algorithms
Article
A Multi-Objective Model and Algorithms of Aggregate
Production Planning of Multi-Product with Early and
Late Delivery
Lanfen Liu and Xinfeng Yang *
Citation: Liu, L.; Yang, X. A
Multi-Objective Model and
Algorithms of Aggregate Production
Planning of Multi-Product with Early
and Late Delivery. Algorithms 2022,
15, 182. https://doi.org/10.3390/
a15060182
Academic Editor: Massimiliano
Caramia
Received: 16 April 2022
Accepted: 23 May 2022
Published: 25 May 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Algorithms 2022, 15, 182. https://doi.org/10.3390/a15060182
School of Traffic & Transportation Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China;
liulanfen@mail.lzjtu.cn
* Correspondence: xinfengyang@mail.lzjtu.cn
Abstract: Due to the influence of insufficient production capacity or shortage of production materials,
production enterprises may produce products in advance or be backordered. In order to improve the
adaptability of enterprises and reduce production costs, the impacts of early delivery and delayed
delivery are analyzed, and the method to determine the loss threshold is put forward. Moreover,
the maximum allowable shortage of customers with different tardiness is calculated, and the cost of
delayed delivery and loss of sales is determined. Considering the production cost, raw material cost,
inventory cost, staff cost, stockout, and lost sales cost, an early/delay multi-objective optimization
model is developed for an aggregate production planning (APP) problem to minimize total production
costs and instability in the workforce. Three algorithms and three different hybrid strategies are
designed to solve the model. Finally, some test experiments are employed in order to validate the
performance of the proposed evaluation of the three algorithms. The results show that: The method
of determining the loss threshold can effectively reflect the double influence of customer satisfaction
with waiting time and shortage quantity. The definition of unit tardiness cost reflects the law that it
increases gradually with waiting time. The determination of the feasible range of product output and
the number of workers in the workforce can reduce the search scope of the algorithm and improve
the efficiency of the algorithm.
Keywords: aggregate production planning; multi objective; multi-product; late delivery; hybrid
algorithms; hybrid strategies
1. Introduction
Aggregate production planning (APP) is an operational activity that determines the
minimum cost, workforce, and production plans required to meet customer demands. APP
simultaneously establishes optimal production, inventory, and employment levels over
a given finite planning horizon to meet the total demand for all products that share the
same limited resources [1]. As the implementation of JIT (Just in Time) practice becomes
increasingly popular, each echelon in a supply chain tends to carry smaller inventories, and
thus the whole supply chain is made more vulnerable to lost sales and/or backorders [2].
Furthermore, with the development of the manufacturing industry, more dynamic char-
acteristics of production need to be considered in practical application. Therefore, the
APP model should include multiple factors related to production and inventory, such as
production cost, labor cost, inventory cost, etc. and should also consider some extension
conditions, such as backorder cost, overtime cost, and lost sales cost.
APP is one of the most critical areas of planning performed in the design of production
systems, and it has attracted considerable interest from both practitioners and academics [1].
Many researchers have considered the backordering decisions in the APP model. We
summarized the most important and related studies that consider backordering for APP in
Table 1.
https://www.mdpi.com/journal/algorithms
Algorithms 2022, 15, 182
Table 1. The related studies considering the backordering for APP.
Article Model Category Objective Function
Total production costs;
Wang and Multiple objective linear Carrying and backordering
Liang [3] programming model costs; Costs of changes in
labor levels
The chance of obtaining the
A fuzzy random APP
Ning et al. [4] profit more than the
model
predetermined profit
The summation of
machine, reconfiguration,
Mahdavi An integer mathematical inter-cell material handling,
et al. [5] programming model inventory holding,
backorder, worker hiring,
firing and salary costs
The production costs; the
Chakrabortty Multiple objective linear carrying and backordering
and Hasin [6] programming model cost; the rate of change in
labor levels
The total production costs;
the total carrying and
Sadeghi Fuzzy Grey Goal
backordering costs; the rate
et al. [7] Programming model
of changes in
workforce level
The total costs of machine
maintenance and overhead,
Saidi-Mehrab Integer linear system reconfiguration,
et al. [8] programming model backorder and inventory
holding, training and
salaryof workers
The total cost composed of
production, setup, raw
Mixed integer linear
Basis et al. [9] material supply, inventory
programming model
holding and
backorder penalty
The operational cost
(including backorder and
Modarres and Linear programming
inventory carrying costs);
Izadpanahi [10] model
the energy cost;
carbon emission
The total costs in terms of
inventory levels, labor
levels, overtime,
Hossain Mixed integer linear subcontracting and
et al. [11] programming model backordering levels, and
labor, machine, warehouse
capacity, incentive and
wastage cost
The costs of machine
breakdown and relocation,
Sakhaii Deterministic nonlinear operator training and
et al. [12] mathematical model hiring, inter-intra cell part
trip, and shortage
and inventory
2 of 24
Solving
Considerations
Approaches
Solution algorithm
Inventory levels; labor levels;
based on linear
machine capacity; warehouse
programming
space; the time value of money
problem
The market demand;
production cost;
subcontracting cost; inventory A hybrid
carrying cost; backorder cost; optimization
product capacity; sales algorithm
revenue; maximum labor level;
maximum capital level
The available time for workers;
Linearized using
capacity of machine; worker
some auxiliary
assignment; worker
variables
assignment
Inventory levels; labor levels;
overtime; subcontracting and
Multi-Objective
backordering levels; labor,
Genetic Algorithm
machine and warehouse
capacity
machine capacity and A goal
warehouse space; labor levels; programming
carrying inventory approach
Demand satisfaction;
Linearized using
machineavailability; machine
some auxiliary
time-capacity; available time of
variables
worker and training
Demand satisfaction; material
balance; inventory capacity; A rolling
the relationship between setup horizon-based
binary and approach
production quantity
Demand satisfaction; limits for
each product in each period The goal attainment
and total production; technique
energy consumption
The time varying demand,
unstable production capacity Genetic Algorithm
and work forces, inventory Optimization
control, wastage reduction, approach and Big
and proper incentive M method
for workforce
The inter-cell layout, machine
reliability and relocation,
Linearized
machine capacity and
operator assignment
Algorithms 2022, 15, 182 3 of 24

Table 1. Cont.

Article Model Category Objective Function
The profit by improving
learning and reducing the
Mehdizadeh Multi-objective
failure cost of the system;
et al. [13] optimization model
the costs associated with
repairs and deterioration
A framework based on a The total revenue; total
Jamalnia set of stochastic, production costs;
et al. [14] nonlinear, multi-objective utilization of production
optimization models resources and capacity
The total cost of machine
maintenance and overhead,
inter- and intra-cell
Xue and Non-linear mixed integer
material handling,
Offodile [15] programming model
inventory holding,
subcontracting, and
backordering
The total costs composed
of regular time labor costs,
overtime labor costs, hiring
Jang and A robust costs, layoff costs, and
Chung [16] optimization model product-related costs that
include producing, holding
inventory, backlogging,
and subcontracting costs
Solving
Considerations
Approaches
Subpopulation
The market demands; the genetic algorithm;
machine capacity; the weighted sum
limitation on the total quantity multi-objective
produced; the workforce levels genetic algorithm
of labor groups; and nondominated
inventory capacity sorting genetic
algorithm II
The multiple
criteria
The production capacity; the decision-making
product demand; workforce methods Additive
value function,
TOPSIS and VIKOR
The production-inventory
balance; production
consistency; the lower and
upper bounds for the
Linearized
production level; the capacity
limits; the machine balance;
the storage space limits; the
maximal backordering level
The workforce level; the
Bi-level
production capacity; the
particleswarm
production balance; overtime
optimization
labor limit

Throughout the review, multi-objective programming has been widely used in this
area, and most of the existing research on APP mainly consider the constraints on the
balance equation for production, inventory capacity, inventory and demand, production
capacity, and labor capacity. Moreover, the production cost, inventory and backorder level
are the other main objectives that have been taken into consideration. The main methods
applied to management science techniques for APP problems are: linear programming,
piecewise linear programming, nonlinear programming, etc. Table 2 shows some of the
characteristics of the existing models.
Furthermore, Florian et al. [17] analyzed the computational complexity of a class
of deterministic production planning problems with various types of cost functions and
proved that the single material problem under the constraints of linear cost function
and time-varying capacity and some special cases are NP-hard. Chen and Thizy [18]
also proved that the multi-item capacitated production planning problem is strongly NP-
hard. In addition, according to Ramezanian et al. [19] and Mehdizadeh et al. [13], APP
problems with multi-phase production are among strongly NP-hard issues. Moreover,
metaheuristics have proved to be efficient techniques for solving APP problems. Among
the metaheuristics, genetic algorithms (GAs) (Chakrabortty et al. [6], Hossain et al. [11],
Mehdizadeh et al. [13], Ramezanian et al. [19], Liu et al. [20]), and particle swarm optimiza-
tion (PSO) (Jang et al. [16], Wang et al. [21], Chakrabortty et al. [22]) have been used to deal
with APP models.
Algorithms 2022, 15, 182 4 of 24

Table 2. The characteristics of the existing models.

Issue Description
Be deterministic, and must be satisfied by product, inventory,
Product demand
or backorder
Strictly linear or piecewise linear in any given planning period (consist of
Production costs regular time and overtime production and costs of inventory
and backorders)
Inventory Be limited over the entire planning horizon
Inventory capacity, production capacity and labor capacity are
Capacity
mainly considered
Backorders May or may not be allowed
Multiple product In most APP models more than one product exists
Some important labor characteristics are considered in some APP models,
Labor characteristics
such as labor skills, labor training, labor productivity, and constant level
The costs associated with meeting a known demand, the total revenue of
Objective
the system or inventory and backorder level
The balance equation for production, inventory capacity, inventory and
Constraints
demand, pro-duction capacity, and labor capacity
Linear programming, piecewise linear programming, nonlinear
Model category
programming, etc.

As a major factor determining the production capacity of enterprises, labor costs

account for an increasing proportion. Therefore, decision makers pay great attention to the
impact of labor changes, and the stability of workers has become more important than ever.
Furthermore, the fluctuation of demand will lead to a change in enterprise employment.
So, the stability of employment is extremely important to the APP problem. Therefore,
in the objective function, the sum of the changes in the number of workers is used to
measure its stability. In reality, due to the decline of actual production capacity caused by
the change of workers, equipment maintenance or damage, holidays and sudden epidemic
situations, or due to the shortage of production materials, production enterprises may
produce products in advance or be backordered. Both of these situations will affect the cost
of the enterprise. In order to improve the adaptability of enterprises and reduce production
costs, it is necessary to properly arrange the production plan to improve the production
management level of the enterprise.
In considering the situation of production in advance or backorder, this paper an-
alyzes the impact of backorders and determines the cost of backorder and loss of sales.
Furthermore, a bi-objective optimization model for an APP problem of multi-product,
multi-stage with early and late delivery minimizes total production costs and instability in
the workforce, considering the balance equation for production, product demand, labor,
inventory and production capacity, and different worker types. In order to improve the
efficiency of solving large-scale multi-objective APP problems, a local search (LS) algorithm
is based on the minimum cost flow (MCF). Further, in order to improve the speed of solving
APP problems, hybrid strategies based on local search-based GA (LS-GA), hybrid genetic
algorithm-particle swarm optimization based on stages (HGA-PSO1) and multi-population
strategy (HGA-PSO2) are adopted to solve this model. Finally, the performances of the
proposed evaluation of the multi-objective algorithm are selected to compare and analyze
each algorithm.
The remainder of this paper is organized as follows: Section 2.1 introduces some
notations. Section 2.2 analyzes the impacts of production in advance and backorder. In
Sections 2.3 and 2.4, the objective functions and constraints are discussed, respectively.
In addition, the GA and the LS algorithm based on MCF are proposed in Sections 3.1
and 3.2, respectively, and a double-layered multi-objective particle swarm optimization
algorithm (DMOPSO) is designed in Section 3.3, and three hybrid strategies are adopted in
Section 3.4. The proposed algorithm is tested via some numerical examples in Section 4.
Finally, Section 5 provides concluding remarks.
Algorithms 2022, 15, 182 5 of 24

2. Bi-Objective Model for APP with Production in Advance and Backordering

On-time delivery (OTD) refers to the delivery of products that meet the quantity and
quality requirements to customers in the period of the promised delivery date, which
is a key metric to measure delivery performance [23]. OTD can ensure the reliability of
delivery and, most importantly, customer loyalty as well as improve the reputation of
enterprises and customer satisfaction. In reality, due to the decline of actual production
capacity caused by the change of workers, equipment maintenance or damage, holidays
and sudden epidemic situations, or due to the shortage of production materials, production
enterprises may produce in advance or be backordered. Production in advance will increase
the inventory cost, while backorders will affect customer satisfaction. Generally speaking,
for customers, the longer the delayed delivery time and the greater the quantity, the lower
the customer satisfaction. If enterprises cannot meet their customers’ expectations, then
they may find a supplier who can. This leads to a loss of sales, which greatly increases the
lost sales cost (LSC). For the multi-product and multi-stage production mode, production
in advance or backorder often occurs, which will directly or indirectly lead to an increase in
enterprise production cost and can irreparably damage the customer relationship. So, it is
necessary to properly arrange the production plan to improve the production management
level of the enterprise. Therefore, a bi-objective optimization model is established in
this section for the APP problem considering the impacts of production in advance and
backorder. These two objective functions are formulated in Section 2.3, and the constraints
are analyzed in Section 2.4 based on the mathematical notations defined in Section 2.1.

2.1. Notations
(1) Sets and indices
The sets and indices used in this paper are listed in Table 3.

Table 3. Sets and indices.

Sets/Indices Description
T Set of periods in planning
t Index of the production planning period, t ∈ T
I Set of product categories
i Index of the product category, i ∈ I
J Set of raw material categories
j Index of the raw material category, j ∈ J
K Set of the worker types
K Index of the worker type, k ∈ K

(2) Parameters
The parameters used in this paper are listed in Table 4.

Table 4. Parameters.

Parameters Description
PDit Demand of product i in period t (units)
PCi Unit production cost of product i
The maximum tolerant backorder quantity for product i in period t
PBit (t0 − t)
for the customer waiting time t0 − t;
The delivery quantity of the product i from period t delayed to
Bit (t0 − t)
period t0
cbi (t0 − t) The backorder cost per unit time and per unit quantity
C1t Total production cost of period t
cli The unit LSC of product i
C5t Total backorder cost or lost sales cost in period t
PRit The unit raw material cost of product i in period t
Algorithms 2022, 15, 182 6 of 24

Table 4. Cont.

Parameters Description
C2t Total raw material cost in period t
PWik Labor hours required for a unit product i of worker k
PNit The production capacity for product i in period t
CIit The inventory of product i in period t
CKi Inventory cost of product i
CNi The inventory capacity of product i
C3t Total inventory cost in period t
Rij The demand of raw material j for producing unit i
RM jt Total demand of raw material j in period t
RCjt The price of raw material j in period t
W Htk The worker number for k type in period t
W Ik The basic salary of k type worker in a planning period
W Ltk The number of laid-off workers for k type in period t
W HC Training cost for a new worker
WR Maximum regular labor hours in a period
WO Maximum overtime labor hours in a period
WRTtk Total regular labor hours for k type worker of period t
WOTtk Total overtime labor hours for k type worker of period t
WRCk Unit regular time labor cost for k type worker
WOCk Unit overtime labor cost for k type worker
C4t Total labor cost in period t

(3) Decision variables

The decision variables are listed in Table 5.

Table 5. Decision variables.

Variables Description
PPit Production quantity of i in period t
Wtk The number of k type workers employed in period t

2.2. Analysis of the Impact of Production in Advance and Stockout

(1) Production in advance
Manufacturing enterprises make use of the remaining production capacity to carry out
production in advance, which can make balanced use of the effective production capacity
of the enterprise in each period and avoid the impact caused by the change of workers,
equipment maintenance or damage, holidays, and other conditions [9]. However, produc-
tion in advance will increase the inventory cost of production, which can be expressed by
Equation (1).
T I
∑ ∑ CIit ·CKi (1)
t =1 i =1

(2) Stockout
Due to the decline of actual production capacity or the shortage of production ma-
terials, production enterprises may be backordered [2]. Stockout will lead to a decline
in customer satisfaction or loss of sales. Generally, the longer the replenishment time
and the greater the backorder quantity, the greater the possibility of a loss of sales [24].
Therefore, the stockout cost, which includes backorder cost (BC) and LSC, depends on the
replenishment time and the backorder quantity. In order to determine the stockout cost, it
is necessary to determine firstly the critical value of the proportion of backorder quantity
to the total demand for different replenishment times that customers can tolerate.
Algorithms 2022, 15, 182 7 of 24

The threshold for customer loss: Let the threshold for customer loss be the critical
value. Referring to the definition of exponential partial backlogging rate in the inventory
model [25], the threshold for customer loss is defined as Equation (2).
0
β ( t 0 − t ) = k 0 · e − k 1 ( t − t −1) (2)

where t0 − t is the customer waiting time (t0 > t), t represents the planning period of
customer demand, and t0 is the period of replenishment. k0 is the backordering intensity
coefficient, that is, the maximum backorder rate acceptable to the customer for one planned
period of delayed delivery. k1 is the waiting time resistance, and 0 < k0 < 1, k1 > 0.
Then, the maximum tolerant backorder quantity of customers for the customer waiting
time t0 − t according to the demand PDit of product i in the period t can be calculated by
Equation (3).
0
PBit (t0 − t) = PDit · β(t0 − t) = PDit ·k0 ·e−k1 (t −t−1) (3)
The backorder cost (BC): Stockout will lead to significant decreases in customer satis-
faction or loss of sales. Generally, the method of penalty cost is adopted to deal with the
BC, and the BC per unit time is fixed, and most of them do not consider the difference in
the time to replenishment. However, in fact, the longer the waiting time and the greater the
quantity of backorder, the worse the customer’s satisfaction. Therefore, the BC needs to
consider the two influencing factors jointly. It is assumed that the BC per unit time and per
unit quantity also increases linearly [2]. Then, the BC per unit time and per unit quantity,
cbi (t0 − t), can be expressed by Equation (4).
2
cbi (t0 − t) = f i + ai ·(t0 − t) + bi ·(t0 − t) (4)

where f i is the fixed cost part of product I unit BC, ai and bi are the cost rate and cost
increasing rate of unit BC, respectively.
If the stockout quantity of each period is less than or equal to the maximum tolerant
backorder quantity of customers, the BC in period t can be obtained by Equation (5).

I
C5t = ∑ Bit (t0 − t)·cbi (t0 − t), Bit (t0 − t) ≤ PBit (t0 − t) (5)
i =1

Lost sales cost (LSC): If the stockout quantity is greater than the maximum tolerant
backorder quantity of customers, the customer will no longer wait. The enterprise loses the
order of this part of the products, and the LSC of this part can be calculated according to
the lost sales volume. So, the LSC in period t can be expressed by Equation (6).

I
C5t = ∑ Bit (t0 − t)·cli , Bit (t0 − t) > PBit (t0 − t) (6)
i =1

2.3. Objective Functions

According to the literature [20], the total production cost and stability in the work-
force are mainly considered, where the total production cost mainly includes the product
production cost, raw material cost, product inventory cost, labor cost, BC, and LSC.
The total production cost in period t can be calculated by Equation (7).

I
C1t = ∑ PPit · PCi (7)
i =1

Total demand of raw material j and the total raw material cost in period t can be
calculated by Equations (8) and (9).

RM jt = R ji · PPit (8)
Algorithms 2022, 15, 182 8 of 24

I I J
C2t = ∑ PPit · PRit = ∑ ∑ PPit · Rij · RCjt (9)
i =1 i =1 j =1

Using Equation (10), the total inventory cost in period t can be calculated based on
Equation (1).
I
C3t = ∑ CIit ·CKi (10)
i =1

If the working hours required for production are less than the maximum labor hours
I
of regular time for k type worker, ∑ PPit · PWik ≤ Wtk ·WR, then the total labor hours for
i =1
I
regular time WRTtk = ∑ PPit · PWik and the total labor hours of overtime WOTtk = 0. Oth-
i =1
I
erwise, WRTtk = Wtk ·WR and WOTtk = ∑ PPit · PWik − Wtk ·WR. Thus, total labor hours
i =1
for k type worker of regular time and overtime can be written as Equations (11) and (12).
( )
I
WRTtk = min ∑ PPit · PWik , Wtk ·WR (11)
i =1
( )
I
WOTtk = max ∑ PPit · PWik − Wtk ·WR, 0 (12)
i =1

The total labor cost in period t can be calculated by Equation (13).

K
C4t = ∑ (W Htk ·W HC + Wtk ·W Ik + WRTtk ·WRCk + WOTtk ·WOCk ) (13)
k =1

The BC and LSC can be calculated according to Equations (5) and (6). Then, the first ob-
jective function that aims to minimize the total production cost is defined in Equation (14).

T
minZ1 = ∑ (C1t + C2t + C3t + C4t + C5t ) (14)
t =1

The diversity of products and fierce competition make the stability of the manufac-
turing industry more important than ever. Therefore, in order to improve the adaptability
of enterprises and reduce production costs, we need to consider the stability of work-
ers. Hence, the sum of changes in the number of workers is used to measure its stability
according to Equation (15).

T K
minZ2 = ∑ ∑ (W Htk + W Ltk ) (15)
t =1 k =1

where for period t, the hired k type worker number, W Htk = max{Wtk − Wtk−1 , 0}, and the
number of laid off k type workers, W Ltk = max{Wtk−1 − Wtk , 0}.

2.4. The Constraints

Since Bit (t0 − t) indicates the delivery quantity of product i from period t delayed to
period t0 , if there is a delivery quantity from the previous period delayed to period t0 , it is
equivalent to the increase in demand in period t0 . Conversely, if there is a delivery period
in period t, it is equivalent to a reduction in demand in period t. In the case of backorder
Algorithms 2022, 15, 182 9 of 24

and loss of sales, the production and inventory balance condition can be expressed by
Equation (16).

t −2 T
CIit = CIit−1 + PPit−1 − ∑ Bit−1 (t − 1 − t1 ) − PDit−1 + ∑ Bit−1 (t2 − t + 1) (16)
t1 =1 t2 = t

Equation (17) represents the workforce balance constraint.

Wtk = Wtk−1 + W Htk − W Ltk (17)

Constraint (18) and Constraint (19) express the inventory and production capacity
constraints, respectively.
0 ≤ CIit ≤ CNi (18)
0 ≤ PPit ≤ PNit (19)
The sum of the initial inventory level, the production volume and the delivery quantity
of the product in each period should be equal to or greater than the demand for the product.
Then, the demand constraint of the product in each period can be expressed as:

T
PPit + CIit +
0
∑ Bit (t0 − t) ≥ PDit (20)
t = t +1

Constraint (21) ensures that the production capacity of the workforce must meet the
labor hours required to produce the production quantity.

I
∑ PPit · PWik ≤ Wtk (WR + WO) (21)
i =1

For the same product in the same period, production in advance and backordering
or loss of sales cannot occur at the same time. That is, if the product inventory is greater
than zero, there will be no backordering or loss of sales. The constraint can be expressed
as follows:
T
CIit ·
0
∑ Bit (t0 − t) = 0 (22)
t = t +1

Then the bi-objective APP problem model with production in advance and partial
backordering can be established.

3. Hybrid Algorithms Design

To the complexity of this APP model, three hybrid algorithms based on the local search
are designed to improve the efficiency of the algorithm. A genetic algorithm (GA) has
broadly applicable stochastic search and optimization techniques, while a simple GA is
prone to premature and has slow convergence. The particle swarm optimization (PSO)
algorithm is a kind of swarm optimization algorithm with a fast search speed and is
easy to implement [26]. A local search (LS) algorithm can improve search ability in the
solution space.
In order to improve the efficiency of solving large-scale multi-objective APP problems,
the three algorithms can be effectively combined. In order to improve the efficiency of
solving large-scale multi-objective APP problems, based on the characteristics of GA, LS,
and PSO, this section designs three hybrid algorithms using different hybrid strategies.
In this section, the GA and the LS algorithm based on the augmented cycle algorithm
are designed to solute the APP problem model in Sections 3.1 and 3.2, respectively. In
Section 3.3, a DMOPSO algorithm is proposed, and, finally, three hybrid strategies are
introduced in Section 3.4.
Algorithms 2022, 15, 182 10 of 24

3.1. Genetic Algorithm

(1) Chromosome Representation
In this study, a chromosome CH contains two sub-chromosomes, P and W, which
represent the production quantity of a product and the number of workers, respectively.
Algorithms 2022, 15, x FOR PEER REVIEW 11 of 25
The genes in each chromosome are integer. Figure 1 is a simple example of a chromosome
with I product categories and T periods.

Period (t) 1 2 ... T

Category 1 PP11 PP12 ... PP1T
P Category 2 PP21 PP22 ... PP2T
... ...
Category I PPI 1 PPI 2 ... PPIT

Worker Type 1 W11 W21 ... WT 1

Worker Type 2 W12 W22 ... WT 2
W ... ...
Worker Type k W1k W2k ... WTk

Figure
Figure1.1.An
Anexample
exampleofofa achromosome.
chromosome.

(2)
(2) Initialization
Initialization
Analysis
Analysisofofthe thefeasible
feasible rangerange PPitPP
of of : For the same product in the same period, pro-
it : For the same product in the same period,
duction
productionin advance and backordering
in advance and backordering or lossorofloss sales ofcannotsales cannotoccur at occurthe same at thetime.
sameIftime.
the
total amount
If the of demand
total amount in the current
of demand in the period
currentand period the quantity of backorders
and the quantity in the pre-in
of backorders
vious period isperiod
the previous greater is than
greaterthethan sumthe of production
sum of production capacitycapacity and inventory at the begin-
and inventory at the
ning of the current period, stockout occurs. Therefore,
beginning of the current period, stockout occurs. Therefore, in this case, the product in this case, the product should be
should
produced
be produced withwith
the the
maximum
maximum production
production capacity
capacity due due to to
thethe existence
existence ofofstockout
stockoutcost.
cost.
Otherwise,
Otherwise,the theminimum
minimum PP PPitit should
shouldmeet meetthe thesum sumofofthe the product
product demand
demand in inthethe plan-
planning
period
ning and and
period the quantity
the quantity of backorder
of backorder in the in previous
the previous period
period according
according to Constraint
to Constraint(20).
(20). Thus, the minimumit PPit in the planning period can be obtained by
Thus, the minimum PP in the planning period can be obtained by

  
( ( ))
t−t −11 
MinPPMinPP
it = max it = max 0, min
0, min PD PD

it + it +∑  Bitit −−1 (1t(−
B t−t2 )t−
2 )CI , PN
−it CI  it
it ,itPN
 
(23)
(23)
 t2 =11
t2 =

Moreover, the maximum production quantity should not exceed the production ca-
Moreover, the maximum production quantity should not exceed the production capac-
pacity and
ity and inventory
inventory capacity
capacity according
according to Constraints
to Constraints (18)(18)
and and (19).
(19).
(   t −1  )
MaxPPit = min CNi + PDit +t− 1 Bit −1 (t − t2 ) − CI it , PN it  (24)
MaxPPit = min CN i + PDit + ∑ t2 =1 B
it−1 ( t − t2 ) − CIit , PN  it (24)
t =1
Analysis of the feasible range of Wtk : The2minimum and maximum Wtk can be ob-
tainedAnalysis
according to the
of the production
feasible range ofquantity
Wtk : Theand the labor
minimum andcost and stability
maximum Wtk canin be
theobtained
work-
force (   indicates
according roundingquantity
to the production up). and the labor cost and stability in the workforce (d e
indicates rounding up).
 I 
MinWtk =&  (  PPit ⋅ PWik ) / (WR + WO )  ' (25)
 Ii =1 
MinWtk = ( ∑ PPit · PWik )/(WR + WO) (25)
i =1  I

MaxWtk = max ( Wtk −1 ,&( PPit ⋅ PWik ) / WR  ')
(26)
  i =1I 
MaxWtk = max Wtk−1 , ( ∑ PPit · PWik )/WR (26)
Obviously, the chromosomes satisfy Constraints
i =1
(18)–(22) in the process of initializa-
tion.
Obviously, the
Calculation chromosomes
of the satisfy Constraints
stockout quantity: If the total (18)–(22)
amountinofthedemand
process of
in initialization.
the current
Calculation of the stockout quantity: If the total amount of demand in the current period
period and the quantity of backorders in the previous period is greater than the sum of
and the quantity of backorders in the previous period is greater than the sum of production
production capacity and inventory at the beginning of the current period, the products
will be out of stock; otherwise, the products will be produced in advance. So, the stockout
quantity can be calculated by

{ t −1
max PDit +  Bit −1 (t − t ") −CI it − PPit ,0
t " =1
} (27)
Algorithms 2022, 15, 182 11 of 24

capacity and inventory at the beginning of the current period, the products will be out of
stock; otherwise, the products will be produced in advance. So, the stockout quantity can
be calculated by ( )
Algorithms 2022, 15, x FOR PEER REVIEW t −1 12 of 25
max PDit + ∑
00
Bit−1 (t − t00 )−CIit − PPit , 0 (27)
t =1

(3) The
Genetic Operators
genetic operators in this paper are almost the same as that of the genetic algo-
rithmThein [20], mainly
genetic including
operators the paper
in this partheno
are crossover
almost the operator
same as(only
thatutilizable for sub-
of the genetic al-
chromosome
gorithm in [20], P ),mainly
arithmetic crossover
including the operator,
parthenoproduction mutation (only
crossover operator operator, mutation
utilizable for
sub-chromosome
operator P), arithmetic
for the number crossover
of workers, operator, infeasible
and repairing production mutation
gene operator,
operators, whichmuta-
will
tionbe
not operator for the
introduced number
in detail of workers, and repairing infeasible gene operators, which
here.
will not
ThebeGA introduced in detail here.
has the characteristics of random multi-point search and implicit parallelism,
whileThe GA has
a simple the characteristics
genetic algorithm is of random
prone multi-point
to being prematuresearch
andand
hasimplicit parallelism,
slow convergence.
while
As a simple
shown genetic
in Figure algorithm
2, the is prone toisbeing
same experiment premature
calculated 10 timesandusing
has slow convergence.
the designed GA,
As shown
and in Figure
it can be seen that 2, the same experiment
sometimes is calculated
the algorithm 10 times
will fall into using
the local the designed
optimal solution.GA,
and it can be seen that sometimes the algorithm will fall into the local optimal solution.
45

40
1
Z

35
5.02 5.04 5.06 5.08 5.1 5.12 5.14
Z 5
2 x 10

Figure 2. Running results of GA for 10 times.

3.2. Local
3.2. Local Search
Search Algorithm
Algorithm
This LS
This LS algorithm
algorithm searches
searches thethe neighborhood
neighborhood from
from aa solution
solution of
of production
production quantity
quantity
under the
under the condition
condition that that the
the number
number of of workers
workers remains
remainsunchanged
unchangedto improve Z
toimprove Z11 .. The
The
problem of optimization of production quantity is a special case of the MCF problem,
problem of optimization of production quantity is a special case of the MCF problem, ac-
according to [20]. Thus, the LS algorithm for production quantity can be designed based
cording to [20]. Thus, the LS algorithm for production quantity can be designed based on
on augmenting cycle.
augmenting cycle.
First, production planning can be dealt with as a network model. Any two unequal
First, production planning can be dealt with as a network model. Any two unequal
periods t1 , t2 ∈ [1, T ] can form a cycle. The maximum adjustment production quantity of
periods
i is limited 2 ∈the
t1 , tby [1, T minimum
] can formvalue
a cycle. Theproduction
of the maximum adjustment
quantity of production quantity
period t1 , the inventory of i
is limitedfrom
capacity by the minimum
period t1 to tvalue
2 , andof theproduction
the productionquantity periodt2t.1 The
quantityofofperiod , the maximum
inventory
adjustment
capacity from amount
periodcan t1 be
to determined according to
t2 , and the production the limit
quantity period t2analysis
of quantity of three
. The maximum
aspects for the formed anticlockwise circle and clockwise circle, respectively.
adjustment amount can be determined according to the limit quantity analysis of three
(1) The
aspects forformed
the formedanticlockwise cycle circle and clockwise circle, respectively.
anticlockwise
(1) The production quantity ofcycle
formed anticlockwise period t1 : The maximum production quantity of the period
t1 will limit the increased flow, which can be expressed as MaxPPit1 − PPit1 .
The production quantity of period t1 : The maximum production quantity of the pe-
Inventory remaining capacity: The minimum remaining inventory capacity from
riod
periodt1 twill limit the increased flow, which can be expressed as MaxPPit1 − PPit1 .
1 to t2 is min{CNi − CIit | t1 < t ≤ t2 }.
The production
Inventory remaining quantity of period
capacity: t2 : If there
The minimum is no stockout
remaining inventory from t1 tope-
periodfrom
capacity t2 ,
riod t1 to t2 is min {CN i − CI it t1 < t ≤ t22 } .
the adjustment amount of the period t limit is PP it2 − MinPP it2 (as shown in Figure 3a).
Otherwise, the adjustment amount can be increased to the total stockout quantity from
The production t2 quantity
T of period t2 : If there is no stockout from period t1 to t2 ,
period t1 to t2 , ∑ ∑ Bit00 −1 (t0 − t00 + 1) (as shown in Figure 3b).
the adjustmenttamount oft00 the period t2 limit is PPit2 − MinPPit2 (as shown in Figure 3a).
00 =t +1 t0 =
1

Otherwise, the adjustment amount can be increased to the total stockout quantity from
t2 T
period t1 to t2 ,  B
t "= t1 +1 t ' = t "
it " −1 (t '− t "+ 1) (as shown in Figure 3b).

The maximum adjustment amount of anticlockwise cycle for product i can be ob-
tained as follows:
 t2 T
FP
Algorithms 2022, 15, x FOR = min
PEER MaxPP
REVIEW
i { it1 − PPit1 , CN i − CI it , PPit2 − MinPPit2 +  B it " −1 (t '− t "+ 1) t1 < t ≤ t2 }. 13 of 25
(28)
t "= t1 +1 t ' = t "
Algorithms 2022, 15, 182 12 of 24

t2 T

{
FPi = mino MaxPPit1 − PPit1 , CN i − CI it , PPit2 − MinPPit2 +  B it " −1 (t '− t "+o1) t1 < t ≤ t2 }. (28)
t "= t1 +1 t ' = t "

θ θ

PPit 2

PP
i
i
it

+
+

2
−θ
PP
it1
−θ i
PP
1
it
i
θi θi
o CIit1 CI it + θi o CI it2
CI it1 CIit2

vt1 θ CI it + θi vt2 vt1 θ

t2

 B
T
−itθi vt2
PPit 2

PP
i
it " −1
i

+
+

t "= t1 +1 t ' = t "

2
−θ
PD

PP
it1
−θ i
PP
1
it

i
θi θi
PD it1

PD
it 2
it2
PD

it 1
CIit1 CI it + θi CI it2
CI it1 CIit2

vt1 d
CI + θ vt2 vt1 dB

t2 T

it " −1 − θi vt2
it i
t "= t1 +1 t ' = t "
PD
(a) No stockout from period t1 to t2 (b) Stockout from period PDt1 to t2
PD it1

it 2
it2
PD

it 1
Figure 3.
Figure Illustration of
3. Illustration of augmenting
augmenting flow
flow in
in anticlockwise
anticlockwise cycle.
cycle.
d d
The formed
(2) period
The maximum adjustment amount of anticlockwise cycle for product i can be obtained
(a) No stockout from t1 to t2 clockwise cycle (b) Stockout from period t1 to t2
as follows:
The production quantity of period t1 : If there is no stockout from period t1 to t2 ,
Figure 3. Illustration of augmenting flow int2anticlockwise T is PPcycle.
the adjustment amount of the period t1 limit it − MinPPit (as shown in Figure 4a).
− PPit1 , CNi − CIit , PPit2 − MinPPit2 + ∑ ∑ Bit00 −11 (t0 − t00 +1 1) |t1 < t ≤ t2 }.

FPi = min MaxPPit1 (28)
Otherwise,
(2) The formedthe adjustment amount tcan
clockwise cycle 00 =t be increased
0 00 to the minimum stockout quantity
1 +1 t = t
fromThe min {Bit "of
period t1 to t2 :quantity '− t "+ 1)tt1 :≤Ift "there
−1 (tperiod < t2 , tis
" ≤no T } (as shown
t ' ≤stockout t1 4b).
in Figure to t2 ,
(2) Theproduction
formed clockwise cycle 1 from period
the adjustment
The productionamount of the period
quantity of period t1 tlimit is PPit − MinPPit1 (as shown in Figure
t1 to4a).
t2 ,
1 : If there 1is no stockout o
from period
o
Otherwise,
the adjustmentthe adjustment
amount of theamount period cant1 belimitincreased
is PPit1 − toMinPP
the minimum stockout
it1 (as shown quantity
in Figure 4a).
periodthet1 adjustment
Otherwise, {B0 it "−1 (00t '−can
to t2 : minamount t "+be t1 ≤ t " < t2 ,totθ "the
1)increased ≤ T } (as shown
≤ t 'minimum
PP
stockout in quantity from
PP

θ from 2
+
Figure 4b).
i

i
it

it 2
0
−
−

+θ

period t1 to t2 : min{ Bit00 −1 (t − t + 1)|t1 ≤ t < t2 , t ≤ t ≤ T } (as shown in Figure 4b).

00 00

θi
θi θi
PP
i1t
PP
i1t

CIit1 o CIit2 CIit1 CI it − θi o CIit2

v t1 CIit − θi v t2 v t1 Bit "−1 − θPi v t2

P
PP

θ 2 θ +
i

i
it

it 2
−
−

+θ
PD it1

PD it 1

θi
θi θi
PP
i1t

2
PP
i1t

2
i

PD
it
PD
it

CIit1 CIit2 CIit1 CI it − θi CIit2

v t1 CI d− θ it i
v t2 v t1 d Bit "−1 − θi v t2
PD it1

(a) No stockout from period t1 to t2 (b) Stockout from period t1 to t2

PD it 1

2 2
PD
it
PD
it

Figure 4. Illustration of augmenting flow in clockwise cycle.

d d
Inventory
(a) No stockout from period t1 to remaining
t2 capacity: The minimum remaining
(b) Stockout inventory
from period t1 tocapacity
t2 from pe-
riod t1 to t2 is min {CI it CI it > 0, t1 < t ≤ t2 } .
Figure
Figure 4. Illustration
Illustration of augmenting
ofquantity
augmenting flow in clockwise cycle.
The production of period t : The maximum production quantity of the pe-
2

t2 will limit
riod Inventory
Inventory the increased
remaining
remaining flow,
capacity:
capacity: Thewhich
The can beremaining
minimum
minimum expressed
remaining as inventory
inventory it2 −capacity
MaxPPcapacity PPit2 .fromfrom
pe-
riod The
periodt1 tomaximum
t 1 to t 2 is min {{CI it CI it > amount
t2 is minadjustment
CI it | CIit > 0, t1 < t ≤ of
0, t 1 < tt2 }clockwise
≤ . t 2 } . cycle for product i can be obtained
by The production quantity of period t2 : The maximum production quantity of the period
The production quantity periodcant2 be
flow,ofwhich : The maximum production quantity of the pe-
t2 will limit the increased expressed as MaxPP it2 − PPit2 .
t2 will
riod The FP
limit the
maximum = min MaxPP
i increasedamount
adjustment flow,
it2 − PP
which
of , {
CI can
it2 clockwise
it , B itbe ( t ' − t
" −1 expressed "+ 1), as MaxPP
cycle for product i canit2 −be
PPobtained
it2 . by
(29)
The maximum adjustment PPit − MinPP
FPi = min MaxPP 1 it1 + min{
amount it " −1 (t ' − t "+01)}|
ofBclockwise
it2 − PPit2 , CIit , Bit00 −1 ( t − t + 1),
cycle CI > 0,product
for
00 it
t1 < t ≤ ti2 can
. be obtained }
by (29)
PPit1 − MinPPit1 + min{ Bit00 −1 (t0 − t00 + 1)}CIit > 0, t1 < t ≤ t2 .

{
FPi = min MaxPPit2 − PPit2 , CI it , Bit "−1 (t '− t "+ 1),
LS algorithm can significantly improve the quality of solution, especially in large-scale (29)
experiments, as shownPP − MinPP
init1 Figure + min{Bit "−the
5.it1However, 1 (t ' − t "+ 1)}| time
running CI it >will < t greatly
0, t1 be ≤ t2 . increased for }
each experiment.
 Algorithms 2022, 15, x FOR PEER REVIEW 14 of 25

LS algorithm can significantly improve the quality of solution, especially in large-

Algorithms 2022, 15, 182 13 of 24
scale experiments, as shown in Figure 5. However, the running time will be greatly in-
creased for each experiment.

30 30 46

29
No LS No LS 44 No LS
25 28 With LS
With LS With LS
27 42
20
26
40

Z2
Z2
Z2

25
15
24 38

23
10 36
22

5 21 34
8.8 8.85 8.9 8.95 9 9.05 9.1 9.15 9.2 2.64 2.65 2.66 2.67 2.68 2.69 2.7 5 5.02 5.04 5.06 5.08 5.1 5.12 5.14
Z1 4 Z1 5 Z1 5
x 10 x 10 x 10

(a) Experiment 1 (I = 2, T = 4, J = 3) (b) Experiment 2 (I = 4, T = 6, J = 3) (c) Experiment 3 (I = 6, T = 8, J = 3)

Figure5.5.Effect
Figure Effectanalysis
analysisof
ofLS
LSalgorithm.
algorithm.

3.3.Double-Layered
3.3. Double-LayeredMulti-Objective
Multi-Objective Particle
Particle Swarm
Swarm Optimization Algorithm
Inthe
In thePSO
PSOmodel,
model,each
each particle
particle needs
needs toto continuously
continuously update
update the
the velocity
velocity and
andthe
the
positionaccording
position according to the
the Pbest
Pbest(the
(thebest
bestposition
positionof of
a particle) andand
a particle) the the
Gbest (the (the
Gbest best best
po-
sition for
position forall
allparticles).
particles).For
Forthe
thesingle-objective
single-objectiveproblem,
problem,these
thesetwo
twosolutions
solutionsare areeasy
easyto
to
choose,but
choose, butfor
forthe
themulti-objective
multi-objective problem,
problem, it
it is difficult to choose the
the local
local optimal
optimalguide
guide
Pbestand
Pbest andthe
theglobal
globaloptimal
optimalguide
guide Gbest.
Gbest. How to choose the PbestPbest and
and thethe Gbest
Gbestto toguide
guide
themoving
the movingof ofeach
eachparticle
particle in
in the
the search
search space
space is a critical issue, and
and itit is
is very
very important
important
forthe
for thepremature
prematureconvergence
convergenceof ofthe
thePSO
PSOalgorithm
algorithm andand the
the diversity of solutions [27].
(1) Generation
(1) Generation and and Update
Update of of the
the Global
Global Optimal
Optimal Guide
Guide
Choosingananappropriate
Choosing appropriate global
global optimal
optimal solution,
solution, Gbest, Gbest, to guide
to guide particles
particles will
will greatly
greatly improve the quality of Pareto solutions and maintain
improve the quality of Pareto solutions and maintain the diversity of nondominated the diversity of nondomi-
nated solutions.
solutions. Firstly,Firstly, this paper
this paper establishes
establishes an external
an external archive,
archive, whichwhich mainly
mainly records
records the
the nondominated solutions found so far. At each iteration, the external
nondominated solutions found so far. At each iteration, the external archive is updated to archive is updated
to ensure
ensure thatthat
the the external
external archive
archive is only
is only nondominated
nondominated solutions.
solutions.
Selection mechanism of Gbest: The following selection mechanism
Selection mechanism of Gbest: The following selection mechanismof ofGbest
Gbestisisestab-
estab-
lished based on the external archive: If the number of nondominated solutions in ex-
lished based on the external archive: If the number of nondominated solutions in the the
ternal archive
external archiveisislesslessthan
thanoror
equal
equalto to
thethe population
population sizesize
of the PSOPSO
of the algorithm,
algorithm,popsize, all
popsize,
the solutions in the external archive are selected as the global optimal
all the solutions in the external archive are selected as the global optimal guides Gbest. Oth- guides Gbest. Oth-
erwise,the
erwise, thecrowding
crowdingdistance-based
distance-based selection
selection algorithm
algorithm is is used
used toto select
select popsize
popsizesolutions
solutions
from the external archive as the global optimal
from the external archive as the global optimal guides Gbest. guides Gbest.
Gbestassignment
Gbest assignmentmechanism
mechanismfor foreach
each particle:
particle: LetLet NDnum
NDnum be be
thethe number
number of solu-
of solutions
in Gbest, and the mechanism for assigning a global optimal guide to each particle isis
tions in Gbest, and the mechanism for assigning a global optimal guide to each particle
asasfollows:
follows:
Step1.1.Calculate
Step Calculate the the distance
distance between
between individual
individual particles
particles and
andthethesolutions
solutionsininGbest.
Gbest.
Step 2. Assign a guide to each particle. If NDnum=popsize, select a guide
Step 2. Assign a guide to each particle. If NDnum=popsize, select a guide with
with the
the
smallest distance
smallest distance for for each
eachparticle
particleasasthetheindividual
individual global optimal
global guideguide
optimal (the global opti-
(the global
mal guides
optimal guides forfor
individual
individual particles
particlescannot
cannotbeberepeated).
repeated).Otherwise,
Otherwise,eacheachglobal
globaloptimal
optimal
guide is assigned to popsize/NDnum particles with the smallest distance as their
guide is assigned to popsize/NDnum particles with the smallest distance as their global
global
optimalguides.
optimal guides.
Figure66shows
Figure showsthe theGbest
Gbestassignment
assignment mechanism
mechanism for for each
each particle
particle ofof the
the bi-objective
bi-objective
optimal problem.
optimal problem.
Algorithms 2022, 15, x FOR PEER REVIEW 15 of 25

Algorithms 2022, 15, x FOR PEER REVIEW 15 of 25

Algorithms 2022, 15, 182 14 of 24
Z2

Z2

Z1

Figure 6. Illustration of global optimal guide assignment for individual particle.

Z1
(2) Selection Mechanism of the Local Optimal Guide Pbest
Figure 6.
Figure Illustration of
6. Illustration of global
global optimal
optimal guide
guide assignment
assignment for for individual
individual particle.
particle.
During the evolution process of the PSO algorithm, the local optimal guide Pbest is
selected
(2) for each
Selection particle according
Mechanism ofthe
theLocal
Local to Optimal
the following Guidemethod:
Pbest Calculate the distance be-
(2) Selection Mechanism of Optimal Guide Pbest
tween each solution in the Pbest set and the global optimal guide corresponding to the
During
During the the evolution
evolutionprocess
processofofthe thePSO PSOalgorithm,
algorithm,the local
the localoptimal
optimal guide
guidePbest is
Pbest
individual in Gbest, and select the individual optimal location with the smallest distance
selected
is selectedforforeach particle
each according
particle according to the
to thefollowing method:
following Calculate
method: the distance
Calculate the distancebe-
as the local optimal guide. Figure 7 visualizes this process. g1 is the global optimal guide
tween
between each eachsolution
solutionin in
the
thePbest
Pbestset
setand
andthe theglobal
globaloptimal
optimalguide
guidecorresponding
corresponding to to the
the
assigned
individualtoin
individual inan individual
Gbest,
Gbest, and particle
and select
select the x1 . p1 , p2optimal
the individual
individual , p3 arelocation
optimal the solutions
location with in smallest
with the
the smallest for x1 .
the Pbestdistance
distance
as the
When
as thelocal
local
selectingoptimal
optimal guide.
the guide.
local Figureguide
optimal
Figure for x1this
77visualizes
visualizes pprocess.
,this process. gg11 is
1 is selected
the global
because optimalisguide
its position guide
clos-
assigned g to an individual particle x . p , p , p are the solutions in the
est to 1 .to an individual particle x1 . p1 , p2 , p3 are the solutions in the Pbest for x1 .
1 1 2 3 Pbest for x 1 . When
assigned
selecting the local optimal guide for x1 , p1 is selected because its position is closest to g1 .
When selecting the local optimal guide for x1 , p1 is selected because its position is clos-
Z 2 to g1 .
est x1

x1
Z2

p1
p2 p3
g1
p1
p2 p3
g1

Z1

Figure 7. Illustration of the selection mechanism of local optimal guide.

Z1
(3) Updating Rules
(3) Updating Rules of of Particle
Particle Velocity
Velocity and and Position
Position
Figure 7. Illustration of the selection mechanism of local optimal guide.
In the standard
In the standard form formof ofPSO,
PSO,the thevelocity
velocityvector vectorfor forparticle
particle k isk is updated
updated according
according to
to three other vectors that are the local optimal guide
three other vectors that are the local optimal guide of the kth particle (Pbest), of the kth particle (Pbest), the current
(3) Updating Rules of Particle Velocity and Position
global optimal
optimal guide
guide(Gbest),
(Gbest),andandthe thecurrent
current position
position of ofthethe particle
particle [28].[28]. In every
In every itera-
iteration,
In the standard form of PSO, the velocity vector for particle k is updated according
tion,
each each particle
particle updates updates its velocity
its velocity and andpositionposition according
according to theto the assigned
assigned globalglobal opti-
optimal
to three other vectors that are the local optimal guide of the kth particle (Pbest), the current
mal
guideguide
and and
locallocal
optimaloptimal
guide guide
according according to thetofollowing
the following rules:rules:
global optimal guide (Gbest), and the current position of the particle [28]. In every itera-
g + 1)h= χits
vk1 (updates ωvelocity
1
k vk ( g ) + cand 1r1 ( pb
1 1
k ( g ) − xkaccording
( g )) + c2 r2 (to gbk1the − xk1 ( g )) global
( g )assigned
v1k ( g + 1) = χ ωk v1k ( g) + c1 r1 ( pbk1 ( g) − xk1 ( g)) + c2 r2 ( gbk1 ( g) − xk1 ( g))
tion, each particle position i (30)
opti-
(30)
mal guide and local optimal guide according to the following rules:
vk1 ( g + 1) h= χ ωk v1k ( g ) + c1r1 ( pb1k ( g ) − x1k ( g )) + c2 r2 ( gb1k ( g ) − x1k ( g )) i
2 2 2 2 2 2 (31)
v ( g + 1) = χ ω v ( g ) + c r ( pb ( g ) − x ( g )) + c r ( gbk (2g ) − xk ( g2))
v2k ( g +k1) = χ ωk v 2k (k gk) + c1 r11 (1pbk2 (k g) − xk2k ( g)) + 2c22r2 ( gb k ( g ) − x k ( g )) (30)
(31)
x 1
( g + 1) = x1
( g ) +  v 1
( g + 1)  (32)
vk2 ( g + 1) = χ ωxk1v(k2g( g+ 1 k ( g ) − xk 1( g )) + c2r2 ( gbk ( g ) − xk ( g )) 
2k l2 k m 2 2
k
) +1)c1=
r1 ( xpb 
(31)
k k ( g ) + v k ( g + 1) (32)

xk2 ( gx+
1
k (g 1) =x2x(1kg()g+
1)+ = ) + vv21k (( gg +
+ 1)1) (32)
l m
k k (33)
 Algorithms 2022, 15, x FOR PEER REVIEW 16 of 25

xk2 ( g + 1) = xk2 ( g ) + vk2 ( g + 1)  (33)

Algorithms 2022, 15, 182 15 of 24
where k is a particle, χ is the constriction factor, and ωk is the inertia weight. c1 , c2
are respectively learning factors. r1 , r2 are two random parameters within [0, 1]. g is
1
the number
where of iterations,
k is a particle, χ is the g ) and vk2 ( gfactor,
v ( constriction
k ) represent
and ωkthe velocity
is the of particle
inertia weight. kcin the
1, c PP
2 are
respectively learning factors. r , r are two random 1 parameters
2 within
1 [0,
layer and W layer of iteration g, respectively. pbk ( g ) , pbk ( g ) , gbk ( g ) and gbk ( g )
1 2 1]. g is 2 the

number of iterations, v1k ( g) and v2k ( g) represent the velocity of particle k in the PP layer
represent the local optimal guide and global optimal guide of particle k in the PP layer
and W layer of iteration g, respectively. pb1 k1 ( g), pbk2 ( g)2, gbk1 ( g) and gbk2 ( g) represent the
and W layer of iteration g, respectively. xk ( g ) and xk ( g ) are the position of particle k.
local optimal guide and global optimal guide of particle k in the PP layer and W layer of
Generally,
iteration a large inertial
g, respectively. xk1 ( g) andweight
xk2 ( g)facilitates the global
are the position search,k.while a small inertial
of particle
weight facilitates
Generally, the inertial
a large local search.
weightTherefore,
facilitates athe
linearly decreasing
global search, whileinertial
a smallweight
inertialis
adopted to adjust ω k .
weight facilitates the local search. Therefore, a linearly decreasing inertial weight is adopted
to adjust ωk . g (ω − ωmin )
ω ( g ) = ω g−(ωmaxmax− ωmin ) (34)
ωk ( g) k= ωmax max − G (34)
G
whereGGisisthe
where themaximum
maximumiteration number,ωω
iterationnumber, maxand
max
and ωmin
ωmin are predefined
are the the predefined maxi-
maximum
mum
and and minimum
minimum inertia inertia
weight,weight, respectively.
respectively.

3.4.
3.4.Hybrid
HybridStrategies
Strategies
In
In ordertotoimprove
order improvethethe
efficiency of solving
efficiency large-scale
of solving multi-objective
large-scale APPAPP
multi-objective problems,
prob-
based
lems, on the on
based characteristics of GA, of
the characteristics LS,GA,
andLS,
PSO,
andthis
PSO,section designsdesigns
this section three different hybrid
three different
strategies.
hybrid strategies.
(1)
(1) Hybrid
Hybrid Strategy
Strategy of
of LS-GA
LS-GA
Theidea
The ideaofofLS-GA
LS-GAhybrid
hybrid strategy
strategy is to
is to useuse
thethe local
local search
search ability
ability of LSofto
LSimprove
to improve
the
the efficiency
efficiency ofalgorithm.
of the the algorithm.
The The
methodmethod
is toisadd
to add ansearch
an LS LS search algorithm
algorithm in evolution
in the the evolu-
tion process
process of eachof offspring.
each offspring. The chart
The flow flow chart
of theof the LS-GA
LS-GA Strategy
Strategy is shown
is shown in Figure
in Figure 8. 8.

Initialize Genetic Algorithm

Start
population
Non-dominated sorting and
crowding distance-based fitness
assignment strategies (NSGA-II)

Crossover
No

Mutation

Yes Stopping
Stop criteria Local Search
met?

Figure8.8.The
Figure Theflow
flowchart
chartof
ofthe
thehybrid
hybridstrategy
strategyfor
forLS-GA.
LS-GA.

(2)
(2) Hybrid
Hybrid Strategy
Strategy of
of HGA-PSO1
HGA-PSO1
Thehybrid
The hybridgenetic
geneticalgorithm-particle
algorithm-particle swarm swarm optimization
optimization based
based onon stages
stages (HGA-
(HGA-
PSO1)combines
PSO1) combinesthe thePSO
PSOalgorithm
algorithmand andLS-GA
LS-GAalgorithm,
algorithm,considering
consideringthe thecharacteristics
characteristics
ofthe
of thePSO
PSOalgorithm,
algorithm,which
whichhas has fast
fast convergence
convergence speed
speed butbut is prone
is prone to premature
to premature con-
conver-
vergence, and the LS-GA algorithm has strong local
gence, and the LS-GA algorithm has strong local search ability. search ability.
Oneof
One ofthe
themethods
methodsof ofthis
thisstrategy
strategyis: is: Firstly,
Firstly,the
theDMOPSO
DMOPSOalgorithm
algorithm isisused
usedfor
for
globalsearch.
global search.When
Whenititfalls
fallsinto
intothe
thelocal
local optimal
optimal solution
solution (by(by setting
setting a certain
a certain evolution-
evolutionary
algebra), thenthen
ary algebra), the LS-GA
the LS-GAalgorithm
algorithm is used forfor
is used local search
local search based
basedononthetheglobal
globaloptimal
optimal
solution.
solution.The
Theflow
flowchart
chartofofthe
thehybrid
hybridstrategy
strategyforforHGA-PSO1
HGA-PSO1isisshown shownininFigure
Figure9.9.
Algorithms 2022,
Algorithms2022,
Algorithms 15,
2022,15, xxFOR
15,182 FORPEER
PEERREVIEW
REVIEW 17 of
17of
16 25
of24
25

Initialize
Initialize External
External
Start
Start population
population archive
archive

Fitness
Fitnessassignment
assignment
and
andselection
selection

Yes Whether
Whetheritit No
Yes falls No
fallsinto
intolocal
local
optimization?
optimization?
Generation
Generationand
and
update No
update Gbest
Gbest Crossover No
Crossover
DMOPSO
DMOPSO

Gbest
Gbestassignment

LS-GA
assignment

LS-GA
Mutation
Mutation
Select
SelectPbest
Pbest
Local
LocalSearch
Search
Update
Updateparticle
particle
velocity
velocityand
andposition
position

Stopping
Stopping
criteria
criteriamet?
met?
Yes
Yes
Stop
Stop

Figure 9.9.The
Figure9. flow
Theflow chart
flowchart of
chartof the
ofthe hybrid
thehybrid strategy
hybridstrategy for
strategyfor HGA-PSO1.
forHGA-PSO1.
HGA-PSO1.
Figure The

(3)
(3) Hybrid
(3) HybridStrategy
Hybrid Strategyof
Strategy of HGA-PSO2
ofHGA-PSO2
HGA-PSO2
The other
Theother method
othermethod
method ofof
ofhybrid
hybridstrategy
strategybased
basedon on LS-GA and PSO is: The population is
The hybrid strategy based on LS-GA
LS-GA and and PSO
PSOis:is:The
Thepopulation
populationis
divided into two subpopulations for LS-GA and PSO, and nondominated sorting and
isdivided
dividedinto
intotwotwosubpopulations
subpopulationsfor for LS-GA
LS-GA and PSO, and
and PSO, and nondominated
nondominated sorting
sortingandand
crowding
crowding distance-based
distance-based fitness
fitness assignment
assignment strategies
strategies of
of the
the NSGA-II
NSGA-II algorithm
algorithm are
are used
used
crowding distance-based fitness assignment strategies of the NSGA-II algorithm are used to
to
toimplement
implement the
thecompetitive
competitive selection.
selection. LS-GA
LS-GA selects
selects the
theinitial
initial population
population from
from the
theex-
ex-
implement the competitive selection. LS-GA selects the initial population from the external
ternal archive
ternal archive by selection
by selection operation
operation to optimize
to optimize the solution
the solution in the external archive, and
archive by selection operation to optimize the solution in the in the external
external archive,
archive, and PSO and
PSO
PSO selects
selectsindividual
individual particles
particles from
from the
the overall
overall parent
parent population
population for
for global
globalsearch.
search. The
The
selects individual particles from the overall parent population for global search. The flow
flow
flow chart
chart of
ofthe
the hybrid
hybrid strategy
strategy of
of HGA-PSO2
HGA-PSO2 isisshown
shown in
in Figure
Figure 10.
10.
chart of the hybrid strategy of HGA-PSO2 is shown in Figure 10.

Initialize
Initialize External
External Fitness
Fitnessassignment
assignment
Start
Start population
population archive
archive and
andselection
selection
No
No
Generation
Generationand
and
Crossover update
update Gbest
Gbest
Crossover
DMOPSO
DMOPSO

Gbest
Gbestassignment
LS-GA

assignment
LS-GA

Mutation
Mutation
Select
SelectPbest
Pbest
Local
LocalSearch
Search Update
Updateparticle
particle
velocity
velocityand
andposition
position

Stopping
Stopping
criteria
criteriamet?
met?
Yes
Yes
Stop
Stop

Figure
Figure 10. The
10.The
Figure10. flow
Theflow chart
flowchart of
chartof the
ofthe hybrid
thehybrid strategy
hybridstrategy for
strategyfor HGA-PSO2.
forHGA-PSO2.
HGA-PSO2.
Algorithms 2022, 15, 182 17 of 24

4. Numerical Experiments
For comparing the performance of LS-GA, HGA-PSO1, and HGA-PSO2, all algorithms
solve nine test experiments 10 times, and the performance measures defined in [20], the
average objectives value (avg(·)), the runtime of the algorithm (Runtime (S)), the number
of nondominated sets (M1 ), set coverage measure (M2 ), and mean ideal distance (MID), are
used to analyze the performance of algorithms.
The main parameters of the test experiments are listed in Table 6 (k = 1), and other
parameters are described in Tables A1–A3 of Appendix A. Several parameters, such as the
population sizes, the maximum number of generations, the probability of the crossover,
and the mutation operator, may influence the performance of an algorithm. In this research,
all the parameters of the GA and PSO algorithms are relatively efficient values obtained
through many tests, which are shown in Table 7. Moreover, the backordering intensity
coefficient k0 = 0.25, the waiting time resistance k1 = 0.3, the fixed cost part of product
i f i = 0.05PCi , the cost rate ai = 0.5 f i , and the cost increasing rate bi = 0.05 f i , respectively.
The rates of customer loss threshold and backorder cost per unit quantity are shown in
Figures 11 and 12, respectively.

Table 6. Parameters setting.

Experiments Parameters of Labors

No. I T J Parameter Value
1 2 4 3 W HC (Yuan/p) 100
2 2 6 3 WR (h) 50
3 2 8 3 WO (h) 10
4 4 4 3 W Ik (Yuan) 800
5 4 6 3 WRCk (Yuan/h) 5
6 4 8 3 WOCk (Yuan/h) 10
7 6 4 3 W1k (p) 10
8 6 6 3
9 6 8 3

Table 7. Parameters of the GA and PSO.

Parameter Setting of GA Parameter Setting of PSO

Parameter Value Parameter Value
Experiments 1~3: 30, each sub-population
of HGA-PSO2: 15
Experiments 4~6: 40, each sub-population
Population sizes popsize Constriction factor 0.73
of HGA-PSO2: 20
Experiments 7~9: 50, each sub-population
of HGA-PSO2: 25
Experiments 1~3: 1000, Set algebra of
HGA-PSO1: 500
Predefined maximum
Experiments 4~6: 1200, Set algebra of
Maximum number of generations G value of inertia weight 0.8
HGA-PSO1: 600
ωmax
Experiments 7~9: 1500, Set algebra of
HGA-PSO1: 750
Predefined minimum
Probability of partheno crossover
Iteration number < 600: 0.2, otherwise, 0.3 value of inertia weight 0.4
operator: Pc1
ωmin
Probability of arithmetic crossover Pc2 Iteration number < 600: 0.1, otherwise, 0.2 Learning factors c1 2.0
Probability of production mutation Pm1 Iteration number < 600: 0.4, otherwise, 0.6 Learning factors c2 2.1
Probability of mutation for the number
Iteration number < 600: 0.5, otherwise, 0.7
of workers Pm2
 c2 0.2
Probability
Probability of
of production
production mutation
mutation Iteration
Iteration number
number << 600:
600: 0.4,
0.4, otherwise,
otherwise, Learning factors c2 2.1
PPm1 0.6 Learning factors c2 2.1
m1 0.6
Probability
Probability of
of mutation
mutation for
for the Iteration
the Iteration number
number << 600:
600: 0.5,
0.5, otherwise,
otherwise,
Algorithms
number 2022,
of workersP
15, 182
number of workers Pmm22 0.7 18 of 24
0.7

loss( β( )β )
0.25 0.25
0.25 0.25

customerloss
0.185
0.2 0.185

forcustomer
0.2 0.137
0.137
0.15 0.102
0.15 0.102
0.075
thresholdfor
0.0750.056
Thethreshold 0.1 0.0560.041
0.1
0.0410.031
0.05 0.031
0.05
The

0
0 1 2 3 4 5 6 7 8
1 2 customer
The 3 4waiting
5 time6 ( t'- 7t) 8
The customer waiting time ( t'- t)
Figure 11. The rate of customer loss threshold.
Figure11.
Figure 11.The
Therate
rateof
ofcustomer
customerloss
lossthreshold.
threshold.

0.5
0.5
0.41
0.4 0.41
( cb) i )

0.4
i

0.3475
cost( cb

0.3475
0.3 0.29
backordercost

0.3 0.29
0.2375
Thebackorder

0.2375
0.2 0.19
0.2 0.19
0.1475
0.1475
0.11
0.1 0.0775 0.11
The

0.1 0.0775

0
0 1 2 3 4 5 6 7 8
1 2 customer
The 3 4waiting
5 time6 ( t'- 7t) 8
The customer waiting time ( t'- t)
Figure 12.
12. The rate
rate of backorder
backorder cost.
Figure 12.The
Figure The rateof
of backordercost.
cost.

The
The experiments are
The experiments are performed
are performedon
performed onDELL
on DELLVostro
DELL Vostro3800-R1846
Vostro (Intel
3800-R1846
3800-R1846 Core
(Intel
(Intel i3
Core
Core i3 4130(3.4
i3 4130
4130(3.4
(3.4
GHz GHz
8 GB 8memory))
GB memory))
10 10 times.
times. The The results
results of of performance
performance measures
measures are are reported
reported in Table
GHz 8 GB memory)) 10 times. The results of performance measures are reported in Table
in Table 8 and13,
88 and Figure 13, respectively.
and Figure
Figure 13, respectively.
respectively.
Table 8. The results of performance measures.

Performance Experiment No.

Hybrid Strategy
Measure 1 2 3 4 5 6 7 8 9
LS-GA 8.96 14.71 19.42 16.47 26.44 34.84 23.32 37.28 50.29
avg( Z1 )/(104 ) HGA-PSO1 8.96 14.71 19.31 16.47 26.44 34.76 23.32 37.26 50.27
HGA-PSO2 8.95 14.71 19.27 16.46 26.45 34.75 23.31 37.26 50.29
LS-GA 15.78 15.66 18.83 25.21 25.75 30.23 32.76 36.00 37.45
avg( Z2 ) HGA-PSO1 15.83 15.52 18.07 25.21 25.39 29.58 32.80 36.50 37.70
HGA-PSO2 16.08 15.61 17.26 25.53 25.75 29.63 33.08 36.47 37.97
LS-GA 75.28 87.86 102.54 152.64 177.57 204.66 276.56 300.18 348.90
Runtime/(S) HGA-PSO1 73.57 69.64 79.40 124.25 126.31 129.35 192.50 227.47 341.08
HGA-PSO2 78.22 106.39 99.46 120.52 123.45 127.95 204.67 230.62 210.89
LS-GA 60 88 115 28 28 110 42 22 31
M1 HGA-PSO1 64 87 110 28 23 107 45 24 33
HGA-PSO2 66 90 98 17 24 101 26 15 34
LS-GA, HGA-PSO1 −0.03 −0.08 −0.87 0.21 −0.20 −0.46 0.18 0.09 −0.28
LS-GA, HGA-PSO2 −0.08 −0.09 −0.97 0.58 0.18 −0.35 0.35 0.06 0.24
HGA-PSO1, LS-GA 0.03 0.08 0.87 −0.21 0.20 0.46 −0.18 −0.09 0.28
M2
HGA-PSO1,
−0.05 −0.08 0.14 0.23 0.39 0.07 0.05 −0.08 0.54
HGA-PSO2
HGA-PSO2, LS-GA 0.08 0.09 0.97 −0.58 −0.18 0.35 −0.35 −0.06 −0.24
HGA-PSO2,
0.05 0.08 −0.14 −0.23 −0.39 −0.07 −0.05 0.08 −0.54
HGA-PSO1
 Algorithms 2022, 15, 182 19 of 24

Table 8. Cont.

Performance Experiment No.

Hybrid Strategy
Measure 1 2 3 4 5 6 7 8 9
LS-GA 23.97 33.43 43.17 41.50 58.83 76.02 57.00 82.81 107.34
Algorithms 2022, 15, x FOR PEER REVIEW 20 of 25
MID HGA-PSO1 23.99 33.36 42.82 41.49 58.67 75.64 57.03 82.99 107.38
HGA-PSO2 24.14 33.40 42.38 41.66 58.83 75.62 57.17 82.97 107.51

5
6x10
LS-GA 60
LS-GA
HGA-PSO1
5 HGA-PSO1
5x10 HGA-PSO2
50 HGA-PSO2

5
4x10 40
avg(Z1)

avg(Z2)
5
3x10 30

2x10
5 20

5 10
1x10

0
0
1 2 3 4 5 6 7 8 9
1 2 3 4 5 6 7 8 9
Experiment No.
Experiment No.

(a) avg( Z1 ) (b) avg( Z2 )

350 120
LS-GA LS-GA
HGA-PSO1 HGA-PSO1
300 HGA-PSO2 100 HGA-PSO2

250
80

200
Runtime(S)

60
MID

150
40
100

50 20

0 0
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
Experiment No. Experiment No.

(c) Runtime (d) MID

120
LS-GA
HGA-PSO1
100 HGA-PSO2

80

60
M1

40

20

0
1 2 3 4 5 6 7 8 9
Experiment No.

(e) M1

Figure
Figure 13.13. Comparativeanalysis
Comparative analysis of
of performance
performancemeasures.
measures.

Table 8. The results of performance measures.

Performance Meas- Experiment No.

Hybrid Strategy
ure 1 2 3 4 5 6 7 8 9
LS-GA 8.96 14.71 19.42 16.47 26.44 34.84 23.32 37.28 50.29
 HGA-PSO2, LS-GA 0.08 0.09 0.97 −0.58 −0.18 0.35 −0.35 −0.06 −0.24
HGA-PSO2, HGA-PSO1 0.05 0.08 −0.14 −0.23 −0.39 −0.07 −0.05 0.08 −0.54
LS-GA 23.97 33.43 43.17 41.50 58.83 76.02 57.00 82.81 107.34
MID HGA-PSO1 23.99 33.36 42.82 41.49 58.67 75.64 57.03 82.99 107.38
Algorithms 2022, 15, 182 20 of 24
HGA-PSO2 24.14 33.40 42.38 41.66 58.83 75.62 57.17 82.97 107.51

From the comparison results of performance measures avg( Z 1 ) , avg( Z 2 ) , M 1

and MID,the
From therecomparison results of
is little difference inperformance
the performancemeasures avg( Z
of LS-GA, 1 ), avg( Z2 )and
HGA-PSO1, , M1HGA-
and
MID, there is little difference in the performance of LS-GA, HGA-PSO1,
PSO2 algorithms. From the algorithm running time, HGA-PSO1 and HGA-PSO2 algo- and HGA-PSO2
algorithms. From the algorithm running time, HGA-PSO1 and HGA-PSO2 algorithms
rithms have less average runtime than the LS-GA strategy, and the larger the problem
have less average runtime than the LS-GA strategy, and the larger the problem scale, the
scale, the greater the difference in running time. Moreover, the M 2 in Table 8 shows that
greater the difference in running time. Moreover, the M2 in Table 8 shows that the overall
the overall performances
performances of HGA-PSO1 ofand
HGA-PSO1
HGA-PSO2 andare
HGA-PSO2 are slightly
slightly better than thatbetter than that
of LS-GA, andof
LS-GA, and there is no significant difference between HGA-PSO1
there is no significant difference between HGA-PSO1 and HGA-PSO2 strategies. The and HGA-PSO2 strate-
gies. Thesolutions
obtained obtainedofsolutions of three
three hybrid hybridare
strategies strategies
compared are in
compared
Figure 14.inThe
Figure 14. The
results show re-
sults show that there is no significant difference in the number of nondominated
that there is no significant difference in the number of nondominated sets obtained by the sets ob-
tained
three by the
hybrid three hybrid strategies.
strategies.
4 5 5
x 10 x 10 x 10
9.15 1.52 2.15
HGA-PSO2 HGA-PSO2 HGA-PSO2
HGA-PSO1 HGA-PSO1 HGA-PSO1
9.1 1.51 2.1
LS-GA LS-GA LS-GA

9.05 1.5 2.05

1.49
9 2

1
Z1

Z
1
Z

1.48
8.95 1.95

1.47
8.9 1.9

1.46
8.85 1.85
12 13 14 15 16 17 18 19 22 11 12 13 14 15 16 17 18 19 20 21 10 11 12 13 14 15 16 17 18 19 20 21 22 24 25 26 27
Z Z Z
2
2 2

(a) Experiment 1 (b) Experiment 2 (c) Experiment 3

5 5 5
x 10 x 10 x 10
1.66 2.66 3.62
HGA-PSO2 HGA-PSO2 HGA-PSO2
1.658 2.658 3.6 HGA-PSO1
HGA-PSO1 HGA-PSO1
LS-GA 2.656 LS-GA
1.656 LS-GA 3.58
2.654
1.654 3.56
2.652
1.652 3.54
2.65
1
1
1

Z
Z
Z

1.65 3.52
2.648
1.648 3.5
2.646
1.646 3.48
2.644

1.644 2.642 3.46

Algorithms
1.642 2022, 15, x FOR PEER REVIEW 2.64
24 25 26 27
3.44
22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 39 40 22
42 of 25
24 25 26 27 28 29
Z Z Z
2 2
2

(d) Experiment 4 (e) Experiment 5 (f) Experiment 6

5 5 5
x 10 x 10 x 10
2.348 3.75 5.06
LS-GA LS-GA
2.346 LS-GA HGA-PSO1 5.055 HGA-PSO1
HGA-PSO1 3.745 HGA-PSO2
2.344 HGA-PSO2
HGA-PSO2 5.05
2.342
3.74
2.34 5.045
1
1

Z
Z

2.338 3.735 5.04

1
Z

2.336
5.035
3.73
2.334
5.03
2.332
3.725
2.33 5.025

2.328 3.72 5.02

30 31 32 33 34 35 36 34 35 36 37 38 39 35 36 37 38 39 40 41 42
Z Z Z
2 2 2

(g) Experiment 7 (h) Experiment 8 (i) Experiment 9

Figure
Figure 14.14. Comparison
Comparison ofof calculation
calculation results
results forfor three
three strategies.
strategies.

5. 5. Discussion
Discussion and
and Conclusions
Conclusions
The increasingly
The increasinglyfierce market
fierce competition
market makes
competition enterprises
makes faceface
enterprises the changing market
the changing mar-
environment. The diversity of products and fierce competition make the stability
ket environment. The diversity of products and fierce competition make the stability of of the
manufacturing industry
the manufacturing and supply
industry chainchain
and supply moremore
important than than
important ever.ever.
In order to adapt
In order to
to adapt
the
torapidly changing
the rapidly marketmarket
changing demand, more and
demand, more
more manufacturing
and enterprises
more manufacturing choose
enterprises
choose the coexistence of order-oriented and inventory-oriented multi-variety and small
batch production. The change in demand level will lead to a change in enterprise employ-
ment, and the change in employees will lead to a change in workers’ labor efficiency and
labor cost. Therefore, decision makers pay great attention to the impact of labor changes,
Algorithms 2022, 15, 182 21 of 24

the coexistence of order-oriented and inventory-oriented multi-variety and small batch

production. The change in demand level will lead to a change in enterprise employment,
and the change in employees will lead to a change in workers’ labor efficiency and labor
cost. Therefore, decision makers pay great attention to the impact of labor changes, and the
stability of workers has become more important than ever.
Due to the decline of actual production capacity or the shortage of production materi-
als, production enterprises may be backordered. Generally, the method of penalty cost is
adopted to deal with the BC, and the BC per unit time is fixed, and most of them do not
consider the difference in the time to replenishment. In fact, the longer the replenishment
time and the greater the backorder quantity, the greater the possibility of a loss of sales
and the worse the customer satisfaction. Therefore, the impact of early and late delivery
is analyzed, and the threshold for customer loss and the BC per unit time is defined by
considering the dual effects of replenishment time and backorder quantity.
Then, the maximum tolerant backorder quantity of customers in different delay peri-
ods is calculated according to the demand for products. In considering the waiting time
and the quantity of backorder, the delayed delivery cost varying with the waiting time is
designed, and the cost of delayed delivery and loss of sales is determined. Considering the
production cost, raw material cost, inventory cost, staff cost, stockout, and lost sales cost,
an early/delayed multi-objective optimization model is developed for an APP problem
of multi-product.
Moreover, three algorithms, GA, LS, and DMOPSO, and three different hybrid strate-
gies, are designed to solve the model, and the experiments are performed on some testing
examples. The computational results indicated that the determination of the feasible range
of product output and the number of workers in the workforce can reduce the search
scope and improve the efficiency of the algorithm. HGA-PSO1 and HGA-PSO2 strate-
gies have less average runtime than the LS-GA strategy, and the larger the problem scale,
the greater the difference in running time. In the future, the uncertainty of product de-
mand and production capacity will be taken into account to design the related APP model
and algorithm.

Author Contributions: Conceptualization, L.L.; methodology, X.Y. and L.L.; validation and for-
mal analysis, L.L.; investigation and data curation, X.Y.; writing—original draft preparation, X.Y.;
writing—review and editing, L.L.; supervision, X.Y.; project administration, X.Y.; funding acquisition,
X.Y. and L.L. All authors have read and agreed to the published version of the manuscript.
Funding: This work is supported by the National Natural Science Foundation of China (Grant
No. 71761024), the Natural Science Foundation of Gansu Province of China (No. 21JR1RA236),
and Double-First Class Major Research Programs, Educational Department of Gansu Province
(No. GSSYLXM-04).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Data sharing is not applicable to this article.
Conflicts of Interest: The authors declare no conflict of interest.
Algorithms 2022, 15, 182 22 of 24

Appendix A

Table A1. Production capacity of experiments.

Experiment t
i
No. 1 2 3 4 5 6 7 8
1 50 90 190 260 – – – –
1
2 40 40 50 100 – – – –
1 50 80 180 250 200 200 – –
2
2 50 50 50 100 100 90 – –
1 50 80 175 260 210 200 150 150
3
2 50 50 50 100 80 80 80 80
1 50 100 150 280 – – – –
2 50 80 50 80 – – – –
4
3 50 50 70 70 – – – –
4 100 110 100 110 – – – –
1 50 110 150 250 250 200 – –
2 50 80 80 80 80 80 – –
5
3 60 60 60 70 70 70 – –
4 100 110 100 100 100 120 – –
1 50 100 150 260 250 200 150 150
2 50 80 80 100 80 80 80 80
6
3 60 60 60 70 80 80 80 70
4 100 100 100 120 120 120 100 100
1 60 100 200 200 – – – –
2 60 70 70 80 – – – –
3 60 60 70 70 – – – –
7
4 120 110 100 120 – – – –
5 70 70 70 70 – – – –
6 90 90 100 90 – – – –
1 70 150 200 250 220 200 – –
2 80 80 80 80 80 80 – –
3 70 70 70 80 80 80 – –
8
4 80 100 100 150 100 150 – –
5 80 80 80 80 80 80 – –
6 90 100 100 90 100 100 – –
1 60 70 150 250 220 200 200 200
2 80 80 100 100 100 100 100 80
3 70 70 70 80 80 80 80 80
9
4 90 150 100 150 150 150 100 100
5 90 85 80 80 90 100 95 90
6 100 95 105 100 100 95 105 100
Algorithms 2022, 15, 182 23 of 24

Table A2. Parameters of production and raw materials.

t
Parameters i, j
1 2 3 4 5 6 7 8
1 85.0 125.0 175.0 250.0 200.0 180.0 150.0 120.0
2 30.0 45.0 50.0 80.0 70.0 65.0 55.0 60.0
PDit 3 50.0 55.0 60.0 70.0 55.0 75.0 60.0 65.0
(unit) 4 100.0 120.0 90.0 110.0 100.0 120.0 90.0 85.0
5 50.0 70.0 65.0 70.0 50.0 80.0 85.0 90.0
6 80.0 85.0 95.0 75.0 80.0 85.0 90.0 95.0
1 28.8 28.8 28.8 28.8 28.8 28.8 28.8 28.8
2 23.0 23.0 23.0 23.0 23.0 23.0 23.0 23.0
PCi 3 20.0 20.0 20.0 20.0 20.0 20.0 20.0 20.0
(Yuan/unit) 4 30.0 30.0 30.0 30.0 30.0 30.0 30.0 30.0
5 20.0 20.0 20.0 20.0 20.0 20.0 20.0 20.0
6 22.0 22.0 22.0 22.0 22.0 22.0 22.0 22.0
1 3.8 3.8 3.8 3.8 3.8 3.8 3.8 3.8
2 5.7 5.7 5.7 5.7 5.7 5.7 5.7 5.7
PWik 3 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0
(h/unit) 4 4.0 4.0 4.0 4.0 4.0 4.0 4.0 4.0
5 5 5 5 5 5 5 5 5
6 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5
1 100.0 170.0 200.0 255.0 220.0 210.0 200.0 200.0
2 80.0 80.0 150.0 110.0 100.0 100.0 100.0 80.0
PNit 3 70.0 75.0 75.0 75.0 80.0 80.0 95.0 90.0
(unit) 4 150.0 150.0 160.0 150.0 150.0 140.0 150.0 140.0
5 90.0 85.0 80.0 80.0 90.0 100.0 95.0 90.0
6 100.0 95.0 105.0 100.0 100.0 95.0 105.0 100.0
1 2.0 2.0 3.0 1.0 2.0 2.0 2.0 3.0
RCjt
2 3.0 2.0 3.0 3.0 2.0 2.0 3.0 2.0
(Yuan/unit)
3 3.0 3.5 3.0 2.8 3.0 4.0 3.0 3.5

Table A3. Parameters of inventory and raw materials.

i
Parameters
1 2 3 4 5 6
CIi1 (unit) 65.0 29.0 10.0 20.0 15.0 25.0
CKi (Yuan) 15.0 2.0 3.0 3.0 4.0 3.0
CNi (unit) 100.0 50.0 70.0 100.0 55.0 60.0
1 0.8 0.3 0.2 0.5 0.4 0.1
Rij (unit) 2 0.5 0.5 0.3 0.2 0.2 0.4
3 0.0 0.4 0.3 0.6 0.1 0.2

References
1. Cheraghalikhani, A.; Khoshalhan, F.; Mokhtari, H. Aggregate production planning: A literature review and future research
directions. Int. J. Ind. Eng. Comput. 2019, 10, 309–330. [CrossRef]
2. Hu, W.T.; Kim, S.L.; Banerjee, A. An inventory model with partial backordering and unit backorder cost linearly increasing with
the waiting time. Eur. J. Oper. Res. 2009, 197, 581–587. [CrossRef]
3. Wang, R.C.; Liang, T.F. Aggregate production planning with multiple fuzzy goals. Int. J. Adv. Manuf. Technol. 2005, 25, 589–597.
[CrossRef]
4. Ning, Y.F.; Tang, W.S.; Zhao, R.Q. Multi-product aggregate production planning in fuzzy random environments. World J. Model.
Simul. 2006, 22, 312–321.
5. Mahdavi, I.; Aalaei, A.; Paydar, M.M.; Solimanpur, M. Designing a mathematical model for dynamic cellular manufacturing
systems considering production planning and worker assignment. Comput. Math. Appl. 2010, 60, 1014–1025. [CrossRef]
6. Chakrabortty, R.K.; Hasin, M.A.A. Solving an aggregate production planning problem by using multi-objective genetic algorithm
(MOGA) approach. Int. J. Ind. Eng. Comput. 2013, 4, 1–12. [CrossRef]
Algorithms 2022, 15, 182 24 of 24

7. Sadeghi, M.; Razavi Hajiagha, S.H.; Hashemi, S.S. A fuzzy grey goal programming approach for aggregate production planning.
Int. J. Adv. Manuf. Technol. 2013, 64, 1715–1727. [CrossRef]
8. Saidi-Mehrab Ad, M.; Paydar, M.M.; Aalaei, A. Production planning and worker training in dynamic manufacturing systems.
J. Manuf. Syst. 2013, 32, 308–314. [CrossRef]
9. Basis, H.; Lu, S.; Su, H.Y.; Wang, Y.; Xie, L.; Zhang, Q.L. Multi-product multi-stage production planning with lead time on a
rolling. IFAC PapersOnLine 2015, 48, 1162–1167. [CrossRef]
10. Modarres, M.; Izadpanahi, E. Aggregate production planning by focusing on energy saving: A robust optimization approach.
J. Clean. Prod. 2016, 133, 1074–1085. [CrossRef]
11. Hossain, M.M.; Nahar, K.; Reza, S.; Shaifullah, K.M. Multi-period, multi-product, aggregate production planning under demand
uncertainty by considering wastage cost and incentives. World Rev. Bus. Res. 2016, 6, 170–185.
12. Sakhaii, M.; Tavakkoli-Moghaddam, R.; Bagheri, M.; Vatani, B. A robust optimization approach for an integrated dynamic cellular
manufacturing system and production planning with unreliable machines. Appl. Math. Model. 2016, 40, 169–191. [CrossRef]
13. Mehdizadeh, E.; Niaki, S.T.A.; Hemati, M. A bi-objective aggregate production planning problem with learning effect and
machine deterioration: Modeling and solution. Comput. Oper. Res. 2018, 91, 21–36. [CrossRef]
14. Jamalnia, A.; Yang, J.B.; Xu, D.L.; Feili, A.; Jamali, G. Evaluating the performance of aggregate production planning strategies
under uncertainty in soft drink industry. J. Manuf. Syst. 2019, 50, 146–162. [CrossRef]
15. Xue, G.S.; Offodile, O.F. Integrated optimization of dynamic cell formation and hierarchical production planning problems.
Comput. Ind. Eng. 2020, 139, 106155. [CrossRef]
16. Jang, J.; Chung, B.D. Aggregate production planning considering implementation error: A robust optimization approach using
bi-level particle swarm optimization. Comput. Ind. Eng. 2020, 142, 106367. [CrossRef]
17. Florian, M.; Lenstra, J.K.; Rinnooy Kan, A.H.G. Deterministic production planning: Algorithms and complexity. Manag. Sci. 1980,
26, 669–679. [CrossRef]
18. Chen, W.H.; Thizy, J.M. Analysis of relaxations for the multi-item capacitated lot-sizing problem. Ann. Oper. Res. 1990, 26, 29–72.
[CrossRef]
19. Ramezanian, R.; Rahmani, D.; Barzinpour, F. An aggregate production planning model for two phase production systems: Solving
with genetic algorithm and tabu search. Expert Syst. Appl. 2012, 39, 1256–1263. [CrossRef]
20. Liu, L.F.; Yang, X.F. Multi-objective aggregate production planning for multiple products: A local search-based genetic algorithm
optimization approach. Int. J. Comput. Intell. Syst. 2021, 14, 156. [CrossRef]
21. Wang, S.C.; Yeh, M.F. A modified particle swarm optimization for aggregate production planning. Expert Syst. Appl. 2014, 41,
3069–3077. [CrossRef]
22. Chakrabortty, R.K.; Hasin, M.A.A.; Sarker, R.A.; Essam, D.L. A possibilistic environment based particle swarm optimization for
aggregate production planning. Comput. Ind. Eng. 2015, 88, 366–377. [CrossRef]
23. Ramachandran, G.M.; Neelakrishnan, S. An approach to improving customer on-time delivery against the original promise date.
S. Afr. J. Ind. Eng. 2017, 28, 109–119. [CrossRef]
24. Liu, X.; Tu, Y. Production planning with limited inventory capacity and allowed stockout. Int. J. Prod. Econ. 2008, 111, 180–191.
[CrossRef]
25. Dye, C.Y.; Hsieh, T.P.; Ouyang, L.Y. Determining optimal selling price and lot size with a varying rate of deterioration and
exponential partial backlogging. Eur. J. Oper. Res. 2007, 181, 668–678. [CrossRef]
26. Yu, B.Y.; Kang, H.W.; Shen, Y.; Sun, X.P.; Chen, Q.Y. Research on quorum sensing particle swarm optimization based on chaos.
IOP Conf. Series: Mater. Sci. Eng. 2020, 768, 072027. [CrossRef]
27. Wang, H.F.; Moon, I.; Yang, S.X.; Wang, D.W. A memetic particle swarm optimization algorithm for multimodal optimization
problems. Inf. Sci. 2012, 197, 38–52. [CrossRef]
28. Clerc, M.; Kennedy, J. The particle swarm-explosion, stability, and convergence in a multidimensional complex space. IEEE Trans.
Evol. Comput. 2002, 6, 58–73. [CrossRef]