A Proactive Material Handling Method For CPS Enabled Shop-Floor
A Proactive Material Handling Method For CPS Enabled Shop-Floor
A Proactive Material Handling Method For CPS Enabled Shop-Floor
com/science/article/pii/S0736584518306343
Manuscript_16d35506b1b99d89f9ba5bfd16ef7495
a
Key Laboratory of Contemporary Design and Integrated Manufacturing Technology,
Ministry of Education, Northwestern Polytechnical University, Xi’an 710072, China;
b
Department of Mechanical and Energy Engineering, Southern University of Science and
Technology
c
Department of Industrial and Manufacturing Systems Engineering, The University of Hong
Kong, Hong Kong, China
*
Corresponding author
Abstract
Cyber physical system (CPS) enables companies to keep high traceability and controllability
in manufacturing for better quality and improved productivity. However, several challenges
including excessively long waiting time and a serious waste of energy still exist on the shop-
floor where limited buffer exists for each machine (e.g., shop-floor that manufactures large-
size products). The production logistics tasks are released after work-in-processes (WIPs) are
processed, and the machines will be occupied before trolleys arrival when using passive
material handling strategy. To address this issue, a proactive material handling method for
CPS enabled shop-floor (CPS-PMH) is proposed. Firstly, the manufacturing resources
(machines and trolleys) are made smart by applying CPS technologies so that they are able to
sense, act, interact and behave within a smart environment. Secondly, a shop-floor digital
twin model is created, aiming to reflect their status just like real-life objects, and key
production performance indicators can be analysed timely. Then, a time-weighted multiple
linear regression method (TWMLR) is proposed to forecast the remaining processing time of
WIPs. A proactive material handling model is designed to allocate smart trolleys optimally.
Finally, a case study from Southern China is used to validate the proposed method and results
show that the proposed CPS-PMH can largely reduce the total non-value-added energy
consumption of manufacturing resources and optimize the routes of smart trolleys.
© 2019 published by Elsevier. This manuscript is made available under the Elsevier user license
https://www.elsevier.com/open-access/userlicense/1.0/
Keywords
Cyber physical system (CPS), Material handling, Shop-floor, Prediction model, Remaining
processing time, Large-size product
Abbreviations
AGV Automatic guided vehicle MLR Multiple linear regression
AR Augmented reality MRP Material requirements planning
CEP Complex event processing NSGA-II Non-dominated sorting genetic
CPS Cyber physical system algorithms-II
CNC Computer numerical control NVD Non-value-added
CPPS Cyber physical production system PI Physical Internet
CPS-PMH Proactive material handling for CPS- PT Processing time
enabled shop-floor PLS Production and logistics system
CSA Centralized storage area RFID Radio frequency identification
DT Digital twin ROI Return on investment
EIS Enterprises information system RPT Remaining processing time
EPL Event processing languages SoLS Self-organising logistics system
IoT Internet of Things TWMLR Time-weighted multiple linear
IoMT Internet of Manufacturing Things regression
KPI Key performance indicator VD Value-added
LR Linear regression WIP Work-in-process
1. Introduction
Nowadays, the emerging cyber physical system (CPS) presents a significant opportunity to
implement smart manufacturing [1]. The CPS can assist manufacturing systems with
advanced technologies, including Internet of Things (IoT) [2,3], big data analytics [4–6],
cloud computing[7], etc. Typical manufacturing resources are thus converted into smart
objects using these technologies so that they are able to sense, act and behave within a smart
environment. In addition, CPS is a collection of modelling, simulation and optimisation [8,9],
and predictive analytics [10,11], which supports manufacturing applications make reasonable
decisions accurately. As a result, CPS enables companies to cope with increasingly
customized products with a shorter lead-time to market and higher quality [12].
Manufacturing sectors are striving to upgrade their current production decision-making
modes to a more transparent, lean and intelligent level, and many new production paradigms
are proposed, such as digital twin shop-floor [13,14], Internet of Manufacturing Things
(IoMT) [15], intelligent Physical Internet (PI) containers [16,17], cloud manufacturing [18,19]
and cloud-based cyber physical production system (CPPS) [20,21].
Material handling on shop-floor is critical to ensure that each machine or workstation is
being fed with the right material in the right quantity and quality at the right point in time. It
has gained great attention to keep a high degree of production fluency. For example, Teruaki
Ito and Abadi proposed an agent-based material handling and inventory planning method in a
warehouse [22]. Chakraborty and Banik designed a material handling equipment selection
model using analytic hierarchy process [23]. Khayat et al. proposed an integrated production
and material handling scheduling method [24]. Pan et al. proposed a self-organising logistics
system (SoLS) which can function without significant intervention by humans and central
control by software [25]. Recently, by applying the CPS technologies in the manufacturing
environment, several real-time material handling strategies are proposed. Ren et al. proposed
a framework for shop-floor material delivery based on real-time manufacturing big data [26].
Zhang et al. proposed an optimization method for shop-floor material handling method,
where each trolley is an active entity which will request the transport tasks [27]. Qu et al.
discussed an IoT-based real-time production logistics synchronization system under smart
cloud manufacturing [28].
Based on these works, the material handling efficiency can be improved from three aspects:
1) real-time and multi-source manufacturing data has become accessible and ubiquitous,
which can provide accurate and comprehensive information support for production logistics
decision-making. 2) production disturbances (such as logistics tasks change and movement
equipment breakdowns) can be detected timely and responded rapidly. 3) material movement
trolleys are made ‘smart’, which can communicate with the management system actively,
execute production logistics decisions accurately and adjust their status dynamically, make
optimal and reliable material handling decisions using intelligent algorithms.
Despite significant progress achieved by the researchers in the field of material handling,
major challenges still exist in a CPS enabled shop-floor. They are summarized as follows:
1) Most of existing material handling systems use passive strategies. The delivery tasks
are always released after work-in-processes (WIPs) are processed, and an excessively
long waiting time and energy will be wasted before the tasks are assigned and executed
by smart trolleys. For the shop-floor that manufactures large-size products, which often
set limited buffer, it is challengeable for material handling. For example, a tire mould
manufacturing factory always set only one centralized storage area (CSA), each WIP
has to be delivered to the CSA or next workstation directly after being processed,
otherwise the workstation will be occupied.
2) Some assumptions, such as constant processing time and continuous facility
availability, are usually made when making material handling decisions. However,
some assumptions are not reasonable since there are many uncertain factors in real-life
manufacturing systems. For example, the proficiency of different workers often varies
largely due to the difference among their working ages and motivations, which may
further lead to different setup time and manufacturing time of a same process. Besides,
machines cannot be available all the time because of the unavoidable occurrences of
unexpected machine breakdown. How to combine real-life production uncertainties in
production decision-making so that the allocation can be more practical should be
further discussed.
To tackle these problems, this paper proposed a proactive material handling method for
CPS enabled shop-floor (CPS-PMH). It integrates three important components. The first is a
CPS-enabled shop-floor where typical resources are made smart which can achieve the
capability of identifying their real-time status and interacting with the upper management
system. The second is a remaining processing time prediction method for WIPs, which can
combine historical manufacturing data and real-time manufacturing state simultaneously in
the prediction model. The third is the proactive material handling strategy, which can
proactively assign logistics tasks to smart trolleys with minimized transport distance of
trolleys and non-value-added energy consumption of machines and trolleys. After the
descriptions of the overall architecture and key technologies of CPS-PMH, a case study from
Southern China is used to validate the feasibility and effectiveness of the proposed method.
The designed material handling model will provide a new paradigm for manufacturing
enterprises to implement more efficient material handling.
The rest of the paper is organized as follows. Section 2 reviews the related literature.
Section 3 outlines the proposed CPS-PMH. Section 4 presents the construction method of
shop-floor digital twin. Section 5 describes the overall solution for predicting the remaining
processing time of WIPs. Section 6 presents the overall solution of proactive material
handling strategy. In section 7, a case study is used to illustrate the implementation of the
proposed CPS-PMH. Conclusions and future works are given in Section 8.
2. Literature review
Three streams of the literature are relevant to this research. They are CPS and their
application in manufacturing, material handling in shop-floor, and processing time prediction.
3. Overview of CPS-PMH
This research focuses mainly on discrete manufacturing shop-floors which have a centralized
storage area and limited buffer around each workstation. The objective of the proposed
method is to change traditional passive material handling strategy by using predictive
analytics and CPS technologies.
The core of shop-floor digital twin model is to give a mirror of physical shop-floor, which
can change the status of its elements with real-life production activities and convey
production decisions to corresponding smart objects timely. Manufacturing data is a main
connection between the two layers. Since raw data sensed in physical shop-floor often carries
many unrelated, duplicate information, it is necessary to analyse production KPIs. Then the
status of shop-floor digital twin can be updated easily with the changes of production KPIs.
Fig. 3 shows the main steps for the construction of shop-floor digital-twin model, which
consists of four main steps, real-time data modelling, data classification, data integration, and
KPIs extraction.
End
In the traditional MLR method, the historical observations are weighted equally. However, in
a real-life production system, the operations of production resources evolve with time passed
by, for example, the senior workers can be more efficient than young ones. Thus, recent
samples often reflect more information than old ones. Meanwhile, exponential smoothing is
one of many window functions commonly applied to smooth time-series data. This paper
uses the exponential functions to assign exponentially decreasing weights over time. The
work logic of TWMLR for predicting remaining processing time is described as follows.
1) New case
In order to represent a new prediction problem, the real-time information obtained by the KPI
analysis is denoted as a new prediction case (N’). The purpose of representing a problem
through a case style matrix is to reflect how a decision-maker would structure a new problem.
A new case is defined as
N ' = { x1' , x2' ,..., x 'j ,..., xn' } (1)
where, n refers to the total number of attributes; x 'j refers to the value of jth attributes, e.g.,
denotes the real processing time of the ith case; T i denotes the occurrence time of case N i ' ; i
refers to the case ID; x ij' refers to the value of jth attributes.
Since the attributes may have a large range of values and different attributes vary in their
types, a data standardization is necessary before creating a data matrix. In general, Z-score
normalization or min-max scaling are the most popular data standardization methods, this
paper selects the min-max normalization to normalize the raw data sets, as follows
x ij' − µ j
x ij = (2)
x 'j ,max − x 'j ,min
where, xij and x ij’ denote the standardized value and raw value of the jth attributes of case,
respectively; µ j , x'j ,max and x 'j ,min denote the average value, maximum value, and minimum
value in the data set, respectively. Based on the data standardization, the data set can be
changed to M = { N i , y i , T i } , where N i = { x1i , x2i ,..., x ij ,..., xni } .
......
T c −Ti
where, WN i = denotes the time windows number between ith case and current time,
tw
T c denotes the current time; t w denotes the time window that the weight changes, which can
be defined according to real-life experience.
4) TWMLR model calculation
After a set of cases is collected and time weights are added to each case, the data set can
be analysed to find a prediction equation which can be described as
hθ ( x) = θ0 + θ1 x1 + ⋅⋅⋅ + θ j x j ⋅⋅⋅ +θn xn (4)
where, hθ (x) refers to the predicted processing time of a case; θ j denotes the parameters for
jth attributes.
Compared with traditional MLR method, the proposed model uses a time-weighted cost
function, which can be described as follows
1 m i n
J (θ ) = ∑
2 m i =1
w (∑ θ j x j ( i ) − y ( i ) ) 2
j =0
(5)
where, J (θ ) denotes the cost function with a set of θ . m denotes total number of cases.
Then, the gradient descent technologies are used to find a fitted model, repeat until
convergence, the equations are listed as follows
∂
θ j := θ j − ∂ J (θ 0 , θ1 ) (6)
∂θ j
where, Tr denotes the predicted remaining processing time; Tc denotes the current time; Ts
denotes the start time of the new case.
6. Proactive material handling method
In this section, a proactive shop-floor material handling method is designed. It aims to find an
optimal combination between delivery tasks and smart trolleys according to their both future
status and current information. Since a huge number of logistics tasks exist in a
manufacturing shop-floor, a two-stage optimization method is designed to reduce the
complexity of the tasks assignment and improve the solving efficiency for real-time decisions.
In this two-stage optimization method, the first stage is used to obtain a candidate task set
from whole move tasks. The second stage is used to assign optimal tasks to smart trolleys.
If the task is delivered from CSA, it will be set as zero; EM i2 is used to define energy cost
that will occur if the task arrives after Di. If the tasks are delivered from CSA, it will be set as
zero.
here, f denotes the selected task number. Then, the smart trolleys whose available time before
the released time of any task will be selected and stored in a matrix Tr, as
Tr = V 1 , V 2 ,..., V e ,..., V p , here, p is the total number of available trolleys and e denotes the
trolley number.
After the top q tasks and p available trolleys are selected, the NSGA-II based optimization
method will be involved to find an optimal allocation of trolleys for the material handling
tasks.
Fdis tan ce = Min(∑e=1 ∑ ge=1 (( FLg − ALe ) + ( FLg +1 − TLg ) + (TLg − FLg )))
p q
(10)
Objective function 2:
Etotal = Min( Eoccupation
machine
+ Eidle
machine
+ Eoperational
trolley
+ Eidle
trolley
)
=Min(∑ e =1 ∑ ge=1 ( EM g1 max(0, Te,start
p q
g − Rg ) + EM g max (0, Te , g − Dg ))
2 finish
(11)
+ ∑ e =1 ∑ ge=1 ( Eeidle ((TFL ,1 − TAL ,e ) + (TFL , g +1 − TTL , g )) + ( Eeoperate (TTL , g − TFL , g ))))
p q
Subject to:
,( g +1) > Te , g
Testart , e = 1, 2,..., p, g = 1, 2,..., qe (12)
finish
Testart
,g ≥ ATe , e = 1, 2,..., p, g = 1, 2,..., qe (13)
Testart
,g ≥ Rg , e = 1, 2,..., p, g = 1, 2,..., qe (14)
Here, qe denotes the total task number that assigned to eth trolley; g denotes the task
number; Te,start
g denotes the time that eth trolley starts to move gth task; Tg
finish
denotes the time
that task is finished; TFL ,g and TTL ,g represent the from and to time of gth task.
In the above formulations, two objectives are used to determine a schedule that minimizes
the total routing distance of smart trolley Fdistance (Eq. (10)) and minimizes the total energy
consumption of non-value-added activities Etotal (Eq. (11)) which consists of the four parts:
occupation energy consumption of last machine, idle energy consumption of next machine,
idle and operational energy consumption of smart trolleys moves without and with tasks.
Constraint (12) ensures that no trolley can move more than one tasks at a time, a new task
need to load after last task is unloaded. Constraint (13) takes care of the requirement that the
start moving time of tasks should after the available time of each trolley. Constraint (14)
ensures that the start moving time of each task should after their release time.
After the problem is defined, the NSGA-II algorithm is used to optimise the allocation. The
NSGA-II algorithm consists of seven main steps which are briefly described as follows.
Step 1: Construct a parent population PoP0 whose size is pop. Each individual of PoP0 can
be seen as a chromosome which consists of q genes. Each gene consists of an integer
e (1 ≤ e ≤ p ) . Based on the genes and chromosome, the decoding schema is also designed.
The integer ‘e’ appeared in chromosome presents the trolley e. The appeared location of
integer ‘e’ refers to the tasks that are assigned to the eth trolley; p is the total available
trolleys; q refers to the total task number. The appeared order of integer ‘e’ also presents
the move sequence of the assigned tasks. For example, the chromosome ‘1-2-1-2’ presents
the processing order of four tasks assigned to two trolleys, i.e., trolley ‘1’ will move tasks
‘1’ and ‘3’ successively, and trolley ‘2’ will move tasks ‘2’ and ‘4’ in turn.
Step 2: Generate one random number between 0 and 1. If the number exceeds a predefined
probability (e.g., 0.25), then adopt the crossover operator; else use the mutation operator.
The crossover operator is used to obtain offspring by recombining parents. In this case, we
adopt a multi-point crossover operator to permute the trolley order as well as keep the legality
of the generated chromosome. Firstly, choose two parents (A and B). Secondly, randomly get
two crossover points to form two crossover-sections. At last, exchange the genes in crossover
sections between the parents.
The mutation operator works on a single chromosome and generates offspring by altering
one or more genes.
Step 3: Combine the offspring and parent population to create a new population whose
size is 2×pop. Calculate the fitness functions of each individual. In this paper, the
objective functions are used as the fitness functions to evaluate the chromosomes.
Step 4: Sort the individuals according to non-dominated sorting procedure which ranks the
solutions into different Pareto fronts.
Step 5: As to the individuals in the same level, crowding distance sorting procedure is
used to rank them. The crowding distance can be calculated based on the following
equation:
where, dispi represents the distance of pith point Ppi; pm represents the total number of
fitness functions, it equals 2 in this paper; f pipj+1 and f pipj-1 represent the value of fitness
respectively; f x1 and f x2 are the specific values of fitness functions of the individual.
After an optimal allocation decision is obtained, the trolleys will receive the decisions and
execute their assigned tasks accordingly. If an exception happens (e.g., a trolley breaks, a task
cancels), the consequence will be evaluated in time, and a new round of optimization will be
involved if the impact of exception will largely influence the normal operation of production
system.
7. Case study
Table 1
A fragment of dataset.
When constructing a TWMLR model for predicting the processing time of a specific case,
only the cases whose occurrence time before the case are used, for example, when forecasting
the 60th case, only previous 59 cases are used. As a result, the time weights need to be re-
calculated, and the prediction model need to be updated when analysing different cases.
Equation (17) presents the TWMLR model learned from the whole dataset.
Y = 21.3439 + 23.3184 x1 + 13.6224 x2 + 9.7926 x3 (17)
To demonstrate the effectiveness of TWMLR, the raw MLR method which do not give the
samples time-related weights is used for comparison. The obtained model is as follows
Y = 23.4634 + 31.3636 x1 + 21.6269 x2 + 15.2050 x3 (18)
Fig. 6 shows the prediction results of the two different models. Fig. 6(a) shows the
comparison between real value and the two predictions. It can be found that most of the
predictions from TWMLR which considers time-weight of cases are more approximate to the
real values than the non-time-weighted method (raw MLR method). Fig. 6(b) shows the
residual error of the two different prediction models. It can be found that the proposed
method performs worse for the first several cases, and the prediction accuracy increases with
time passed by. For the latest several cases, the proposed method performs good enough than
non-time-weighted method. Based on these analyse, it can be concluded that the TWMLR-
based prediction gives more accurate prediction than raw MLR method, and with time passed
by, the prediction becomes more and more reliable.
Table 3
Task ID 1 2 3 4 5 6
Available time 4 1 0 1 1 0
Available location 90, 20 66,40 17,40 83,60 50,60 17,20
Based on the two datasets, an optimal allocation for material handling can be found using
the NSGA-II algorithm. The maximum number of iterations is set as 400. The results are
shown in Fig. 7. Fig. 7(a) shows the Pareto front of the optimization which presents the
relationship between the distance of smart trolleys and energy consumption. From the Pareto
front, it can be seen that the solution could lead to a lower distance with larger energy
consumption when a higher weight is assigned to trolley distance. In contrast, if a higher
weight is assigned to energy consumption, the solution could have a lower energy
consumption with a higher distance. As a result, the managers need to make a balance
between logistics distance and energy consumption according to their actual requirements.
When the default output set is used, the Gantt chart of the optimal schedule is shown in Fig.
7(b). The horizontal axis gives the time and vertical axis refers to the task ID. The numbers
embedded in the coloured bars of Fig. 7(b) presents the assigned trolley for each task. After
an allocation decision is obtained, the information about the tasks can be transferred to each
trolley, and the trolley will deliver the materials at the specific time points.
Fig. 7. Optimal allocation results of the proposed case
To illustrate the efficiency of the proactive material handling strategy, the passive material
handling strategy is simulated to give a comparison. In the passive material handling strategy,
the material handling tasks can only be released to the system after the WIPs are processed,
and only the idle smart trolleys can be assigned to move the WIPs. The NSGA-II algorithm is
used to solve the problem of the passive material handling and the Euler distance method is
used to give a default result too. The comparison results of the two strategies are shown in
Table 4. From the comparison, it can be found that the proactive material handling method
can largely reduce the total energy consumption (52.7%) and distance of trolleys (66.7%)
simultaneously.
Table 4
8. Conclusion
In order to cope with frequent changes and disturbances, many companies applying advanced
CPS technologies to keep high-level production traceability and controllability about its
shop-floor. However, most of the shop-floors still use passive material handling strategies,
which lead to excessively long occupation or idle time of machines. This situation is more
serious on the shop-floor that have a central material storage area and limited buffers for each
machine. To address this issue, a proactive material handling method for CPS enabled shop-
floor is proposed in this paper. The main contributions of this work include 1) an overall
architecture of a proactive material handling method for CPS-enabled shop-floor. 2) a shop-
floor digital twin model which can simulate real-life production in a virtual environment and
obtain the production KPIs easily. 3) an innovative combination of time series technology
and multiple linear regression method (i.e., TWMLR) in the prediction of remaining
processing time. 4) a proactive material handling strategy which can consider both real-time
and future status of smart trolleys and WIPs at the same time.
The proposed CPS-PMH and its key technologies provide an opportunity to improve shop-
floor production fluency and efficiency. It is of great significance in the following aspects.
Firstly, by establishing a shop-floor digital twin model, the physical shop-floor can be
mirrored as a dynamic autonomous system, the output of production decisions can be easily
simulated and checked. As a result, individualized manufacturing demands can be met
without costly and long-term analysis, which is meaningful for the current mass-
individualization production paradigm. Secondly, in the aspects of return on investment
(ROI), with the cost reduction of smart sensing devices and the significant promotion of
computational ability, the proposed CPS-PMH can be fulfilled with rather low investment,
while the return in energy saving and production efficiency is large. Thirdly, the optimization
and comparison results of proactive material handling strategy show that making decisions in
advance can largely reduce the energy consumption and logistics distance in shop-floor,
which is of referential significance for other production decision-making aspects such as
machine maintenance, schedule adjustment, etc.
There are three main challenges waiting to be solved. The first challenge is to build a high-
accurate and comprehensive digital-twin model for a physical shop-floor which consists of a
large amount of dynamic resources. The second challenge is how to incorporate the proposed
CPS-PMH with other enterprise information systems (EISs) (such as advanced planning
system, MRP, etc.) so that more feasible and optimal production and logistics schedule can
be made. The third challenge is how to combine the more comprehensive big data analytics
that could provide a more accurate prediction of production KPIs so that more robust decision
support can be achieved. Consequently, future research is necessary to focus mainly on the
construction of complex digital twin model, the integration of CPS-PMH with advanced EISs,
and the deep fusion of production KPIs prediction algorithms with big data analytics.
Acknowledgements
The authors would like to acknowledge the financial supports of National Science Foundation
of China (51675441), the National Key Research and Development Program of China
(2018YFB1703402), and the 111 Project Grant (B13044).
Reference
[1] S. Wang, J. Wan, D. Zhang, D. Li, C. Zhang, Towards smart factory for Industry 4.0:
A self-organized multi-agent system with big data based feedback and coordination,
Comput. Networks. 101 (2016) 158–168. doi:10.1016/j.comnet.2015.12.017.
[2] F. Tao, Y. Zuo, L. Da Xu, L. Zhang, IoT-Based intelligent perception and access of
manufacturing resource toward cloud manufacturing, IEEE Trans. Ind. Informatics. 10
(2014) 1547–1557. doi:10.1109/TII.2014.2306397.
[3] J. Wang, Y. Zhang, Y. Liu, N. Wu, Multi-agent and Bargaining-game-based Real-time
Scheduling for Internet of Things-enabled Flexible Job Shop, IEEE Internet Things J.
6 (2019) 2518–2531. doi:10.1109/JIOT.2018.2871346.
[4] R.Y. Zhong, S. Lan, C. Xu, Q. Dai, G.Q. Huang, Visualization of RFID-enabled
shopfloor logistics Big Data in Cloud Manufacturing, Int. J. Adv. Manuf. Technol. 84
(2016) 5–16. doi:10.1007/s00170-015-7702-1.
[5] R.Y. Zhong, C. Xu, C. Chen, G.Q. Huang, R.Y. Zhong, C. Xu, C. Chen, G.Q. Huang,
B. Data, Big Data Analytics for Physical Internet-based intelligent manufacturing shop
floors, Int. J. Prod. Res. 5 (2017) 2610–2621. doi:10.1080/00207543.2015.1086037.
[6] S. Ren, Y. Zhang, Y. Liu, T. Sakao, D. Huisingh, C.M.V.B. Almeida, A
comprehensive review of big data analytics throughout product lifecycle to support
sustainable smart manufacturing: A framework, challenges and future research
directions, J. Clean. Prod. (2019) 1343–1365. doi:10.1016/j.jclepro.2018.11.025.
[7] X. Xu, From cloud computing to cloud manufacturing, Robot. Comput. Integr. Manuf.
28 (2012) 75–86. doi:10.1016/j.rcim.2011.07.002.
[8] R. Liu, X. Xie, K. Yu, Q. Hu, A survey on simulation optimization for the
manufacturing system operation, Int. J. Model. Simul. 38 (2018) 116–127.
doi:10.1080/02286203.2017.1401418.
[9] R.Y. Zhong, G.Q. Huang, S. Lan, Q.Y. Dai, T. Zhang, C. Xu, A two-level advanced
production planning and scheduling model for RFID-enabled ubiquitous
manufacturing, Adv. Eng. Informatics. 29 (2015) 799–812.
doi:10.1016/j.aei.2015.01.002.
[10] S. Stoyanov, M. Ahsan, C. Bailey, T. Wotherspoon, C. Hunt, Predictive analytics
methodology for smart qualification testing of electronic components, J. Intell. Manuf.
30 (2019) 1497–1514. doi:10.1007/s10845-018-01462-9.
[11] B. Huang, W. Wang, S. Ren, R.Y. Zhong, J. Jiang, A proactive task dispatching
method based on future bottleneck prediction for the smart factory, Int. J. Comput.
Integr. Manuf. 32 (2019) 278–293. doi:10.1080/0951192X.2019.1571241.
[12] R.Y. Zhong, X. Xu, E. Klotz, S.T. Newman, Intelligent Manufacturing in the Context
of Industry 4.0: A Review, Engineering. 3 (2017) 616–630.
doi:10.1016/J.ENG.2017.05.015.
[13] F. Tao, M. Zhang, Digital Twin Shop-Floor: A New Shop-Floor Paradigm Towards
Smart Manufacturing, IEEE Access. (2017). doi:10.1109/ACCESS.2017.2756069.
[14] Q. Qi, F. Tao, Digital Twin and Big Data Towards Smart Manufacturing and Industry
4.0: 360 Degree Comparison, IEEE Access. 6 (2018) 3585–3593.
doi:10.1109/ACCESS.2018.2793265.
[15] Y. Zhang, G. Zhang, J. Wang, S. Sun, S. Si, T. Yang, Real-time information capturing
and integration framework of the internet of manufacturing things, Int. J. Comput.
Integr. Manuf. 28 (2015) 811–822. doi:10.1080/0951192X.2014.900874.
[16] Y. Sallez, S. Pan, B. Montreuil, T. Berger, E. Ballot, On the activeness of intelligent
Physical Internet containers, Comput. Ind. 81 (2016) 96–104.
doi:10.1016/j.compind.2015.12.006.
[17] Y. Yang, S. Pan, E. Ballot, Mitigating supply chain disruptions through interconnected
logistics services in the Physical Internet, Int. J. Prod. Res. 55 (2017) 3970–3983.
doi:10.1080/00207543.2016.1223379.
[18] G. Zhang, Y. Zhang, R.Y. Zhong, Y. Wu, Extending augmented Lagrangian
coordination for the optimal configuration of cloud-based smart manufacturing
services with production capacity constraint, Robot. Comput. Integr. Manuf. 58 (2019)
21–32. doi:10.1016/j.rcim.2019.01.009.
[19] G. Zhang, Y. Zhang, X. Xu, R.Y. Zhong, An augmented Lagrangian coordination
method for optimal allocation of cloud manufacturing services, J. Manuf. Syst. 48
(2018) 122–133. doi:10.1016/j.jmsy.2017.11.008.
[20] Y. Lu, X. Xu, Cloud-based manufacturing equipment and big data analytics to enable
on-demand manufacturing services, Robot. Comput. Integr. Manuf. 57 (2019) 92–102.
doi:10.1016/j.rcim.2018.11.006.
[21] D. Mourtzis, E. Vlachou, A cloud-based cyber-physical system for adaptive shop-floor
scheduling and condition-based maintenance, J. Manuf. Syst. 47 (2018) 179–198.
doi:10.1016/J.JMSY.2018.05.008.
[22] Teruaki Ito, S.M.M. Abadi, Agent-based material handling and inventory planning in
warehouse, J. Intell. Manuf. 13 (2002) 201–210. doi:10.1023/A:1015786822825.
[23] S. Chakraborty, D. Banik, Design of a material handling equipment selection model
using analytic hierarchy process, Int. J. Adv. Manuf. Technol. 28 (2006) 1237–1245.
doi:10.1007/s00170-004-2467-y.
[24] G. El Khayat, A. Langevin, D. Riopel, Integrated production and material handling
scheduling using mathematical programming and constraint programming, Eur. J.
Oper. Res. 175 (2006) 1818–1832. doi:10.1016/j.ejor.2005.02.077.
[25] S. Pan, D. Trentesaux, Y. Sallez, Specifying self-organising logistics system:
Openness, intelligence, and decentralised control, in: P.L. and J.B.O. T. Borangiu, D.
Trentesaux, A. Thomas (Ed.), Serv. Orientat. Holonic Multi-Agent Manuf. Proc.
SOHOMA 2016, Springer, 2017: pp. 93–102. doi:10.1007/978-3-319-51100-9_9.
[26] S. Ren, X. Zhao, B. Huang, Z. Wang, X. Song, A framework for shopfloor material
delivery based on real-time manufacturing big data, J. Ambient Intell. Humaniz.
Comput. 10 (2019) 1093–1108. doi:10.1007/s12652-018-1017-7.
[27] Y. Zhang, G. Zhang, W. Du, J. Wang, E. Ali, S. Sun, An optimization method for
shopfloor material handling based on real-time and multi-source manufacturing data,
Int. J. Prod. Econ. 165 (2015) 282–292. doi:10.1016/j.ijpe.2014.12.029.
[28] T. Qu, S.P. Lei, Z.Z. Wang, D.X. Nie, X. Chen, G.Q. Huang, IoT-based real-time
production logistics synchronization system under smart cloud manufacturing, Int. J.
Adv. Manuf. Technol. 84 (2016) 147–164. doi:10.1007/s00170-015-7220-1.
[29] E.A. Lee, Cyber-Physical Systems-Are Computing Foundations Adequate?, in: NSF
Work. Cyber-Physical Syst. Res. Motiv. Tech. Roadmap, Austin, Texas, 2006.
[30] L. Wang, M. Törngren, M. Onori, Current status and advancement of cyber-physical
systems in manufacturing, J. Manuf. Syst. 37 (2015) 517–527.
doi:10.1016/j.jmsy.2015.04.008.
[31] M. Zhou, P. Guo, T. Wang, L. Gao, H. Yin, C. Cai, J. Gu, X. Lü, Biotechnology for
Biofuels Metagenomic mining pectinolytic microbes and enzymes from an apple
pomace ‑ adapted compost microbial community, Biotechnol. Biofuels. 10 (2017) 1–
15. doi:10.1186/s13068-017-0885-y.
[32] L. Monostori, B. Ka, T. Bauernhansl, S. Kondoh, S. Kumara, Cyber-physical systems
in manufacturing, CIRP Ann. - Manuf. Technol. 65 (2016) 621–641.
doi:10.1016/j.cirp.2016.06.005.
[33] Y. Zhang, C. Qian, J. Lv, Y. Liu, Agent and cyber-physical system based self-
organizing and self-adaptive intelligent shopfloor, IEEE Trans. Ind. Informatics. 13
(2017) 737–747. doi:10.1109/TII.2016.2618892.
[34] S. Ma, Y. Zhang, J. Lv, H. Yang, J. Wu, Energy-cyber-physical system enabled
management for energy-intensive manufacturing industries, J. Clean. Prod. 226 (2019)
892–903. doi:10.1016/j.jclepro.2019.04.134.
[35] Y. Zhang, S. Ma, H. Yang, J. Lv, L. Yang, A big data driven analytical framework for
energy-intensive manufacturing industries, 197 (2018) 57–72.
doi:10.1016/j.jclepro.2018.06.170.
[36] A.J. Álvares, L. Eduardo, S. De Oliveira, J. Carlos, A. José, L. Eduardo, S. De Oliveira,
J. Carlos, E. Ferreira, Development of a Cyber-Physical framework for monitoring and
teleoperation of a CNC lathe based on MTconnect and OPC protocols, Int. J. Comput.
Integr. Manuf. 31 (2018) 1049–1066. doi:10.1080/0951192X.2018.1493232.
[37] F. Tao, J. Cheng, Q. Qi, IIHub: An Industrial Internet-of-Things Hub Toward Smart
Manufacturing Based on Cyber Physical System, IEEE Trans. Ind. Electron. 14 (2018)
2271–2280. doi:10.1109/TII.2017.2759178.
[38] F. Tao, J. Cheng, Q. Qi, M. Zhang, H. Zhang, F. Sui, Digital twin-driven product
design, manufacturing and service with big data, Int. J. Adv. Manuf. Technol. (2017).
doi:10.1007/s00170-017-0233-1.
[39] F. Tao, M. Zhang, Y. Liu, A.Y.C. Nee, Digital twin driven prognostics and health
management for complex equipment, CIRP Ann. Technol. 67 (2018) 169–172.
doi:10.1016/j.cirp.2018.04.055.
[40] J. Leng, H. Zhang, D. Yan, Q. Liu, X. Chen, D. Zhang, Digital twin-driven
manufacturing cyber-physical system for parallel controlling of smart workshop, J.
Ambient Intell. Humaniz. Comput. 10 (2019) 1155–1166. doi:10.1007/s12652-018-
0881-5.
[41] J. Leng, P. Jiang, Granular computing–based development of service process reference
models in social manufacturing contexts, Concurr. Eng. Res. Appl. 25 (2017) 95–107.
doi:10.1177/1063293X16666312.
[42] C. Zhuang, J. Liu, H. Xiong, Digital twin-based smart production management and
control framework for the complex product assembly shop-floor, 96 (2018) 1149–1163.
doi:10.1007/s00170-018-1617-6.
[43] Y. Zheng, S. Yang, H. Cheng, An application framework of digital twin and its case
study, J. Ambient Intell. Humaniz. Comput. 10 (2019) 1141–1153.
doi:10.1007/s12652-018-0911-3.
[44] U.A. Umar, M.K.A. Ariffin, N. Ismail, S.H. Tang, Hybrid multiobjective genetic
algorithms for integrated dynamic scheduling and routing of jobs and automated-
guided vehicle (AGV) in flexible manufacturing systems (FMS) environment, Int. J.
Adv. Manuf. Technol. 81 (2015) 2123–2141. doi:10.1007/s00170-015-7329-2.
[45] Y. Zhang, Z. Guo, J. Lv, Y. Liu, A Framework for Smart Production-Logistics
Systems based on CPS and Industrial IoT, IEEE Trans. Ind. Informatics. 14 (2018)
4019–4032. doi:10.1109/TII.2018.2845683.
[46] B. Veeravalli, G. Rajesh, N. Viswanadham, Design and analysis of optimal material
distribution policies in flexible manufacturing systems using a single AGV, Int. J. Prod.
Res. 40 (2010) 2937–2954. doi:10.1080/00207540210137648.
[47] E.K. Xidias, A.C. Nearchou, N.A. Aspragathos, E.K. Xidias, A.C. Nearchou, N.A.
Aspragathos, Integrating Path Planning, Routing,and Manufacturing Facilities, Cybern.
Syst. 43 (2012) 143–162. doi:10.1080/01969722.2012.659965.
[48] Y. Zhang, Z. Zhu, J. Lv, CPS-Based smart control model for shopfloor material
handling, 14 (2018) 1764–1775. doi:10.1109/TII.2017.2759319.
[49] B. Zhou, C. Shen, Multi-objective optimization of material delivery for mixed model
assembly lines with energy consideration, J. Clean. Prod. 192 (2018) 293–305.
doi:10.1016/j.jclepro.2018.04.251.
[50] M. Abdoli, F.F. Choobineh, Empirical Bayes forecasting methods for job flow times
Empirical Bayes forecasting methods for job flow times, IIE Trans. 37 (2005) 635–649.
doi:10.1080/07408170590948495.
[51] A. Alenezi, S.A. Moses, T.B. Trafalis, Real-time prediction of order flowtimes using
support vector regression, Comput. Oper. Res. 35 (2008) 3489–3503.
doi:10.1016/j.cor.2007.01.026.
[52] S.M. Asadzadeh, A. Azadeh, A. Ziaeifar, A Neuro-Fuzzy-Regression Algorithm for
Improved Prediction of Manufacturing Lead Time with Machine Breakdowns, Concurr.
Eng. 19 (2011) 269–281. doi:10.1177/1063293X11424512.
[53] G. Ioannou, S. Dimitriou, Lead time estimation in MRP/ERP for make-to-order
manufacturing systems, Int. J. Prod. Econ. 139 (2012) 551–563.
doi:10.1016/j.ijpe.2012.05.029.
[54] N. Wu, F. Chu, C. Chu, M. Zhou, Petri net modeling and cycle-time analysis of dual-
arm cluster tools with wafer revisiting, IEEE Trans. Syst. Man, Cybern. Part ASystems
Humans. 43 (2013) 196–207. doi:10.1109/TSMCA.2012.2187890.
[55] D. Mourtzis, M. Doukas, K. Fragou, K. Efthymiou, V. Matzorou, Knowledge-based
estimation of manufacturing lead time for complex engineered-to-order products,
Procedia CIRP. 17 (2014) 499–504. doi:10.1016/j.procir.2014.01.087.
[56] J. Wang, J. Zhang, X. Wang, A Data Driven Cycle Time Prediction With Feature
Selection in a Semiconductor Wafer Fabrication System, IEEE Trans. Semicond.
Manuf. 31 (2018) 173–182. doi:10.1109/TSM.2017.2788501.
[57] Y. Zhang, W. Wang, W. Du, C. Qian, H. Yang, Coloured Petri net-based active
sensing system of real-time and multi-source manufacturing information for smart
factory, Int. J. Adv. Manuf. Technol. 94 (2018) 3427–3439. doi:10.1007/s00170-017-
0800-5.
[58] Y. Zhang, G.Q. Huang, T. Qu, O. Ho, S. Sun, Agent-based smart objects management
system for real-time ubiquitous manufacturing, Robot. Comput. Integr. Manuf. 27
(2011) 538–549. doi:10.1016/j.rcim.2010.09.009.
[59] Y. Zhang, W. Wang, N. Wu, C. Qian, IoT-Enabled Real-Time Production
Performance Analysis and Exception Diagnosis Model, IEEE Trans. Autom. Sci. Eng.
13 (2016) 1318–1332. doi:10.1109/TASE.2015.2497800.
[60] W. Wang, H. Yang, Y. Zhang, J. Xu, IoT-enabled real-time energy efficiency
optimization method for energy-intensive manufacturing enterprises, Int. J. Comput.
Integr. Manuf. 31 (2018) 362–379. doi:10.1080/0951192X.2017.1337929.
[61] K. Deb, A. Pratap, S. Agarwal, T. Meyarivan, A fast and elitist multiobjective genetic
algorithm: NSGA-II, IEEE Trans. Evol. Comput. 6 (2002) 182–197.
doi:10.1109/4235.996017.