy.

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2287

region. There may be voids without solutions, which con-

{ }
min F(𝐗) = f1 (𝐱), f2 (𝐱), ..., fn (𝐱)
tribute to the discontinuity of the problem [17, 31]. Figure 2
subject to ∶ hi (𝐱) = 0, i = 1, 2, ..., p
illustrates this paragraph well.
gi (𝐱) ≤ x 0, i = 1, 2, ..., q In mono-objective optimization, the specific param-
xmin ≤ 𝐱 ≤ 𝐱max (1) eters of each meta-heuristic must be correctly chosen for
each problem to have the best exploration and exploitation
This leads to a set of solutions called Pareto-optimal, response. This is respectively, being able to escape from
which according to Rao [147] is a feasible region X that local minimums and still being able to improve the preci-
there is no other feasible region Y such that fi(Y) ≤ fi(X) for sion of the solutions already found [125]. In MOP, this alone
i = 1, 2, …, k with f j(Y) ≤ f j(X) for at least one j. That is, is not enough. A good multi-objective optimization algo-
there is no other feasible solution Y that would reduce some rithm when looking for a set of possible solutions forming
objective function without causing an increase in at least one the Pareto front, must be able to find a PF with precision
other objective function at the same time. The method most (convergence) and good distribution of possible solutions
commonly used to compare solutions is Pareto Dominance throughout the solution space (coverage) [22, 114, 115], as
Relationship, which instead of determining a single optimal shown in Fig. 3.
solution, leads to a set of optimal alternatives between the Accurately determine the set of non-dominated solutions to
objectives. These solutions are also called non-dominated a problem is a hard task made possible by the development of
solutions or Pareto Front (PF) [78] and any of these solu- better computers and new techniques and algorithms. Zitzler
tions are optimal and it is up to the operator of the problem et al. [193] interested in testing and comparing evolutionary
to choose the best one according to his preference. See the multi-objective optimization algorithms has created six test
variable decision space and the objective or solution space
with examples of non-dominated solutions in blue (PF) in
Fig. 1, where n is the number of design variables and k is the
number of objective functions of the problem.
Possible solutions in variable decision space generate
solutions in the objective space and dominance relations are
analyzed to eliminate those that are not Pareto-optimal. As
seen, this front is built through iterations and before having
a final Pareto front, there are local Pareto fronts behind this.
Each MOP has a characteristic PF, which can be continuous
or discontinuous (disconnected) and convex or concave. A
MOP will be considered convex if the viable set and the
individual objective functions are convex (as in Fig. 1) and
this leads to a convex PF. If the viable set is not convex, or at
least one of the functions is non-convex, the problem will be
considered concave. In general, for non-convex MOP, PF can
be concave and disconnected [36]. Not every region of the
Fig. 2  Regions of a design problem with two-variable and two objec-
objective space, including those above the PF, is a feasible
tive functions [66]

Fig. 1  Regions of a design

problem with two-variable and
two objective functions [66]

13
2288 J. L. J. Pereira et al.

2.1.1.1 Global Criterion Method The ideal point or ideal

vector is the utopia solution that contains each separately
minimum objective function value achieved at the same
point; see Eqs. (2) and (3) [31]. The Nadir point is the exact
opposite, is the vector that contains each separately maxi-
mum objective function value found.

fi0 (x0i ) = min fi (x) (2)

]T
f 0 = f10 , f10 , f10 , ..., fk0 (3)
[

Therefore, the best solution found in this method is seen in

Eq. (4) [113]:
k
∑
Lp = min( |fi (x) − fi0 |)1∕p (4)
| |
| |
i=1
Fig. 3  Capabilities that a meta-heuristic must have to be successful in
a MOP where fi(x) is the objective function i evaluated for all x with
i = 1, 2, …, k the number of objectives and fi0 the minimum
value found in the minimization separately of function i.
functions that are still one of the most popular ones for testing The solution obtained depends greatly on the value chosen
new algorithms [114, 115]. The ZDT (Zitzler-Deb-Thiele) test for p, widely used choices are 1, 2 or ∞. This is the function
functions are a set of problems with very diverse Pareto fronts that provides the shortest distance between the PF and the
in which the best algorithm is the one that has the Pareto front ideal point. If p is one, there is a linear distance and if 2, a
found closest to the true Pareto front. Figure 4 shows the true Euclidean distances [113].
Pareto front for some of these functions. The main drawback of these types of methods is to
Note that the Pareto fronts of the ZDT1 (Fig. 4a) and neglect all other solutions.
ZDT2 (Fig. 4b) functions are continuous, however the first
is convex and the second is concave. The ZDT3 (Fig. 4c) 2.1.2 A Priori Preference Articulation
represents the discreteness feature; its Pareto-optimal front
consists of several noncontiguous convex parts and the In a priori methods the DM need to input his preference
ZDT4 (Fig. 4d) contains ­219 local Pareto-optimal front and, before optimization starts. This method have some difficul-
therefore, tests for the algorithm´s ability to deal with multi- ties: (i) The DM, when initially taking his preferences, can
modality. The first three have thirty design variables and the neglect important aspects of the problem and consequently
last ten. Those functions that are well defined can check how arrive at ineffective or even confusing results [167], (ii) An
well a new algorithm has convergence and coverage capa- algorithm like this should be run multiple times to determine
bilities. At the same time, it shows how diverse the Pareto the PF and (iii) some special PF´s can not be determined
fronts of a real engineering problem can be. with this approach [36, 111]. The main drawbacks of these
methods are that: (i) an algorithm should be run multiple
2.1 Main Methods to Approach Multi‑objective times to determine the Pareto optimal set, (ii) there is a need
Problems to consult with an expert because an inexperienced DM can
select bad regions for exploration and neglect better ones by
There are several techniques for dealing with MOPs. In gen- inducing the optimization process in the wrong direction and
eral, they are classified according to when the preferences of (iii) some special Pareto optimal fronts cannot be determined
the problem operator (or decision maker, DM) are inserted with this approach [114].
in the problem: a priori, interactive, a posteriori methods or
no preference method [32]. 2.1.2.1 Lexigraphic Method In Lexigraphic method the
DM must arrange the objective function according to their
2.1.1 No Preference Methods absolute importance (best to worst). After, the most impor-
tant objective function is minimized subject to the original
In no preference methods the DM is not needed. Only one constraints. If this problem has a unique solution, is the
solution is computed and is usually as close as possible to solution of the whole MOP. Otherwise, the second most
the ideal point. important objective function is minimized. Now, in addition

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2289

a) Convex PF b) Concave PF

c) PF with several convex parts d) PF with two difficulties: no uniformity of

search space and lowest density of the
solutions near the true PF

Fig. 4  Main ZDT test functions for MOP

to the original constraints, a new constraint is added to guar- tries to minimize the absolute deviations from the targets
antee that the most important objective function preserves to the objectives. The simplest form of this method is in
its optimal value. If this problem has a unique solution, is Eq. (5) [31, 42]
the solution of the original problem. Otherwise, the process k
goes on as above [145] ∑
min |fi (x) − Ti |
| | (5)
i=1
2.1.2.2 Goal Programming In this method, DM has to
assign targets or a goal that wishes to achieve for each where Ti denotes the target or goal set by the DM for the i-th
objective. These values are incorporated into the problem objective function fi(x).
as additional constraints and then the objective function

13
2290 J. L. J. Pereira et al.

Other less used a priori methods and references for need to consult with an expert to find the best weights [116].
knowing hands are: Min–Max Optimization [127, 146], According to Miettinen [113], this method can be an a priori
Multi-attribute Theory [121], ELECTRE (elimination and method if the DM defines the weight he wants to transform
choice translating algorithm) [19] and it is derivations; a MOP into a mono-objective problem.
PROMETHEE (Preference Ranking Organization Method
for Enrichment Evaluations) [21], among others. 2.1.4.2 ∈‑Constraint Method This method was introduced
by Haimes et al. [70] and only one of the objective functions
is selected to be optimized and all other objective functions
2.1.3 An Interactive Preference Articulation
are converted into constraints by setting an upper bound
to each of them. The problem to be solved has the form in
In Interactive methods the DM can to articulate his prefer-
Eq. (7):
ences during the optimization process, usually based on the
domain knowledge acquired during the optimization [50, minimize fl (x)
82]. These techniques normally operate in three stages: (i) (7)
subject to fj (x) ≤ 𝜀j
find a non-dominated solution, (ii) get the reaction of the
DM regarding this solution and modify the preferences of where for all j = 1, …, k, j ≠ l, l ∈{1, …,k}.
the objectives according to his need and (iii) repeat the two
previous steps until the DM is satisfied [32]. 2.1.4.3 Hybrid Method This method combines the Weight-
The main algorithms that use this technique are: (i) Prob- ing Method and the ∈-Constraint Method [113].
abilistic Trade-Off Development Method (PROTRADE), (ii)
STEP Method and (iii) Sequential Multi-objective Problem 2.1.4.4 Normal Boundary Intersection The Normal Bound-
Solving Method, (iv) Interactive Surrogate Worth Trade- ary Intersection (NBI) is proposed by Das and Dennis [36]
Off Method (ISWT), (v) Geoffrion–Dyer–Feinberg Method and according to the author, the method is independent of
(GDF), (vi) Sequential Proxy Optimization Technique the relative scales of the functions and is successful in pro-
(SPOT), (vii) Tchebycheff Method, (viii) Reference Point ducing an evenly distributed set of points in the Pareto set
Method, among others [63, 31, 33, 113, 118]. given an evenly distributed set of parameters, a property
which the popular method of minimizing weighted combi-
2.1.4 A Posteriori Preference Articulation nations of objective functions lacks. However, according to
Brito et al. [24], this method is extremely sensitive to the
In a posteriori methods, a set of representatives Pareto opti- presence of correlation between objective functions that are
mal is obtained and the DM can analyze the trade-off rela- used in the construction of PF.
tionships between the objectives [32]. This method is the This method starts with the determination of the payoff
most used in the literature to solve real problems, since one matrix (Φ). The vector of decision variables that minimizes
of the advantages is to find PFs that no other method can find the objective function fi (x) is represented by xi* and conse-
and with just one program run [114, 115]. quently the minimum value of fi (x) at this point is fi* (xi*).
When replacing the individual point xi* in the other func-
2.1.4.1 Weighting Method The idea is associate each tions, there is fi (xi*), this is a non-optimal value of this func-
objective function with a weighting coefficient and mini- tion. Repeating the algorithm for all m functions, is obtained
mize the weighted sum of the objectives. The multiple the payoff matrix represented in Eq. (8):
objective functions are transformed into a single objective
⎡ f1∗ (x1∗ ) ⋯ f1 (xi∗ ) ⋯ f1 (xm ∗ ⎤
)
functions. The weights wi are positive real numbers for al . .
⎢ ⋮ ⋱ ⋮ . ⋮ ⎥
i = 1,…, k objective functions. The weights are normalized,
Φ = ⎢ fi∗ (x1∗ ) ⋯ fi∗ (xi∗ ) ⋯ fi∗ (xm
∗
(8)
⎢ ⎥
k )⎥
that is, wi = 1. So, here is the Eq. (6) [184]: .
..
∑
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
i=1 ⎢ f ∗ (x∗ )
k ⎣ m 1 ⋯ fm (xi∗ ) ⋯ fm∗ (xm∗ ⎥
)⎦
∑
min wi fi (x) (6)
i=1 Each line of Φ is composed of minimum and maximum
values of f i (x). These sets of extreme points are used to
However, the main drawbacks of this approach are the normalize objective functions. Considering a set of weights
need to run an algorithm multiple times to find multiple wi, the Φ* wi will represent a point on the utopia line. Since
Pareto optimal solutions, dealing with all the challenges ƞ is a unit vector in the direction of origin and normal to the
in every run, lack of information exchange between Pareto utopia line at points Φ* wi, is obtained that Φ* w + D* ƞ will
optimal solutions during optimization cause weights are represent the set of points in that normal. The point where
always positive and concave PF can not be found and the

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2291

the normal intersects the boundary of the viable region making the MOP solution extremely costly and slows even
closest to the origin will be the point corresponding to the computationally [31]. Deterministic methods are those based
maximization of the distance between the utopia line and on gradient or derivatives and the most used in MOP are:
the PF [36, 24]. Therefore, the NBI method can be written (i) Greedy, (ii) Hill Climbing algorithms, (iii) Branch and
as a constrained nonlinear maximization problem defined Bound, (iv) Depth-First and (v) Breadth-First, (vi) best-first
as in Eq. (9): and (vii) calculations-based [23, 64]. However according
Parkinson et al. [136] and Coello et al. [31], deterministic
Max(x,t) D
(9) algorithms have difficulty for optimization problems with: (i)
s.t. ∶ Φ ∗ w + D ∗ 𝜂 = F(x) discrete-valued design variables,(ii) large number of design
variables; (iii) multiple local minima, maxima, and saddle
This problem must be solved iteratively for different val-
points (multimodal); (iv) no differentiable objectives and
ues of w to generate an equally spaced PF.
constraints; (vi) discontinuities of functions or regions (vi)
analysis programs which crash for some designs. Therefore,
2.1.4.5 Technique for Order of Preference by Similarity
enumerative and deterministic search techniques are unsuit-
to Ideal Solution The Technique for the Order of Preference
able in real-world and engineering MOPs. Therefore they
by Similarity to Ideal Solution (TOPSIS) was introduced by
will not be addressed in this paper.
Hwang and Yoon [76] and became a classic multiple attrib-
Stochastic techniques have demonstrated great poten-
ute decision making method with more than 4500 citations
tial for the solution of complex MOPs and are increasingly
[183]. TOPSIS determines the positive ideal solution (A+)
gaining space with the increase in the speed and processing
(Utopia point) as well as the negative ideal solution (A−)
capacity of computers. Today this is the main technique for
(Nadir Point) and normalizes each of the objectives accord-
engineers and designers and although there are several algo-
ing to itself and multiplies them by the weight assigned to
rithms, the basis of all of them consists in the initialization
each objective, being the sum of all weights wi always one.
of the optimization process with a set of random candidate
Then it calculates the Euclidian distance of each solution
solutions for a given problem and improves them over a pre-
(Ai) in the Pareto front to the utopia point (originating Si+)
defined number of steps. To address the real-world issues,
and to the Nadir point (originating Si−) and calculates the
these algorithms should be equipped with different operators
score Pi using Eq. (10):
[114, 115].
Si− The literature shows that almost all stochastic algorithms
Pi = (10) used in multi-objective optimization were inspired by some
Si+ + Si−
optimal phenomena found in nature and are commonly
If Pi = 1, Ai = A+ and if Pi = 0, Ai = A  (that is, the higher called meta-heuristics. In general, there are four main groups
Pi, the better the solution). In few words, TOPSIS deter- that divide meta-heuristics according to the inspiration for
mines the best compromised solution, which was the closest their creation: (i) based on evolution, (ii) based on physical
to (A+) and the furthest from (A−) based on the Pareto set phenomena, (iii) based on behaviors related to humans and
according to the objective weights and the normalization of (iv) based on swarms [72]. For Yang [179], there is also a
these solutions. Note that with the Score Pi of each solution, classification for meta-heuristic algorithms that can be based
it is possible to rank all solutions on the Pareto front from on trajectories or population.
best to worst according to TOPSIS. In the literature, an immense variety of meta-heuristics
can be found that are capable of solving mono-objective
optimization problems, but the number of algorithms capa-
2.2 Main Algorithms to Approach Multi‑objective ble of solving a MOP is much less.
problems

Regardless of the way the DM approaches MOP, this as 2.2.1 Evolutionary Algorithms

an optimization problem and must have some technique
of general search that should lead to the maximization or Evolutionary algorithms were the first meta-heuristics
minimization of the functions or their compositions. Just created to deal with multi-objective optimization prob-
like optimizing a single objective, The MOP can be clas- lems and they drastically broke (and at the same time
sified through 3 categories according to the applied search can be used together) with most of the classic methods
technique: enumerative, deterministic and stochastic. presented earlier. They deal simultaneously with a set of
Enumerative is the simplest search strategy where each possible solutions (the so-called population). This allows
possible solution is evaluated. However, this technique is finding several members of the Pareto optimal set in a
inefficient or even unfeasible as search space becomes large,

13
2292 J. L. J. Pereira et al.

single run of the algorithm and are less susceptible to the 2.2.2 Others Meta‑heuristics
shape or continuity of the PF [31].
In short, they use paradigms from natural evolution, Although there are good evolutionary algorithms for cer-
such as selection, recombination and mutation to lead a tain types of problems, that are population-based and the
population (set) of individuals (decision vectors) towards inspiration is natural evolution, it has been shown that there
optimal or near-optimal solutions [15]. The first meta- is still room for improvement for specific problems. Based
heuristic created in this sense was by Holland [73], it on this, other important multi-objective optimization algo-
dealt with mono-objective problems and is called Genetic rithms found in the literature are: (i) Simulated Annealing
Algorithm (GA). The first to be developed to deal with for Multi-objective Optimization (SAMO) [157],(ii) Multi-
MOP´S is now called multi-objective evolutionary algo- objective Tabu Search (MOTS) [56]; (iii) Multi-objective
rithm (MOEA) or Vector Evaluation Genetic Algorithm Ant-Q (MOAQ) [109], (iv) Vector Evaluated Particle
(VEGA) [159]. VEGA was mainly aimed for solving prob- Swarm (VEPSO) [56]; (v) MOPSO [120]; (vi) Adaptive
lems in machine learning [31]. Weighted Particle Swarm Optimization (AWPSO) [106].
After this, many other derivations with attempts at 2004); (vii) Artificial Immune Systems (AIS) [30]; (viii)
improvement came. The main ones are: (i) VEGA, (ii) Multi-objective Water Cycle Algorithm (MOWCA) [150],
Lexigraphic Ordering GA (LOGA- a priori preference) (ix) Multi-Objective Grey Wolf Optimizer (MOGWO) [114,
[52], (iii) Vector Optimized Evolution Strategy (VOES) 115],(x) Multi-objective Imperialist Competitive Algorithm
[91], (vi) Weight-Based GA (WBGA) [71], (v) Multi- (MOICA) [20], (xi) Self-adaptive Multi-objective Brain
ple Objective GA (MOGA) [50], (vi) Niched Pareto GA Storm Optimization (SMOBSO) [69], (xii) Multi-objective
(NPGA, NPGA 2 [74, 75], (vii) Non-dominated Sorting Ant Lion Optimization (MOALO) [115], (xiii) MOGOA
GA (NSGA, NSGA II) [38, 163], (viii) Strength Pareto [116], (xiv) Multi-objective Sine–Cosine algorithm (MO-
Evolutionary Algorithm (SPEA, SPEA II [193, 194], (ix) SCA) [166], (xv) Multi-objective Stochastic Fractal Search
Multi-objective Evolutionary algorithm Based on Decom- (MOSFS) (Khalilpourazary et al. 2019), (xvi) Multi-objec-
position [189], (x) Pareto Archived Evolution Strategy tive Seagull Optimization Algorithm (MOSOA) [39], (xvii)
(PAES) [89], among others. Evolutionary MOSOA (EMOSOA) [39], (xviii) Differential
With some exceptions, the distinction between all evo- Evolution-Crossover Quantum Particle Swarm Optimization
lutionary multi-objective algorithms is mainly due to the (DE-CQPSO) [171] (xix) Multi-objective Sunflower Opti-
differences in the paradigms used to define the selection mization [53], etc. As well as the evolutionary algorithms,
operators, whereas the choice of the variation operators is all of them have their internal parameters that regulate their
generic and dependent on the problem. As example, one search according to their inspiration and then, these solu-
of the most popular is the NSGA II which can be applied tions are compared using a Pareto dominance operator. All
to continuous search spaces as well as to combinatorial are stochastic algorithms and most of them has a posteriori
search spaces, whereas the selection operators stay the preference.
same, the variations operators (mutation and combination)
must be adapted to the representations of solutions in the
decision space [45]. All population-based multi-objective
algorithms are almost identical. They start the optimiza-
tion process with multiple candidate solutions and such
solutions are compared using the Pareto dominance opera-
tor. The most well regarded ones are: SPEA, NSGA-II,
MOEA/D and PAES [114, 115]
Although most of these algorithms are divided between
a priori, interactive or a posteriori methods, most of them
are like the latter. All are stochastic. According to the no-
free-lunch theorem, none of them can be excellent at solv-
ing any type of problem. Zitzler et al. [193] applied several
evolutionary algorithms to identify the PF of the ZDT3
function shown in Fig. 4c. Figure 5 has the result for each
of these algorithms and also that of a random search. As
evidenced, any evolutionary algorithm can be better than
a rand, but there are significant differences between them
and these differences can be very different from problem Fig. 5  Performance of the main evolutionary algorithms in the ZDT3
to problem. function [ adapted from Zitzler et al. [193]]

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2293

It is possible to observe a huge variety of stochastic Fan et al. [47] designed a new mixed flow flan improv-
algorithms found in the literature. For Yang [179], there ing a currently standard model. The author increased the
are no good or bad algorithms, but one more appropriate efficiency by 11.71% and the pressure raised by 50.15%
for a given optimization problem. It is a difficult task for a using Orthogonal Method (OM). The purpose of this
single algorithm to face any type of MOP with good bal- method is to test the influence of a particular factor over
ance of exploration, exploitation, convergence, coverage and the whole outcome, with the view of obtaining optimal
low computational cost at the same time. This fact suggests configuration in terms of the performance levels [187].
that there is always an opening for the development of new Zhang et al. [186, 188, 191] also applied multi-objective
meta-heuristics. optimization using OM in design of Axial Flow Pump
where the new design’s head and efficiency increased by
17.8% and 4.26%, whilst the shaft power and the pressure
3 Literature Review pulsation coefficient reduced by 1.22% and 11%, respec-
tively. Wang and Huo [177] used the same approach to
Multi-objective optimization has been improving and devel- improve the hydraulic performance of a centrifugal pump.
oped for some decades as it faces several problems that may Karimi et al. [84] proposed a design optimization approach
be encountered. The number of applications has been grow- for floating offshore wind turbine support structures,
ing in Mechanical Engineering and the following are those where the author found the locus of platform cost minima
that are considered most relevant or recent, divided into and wind performance maxima for a given environmental
the subareas where multi-objective optimization is present: condition and sea state spectrum modifying the geometry
design optimization, manufacturing and structural health of the structure (multi-objective constrained problem).
monitoring. Moleiro et al. [117] studied metal-ceramic function-
ally graded (FG) plates, which are composed of a main
3.1 Multi‑objective Problems in Design functionally graded material layer and may include metal
Optimization and/or ceramic faces under thermo-mechanical loadings.
Other authors who also studied this type of structure with
Undoubtedly one of the greatest applications in mechanical a pit in multi-objective optimization were: Ashjari and
engineering is in determining the geometry of structures to Khoshravan [14] and Franco Correia [54, 55]. Both con-
obtain the best possible performances. Generally related to sidered constraints on the problem.
weight, stiffness, aerodynamics and/or modal parameters. Multi-objective optimization is widely found in the
The more objectives that must be met at the same time, the design of collision structures. Asanjarani et al. [13] pre-
more complex MOP becomes. The most relevant applica- sented a crashworthiness optimization of the tapered thin-
tions are cited in the text and the objective functions, deci- walled square tube with indentations using one and multi-
sion variables and algorithm employed are in Table 1. objective approach, where the author found an optimized
Grachi et al. [62] found a design for a meta-material that collision geometry. Zhang et al. [186] developed a hybrid
found the viscoelastic with the highest vibration attenua- multi-objective optimization approach for absorbing struc-
tion in a low-frequency range. Meta-materials are compos- tures in train collisions that brought good results using
ites that are artificially engineered to have unconventional multi-objective artificial bee colony. Peng et al. [132]
mechanical properties that stem from their microstructural also studied structures for collision using NSGA-II with
geometry rather than from their chemical composition. an approach using finite element methods and design of
Schlieter and Dlugosz [155] optimize the design of an experiments (DOE).
airfoil by proposing a new optimization methodology and Ebrahimi-Nejad et al. [44] tried to found the best design
obtaining many optimal Pareto solutions for a large number of a sports car suspension system using simplified quarter-
of decision variables. Li et al. [97] proposed a new method car models and TOPSIS. Other authors who studied sus-
for optimizing the design of wind turbine blades that sur- pension designs using a multi-objective approach were: (i)
passed traditional methods, since in addition to optimizing Zhang et al. [188] performed the multi-objective suspension
structural strength and stiffness of the blade also considers system optimization for an in-wheel-motor driven electric
the noise and power generation efficiency. Wang et al. [173, vehicle, (ii) Zhang and Wang [189] conducted a parametric
175] also studied multi-objective wind turbine blade design study to optimize a half-vehicle suspension system model,
proposing a novel gradient-based multi-objective evolution (iii) Fossati et al. [51] used NSGA-II and numerical compu-
algorithm based on both uniform decomposition and differ- tational studies comprising the multi-objective optimization
ential evolution (MODE/D), which according to the author, of a full-vehicle suspension, (iv) Jiang and Wang [79, 80]
performed better than NSGA-II mainly when the number of used TOPSIS to optimize the suspension system of a truck
objectives is increased. and also to optimize handling stability and ride comfort.

13
 Table 1  Multi-objective optimization system in design problems
2294

Authors Objective functions Decision variables Optimization method Structure

13
Panagant et al. [130] Structural mass and Compliance Several variables related to the size and Fourteen meta-heuristics Truss
number of trusses
Panagant et al. [129] Strucutural mass, welding cost and Topology, shape and size Fourteen meta-heuristics Floor-flame
compliance
Grachi et al.[62] Metamaterial inertia and Bragg scatter- Layer thickness and number of layers MO based on GA Metamaterial
ing effect
Schlieter and Dlugosz [155] Equivalent stress, displacement, fre- 24 design variables (geometry) MO based on DE Airfoil
quency, mass
Moleiro et al. [117] Mass, displacement and Tsai-Hill failure Thickness, power-low distribution, the Direct MultiSearch (DMS) FG blades
criteria thickness of metal
Ashjari, Khoshravan [14] Mass and deflection FG core volume fractions and thickness NSGA II FG blades
of the face sheets
Franco Correia et al. [54, 55] Mass, cost, natural frequency Index of the power-law of volume frac- DMS FG blades
tions, thickness of FGM layer and face
sheets
Li et al. [97] Structural strength, stiffness, noise reduc- Chord length and twist angle (for each MOPSO and finite volume method Wind turbine blade
tion and aerodynamic performance cross section)
Wang et al. [173, 175] Energy production, blade mass, root 30 related to design MODE/D Wind tubine blade
thrust, cost
Fan et al. [47] Efficiency and Pressure Hub angle of impeller (and wrap) and Orthogonal method Mixed flow FAN
diffuser
Zhang et al. [191] Head, efficiency, shaft power and pres- Number of blades, blade setting angle, Orthogonal method Axial flow pump
sure pulsation hub ratio, distance between the blade
and the guide vane
Wang and Huo [177] Indexes head, efficiency, shaft power and Impeller outlet width, blade inlet angle, Orthogonal method Centrifugal pump
pump net positive suction head blade outlet angle and cape angle
Karimi et al. [84] Cost model and wind turbine perfor- Nine geometric variables of multi-body NSGA-II Offshore support
mance metric platform
Asanjarani et al. [13] Specific energy absorption and ratio Cross section, thickness, taper angle, RSM, NSGA-II and desirability function Tapered thin-walled square
between average and maximum crush- number and radius of indentations tube
ing forces
Zhang et al. [186] Capability of absorbing impact and Side length and wall thickness of hexago- MOABC Train collision piece
energy nal tube
Ebrahimi-Nejad et al. [44] unsprung and sprung mass accelerations, Stiffness and damping TOPSIS Sports car suspension
displacement and suspension travel
Zhang et al. [187] Sensitive of the front double pivot and the Eighteen parameters related to stiffness NSGA-II Suspension system
rear double wishbone suspensions and damping coefficients
Fossati et al. [51] Three objective functions related to com- Six parameters being stiffness and damp- NSGA-II Full-vehicle suspension
fort and safety ing coefficients of each suspension
Rao et al. [142, 144] Total surface area, total annual cost, total Seven design variables related to geom- Jaya algorithm Plate-fin heat exchangers
pressure drop and effectiveness etry
J. L. J. Pereira et al.
 A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2295

Panagant et al. [130] conducted a research using 14 types

of meta-heuristics in 8 types of classical trusses subject to
bound and stress constraints and compared the results of

Composite cylindrical
each one, concluding that the algorithm proposed by them
was the one that had the best performance, the Success His-
Radial expanders

Composite plate
Composite plate
Composite plate
Composite plate
Composite plate
Beam structure tory-based Adaptive Multi-objective Differential Evolution
(SHAMODE-WO). Panagant et al. [129] also used 14 meta-
Structure

heuristics, this time in an automotive floor-frame. The author

concluded that the meta-heuristic proposed in his work was
one of the best algorithms, the Real-code Population-Based
Incremental Learning hybridized with Adaptive Differential
evolution (RPBILADE).
Multi-objective firefly algorithm

In the thermal area some designs can be found. Rao et al.

[142, 144] applied single and multi-objective optimization
in design of plate-fin heat Exchangers, which the design
Optimization method

involves a number of geometric and physical parameters

with high complexity. The general approaches are based
and trial and few studies before him used multi-objective
NSGA-II
NSGA-II

NSGA-II
NSGA-II

optimization. One of these was Ahmadi et al. [5] who used

VEPSO
MOGA

NSGA-II. However, Rao compared this study to his and

GA

concluded that Jaya had a better result. Bahadormanesh

et al. [16] applied multi-objective optimization in improve-
Stacking sequence, thickness, material
Different organic working fluids, mass

ment design of radial expanders of Organic Rankine Cycles

Volume fractions, thickness and fiber

Cylinder thickness, radius and lenght

flow rate, evaporator temperature,

(ORC) using firefly algorithm, where it was possible to pre-

select better parameters for the construction of these rotors.
Wang et al. [174] also applied multi-objective optimization
Volume fraction, thickness
Thickness, fiber and resin

in ORC, but aiming to improve aspects thermodynamics and

maximum pressure

economics using NSGA-II.

orientation angles
Decision variables

Fiber orientation
Fiber orientation

Works in the field of composite materials can also be

found. Vo-duy et al. [170] optimized a beam structure made
by laminated composite using finite element method. Gha-
semi and Hajmohammad [59] applied multi-objective opti-
mization in design of laminated composite cylindrical shell
under external hydrostatic pressure for minimum cost and
Thermal efficiency and size parameter

maximum buckling pressure. Other actors who also worked

with the optimization of carbon fiber structures using a
multi-objective approach were: Arian Nik et al. [12], Lee
et al. [92], Omkar et al. [126], Kalantari et al. [83], Ikeya
Weight, natural frequency

et al. [77] and Diniz et al. [41].

Weight, deformation
Mass/cost, buckling
Objective functions

Mass, compliance
Rigidity, buckling

3.2 Multi‑objective Problems in Manufacturing

Weight, cost
Weight, cost

Processing engineering is an extensive area in Mechani-

cal Engineering, the largest areas of which are Welding,
Machining and Molding.
Bahadormanesh et al. [16]

3.2.1 Welding
Arian Nik et al. [12]
Table 1  (continued)

Kalantari et al. [83]

Vo-duy et al. [170]

Omkar et al. [126]

Welding is one of the most important areas of engineering

Ikeya et al. [77]

and there are currently many types of processes. Some

Lee et al. [92]
Ghasemi [59]

of these processes can have a large number of decision

Authors

variables controlling the process, with high complexity

of correlation or analytical solutions. Still, you can have

13
2296 J. L. J. Pereira et al.

numerous objectives to be optimized simultaneously. Saha and Mondal [151] studied the optimization of
Some of them are to increase productivity, decrease costs, manual metal arc welding (MMAW) process parameters for
reduce the thermally affected zone, increase the reinforce- nanostructured hardfacing material using hybrid approach
ment, decrease (or increase) the hardness, decrease the with Taguchi, TOPSIS and PCA (Principal Component
emission of toxic gases, decrease the consumption of Analysis) identifying the optimal process parameters. The
electric energy, increase the impact strength, increase study of the optimization of welding parameters by a multi-
the tensile strength, increase elongation in the weld area, objective approach was found less frequently in other pro-
reduce noise pollution, among others. As seen, the use of cesses because they are not so common or are recent: (i)
Meta-heuristics has collaborated a lot to deal with this Laser-magnetic hybrid welding (LMW) by Yang et al. [180],
extensive area. The most relevant applications are cited in (ii) hot wire laser welding (HLW) by Yang et al. [181] (iii)
the text and the objective functions, decision variables and Hybrid laser-arc welding (HLAW) by Gao et al. [57], (vi)
algorithm are in Table 2. laser welding process (LW) by Jiang et al. [81] using FEM,
Ahmad et al. [4] studied Submerge Arc Welding (SAW) Kriging (a meta-model) and NSGA-II, (v) Micro resistance
looking for optimal parameters to achieve productivity and spot welding (MRSW) by Chen et al. [27] and (vi) Laser
weld quality. SAW is a versatile welding process widely beam machining by Belinato et al. [18].
used in fabrication and manufacturing of marine and pres- Another type of multi-objective problem related to
sure vessels, pipelines and offshore structures. Other authors welding is the local scheduling. Lu et al. [104] studied an
who also studied the SAW process aiming at multi-objective approach to welding shop scheduling that according to the
optimization, generally as weld quality, strength, hardness authors, should simultaneously consider economic, environ-
and/or productivity, were: Choudhary et al. [29], Ahire et al. mental and social impacts. In this way, the authors proposed
[3], Sailender et al. [153], Silva et al. [154], Rao et al. [142, a multi-objective approach using a novel hybrid multi-objec-
144], Al Dawood et al. [6], Yifei et al. [182] (welding robot tive grey wolf optimizer (HMOGWO) for makespan (total
parameters), Torres et al. [168], among others. sum time of each process), energy consumption and noise
Sowrirajan et al. [162] applied multi-objective optimiza- pollution (ignores in previous studies). The same author
tion to find the optimum clad layer dimensions in pressure in Lu et al. [105] applied the same algorithm for dynamic
vessels using stainless steel that maximize clad height and scheduling in a real-world welding industry. In both cases
clad width and minimize depth of penetration in a process the authors concluded that the algorithm used outperforms
FCAW (Flux Cored Arc Welding). Paula et al. [131] and known EA’s.
Almeida et al. [11] also studied the optimization of the
parameters of this process. 3.2.2 Machining
Shao et al. [161] studied the optimization of gas metal
arc welding (GMAW) parameters and sequences for low- Another major area of process engineering is machining.
carbon steel (Q345D) T-joint using FEM and DOE conclud- As with welding, machining has a wide range of processes
ing that the welding residual deformation and stress always like milling, turning, drilling or cutting and each can have
have opposite behavior and are very influence by the process a great number of decision variables that control the pro-
parameters. Lorza et al. (2018) also studied the optimiza- cesses. Some of the conflicting objectives to be optimized
tion of welded joints in GMAW using FEM, but another simultaneously are minimizing roughness, minimizing cost,
algorithm, the NSGA II. Like the previous author, this work minimizing cutting force, increasing productivity, increasing
approached a methodology to reduce the error between the material removal, decreasing process variability, reducing
FEM and a real case. Both seek the selection of optimal residual stress, among others. The most relevant applications
process parameters. are cited in the text and the objective functions, decision
Another welding process that many authors approached variables and algorithm employed are in Table 3.
to optimize their parameters was the friction stir weld- Rao et al. [141] applied multi-objective optimization
ing (FSW). Gupta et al. ([60] and [61]) studied this pro- using Non-dominated sorting Teaching–Learning Based
cess for joining different alloys in two works with different algorithm (NSTLBO) in three machining process (turning,
approaches, but the same materials. In both studies, the opti- wire-eletric-discharge machining and laser cutting) and
mal parameters found were the same. Shanjeevi et al. [160] two micro-machining process (ion beam micro-milling and
studied for AISI 304 l austenitic stainless steel and copper micro wire-electric-discharge machining) looking for the
points. Wakchaure et al. [172] for Alloy 6082 using Tagu- best process parameters. The author compared this algorithm
chi-GRA and artificial neural network (ANN) and Senthil with NSGA-II and others algorithms.
et al. [156] for AA6063-T6 pipes using Analysis of Variance Lin et al. [98] studied machining parameters in multi-
(ANOVA) and RSM. pass turning operations for low carbon manufacturing
considering reducing machine cost, energy consumption

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2297

Table 2  Multi-objective optimization system in welding process

Authors Objective functions Decision variables Optimization method Process

Choudhary et al. [29] Bead width, reinforcement and Voltage, feed, speed, nozzle to GA, JAYA Algorithm and desir- SAW
penetration plate distance, flux condition and ability function
plate distance
Ahmad et al.[4] UTS, Hardness, deposition rate, Current, voltage, speed and heat Taguchi-desirability function SAW
reinforcement height and bead input
width
Ahire et al. [3] Welding strength, welding deposi- Current, speed, root gap and elec- Response Surface and GA SAW
tion rate trode angle
Sailender et al. [153] UTS and Hardness Voltage, feed, speed and nozzle to Taguchi SAW
plate distance
Silva et al. [154] Dilution, reinforcement and bead Voltage, feed and nozzle to plate ANOVA SAW
width radio distance
Rao et al. [142, 144] Bead width, weld reinforcement, Current, speed and feed JAYA, GA, PSO and Imperialist SAW
weld penetration, tensile strength Competitive Algorithm
and weld hardness
Al Dawood et al. [6] UTS and hardness Current, voltage, speed Taguchi-fuzzy interference system SAW
Yifei et al. [182] Productivity and cost Welding path GA, Particle Swarm optimization SAW
Torres et al. [168] Joint dimensions and dilution Voltage, speed, wire feed rate, Generalized reduced gradient SAW
contact distance (GRG)
Rivas et al. [149] Carbon dioxide emissions, slag, Current, voltage, welding speed NSGA II, MOEA/D, MOPSO, SAW
wastes, electric power, material, SPEA II and PESA II
labor and energy cost,
Sowrirajan et al. [162] Clad height, clad width and depth Open circuit voltage, wire feed RSM and NSGA-II FCAW​
of penetration rate, welding speed, distance and
electrode angle
Shao et al. [161] Welding stress and deformation Current, voltage, speed, sequence DOE and MOPSO GMAW
and direction
Lorza et al. [103] Temperature field and angular Current and voltage NSGA II GMAW
distortion
Gupta et al. [60] Tensile strength, average hardness Rotational speed, welding speed, Grey relational analysis coupled FSW
and average grain size at weld shoulder and pin diameter with PCA and Taguchi
nugget zone
Gupta et al. [61] Tensile strength, micro-hardness Idem Gupta et al. [60] Artificial Intelligence and NSGA FSW
and grain size II
Shanjeevi et al. [160] Tensile strength, metal loss and Friction and upset pressure, rota- TOPSIS and Taguchi FSW
weld time tional speed
Wakchaure et al. [172] Tensile strength and impact Tool rotation speed, welding speed ANN, GRA and Taguchi FSW
strength and tilt angle
Senthil et al. [156] Yield, tensile strength and elonga- Rotational and weld speed ANOVA and RSM FSW
tion
Saha and Mondal [151] Weld bead width, reinforcement Current, voltage and welding TOPSIS-PCA MMAW
and bead hardness speed
Yang et al. [180] Macro-weld profile, microstructure Magnetic flux density, laser power, NSGA II and Taguchi LMW
and hardness welding speed
Yang et al. [181] Welding depth and reinforcement Laser power, welding speed, hot- Meta-models and NSGA II HLW
and strength wire current
Chen et al. [27] Tensile-shear, weld nugget size, Ramp time, welding time, current NSGA-II MRSW
failure energy and force
Gao et al. [57] Depth of penetration, bead width Laser power, current, distance NSGA II and Taguchi HLAW
and bead reinforcement between laser and arc and travel-
ling speed
Jiang et al. [81] Bead width and depth of penetra- Laser power and position and Kriging and NSGA-II LW
tion welding speed
Lu et al. [104] Makespan, total penalty of Permutation of tasks, actual HMOGWO –
machine load and instability quantify of the welders and the
starting time

13
2298 J. L. J. Pereira et al.

Table 2  (continued)
Authors Objective functions Decision variables Optimization method Process
Lu et al. [105] Makespan, noise pollution, energy Actual, start and finish processing HMOGWO –
consumption time

Table 3  Multi-objective optimization system in machining process

Authors Objective functions Decision variables Optimization method Process

Rao et al. [141] Tool flank wear and surface Cutting speed, feed and depth NSTLBO and NSGA II 5 different
roughness of cut machining
process -txt
Lin et al. [98] Carbon emissions, operation Velocity, feed rate, dept of MOTLBO Turning
time and cost cut and spindle speed of
machine tools
Sahu and Andhare [152] Roughness and force of Speed of cutting, feed rate TLBO, JAYA, GA and RSM Turning
cutting and depth of cut
Sivaiah and Chakradhar [158] roughness, flank wear and Speed of cutting, feed rate Taguchi and TOPSIS Cryogenic turning
remove rate and depth of cut
Mia et al. [112] Cutting force, specific energy, Cutting speed, feed rate and Gray-Taguchi Cryogenic turning
temperature, surface rough- number of jets
ness, material removal
Gaudêncio et al. [58] Roughness and MMSE Cutting speed, cutting feed, RSM, GRG and NBI Turning
machining depth
Almeida et al. [10] Mean and deviation of rough- Cutting speed, feed and depth RSM
ness of cut
Park et al. [135] Cutting energy and energy Cutting speed, feed rate, nose RSM, NSGA-II and TOPSIS Turning
efficiency radius, edge radius, rake
and relief angles
Warsi et al. [178] Cutting energy, material Cutting speed, feed and depth Gray-RSM Turning
removal and roughness of cut
Qu et al. [139] Cutting force, roughness, Spindle speed, feed per tooth, NSGA-II Milling
milling efficiency axial depth of cut
Montalvo-Urquizo et al. [119] Deformation, stress, shape Cutting velocity and axial Simulation-based multi- Milling
error and tool wear cutting depth objective optimization
Niu et al. [122] Residual stress, material Milling speed, feed per tooth, NSGA II Milling
removal rate, roughness and width of cut, depth of cut
surface hardness and amplitude ultrasonic
Prakash et al. [138] Surface roughness and micro- Peak current, pulse dura- Taguchi, RSM and NSGA-II PMEDM
hardness tion, duty cycle and silicon
powder concentration
Tripathy and Tripathy [169] Material removal rate, tool Powder concentration, peak ANOVA, GRA and TOPSIS PMEDM
wear rate, electrode wear current, pulse on time, duty
ratio and surface roughness cycle and gap voltage
Abidi et al. [1] Material removal rate, rough- Capacitance, electrode mate- MOGA II EDM
ness and tool wear rate rial and discharge
Prakash et al. [137] Material removal rate and Peak current, pulse-on, duty MOPSO Mixed-EDM
surface roughness cycle and powder concen-
tration
Dumbhare et al. [43] Surface roughness and kerf Abrasive flow rate, standoff ANOVA and RSM AWJM
taper angle distance and transverse
speed
Rao et al. [142, 144] Surface roughness and kerf Transverse speed, pressure, MO-Jaya and PROMETHEE AWJM
taper angle stand-off distance, tilt angle,
surface speed and abrasive
flow rate

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2299

and environmental impacts using a multi-objective teach- Other welding processes can be found in the literature
ing–learning-based optimization algorithm (MOTLBO) that have gained more space and multi-objective approaches,
in dry and wet cut. Sahu & Andhare [152] applied multi- but even less common, such as: (i) powder mixed electric
objective optimization to improve the machinability of discharge (PMEDM)—studied by Prakash et al. [138](first
Titanium alloy in cryogenic turning process using Teach- in this process to use NSGA-II) and Tripathy and Tripathy
ing–Learning Based Optimization (TLBO), JAYA, GA and [169], (ii) electrical discharge machine—studied by Abidi
RSM. Mia et al. [112] also studied cryogenic turning of a et al. [1] and Prakash et al. [137] and (iii) Abrasive Water
Titanium alloy. Sivaiah and Chakradhar [158] improved the jet Machine (AWJM)—studied by Dumbhare et al. [43] and
cryogenic turning process during machining of hardened Rao et al. [142, 144].
stainless steel.
Gaudêncio et al. [58] proposed a model for machin- 3.2.3 Molding
ing quality in the AISI 12L14 steel turning process using
fuzzy multivariate mean square error, that is, the objectives Molding is another area in which multi-objective optimiza-
are decrease roughness to minimize the roughness and its tion is not yet fully explored. It also has a significant num-
own variability (Multivariate mean square error—MMSE). ber of decision variables that control the process and many
Almeida et al. [10] used this same steel and mean and stand- objectives to be optimizing simultaneously. Among the ana-
ard deviation roughness objectives to optimize parameters in lyzed processes, this process had the least number of studies.
turning. Park et al. [135] studied turning process of hardened Most were in the injection process and aimed at reducing
material aiming to resolve environmental issues reducing the warpage, minimizing the cycle time or reducing costs,
the consumed energy and improving energy efficiency. The materials or energy used. The most relevant applications are
energy decreased 16% and the efficiency could be improved cited in the text and the objective functions, decision vari-
11% compared to the non-optimized system. Warsi et al. ables and algorithm employed are in Table 4.
[178] studied a sustainable turning of Al 6061 T6 where the Li et al. [96] proposed an approach to optimize the fiber-
proposed parameters resulted in reduction of specific cutting reinforced composite injection molding process. Kitayama
energy by 5% and improvement of 33% in material removal et al. [87] proposed another methodology to multi-objective
rate while surface roughness remained unaffected. optimization of injection parameters for short cycle time and
Another machining process is the milling. Qu et al. [139] warpage reduction using conformal cooling channel. The
applied multi-objective optimization to select optimum same author in Kitayama et al. [88] used the same method-
parameters in milling thin-walled plates. Montalvo-Urquizo ology to found the optimal parameters in plastic injection
et al. [119] also studied milling creating a simulation model molding for minimizing warpage and cycle time. Zhang
with FEM that the accuracy of the method compares very et al. [192] studied the optimization of the injection molding
well with experimental data. process parameters for diesel engine oil cooler. Okabe et al.

Table 4  Multi-objective optimization system in molding process

Authors Objective functions Decision variables Optimization method Process

Li et al. [96] Warpage, volumetric shrinkage Fiber content, fiber aspect ratio, Taguchi, RSM and NSGA-II Injection (fiber-
and residual stress melt temperature, injection pres- reinforce
sure and cooling time composite)
Kitayama et al. [87] Cycle time and warpage Melt temperature, injection time, Sequential approximate opti- Injection (plastic)
packing pressure, packing time, mization and radial basis
cooling time, cooling tempera- function
ture
Kitayama et al. [88] Packing pressure profile and Packing pressure profile, melt Sequential approximate opti- Injection (plastic)
warpage temperature, injection time, cool- mization and radial basis
ing temperature of coolant and function
cooling time
Zhang et al. [192] Warpage and clamping force Gate open time, molding tempera- Neural network and MOPSO Injection (plastic)
ture, melt temperature, injection
time, packing pressure and time,
cooling time
Li et al. [94] Stent expansion in a stenotic artery Length, thickness and outer Kriging Surrogate Microinjection
and molding quality diameter
Okabe et al. [124] Fill time, dry spot, weld line and Injection points MOGA Resin-transfer
wasted resin

13
2300 J. L. J. Pereira et al.

[124] proposed a new multi-objective optimization approach methodology. The first stage is to detect the occurrence of
for resin transfer molding process using FEM, MOGA, self- damage and the location of damaged zones by FE model
organizing map and scatter plot matrix. updating using damage functions [164]. The use of damage
Li et al. [94] applied multi-objective optimization in functions avoids adjusting the possible damage values of all
a biodegradable polymer stent structure and stent micro- the elements separately. This results in a reduced number
injection molding process using FEM, DOE and Kriging of parameters to determine which contributes to avoiding
surrogate method. The author improved the expansion per- optimization numerical problems and makes its application
formance and the microinjection molding process of the to large-scale structures easier. In the second stage, refined
biresorbable polymeric stent with tiny struts. location of damage and estimation of severity, a standard FE
model updating technology with independent adjustment of
3.3 Multi‑objective Problems in Structural Health the design variables is applied but, in this case, the num-
Monitoring ber of variables is much reduced because only the elements
belonging to the zones identified previously as damaged are
Structural Health Monitoring (SHM) is a structural inspec- now assumed to be damaged.
tion methodology that allows an early diagnosis of structural Kim and Park [85] made a very interesting study of iden-
health through non-destructive techniques, algorithms and tifying a crack on a rectangular plate where they addressed
the use of sensors in real time. The application of multi- the multi-objective problem in two ways: comparing the
objective optimization can be in the numerical analysis of two objectives into one using the method of weighting sum
the collected data and also in the optimization of the reading method and without composing them, leaving them free and
of the structure through, for example, positioning of the sen- on analysis of the dominance relationship. They reported
sors [66]. The most relevant applications are cited in the text that the proposed multi-objective EA was more efficient than
and the objective functions, decision variables and algorithm the single objective EA.
employed are in Table 5. Some authors have analyzed three-dimensional steel
The study of the positioning of the sensors aims cost structures: (i) Cha and Buyukozturk [26] showed a method-
reduction by minimizing the number of sensors by improv- ology that could be used effectively for detecting minor local
ing the efficiency of data reading. In the literature some damage, although real-world validation using experimental
cases can be found considering multi-objective optimization. data is still needed, (ii) Alkayem et al. [8] studied 3D steel
In highlight are: (i) Ferentinos and Tsiligiridis [48] (wire- structures through two algorithms applying them in their
less sensors subject to the connectivity and spatial density mono and multi-objective versions and concluded that in
constraints), (ii) Gomes et al. [66] (composite plates), (iii) this case PSO and MOPSO had more accurate results and
Lin et al. [99] (truss structure—3D), (iv) Zhou et al. [185] less computational cost, (iii) Wang et al. [176] compared
(wireless sensors) and (v) Li et al. [95, 96] also studied com- FE model updating using NSGA-II, differential evolution
posite plates and used wavelet decomposition. for multi-objectives (DEMO) and multi-objective particle
With positioned sensors, the numerical approach to the swarm optimization (MOPSO) and noted that DEMO out-
treatment of these data considering inverse methods can performed NSGA-II and MOPSO for all damage patterns
have a multi-objective approach. It is important to note that and (iv) Alkayem et al. [9] used a methodology that uses at
these data can be acquired directly in the structure and/or the same time two meta-heuristics and the TOPSIS getting
through finite element (FE) model updating procedure. As good results even with bad conditions or with incomplete
can be seen in Alexandrino et al. [66], where the authors date.
proposed a robust method of identifying ellipses and circu- Slipping away from the application of multi-objective
lar holes in composite plates. The ellipse having one of its optimization in the positioning of sensors or in the iden-
rays much larger than the other could be considered in the tification of damage itself, a relevant article was found on
authors' work as a crack. the planning application of the SHM methodology: Kim
Figure 6 shows a summary of a standard methodology and Frangopol [86] proposed a probabilistic optimum SHM
applied in SHM when using a multi-objective approach. planning based on the probabilistic fatigue damage assess-
First, the structure to be analyzed is defined. Then, a study of ment which aimed to determine the most important param-
the strategy to be adopted in relation to the sensors is made. eters and their weights in order to have the best possible
Finally, an iterative process takes place using an optimiza- planning of methodology.
tion algorithm, usually a meta-heuristic, until some conver- According to Alkayem et al. [8], some of the studies
gence criterion is reached (number of iterations or stipulated in Table 5 have effectively demonstrated the advantage of
difference). multi-objective functions rather than conversion into sin-
Perera and Ruiz [133] successfully applied a methodol- gle-objective functions using the weighted sum method.
ogy to large structures such as bridges through a two-step This is because when using a combination between a

13
 Table 5  Multi-objective optimization system in structural health monitoring
Authors Objective functions Decision variables Optimization method Object

Ferentinos and Tsiligiridis [48] Total energy related to parameters and sens- Uniformity of sensors points, number of GA (mono-objective) with weighted sum Composite
ing points uniformity sensors, operational energy consumption method plates
battery energy consumption
Gomes et al. [66] Fisher information matrix and difference Location and number of sensors NSGA Composite
module of mode shapes plates
Lin et al. [99] Response covariance sensitivity and correla- Location and number of sensors NSGA-II Truss structure
tion analysis
Zhou et al. [185] Information effectiveness (mode shapes) and Location and number of sensors MO discrete firefly algorithm based on Composite
network performance (standard deviation of neighboring information (MDFA/NS) and plates
total energy) NSGA II
Li et al. [95, 96] Number of sensors and vibration response Location and number of sensors NSGA-II and Wavelet decomposition Composite
plates
Alexandrino et al. [66] Differences in the mean stress and variance x, y, r (circular hole) x, y, a, b, θ (elliptical NSGA II, neural networking and fuzzy deci- Composite
hole) sion making plates
Kim and Park [85] Difference module of the natural frequencies Element position and stiffness reduction fac- GA with sum method and MOGA Composite
and Modal Assurance Criterion (MAC)— tor in this element plates
related to mode shapes
Perera and Ruiz [133] Modal frequencies and mode shapes Element location and damage index SPGA Truss structure
Cha and Buyukozturk [26] Modal Strain Energy and mode shapes Structural element and Young´s modulus Implicit Redundant Representation GA and Truss structure
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering…

reduction NSGA-II
Wang et al. [176] Natural Frequencies and accumulative MAC Truss element and reduction ration MOPSO, NSGA-II and MODE Truss structure
Alkayem et al. [8] Difference module of the natural frequencies Truss element and Young’s modulus reduc- GA, MOGA, PSO and MOPSO Truss structure
and MAC tion
Alkayem et al.[9] Modal strain energy and mode shape Truss element and Young’s modulus reduc- MOPSO and Lévy flights Truss structure
tion
Kim and Frangopol [86] Expected damage detection delay, expected Uncertainties of fatigue damage, maintenance Multi-objective probabilistic optimization –
maintenance delay, damage detection time- delay, damage detection delay, effects if process (MOPOP)
based reliability index, expected total life maintenance actions, service life and costs,
extension and expected life cycle cost maintenance and structural failure

13
2301
2302 J. L. J. Pereira et al.

Fig. 6  Framework for SHM in a

multi-objective approach

single-objective optimization algorithm (such EA) and the However, the amount of information in the literature can
weighting sum method to solve multiple objectives, the out- make the choice of a methodology to address a problem
come is a sub-set of total Pareto optimal solutions, while confusing and this work aimed to show through a system-
a powerful multi-objective meta-heuristic algorithm can atic and detailed review of the literature what are the most
generate the whole Pareto optimal solutions or at least the used algorithms, techniques, decision variables and objec-
majority of them. tive functions employed in ninety different research papers
in mechanical engineering.
3.4 General Discussion It can be seen that classic optimization methods, such
as gradient-based methods, that had their importance in
All recent or relevant publications found for multi-objective the past lost space for new algorithms that emerged with
optimization in mechanical engineering in this work are in the advancement of computing, better able to deal with a
Tables 1, 2, 3, 4 and 5, where it contains the decision varia- greater number of variables, multiple objectives and nonlin-
bles, the objective functions, the algorithms and the structure earities. Meta-heuristics are the most suitable to deal with
or process to be optimized employed. In total, 90 researches multi-objective optimization problems, being evolution-
were detailed from the point of view of the multi-objective ary algorithms the most used in the literature, mainly the
approach. NSGA-II. Next comes a swarm algorithm, MOPSO. Even
Of these works, 23 different algorithms were used and the evolutionary and swarm algorithms are being challenged
since in some works more than one is used, it totaled 102 by new and more powerful meta-heuristics with more appro-
applications. NSGA-II was the most used, appearing in 32 priate routines for specific problems, such as Jaya algorithm,
researches, just behind came MOPSO with 11 appearances, Multi-objective Grey Wolf Optimizer, Success History-
MOGA with 5, Jaya with 5, MODE with 3 and Ortoghonal based Adaptive Multi-objective Differential Evolution,
Method with 3. The other algorithms found in the Table Real-code Population-Based Incremental Learning hybrid-
appeared less than twice and this includes the only gradient- ized with Adaptive Differential evolution or Non-dominated
based method in the list: the GRG. Note that of the 23 algo- sorting Teaching–Learning Based algorithm. Since most of
rithms found, 18 appeared less than twice. these algorithms use the Pareto dominance relationship to
A widely used methodology was that of RSM, where the find non-dominated solutions, the overwhelming majority of
authors performed analysis using DOE or Taguchi and found recent problems use a posteriori decision-making technique,
polynomial regression curves that were optimized using where the decision maker tries to come up with all possible
desirability function or other algorithms. In all, 18 studies optimal solutions to the problem before choosing the best
followed this methodology. Of those who used and detailed one. A widely used a posteriori decision-making technique
the decision-making technique adopted, TOPSIS was the is TOPSIS.
most present, being in 6 of the works found. Even so, it was possible to verify that these powerful
tools of multi-objective optimization are still little used in
mechanical engineering and who made their use, obtained
4 Conclusion great improvement in the object in which they work.

Multi-objective optimization is an area that has been highly Acknowledgements The authors would like to acknowledge the
financial support from the Brazilian agency CNPq (Conselho
developed in the last decades and several methodologies, Nacional de Desenvolvimento Científico e Tecnológico), CAPES
algorithms and decision techniques have been created.

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2303

(Coordenação de Aperfeiçoamento de Pessoal de Nível Superior) loading in the presence of stress constraint. J Mech Behav Mater
and FAPEMIG (Fundação de Amparo à Pesquisa do Estado de Minas 26(3–4):79–93
Gerais—APQ-00385-18). 15. Back T (1996) Evolutionary algorithms in theory and practice:
evolution strategies, evolutionary programming, genetic algo-
rithms. Oxford University Press, Oxford
Declarations 16. Bahadormanesh N, Rahat S, Yarali M (2017) Constrained
multi-objective optimization of radial expanders in organic
Conflict of interest The authors declare that they have no conflict of Rankine cycles by firefly algorithm. Energy Convers Manage
interest. 148:1179–1193
17. Baril C, Yacout S, Clément B (2011) Design for Six Sigma
through collaborative multi-objective optimization. Comput
Ind Eng 60:43–55
References 18. Belinato G, Almeida FA, Paiva AP, Freitas Gomes JH, Bal-
estrassi PP, Rosa PARC (2018) A multivariate normal bound-
1. Abidi MH, Al-Ahmari AM, Umer U, Rasheed MS (2018) Multi- ary intersection PCA-based approach to reduce dimensional-
objective optimization of micro-electrical discharge machining ity in optimization problems for LBM process. Eng Comput
of nickel-titanium-based shape memory alloy using MOGA-II. 35(4):1533–1544
Measurement 125:336–349 19. Benayoun B, Sussman B (1966) Electre: Une methode pour
2. Afshari H, Hare W, Tesfamariam S (2019) Constrained multi- guider le choix en presence de points de vue multiple. Direc-
objective optimization algorithms: review and comparison with tion Scientifique, Note de Travail, No. 49
application in reinforced concrete structures. Appl Soft Comput 20. Bilel N, Mohamed N, Zouhaier A, Lotfi R (2016) An improved
83:105631 imperialist competitive algorithm for multi-objective optimiza-
3. Ahire PG, Patil US, Kadam MS (2018) Genetic algorithm based tion. Eng Optim 48(11):1823–1844. https://​doi.​org/​10.​1080/​
optimization of the process parameters for manual metal arc 03052​15x.​2016.​11412​04
welding of dissimilar metal joint. Procedia Manuf 20:106–112 21. Brans JP, Vincke P, Mareschal B (1986) How to select and how
4. Ahmad MA, Sheikh AK, Nazir K (2019) Design of experiment to rank projects: the PROMETHEE method. Eur J Oper Res
based statistical approaches to optimize submerged arc welding 24(2):228–238
process parameters. ISA Trans. https://​doi.​org/​10.​1016/j.​isatra.​ 22. Branke J, Kaußler T, Schmeck H (2001) Guidance in evolution-
2019.​04.​003 ary multi-objective optimization. Adv Eng Softw 32:499–507
5. Ahmadi P, Hajabdollahi H, Dincer I (2011) Cost and entropy 23. Brassard G, Bratley P (1988) Algorithmics: theory and prac-
generation minimization of a cross flow plate fin heat exchanger tice. Prentice-Hall, Englewood Cliffs, New Jersey
using multi-objective genetic algorithm. J Heat Transfer 24. Brito TG, Paiva AP, Ferreira JR, Gomes JHF (2014) Balestrassi
133(2):021801–021810 PPA normal boundary intersection approach to multiresponse
6. Al Dawood ZIA, Saadoon AM (2017) Multi response optimiza- robust optimization of the surface roughness in end milling
tion of submerged arc welding using Taguchi fuzzy logic based process with combined arrays. Precis Eng 38(3):628–638
on utility theory. Int J Sci Res 6(12):475–481 25. Brito TG, Paiva AP, Paula TI, Dalosto DN, Ferreira JR,
7. Alexandrino PDSL, Gomes GF, Cunha SS Jr (2020) A robust Balestrassi PP (2016) Optimization of AISI 1045 end mill-
optimization for damage detection using multi-objective genetic ing using robust parameter design. Int J Adv Manuf Technol
algorithm, neural network and fuzzy decision making. Inverse 84(5–8):1185–1199
Prob Sci Eng 28(1):21–46 26. Cha Y, Buyukozturk O (2015) Structural damage detection
8. Alkayem NF, Cao M, Zhang Y, Bayat M, Su Z (2017) Struc- using modal strain energy and hybrid multi-objective optimiza-
tural damage detection using finite element model updating tion. Comput Aided Civ Infrastruct Eng 30:347–358
with evolutionary algorithms: a survey. Neural Comput Appl 27. Chen F, Wang Y, Sun S, Ma Z, Huang X (2018) Multi-objec-
30(2):389–411 tive optimization of mechanical quality and stability during
9. Alkayem NF, Cao M, Ragulskis M (2018) Damage diagnosis in micro resistance spot welding. Int J Adv Manuf Technol
3D structures using a novel hybrid multiobjective optimization 101(5):1903–1913
and FE model updating framework. Complexity. https://​doi.​org/​ 28. Chiandussi G, Codegone M, Ferrero S, Varesio FE (2012)
10.​1155/​2018/​35416​76 Comparison of multi-objective optimization methodologies
10. Almeida FA, Gomes GF, Paula VR, Correa JE, Paiva AP, for engineering applications. [S.I.]. pp 912–942, Elsevier,
Gomes JHF, Turrioni JB (2018) A weighted mean square error Amsterdam
approach to the robust optimization of the surface roughness in 29. Choudhary A, Kumar M, Unune DR (2019) Experimental
an AISI 12L14 free-machining steel-turning process. J Mech Eng investigation and optimization of weld bead characteristics
64(3):147–156 during submerged arc welding of AISI 1023 steel. Defence
11. Almeida FA, Santos ACO, Paiva AP, Gomes GF, Gomes JHF Technol 15(1):72–82
(2020) Multivariate Taguchi loss function optimization based on 30. Coello Coello CA, Corte´s NC, (2005) Solving multi-objec-
principal components analysis and normal boundary intersection. tive optimization problems using an artificial immune system.
Eng Comput, pp 1–17 Genet Program Evol Mach 6(2):163–190
12. Arian Nik M, Fayazbakhsh K, Pasini D, Lessard L (2012) Surro- 31. Coello C, Lamont B, Van Veldhuizen D (2007) Evolutionary
gatebased multi-objective optimization of a composite laminate algorithms for solving multi-objective problems. Springer,
with curvilinear fibers. Compos Struct 94(8):2306–2313 New York. https://​doi.​org/​10.​1007/​978-0-​387-​36797-2.
13. Asanjarani A, Dibajian SH, Mahdian A (2017) Multi-objective 32. Cohon JL, Marks DH (1975) A review and evaluation of
crashworthiness optimization of tapered thin-walled square tubes multi-objective programming techniques. Water Resour Res
with indentations. Thin-Walled Struct 116:26–36 11(2):208–220
14. Ashjari M, Khoshravan MR (2017) Multi-objective optimiza- 33. Cohon JL (1978) Multi-objective programming and planning.
tion of a functionally graded sandwich panel under mechanical Academic Press, Cambridge

13
2304 J. L. J. Pereira et al.

34. Cui Y, Geng Z, Zhu Q, Han Y (2017) Review: multi-objective and their applications: proceedings of the first international con-
optimization methods and application in energy saving. Energy ference on genetic algorithms, pp 141–153. Lawrence Erlbaum,
125:681–704 Hillsdale, New Jersey
35. Custódio AL, Madeira JFA, Vaz AIF, Vicente LN (2011) Direct 53. Francisco MB, Pereira JLJ, Oliver GA, Silva FHS, Cunha SS,
multisearch for multi-objective optimization. SIAM J Optim Gomes GF (2021) Multi-objective design optimization of crp
21(3):1109–1140 isogrid tubes using sunflower multi-objective optimization based
36. Das I, Dennis JE (1998) Normal-boundary intersection: a new on metamodel. Comput Struct 249:106508
method for generating the pareto surface in nonlinear multicri- 54. Franco Correia VM, Aguilar Madeira JF, Araújo AL, Mota
teria optimization problems. SIAM J Optim 8(3):631–657 Soares CM (2018) Multi-objective optimization of ceramic-metal
37. Deb K (2008) Introduction to evolutionary multi-objective opti- functionally graded plates using a higher order model. Compos
mization. multi-objective optimization. Springer, Berlin, pp Struct 183:146–160
59–96 55. Franco Correia VM, Aguilar Madeira JF, Araújo AL, Mota
38. Deb K,Agrawal S, Pratab A,Meyarivan T (2000) A fast elitist Soares CM (2019) Multi-objective optimization of functionally
non dominated sorting genetic algorithm for multi-objective graded material plates with thermo-mechanical loading. Compos
optimization: NSGA-II. In:Schoenauer M, Deb K,Rudolph G, Struct 207:845–857
Yao X, Lutton E, Merelo JJ, Schwefel H-P (eds) Proceedings 56. Gandibleux X, Mezdaoui N, Fr´eville A (1997) A tabu search
of the parallel problem solving from nature VI conference, pp procedure to solve combinatorial optimisation problems. In:
849–858, Paris, France. Lecture Notes in Computer Science No. Caballero R, Ruiz F, References 667 Steuer RE (eds) Advances
1917, Springer, Berlin in multiple objective and goal programming, volume 455 of lec-
39. Dhiman G, Singh KK, Soni M, Nagar A, Dehghani M, Slowik A, ture notes in economics and mathematical systems, pp 291–300.
Cengiz K (2020) MOSOA: a new multi-objective seagull optimi- Springer-Verlag
zation algorithm. Expert systems with applications, p 114150 57. Gao Z, Shao X, Jiang P, Wang C, Zhou Q, Cao L, Wang Y (2016)
40. Dhiman G, Singh KK, Slowik A, Chang V, Yildiz AR, Kaur A, Multi-objective optimization of weld geometry in hybrid fiber
Garg M (2020) EMOSOA: a new evolutionary multi-objective laser-arc butt welding using Kriging model and NSGA-II. Appl
seagull optimization algorithm for global optimization. Int J Phys A 122(6):1–12
Mach Learn Cybern 12(2):571–596 58. Gaudêncio DJH, Almeida FA, Turrioni JB, Costa Quinino R,
41. Diniz CA, Mendez YAD, Almeida FA, Cunha Jr SS (2021) Opti- Balestrassi PP, Paiva AP (2019) A multiobjective optimization
mum design of composite structures with ply dropp-offs using model for machining quality in the AISI 12L14 steel turning
responde surface methodology. Eng Comput process using fuzzy multivariate mean square error. Precis Eng
42. Duckstein L (1984) Multi-objective optimization in structural 56:303–320
design: the model choice problem. In: Atrek E, Gallagher RH, 59. Ghasemi AR, Hajmohammad MH (2016) Multi-objective optimi-
Ragsdell KM, Zienkiewicz OC (eds) New directions in optimum zation of laminated composite shells for minimum mass/cost and
structural design. John Wiley and Sons, New Jersey, pp 459–481 maximum buckling pressure with failure criteria under external
43. Dumbhare AP, Dubey S, Deshpande V, Y., Andhare, A. B., & hydrostatic pressure. Struct Multidiscip Optim 55(3):1051–1062
Barve, P. S. (2018) Modelling and multi-objective optimization 60. Gupta SK, Pandey K, Kumar R (2016) Multi-objective opti-
of surface roughness and kerf taper angle in abrasive water jet mization of friction stir welding process parameters for join-
machining of steel. J Braz Soc Mech Sci Eng 40(5):259 ing of dissimilar AA5083/AA6063 aluminum alloys using
44. Ebrahimi-Nejad S, Kheybari M, Borujerd SVN (2020) Multi- hybrid approach. Proc Inst Mech Eng Part L J Mater Des Appl
objective optimization of a sports car suspension system using 232(4):343–353
simplified quarter-car models. Mech Ind 21(4):412 61. Gupta SK, Pandey K, Kumar R (2016) Artificial intelligence-
45. Emmerich MTM, Deutz AH (2018) A tutorial on multi-objective based modelling and multi-objective optimization of friction
optimization: fundamentals and evolutionary methods. Nat Com- stir welding of dissimilar AA5083-O and AA6063-T6 alu-
put 17(3):585–609. https://​doi.​org/​10.​1007/​s11047-​018-​9685-y minium alloys. Proc Inst Mech Eng Part L J Mater Des Appl
46. Fadaee M, Radzi MAM (2012) Multi-objective optimization 232(4):333–342
of a stand-alone hybrid renewable energy system by using 62. Ghachi RF, Alnahhal WI, Abdeljaber O, Renno J, Haque ABMT,
evolutionary algorithms: a review. Renew Sustain Energy Rev Shim J, Aref A (2020) Optimization of viscoelastic metamateri-
16(5):3364–3369 als for vibration attenuation properties. Int J Appl Mech
47. Fan H, Zhang J, Zhang W, Liu B (2020) Multiparameter and 63. Goicoechea A, Duckstein L, Fogel M (1976) Multi-objective pro-
multiobjective optimization design based on orthogonal method gramming in watershed management: a study of the Charleston
for mixed flow fan. Energies 13(11):2819 watershed. Water Resour Res 12(6):1085–1092
48. Ferentinos KP, Tsiligiridis TA (2007) Adaptive design optimiza- 64. Goldberg DE (1989) Genetic algorithms in search, optimization
tion of wireless sensor networks using genetic algorithms. Com- and machine learning. Addison-Wesley Publishing Company,
put Netw 51(4):1031–1051 Reading, Massachusetts
49. Fonseca CM, Fleming PJ (1994) An overview of evolutionary 65. Gomes GF, Giovani RS (2020) An efficient two-step damage
algorithms in multi-objective optimization. Technical report, identification method using sunflower optimization algorithm
department of automatic control and systems engineering, Uni- and mode shape curvature (MSDBI–SFO). Eng Comput. https://​
versity of Sheffield, Sheffield, UK doi.​org/​10.​1007/​s00366-​020-​01128-2
50. Fonseca CM, Fleming PJ (1993) Genetic algorithms for multi- 66. Gomes GF, Almeida FA, Alexandrino PSL, Cunha SS, Sousa BS,
objective optimization: Formulation discussion and generaliza- Ancelotti AC (2018) A multi-objective sensor placement optimi-
tion. In: Proceedings of the international conference on genetic zation for SHM systems considering Fisher information matrix
algorithms, vol 93. Citeseer, pp 416–423 and mode shape interpolation. Eng Comput 35(2):519–535
51. Fossati GG, Miguel LFF, Casas WJP (2019) Multi-objective opti- 67. Gomes J (2013) Método dos polinômios canônicos de misturas
mization of the suspension system parameters of a full vehicle para otimização multi-objetivo. Itajubá, Minas Gerais, Brasil:
model. Optim Eng 20:151–177 Doctoral Thesis - Postgraduate Program in Production Engineer-
52. Fourman MP (1985) Compaction of symbolic layout using ing – Universidade Federal de Itajubá
genetic algorithms. In: Grefenstette JJ (ed) Genetic algorithms

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2305

68. Gunantara N (2018) A review of multi-objective optimization: 88. Kitayama S, Yokoyama M, Takano M, Aiba S (2017) Multi-
Methods and its applications. Cogent Eng 5(1):1–16 objective optimization of variable packing pressure profile
69. Guo X, Wu Y, Xie L, Cheng S, Xin J (2015) An adaptive brain and process parameters in plastic injection molding for mini-
storm optimization algorithm for multi-objective optimization mizing warpage and cycle time. Int J Adv Manuf Technol
problems. Lecture notes in computer science, pp 365–372 92(9–12):3991–3999
70. Haimes YY, Lasdon LS, Wismer DA (1971) On a bicriterion for- 89. Knowles JD, Corne DW (2000) Approximating the nondomi-
mulation of the problems of integrated system identification and nated front using the Pareto archived evolution strategy. Evol
system optimization. IEEE Trans Syst Man Cybern 1(3):296–297 Comput 8:149–172
71. Hajela P, Lin CY (1992) Genetic search strategies in multicrite- 90. Kumar R, Singh S, Bilga PS, Singh J, Singh S, Scutaru M-L,
rion optimal design. Struct Optim 4:99–107 Pruncu CI (2021) Revealing the benefits of entropy weights
72. Heidari AA, Mirjalili S, Faris H, Aljarah I, Mafarja M, Chen H method for multi-objective optimization in machining operations:
(2019) Harris hawks optimization: algorithm and applications. a critical review. J Mater Res Technol 10:1471–1492
Futur Gener Comput Syst 97:849–872 91. Kursawe F (1991) A variant of evolution strategies for vector
73. Holland J (1975) Adaptation in natural and artificial systems. optimization. In: Schwefel H-P, Manner R (eds) Parallel prob-
University of Michigan Press, Ann Arbor lem solving from nature. 1st Workshop, PPSN I, pp 193–197,
74. Horn J, Nafpliotis N (1993) Multi-objective optimization using Dortmund, Germany. Springer-Verlag. Lecture notes in computer
the niched pareto genetic algorithm. Technical Report IlliGAl science No. 496
Report 93005, University of Illinois at Urbana-Champaign, 92. Lee D, Morillo C, Bugeda G, Oller S, Onate E (2012) Multilay-
Urbana, Illinois, USA ered composite structure design optimisation using distributed/
75. Horn J, Nafpliotis N, Goldberg DE (1994) A niched pareto parallel multi-objective evolutionary algorithms. Compos Struct
genetic algorithm for multi-objective optimization. In: Proceed- 94(3):1087–1096
ings of the first IEEE conference on evolutionary computation, 93. Leguizamón G, Coello CAC (2011) Multi-objective ant colony
IEEE world congress on computational intelligence, vol 1, pp optimization: a taxonomy and review of approaches. Series in
82–87, Piscataway, New Jersey. IEEE Service Center machine perception and artificial intelligence, pp 67–94
76. Hwang CL, Yoon KP (1981) Multiple attribute decision making: 94. Li H, Wang X, Wei Y, Liu T, Gu J, Li Z, Liu Y (2017) Multi-
methods and applications. Springer-Verlag, Berlin/Heidelberg/ objective optimizations of biodegradable polymer stent structure
New York and stent microinjection molding process. Polymers 9(12):20
77. Ikeya K, Shimoda M, Shi J-X (2016) Multi-objective free-form 95. Li P, Huang L, Peng J (2018) Sensor distribution optimization
optimization for shape and thickness of shell structures with for structural impact monitoring based on NSGA-II and wavelet
composite materials. Compos Struct 135:262–275 decomposition. Sensors 18(12):4264
78. Jaimes AL, Martınez SZ, Coello CAC et al (2009) An introduc- 96. Li K, Yan S, Zhong Y, Pan W, Zhao G (2018) Multi-objective
tion to multi-objective optimization techniques. Optim Polym optimization of the fiber-reinforced composite injection molding
Process, pp 29–57. process using Taguchi method, RSM, and NSGA-II. Simulation
79. Jiang R, Wang D (2016) Optimization of suspension system modelling practice and theory
of self-dumping truck using TOPSIS-based Taguchi method 97. Li Y, Wei K, Yang W, Wang Q (2020) Improving wind turbine
coupled with entropy measurement, SAE Technical Paper, blade based on multi-objective particle swarm optimization.
2016-2001-1385 Renew Energy. https://​doi.​org/​10.​1016/j.​renene.​2020.​07.​067
80. Jiang R, Wang D (2015) Optimization of vehicle ride comfort 98. Lin W, Yu D, Zhang C, Zhang S, Tian Y, Liu S, Luo M (2016)
and handling stability based on TOPSIS Method, SAE Technical Multi-objective optimization of machining parameters in multi-
Paper, 2015-200-1348 pass turning operations for low-carbon manufacturing. Proc Inst
81. Jiang P, Wang C, Zhou Q, Shao X, Shu L, Li X (2016) Optimiza- Mech Eng Part B J Eng Manuf 231(13):2372–2383
tion of laser welding process parameters of stainless steel 316L 99. Lin J-F, Xu Y-L, Law S-S (2018) Structural damage detection-
using FEM, Kriging and NSGA-II. Adv Eng Softw 99:147–160 oriented multi-type sensor placement with multi-objective opti-
82. Jin Y, Sendhoff B (2002) Fuzzy preference incorporation into mization. J Sound Vib 422:568–589
evolutionary multi-objective optimization. In: Proceedings of the 100. Liu H, Li Y, Duan Z, Chen C (2020) A review on multi-objective
4th Asia-pacific conference on simulated evolution and learning, optimization framework in wind energy forecasting techniques
vol 1, pp 26–30 and applications. Energy Conv Manage 224:113324
83. Kalantari M, Dong C, Davies IJ (2016) Multi-objective robust 101. Liu Q, Li X, Liu H, Guo Z (2020) Multi-objective metaheuristics
optimisation of unidirectional carbon/glass fibre reinforced for discrete optimization problems: a review of the state-of-the-
hybrid composites under flexural loading. Compos Struct art. Appl Soft Comput 93:106382
138:264–275 102. Long Q, Wu X, Wu C (2021) Non-dominated sorting methods for
84. Karimi M, Hall M, Buckham B, Crawford C (2016) A multi- multi-objective optimization: review and numerical comparison.
objective design optimization approach for floating offshore J Ind Manage Optim 17(2):1001
wind turbine support structures. J Ocean Eng Marine Energy 103. Lostado Lorza R, Escribano García R, Fernandez Martinez R,
3(1):69–87 Martínez Calvo M (2018) Using genetic algorithms with multi-
85. Kim G, Park Y (2004) An improved updating parameters method objective optimization to adjust finite element models of welded
and finite element model updating using multi-objective optimi- joints. Metals 8(4):230
sation technique. Mech Syst Signal Process 18(1):59–78 104. Lu C, Gao L, Li X, Zheng J, Gong W (2018) A multi-objective
86. Kim S, Frangopol DM (2016) Efficient multi-objective optimisa- approach to welding shop scheduling for makespan, noise pollu-
tion of probabilistic service life management. Struct Infrastruct tion and energy consumption. J Clean Prod 196:773–787
Eng 13(1):147–159. https://​doi.​org/​10.​1080/​15732​479.​2016.​ 105. Lu C, Gao L, Li X, Xiao S (2017) A hybrid multi-objective grey
11984​05 wolf optimizer for dynamic scheduling in a real-world welding
87. Kitayama S, Miyakawa H, Takano M, Aiba S (2016) Multi-objec- industry. Eng Appl Artif Intell 57:61–79
tive optimization of injection molding process parameters for 106. Mahfouf M-Y, Linkens DA (2004) Adaptive weighted particle
short cycle time and warpage reduction using conformal cooling swarm optimisation for multi-objective optimal design of alloy
channel. Int J Adv Manuf Technol 88(5–8):1735–1744 steels. In: Parallel problem solving from nature - PPSN VIII, pp

13
2306 J. L. J. Pereira et al.

762–771, Birmingham, UK. Springer-Verlag. Lecture notes in congress on evolutionary computation (IEEE World Congress on
computer science vol 3242 Computational Intelligence), IEEE, pp 1128–1134
107. Mane SU, Rao MRN (2017) Many-objective optimization: prob- 126. Omkar S, Mudigere D, Naik GN, Gopalakrishnan S (2008)
lems and evolutionary algorithms – a short review. Int J Appl Vector evaluated particle swarm optimization (VEPSO) for
Eng Res 12(20):9774–9793 multi-objective design optimization of composite structures.
108. Marler RT, Arora JS (2004) Survey of multi-objective opti- Comput Struct 86(1):1–14
mization methods for engineering. Struct Multidiscip Optim 127. Osyczka A (1978) An approach to multicriterion optimization
26(6):369–395 problems for engineering design. Comput Methods Appl Mech
109. Mariano CE, Morales E (1999) MOAQ an ant-q algorithm for Eng 15:309–333
multiple objective optimization problems. In: Banzhaf W, Daida 128. Paiva AP, Gomes JHF, Peruchi RS, Leme RC, Balestrassi PP
J, Eiben AE, Garzon MH, Honavar V, Jakiela M, Smith RE (eds) (2014) A multivariate robust parameter optimization approach
Genetic and evolutionary computing conference (GECCO 99), based on principal component analysis with combined arrays.
vol 1, pp 894–901, San Francisco, California, Morgan Kaufmann Comput Ind Eng 74:186–198
110. Song M-P, Gu G-C (2004) Research on particle swarm optimiza- 129. Panagant N, Pholdee N, Wansasueb K, Bureerat S, Yildiz AR,
tion: a review. In: Proceedings of 2004 international conference Sait SM (2019) Comparison of recent algorithms for many-
on machine learning and cybernetics (IEEE Cat. No.04EX826) objective optimisation of an automotive floor-frame. Int J Vehi-
111. Messac A, Mattson CA (2002) Generating well-distributed sets cle Des 80(2/3/4):176
of Pareto points for engineering design using physical program- 130. Panagant N, Pholdee N, Bureerat S et al (2021) A comparative
ming. Optim Eng 3:431–450 study of recent multi-objective metaheuristics for solving con-
112. Mia M, Gupta MK, Lozano JA, Carou D, Pimenov DY, Król- strained truss optimisation problems. Arch Computat Methods
czyk G, Dhar NR (2018) Multi-objective optimization and life Eng. https://​doi.​org/​10.​1007/​s11831-​021-​09531-8
cycle assessment of eco-friendly cryogenic N2 assisted turning 131. Paula TI, Gomes GF, Freitas Gomes JH, Paiva AP (2019). A
of Ti-6Al-4V. J Clean Prod 210:121–133 mixture design of experiments approach for genetic algorithm
113. Miettinen K (1998) No-preference methods. Int Ser Oper Res tuning applied to multi-objective optimization. Optim Complex
Manage Sci, pp 67–76 Syst Theory Models Algorithms Appl, pp 600–610.
114. Mirjalili S, Jangir P, Saremi S (2016) Multi-objective ant lion 132. Peng Y, Wang X, Xiong X, Xu P (2016) Crashing analysis and
optimizer: a multi-objective optimization algorithm for solving multi-objective optimisation of duplex energy-absorbing struc-
engineering problems. Appl Intell 46(1):79–95. https://​doi.​org/​ ture for subway vehicle. Int J Crashworthiness 21(4):338–352
10.​1007/​s10489-​016-​0825-8 133. Perera R, Ruiz A (2008) A multistage FE updating procedure for
115. Mirjalili S, Saremi S, Mirjalili SM, Coelho LDS (2016) Grass- damage identification in large-scale structures based on multi-
hopper optimization algorithm for multi-objective optimization objective evolutionary optimization. Mech Syst Signal Process
problems. Expert Syst Appl 47:106–119 22(4):970–991. https://​doi.​org/​10.​1016/j.​ymssp.​2007.​10.​004
116. Mirjalili SZ, Mirjalili S, Saremi S, Faris H, Aljarah I (2017) 134. Parsopoulos K, Vrahatis M (2002) Particle swarm optimiza-
Grasshopper optimization algorithm for multi-objective optimi- tion method in multi-objective problems. In: Proceedings of the
zation problems. Appl Intell 48(4):805–820 2002 ACM symposium on applied computing (SAC’2002), pp
117. Moleiro F, Madeira JFA, Carrera E, Reddy JN (2020) Design 603–607, Madrid, Spain, ACM Press, New York
optimization of functionally graded plates under thermo-mechan- 135. Park H-S, Nguyen T-T, Dang X-P (2016) Multi-objective opti-
ical loadings to minimize stress, deformation and mass. Compos mization of turning process of hardened material for energy effi-
Struct p 112360 ciency. Int J Precis Eng Manuf 17(12):1623–1631
118. Monarchi DE, Kisiel CC, Duckstein L (1973) Interactive multi- 136. Parkinson AR, Balling RJ, Hedengren JD (2013) Optimiza-
objective programming in water resources: a case study. Water tion methods for engineering design: applications and theory.
Resour Res 9(4):837–850 Brigham Young University
119. Montalvo-Urquizo J, Niebuhr C, Schmidt A, Villarreal-Marro- 137. Prakash C, Singh S, Singh M, Antil P, Aliyu AAA, Abdul-Rani
quín MG (2018) Reducing deformation, stress, and tool wear AM, Sidhu SS (2018) Multi-objective optimization of MWCNT
during milling processes using simulation-based multiobjective mixed electric discharge machining of Al–30SiCp MMC using
optimization. Int J Adv Manuf Technol 96(5–8):1859–1873 particle swarm optimization. materials horizons: From Nature to
120. Mostaghim S, Teich J (2003) Strategies for finding good local Nanomaterials, pp 145–164
guides in multi-objective particle swarm optimization (MOPSO). 138. Prakash C, Kansal HK, Pabla BS, Puri S (2016) Multi-objective
In: 2003 IEEE swarm intelligence symposium proceedings, pp optimization of powder mixed electric discharge machining
26–33, Indianapolis, Indiana, USA, IEEE Service Center parameters for fabrication of biocompatible layer on β-Ti alloy
121. Neumann and Morgenstern (1944) Theory of games and eco- using NSGA-II coupled with Taguchi based response surface
nomic behavior. Princeton University Press, Princeton, New methodology. J Mech Sci Technol 30(9):4195–4204
Jersey 139. Qu S, Zhao J, Wang T (2016) Experimental study and machining
122. Niu Y, Jiao F, Zhao B, Wang D (2017) Multiobjective optimi- parameter optimization in milling thin-walled plates based on
zation of processing parameters in longitudinal-torsion ultra- NSGA-II. Int J Adv Manuf Technol 89(5–8):2399–2409
sonic assisted milling of Ti-6Al-4V. Int J Adv Manuf Technol 140. Rangaiah GP, Zemin F, Hoadley AF (2020) Multi-objective opti-
93(9–12):4345–4356 mization applications in chemical process engineering: tutorial
123. Ojstersek R, Brezocnik M, Buchmeister B (2020) Multi-objective and review. J Process. https://​doi.​org/​10.​3390/​pr805​0508
optimization of production scheduling with evolutionary compu- 141. Rao RV, Rai DP, Balic J (2016) Multi-objective optimization of
tation: a review. Int J Ind Eng Comput 11(3):359–376 machining and micro-machining processes using non-dominated
124. Okabe T, Oya Y, Yamamoto G, Sato J, Matsumiya T, Matsu- sorting teaching–learning-based optimization algorithm. J Intell
zaki R, Obayashi S (2017) Multi-objective optimization for resin Manuf 29(8):1715–1737
transfer molding process. Compos A Appl Sci Manuf 92:1–9 142. Rao RV, Saroj A, Ocloń P, Taler J, Taler D (2017) Single- and
125. Olorunda O, Engelbrecht AP (2008) Measuring exploration/ multi-objective design optimization of plate-fin heat exchangers
exploitation in particle swarms using swarm diversity. In: IEEE using jaya algorithm. Heat Transfer Eng 39(13–14):1201–1216

13
A Review of Multi‑objective Optimization: Methods and Algorithms in Mechanical Engineering… 2307

143. Rao RV, Rai DP (2017) Optimization of submerged arc welding 162. Sowrirajan M, Koshy Mathews P, Vijayan S (2018) Simultane-
process parameters using quasi-oppositional based jaya algo- ous multi-objective optimization of stainless steel clad layer on
rithm. J Mech Sci Technol 31(5):2513–2522 pressure vessels using genetic algorithm. J Mech Sci Technol
144. Rao RV, Rai DP, Balic J (2017) Multi-objective optimization of 32(6):2559–2568
abrasive waterjet machining process using Jaya algorithm and 163. Srinivas N, Deb K (1994) Multi-objective optimization using
PROMETHEE Method. J Intell Manuf 30(5):2101–2127 nondominated sorting in genetic algorithms. Evol Comput
145. Rao S (1984) Multi-objective optimization in structural design 2(3):221–248
with uncertain parameters and stochastic processes. AIAA J 164. Teughels A, Maeck J, Roeck G (2002) Damage assessment by
22(11):1670–1678 FE model updating using damage functions. Comput Struct
146. Rao S (1986) Game theory approach for multi-objective struc- 80:1869–1879
tural optimization. Comput Struct 25(1):119–127 165. Tamaki H, Kita H, Kobayashi S (1996). Multi-objective opti-
147. Rao SS, Rao SS (2009) Engineering optimization: theory and mization by genetic algorithms: a review. In: Proceedings of
practice. John Wiley and Sons, New Jersey IEEE international conference on evolutionary computation, pp
148. Ridha HM, Gomes C, Hizam H, Ahmadipour M, Heidari AA, 517–522
Chen H (2021) Multi-objective optimization and multi-criteria 166. Tawhid MA, Savsani V (2017) Multi-objective sine-cosine algo-
decision-making methods for optimal design of standalone rithm (MO-SCA) for multi-objective engineering design prob-
photovoltaic system: a comprehensive review. Renew Sustain lems. Neural Comput Appl 31(2):915–929
Energy Rev 135:110202 167. Thiele L, Miettinen K, Korhonen PJ, Molina J (2009) A prefer-
149. Rivas D, Quiza R, Rivas M, Haber RE (2020) Towards sustain- ence-based evolutionary algorithm for multi-objective optimiza-
ability of manufacturing processes by multi-objective optimi- tion. Evol Comput 17(3):411–436
zation: a case study on a submerged arc welding process. IEEE 168. Torres AF, Rocha FB, Almeida FA, Gomes JHF, Paiva AP, Bal-
Access 8:212904–212916 estrassi PP (2020) Multivariate stochastic optimization approach
150. Sadollah A, Eskandar H, Kim JH (2015) Water cycle algorithm applied in a flux-cored arc welding process. IEEE Access
for solving constrained multi-objective optimization problems. 8:61267–61276
Appl Soft Comput 27:279–298 169. Tripathy S, Tripathy DK (2017) Multi-response optimization
151. Saha A, Mondal SC (2017) Multi-objective optimization of of machining process parameters for powder mixed electro-dis-
manual metal arc welding process parameters for nano-struc- charge machining of H-11 die steel using grey relational analysis
tured hardfacing material using hybrid approach. Measurement and topsis. Mach Sci Technol 21(3):362–384
102:80–89 170. Vo-Duy T, Duong-Gia D, Ho-Huu V, Vu-Do HC, Nguyen-Thoi
152. Sahu NK, Andhare AB (2018) Multiobjective optimization T (2017) Multi-objective optimization of laminated compos-
for improving machinability of Ti-6Al-4V using RSM and ite beam structures using NSGA-II algorithm. Compos Struct
advanced algorithms. J Comput Des Eng 6(1):1–12 168:498–509. https://​doi.​org/​10.​1016/j.​comps​truct.​2017.​02.​038
153. Sailender M, Reddy GC, Venkatesh S (2018) Influences of 171. Xin-Gang Z, Ji L, Jin M, Ying Z (2020) An improved quantum
process parameters on weld strength of low carbon alloy steel particle swarm optimization algorithm for environmental eco-
in purged SAW. Mater Today Proc 5(1):2928–2937 nomic dispatch. Exp Syst Appl 152:113370
154. Silva MM, Batista VR, Maciel TM, dos Santos MA, Brasileiro 172. Wakchaure KN, Thakur AG, Gadakh V, Kumar A (2018) Multi-
TL (2018) Optimization of submerged arc welding process objective optimization of friction stir welding of aluminium
parameters for overlay welding. Weld Int 32(2):122–129 alloy 6082–T6 using hybrid taguchi-grey relation analysis- ANN
155. Schlieter T, Dlugosz A (2020) Structural optimization of aero- method. Mater Today Proc 5(2):7150–7159
foils for many criteria. 26°International Conference. Engineer- 173. Wang H, Olhofer M, Jin Y (2017) A mini-review on preference
ing Mechanics 2020. Brno, Czech Republic modeling and articulation in multi-objective optimization: cur-
156. Senthil SM, Parameshwaran R, Ragu Nathan S, Bhuvanesh rent status and challenges. Complex Intell Syst 3(4):233–245
Kumar M, Deepandurai K (2020) A multi-objective optimization 174. Wang J, Yan Z, Wang M, Li M, Dai Y (2013) Multi-objective
of the friction stir welding process using RSM-based-desirability optimization of an organic Rankine cycle (ORC) for low grade
function approach for joining aluminum alloy 6063–T6 pipes. waste heat recovery using evolutionary algorithm. Energy Con-
Struct Multidiscip Optim 62(3):1117–1133 vers Manage 71:146–158
157. Serafini P (1994) Simulated annealing for multiple objective opti- 175. Wang L, Wang T, Wu J, Chen G (2017) Multi-objective differ-
mization problems. In: Tzeng G, Wang H, Wen U, Yu P (eds) ential evolution optimization based on uniform decomposition
Proceedings of the tenth international conference on multiple cri- for wind turbine blade design. Energy 120:346–361
teria decision making: expand and enrich the domains of thinking 176. Wang Y, Ma Q, Li W (2012) Structural damage detection by
and application, vol 1, pp 283–292, Springer-Verlag, Berlin multi-objective intelligent algorithm. In: The 15th world confer-
158. Sivaiah P, Chakradhar D (2018) Performance improvement of ence on earthquake engineering, Lisbon
cryogenic turning process during machining of 17–4 PH stainless 177. Wang Y, Huo X (2018) Multiobjective optimization design and
steel using multi objective optimization techniques. Measure- performance prediction of centrifugal pump based on orthogonal
ment. https://​doi.​org/​10.​1016/j.​measu​rement.​2018.​12.​094 test. Adv Mater Sci Eng 2018:1–10
159. Schaffer JD (1984) Multiple objective optimization with vector 178. Warsi SS, Agha MH, Ahmad R, Jaffery SHI, Khan M (2018)
evaluated genetic algorithms. PhD thesis, Vanderbilt University, Sustainable turning using multi-objective optimization: a study
Nashville, Tennessee of Al 6061 T6 at high cutting speeds. Int J Adv Manuf Technol
160. Shanjeevi C, Satish Kumar S, Sathiya P (2014) Multi-objective 100(1–4):843–855
optimization of friction welding parameters in AISI 304L auste- 179. Yang X-S (2014) Nature-inspired optimization algorithms. Else-
nitic stainless steel and copper joints. Proc Inst Mech Eng Part vier, Amsterdam
B J Eng Manuf 230(3):449–457 180. Yang Y, Cao L, Zhou Q, Wang C, Wu Q, Jiang P (2018) Multi-
161. Shao Q, Xu T, Yoshino T, Song N (2017) Multi-objective opti- objective process parameters optimization of Laser-magnetic
mization of gas metal arc welding parameters and sequences hybrid welding combining Kriging and NSGA-II. Robot Comput
for low-carbon steel (Q345D) T-joints. J Iron Steel Res Int Integr Manuf 49:253–262
24(5):544–555

13
2308 J. L. J. Pereira et al.

181. Yang Y, Cao L, Wang C, Zhou Q, Jiang P (2018) Multi-objective 190. Zhang R, Wang X (2019) Parameter study and optimization of
process parameters optimization of hot-wire laser welding using a half-vehicle suspension system model integrated with an arm-
ensemble of metamodels and NSGA-II. Robot Comput Integr teeth regenerative shock absorber using Taguchi method. Mech
Manuf 53:141–152 Sys Signal Process 126:65–81
182. Yifei T, Meng Z, Jingwei L, Dongbo L, Yulin W (2018) Research 191. Zhang Y, Xu Y, Zheng Y, Fernandez-Rodriguez E, Sun A, Yang
on intelligent welding robot path optimization based on GA and C, Wang J (2019) Multiobjective optimization design and experi-
PSO algorithms. IEEE Access 6:65397–65404 mental investigation on the axial flow pump with orthogonal test
183. Yoon KP, Kim WK (2017) The behavioral TOPSIS. Expert Syst approach. Complexity 2019:1–14
Appl 89:266–272 192. Zhang J, Wang J, Lin J, Guo Q, Chen K, Ma L (2015) Multi-
184. Zadeh LA (1963) Optimality and nonscalar-valued performance objective optimization of injection molding process parameters
criteria. IEEE Trans Autom Control 8(1):59–60 based on Opt LHD, EBFNN, and MOPSO. Int J Adv Manuf
185. Zhou GD, Yi T-H, Xie M-X, Li H-N, Xu J-H (2021) Optimal Technol 85(9–12):2857–2872
wireless sensor placement in structural health monitoring empha- 193. Zitzler E, Deb K, Thiele L (2000) Comparison of multi-objec-
sizing information effectiveness and network performance. J tive evolutionary algorithms: empirical results. Evol Comput
Aerosp Eng 34(2):04020112 8(2):173–195
186. Zhang H, Peng Y, Hou L, Tian G, Li Z (2019) A hybrid multi- 194. Zitzler E, Laumanns M, Thiele L (2001) SPEA2: improving the
objective optimization approach for energy-absorbing structures strength pareto evolutionary algorithm. In: Giannakoglou K, Tsa-
in train collisions. Inf Sci 481:491–506 halis D, Periaux J, Papailou P, Fogarty T (eds) EUROGEN 2001.
187. Zhang J, Zhu H, Yang C, Li Y, Wei H (2011) Multi-objective Evolutionary methods for design, optimization and control with
shape optimization of helico-axial multiphase pump impel- applications to industrial problems, pp 95–100, Athens, Greece
ler based on NSGA-II and ANN. Energy Convers Manage 195. Zitzler E, Thiele L (1999) Multi-objective evolutionary algo-
52(1):538–546 rithms: a comparative case study and the strength pareto
188. Zhang L, Zhang S, Zhang W (2019) Multi-objective optimization approach. IEEE Trans Evol Comput 3(4):257–271
design of in-wheel motors drive electric vehicle suspensions for
improving handling stability. Proc Inst Mech Eng Part D J Autom Publisher's Note Springer Nature remains neutral with regard to
Eng 233:2232–2245 jurisdictional claims in published maps and institutional affiliations.
189. Zhang Q, Li H (2007) MOEA/D: a multi-objective evolutionary
algorithm based on decomposition. IEEE Trans Evol Comput
11:712–731

13