Download the Full Ebook and Access More Features - ebookmass.

com

Multi-Objective Passing Vehicle Search algorithm

for structure optimization Sumit Kumar

https://ebookmass.com/product/multi-objective-passing-
vehicle-search-algorithm-for-structure-optimization-sumit-
kumar/

OR CLICK HERE

DOWLOAD NOW

Download more ebook instantly today at https://ebookmass.com

 Instant digital products (PDF, ePub, MOBI) ready for you
Download now and discover formats that fit your needs...

Objective Physics for the JEE Main 2015 Sanjeev Kumar

https://ebookmass.com/product/objective-physics-for-the-jee-
main-2015-sanjeev-kumar/

ebookmass.com

Objective Physics for NEET and Other Medical Entrance

Examinations Abhay Kumar

https://ebookmass.com/product/objective-physics-for-neet-and-other-
medical-entrance-examinations-abhay-kumar/

ebookmass.com

Meeting EURO6 emission regulations by multi-objective

optimization of the injection strategy of two direct
injectors in a DDFS engine Sasan Shirvani
https://ebookmass.com/product/meeting-euro6-emission-regulations-by-
multi-objective-optimization-of-the-injection-strategy-of-two-direct-
injectors-in-a-ddfs-engine-sasan-shirvani/
ebookmass.com

. Four-Dimensional Printing for Clinical Dentistry

Rupinder Singh & Ravinder Sharma & Nishant Ranjan

https://ebookmass.com/product/four-dimensional-printing-for-clinical-
dentistry-rupinder-singh-ravinder-sharma-nishant-ranjan/

ebookmass.com
(eBook PDF) Mosby’s Dictionary of Medicine, Nursing &
Health Professions 10th Edition

https://ebookmass.com/product/ebook-pdf-mosbys-dictionary-of-medicine-
nursing-health-professions-10th-edition/

ebookmass.com

MKTG5 5th Edition Joe F. Hair

https://ebookmass.com/product/mktg5-5th-edition-joe-f-hair/

ebookmass.com

The Archaeology of Loss: Life, love and the art of dying

Sarah Tarlow

https://ebookmass.com/product/the-archaeology-of-loss-life-love-and-
the-art-of-dying-sarah-tarlow/

ebookmass.com

Human Security In China: A Post-Pandemic State 1st Edition

Chi Zhang

https://ebookmass.com/product/human-security-in-china-a-post-pandemic-
state-1st-edition-chi-zhang/

ebookmass.com

5 Steps to a 5: AP Calculus BC 2024 William Ma

https://ebookmass.com/product/5-steps-to-a-5-ap-calculus-
bc-2024-william-ma/

ebookmass.com
Facharztprüfung Innere Medizin 5th Edition Stefan Endres

https://ebookmass.com/product/facharztprufung-innere-medizin-5th-
edition-stefan-endres/

ebookmass.com
 Expert Systems With Applications 169 (2021) 114511

Contents lists available at ScienceDirect

Expert Systems With Applications

journal homepage: www.elsevier.com/locate/eswa

Multi-Objective Passing Vehicle Search algorithm for

structure optimization
Sumit Kumar a, 1, Ghanshyam G. Tejani b, *, 2, Nantiwat Pholdee c, Sujin Bureerat c
a
Department of Mechanical Engineering, GPERI, Gujarat Technological University, Gujarat, India
b
Department of Mechanical Engineering, School of Technology, GSFC University, Vadodara, Gujarat, India
c
Sustainable and Infrastructure Research and Development Center, Department of Mechanical Engineering, Faculty of Engineering, Khon Kaen University, 40002,
Thailand

A R T I C L E I N F O A B S T R A C T

Keywords: A novel Multi-Objective Passing Vehicle Search (MOPVS) algorithm is proposed for structural design optimi­
Metaheuristic zation. MOPVS is inspired by the two-lane highway passing vehicle mechanism. This multi-objective version is
Multi-objective optimization modified and further improved from the single-objective version of passing vehicle search through a Pareto
Structural design
dominance-based approach. For performance evaluation of MOPVS, five daunting benchmark structural design
Discrete variables
Constraint optimization
problems have been used. Two conflicting objectives i.e. structure weight minimization and minimization of
maximum nodal displacement along with discrete design variables have been considered to ensure its real-world
applications. For fitness and efficiency evaluation of the proposed algorithm, the results obtained from the new
algorithm are compared with four other state-of-the-art multi-objective algorithms. Moreover, two performance
indicators test called Hypervolume and Spacing-to-Extent were performed for the rigorous evaluation of the
performance and feasibility of the proposed algorithm. The findings demonstrate the superiority of the MOPVS
algorithm over the others while the potential to find a non-dominated solution set with diverse individual so­
lutions. Present work considers the Friedman’s rank test for the statistical investigation of the experiment work.
The solutions and convergence behavior achieved by MOPVS show its high efficiency in solving challenging
design problems.

1. Introduction at the same time, maximization of strength or performance which, by

itself, is a very challenging design issue (Kumar, Tejani, & Mirjalili,
Real-world engineering design problems are typically complicated 2019). From the last three decades, there have been lots of methodol­
and multidimensional, thus, finding its solution is a very challenging ogies like deterministic optimization techniques developed for solving
job. All such sectors are mainly linked to many goals which are called such kind of design issues; however, they are struggling to find optimal
Multi-objective (MO). More often, those objectives are conflicting and solutions due to their limitations like premature convergence, complex
demand a trade-off between them (Blondet, Duigou, & Boudaoud, 2019; programming approach, inefficiency in solving non-linear, and the is­
Wu et al., 2020; Chinchuluun, Pardalos, Migdalas, & Pitsoulis, 2008). sues of non-convex functions and regions (Pardalos, Žilinskas &
This leads to a set of two or more solutions rather than a single optimal Žilinskas, 2017). Often, an inaccurate derivative calculation can lead the
solution as with a problem with one objective function (Anita, Yadav, & optimizers to wrong solutions and may cause linear and quadratic pro­
Kumar, 2020). Optimization is an integral part of any system design as it gramming sub problems being ill-conditioned.
works to maximize/minimize the system outcomes by making the least The above demerits make most scholars turn towards more robust
use of resources possible. In the case of a structure optimization, these and gradient-free algorithms for solving challenging design problems,
conflicting objectives are generally minimization of cost or weight and, particularly Metaheuristics (MHs). The MHs are the algorithms which

* Corresponding author.
E-mail addresses: [email protected] (S. Kumar), [email protected] (G.G. Tejani), [email protected] (N. Pholdee), [email protected]
(S. Bureerat).
1
ORCID ID: 0000-0003-3042-3779.
2
ORCID ID: 0000-0001-9106-0313.

https://doi.org/10.1016/j.eswa.2020.114511
Received 7 June 2020; Received in revised form 5 October 2020; Accepted 14 December 2020
0957-4174/© 2020 Elsevier Ltd. All rights reserved.
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

are inspired by natural events in our surroundings and have advantages was modified, improved, and hybridized with other algorithms by a few
like simplicity, almost parameter-free ability, and flexibility. Owing to scholars to check its performance. Savsani, Tejani, Patel, and Savsani
those advantages, the algorithms have been implemented to a wide (2017) modified PVS based on random mutation and applied for to­
range of applications in engineering and industrial sectors for solving pology optimization of structures with static and dynamic constraints.
real-word optimization problems (Tejani, Kumar, & Gandomi, 2019). The results demonstrate the superiority of the modified PVS algorithm
Moreover, from the last two decades, MHs have seen a rise in their ap­ over other compared MHs. Tejani, Pholdee, Bureerat, Prayogo, and
plications in structural optimization problems for investigation of its Gandomi (2019) and Tejani, Savsani, Bureerat, Patel, and Savsani
potential to solve many objectives with conflicting nature. As opposed to (2019) modified the PVS based on a parallel run mechanism with the
the single-objective design problem which has only one solution, the MO incorporation of simulated annealing methodology which is used for
design issue has a set of optimal solutions traditionally called the Pareto- prevention from local optima trap and enhancement of population di­
optimal set (Ho-Huu et al., 2018; Kumar, Tejani, Pholdee, & Bureerat, versity. In a similar study, to improve the exploration and exploitation
2020). Generally, there are two steps in dealing with a design problem potential of the basic PVS algorithm, Tejani, Pholdee, et al. (2018),
with more than one objective function. The first one is to explore all Tejani, Savsani, Patel, and Mirjalili (2018) and Tejani, Savsani, Patel,
possible Pareto optimal solutions by using a powerful optimizer. In the and Savsani (2018) incorporated a mutation-based search technique for
second task, the selection of one solution from the obtained Pareto the optimization of the truss under dynamic bounds. The results were
optimal set is performed usually employing a decision-making process. compared with other popular MHs and proved the dominance of the
Therefore, the best solution must be from the Pareto optimal set, and at improved PVS version over the others. Literature also witnessed few
the same time, it should be the chosen one or more from the designer hybridizations like enhanced PVS algorithm for wind farm layout opti­
(Kaveh & Laknejadi, 2011). mization (Patel, Savsani, Patel, & Patel, 2019), and fuzzy-based sym­
Meanwhile, as metaheuristics (initially named evolutionary algo­ biotic organism search with PVS for optimization of automatic
rithms) have been established for single-objective optimization, their generation control (Nayak, Shaw, & Sahu, 2020) which proves the better
MO versions can be created exploiting the nature of MH search, which is performance characteristics of PVS in challenging design issues.
based on a population or set of design solutions and randomization A very few scholars extended PVS to MO design issues. Parsana et al.
(Pardalos & Romeijn, 2013). The so-called non-dominated sorting (2018) in their study applied the PVS algorithm for the parametric MO
operator is a key to upgrade existing single-objective MHs to become MO optimization of Electric Discharge Machine (EDM) on Mg–RE Alloy. The
ones. At the very beginning, they were termed multi-objective evolu­ obtained Pareto front was better and the performance metric results
tionary algorithms (MOEAs) and later multi-objective metaheuristics were found to be acceptable. In a similar parametric study of EDM, Ram
(MOMHs) to cover all possible philosophical aspects of MH search. Some Prabhu et al. (2018) investigated the performance of PVS with MO in
of the first generation MOMHs are MO genetic algorithms (MOGA) different mathematical models and also compared it with other intelli­
(Fonseca & Fleming, 1993), non-dominated sorting genetic algorithms gent MHs. The findings show better results. However, as per the
(NSGAII) (Deb, Pratap, Agarwal, & Meyarivan, 2002), and strength knowledge of the authors, the PVS has limited investigation so far
Pareto evolutionary algorithms (SPEA2) (Zitzler, Laumanns, & Thiele, especially in the case of MO structural design issues. Moreover, PVS is a
2001). Then, there have been numerous MOMHs invented in parallel to novel population-based algorithm which was just suggested, thus, there
their single-objective counterparts. The concepts for upgrading single- is room for even more research and need to explore its potential in
objective MHs to be MO versions include, for example, a non- different analyses. Furthermore, as per the no free lunch (NFL) theorem
dominance approach (Tejani, Pholdee, Bureerat, & Prayogo, 2018; (Wolpert & Macready, 1997), one MH cannot be efficient for all design
Tejani, Savsani, Patel, & Mirjalili, 2018; Tejani, Savsani, Patel, & Sav­ problems. Thus, there is always a need for a novel efficient algorithm for
sani, 2018), a decomposition-based approach (Zhang & Li, 2007), and more challenging design problems posed every year. PVS, with its suc­
an elitism strategy (Deb, Agrawal, Pratap, & Meyarivan, 2000), a grid- cessful implementation to many engineering applications, is a new
based approach (Knowles & Corne, 1999), a preference-based metaheuristic that deserves this attention.
approach (Bureerat & Srisomporn, 2010), a guided population-based Therefore, this study suggests a Multi-Objective Passing Vehicle
approach (Got, Moussaoui, & Zouache, 2020), and an opposition- Search (MOPVS) algorithm for structure optimization. The two objec­
based learning approach (Gupta, Deep, Heidari, Moayedi, & Wang, tives are mass minimization and minimization of maximum nodal
2020). The most outstanding feature of MOMHs is that they can explore deflection which are diverse in nature and simultaneously optimized.
a Pareto front within one optimization run. Nevertheless, to achieve Five challenging constrained test examples i.e. the 10-, 25-, 60-, 72-, and
such a feat, the reproduction must be very powerful for both exploitation 942-bar trusses have been examined in this investigation for the per­
and exploration. It is even more difficult when the number of objective formance measurement of the proposed MOPVS algorithm. Also, the
functions is higher than 3, which is usually called many-objective findings of MOPVS are compared with other noteworthy algorithms for
optimization. As a result, the seek for new better MOMHs is still a efficiency verification.
challenging task. The rest of the article is convened as follows: Section 2 provides the
One of the newly developed state-of-the-art algorithms is Passing details about the fundamental PVS algorithm and the proposed MOPVS
Vehicle Search (PVS) which was proposed by Savsani and Savsani algorithm description has been shared in Section 3; Section 4 presents
(2016). This algorithm is typically a human action-based system in the mathematical formulation of the MO design problem; the findings
which the behavior of passing vehicles on the two-lane highway has and detail discussion of all the test examples are presented in Section 5;
been mathematically modelled. This is one of the population-based Section 6 puts forward the concluding remarks with prospects of the
global optimization algorithms which is designed for finding global proposed algorithm.
optima. It is motivated by the overtaking mechanism of vehicles and has
advantages like controlling parameter-free and fast convergence in 2. The Passing Vehicle Search algorithm
comparison to other MHs. This novel algorithm is investigated by a few
researchers in numerous design problems for its performance evalua­ PVS is a novel single-objective population-based MH suggested by
tion. Sheth et al. (2019) investigated PVS for parametric optimization of Savsani and Savsani (2016) as a global optimization algorithm that is
the Wire Electric discharge machine on Mg-Zn-RE-Zr alloy. Results were inspired by a certain natural occurrence. Like other population-based
compared with a standard GA code while outstanding performance was algorithms, PVS also initializes its search mechanism with a random
noticed which can be explored in the aerospace industry. To control the set of solutions traditionally called a population, and updates the current
network flow, PVS was used and the results show the least time delay of population as per the other searched solutions and finally end with the
the vehicles at the traffic signals. Furthermore, the basic version of PVS optimized set of solutions. PVS mimics the two-lane highways

2
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

overtaking or passing behavior of vehicles which is simplified into a In this PVS algorithm, three different vehicles i.e. Front vehicle (FV),
mathematical model. It is a simple, efficient, and parameter-free algo­ Back vehicle (BV), and Oncoming vehicles (OV) are considered which
rithm and modifies the population-based on three simple mathematical are engaged in the mechanism of overtaking of the vehicle on the
conditions of passing vehicles with numerous probabilities. Having a highway. The process of overtaking will be executed only when the BV
safe passing opportunity is the most significant criterion for the over­ velocity will become greater than the speed of FV and if not then the
taking mechanism of two-lane. This overtaking mechanism depends on overtaking process of vehicles will not be possible. This feature clearly
various criteria like the gap between the vehicles, individual vehicle illustrates the dependency of the overtaking process on the speed and
velocity, road, driver skills, traffic behaviors, and weather which are the position of OV and the distance between each vehicle and their
highly interrelated and have complex characteristics. corresponding velocities. This generates different conditions of the
Three sets of conditions will arise when a slow vehicle is approached overtaking process on the two-lane highway as follows:
by a vehicle: Assume,

(a) the fast vehicle will overtake the slower vehicle; s– Distance between Back vehicle and Front vehicle
(b) until a suitable opportunity arises, the vehicle will follow the k – Distance between Front vehicle and Oncoming vehicles
slow vehicle; S1 , S2 , S3 – Distance from the reference line
(c) without the objective of overtaking, the vehicle continues to v1 , v2 , v3 – Velocities of Back vehicle, Oncoming vehicles, and Front
follow the slower vehicle. vehicle, respectively,

If the vehicle is not able to overtake other slower vehicles then pla­ At any instant, the velocity of the BV, OV, and FV vehicles are v1 , v2 ,
toon rise to form which affects the desired speed of the vehicle again. and v3 respectively on the two-lane highway. Thus, based on the ve­
This makes mathematical modeling of the overtaking mechanism of locity of FV, two specific conditions will arise i.e. FV is slower than BV
two-lane traffic more challenging. However, Savsani and Savsani (2016) and vice versa. At this particular condition, the overtaking phenomena
suggested a simplified model for easy understanding as illustrated in possibly mean BV can pass the FV. Passing is only feasible in that state
Fig. 1. when the distance from FV at which passing takes place is shorter than

Fig. 1. Two-lane highway vehicle overtaking mechanism.

3
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

the distance covered by OV. However, if the BV speed is slower than FV, prevented if BV does not switch lanes before OV passes BV. The
BV cannot overtake the FV and it only moves with its speed. Thus, based encounter between BV and OV will happen at distance somewhere
on the three selected vehicles, the following conditions will arise: located in between BV and OV initial positions.
Case1. When FV is slower than BV (v3 < v1 ) (termed as Primary The BV has the position changed as
condition-1)
BV c2 = rand(s + k) (12)
(a) (k − k1 ) > s1 (Secondary condition-1) In the above equation, rand is a uniform random number, rand ∈ [0,
(b) (k − k1 ) < s1 (Secondary condition-2) 1].
The change of BV location from the line of reference is
Case 2. FV is faster than BV (v3 > v1 ) (termed as Primary condition-2)
S1 + BV c2 = S1 + rand(s + k) = S1 + rand(S2 − S1 ) (13)

2.1. Primary condition-1

2.2. Primary condition-2

This is the case when FV has a slower velocity than BV which further
As per Fig. 1, BV cannot overtake FV when FV is faster than BV. Thus,
leads to two sub-cases i.e. secondary conditions (1) and (2). The details
the BV position change can be written as:
are in the following mathematical formulation form:
S1 + BV c3 = S1 + rands = S1 + rand(S3 − S1 ) (14)
2.1.1. Secondary condition-1
In the PVS algorithm, two-lane highway different vehicles are
It is assumed that after traveling distance s1 by FV the BV can catch
assumed to correspond to different solutions set. Vehicle velocities are
the FV and simultaneously pass it and the corresponding time is ‘t’ to
corresponding to objective function or fitness values where the vehicle
execute this passing mechanism which is depicted in Fig. 1.
with the highest velocity is assumed as the best fitness value. The ve­
Hence, in time interval ‘t’ the distance travelled by FV is
hicle’s position on a motorway is analogous to design variables. Thus,
s1 = v3 t (1) PVS initializes its search with the vehicle population (set of solutions). In
the reproduction phase, three vehicles (solutions) are randomly
Similarly, in time ‘t’ the distance covered by BV will be
selected. BV corresponds to the current solution out of the three selected
s + s1 = v 1 t (2) vehicles while FV and OV are other two different solutions. Depending
on the size of the population and its fitness values, the relative distances
On substitution of Eq. (1) in Eq. (2), the following result will generate
between the vehicles and their corresponding velocities are allocated.
s1 =
v3 s
(3) Then the overtaking conditions are inspected after velocity and distance
v1 − v3 allocation. Subsequently, vehicles alter their positions on the highway
based on the condition applied. The detailed explanation of the PVS
and algorithm is shown in Fig. 2.
s
t= (4)
v1 − v3 3. The Multi-Objective Passing Vehicle Search (MOPVS)
At the same time, ‘t’ OV covered the distance of algorithm

k1 = v2 t (5) MO engineering design problems simulate the more practical con­

ditions and that is why they are more widely applied and investigated.
After replacing the value of ‘t’ from Eq. (4) in Eq. (5), one can have
However, because of the challenges like diverse objectives, non-
y1 =
v2 x
(6) linearity, stringent bounds, and multi-modal search domains, it be­
v1 − v3 comes difficult for a designer to find the best solutions (Tejani, Kumar,
The corresponding change in positions of BV is et al., 2019). MHs emerged as arguably the best alternative for solving
these intricate MO design issues. MHs are the remarkably efficient and
BV c1 = s + s1 (7) robust algorithms that mimic the basic physical and biological phe­
After putting the value of s1 from Equation (3), nomena and have widely been applied for solving several industrial,
( ) science, and engineering design problems.
BV c1 = s
v1
(8) MHs pose several advantages over traditional algorithms i.e.
v1 − v3 gradient-free, can handle any kind of variables, free from problem-
The value of s that relates the reference line is specific knowledge, least setup time, and objective convexity require­
ment. However, they face problems like premature convergence, lack of
s = S3 − S1 (9) population diversity, and local optima trap which needs to be addressed
Thus, properly (Kumar et al., 2019). For large-scale problems, their perfor­
( ) mance does not scale well and usually crumbled in the aspect of both
BV c1 = (S3 − S1 )
v1
(10) computing time and quality of the solution. Moreover, most of the MHs
v1 − v3 required controlling parameters that require proper tuning, which
The change in the position of BV from the reference line is makes them inefficient for global optimization design issues.
( ) Another aspect for MHs which plays a crucial role in governing their
v1 efficiency in finding the optimal solution is the proper balance between
S1 + BV c1 = S1 + (S3 − S1 ) (11)
v1 − v3 two key components i.e. local intensification (exploitation) and global
diversification (exploration) (Yang, Deb, & Fong, 2014). Diversification
2.1.2. Secondary condition-2 implies the production of heterogeneous and diverse solutions to explore
This case pertains to two separate possibilities, one being positive the global search space whereas intensification denotes the focusing on
and the other being negative which is represented in Fig. 1. In case of the searching in an immediate territory via exploiting the current found
existence of any one of these possibilities, the passing of FV by BV cannot good solution knowledge. The interaction between these two key com­
occur before OV passes BV. A motor vehicle accident can, therefore, be ponents will help in finding the best feasible solution out of all possible

4
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

Fig. 2. Flowchart of fundamental PVS algorithm.

solutions within acceptable computational time. However, there is no issues are challenging, thus a high-performance algorithm is always in
predefined set of conditions to balance these two performance defining demand. Also, the NFL theorem (Wolpert & Macready, 1997) states that
parameters. Thus, it is required to find an algorithm that can find a good an MH cannot solve all design issues competently. That means one al­
quality solution most of the time and simultaneously efficient and gorithm may be seen effective in solving one particular design problem,
practical (Blum & Roli, 2003). In general, for efficient MHs two poten­ however, it may fail in another kind of optimization issue. Thus, there is
tials are a must. The first one is the potential of new solution generation a need for a successful MH which can solve MO design issues and has the
which has a greater chance to replace the existing or previous solution potential to find the global or near optimum solution with high
along with the ability to search all-important arena where the global accuracy.
optimum solution possibility lies. The second potential is to escape from Hence to resolve all aforementioned problems, in the proposed study
any local optima solution so that the MH can avoid the local mode trap. authors applied a highly efficient population-based novel PVS algorithm
However, the previous MHs studies and analyses of their conver­ for solving MO structural optimization problems named MOPVS. The
gence behavior, advocate that mere diversification leads to convergence dominance approach is incorporated into PVS to create a new MO
rate reduction whereas only intensification results in the enhancement version of PVS. Given that design solutions X1 and X2 result in function
of convergence speed (Kumar et al., 2019). Notwithstanding, excessive vectors f1 and f2 respectively, X1 is said to dominate X2 (for minimiza­
diversification enhances the likelihood of the global optimal solution, tion) if i) all elements in f1 are less than or equal to their corresponding
however, it reduces the efficiency of an algorithm. Whereas too much elements in f2 and ii) at least one element of f1 is strictly less than its
intensification has an inclination toward the local optima trap of the corresponding element in f2. With the definition of dominance, having
algorithm. Thus, it is required for powerful MHs to have a good harmony had a population of solutions, those who are not dominated by any so­
between the fine level of local intensification and well magnitude of lution in the set are said non-dominated solutions. Thus, MOMHs work
global diversification (Tejani, Kumar, et al., 2019; Yang et al., 2014). by iteratively reproducing a population, classifying non-dominated so­
In MO problems, there are a set of optimal solutions called Pareto- lutions, and saving them to the so-called Pareto archive. The set of non-
optimal set instead of a single solution as in single objective design dominated solutions at the final iteration is regarded as an approximate
problems. In the absence of any prior knowledge, it is difficult for the Pareto optimal set. MOPVS works on the overtaking mechanism of the
designer to find which one is the best solution out of all. As MO design vehicles on the freeway which helps in updating the population-based

5
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

on vehicle velocity and the relative distance between the vehicles. It is a (Kanyakam & Bureerat, 2007). The normal line method is one of the
highly efficient algorithm that poses a good convergence rate which archiving techniques used to screen out some non-dominated solutions
helps in finding globally optimal solutions with less computational time. from the archive in order to save computer memory. This is because it is
At the initial stage, an initial population with function evaluations is usual that MOMHs can explore an excessively large number of non-
created at random while the non-dominated solutions are sorted and dominated solutions often leading to insufficient memory for the used
saved to the initial Pareto archive. The PVS reproduction operator is computer. The concept is to screen out some solutions whilst main­
then activated to produce a new set of offspring. The offspring and the taining as high as possible diversity of the remaining solutions. The
members in the Pareto archive are combined to sort for new non- detailed process of the MOPVS algorithm is illustrated in Fig. 3.
dominated solutions. The population and the archive are iteratively In this study, MOPVS is applied for solving five challenging bench­
updated until reaching a termination criterion. In cases that the archive mark structure design examples i.e. the 10-bar, 25-bar, 60-bar, 72-bar,
size is larger than a predefined size, some non-dominated solutions are and 942-bar trusses with discrete sections as design variables, to eval­
removed from the Pareto archive by using the normal line method uate the performance of the algorithm. MOPVS algorithm creates a

Fig. 3. Flow chart of the MOPVS algorithm.

6
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

balance between intensification and diversification of the search and

simultaneously results in non-dominated solutions keeping the diversity Cross − sectional area constraints, Zimin ≤ Zi ≤ Zimax
in the Pareto fronts. For comparison purposes, the optimal Pareto front
obtained by the proposed MOPVS algorithm is compared with the other where, i = 1, 2, ⋯, m; j = 1, 2, ⋯, n
five distinct MHs i.e. MO ant system (MOAS), MO ant colony system Here,Zi is a design variable vector; ρi and Li are the mass density and
(MOACS), MO symbiotic organism search (MOSOS), and MO heat length of the elements respectively;Ei and σi corresponds to the ‘i’
transfer search (MOHTS). Moreover, two performance indicators called element Modulus of elasticity and stress respectively. Moreover, the
Hypervolume and Spacing-to-Extent were performed for the exact allowable upper and lower bounds are represented by superscripts
evaluation of the performance and feasibility of the proposed algorithm. ‘max’ and ‘min’ respectively.
The present work also considers the Friedman’s rank test for the sta­
tistical investigation of the experiment work. Finally, the comparison is
made based on the ranking and the nature of the best Pareto front ob­
tained by all considered algorithms for all test examples. 4.1. The dynamic penalty function

4. The Multi-objective optimization problem formulation For processing the constrained problems like Eq. (7), one of the best
ways is to transform these design issues into dynamic unconstrained
Realistic industry and engineering design issues are relevant to optimization problems using a dynamic penalty approach. Considering
modeling as MO problems having conflicting objectives. These MO that in the design arena all the objective function values being positive,
challenging issues have characteristics like non-linearity, large search the multiplication-based penalty function (Tejani, Savsani, & Patel,
domain, and sometimes multi-modal functions. Moreover, instead of a 2016) is among the most effective penalty functions.
single best solution as in the case of single objectives, MO has an optimal ∑q ⃒
⃒
⃒
p⃒
solution set which makes the decision process more intricate for a f (X)*(1 + ε1 *∁)ε2 , ∁ = ∁i , ∁i = ⃒⃒1 − *i ⃒⃒ (16)
pi
designer. Out of all solutions in the set (also known as a Pareto optimal i=1

set), a designer has to choose one best solution based on predefined

where pi is the dimension of requirement infringement having the bound
rules. For any specific design issue, these optimal Pareto solutions are
as p*i . The parameter q is the number of constraints. The factors ε1 and ε2
usually unexplored. Therefore, it is essential to search for more optimal
are pre-specified by a user. In this investigation, the estimations of both
solutions rather than being limited to a single solution.
The mathematical formulation for the MO structure optimization
ε1 and ε2 are set at 3, which were obtained from testing their impact on
the parity of the investigation balance (Tejani, Pholdee, et al., 2018,
problem is as follows:
2019; Tejani, Savsani, Patel, & Mirjalili, 2018; Tejani, Savsani, Patel,
Find, Z = {Z1 , Z2 , .., Zm } (15) and Savsani, 2018; Tejani, Kumar, et al., 2019; Tejani, Savsani, et al.,
2019; Kumar et al., 2020).
to minimize, the structure mass and maximum nodal deflection
5. Test examples, results, and discussion
∑
m
(⃒ ⃒)
f1 (Z) = Zi ρi Li and f2 (Z) = max ⃒δj ⃒
i=1 The performance evaluation of the considered algorithms is made by
the application in five challenging structure examples which are being
Subjected to : investigated by many scholars (Angelo, Bernardino, & Barbosa, 2015;
Tejani, Pholdee, et al., 2018, 2019; Tejani, Savsani, Patel, and Mirjalili,
2018; Tejani, Savsani, Patel, and Savsani, 2018; Tejani, Kumar, et al.,
Behavior constraints:
2019; Tejani, Savsani, et al., 2019; Kumar et al., 2020). Also, some
g(Z) : Stress constraints, |σ i | − σmax
i ≤0 multi-objective standard convex test functions are used to examine the
search performance of the considered algorithm. The population size is
taken as 100 and functional evaluations are considered 50,000 in this
Side constraints:
investigation according to previous studies. Each algorithm tested with

Fig. 4. The 10-bar truss.

7
 S. Kumar et al.
Table 1
Design considerations of the truss problems.
The 10-bar truss The 25-bar truss The 60-bar truss The 72-bar truss The 942-bar truss

Design Ai, i = 1, 2, ⋯, 10 Ai, i = 1, 2, ⋯, 8 Ai, i = 1, 2, ⋯, 25 Ai, i = 1, 2, ⋯, 16 Ai, i = 1, 2, ⋯, 59

variables
Constraints σmax = 25 ksi σmax = 40 ksi σmax = 40 ksi σmax = 25 ksi σmax = 25 ksi
Density ρ = 0.1 lb/in3 ρ = 0.1 lb/in3 ρ = 0.1 lb/in3 ρ = 0.1 lb/in3 ρ = 0.1 lb/in3
Young modules E = 104 ksi E = 104 ksi E = 104 ksi E = 104 ksi E = 104 ksi
Size variables Ai, ∈ S, i = 1, 2, ⋯, 10 Ai, ∈ S, i = 1, 2, ⋯, 8 Ai, ∈ S, i = 1, 2, ⋯, 25 Ai, ∈ S, i = 1, 2, ⋯, 16 Ai, ∈ S, i = 1, 2, ⋯, 59
S = [1.62,1.8,1.99,2.13,2.38, S = [1,0.2,0.3,0.4,0.5,0.6,0.7,0.8, S = [0.5, 0.6, 0.7, …, 4.9] in2 S = [0.1, 0.2, 0.3,…, 2.5] in2 S = [1, 2, 3,…, 200] in2
2.62,2.63,2.88,2.93,3.09, 3.13, 0.9,1,1.1,1.2,1.3,1.4,1.5,1.6,1.7,1.8,
3.38,3.47,3.55,3.63,3.84,3.87, 1.9,2,2.1,2.2,2.3,2.4,2.5,2.6,2.8,3,3.2, 3.4]
3.88,4.18,4.22,4.49,4.59,4.8,4.97, in2
5.12, 5.74, 7.22,7.97, 11.5,13.5,13.9,14.2,
15.5,16,16.9,18.8,19.9,22,22.9,26.5,30, 33.5]
in2
Loading Py2 = Py4 = − 100 Klb Px1 = 1 Klb, Case 1: Px1 = − 10 Klb, Px7 = 9 Klb Case 1: F1x = F1y = 5kips, F1z = At each node: vertical
conditions − 5kips loading:
Py1 = Pz1 = Py2 = Pz2 = − 10 Klb, Case: : Px15 = Px18 = − 8 Klb, Py15 = Py18 = Case 2:F1z = F2z = F3z = F4z = Section 1;
8

3 Klb − 5 kips Pz = − 3kips

Section 2;
Px3 = 0.5 Klb, Px6 = 0.6 Klb Case 3:Px22 = − 20 Klb and Py22 = 10 Klb
Pz = − 6kips
Section 3;
Pz = − 9kips
Lateral loading:
Right-hand side;
Px = 1.5kips
Left-hand side;
Px = 1.0kips
Lateral Loading:
Py = 1.0kips

Expert Systems With Applications 169 (2021) 114511

S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

Table 2
The hypervolume values of results obtained for the 10-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 48495.37 54326.74 52094.31 1304.90 100 5

MOACS 53839.12 55653.40 54864.00 321.26 200 4
MOSOS 56538.52 57796.10 57352.53 241.81 300 3
MOHTS 58208.51 58612.36 58491.60 81.86 400 2
MOPVS 58602.79 58732.49 58684.42 27.57 500 1

Table 3
The front Spacing-to-Extent values of results obtained for the 10-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 0.0054 0.0247 0.0106 0.0038 356 4

MOACS 0.0072 0.0296 0.0142 0.0046 456 5
MOSOS 0.0080 0.0142 0.0103 0.0010 381 3
MOHTS 0.0049 0.0066 0.0058 0.0004 153 1
MOPVS 0.0057 0.0063 0.0059 0.0001 154 2

100 independent runs to find the solutions. The Pareto front Hyper­ MOACS, MOSOS, MOHTS, and MOPVS algorithms are 52094.31,
volume (PFHV) test is conducted for all algorithms where the mean 54864.00, 57352.53, 58491.60, and 58684.42 respectively. Similarly,
value is treated as equivalent to the rate of convergence and the standard the STD of the PFHV values for the MOAS, MOACS, MOSOS, MOHTS, and
deviation (STD) value is taken as the measure of reliability. MOPVS algorithms are 1304.90, 321.26, 241.81, 81.86, and 27.57
A blend of two testing indicators, i.e. a front spacing (S) scale and respectively. These results show that the proposed MOPVS algorithm has
Extent called the Front Spacing-to-Extent (FSTE) (Tejani, Pholdee, et al., better values of mean and STD than others in terms of convergence and
2018, 2019; Tejani, Savsani, Patel, and Mirjalili, 2018; Tejani, Savsani, search consistency. The Friedman’s rank test is performed for statistical
Patel, and Savsani, 2018; Tejani, Kumar, et al., 2019; Tejani, Savsani, analysis for each algorithm based on their PFHV values. The Friedman
et al., 2019; Kumar et al., 2020) is used to measure algorithm perfor­ test values obtained by MOAS, MOACS, MOSOS, MOHTS, and MOPVS
mance. The lower value of FSTE specifies the superior result. Friedman’s are 100, 200, 300, 400, and 500 respectively. At a 95% significance level
rank test is considered to rank the algorithms statistically. of Friedman’s rank test, MOPVS is found to be the best algorithm that
justifies the Hypervolume results. MOHTS and MOSOS followed the
5.1. 10-bar test example MPOVS and ranked second and third respectively among all.
Table 3 demonstrates the FSTE metric values which signify the
The first test example taken into account for investigation in this relative spacing between the solutions in the non-dominated set for the
study is a 10-bar structure which was widely applied in numerous re­ 10-bar example. The corresponding mean values for MOPVS and
searches. The details about this benchmark are illustrated in Fig. 4 which MOHTS are 0.0059 and 0.0058. The best values of STD are obtained by
also represents the loading conditions, constraints, nodes, and structure MOPVS i.e. 0.0001 followed by that of MOHTS and MOSOS, which are
dimensions. The design consideration of all the benchmarks is depicted 0.0004 and 0.0010 respectively. As per the Friedman’s test ranking, it is
in Table 1. evident that MOHTS and MOPVS are the best performers.
The statistical results obtained for the 10-bar test example are rep­ The obtained median Pareto fronts for the 10-bar example from all
resented in Table 2 in the PFHV value form. Here the mean and STD the considered algorithms are illustrated in Fig. 5 which is obtained from
optimal values are considered as the measure of performance for all the 100 independent runs of each optimizer. It is understood from the figure
considered algorithms. The optimal mean PFHV values of MOAS, that the solutions for MOAS and MOACS are scattered and non-
continuous. However, MOPVS, MOHTS, and MOSOS Pareto fronts are
smooth, well-distributed, and steady. Also, these algorithms have an
array of heterogeneous solutions. Considering all aspects, we can say
that MOPVS is the superior algorithm in comparison to other considered
algorithms in solving the MO 10-bar structure design problem.

5.2. 25-bar test example

The second test example considered in this investigation for perfor­

mance evaluation of the considered algorithms is the 25-bar 3D struc­
ture which is shown in Fig. 6 with all the dimensional conditions. The
design consideration of this truss is depicted in Table 1. Based on the
symmetry about planes x–z and y–z, all the 25 elements of the structure
are divided into 8 groups.
The PFHV test optimal results for the 25-bar example is presented in
Table 4. In this test problem, the mean and STD values are considered as
the performance evaluation parameters. The mean values obtained for
MOAS, MOACS, MOSOS, MOHTS, and MOPVS algorithms are 1881.80,
1893.66, 1942.58, 1947.23, and 1950.25 respectively. Likewise, the
obtained STD values from the PFHV test are 9.79, 14.43, 0.54, 0.51, and
0.21 for the MOAS, MOACS, MOSOS, MOHTS, and MOPVS algorithms
Fig. 5. Median Pareto fronts of the 10-bar truss. respectively. MOPVS shows better results in terms of convergence rate
and search consistency. Moreover, for a fair comparison of all the

9
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

Fig. 6. The 25-bar space truss.

Table 4
The hypervolume values of results obtained for the 25-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 1851.02 1905.47 1881.80 9.79 123 5

MOACS 1853.57 1922.08 1893.66 14.43 177 4
MOSOS 1940.91 1943.59 1942.58 0.54 300 3
MOHTS 1946.23 1948.11 1947.23 0.51 400 2
MOPVS 1949.32 1950.49 1950.25 0.21 500 1

algorithms, a statistical test is performed (Friedman rank test) at a 95%

significant level. The test values obtained by MOAS, MOACS, MOSOS,
MOHTS, and MOPVS are 123, 177, 300, 400, and 500. As per the rank,
MOPVS outperforms other followed by MOHTS and MOSOS algorithm.
Also, the results values of all the five algorithms are different from each
other.
Table 5 shows the FSTE results obtained from all examined algo­
rithms for the 25-bar test example. The algorithms MOHTS and MOPVS
have the best values of mean and STD values i.e. 0.0065 and 0.0004,
0.0067, and 0.0001 respectively. Based on the Friedman rank values
MOHTS and MOPVS are the best algorithms in terms of the relative
spacing between the solutions in a non-dominated set for the 25-bar
example.
The median Pareto fronts for the 25-bar example from all the
considered algorithms are illustrated in Fig. 7 which is obtained after
100 independent runs. It is understood from the figure that solutions
from MOAS and MOACS are scattered and non-continuous. However,
MOPVS, MOHTS, and MOSOS Pareto fronts are smooth, well-
distributed, and steady. Also, these algorithms have an array of het­
erogeneous solutions. Considering all we can say that MOPVS is the
superior algorithm in comparison to other considered algorithms in Fig. 7. Median Pareto fronts of the 25-bar truss.
solving the MO 25-bar structure design problem. presented in Table 1. Motivated by the previous studies, 60 elements of
the structure are grouped into 25 members.
5.3. 60-bar test example The PFHV test results for the 60-bar test example is demonstrated in
Table 6. The mean and STD values are considered as the parameters of
For performance evaluation of all five considered algorithms, the efficiency measurement. The STD values obtained by the MOAS,
third benchmark test example examined is the 60-bar truss, which is MOACS, MOSOS, MOHTS and MOPVS algorithms are 166.65, 74.18,
depicted in Fig. 8. All the details of the design consideration are 5.92, 2.04, and 1.38 respectively. The PFHV mean values obtained from

Table 5
The front Spacing-to-Extent values of results obtained for the 25-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 0.0079 0.0590 0.0226 0.0084 455 5

MOACS 0.0053 0.0449 0.0170 0.0084 393 4
MOSOS 0.0118 0.0138 0.0133 0.0004 350 3
MOHTS 0.0054 0.0075 0.0065 0.0004 134 1
MOPVS 0.0063 0.0070 0.0067 0.0001 168 2

10
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

mean and STD values are the point of interest in the performance
evaluation. As per the PFHV test, the mean values obtained from the
algorithms MOAS, MOACS, MOSOS, MOHTS, and MOPVS are 2140.24,
2142.38, 2270.93, 2282.66, and 2285.46 respectively. The STD values
are 10.17, 19.50, 1.81, 0.63, and 0.90 respectively for MOAS, MOACS,
MOSOS, MOHTS, and MOPVS algorithms. The MOPVS algorithm pre­
sents better results in terms of convergence and search consistency. Also,
for a fair comparison of all the algorithms, a statistical test is performed
(Friedman rank test) at a 95% significant level. The test values obtained
by MOAS, MOACS, MOSOS, MOHTS, and MOPVS are 145, 155, 300,
403, and 497. As per the rank, MOPVS dominates the others followed by
the MOHTS and MOSOS algorithms. Also, the results values of all the
five algorithms are different from each other.
Table 9 demonstrates the FSTE metric values which signify the
relative spacing between the solutions in the non-dominated set for the
72-bar example. The corresponding mean values for MOPVS and
MOHTS are the same i.e. 0.0072. The best values of STD are obtained by
MOPVS i.e. 0.0002 followed by MOSOS and MOHTS which are 0.0004
and 0.0005 respectively. As per the Friedman’s test ranking, it is evident
that MOPVS is the best performer.
The obtained best Pareto fronts for the 70-bar example are illustrated
in Fig. 11 which is obtained from running the algorithms 100 times. The
figure demonstrates that the solutions from MOAS and MOACS are
scattered and non-continuous. However, MOPVS and MOHTS Pareto
fronts are smooth, well-distributed, and steady. Also, these algorithms
Fig. 8. The 60-bar ring truss. have an array of heterogeneous solutions. Considering all, we can say
that MOPVS is the superior algorithm in comparison to other considered
using the MOAS, MOACS, MOSOS, MOHTS, and MOPVS algorithms are algorithms in solving the MO 72-bar structure design problem.
3179.88, 3106.68, 4293.25, 4316.10, and 4323.02 correspondingly.
MOPVS and MOHTS algorithm obtained the Friedman test values of 500
5.5. 942-bar test example
and 400. The ranks allotted to algorithms conclude the dominance of
MOPVS over other algorithms. MOHTS and MOSOS are ranked second
The tower structure problem (942-bar) is the fifth benchmark
and third respectively in the Friedman ranking.
example considered in this study depicted in Fig. 12 with its geometric
The second test i.e. FSTE is performed for comparative analysis of
parameters. All accounted design aspects are presented in Table 1 for
algorithms and the obtained results are depicted in Table 7. The MOPVS
this test example. From a total of 200 structural elements, 59 groups
achieves the best mean value 0.0060 followed by MOHTS and MOSOS i.
were formed depending on the structural symmetry.
e. 0.0062 and 0.0120 respectively. Also, in terms of STD, MOPVS out­
The PFHV results for the 942-bar example are presented in Table 10.
performs the others with a value of 0.0001. The Friedman’s rank test is
The mean values obtained from the MOAS, MOACS, MOSOS, MOHTS
performed for all algorithms and the results show the ascendancy of
and MOPVS algorithms are 60795099.20, 61174938.07, 71950729.01,
MOPVS among all others whereas the MOHTS and MOSOS settled at
76064884.48, and 76279259.43 respectively. Likewise, the obtained
second and third ranks individually.
STD values from the PFHV test are 4084186.41, 1116525.94,
The median Pareto fronts of the third test example are illustrated in
696707.86, 113801.21, and 122569.23 for the MOAS, MOACS, MOSOS,
Fig. 9 that is obtained from 100 independent runs of all algorithms. The
MOHTS, and MOPVS algorithms respectively. MOPVS shows better re­
Pareto fronts obtained by MOAS and MOACS are broken and also
sults in terms of convergence rate and consistency. Moreover, for a fair
discontinuous whereas that from MOPVS is uniform and steady with
comparison of all the algorithms, a statistical test is performed (Fried­
well-diverged solutions. Hence the MOPVS is the better algorithm in
man rank test) at a 95% significant level. The test values obtained by
terms of non-dominated solutions.
MOAS, MOACS, MOSOS, MOHTS, and MOPVS are 152, 148, 300, 413,
and 487. As per the rank, MOPVS outperforms the others followed by the
5.4. 72-bar test example MOHTS and MOSOS algorithms. Also, the results values of all the five
algorithms are different from each other.
This subsection comprised the 72-bar 3D test example as the fourth Table 11 presents the FSTE results obtained from all examined al­
benchmark (Fig. 10), which was evaluated according to five algorithms. gorithms for the 942-bar test example. The algorithms MOHTS and
Table 1 represents all the loading, elements, and dimension details of MOPVS have the best values of mean and STD values i.e. 0.0066 and
this example. From 72 elements of the structure, 16 groups were formed 0.00038, 0.0068, and 0.00040 respectively. Based on the Friedman rank
based on structural symmetry. values MOHTS and MOPVS are the best algorithms in terms of the
The PFHV values of the 72-bar example obtained by the various al­ relative spacing between the solutions in the non-dominated set.
gorithms are reported in Table 8. Similar to the previous examples, the The median Pareto fronts for the 942-bar example from all the

Table 6
The hypervolume values of results obtained for the 60-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 2465.08 3397.56 3179.88 166.65 173 4

MOACS 2905.27 3276.04 3106.68 74.18 127 5
MOSOS 4271.94 4304.66 4293.25 5.92 300 3
MOHTS 4310.91 4320.11 4316.10 2.04 400 2
MOPVS 4319.62 4326.55 4323.02 1.38 500 1

11
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

Table 7
The front Spacing-to-Extent values of results obtained for the 60-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 0.0100 0.1339 0.0349 0.0195 456 5

MOACS 0.0079 0.0745 0.0299 0.0137 438 4
MOSOS 0.0101 0.0129 0.0120 0.0005 306 3
MOHTS 0.0054 0.0072 0.0062 0.0004 172 2
MOPVS 0.0056 0.0061 0.0060 0.0001 128 1

considered algorithms are illustrated in Fig. 13 which is obtained from

the 100 runs. It is understood from the figure that the solutions for
MOAS, MOACS, and MOSOS are scattered and non-continuous. How­
ever, MOPVS and MOHTS Pareto fronts are smooth, well-distributed,
and steady in nature. Also, these algorithms have an array of hetero­
geneous solutions. Comprehensively, MOPVS outperforms the other
considered algorithms in solving the MO large structure design problem.
Fig. 14 shows the convergence plot based on the mean hypervolume
values against the number of function evaluations. Only the top three
best algorithms are considered. From the figure, MOPVS and MOHTS
performances are equally good. When zooming into the figure at an
early stage, it was found that MOPVS is slightly better than MOHTS at
the number of function evaluations (FEs) before 2000. This paves the
way for further improvement of the algorithm through hybridization.

5.6. Multi-objective convex test functions

Further performance investigation has been made. Rather than the

five real engineering applications of truss optimisation problems, six
traditional convex test functions; ZDT1, ZDT3, ZDT4 (Zitzler, Deb &
Fig. 9. Median Pareto fronts of the 60-bar truss. Thiele, 2000), Four Bar Truss design, Hatch Cover design, and Rein­
forced Concrete Beam design (Tanabe & Ishibuchi, 2020) are also

Fig. 10. The 72-bar 3D truss.

12
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

Table 8
The hypervolume values of results obtained for the 72-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 2112.10 2162.75 2140.24 10.17 145 5

MOACS 2082.81 2176.18 2142.38 19.50 155 4
MOSOS 2267.43 2274.65 2270.93 1.81 300 3
MOHTS 2280.93 2283.95 2282.66 0.63 403 2
MOPVS 2281.32 2286.28 2285.46 0.90 497 1

Table 9
The front Spacing-to-Extent values of results obtained for the 72-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 0.0109 0.0436 0.0227 0.0072 434 4

MOACS 0.0079 0.0761 0.0268 0.0138 441 5
MOSOS 0.0134 0.0156 0.0144 0.0004 324 3
MOHTS 0.0056 0.0083 0.0072 0.0005 154 2
MOPVS 0.0065 0.0075 0.0072 0.0002 147 1

Fig. 11. Median Pareto fronts of the 72-bar truss.

employed for performance investigation. For each problem, population

size and the number of iterations is set to be 50 and 100, respectively.
For the algorithm using different population sizes, they will be termi­
nated at 50 × 100 FEs. The top three algorithms from the truss opti­
misation problems; MOSOS, MOHTS, and MOPVS are used to solve the
standard test functions for 20 optimization runs while the results based
on hypervolume are reported in Table 12. From the table, the most
efficient optimizer for the ZDT1, ZDT3, and Reinforced Concrete Beam
design is MOHTS while the most efficient method for the ZDT4 and
Hatch Cover design is MOPVS. For the Four Bar truss design, all opti­
misers perform similar mean hypervolume values while the MOSOS
gives a slightly better standard deviation.
Fig. 15 shows the search history based on the normalised mean
hypervolume indicator versus FEs for all test functions. For ZDT1 and
ZDT3, MOHTS and MOSOS are said to be fast at the beginning while
MOSOS slows down in the second half of the run. Meanwhile, MOHTS
remains consistent and converges to the final values after about 200
function evaluations. For the ZDT4, MOPVS shows its clearly fastest
convergence rate from the beginning until reaching the termination Fig. 12. The 942-bar tower truss.
criterion while the second fastest is MOHTS. MOPVS and MOHTS show a
somewhat similar convergence rate to the final values for the Four Bar
Truss design and the Hatch Cover design problems while, for the Rein­
forced Concrete Beam problem, the MOHTS has the fastest convergence
rate at the beginning while the MOPV can later converge and catch up

13
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

Table 10
The hypervolume values of results obtained for the 942-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 52814995.24 67170665.89 60795099.20 4084186.41 152 4

MOACS 57425914.37 63616875.48 61174938.07 1116525.94 148 5
MOSOS 70108177.16 73682236.85 71950729.01 696707.86 300 3
MOHTS 75780010.21 76306770.47 76064884.48 113801.21 413 2
MOPVS 76013417.23 76536083.23 76279259.43 122569.23 487 1

Table 11
The front Spacing-to-Extent values of results obtained for the 942-bar truss.
Algorithms Min Max Mean STD Friedman test Friedman rank

MOAS 0.0143 0.1207 0.0427 0.02098 471 5

MOACS 0.0104 0.0799 0.0290 0.01468 415 4
MOSOS 0.0122 0.0169 0.0149 0.00076 314 3
MOHTS 0.0057 0.0078 0.0066 0.00038 137 1
MOPVS 0.0061 0.0081 0.0068 0.00040 163 2

Table 12
The comparison of standard test convex problems based on hypervolume
indicator.
Test functions HTS PVS SOS

ZDT1 0.7419 ± 0.64669 ± 0.73238 ±

0.000535 0.042206 0.0016155
ZDT3 0.9509 ± 0.80164 ± 0.91527 ±
0.0010863 0.056294 0.011625
ZDT4 0.34753 ± 0.73765 ± 0.17702 ±
0.050306 0.071183 0.050891
Four Bar Truss 0.98559 ± 0.98559 ± 0.98559 ±
1.5661e-08 1.8594e-08 1.6247e-09
Hatch Cover 0.83792 ± 0.83823 ± 0.83817 ±
0.0011686 0.00083907 0.00055051
Reinforced 0.365 ± 0.36427 ± 0.36462 ±
Concrete Beam 0.00032971 0.00034434 0.00042715

Bold value signifies the best solution.

MOHTS. This implies that the proposed MOPVS is acceptable for

general-purpose multi-objective optimization.

6. Conclusions
Fig. 13. Median Pareto fronts of the 942-bar truss.
A novel MOPVS algorithm is developed and investigated for multi-
objective structure optimization design issues. Inspired from the funda­
mental two-lane highway overtaking mechanism of the vehicles, the
MOPVS is a global optimization algorithm. For performance evaluation,

Fig. 14. The 942 bar search history based on mean hypervolume indicator (a) original view (b) zoom in.

14
S. Kumar et al. Expert Systems With Applications 169 (2021) 114511

Fig. 15. Multi-objective standard convex test functions search history based on mean hypervolume indicator (Normalised).

five challenging structure test examples are accounted with discrete relatively new in this field, more research is needed for prediction of its
cross-sections under multiple loading conditions. Also, six convex test global optimization behavior which can be contrasted with other
functions are added for further performance investigation. For the prominent algorithms.
comparative analysis, the outcomes of the MOPVS can be compared with
four distinguished algorithms. Moreover, for measuring the effectiveness CRediT authorship contribution statement
of the proposed strategy, PFHV and FSTE tests were conducted. Simul­
taneously for statistical analysis of considered algorithms, the Friedman Sumit Kumar: Conceptualization, Methodology, Writing - original
rank test is performed. Outcomes of this test demonstrated the superiority draft. Ghanshyam G. Tejani: Conceptualization, Methodology, Soft­
of MOPVS over the other accounted algorithms in terms of obtaining a ware, Formal analysis, Data curation, Writing - original draft, Writing -
good balance between global diversification and local intensification. review & editing, Supervision, Project administration. Nantiwat Phol­
Moreover, MOPVS results are significantly diverse from the other MHs, dee: Conceptualization, Methodology, Software, Formal analysis, Data
especially for large-scale truss design problems. From the proposed curation, Writing - original draft, Writing - review & editing, Supervi­
investigation, it is understood that MOPVS achieves the better non- sion, Project administration. Sujin Bureerat: Funding acquisition,
dominated Pareto front with a diverse optimal solution set in contrast Conceptualization, Methodology, Software, Formal analysis, Data
to MOAS, MOACS, MOSOS, and MOHTS algorithms. curation, Writing - original draft, Writing - review & editing, Supervi­
In the future, one can investigate this algorithm for higher di­ sion, Project administration.
mensions and intricate engineering design problems. As this algorithm is

15
Discovering Diverse Content Through
Random Scribd Documents
[Inhoud]
 ZEVENDE HOOFDSTUK.

In ’t hart van ’t haventje, en op Baanwijk-boulevard,

was plots, tusschen zonnelaat licht en gewoel van
groentekarren, onder kastanjelaan en boomlommer,
vreemd gedruisch van kermiswagens komen
òplawaaien; brokken van tenten, ingangen, poortig-
cirkusachtig beschilderde deuren; warrel van paarden,
tusschen schonkige kerels en meiden,
zigeunerstoeten, overwaaid van vreemden gloed, en
kleur-fel bont gekeer.—Langs de Haven en Baanwijk
ratel-wielden en bonkerhotsten de gloeierige
kermiskarren áán, met de schooierige kerels vóórop,
paardje of ezel vast aan den toom, zwenkend en
kronkelend door havenrumoer en sjacher-gedrang van
venters.

Op de kar-balustrades waggelden morsige vrouwen in

rotplunje, met onooglijke kinders, borst-lurkend.—En
achter, en òm de karren hortte gore nasleep van
beluisde wezentjes, meestrompelend over kei en grint.
—

Onder ’t jongezellen-landvolk van Wiereland furiede

de kerels-uitgelatenheid los, lichtelijk flirtend met
woelig wellustige meiden, brutaal en fel van passie.
Troepjes werkers sprongen in klepperenden
klompendans en boeren-kankan, met beenen-
guitigheid en komieke handgebaren op de meiden
áán, midden op straat, onder erotisch bombardement
van zoenen. Volkje van Wiereland, ’t zuipende
 verhitte, uitbarstende, hoorde van àf de akkers de
meiden gillen en heet schateren, hoorde ’t gewoel van
kermisgangers en daverend belalden ze elkaar, in ’t
voorbijgaan. Ze hoorden geklop, getimmer op de
Haven, door de straten heenhamerend in verwaaid
gerucht, uit tenten en kramen die in ’n paar uurtjes,
geraamte-ribbig en naakt opgegroeid [272]stonden. Ze
hooren ’t lawaai van de kermisrazernij, de vreemde
stemmen en spraken.—’t Zien aandrijven van
schuiten, met ’t kleurig spul van brokken schiettenten,
draaimolens, bont gestraal van zwaar beverfde
schommelbakken, goot scheuten gierende pret over
de popelende kijkers.—

Vreemd getrappel van hitjes uit ’t paardenspel

hoefklakkerde over de Haven, en de kinder-kijkertjes
omkringden, wild en verbluft in verbijstering, de
beesten. Gemompel en gekonkel tusschen de
Wierelanders klonk overal door, al maar over de
kermisjool; wat er zijn zou, wat er wàs, wat ’r komen
ging.

Schetterende voorklank en rauwe jubel van lol,

brachten de straatorgels, twee dagen vóór ’t openen
van kermis,—dertig levende klank-orkanende wezens,
elken ochtend vóór ’t stadhuis bijeenwaggelend in rij,
om van die gewijd-officiëele plek uit, heel Wiereland
van alle kanten vol te roffelen, te betrompetten, te
bepauken in schellen schater en daver, te omgolven
met heet getjink, geraas en gedreun van dooreen
rammelende klanken.—Midden in tragen zwoeg der
landwerkers omdaverden ze de straatjes in hellen joel
en roffel. Soms rammelden drie orgels naast elkaar in
 de deftige Bikkerstraat, in heeten donder van
trombones, de stedekestilte met klankrazernij
verpletterend. Aan de huisjes tingelde ’t den
ganschendag; gingen schuwe armoekinders, in
doorvreten plunje, naakt-beenig en stank-
verwasemend met ’t bakje rond, liepen moeders
achter de draaiorgels áán, in gore ellende van
stoelenmatters-scharrelaars. De wreed-snuitige
kerels, ’t straatorgel verrollend van deur naar deur,
bleven door-draaien onder ’t voortwielen, en in woeste
gluiperigheid knorden hun harige dier-woeste tronies,
snauwen naar wijf en bedelkinders. Als verstrooide
optocht van melaatschen schooierden ze achter
elkaar aan, de klankengekke orgellui. En ’r òm, de
angstig-verschuwde vrouwenhoofden, mottig-wreede
en blondharig, verwilderd-romantische, op zwaren
borsthang kinders sjokkend; en de rossige en zwart-
besnorde boeventronies, van mannen, jukkig gekaakt,
doodelijk uitgeput met schuwen oogenkijk van
gluipende overrompelaars en hongerende dieven; of
zwaar krachtpatserig d’r lijven, in landloopers-
gezondheid, [273]tusschen de wijven, dralend en
bedelend, versmald in den knel hunner rood-vale en
groene misère-jassen.—

Daartusschen in forschten òp, donker bronze kerels,

met koperig ringwerk in d’ooren gedraaid, zwart-
glanzend haar, kleurige doeken om den hals,
melodramatiekerig-romantisch in hun zuidelijk
armoekoloriet, hun tambourijn in knuisgedruisch van
rinkel en bellenzang, met tenoor-beverig
weemoedslied begeleidend.
 En telkens de romantiekerige Mignon-schooiers,
snauwden scherp-radde woordjes neer op de
bedelkindertjes, als kleurige deurtjes van cynische
Bikkerstraters gesloten bleven en bakjes niet meer
kletterden van rondgestrooide centen.

Zoo ging er al twee dagen, voor opening van kermis,

sjofele woel en rondgang in alle Wierelandsche wijkjes
en achterafjes; bleef jubelende klinkklank schetteren
in kleurige deunen; klompklepperden om de orgels al
meer kerels en meiden hun boerenkankan. ’s Avond’s
vooral, als op de drukke, breede Baanwijk, tusschen
duisterende geraamten van tenten en kramen,
vlammig fel licht uitschoot van de winkels, de
stoeiende paren overbeefde in rood-gelen walm, en
kleurige vóórjubel van kermis demonisch tegen de
half-aangegloeide boomstammen oplaaide, schroeide
rond in avondbrio, de duivelsche zeng van
karnavalgloed.

Op de Haven was de voorwoel ’t hevigst. Daar zeilden

en dobberden de schuiten áán, voor den wal.—
Woeste hunkerende jochies, doortrild van
kermisrazernij, smakten zich als wroet-miertjes op de
spul-brokken, van schuitdekken afgevoerd naar
standplaats. En al luider, door den nog drukken sjouw
en markthandel der groenboeren, dwars door vruchten
en voerkarren, ging de gistende kermiswoel.
Werkelooze sjouwjochies en schoolkinders hijgden
onder de vrachten die ze te verdragen kregen,
planken en fundamenten van draaimolens,
vuilkoperen stangen, paarden en ringwerken, staven,
wilde dieren en kleurige lompen. Ze hijgden, met
kramperige pret van aandoening, en dolle lol-stralende
oogen. Ze gierden dat ze zoo vlak bij, in hand-
tastende aanraking leefden met de heilige kermis-
brokken. [274]Ze verbluften elkaar met ’t tillen van al
zwaarder vrachten, en dwars door den
groenboerzwoeg en kroegsjacher, hielpen ze de
vreemde, norsche kermiskerels, die gretig de ventjes
lieten sloven en sjouwen, zich vertillen en half-
verrekken. Want diè merkten den pret-duizel in hun
helle oogen, den groei van hun genot bij elk grooter
stuk dat uit de schuiten opgedragen werd.—En
angstvoller van ontroerende pret, verbreedde de
echtheid van hun kermisvizie, als ze uitzinnig, plots te
dragen kregen, groote brokken dekoratie, bont
schilderwerk, verf-riekende doeken waarop
moordtafreelen bloedden, waarop bulderzeeën
schuimden met lijk-paarse schipbreukelingen,
uitgeteerde weggezonken juffers en meneeren in
nood; waarop kruitdampten, veldslagen van
Transvaal, Boeren in bont rumoer, onder angstige
hemelen, stinkend naar bakolie en gemeene
uitgedroogde verf.—

Met bevend ontzag, staarden ze op de halve lijven van

moordenaars, op den zeebruis en ’t rottende
zwalpende vlot neer, op hakkende Boeren en
doorgevlijmde Engelschen.—

Een groepje droeg ’n griezelig moordtafreel op den

kop, en gestreden werd er hevig, met heerlijken
griezel-angst voor de werkelijkheid der beelden, dat
Jan de beenen had van de moordenaars, en Gijs de
bovendeelen van de schipbreukelingen, Piet de
 geweren van de Boeren, en Toon den grond. En vlak
voor hun oogen gebeurde ’t, al die gruweldingen. In
rillender suggestie liepen de jochies vlak achter elkaar,
staarde de een naar den ander, niet wetend wat ’t
worden ging, als de brokstukken op elkaar ingevoegd
gebouwd stonden. En al meer kermisdingen sjokten
ze áán. Komieke harelekijnen, met streepkleurige
broeken voor schiettent, wapenschilden, kachelpijpen
en hoededoozen. Dan schoot schater los, ontspande
hun angst voor de front-dramatiek der groote kijk-tent.
Al komischer mengeling van grappige en angstige
voorwerpen droegen ze uit, en zóó voelden de jochies
de kermis in hùn gretige pootjes, nu zij hielpen
opzetten en bijeengroepen. De kermisklanten beukten
en vervloekten de kinders, als ze in hun verbijstering
en lol, verkeerd den rommel uitdroegen. Dan joegen
ze de kereltjes plots wèg met ’n woest lippengebrul
om angst te wekken. In bangheidsstuip holden [275]ze
heen naar huizenkant van Haven, tusschen karren en
geboomte doorzwenkend, uit schrik brutaal scheldend
en hoonzwaar uitfluitend de kermisgasten. Zoo van
ver, keken ze dan met stil genot, naar groei en
vleeschwording van al de tenten, draaimolens en
spellen.

Eén breed en hoog lattengeraamte stond er al boven

alles uit, ’n schouwburg waar kerels tot in den nok
hurkend hamerden en klopten, in davering van
scheepstimmerwerf. Schouwburg-geraamte was zoo
vlug verrezen dat de tuinders ’t bijna niet hadden zien
bouwen.—
 Aan stillen kant van ’t haventje, ònder polderdijk,
schooierden bijéén, op ’t hel-groene grasveldje, de
bonte kermiswagens. Als melaatsch menschgroepje
hurkten daar saâm, de verluisde kinders, meisjes en
vrouwen met hangende haren en vervuilde ragebollen.
Hun morsige bloote voeten ploeterden in ’t warm-
zonnige gras.

De wijven daar, met rottende lompen aan ’t slobberig

lijf, zwijnden rond hun kermiskar, en boevige kerels,
rossige en donkere, luierden op de balustraden-
trapjes, pijpen den mondhoek ingebeten, rookten
vadsig in zonnebroei. Rond hen zwirrelde maar ’n
trage warrel van ’t beestig-morsige kinderkriel,
krijschend, joelend, bevuilend ’t grasveld, dat helgroen
vlamde ònder kleurig-melaatschen menschenkluit.

Eén dag vóór de kermis was Augustus-zon weer

doorgevlamd achter wolkgrauw, zette ’t plots
bewoonde, anders doodstille grasveldje, met z’n
stinkende luier-ellende, in groenen lichtdaver, en
kleurenbrand. De fel-bonte kermiskarren, eerst dof en
druilig onder grijs-stugge hemelstolping, blakerden en
vlamden nu òp in ’t gras, de plankige kasten, hard
geel, wreed-fel geel en rauw Zaandamsch
hutjesgroen, féros en wrang in den zeng-gloei van
smoor-dampende Augustuszon.

Rood-bruin de hekjes en balustrades, menierood-helle

pilastertjes en wielspul, kanarie-gele en meloengoûd-
vlammige zijwanden, scharlaken assen-spul en
spaken, zoo gloeiden ze tegen elkaar òp, de kist-lage
kermiskarren, als één brandende kleur-rij, [276]één
braking van rauwen gloed, van ònder zengend ’t gras,
van bòven zengend de lucht, ’t eindeloos-aangroeiend
polderblauw.

Als ’n drom harig begroeide apen kropen de trage

wezens, loom dooréén in den zonneblaker;
kammende wijven, hekserig en brutaal, naakte kinders
en verlompte meisjes. Tusschen blauwig-triest gerook
van kleine straat-kacheltjes, midden op ’t veldje,
hurkten smoezelige tooverkollen met sproetige
zuigelingen. Daar naast rengelden, wijdbeensch-
achterover en lig-zittend ’n paar jonge meiden in
oostersche sfeer van siësta-traagheid, schreeuwend
om beurten tegen ’n stoetje kerels, zwart als
kolenbranders. Tusschen de kacheltjes aan eindhoek,
stonden traag konkelend, donkere kermisgangers in ’t
trillende walm-zwart van den rook, elkaar te
beschelden, en jonge meiden, in verkaalde
fluweelmisere-jakjes,—paarsig en donker
smartenrood, gore restjes van pronk’rige
akrobatenplunje,—scholden méé, woelden zich in de
loshangende ragebollen, omzengd van karren-gloed
en kleurenvlam van knetterend-heeten zonnedag. Al
meer belompte vrouwen waggelden áán uit
miniatuurdeurtjes van karren. ’n Walg-stank van zurige
ellende walmde òp uit de kleur-heete kermisplek. En
vreemd, in Spaansche cier, ging tusschen den
zigeunerigen kleurgloei van hun wagens en rommel ’t
brio van hun versleten opschik en rauwe behaagzucht;
Spaansche cier daar wreed neergeluierd in de
schroei-felle zon, overgolfd van licht, naakt-gezengd in
den dag, versmart en rampzalig in den verbijsterenden
vloek van hun helsch-bonte vurigheid; rampzalig in het
 verzweetende blanketsel van hun cirkuspronk en
passie, overlaaid in zonnigen smoor met de
bloedroode en geel-férose verf van hun huiverig-
opgesmukt, moordend ellende-bestaan.

Vrouwen met bloote borsten sleurden kinderen mee bij

de haren, in de krot-nauwe karren. Andere, spiernaakt,
werden weer de holletjes uitgestooten, schreeuwden
en grienden, vluchtend naar versten hoek van
grasveld. Telkens monsterden nieuwe harige koppen
òp uit de lage wagen-deurtjes, plots wègduikend in ’t
vuil-diepe kamertjes-binnen van kar; verschenen dan
weer, kiekeboeënd in gesprek-schreeuw met
[277]buitenhurkenden.—Ontkleed boenden anderen
zich op ’t gras, tusschen en vóór de wagens, dierlijk
naast elkaar krijscherig doorscheldend.

Van woelige Haven af, klonterden de karren en

menschen in wriemel van kruipende kleuren, onder de
drijvende stilte van polderlucht en land, eindloos wijd-
om. Dìchterbij ronkte de hurrie uiteen, verwalmde
kachelstank en rook, rond stoetjes en rommelige
wagenboedels. Onder de karwielen kropen al meer
klierige kinders uit, met moddertronietjes, zwart-
verbakken, als vervuilde duiveltjes zich opgravend uit
aard-ingewanden. Tusschen de paardjes en
trekhonden vochten ze, en één morsig groepje
ranselde ’n vuil-zinlijk aapje, dat grijnsde en wild
flikkerde met oogenwit, bij elken mep en ruk aan z’n
staart z’n hekserig rimpelkopje nijdig plooide,
doorjeukend met z’n harige handen z’n luizige lijfje,
bedacht tòch op iederen kinder-aanval.
Wild gekonkel ver-ratelde onder de kakelwijven, rauwe
stemmen in alsoortig dialekt, Vlaamsch, Fransch,
Duitsch en Amsterdamsch woordgewarrel; spraak van
wezens die in heete drift van samenbroei maand aan
maand, elkaar al verstonden door stem-intonatie.
Vloeken en dreigementen bulderden rond in plotsen
overgang van zacht-vleierig gesprek, verdoffend in
klankloos geteem.

Achter bassenden en jankenden hondenstoet lagen

jonge meiden in luierkring op d’r buik, op ’t heet-
zonnige gras, de handen ingehaakt onder kin, met
onderbeenen de lucht inzwabberend, zich naakt
schoppend tot den rug. Ze begierden elkaar, groeven
zich vuilnisbelten van zinlijke lol, stootten de handen
onder de kaken wèg, dat hun tronies ’t zengende gras
insmakten. Ze beranselden elkaar in woesten stoei,
dat hun flodderige lompen scheur-risten en kraakten
van de half-naakte lijven. Los de haren over
schouders en ontbloote borsten, droogden ze zich de
zweettronies met d’r onderrokken of gore
hemdfladders.—Een brandende woestijnige hette,
schroeide over den meidenstoet in ’t gloei-gras, dat
schitter-vonkte en hel-groende. Een luiïge loomheid
sufdutte dan plots weer in de snikheete Augustus-
zonning.[278]—En rookige walmbenauwing van de
straatkacheltjes, verwoei wolkerig over hun verstoofde
bloote armen en beenen, en morsige lijven.

Akrobaatkerels en draai-orgelaars schuimden daar

samen met uitgestooten schooiers en landloopers van
den omtrek, bedelaars en manke mirakels, zich veilig
voelend in de boevige broederschap der
 kermisklanten, hinkend en lollend onder de
goochelaars en kijvende waarzegsters.

In Oosterschen gloei zwirrelde tegenover den

meidenstoet ’n vrouwenkeet, afgebeulde kijfkoppen,
waaronder zeerig-beklierde, bevrat en bepokt, met
bloed’rige vlekken onder d’oogen als moordmaskers in
melodrama. Afgezonderd op kachelrand zat één jonge
vrouw, mooi-Carmenachtig gloeiend ’t zwart-blauwe
prachthaar, met ’r droef-verliefde hondenoogen te
staren in de lucht, verwiegelend ’n versnotterden
slapenden zuigeling op d’r knie. Als ’n paar
orgeldraaiers van hun rondgang door ’t stedeke, ’t
grasveld opreden, wuifde en kruifde plots ’n
waaierspel van hel-rooie, blauwe, gele en wit-gore
rokken òp achter de wagens van den meidenstoet,
bradend op den buik in zonnevuur. Dan kwam er
danspassie onder de heksen en jonge meiden,
kiekeboeden wat tronies en halve bovenlijven uit
kardiep, ontblootten zich brokken van kammende en
kleedende armen. En achter de orgeldraaiers liepen
mee, centenophaalsters, donker-harige meiden, slank
en kanaljeus in hun groen en paarse afgesleten
fluweelen lijfjes, vurig met hun kleur-gloeiende
hoofddoeken als zigeunersche kappenbrand van
Spaansche cier op de pracht-donkere haren. En langs
de slanke heupen, tenger en wulpsch-elastisch,
krioelde ’t van kleurige doeken, en om de
meidenschouders vonkte ’t van goudspatjes, vurig
gevlam van roode en groene sjaal-franjes, bonte en
warme cier van zuidelijke passie-beesten. Er ging
geschuifel tusschen de karren bij hun ommetocht, er
verklonk gevloek en gelach op de kar-trapjes, in ’t
 wagendiep, en uit den meidenstoet, met d’r kinnen
ingehaakt op handen, rolden al meer paren wèg, ’t
grasveld òver in brandenden zonnezwijmel.

Geen sterveling van fatsoenlijk Wiereland durfde over

de [279]spoorbrug ’t grasveld op, te gaan kijken naar
den luizentroep, die van verre, stil in eigen sfeer, z’n
brandend brio en traag-geschooier bleef verzwijmelen,
—

Van woeligen havenwal àf, tegenover ’t polderstille

wèggeschuwde zonnige grasveld, stroomde ’t
zonnelicht in kleurigen klots néér op ’t bruin-groene
watertje. In troeb’lige golfjes kabbelden daar in
zonnewiegel, de kleur-zengende kermiskarren,
onderste boven. Ze gloeiden en plasten daar in ’t hel-
bezonde havenwalletje, als met vurig zwam
bestreken; de wijven òmgekeerd met hun roode,
paarse en groen fluweelen rokken, hun vleesch-gore
malots, de blonde en roode kinderragebollen; de
omgekeerde naakte morsige lijfjes; de wagens met
hun rauwen gloed en uitgespuwd verf-bloed, schel
vernist in de Augustus-zeng, strakke trillende
polderhitte en warmtenevel, inzuigend ’t helle licht in
den alkleurigen zongloei, verscherpend de vlam-felle
ommetrekken van karren en wezens, tegen den
blauwenden polderhemel in.

Zoo dreef de melaatsche hellestoet, in vulkaan-

kleurige hevigheid neergezwamd op ’t nat van den
walstroom, zwabberend en verkruivend in de
waterrimpels en ’t golf geklots, omkabbelend pramen,
tjalken en schuit-vervoer. Zoo leefde daar twéé keer,
 brok bestaan van zomersche kermismisere, in de
trage zonnige slaperigheid en den loomen zwijmel van
niets-doen; één keer op ’t doodstille grasveld,
aangeschuimd in de kokende kleur-klater van licht, de
hel-kleurige karrenrij in den smorenden zeng van
zonnedag; àndere keer, in ’t walletje, ’t overgloeide
water, karren en wezens omdraaiend in de spiegeling
van ’t brandende licht, dooreenvloeiend in heeten
kleurenwarrel.

En rondomme van de Haven roffelden de orgels,

snerpten de fluitenaars, trompetten en basten de
registers hun demonischen klankenorkaan, tusschen
rauwen menschenzang en schetter van
kermisinstrumenten. [280]

[Inhoud]

II.

Zwellend gerucht van kermislawaai ging door ’t

stedeke. De orgels bleven doorrazen in valdreun van
joelklanken. Zon bleef gloeien, haalde de Kermis in,
met helle lichtflonkering. In den roes van ’t
orgelgeschetter verkookten de meiden en kerels hun
passie voor nachtpret. Ze beefden verbijsterd onder
dreun, deun en zangschetter, en in bloedbruising
gistten ze ten dans, in wilde extaze, vastgrijpend wat
er voorkwam, de brandende wellustmonden open, om
 telkens op nieuw weg te duizelen in ’t rond, al in ’t
rond, tot ze neerstortten tegen de keien en karren.

Van ’t land en uit de werkerswijkjes kwam sterker

druisch en jubel, op den dag telkens gesmoord. Maar
’s avonds in de stille straatjes, aangehitst door ’t van
verre aanklankende orgelgerucht, barstte de jool òpen
als ’n vuurlong van krater. Hoog spoog de kleurhitte
van hun passie boven huisjes, tenten en boomen, en
gierende zang en zuip van een en anderen kant,
botsten al op elkaar in.

De kanaljeuze meiden van de groentenfabriek, de

groenboerenknapen, de wijven en vrijers, jochies en
kinders, alles trok op, in den nabroeienden
zomeravond, hurkten, hortten en joolden naar de
kermis, de heilige kermis.

Pastoors en dominees waarschuwden plechtiglijk en

star, maar ’n heete hoon-schater van de knapen
raasde hun tegemoet, dat ze zwegen en stom
aftrokken.

Eindelijk dan stond de heele lange Baanwijk-

boulevard vol kramen, tenten en spullen. Ze leek ’n
diepe allée van flonkerende kleuren, schijnsels en
lichtvegen; flonker van poffertjeskraamruiten en
gebrand glas, met kobalt, brons-gouden en rood-
groene sterreling van deuren; flik-flak-zoetige toetsjes
van [281]rose en goud-randige ornamentiek, vèr-
echoënd in kleurigen warrel tusschen vuilig
vergroende spiegels. Overal glom gloei van koperen
bakken en meelpotten, ’n schemer en flonker van
goud en hel licht, tusschen de hooge, lage, dwarse en
hoekige tenten en spullen, aan weerszij, dol van
schitterig bonte fleurigheid.

’s Middags opende de kermis met klein gejoel pas van

kinders, jochies en meisjes, feestelijk opgesmukt in
Zondagsplunje.—

Op havenkant was schorriemorriënder spullenlawaai

saamgedromd; de schelle klingeling en ’t heete
galmend belgeraas van draaimolens, ’t gebons van
karren tusschen ’t eerste gemoker en gedreun van
vaste stàndplaatsorgels, vóór of in de spellen.—
Dienstmeisjes koekeloerden, jongens schaterden en
kinders gilden en sprongen als razende konijntjes,
overal rond.

Tegen den middag was de zon weggebleekt, na de

kermis met fonkel en gouden glans te hebben
ingehaald. Maar zwoel en broeierig bleef
Augustushitte zwangeren uit druilgrauwe lucht, boven
’t woelende stedeke, grauw-paars boven orgelenden
havenwoel en Baanwijk-boulevard.

Dirk, Piet en Rink van den polder, hadden zich voor de

kermis samengevrindschapt, loerend op de meiden
van Grint. Jan Grint, uit eigenbelang, om Guurt meer
te ontmoeten, school bij hèn; Geert en Trijn, Corrie en
Annie weer bij Jàn, om vrijer overal in te draaien en
mee te lollen.—Klaas Koome, de Hazewind, was bij
de nèven Hassel getrokken, half gelokt met ruim geld,
als spotkracht en durfkerel, en om sterker dan zij
durfden, de meiden van Grint naar hèn toe te tronen.
 De meisjes voelden den aandrang van twee kanten,
hurkten bangelijk in den knel. Maar Trijn wou niet dat
Geert d’r hoofd zou breken, om die narigheidjes. Wat
kon hun ’t schelen hoe ’t gaan zou met die vijandelijke
kerels, nou ze zelf overborrelden van heerlijken
kermiswellust. Zou wèl losloopen.

’s Avonds stapten Dirk, Piet en Rink op ’t pad, gingen

ze Jan Grint halen.—’t Speet de meiden dat Jàn niet
bij de nèven hoste, de neven met den loodpot! Maar
Jan was zèker bij [282]Dirk minstens ’n paar keer Guurt
te treffen. Nou moesten de meisjes mee met hèm. Ze
stonden klaar aan de deur, liepen naar voor, naar
achter, popelend zenuwdruk, gejaagd.—

Toen ze de mannen hoorden aanstappen, en de

vooruit schallende schreeuw van hun zangkelen ’t pad
òpklaroende, stapten ze hun huisje uit, zoenden
moeder Grint, vader en kleuters, en volgden vlug de
lustige bazuin-juichende knapen.

Dwars door hobbeligen duinweg, op polderkant aan,

ging ’t los.

Piet, Rink en Dirk joedelden in dolle dwarse sprongen,

Piet nog op z’n klompen, om lekkerder straks te
klakkeren en klos-meppen op de tentplanken.

In slinger-rij, de drie meisjes omarmd-ingehaakt,

zongen ze den duisteren avondweg àf, die zomerstil
lag te suizelen in heimvol geruisch, onder donker
bleek-sterrenglanzend azuur.
Rauw krijschten de kerelsstemmen door den gil’rigen
vrouwenzang, in opgewonden passie ’t kermislied uit.

Arreme frouw en kin.. d’re

Stuurt men noàr de moordenoarskampe haine..
Aooaauuw! waa’t ’n skande
Loage wroak van En-gè-land!

En telkens met iets bloeddorstigs hysterisch in den

nationalen zangkrijsch scheurden door ’t avondstille,
week-golvende duinpad, in zwenk van alle stemmen
plots, de rauwe geluiden bijéén, een sleurige zeil van
gloeiende passie naar ’t refrein:

Aooaauw! waa’t ’n ska-ande!

heel ’n leven van wraak-bloedige dierlijkheid en woede

er in uitbrallend.

Midden op den dwarsweg trad plots ’n donker stoetje,

ook brullend en zangbruisend bijeengehaakt in
slingerrij, woest tegen de Hassels en Grintjes
insliertend. [283]

Ikke sel d’r bai stoan op wè-el en wai!

Deesie.. Deesie!

verkalden daar krollige mannestemmen. En vlak voor

de meisjes hielden stand de zangers, donker-groot in
’t avondduister van duinpad.

—Saa’k stikke! da hai je de maide van Grint, spoog

Willem Hassel er uit, met ’n spatstraal van z’n pruim
tusschen z’n tanden.—
—Kaik! klonk ’t verrast van andere zij, da binne dur
Willem.… Jaa’n.… Henk! en Klaas Koome! telde Piet
vagelijk in ’t duister.

—Kaike?.. Hoasewind! f’rek.… en.… enne.. waa’s dá’

veur ’n snaiboon.… wâ? Bolk, Platneus jai? jài! gierde
Dirk, die eerst nijdig had willen worden maar in
verglijing van bui, in ’n woesten schater schoot.

—Is je waif dur t’met protekollig Ouë, spotte Piet, kèn

d’r dâ ouë bier nog werreke?

De neven Hassel hadden met Klaas Grint afgesproken

dat zij de meiden zouden halen, en nou had de vent
ze toch met die vervloekte klus meegegeven.—

—Saa’k f’rbrande! brulde Willem, daa’s ’n

judaàsstreek, t’met hep ie vaif moal d’r aige sait da
wài.. wài de maide hebbe daa’s kapsies soeke.…

—Die vint hep dur puur gain koorakter.…

—Nou, hoonde Dirk terug, de meisjes naar zich

toesleurend met twee armen tegelijk, wâ motte de
kooters mit sukke koale jasse van doen?

Beefwoedend klonk z’n stem in ’t duister.

—Daa’s net, jai hep ’t op je ruiker, ironiseerde Rink, de

reuzige polderkerel, ’n kittelstreekje onder Geert’s kin
smerend, dat ze schrok.

—Hande thuis, snauwde Geert gemaakt, zich erg in

de knel voelend, nu de andere knapen op ’t donkere
pad daar dreigden.
 —De deftighaid komp! moak ruimte! hoonde Dirk
weer.

—Die heere binne d’r soo ellendig pienter, trampelde

Piet. [284]

Klaas Koome, nù vriend van den guldenszwaren

Hassel, hoonde stekelig en driftig terug, schold op
Kees den „dief”, den moordenaar en strooper,—
Willem gierde wat gemaakt mee bij elken gemeenen
uitflap van Hazewind; dikbullige Jan vloekte pruttelend
en Bolk dwergde onrustig achter de dreigkerels.
Willem drong weer òp.

—He maide.… wâ motte jullie mit ’n vint van wit hoar..

d’r uitfoere? bai sain in de luure? daa’s niks gedaan!
protekollig veur vaif en ’n broer die d’r vaif joar sete
hep!.. s’n aige waif half doodklopt, en s’n aiges
kooters loat f’r hongere!.. en s’n skoonfoar mit ’t mis
kittelt! daa’s de femillie!—Rink reuzigde vóór,
dreigend, valsche schaterlachen verproestend. Maar
de meisjes wouên geen ruzie, voelden zich veel te
lekker in de aanzwellende lol en ’t zanggerucht van de
kermis. Ze snakten er naar. De koppen bij elkaar
gedonkerd fluisterden ze ’n besluit.—

—Hoor rais manne.… zei plots kloek Geert, wai hebbe

d’r vast niks àn,.. aa’s jullie d’r aige skelde en kriefele..
Wai.. wille d’r vast mi de haire meegoan.. aa’t ’t in
frinskap lait! jullie mot d’r malkoar f’rstoan.—

—Daa’s net!

—Juistíg.…
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebookmass.com