Journal of the Air & Waste Management Association

ISSN: 1096-2247 (Print) 2162-2906 (Online) Journal homepage: www.tandfonline.com/journals/uawm20

Disassembly sequence planning of waste auto

parts

Jia Mao, Dou Hong, Zhe Chen, Ma Changhai, Li Weiwen & Ju Wang

To cite this article: Jia Mao, Dou Hong, Zhe Chen, Ma Changhai, Li Weiwen & Ju Wang (2021)
Disassembly sequence planning of waste auto parts, Journal of the Air & Waste Management
Association, 71:5, 607-619, DOI: 10.1080/10962247.2020.1871444

To link to this article: https://doi.org/10.1080/10962247.2020.1871444

Published online: 08 Feb 2021.

Submit your article to this journal

Article views: 2197

View related articles

View Crossmark data

Citing articles: 9 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=uawm20
JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION
2021, VOL. 71, NO. 5, 607–619
https://doi.org/10.1080/10962247.2020.1871444

TECHNICAL PAPER

Disassembly sequence planning of waste auto parts

Jia Maoa, Dou Honga, Zhe Chena, Ma Changhaia, Li Weiwena, and Ju Wangb
a
School of Transportation, Jilin University, Changchun, Jilin, People’s Republic of China; bCollege of New Energy and Environment, Jilin
University, Changchun, Jilin, People’s Republic of China

ABSTRACT PAPER HISTORY

The disassembly of used products is a critical procedure in remanufacturing, and different disas­ Received July 3, 2019
sembly strategies are often obtained from different perspectives. To describe the disassembly Revised December 2, 2020
process more accurately, the uncertainty of the information in the disassembly process should be Accepted December 5, 2020
considered. Therefore, random variables are introduced for disassembly time, cost, and effort. Based
on the extended stochastic Petri net modeling method and stochastic programming theory,
a stochastic optimization algorithm combined with artificial intelligence technology and
a multiobjective genetic algorithm are designed, and a multiobjective optimization model for the
disassembly sequence of used car parts under uncertain conditions is successfully constructed. This
model considers the viewpoint of the decision maker. Moreover, the Monte Carlo method is applied
to solve the multiobjective optimization model, and the validity and practicability of the model are
verified by an example of an automotive transmission.
Implications: With the rapid development of the economy and the shortening of the product life
cycle, the rate of product renewal is getting faster and faster, which also leads to the production of
a large number of waste products. According to the forecast of the relevant departments, it is
estimated that, in 2020, there will be about 35 million used televisions, 15 million used refrigerators,
13 million used washing machines, 12 million used air conditioners, 57 million used computers, and
8.3 million scrapped cars. Waste products contain a lot of renewable resources. If they cannot be
effectively recycled, it will be a great waste of resources, and unreasonable disposal of waste
products may have a negative impact on the environment. Therefore, due to environmental
pressure and economic drive, product recycling and remanufacturing activities have caused wide­
spread concern in society. Disassembly is defined as the operation or activity of disassembling an
assembly such as a product, assembly, or component, and is the result of multiple removal
operations of the product. It is a prerequisite for the efficient recycling of products and the first
link in remanufacturing, that is, disassembly as a new production activity, which can provide raw
materials for the smooth progress of the remanufacturing production plan, namely, old rough or
used parts. Efficient dismantling not only saves natural resources and energy, but also effectively
reduces environmental pollution. It is also an important guarantee for promoting the healthy
development of the circular economy and achieving sustainable industrial development. The
length of time required for the dismantling process, the level of costs, and the amount of profits
obtained will directly affect the economic benefits of the recycling of end-of-life products.
Therefore, the evaluation and optimization of the dismantling process of waste products have
become one of the current hot issues. The research on the dismantling of waste products is
conducive to speeding up the recycling process of waste, to a greater extent, the rapid and full
recovery of resources, and to a certain extent, it will provide value basis and theoretical significance
for subsequent research.

Introduction
et al. (2012) emphasized the difference between disas­
Disassembly planning is a decision-making process of sembly and dismantlement in the process of disassembly
disassembly sequence planning based on the optimal of waste products, defined the concept of depth and
objective function. From the whole process of product degree of disassembly, and put forward the evaluation
design, Kroll and Carver (1999) studied the difficulty of method and planning model based on the probability of
product disassembly and put forward the evaluation product disassembly.
method of product disassembly, including disassembly Homem de Mello and Sanderson (1991) proposed an
type, disassembly accessibility, disassembly strength, algorithm for generating the sequence of all mechanical
and other indicators, which laid a theoretical foundation devices for a given product by using the component
for the generation of other evaluation indicators. Tian relationship model. At this time, the disassembly

CONTACT Ju Wang [email protected] College of New Energy and Environment, Jilin University, Changchun, Jilin, People’s Republic of China.
© 2021 A&WMA
608 J. MAO ET AL.

problem was decomposed into different disassembly stochastic weight method to transform the multiobjec­
subproblems, and the disassembly task was transformed tive optimization function to the single-objective opti­
into an inverse process of the assembly task. Zussman mization function, in order to obtain the compromise or
and Zhou (2000) proposed a disassembly planning Pareto solution of the multiobjective stochastic optimi­
method based on Petri nets model, and proposed three zation of product disassembly. In view of the increasing
kinds of planning and execution algorithms (global fix­ environmental and economic pressures in the process of
ing, finite adaptive, and adaptive) for disassembly pro­ disassembly, and the influence of human interference on
cess, which realized the optimization of product disassembly, Tang, Zhou, and Gao (2006) developed
disassembly process and provided theoretical guidance a fuzzy attribution Petri nets (FAPN) model to mathe­
for product remanufacturing process. In order to max­ matically express the disassembly uncertainty caused by
imize the disassembly profit, Lambert (Lambert 2006a, a large number of human interventions, and proposed
2006b, 2007) used dynamic planning to optimize the an algorithm on this basis for optimal disassembly plan­
disassembly sequence, and proposed an improved two- ning, which is more suitable for actual industrial pro­
commodity network flow method, which reduces the cesses. Gao, Zhou, and Tang (2004) proposed a fuzzy
number of integer variables and is suitable for network reasoning Petri nets (FRPN) model to represent the
graph model with disassembly priority Then, by solving relevant decision-making rules in the process of disas­
the binary integer linear programming problem, he sembly in order to solve the problem of great uncer­
completely avoided the use of integer variables, and tainty in the products used, and used the fuzzy reasoning
proved the effectiveness and accuracy of the method. algorithm based on FRPN model to automatically and
Li, Khoo, and Tor (2005) proposed an object-oriented quickly make the most decision considering multicri­
modeling method for generating disassembly sequence teria disassembly rules. Grochowski and Tang (2009)
from disassembly constraint graph, and applied genetic used a machine learning method based on disassembly
algorithm to solve the optimization problem of feasible Petri nets (DPN) and hybrid Bayesian networks (HBN),
disassembly sequence. Adenso-Díaz, García -arbajal, which modeled the disassembly process by checking the
and Gupta (2008) proposed a stochastic adaptive greedy probabilistic relationship between different aspects of
search algorithm and a heuristic algorithm for path the disassembly process and predicted the results of
reengineering, aiming at the lowest cost and shortest each disassembly process. The above researchers have
time of disassembly, which is the first key link in the mainly discussed and studied the disassembly evalua­
product recycling process, and verified the effectiveness tion, disassembly sequence planning and disassembly
of the algorithm, which can be used to solve the bi- solutions under uncertain conditions, which provided
objective disassembly path problem. Aguinaga et al. some theoretical support for the writing of this paper.
(2008) analyzed the similarities and differences between Disassembly planning consists of two major steps:
the disassembly problem and the general path planning disassembly model construction and disassembly
problem, and proposed an improved fast growing ran­ sequence planning (DSP) (Tian et al. 2019,).Although
dom tree algorithm to solve the problem that the extrac­ some researchers have paid attention to the problem of
tion trajectory of parts may not be predicted in the random planning of product disassembly, on the basis of
disassembly process. this, this dissertation is mainly based on the construc­
It is noteworthy that the above research on disassem­ tion of a disassembly information model based on
bly of mechanical devices belongs to the category of advanced Petri nets; according to the actual disassembly
certainty. In the actual disassembly process of mechan­ decision requirements, a typical mathematical planning
ical devices, because the products to be disassembled are model is established; design combining artificial intelli­
used in different environments for different periods of gence technology with stochastic optimization algo­
time, human or objective uncertainties will be encoun­ rithm and multiobjective genetic algorithm (Hou et al.
tered in the disassembly process, such as the quality of 2017); analysis of disassembly modeling and compre­
disassembly personnel, the degree of damage to objects hensive optimization of waste products in uncertain
to be disassembled, disassembly tools, disassembly environments; and analysis and application verification
environment, and so on, which cannot be ignored in of related theories and methods.
the process of product disassembly. The rest of the article is divided into the following
Tian, Zhou, and Chu (2013) proposed a disassembly aspects. The second part describes the establishment of
multiobjective stochastic optimization model that com­ the multiobjective disassembly optimization model. The
bines stochastic simulation, genetic algorithm, and third part is the Monte Carlo algorithm design. The
neural network, and analyzed its convergence and fourth part mainly describes the analysis of specific
other mathematical properties. They also use the cases. The fifth part describes the solution of the
 JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION 609

established model. Finally, it summarizes and describes device, where PN = (P,T,A,S,W,R,M0), and now describes
the deficiencies of the research. the parameters of the six-tuple variable PN as follows.
(1) P = {P1, P2, . . ., Pm} is a collection of finite ware­
houses representing m parts, and the element Pi con­
Multiobjective stochastic disassembly
tained in it is the i-th part of the mechanical device to be
optimization model
dismantled, and i satisfies the relationship 1 ≤ i ≤ m,
In order to present the different requirements of decision where m is the total number of parts of the product.
makers for the disassembly process of waste products in (2) T = {T1, T2, . . ., Tn} is a finite transition set repre­
actual disassembly activities, the minimum expected value senting n disassembly operation steps of m parts, Tj repre­
model, the maximum disassembly degree model, and the sents a disassembly operation between two parts, and
minimum optimization model with disassembly time, dis­ j satisfies the relationship 1 ≤ j ≤ n, where n is the sum of
assembly fee, and disassembly energy as multi objectives all disassembly steps of the product to be disassembled.
are established in this section. In order to simplify the (3) A represents the directed arc between the place
model and express it conveniently, the fee of disassembly, and the transition. The mathematical expression of
the time of disassembly, and the energy of disassembly are matrix A can be expressed as follows
collectively referred to as the fee of disassembly–the fee of
disassembly time, the fee of disassembly fee, and the fee of A ¼ a8½ij�
disassembly energy. It should be noted that in the process < wði; jÞðWhen the directed arc points from pi to tj Þ
of solving the multiobjective stochastic optimization model ¼ wði; jÞðWhen the directed arc points from tj to pi Þ
:
(Kang et al. 2017), it is impossible to obtain the theoretical 0; ðotherÞ
optimal solution, so it is necessary to apply the Monte
(4) S = {S(P1), S(P2), S(P3), . . .} means a finite set of
Carlo method to solve the Pareto optimal solutions of the
places in the constraint state of each part, and S(Pi) = 1
model.
means that the constraint of part i has been lifted, S(Pi)
In practical applications, there are many modeling
= 0, the part i is still in a constrained state, which is as
methods for the dismantling of mechanical devices.
follows:
Directed graph models are generally expressed using
�
the binary variable G (V, E), and undirected graphs are 0; Unconstrained
also generally expressed using the binary variable G ‘(V, Sij ¼ ; 1 � i � m; 1 � j � n:
1; Bound
E). The AND or graph model is generally expressed by
the binary variable G (W, D), which is a modeling (5) R = {R(T1), R(T2), R(T3), . . . } is a finite set repre­
method that can effectively express the logical relation­ senting the changes that have occurred, obviously. If Ti ∈
ship between elements and the internal structure infor­ R, it means that Ti is a change that has occurred; if, it means
mation of the model, and the Petri net model is a highly that Ti is a change that has not yet occurred.
efficient, A modeling method for quantitatively and gra­ (6) M0 = (a1, a2, . . ., ai, . . ., am)T, where i∈[1,m], ai
phically processing discrete events (Tian et al. 2018). ∈{0,1}. That is, M0 defines the initial identity of the Petri
The Petri net model is a modeling method that can net. Each element in the vector M0 represents the number
process discrete events efficiently, quantitatively, and gra­ of tokens contained in each library in the initial state, and
phically. The Petri net model is mainly composed of a place the number of tokens in the initial state is at most 1.
(P), a transition (T), a token, and a directed arc. It can (7) W: W is the weight function of each transition.
vividly represent the direction of information flow and What W reflects in the extended stochastic Petri net is
qualitatively calculate the resources occupied by each link. the probability distribution subject to various changes
The construction of a disassembly information model is (such as dismantling time, dismantling cost, dismantling
the basis for product disassembly optimization. The con­ energy). The value method of the weight function is as
struction of product disassembly information model under follows: first, determine the probability distribution that
uncertain environment is a difficult point of the article. each change should obey, and then solve the optimal
Considering that Petri nets can simply and accurately solution of the planning model according to the stochas­
represent the advantages of concurrency, conflict, uncer­ tic optimization theory.
tainty, and dynamic process description of dismantling
systems, this article intends to use extended random con­
struction of product disassembly information model in
Model assumption
uncertain environment based on the Petri net theory.
This article uses the Petri net six-tuple variable PN to 1. Only when the parts are unconstrained can they be
describe the dismantling Petri net model of the mechanical disassembled from the mechanical device, and the
610 J. MAO ET AL.

variable S(Pi) is used to indicate that the parts W = Vector of disassembly energy, the elements in
(place) Pi is unconstrained. the vector are the random disassembly energy amounts
2. For such parts that are not necessary to be scat­ needed to complete each disassembly operation;
tered sometimes in the actual process of disassem­ M = Initial state identification column vector for the
bling mechanical devices, in order to prevent disassembly of a used product;
unnecessary wear and tear during disassembly M0 = Final state identification column vector for the
and reduce disassembly time, they can be disas­ disassembly of used product;
sembled as a whole. A = Incidence matrix, the incidence matrix in the
3. In the process of disassembling parts, the time stochastic extended Petri net model;
consumed by changing disassembly tools is U = Vector of decision variables, if U(i) = 1, the i-th
neglected. transition in the Petri net occurs; otherwise, it does not
4. All parts are rigid. In the process of disassembling occur.
parts, the disassembly direction is the action direc­
tion of two parts, but it is difficult to determine the
disassembly direction of non-rigid parts (such as Minimum expected cost disassembly model
wires), so the disassembly direction of non-rigid First, taking the minimum disassembly time, cost and
parts is not considered in this article. energy as objective functions, the following multiobjec­
5. All parts are removable. In practice, the disassem­ tive minimum expected disassembly fee model is
bly of all the parts connected to it may be affected constructed:
by the damage of some parts. For convenience of
research, it is assumed that all parts can be disas­ min½Eð~uk Þ; Eð~
T~ S~ ~ uk Þ�
uk Þ; EðW~ (1)
sembled in this article.
8
6. Disassembly time, disassembly fee, and disassem­ > ~A~u ¼M ~ M ~0
>
< k
bly energy of each component of waste products ~
u ðiÞ 2 f0; 1g; i ¼ 1; 2; � � � ; n
st: ~k :
are independent random variables. >
> MðjÞ 2 f0; 1g; j ¼ 1; 2; � � � ; m
: �� ~ �� �� ~ ��
M ¼ M0
Parameter definition
m = Total number of places, m = 1,2,3,.;
Maximum disassembly degree model
n = Total number of transitions, n = 1,2,3,;
k = Feasible disassembly sequence number, k = 1,2,3,; In the actual disassembly process, the decision-maker
i = Removal operations corresponding to transition may pay more attention to the possibility that a part can
i, i = 1,2,3,n; be disassembled within a specified time. The maximum
j = Parts to be removed corresponding to place j, cost disassembly model seeks the disassembly sequence
j = 1,2,3,m; uk which is most likely to disassemble the target parts or
t = Disassembly time constraints in the maximum completely disassemble the products within the specified
disassembly degree model; disassembly time t, disassembly fee s, and disassembly
s = Disassembly fee constraints in the maximum energy w:
disassembly degree model; n o n � �
w = Disassembly energy constraints in the maximum max½P ~ uk � t ; P ~
T~ ~ uk � w � (2)
uk � sg; P W~
S~
disassembly degree model;
8
p1 = Disassembly time opportunity constraints in the > ~A~u ¼M ~ M ~0
>
< k
minimum disassembly cost model; ~
u ðiÞ 2 f0; 1g; i ¼ 1; 2; � � � ; n
p2 = Disassembly fee opportunity constraints in the st: ~k :
>
> MðjÞ 2 f0; 1g; j ¼ 1; 2; � � � ; m
: �� ~ �� �� ~ ��
minimum disassembly cost model; M ¼ M0
p3 = Disassembly energy opportunity constraints in the
minimum disassembly cost model;
S = Vector of disassembly fee, the elements in the
Minimum cost disassembly model
vector are the random disassembly fee needed to com­
plete each disassembly operation; In the actual disassembly process, if the decision-maker
T = Vector of disassembly times, the elements in the is willing to take risks, he can choose the sequence that
vector are the random disassembly times needed to may produce the minimum disassembly fee under
complete each disassembly operation; a certain probability for disassembly operation:
 JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION 611

min½~uk ; ~
T~ S~ ~ uk �
uk ; W~ (3) disassembly degree, maximum fee disassembly degree,
maximum energy disassembly degree of the i-th feasible
The constraints are as follows: disassembly sequence under the constraints t, s, and w
8
> ~
A~uk ¼ M ~ M ~0 Step 7: Find the N●p1-th, N●p2-th, and N●p3-th ele­
>
>
>
> ~
u ðiÞ 2 f0; 1 g; i ¼ 1; 2; � � � ; n ments in the Ci1, Ci2, and Ci3 arrays, respectively, which
>
> k
> MðjÞ
< ~ 2 f0; 1g; j ¼ 1; 2; � � � ; m are the minimum disassembly times of the i-th feasible
� � � �
st: �M ~ � ¼ �M~ 0� : disassembly sequence under the constraints p1, p2, and
>
> p3, the minimum disassembly costs and the minimum
> PfTuk � t g � p1
>
>
>
> disassembly energy.
: PfSuk � sg � p2
>
PfWuk � wg � p3
After applying the above algorithm, the Pareto solu­
tion set of the disassembled sequence can be obtained by
comparing the solution results of the k feasible disas­
Algorithm design
sembled sequences.
After the construction of the extended random Petri net The basic elements and operating rules of the disas­
model, the feasible disassembly sequence can be calculated sembled Petri net modeling theory and the basic Petri
according to the equation~ AU~¼M ~ M ~ 0 . The Monte net modeling theory are roughly the same, but the dis­
Carlo method is to solve a problem. First, a probability assembled Petri net has the following characteristics.
model or a random process is established, so that its para­ 1) In the process of dismantling the product, there are
meters or numerical characteristics are equal to the solution dismantling in multiple directions, and there are com­
of the problem: then these parameters or digital character­ plex inter-relationships and spatial constraints between
istics are calculated by observation or sampling experiments parts. Therefore, when constructing the Petri net model,
on the model or process. Finally, give the approximated it is necessary to judge the connection relationship
value. The Monte Carlo method is used to simulate the (AND/OR relationship) between the parts according to
distributions of the total disassembly time, total fees, and the actual design and structure of the product, and
energy for each disassembly sequence. The multiobjective construct the disassembly priority matrix.
Pareto solution sets of the minimum expected cost model, 2) When generating the disassembled Petri net, in
maximum disassembly degree model, and minimum order to reflect the priority of disassembly and the spatial
disassembly cost model are calculated according to the characteristics of the disassembled structure, it is neces­
simulation results. Assuming that there are k feasible dis­ sary to insert appropriate places and changes in the
assembly paths, the specific algorithm is as follows: disassembled Petri net to express this structural feature.

Step 1: Initialize the random number, N, for the total

number of simulation trials. Case analysis
Step 2: Calculate the total disassembly time, disas­
Case description
sembly fees, and disassembly energy of the correspond­
ing probability distribution functions, φi ð�1 Þ, φi ð�2 Þ, In this section, taking automotive transmission as an
andφi ð�3 Þ, and the definition domain of the i-th feasible example, a comprehensive disassembly optimization
disassembly sequence, where 1 � i � k andi 2 Z. model for an uncertain environment is established
Step 3: Generate the random variables, �1 , �2 , and �3 based on the Stochastic Extended Petri Net Theory
in the probability distribution function domain (SEPNT) (Grochowski and Tang 2009; Tian et al.
Step 4: Repeat step 3 N times, and arrange the ran­ 2012), and its solution is analyzed. To facilitate the
dom variables, �1 ,�2 , and�3 , in ascending order accord­ analysis, according to the actual situation, the profile of
ing to their numerical values and store them in the the automobile transmission device is drawn, as shown
arrays, Ci1, Ci2, and Ci3, respectively in Figure 1.
Step 5: Calculate the averages, μi1 , μi2 , andμi3 , of the In the process of setting up the bolt disassembly experi­
arrays, Ci1, Ci2, and Ci3, for the expected disassembly and ment, this project mainly considers the influence of the
the disassembly time, expected disassembly fees, and the following factors: (1) the disassembly state of the product,
desired disassembly energy of the i-th feasible disassem­ which is simulated by the tightening torque of the bolt; (2)
bly sequence are obtained. the two types of disassembly tools are different. The disas­
Step 6: mi1, mi2, and mi3 are the values of the elements sembly tool disassembles the bolt and performs disassembly
that are smaller than t, s, and w in the Ci1, Ci2, and Ci3 time statistics under different experimental conditions. The
arrays; and mNi1 , mNi2 , mNi3 are the maximum time first is to use a ratchet wrench, which is represented by the
612 J. MAO ET AL.

Figure 1. Typical automotive transmission profile. 1, 2, 3, 4 – Bolt I; 5, 6, 7, 8 – Bolt II; 9 – Left End Cap; 10 – Bearing I; 11 – Axis; 12 –
Right End Cap; 13 – Bearing II; 14 – Bushing; 15 – Gear; 16 – Key; 17 – Cover.

tool code I; the second is to remove the plum wrench, which Table 1. Disassembly experiment results.
is represented by the tool code II; (3) the quality of the Data Code Tightening Torque Tool Code Pull Force Time
dismantling staff, the dismantling staff with different max­ 1 25 I 28 18.07
2 50 I 28 18.12
imum pulling forces are used to perform the disassembly, 3 75 I 28 22.34
and the largest Torque is measured with a dynamometer. 4 25 II 28 29.56
5 50 II 28 30.49
Note that the disassembling staff are undertaken by 5 6 75 II 28 34.37
randomly selected students. Since none of them have 7 25 I 24 18.06
professional training, they believe that their disassembly 8 50 I 24 18.56
9 75 I 24 22.38
proficiency is basically the same, and in order to prevent 10 25 II 24 30.77
some additional factors from causing errors in disas­ 11 50 II 24 31.02
12 75 II 24 35.73
sembly time. The impact of these bolts was 4 groups of 13 25 I 20 18.25
disassembly per person, 6 bolts are disassembled per 14 50 I 20 18.94
15 75 I 20 23.17
group, a total of 24 disassembly, and their average 16 25 II 20 29.18
value was taken as the standard disassembly time 17 50 II 20 30.86
18 75 II 20 27.04
under the conditions specified for the bolt. According 19 25 I 15 18.21
to the above disassembly methods, the disassembly time 20 50 I 15 18.80
of the 30 disassembly schemes of the bolt was calculated, 21 75 I 15 23.65
22 25 II 15 33.41
and the statistical results are shown in Table 1. 23 50 II 15 34.71
As can be seen from Table 1, the factors affecting the 24 75 II 15 39.92
25 25 I 12 19.05
disassembly process are qualitative factors. Due to 26 50 I 12 20.17
quantitative factors, in order to better compare the 27 75 I 12 24.63
28 25 II 12 34.05
impact of each factor on the disassembly time, this 29 50 II 12 34.87
article uses a combination of direct observation and 30 75 II 12 40.09
comparative analysis and gray correlation analysis
from disassembly data. The analysis method is used to are both qualitative factors, such as the type of disas­
compare the primary and secondary effects of various sembly tool, and quantitative factors, such as the tigh­
factors. In order to facilitate the comparative analysis of tening torque of the bolt and the pulling force of the
the factors in the disassembly process, the calculation dismantling staff. The analysis of the factors affecting
principle of the gray correlation is introduced as time becomes more difficult, so the combined analysis
follows. method of direct observation and comparative analysis
Based on the results in Table 1, it can be seen that the and gray correlation analysis is used for the disassembly
factors that affect the analysis of bolted disassembly time data described above. With the data results, it can be
 JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION 613

easily seen that when the tools are different, the data
changes greatly, indicating that the impact is greatest.
The changes in the tightening torque and the strength
of the disassembly workers are not easy to distinguish. In
order to quantify the impact of these two factors, and to
analyze the primary and secondary factors of the two
factors, the gray of disassembly tools I and II The asso­
ciation analysis process is demonstrated as follows:
(1) The gray correlation analysis for tool I is as follows:
The results of 15 sets of data of the test of the dis­
assembly tool I with the results of Table 1, combined
with the basic idea of gray correlation analysis, can be
obtained from the analysis of the parent factor; that is,
the disassembly time data is:
y0 ¼ ðy
01 ; y2 ; n; yn Þ 1
18:07; 18:12; 22:34; 18:06; 18:56; 22:38; 18:25;
¼ @ 18:94; 23:17; 18:21; 18:80; 23:65; 19:05; 20:17; A:
24:63
The subfactor has two components, namely the tigh­
tening torque of the bolt and the maximum pulling force
of the disassembler, as follows:
x1 ¼ ðx
� 11 ; x12 ; n; x1n Þ �
25; 50; 75; 25; 50; 75; 25; 50; 75; 25; 50;
¼ ;
75; 25; 50; 75 Figure 2. Multiobjective genetic algorithm flowchart.

x2 ¼ ðx
�21 ; x22 ; n; x2n Þ � the automobile transmission device shown in Figure 2,
28; 28; 28; 24; 24; 24; 20; 20; 20; 15; 15; the disassembly of all the parts can be completed only in
¼ :
15; 12; 12; 12 the horizontal directions, x and -x, and therefore, Equation
(4) can be used. The disassembly priority matrix indicates
Applying the above data and the calculation process
the constraint relationship and disassembly priority rela­
of gray correlation, we can obtain these two sub-
tionship between the various parts of the automotive
factors, namely the tightening torque and the maxi­
transmission.
mum pulling force of the dismantling personnel, and
the correlations with the parent factor, namely the 2 3
0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
disassembly time, are: 0.5815 = 0.5505. From the
60 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 07
above data, it can be known that, for the tool I, the 6 7
60 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 07
effect of the tightening torque on the disassembly time 6 7
60 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 07
is greater than the effect of the pulling force of the 6 7
60 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 07
disassembly personnel. 6 7
60 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 07
Based on the above analysis, the factors that affect the 6 7
60 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 07
size of the bolt removal time can be obtained as follows: 6 7
60 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 07
the type of removal tool > tightening torque > the size of 6 7
W ¼6
60 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 077 (4)
the puller’s pulling force. 60 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 07
6 7
60 0 0 0 0 0 0 0 0 0 0 0 0 0 x x 07
6 7
60 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 07
6 7
Disassembly priority matrix 60 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 07
6 7
60 0 0 0 0 0 0 0 0 0 0 0 0 0 þx 0 07
Before constructing the Petri net model, it is first necessary 6 7
60 0 0 0 0 0 0 0 0 0 0 0 0 0 0 þx 07
to construct a disassembly priority matrix for the various 6 7
40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 15
parts of the automotive transmission. According to actual
work experience in disassembly, it can be seen that for 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
614 J. MAO ET AL.

Figure 3. Automotive transmission expanded stochastic Petri net.

Construction of extended stochastic petri nets meaning of each transition is shown in Table 2, where
model Nðμ; σ 2 Þ is a normal distribution, Uða; bÞis a uniform
Based on the basic theory of Petri nets modeling method, distribution, and Logn(log(μ),σ 2 ) is a lognormal
combined with the composition of typical automobile distribution.
transmission device, referring to the disassembly priority
matrix shown in formula (4), dredging the constraints
Incidence matrix
between parts and disassembly sequence, an extended
stochastic Petri nets model of automobile transmission According to the actual characteristics of automotive
device is constructed, as shown in Figure 3. Among them, transmission device and the construction method of
the places represent the parts to be removed, and the disassembly priority matrix in disassembly Petri nets
transitions represent the removal process and the con­ theory, the Petri nets connection matrix of automotive
straint relationship. The label 2 in Figure 5.1 shows the transmission device shown in Figure 2 is obtained as
four bolts on the left end cap. For convenience, the four shown in the following equation:
bolts are represented by bolt I-1, bolt I-2, bolt I-3, and bolt
I-4, respectively. The same is true for bolt 10.
The place in the Petri network represents the spare Table 2. Meaning of each place.
Place Meaning Place Meaning
parts of the used products. The specific meaning of each
P1 Start P11 Bearing I
place is shown in Table 1. The transition in the Petri net P2 Bolt I-1 P12 Axis
represents the disassembly process of the used product. P3 Bolt I-2 P13 Right End Cap
P4 Bolt I-3 P14 Bearing II
Due to the uncertainty of the disassembly environment, P5 Bolt I-4 P15 Bushing
when assigning the disassembly time, cost, and energy P6 Bolt II-1 P16 Gear
for the disassembly process of each part, this article fits P7 Bolt II-2 P17 Key
P8 Bolt II-3 P18 Cover
the experimental data to obtain the corresponding ran­ P9 Bolt II-4 P19 Finished
dom variables according to actual work experience. The P10 Left End Cap
 JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION 615

2 3
1 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 relevant parameters are set according to the relevant
6 1 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 constraints of minimum expected disassembly fee
6 7
6 1 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 model, maximum disassembly degree model, minimum
6 7
6 1 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 expected cost model and the theory of multiobjective
6 7
6 1 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 7 genetic algorithm, and the optimal solutions are
6 7
6 0 1 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 7 obtained respectively.
6 7
6 0 1 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 7 Genetic algorithm is a kind of mimicking a series of
6 7
6 0 1 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 7 changes in the process of biological evolution, such as
6 7
6 0 1 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 7 chromosome replication, crossover, mutation, etc., and
6 7
A¼6
6 0 0 1 1 1 1 0 0 0 0 10 0 0 0 0 0 0 0 0 77 abstracts the behavior of this chromosome into
6 0 0 0 0 0 0 0 0 0 0 1 10 0 0 0 0 0 0 0 7
6 7 a mathematical model, which is applied to practical
6 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 10 0 0 0 7
6 7 problems.
6 0 0 0 0 0 0 1 1 1 1 0 0 10 0 0 0 0 0 0 7
6 7 When dismantling a waste product with a relatively
6 0 0 0 0 0 0 0 0 0 0 0 0 1 10 0 0 0 0 0 7
6 7 simple structure, it is generally possible to determine
6 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 10 0 0 7
6 7 a feasible disassembly sequence based on actual experi­
6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 10 0 7
6 7 ence, and subjectively determine the optimal disassem­
6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 10 7
6 7 bly sequence. However, it is difficult to subjectively find
4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 15
a feasible disassembly sequence for dismantling disused
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
structures with complex structures. Therefore, it is
(5) necessary to construct a disassembly network graph
Because of the particularity of the transmission device, it model to find a feasible disassembly sequence.
is considered in this article However, when there are many paths for disassembling
that when disassembling the parts on the transmis­ the target parts, it is difficult to obtain a specific optimal
sion device of an automobile, all the parts on it can be disassembly sequence.
disassembled only by disassembling along the horizontal In order to solve the multiobjective stochastic com­
direction. prehensive optimization model with the lowest cost, the
shortest time, and the least energy, the thesis designs
a multiobjective optimization algorithm based on
Disassembly state identification vector genetic algorithm to solve the optimal disassembly
sequence.
Each disassembly state of the product can be represented
by the corresponding disassembly state identification
vector. The initial disassembly state identification vector Definition of the parameters
is set to M0, and the final disassembly state identification
The main consideration is to construct a comprehensive
vector of the product is set to M after the disassembly
optimization model with the shortest disassembly time,
task is completed. Each disassembly operation must not
the lowest disassembly cost, and the lowest disassembly
violate the disassembly constraints between parts.
energy under the premise of meeting the disassembly
According to the definition of disassembly state iden­
constraints.
tification vector, the initial disassembly state identification
vector M0 of the product to be disassembled is as follows:
Minimum expected disassembly time model
~ 0 ¼ ½1; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0�T :
M In a given random disassembly network diagram, if
there is a disassembly path from the disassembly starting
If the object of disassembly is the part 17-case cover in
point to the target disassembly node, for any other dis­
Figure 2, the final disassembly state identification vector
assembly path, the following inequality holds, namely:
M of the product is as follows:
E½Tðt; xÞ� � E½Tðt; x0 Þ�:
~ ¼ ½0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 1�T :
M
The path is called the minimum expected value of the
disassembly of the product from the disassembly start
point to the target disassembly end point, where there is
Solution method for model
the expected value of the disassembly time required for
In this article, the object of disassembly is the part 17- the target disassembly along the path, that is, the average
case cover-in the automobile transmission device shown disassembly required for the target part disassembly
in Figure 2. In order to get part 17 by disassembly, along the path. The model that solves the minimum
616 J. MAO ET AL.

expected value path is called the disassembled minimum target disassembly node. Wherein, the disassembly time
expected value model. of the feasible disassembly path at the disassembly
The objective function for disassembling the mini­ degree D is given. The model that solves the D-shortest
mum expected value model is: disassembly time path is called the D shortest disassem­
X bly time model.
min E½ tij xij �: For the convenience of solving, the mathematical
i;j
expression of the D-shortest disassembly time model is
The constraints it meets are as follows: now given, which is as follows. According to the related
8 concepts of the opportunity constraint programming
< xij 2 f0; 1g
model, the objective function of the model can be
"i; j 2 V :
: expressed as
"ði; jÞ 2 E
min T :

The constraints it meets are as follows:

Maximum disassembly model
8
In a given random disassembly network diagram, if > � �D
PfTðt; xÞ � Tg
>
<
there is a disassembly path from the disassembly starting xij 2 f0; 1g
point to the target disassembly node, for any other dis­ :
> "i; j 2 V
>
assembly path, the following inequality holds: :
"ði; jÞ 2 E
PfTðt; xÞ � T 0 g � PfTðt; x0 Þ � T 0 g:
The disassembly energy can be expressed by power
The path is said to be the maximum disassembly because of the disassembly of the old vehicle transmis­
degree path from the disassembly start point to the sion. At the same time, according to the principle of
target disassembly node. The model that solves the stochastic simulation algorithm and the constraints of
maximum disassembly degree path is called the max­ planning model, the parameters of the algorithm are
imum disassembly degree model. Among them are calibrated as follows:
the disassembly time of a given product, and the Times of stochastic simulation N = 300,000;
disassembly path and the disassembly degree at Initial state identification vector: M0
a given time, respectively. = ½1; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0�T ;
Therefore, the objective function of the maximum The target disassembly status identification vector is
disassembly model of the disassembly problem can be set as M =½0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 0; 1�T ;
expressed as follows: Given disassembly time in maximum disassembly
degree model t = 220s, disassembly fee s = CNY40, dis­
max PTðt; xÞ � T 0 :
assembly energy w = 0.085 kW●h;
The constraints it meets are as follows: Given probability in minimum cost disassembly
8 model p1 = p2 = p3 = 95%;
< xij 2 f0; 1g
The correlation matrix A is shown in formula (5).
"i; j 2 V :
: Using the multiobjective disassembly optimization
"ði; jÞ 2 E
algorithm based on stochastic simulation, the three pro­
gramming models proposed in this article are solved,
and the results are listed in the next section.
D–Minimum disassembly time model
In a given random disassembly network diagram, if
there is a disassembly path from the disassembly starting
point to the target disassembly node, for any other dis­ Solutions
assembly path, the following inequality holds: According to the constraint equation, ~ A~ uk ¼ M ~ M ~ 0,
� the decision vector U is solved as follows:
�
minfT �PfTðt; xÞ � T g � Dg U1 =½1; 0; 1; 1; 1; 1; 0; 0; 0; 0; 1; 1; 0; 0; 1; 0; 0; 0; 1; 1�T ;
�
� U2 =½1; 0; 1; 1; 1; 1; 0; 0; 0; 0; 1; 1; 0; 0; 0; 1; 0; 1; 1; 1�T ;
� min PfT �PfTðt; xÞ � T g � Dg:
U3 =½0; 1; 0; 0; 0; 0; 1; 1; 1; 1; 0; 0; 1; 1; 0; 0; 1; 1; 1; 1�T °
The path is called the D-minimum time model that is The disassembly sequences corresponding to the
disassembled from the disassembly starting point to the above three decision vectors are as follows:
 JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION 617

Sequence 1: P1–><P2, P3, P4, P5>–>P10–>P11–

>P12–>P17–>P18–>P19;
Sequence 2: P1–><P2, P3, P4, P5>–>P10–>P11–
>P12–>P16–>P17–>P18–>P19;
Sequence 3: P1–><P6, P7, P8, P9>–>P13–>P14–
>P15–>P16–>P17–>P18–>P19.
The probability density curves of disassembly time
corresponding to disassembly Sequences 1, 2, and 3 are
shown in Figure 4, respectively.
The disassembly fee probability density curves corre­
sponding to disassembly sequences 1, 2, and 3 are shown in
Figure 5.
The disassembly energy probability density curves
corresponding to disassembly sequences 1, 2, and 3 are
shown in Figure 6.
According to the objective function and constraints Figure 6. Probability density curve of the disassembly energy.
of the model at the minimum expected cost, the Pareto

solution set of the disassembly sequences of the vehicle

transmission device shown in Figure 2 can be obtained
as shown in Table 4.
According to the objective function and maximum
disassembly degree model constraints and given the dis­
assembly times, t1 = 220 s and t2 = 200 s, disassembly feess,
s1 = ¥ 40 and s2 = ¥ 35, and disassembly energy usages, w1
= 0.090 kW.h and w2 = 0.085 kW.h, the probabilities that
the disassembly task can be completed at a given cost are
shown in Table 5.
According to the objective function, maximum disas­
sembly degree model constraints, and given the disassem­
bly times, disassembly feess and disassembly energy
usages, the probabilities are p1 = p2 = p3 = 95% and p1ʹ
= p2ʹ = p3ʹ = 90%. The minimum costs of the three feasible
Figure 4. Probability density curves for the disassembly times. disassembly sequences given the disassembly probabilities
are shown in Table 6.

Conclusion
The results of Tables 3, 4, and 5 show that the objective
functions of disassembly sequences 1, 2, and 3 do not
dominate each other. Therefore, the three disassembly
sequences are Pareto solutions of the minimum
expected disassembly fee model, the maximum disas­
sembly degree model, and the minimum cost model
proposed in this article.
When the decision-makers have different prefer­
ences, different disassembly optimization results will be
obtained when solving the disassembly sequence of
waste products. In this article, the minimum disassem­
bly fee model, maximum disassembly degree model, and
minimum disassembly fee model are proposed so that
Figure 5. Probability density curves for the disassembly fees. decision makers can choose different decision schemes
618 J. MAO ET AL.

Table 3. Meaning of each transition. according to market demand and internal needs of
Removal Removal Removal Energy enterprises. More intelligent methods is further to test
Transition Meaning Time (s) Fees (¥) (10−3kW●h)
the proposed models.
T1 Ready N (3,1) N (1,0.2) U (0.4,0.6)
T2 Ready N (3,1) N (1,0.2) U (0.4,0.6)
T3 Remove bolt Logn(log Logn(log N (8,4)
I-1 (15),0.5) (4),0.3) Acknowledgment
T4 Remove boltI- Logn(log Logn(log N (8,4)
2 (15),0.5) (4),0.3)
T5 Remove boltI- Logn(log Logn(log N (8,4)
The authors acknowledge Science and Technology Research
3 (15),0.5) (4),0.3) Program of Education Department of Jilin Province (Grant
T6 Remove boltI- Logn(log Logn(log N (8,4) No. JJKH20190239SK) for supporting this study.
4 (15),0.5) (4),0.3)
T7 Remove Logn(log Logn(log N (8,4)
boltII-1 (15),0.5) (3),0.2)
T8 Remove Logn(log Logn(log N (8,4) Funding
boltII-2 (15),0.5) (3),0.2)
T9 Remove Logn(log Logn(log N (8,4) This work was supported by the Education Department of Jilin
boltII-3 (15),0.5) (3),0.2) Province [JJKH20190239SK].
T10 Remove Logn(log Logn(log N (8,4)
boltII-4 (15),0.5) (3),0.2)
T11 Remove left U (4,6) U (0.8,1.0) Logn(log(3.5),0.8)
End Cap About the authors
T12 Remove Logn(log Logn(log N (10.5,0.64)
bearing I (20),0.6) (4),0.2) Jia Mao was born in Nanguan District, Changchun, JILIN,
T13 Remov right U (4,6) U (0.8,1.0) Logn(log(3.5),0.8)
End Cap China, in 1973. He received B.S. and M.S. degrees in Vehicle
T14 Remove Logn(log Logn(log N (10.5,0.64) Engineering from the Jilin University of Technology,
bearing II (20),0.6) (4),0.2) Changchun, in 1997 and a Ph.D. degree in Communication
T15 Remove axis N (10,0.8) N (2,0.3) U (3,6) and Information Systems from Jilin University, Changchun,
T16 Remove axis N (10,0.8) N (2,0.3) U (2,6)
T17 Remove N (12,0.7) N (3,0.2) N (5,2)
JILIN, in 2004. From 2005 to 2008, he was a Lecturer with the
bushing Logistics Engineering Laboratory of Jilin University. Since
T18 Remove gear N (6,0.8) N (1.5,0.2) U (4,8) 2008, he has been an Associate Professor with the School of
T19 Remove key N (8,1) N (3,0.4) N (4,1) Transportation, Jilin University. Now, he has been an
T20 Remove cover N (5,0.5) N (1,0.1) U (2,3) Professor with the School of Transportation, Jilin University.
From 2017 to 2018, as a visiting scholar from the Ministry of
Education of China, he studied at Oakland University in the
USA. He is the author of two books, more than 80 articles, and
Table 4. Solution of the minimum expected cost disassembly more than 20 inventions. More than 30 of his articles have
model. been indexed by the ISI or EI. He also holds three patents. His
Expected research interests include logistics information technology,
Expected Expected Disassembly traffic information engineering, big data analysis, and innova­
Sequence Disassembly Disassembly Energy
tion internet applications. Dr. Mao received the Best Paper
No. Time (s) Fee (CNY) (10−3kW●h)
Award from the Chinese Society of Logistics Symposium in
Sequence 1 117.4864 33.6538 61.4924
Sequence 2 123.4864 35.1538 59.9926 2008.
Sequence 3 125.4864 36.1538 56.9926 Dou Hong was born in Hanzhong City, Shaanxi Province in
1996. She obtained a bachelor's degree in engineering from
Shaanxi University of Science and Technology. She currently
Table 5. Solution of the maximum disassembly degree model. holds a master's degree in engineering from Jilin University.
Limited
Since 2019, she has been a graduate student in the Jilin
Limited Limited Disassembly University Logistics Engineering Laboratory. She is the author
Sequence Disassembly Disassembly Energy: of two papers, who has assisted multiple research projects, and
No. Time: 220 s Fee: CNY40 0.085 kW●h has achieved outstanding academic results during the school.
Sequence 1 97.4287% 96.8190% 92.1220%
Sequence 2 96.9820% 95.1833% 93.4010% Zhe Chen was born in Minqin City, Gansu Province, China in
Sequence 3 96.7790% 93.8940% 95.3320% January 2000. He is now studying in School of Transportation,
Jilin University. His main research interests include supply
chain management and reverse logistics. He has published two
academic papers in public.
Table 6. Solution of the minimum disassembly cost model. Ma Changhai was born in Hulunbuir, Inner Mongolia, China
Minimum in November, 2000. He is now studying in School of
Minimum Minimum Disassembly Transportation, Jilin University. His main research interests
Sequence Disassembly Disassembly Energy include logistics information technology and reverse logistics.
No. Time (s) Fee (CNY) (10−3kW●h)
He has published three academic papers in public.
Sequence 1 197.6627 38.2920 88.9595
Sequence 2 203.1698 39.7808 87.3564 Li Weiwen was born in Huanggang City, Hubei Province,
Sequence 3 205.2454 40.7971 84.5262 China in 1996. He received a bachelor's degree in engineering
 JOURNAL OF THE AIR & WASTE MANAGEMENT ASSOCIATION 619

from the Harbin Institute of Technology, and a master's Intell. Transport. Syst. 18(9):2557–68. doi: 10.1109/
degree in engineering from Jilin University. Now, he is work­ TITS.2016.2638898.
ing in Shanghai Automobile Gear Works Co., Ltd. His Kroll, E., and B. Carver. 1999. Disassembly analysis through
research interests include system optimization and reverse time estimation and other metrics. Robot.Cim-Int. Manuf.
logistics. He has published three academic papers, including 15 (3):191–200. doi:10.1016/S0736-5845(99)00026-5.
a journal article indexed by SCI. Lambert, A. J. D. 2006a. Generation of assembly graphs by
systematic analysis of assembly structures. Eur. J. Oper. Res.
Ju Wang received the B.S. degree in Environmental Planning
168:932–51. doi:10.1016/j.ejor.2004.07.038.
and Management from Jilin University, China, in 1993, the M.
Lambert, A. J. D. 2006b. Exact methods in optimum disas­
S. degree and Ph.D. degree in Environmental Science, from
sembly sequence search for problems subject to sequence
Jilin University, China, in 1996 and 2002. She is currently a
dependent costs. Omega-Int. J Manage. S. 34:538–49.
Professor in Department of Environmental Science, College of
doi:10.1016/j.omega.2005.01.005.
New Energy and Environment in Jilin University. Her
Lambert, A. J. D. 2007. Optimizing disassembly processes
research interests include Environmental Planning and
subjected to sequence dependent cost. Comput. Oper. Res.
Management and Environmental Economics.
34:536–51. doi:10.1016/j.cor.2005.03.012.
Li, J. R., L. P. Khoo, and S. B. Tor. 2005. An object-oriented
intelligent disassembly sequence planner for maintenance.
References Comput. Ind. 56:699–718. doi:10.1016/j.compind.2005.03.005.
Tang, Y., M. C. Zhou, and M. Gao. 2006. Fuzzy-Petri-net
Adenso-Díaz, B., S. García -arbajal, and S. M. Gupta. 2008. A
based disassembly planning considering human factors.
path-relinking approach for a bi-criteria disassembly
IEEE T Syst. Man. Cy. A. 36:718–26. doi:10.1109/
sequencing problem. Comput. Oper. Res. 35:3989–97.
TSMCA.2005.853508.
doi:10.1016/j.cor.2007.06.002.
Tian, G., Y. Ren, Y. Feng, et al. 2019, April. Modeling and
Aguinaga, I., D. Borro, and L. Matey. 2008. Parallel RRT-based
planning for dual-objective selective disassembly using
path planning for selective disassembly planning. Int. J Adv.
and/or graph and discrete artificial bee colony. IEEE
Manuf. Tech. 36:1221–33. doi:10.1007/s00170-007-0930-2.
Trans. Ind. Inf. 15 (4):2456–68. doi:10.1109/
Gao, M., M. C. Zhou, and Y. Tang. 2004. Intelligent decision
making in disassembly process based on fuzzy reasoning TII.2018.2884845.
petri netss. IEEE T Syst. Man. Cy. B. 34:2029–34. Tian, G., M. Zhou, and J. Chu. 2013. A chance constrained
doi:10.1109/TSMCB.2004.833331. programming approach to determine the optimal disassem­
Grochowski, D. E., and Y. Tang. 2009. A machine learning bly sequence. IEEE T Autom. Sci. Eng. 10:1004–13.
approach for optimal disassembly planning. Int. J.Comput. doi:10.1109/TASE.2013.2249663.
Integ. M. 22:374–83. doi:10.1080/09511920802024176. Tian, G., M. Zhou, J. Chu, and Y. Liu. 2012. Probability
Homem de Mello, L. S., and A. C. Sanderson. 1991. A correct evaluation models of product disassembly cost subject
and complete algorithm for the generation of mechanical to random removal time and different removal labor
assembly sequences. IEEE T Robotic. Autom. 7:228–40. cost. IEEE T Autom. Sci. Eng. 288–95. doi:10.1109/
doi:10.1109/70.75905. TASE.2011.2176489.
Hou, Y., N. Wu, M. C. Zhou, and Z. Li. 2017, March. Pareto- Tian, G., M. C. Zhou, P. Li. 2018, April. Disassembly sequence
optimization for scheduling of crude oil operations in refin­ planning considering fuzzy component quality and varying
ery via genetic algorithm. IEEE Trans. Syst.| Man Cybern. operational cost. IEEE Trans. Autom. Sci. Eng. 15
Syst. 47(3):517–530. doi: 10.1109/TSMC.2015.2507161. (2):748–60. doi:10.1109/TASE.2017.2690802.
Kang, Q., S. W. Feng, M. C. Zhou, A. C. Ammari, and Zussman, E., and M. C. Zhou. 2000. Design and implementa­
K. Sedraoui. 2017, September. Optimal load scheduling of tion of an adaptive process planner for disassembly
plug-in hybrid electric vehicles via weight-aggregation processes. IEEE T Robotic. Autom. 16:171–79.
multi-objective evolutionary algorithms. IEEE Trans. doi:10.1109/70.843173.