Journal of Ambient Intelligence and Humanized Computing

https://doi.org/10.1007/s12652-018-0832-1

ORIGINAL RESEARCH

A cloud-terminal-based cyber-physical system architecture for energy

efficient machining process optimization
X. X. Li1 · F. Z. He2 · W. D. Li3

Received: 30 December 2017 / Accepted: 3 May 2018

© Springer-Verlag GmbH Germany, part of Springer Nature 2018

Abstract
An innovative cloud-terminal-based cyber-physical system (CTCPS) architecture is presented to support energy efficient
machining process optimization. The CTCPS consists of four levels: machine level, control level, data level and decision
support level. The machine level and control level are composed of all kinds of terminals related to machining process and
mainly responsible for monitoring machines and controlling machines to execute the optimal solutions. The data level and
decision support level are deployed in cloud for data storage, management and analysis and optimization. Experiments on
the practical case shows that the CTCPS is promising and has significant potential of implementation in practice.

Keywords Cyber-physical system · Sustainable manufacturing · Energy efficient machining · Machining process
optimization

1 Introduction process optimization to support the practitioners to enhance

manufacturing energy efficiency and achieve sustainable
In recent years, the mounting demands for production cus- manufacturing.
tomization have incurred more manufacturing activities The energy-aware machining process optimization
(Jania and Magdiel 2017) and then more energy consump- mainly includes four strategies: energy-aware machining
tion in the manufacturing sector. Statistics have shown that parameter optimization, energy-aware toolpath optimiza-
manufacturing contributes to 90% of the energy consump- tion, energy-aware process planning and energy-aware job
tion in the industrial sector, which is the most energy-con- scheduling. In the energy-aware parameter optimization,
suming sector in the world with a 52% share of the global the relationship between the machining parameters such
energy consumption (Wang and Lin 2014). However, the as spindle speed, federate, cutting depth and so on and the
manufacturing energy efficiency is surprisingly low. A large energy consumed in the machining processes are studied.
number of studies have indicated that the specific machin- And then, the machining energy consumption is further
ing energy consumption accounted for less than 15% of the reduced by choosing the optimum machining parameters
total energy consumed by a machine tool during machin- for the machine. In the energy-aware toolpath optimization,
ing processes (Salonitis and Ball 2013). Therefore, it is the influence of tool path on the machining energy consump-
imperative to research into the energy-aware machining tion is explored, and then, the further optimum strategies for
tool path generation are developed to achieve energy sav-
ing. In the energy-aware process planning, the machining
process routings with multiple flexibilities (i.e., alternative
* F. Z. He
[email protected] machining methods, machines, cutting tools and operation
sequences) are optimized to reduce the energy consumed in
1
College of Informatics, Huazhong Agricultural University, the machining processes. In the energy-aware job schedul-
Wuhan, Hubei, China ing, the whole energy consumption of all the machines in
2
State Key Laboratory of Digital Manufacturing Equipment a shop floor is studied, and further optimized by taking the
and Technology, School of Computer, Wuhan University, optimum scheduling solution. It can be seen that the above
Wuhan, Hubei, China
four optimization strategies contribute to the improvement of
3
Faculty of Engineering and Computing, Coventry University,
Coventry, UK

13
Vol.:(0123456789)
 X. X. Li et al.

manufacturing energy efficiency at different level: a machine obtained in the data level is further used by the optimization
level, a machining process level, or a shop floor level. algorithm services at the decision support level to obtain the
A large number of studies (Li and McMahon 2007; Li integrated optimal solution. Based on these cloud services,
et al. 2017b) have indicated that the machining process the integrated optimal solution can be obtained and fed back
optimization strategies at different levels are usually com- to the control level and the machine level to achieve sustain-
plementary. For a part to be machined, process planning is able manufacturing.
used to plan its operations and its manufacturing resources The remainder of this paper is organized as follows. In
including machines, tools and other configuration. For Sect. 2, related work is reviewed. In Sect. 3, the whole sys-
each operation, machining parameters such as spindle tem framework is presented. The details of each level in the
speed, feed rate, the depth of cutting and son on are cho- system framework are described in Sect. 4. In Sect. 5, case
sen to execute it on the determined resource. Obviously, studies are given. Finally, a conclusion is drawn in Sect. 6.
machining parameters of each operation highly depend on
the determined process plan. Meanwhile, the tool path of
each operation is influenced by the determined machin- 2 Related work
ing parameters and process plan. Also, the job scheduling
which is used to handle a group of parts with temporal In recent years, sustainable manufacturing has attracted
constraints and competitive manufacturing resources is many researchers’ attention. Different optimization strate-
significantly influenced by the determined process plan gies have been proposed. This section presents a review of
and machining parameters. As thus, it is ideal to integrate the related work from two different perspectives, i.e. single
these optimization strategies to achieve a global optimum energy efficient optimization strategy and integrated energy
in manufacturing, and improve the manufacturing energy efficient optimization strategy.
efficiency.
However, most of the existing research work involves 2.1 Single energy efficient optimization strategy
only one or two of the above optimization strategies. The
other optimization strategies have been neglected or are The single energy efficient optimization strategy optimizes
assumed having been predetermined. Additionally, both the manufacturing energy consumption from four different
the generation information technologies and big data col- aspects, i.e. machining parameters, process planning, sched-
lected in manufacturing drive the traditional manufacturing uling or tool path. The related research work of energy effi-
to change into smart manufacturing demanding the integra- cient optimization with one strategy consideration can be
tion and cyber-physical system of manufacturing (Tao and viewed from three different perspectives which are summa-
Qi 2017; Tao et al. 2018a; Li et al. 2015a). Consequently, the rized in the following.
integration of the above optimization strategies is required.
Nevertheless, further research is imperative to develop 1. Exploring the impact of the optimization aspects on
new cyber-physical system to integrate energy monitoring, energy consumption
energy data analysis into the machining process optimization
and address the dynamics in the machining process. Also, Most of the research works on this degree focuses on
the cloud-based paradigma become a research hotspot of designing some experiments and doing the experiments.
manufacturing (Wu et al. 2015; Zhang et al. 2017a). And then, based on the experimental data, the impact of the
In this paper, a cloud-terminal-based cyber-physical sys- optimization aspect on the energy consumption is analyzed.
tem is presented to achieve energy efficient machining pro- The related works mainly focuses on exploring the impact of
cess optimiza integrated process planning and scheduling machining parameters and tool path on energy consumption.
systems consider productivity, production cost, or quality The details are described below.
with the energy tion. The CTCPS is divided into four levels:
machine level, control level, data level and decision support • In order to explore the impact of machining parameters
level. The machine level and control level are composed of on the energy consumption, Taguchi method, ANOVA
all kinds of terminals such as CNC machines, computers, analysis and grey relational analysis are most commonly
mobile devices, sensors and so on. Based on these terminals, used to design the experiments and analyze the energy
the data related to the machining processes can be collected consumption experiment data. Series of turning experi-
and the optimal solution from cloud can be executed. Both ments have been done (Hanafi et al. 2012; Camposeco-
the data and decision support level are in cloud. Their func- Negrete 2013; Zhang et al. 2015b; Sarıkaya et al. 2016;
tions are provided as the cloud services. First, the storage Bilga et al. 2016) and the machining parameters includ-
and analysis services at the data level are used to deal with ing feed rate, cutting depth and cutting speed have been
the collected data transferred to cloud. Then, the information revealed to be the significant influence factors for the

13
A cloud-terminal-based cyber-physical system architecture for energy efficient machining…

energy consumed in the machining processes. Similarly, However, in the model, the specifications for the fixed power
some other cutting experiments have also been done P0 and the constant k were not given. Taking up the miss-
to investigate the impact of cutting parameters on the ing specification in the model of Gutowski et al. (2006),
energy consumed in the milling or drilling processes researchers developed some improved energy consumption
(Fratila and Caiza 2011). The experiment results also models such as the unit process energy consumption model
showed that the selection of optimal feed rate, cutting created by Kara and Li (2011), the SEC model provided by
depth and cutting speed is an effective way to reduce Diaz et al. (2011) and the energy consumption model for
energy consumed in the cutting process. milling processes developed by Li et al. (2013). Addition-
• In addition, the impact of tool path on energy consump- ally, Kuram et al. (2013) proposed the SEC model based
tion was also studied. Several tool path generation on the cutting force. Newman et al. (2012) developed an
schemes were tested to explore the impact of tool path empirical model to establish the relationship between the
generation schemes on the amount of energy required power consumption and the process parameters including
to machine the same part (Kong et al. 2011). Different federate and depth of cut.
toolpath selection strategies have been adopted to execute Some other energy consumption models were established
the same pocket milling of stainless steel and the power to characterize the relationship between the machining
consumption of the machine tool was captured (Aram- parameters and the total cutting energy consumption. For
charoen and Mativenga 2014). The experiment results instance, an experimental design method was adopted by
showed that selection of optimal toolpath strategy can Lin et al. (2016) to establish a machining parameter opti-
bring significant energy savings in the machining pro- mization model of multi-pass turning operations in dry and
cesses. Additionally, it has been shown that an optimal wet cut environments. Design Expert was used to optimize
tool path can shorten the cycle time which can reduce the cutting parameters of the turning operation by Anand
the constant energy consumed by the auxiliary devices et al. (2016), the optimized values were further checked and
of the machine tool (Rangarajan and Dornfeld 2004). All compared by those being generally used. A response surface
these studies have indicated that tool path has a signifi- method was used by Campatelli et al. (2014) to optimize
cant impact on energy saving in machining process. the process parameters to minimize the power consumption
in the milling of carbon steel. Kant and Sangwan (2014)
considered power consumption and surface roughness by
2. Building energy consumption models optimizing the machining parameters. On the basis of the
experimental data obtained by the sensors mounted on the
The research works concentrated on quantifying the cutter, another empirical model involving spindle speed, fed-
machining energy consumption to characterize the relation- erate and cutting depth was presented by Hu et al. (2012).
ship between the influence factors and energy consumption. Based on weighted grey relational analysis and response sur-
Different energy consumption models have been established face methodology, a multi-objective method was developed
to support the corresponding energy efficient optimization by Yan and Li (2013) to evaluate trade-offs between sustain-
strategy and are listed in the following. ability, production rate and cutting quality. In Wang et al.
(2015a, b), artificial neural networks were used to establish
• Energy consumption modelling for machining parameter the complex nonlinear relationships between the process
optimization parameters including spindle speed, federate, cutting depth
and cutting width, and energy consumption.
As mentioned above, the machining parameters are the
significant influence factors for cutting energy consumption. • Energy consumption modelling for process planning
Thus, energy consumption models for machining parameter optimization
optimization strategy is mainly used to evaluate the energy
consumed in the cutting processes. The relevant research Process planning influences multiple energy consumption
works are summarized below. stages. In the existing research work, the energy consumed
Some energy consumption models have been built to in three stages is mainly considered. The first is the energy
characterize the relationship between the machining param- consumption of the cutting process. The second is the energy
eters and the specific energy consumed per unit volume of consumed in the setup stages including fixture change, work-
removed material of the cutting processes (i.e. the specific piece setup, tool change, machine change etc. The third is
energy consumption SEC). The most representative one was the energy consumed in transportation stage mainly includ-
developed by Gutowski et al. (2006). The SEC was mod- ing chip transportation and tool transportation. The energy
eled as a function of the process rate by analyzing the unit consumption of these stages is modeled by multiplying the
process energy for the processes in an exergy framework.

13
 X. X. Li et al.

power of the corresponding stage by the time used in the presented by Kong et al. (2011) to evaluate the total energy
same stage. of machining process.
For instance, Tao et al. (2017) proposed an energy con-
sumption model of a process plan, by considering mul- 3. Developing the further optimization method to identify
tiple energy consuming resources: the energy consumed the optimal solution
in cutting process, machine change energy consumption,
tool change energy consumption and configuration change Based on the energy consumption models, some research
energy consumption. Zhang and Ge (2015a) proposed an works developed a number of intelligent optimization algo-
energy consumption model to support process planning rithms to further identify the optimal machining parameters,
optimization. In this model, the energy consumed in the process planning and scheduling. For instance, In Wang
cutting process is evaluated by multiplying the specific et al. (2015a, b), based on the energy consumption model
energy per volume by the total volume of the material established by artificial neural networks, a set of intelligent
removed. However, the energy consumed in the setup stage optimization algorithms including pattern search, genetic
is neglected. Additionally, Choi and Paul (2014) proposed algorithm and simulated annealing algorithm were further
an energy consumption model to evaluate the energy con- applied to identify the optimal process parameters. Tao et al.
sumed in the cutting, setup and transportation stages. (2017) developed a cooperative co-evolutionary algorithm,
which hybridizes the artificial bee colony algorithm and
• Energy consumption modelling for scheduling optimi- Tabu search, to obtain the energy efficient process plans.
zation Dai et al. (2013) presented an improved genetic-simulated
annealing algorithm to search for the energy efficient sched-
The energy consumption models for scheduling optimi- uling solution. Similarly, some research works employed the
zation are used to evaluate the total energy consumption popular intelligent optimization algorithms such as genetic
of all the machines in a shop floor. In the existing studies, algorithm (Yi et al. 2015; Liu et al. 2016), particle swarm
the total energy consumption for scheduling optimization optimization (Li et al. 2016; Petrovic et al. 2016), ant colony
is modeled by adding the energy consumption of all the optimization (Wang et al. 2015a, b) and so on, to search for
machines in the shop floor. For the total energy consump- the optimal solution.
tion of a machine, the energy consumed in all stages of For the energy efficient tool path optimization, the further
the machine is involved to evaluate it. Thus, the energy research works have been done from two aspects. One is
consumed in different stages of the machines is considered developing the energy efficient toolpath generation method.
to reduce the total energy consumption of all the machines For instance, an energy efficient toolpath generation method
in a shop floor. for pocket milling has been proposed by Pavanaskar et al.
For instance, the energy consumed in the machine idle (2015). Compared with several commercial software-gener-
stage is minimized in single machine scheduling (Mouzon ated toolpaths, the proposed toolpath consumed at least 20%
and Yildirim 2008) and job shop scheduling (Liu et al. less energy consumption. The other is adjusting the existing
2014). The energy consumed in the cutting process is con- toolpath. For instance, based on toolpath in CNC codes, the
sidered in (Zhang et al. 2014; Fang and Lin 2013; Salido traveling toolpaths among features were optimized to reduce
et al. 2016; Artigues et al. 2013; Moon et al. 2013). Also, the energy consumed by travelling among the features (Li
multiple energy consumption stages are considered simul- et al. 2018c).
taneously in some research work. For instance, the energy
consumed in the idle, turn on/turn off, and cutting stage is 2.2 Integrated energy efficient optimization
reduced in job shop scheduling (Le and Pang 2013), flow strategy
shop scheduling (Dai et al. 2013) and flexible flow shop
scheduling (Tang et al. 2016). Different from the single energy efficient optimization, the
integrated energy efficient optimization involves two or more
• Energy consumption modelling for tool path optimiza- optimization strategies to improve the energy efficiency of
tion machining processes. Currently, the integration mainly
focused on the integration of scheduling and another strat-
Based on tool path, some energy consumption models egy. The details are described below.
have been established to evaluate the energy consumed in
machining process. For instance, Avram and Xirouchakis • The integration of scheduling and process planning
(2011) developed an energy consumption model to assess
the energy consumed by the spindle axis, feed axis and One of our previous studies as presented in Li et al.
load/unload cycle. Another energy consumption model was (2015b), proposed a hybrid optimization approach for

13
A cloud-terminal-based cyber-physical system architecture for energy efficient machining…

sustainable process planning and scheduling. However, • Most of the previous work addressed the energy efficient
machining parameters optimization is neglected in this work. optimization in a static machining process. However,
A further integration approach was presented by Wang et al. there are many dynamic events which may happen in the
(2015a, b) to consider machining parameter optimization. machining processes. The previous work cannot meet the
The integration approach consists of two stages: a process optimization requirements from the dynamic machining
stage and a system stage. In the process stage, the machining processes.
parameters including spindle speed, feed rate, cutting depth
and cutting width were optimized to improve the energy As a consequence of the above observation, it is impera-
efficiency of the milling process. In the consecutive sys- tive to develop a cyber-physical system to integrate more
tem stage, process planning and scheduling were optimized optimization strategies and address various dynamics in the
to achieve higher energy efficiency in the entire shop floor. machining processes.
However, it is obvious that the optimization in the process
and system stage is carried out separately and sequentially.
Zhang et al. (2016b) presented an integration model of pro- 3 CTCPS 4‑level architecture overview
cess planning and scheduling to select the optimal process
plan and machines for each operation and the start time of The proposed architecture is based on cloud-terminal. The
each operation to achieve the optimization of the total energy terminal consists of two levels: machine level and control
consumption. Similarly, Dai et al. (2015) developed a mixed level. The cloud is composed of data level and decision sup-
programming model for the integration of process planning port level. As illustrated in Fig. 1, the detailed architecture
and scheduling to improve the machining energy efficiency. is outlined as follows:

• The integration of scheduling and machining parameters 1. Machine level The machine level is composed of CNC
machines and the sensors deployed on the machines. At
Some research work focused on integrating scheduling this level, the machining processes of the CNC machines
and machining parameters by assuming the process planning are monitored and data in the machining processes is
having been predetermined. For instance, Lin et al. (2015) measured by sensors.
presented a model for the integration of cutting parameter 2. Data level This level focuses on dealing with the data
optimization and scheduling by considering air-cutting collected in the machine level. It has two functions
energy, cutting energy and energy consumed in the machine which are listed in the following:
idle state. Based on the integrated model, three strategies are • Data storage and management Store the data from
proposed to minimize energy consumption. The first is slow-
the physical world in the cloud data central servers
ing down the processing speed. The second is turning on/
and manage the stored data.
off the idle machines. The third is pre-optimizing the cutting • Data analysis The data-to-information conversion
parameters. It has been proved that the integration optimiza-
is achieved in this level. Based on the data stored in
tion has more energy-saving potential than the single energy
the cloud data central servers and the real-time data
efficient optimization strategy. Similarly, Yan et al. (2016)
from the machine level, the states of the involved
presented a multi-level optimization method to integrate cut-
machines and the prediction models for the optimiza-
ting parameter optimization and scheduling optimization.
tion performance criteria can be obtained.
There are two levels in the method. One is the machine tool
level where cutting parameters of each operation are opti- 3. Decision support level On the basis of the information
mized to reduce cutting energy consumption. The other is obtained in the data level, the initial integrated optimi-
the shop floor level where the scheduling optimization are zation solution can be generated in this level using the
made to minimize the total energy consumption. integrated optimization algorithm services. When the
Based on the above detailed literature survey, it can be dynamic events occur in the machining process, the
observed: adaptive optimization services will be triggered to adjust
the existing optimal solution.
• Many of the above research works focused on the sin- 4. Control level This level consists of all kinds of intel-
gle energy efficient optimization strategy. Though a few ligent terminals such as the CNC machines’ controllers,
studies reported as above have integrated some energy computers, mobile devices and so on. It acts as supervi-
efficient optimization strategies, some other optimiza- sory control to make machines self-configure and self-
tion strategies are not integrated yet. Hence, the research adaptive. It has two main functions which are described
works on the integration are still preliminary. as follows:

13
 X. X. Li et al.

Terminal Cloud

Clustering services
…… Data storage
Prediction services

State detection
services
Machines Wireless sensors networks
deployed on machines Data analysis services
Machine level Data level

Integrated optimization services

;;;;;;
Intelligent terminals:
Adaptive optimization services
controllers, computers, mobile
devices and so on……
Decision support level
Control level
Physical world Cyber world

Fig. 1  Four-level architecture for CTCPS

• Presentation Properly present the results which has information acquisition system (Vijayaraghavan and Dorn-
been obtained by analyzing the real-time data in the feld 2010; Mohamed et al. 2017, Su and Zhao 2017). The
data level, on the terminals to support expert users monitoring infrastructure is formed using the WSNs to mini-
to take part in the correct decision making. mize the system cost from firmware, implementation and
• Application Apply the optimal decision solution maintenance perspectives and maximize the system flex-
made in the decision support level to the monitored ibility and availability to support heterogeneously industrial
system (i.e. the machines in the machine level). application scenarios along with different design require-
ments and limitations.
The architecture of the developed WSN system is shown
in Fig. 2. As depicted, internet protocol version-6 over low
4 Cloud‑terminal‑based four‑level power wireless personal area network (6LoWPAN) is used
cyber‑physical system to develop the WSN system because it supports IPv6 packets
to be sent to and received from over 802.15.4 based net-
As mentioned in the previous section, the proposed CPS is works. In the 6LoWPAN based WSN system, a variable
based on cloud-terminal which consists of four levels. More number of sensor nodes of different types with extended
details on these levels are discussed in this section. capabilities to sense information from the physical environ-
ment can eventually cooperate to form a WSN via a network
4.1 Machine level coordinator, whose main responsibilities are to create and
maintain the WSN. The coordinator can also be seemed
Machine monitoring and data collection are the main func- as a network bridge or gateway to bridge communication
tions of this level. To achieve these functions, wireless between the WSN to Internet because all communication
sensor networks (WSNs) are adopted in this level because between end devices and the components in the cyber level
WSNs have been proven to be the promised ubiquitous propagates through it.
technologies to develop a low-power, low-cost, and flexible

13
A cloud-terminal-based cyber-physical system architecture for energy efficient machining…

Fig. 2  The architecture of WSN

system

Data computing Spark

Resource
Yarn
management

Data storage HDFS

Fig. 4  The structure of data storage and management

employed to form the network gateway. In this mecha-

nism, the USB dongle acts as wireless network coordina-
tor to manage the 6LoWPAN Network and meanwhile the
board router provides the interface with Internet.

4.2 Data level
Fig. 3  6LoWPAN based energy monitoring system
The volume of the data collected in manufacturing is vast
(Tao et al. 2018b; Kusiak 2017). The data level is just
responsible for dealing with the big data collected in the
machine level. The details are described in the following.
The implementation of the presented system is illus-
trated in Fig. 3. As depicted, three types of devices 4.2.1 Data storage and management
including a three-phases current sensor node, a three-
phase voltage sensor node and a WSN coordinator con- The data collected in the physical world can be divided into
nected with an IPv6 Internet Router, are employed to form three types. The first is the structured data which can be
the 6LoWPAN network. All these devices are developed directly described as a two-dimensional table file. For exam-
on Jennic JN5168 developing kits, which offer a power- ple, the data related to a machine contains the machine’s
ful AMR core along with IEEE 802.15.4 radio to fully type, rated power, spindle speed and so on. The machine
support the 6LoWPAN stack, ZigBee pro and 802.15.4 data can be organized into a two-dimensional table. As
stacks. In order to provide interoperability between the thus, the machine data is structured. The second is the semi-
Internet based on Wi-Fi or Ethernet, and the 6LoWPAN structured data which are more flexible than the former.
network based on 802.15.4, the combination of a USB For instance, the process plan for a part is flexible in three
dongle and a Linksys Wireless-N board router supplied aspects: feature sequence, operation selection and resource
from JN5168 developing kits, as depicted in Fig. 3, are selection. Thus, the data for a part’s process plan is flexible

13
 X. X. Li et al.

and XML format is employed to describe this type of data. 4.2.2 Data analysis
The third is the non-structured data such as the pictures,
videos of the machining processes and some sensor data etc. There are two types of data to be analyzed. One is the his-
As is shown in Fig. 4, Hadoop based structure is adopted toric data which has been stored in the HDFS. The other is
to achieve the storage and management of the above three the real-time data measured by the sensors in the machine
types of data. There are three parts in this structure, which level. In order to analyze both types of data, the corre-
are listed in the following. sponding analysis services are developed. The details are
as follows:
1. The Hadoop Distributed File System (HDFS). It is used
to achieve data storage. In the HDFS, the data is stored 1. For the historic data, the cluster analysis services (Li
in the format of file because it exposes a file system et al. 2017a) are first adopted to obtain the data to be fur-
namespace. The HDFS consists of a NameNode, many ther used by the prediction analysis services. And then,
DataNodes and a Secondary NameNode (Lee et al. the prediction analysis services (Li et al. 2018a) are used
2014). The NameNode manages the entire file system, to build the prediction model for the performance crite-
splits the files into many same sized blocks and allocates ria which include the energy consumed in the machining
them to the DataNodes. The DataNodes are responsible processes. For instance, the historic energy consumption
for storing these data blocks. The Secondary NameNode data is corresponded to the whole machining processes.
is just a backup of the NameNode and used to improve The machining parameters mainly influence the energy
the stability of the HDFS. consumed during the cutting and air cutting stages.
2. The Hadoop Yet Another Resource Negotiator (YARN). Thus, the energy consumption data in the cutting and
It is used as the data operation system to manage and air cutting stages should be first selected using the clus-
schedule the resources. YARN adopts master/slave ter analysis service. Based on the data obtained by the
architecture, in which master is a global ResourceMan- cluster analysis, the further prediction can be made and
ager and slaves are many NodeManagers. The Resource- energy consumption model can be obtained.
Manager is responsible for arbitrating the resources on 2. For the real-time data, the state detection analysis ser-
each NodeManager. It contains two main components: vice is adopted to judge the machine’s state because the
scheduler and ApplicationManager. The scheduler machine’s state can be reflected by the real-time data.
is responsible for allocating resources to the various As mentioned above, based on the historical data, the
running applications subject to familiar constraints machine’s normal and abnormal situations have been
of capacities, queues etc. When a job is submitted, an clustered. As thus, the machine’s real-time state can be
ApplicationMaster is started by the ApplicationMan- obtained by analyzing the real-time data.
ager to track and manage the job. The NodeManager
are the per-machine framework agents and responsible Based on the above data analysis services, the data has
for monitoring their resource usage and reporting the been converted to information which can be used to support
results back to the ResourceManager. the further decision making.
3. Spark. It is used to achieve data computing. The clus-
ter mode is used to launch Spark applications on the 4.3 4.3 Decision support level
YARN. In the cluster mode, the Spark driver runs inside
an application master process which is managed by the Based on the information obtained by data analysis, the fur-
YARN on the cluster. The Spark client submits a job to ther optimization algorithm services are developed to search
the ResourceManager and writes the Jar in the HDFS. for the initial optimization solution for a new machining task
The ResourceManager obtains the resource from the or adapt the existing machining solution for the dynamic
NodeManager and build a Spark ApplicationMaster scenarios in the machining processes.
to be started by the NodeManager. The started Spark For a new job, the process planning, scheduling, tool
ResourceManager is used to find the Jar file and start path and machining parameters are considered together to
the DAGScheduler and YARN Cluster Scheduler. Then, minimize the energy consumed in the machining processes.
the container is applied, allocated and started. The job As is shown in Fig. 5, aiming at minimizing the energy
is finished by the cooperation of the Spark Application- consumption, the features in each part are sequenced, the
Master and container (Executor). operations are selected for each feature, the machines to be
used to execute the selected operations are scheduled and
the machining parameters are chosen for each machine. It
should be pointed out that not only the energy consumed
by the machine change and tool change but also the energy

13
A cloud-terminal-based cyber-physical system architecture for energy efficient machining…

Fig. 5  Integration optimization

……
strategy to minimize energy Parts Part1 Part2 Partn
consumption

Features F1 F12 …… F1i F2 F22 …… F2 …… Fn Fn2 …… Fnj

Feature F1 F2 F3 F4 F6 …… Fm …… Fi+k+j
sequence
……

Operation
selection(Os)
Os1 Os Os Os Os …… Os …… Os,i+k+j

…… ……
……

Scheduling Machine1 Machine2 Machine3 …… Machinem

……
Parameter
n,f,ap,ae n,f,ap,ae n,f,ap,ae …… n,f,ap,ae
selection

consumed by the tool’s travelling among the features and scheduling, tool path with minimum tool’s air travelling and
operations are taken into account. That is, the air cutting machining parameters is obtained for the machining task and
segments in tool path are reduced to save energy. Addition- fed back to the control level.
ally, for each machine selected to execute the operations of For the dynamic scenarios reflected by the real-time
each part, the involved machining parameters contain the data, the optimization services are reused to update the
machine’s spindle speed, feed rate, cutting depth and cut- existing machining solution. For instance, the machine
ting width. During the optimization of these machining breakdown or the new different job arrivals will trigger
parameters, both the rated parameters of the machine and the the optimization services to update the existing machining
machining accuracy requirements of the part are considered solution which involves not only the scheduling solution
to constrain the optimization. Based on the above integrated but also process planning, tool path and machining param-
optimization, an optimal solution for process planning, eters. And then, a new optimal solution is obtained and fed
back to the control level.

Fig. 6  PSO service accelerated

Device
on GPUs Multiprocessor N
……
Multiprocessor 2
Multiprocessor1

The global best solution

Shared memory

The local best solution1 The local best solution2 The local best solutionn
Registers Registers Registers
……
Thread1 Thread2 Threadn
Processor1 Processor2 Processorn

Constant cache

Texture cache

13
 X. X. Li et al.

To improve the algorithm services’ efficiency and 2015). Thus, it is adopted in our research to apply the opti-
respond to the dynamic events, multi-core/many-core mal decision solution on the machines and achieve on-board
graphic process units (GPUs) can be employed to support process execution and control.
the particle swarm optimization (PSO) algorithms (Yan FB is described in the IEC 61499 and can be classified
et al. 2017, 2018; Zhang et al. 2016a). For instance, a into two types: basic function block and composite function
fine-granularity method is adopted to accelerate algorithm block. A basic FB is composed of inputs and outputs of data
service on GPUs (Zhou et al. 2017, 2018). As shown in and events, execution control chart (ECC) and algorithms.
Fig. 6, the details are as follows: The algorithm triggering is based on an explicit event-driven
model. That is, the algorithms are executed when triggered
• One of the multiprocessor is allocated to run the PSO. The by the corresponding input events. Meanwhile, the input
current global best solution is saved in the shared memory. data is read. The algorithm scheduling is controlled by the
• A thread running in one of the processors is allocated for ECC which is a finite state machine. At the end of algorithm
each particle which is just an integrated optimal solu- execution, new output data is produced and an output event
tion. Three main works are done simultaneously in each is created. Figure 7 shows one example of FB for blind hole.
thread. First, the location and velocity of the particle is As depicted (cf. Fig. 7a), it consists of input events (cf. EI_
refreshed based on the local best solution in the register INI, EI_ROU, EI_FIN and EI_ESR), input data (cf. EMT,
of the processor and the global best solution in the shared DEPTH, POSITION, SURF_ROUGH etc.), output events
memory to generate a new particle. Secondly, the fitness (cf. EO_INI, EO_ROU, EO_FIN and EO_ESS), output data
of the new particle is computed using the prediction ser- (cf. EMT_ROU, EMT_FIN and FB_EXE), and algorithms
vice in the data level. Thirdly, based on the new particle (cf. ALG_INI, ALG_ROU, ALG_FIN and ALG_MON). Fol-
obtained in each processor, both the local best solution in lowing the ECC shown in Fig. 7b, EI_INI triggers algorithm
each processor and the global best solution are refreshed. ALG_INI which is for initialization such as tool change, last-
The above three works will be repeated until the number minute local optimization and control code preparation and
of iteration is more than the predetermined value. so on. Upon the ALG_INI’s completion, an output event EO_
• The resources of GPUs used by the PSO are released. INI will be fired to indicate the success of initialization. Simi-
larly, ALG_ROU, ALG_FIN are triggered by EO_ROU and
EO_FIN to achieve rough and finish machining respectively.
4.4 Control level For ALG_MON (FB monitoring), EI_ESR is its correspond-
ing triggering event. It can be seen that a basic FB can be
To control the machines in the machine level, function block considered as a class, and a particular FB is its instance. The
(FB) is adopted. As an IEC standard, FB possesses dynamic functionality of multi-inheritance, overriding, and dynamic
control capabilities that are able to handle, in a responsive binding can be used to form a composite function block.
and adaptive way, different kinds of uncertainty (Wang

EI_INI EO_INI

EI_ROU EO_ROU
Execution Control
EI_FIN Chart EO_FIN

EI_ESR EO_ESS

INI ALG_INI EO_INI

1
EMT_ROU EMT_ROU

EMT_FIN
EI_INI
EMT_FIN
Internal Algorithm 1
Hole_Dia START ROU ALG_ROU EO_ROU
ALG_INI FB_EXE EI_ROU
Hole_Position
ALG_ROU EI_FIN
Hole_Depth ALG_FIN 1
Surf_rough ALG_MON 1 EI_ESR FIN ALG_FIN EO_FIN
Dim
Geom
Material MON ALG_MON EO_ESS

(a) Structure designing (b) Execution control chart

Fig. 7  A blind hole FB

13
A cloud-terminal-based cyber-physical system architecture for energy efficient machining…

Fig. 8  Structure designing of function blocks

As is shown in Fig. 8, four kinds of function blocks are 5 Case study and analysis
designed in the present research: (1) machining feature
function block (MF-FB), (2) event switch function block To build the system, the Elastic Compute Service provided
(ES-FB), (3) response control function block (RC-FB) and by Alibaba cloud computing Co. Ltd. is rent. Following the
(4) customized service function block (CS-FB). These are methods in the data level, the data related to each machine is
the basic function block types for adaptive process plans stored and managed in the cloud. The data analysis algorithms
creation and its execution control. The first two types of FB and the optimization algorithms have been serviced as many
are designed to deal with the corresponding machining fea- WSDL files to be used. The code is programmed in JAVA and
ture and the event flows of those parallel features which are C programming language.
not critical operations respectively (Wang et al. 2009). The The assumptions for the case study include:
RC-FB is designed to select the next FB to be executed. As
depicted in Fig. 8, the events of RC-FB is divided into two • A penalty function is used to ensure that the sequence of
types: EI_D1_X and EI_D2_X. The former is those whose the features of each part complies with manufacturing con-
response is to regenerate the corresponding process plan for straints;
each machining feature. The latter is those special events • Each operation can be performed on a single machine, and
which need some customized services. For example, for each machine can only execute an operation at a time;
the tool change event, the algorithm for cutting tool selec- • The time for a set-up is identical and independent of spe-
tion is required. The CS-FB is customized for some special cific operations. The time for a machine change or a tool
dynamic condition such as cutting tool change. Finally, a change follows the same assumption.
composite function block can be obtained by establishing
all event and data connections properly. The total energy consumption in the machining process
is used as the performance criterion of the optimization.

13
 X. X. Li et al.

The total energy consumption can be evaluated using the where EC(Mi) is the energy consumption of the i-th machine,
following models: EC(Mi).stage_j is the energy consumed by the ith machine
in its jth stage, P(Mi).stage_j is the power demand of the
n
∑ ith machine during the its jth stage, and T(Mi).stage_j is the
ECtotal = EC(Mi ) (1) total time of the ith machine in its jth stage.
i=1
Three parts shown in Fig. 9 are selected as the case
because they are widely used in many studies (Li and
m
∑ McMahon 2007; Wen et al. 2014). As is shown in Fig. 9,
EC(Mi ) = EC(Mi ).stage_j (2) the parts from the left to the right have 14, 15 and 10
j=1
machining features respectively. The machines tested by
Kara and Li (2011) were used as the machining resources.
EC(Mi ).stage_j = P(Mi ).stage_j × T(Mi ).stage_j (3) The initial integrated optimal solution is generated in the
cloud and fed back to the terminals to be executed. During
the execution of the optimal solution, the dynamic scenarios
are detected and addressed. The detailed test steps are listed
in the following:

1. Based on the historic data, the clustering service is used

to obtain the cluster center. As is illustrated in Fig. 10,
based on the data collected on a machine (cf. Fig. 10a),
Fig. 9  The use case of three parts the K-means (Huang et al. 2017; Wu et al. 2017; Sharma

(a) The input data for the K-means service

EC=2.5814×104

EC=2.0757×104

EC=1.6125×104
EC=1.3371×104

EC=1.297×103

(b) The centers of the clustering (c) The result of the clustering

Fig. 10  The illustration of the K-means service

13
A cloud-terminal-based cyber-physical system architecture for energy efficient machining…

convergence curves 6. During executing the optimal solution, some dynamic

scenarios may occur. For instance, machine abnormali-
30000
ties are one of the most representative scenarios. For
20000
this kind of dynamic scenario, the mapping relationship
10000 between the machine’s state and the energy consumption
0 has been established by the clustering service. As thus,
1
47
93
139
185
231
277
323
369
415
461
507
553
599
645
691
737
783
829
875
921
967
the machine’s state can be mirrored in the time varia-
GPU-based PSO PSO tions of real-time energy consumption. As is shown in
Fig. 10c, five normal states of the tested machine has
Fig. 11  The comparison of the GPU-based PSO and PSO been identified and the clustering centers for these nor-
mal states have also been determined. When the real-
time energy consumption data transferred to the data
Table 1  comparison studies of GPU-based PSO service and PSO ser- level, the state detection service will be used to com-
vice for the case study pare it with the clustering centers. As thus, the abnormal
Optimization algo- Running The energy con- Acceleration ratio states of the machine can be obtained and the optimiza-
rithm service time (s) sumption (J) tion algorithm service will be driven to timely update
the scheduling in the integrated optimal solution.
GPU-based PSO 64 1.0144 × 104 2.31
PSO 148 1.0455 × 104

6 Conclusions
et al. 2018) service is used to achieve clustering. As is
shown in Fig. 10b, five clustering centers are obtained. In this paper, a CTCPS architecture for high energy efficient
On the basis of these clustering centers, the further machining is proposed. The CTCPS is composed of four
analysis is done to cluster the collected data into the levels: machine level, control level, data level and decision
corresponding cluster (cf. Fig. 10c). support level. The machine and control level are mainly
2. Based on the clustering center obtained in step (1) and responsible for machining process data collection, machin-
the collected real-time data, the state detection analy- ing process monitoring and machining operation execution.
sis service is used to identify the current state of the The collected data is transferred to the data level to be fur-
machines in the resource pool to support the further ther stored, managed and analysis. The prediction models
scheduling. and the clustering results obtained in the data level is then
3. Meanwhile, the energy consumption data corresponding used by the optimization algorithm services in the decision
to different machining stage, which has been obtained support level to obtain the integrated optimal solution and
using the clustering service are used by the prediction fed back to terminal. The prototype implementation verified
service such as the artificial neural network (ANN) algo- the potential of the CTCPS.
rithm service (Wang et al. 2017) to predict the power of Different from other existing solutions in the literature,
each machining stage to support the further computing our architecture integrates the optimization of process
the fitness of each solution. planning, scheduling, machining parameters and tool path
4. The GPU-based optimization algorithm services are simultaneously to achieve high energy efficient machining.
used to obtain the initial optimal solution which inte- Compared with other existing optimization methods for high
grates the process planning, scheduling, machining energy efficient machining, our method integrates more opti-
parameters and tool path. As illustrated in Fig. 11, both mization strategies and forms a closed loop to reflect the
the GPU-based PSO and the normal PSO (Qian et al. real-time dynamic scenarios.
2017) under the condition of energy consumption can In the future, the platform will be further expanded in
reach convergence in 1000 iterations. Moreover, Table 1 each level. Now, the electric current sensors have been
shows the results got by both the GPU-based PSO and deployed in the machine level. New types of sensors will be
PSO are similar. However, it is obvious that the former added in the machine level. New data analysis and manipula-
ran faster than PSO. Thus, the comparison studies in tion services (Li et al. 2018b; Lv et al. 2017, 2018; Ni et al.
this case showed that GPU-based PSO exceeds PSO in 2016, Sun et al 2016; Yu et al. 2018; Zhang et al. 2017a) will
efficiency. be added in the data level. The presentation of the analy-
5. The optimal solution is then fed back to the terminals to sis results will be added on the mobile device terminals.
be executed. More efficient optimization algorithm services (Chen et al.
2017; Zhang et al. 2017b) will be added at the decision level.

13
 X. X. Li et al.

Additionally, the further test and discussion will be done to Gutowski T, Dahmus J, Thiriez A (2006) Electrical energy require-
further verify the proposed CTCPS. ments for manufacturing processes. In: 13th CIRP international
conference on life cycle engineering, vol 31, pp 623–638
Hanafi I, Khamlichi A, Cabrera F, Almansa E, Jabbouri A (2012)
Acknowledgements This research was supported by the 7th European Optimization of cutting conditions for sustainable machining
Community Framework Programme (Grant no. 610675), Natural Sci- of PEEK-CF30 using TiN tools. J Clean Prod 33:1–9
ence Foundation of China (Grant nos. 31771683, 61472289), Hubei Hu S, Liu F, He Y, Hu T (2012) An on-line approach for energy effi-
Province Natural Science Foundation (Grant nos. 2016CFB555 and ciency monitoring of machine tools. J Clean Prod 27:133–140
2016ADC073) and the Fundamental Research Funds for the Central Huang X, Yin C, Dadras S, Cheng Y, Bai L (2017) Adaptive rapid
Universities (Grant no. 2662016PY119). The paper reflects only the defect identification in ECPT based on K-means and automatic
authors’ views and the Union is not liable for any use that may be made segmentation algorithm. J Ambient Intell Humaniz Comput.
of the information contained therein. https​://doi.org/10.1007/s1265​2-017-0671-5
Jania A, Magdiel P (2017) Industry 4.0 framework for management
and operations: a review. J Ambient Intell Humaniz Comput.
References https​://doi.org/10.1007/s1265​2-017-0533-1
Kant G, Sangwan K (2014) Prediction and optimization of machin-
ing parameters for minimizing power consumption and surface
Anand Y, Gupta A, Abrol A, Gupta Y, Kumar V, Tyagi S (2016) roughness in machining. J Clean Prod 83:151–164
Optimization of machining parameters for green manufactur- Kara S, Li W (2011) Unit process energy consumption models
ing. Cogent Eng 3(1):1–16. https​: //doi.org/10.1080/23311​ for material removal processes. CIRP Ann Manuf Technol
916.2016.11532​92 60(1):37–40
Aramcharoen A, Mativenga P (2014) Critical factors in energy Kong D, Choi S, Yasui Y, Pavanaskar S, Dornfeld D, Wright P
demand modelling for CNC milling and impact of toolpath (2011) Software-based tool path evaluation for environmental
strategy. J Clean Prod 78:63–74 sustainability. J Manuf Syst 30(4):241–247
Artigues C, Lopez P, Hait A (2013) The energy scheduling problem: Kuram E, Ozcelik B, Bayramoglu M, Demirbas E, Simsek B (2013)
industrial case-study and constraint propagation techniques. Int Optimization of cutting fluids and cutting parameters during
J Prod Econ 143(1):13–23 end milling by using D-optimal design of experiments. J Clean
Avram O, Xirouchakis P (2011) Evaluating the use phase energy Prod 42(3):159–166
requirements of a machine tool system. J Clean Prod Kusiak A (2017) Smart manufacturing must embrace big data.
19:699–711 Nature 544:23–25
Bilga P, Singh S, Kumar R (2016) Optimization of energy consump- Le C, Pang C (2013) Fast reactive scheduling to minimize tardiness
tion response parameters for turning operation using Taguchi penalty and energy cost under power consumption uncertainties.
method. J Clean Prod 137:1406–1417 Comput Ind Eng 66(2):406–417
Campatelli G, Lorenzini L, Scippa A (2014) Optimization of pro- Lee C, Hsieh K, Hsieh S, Hsiao H (2014) A dynamic data placement
cess parameters using a response surface method for minimizing strategy for Hadoop in heterogeneous environments. Big Data
power consumption in the milling of carbon steel. J Clean Prod Res 1:14–22
2014 66:309–316 Li W, McMahon C (2007) A simulated annealing-based optimization
Camposeco-Negrete C (2013) Optimization of cutting parameters approach for integrated process planning and scheduling. Int J
for minimizing energy consumption in turning of AISI 6061 Comput Integr Manuf 20(1):80–95
T6 using Taguchi methodology and ANOVA. J Clean Prod Li L, Yan J, Xing Z (2013) Energy requirements evaluation of mill-
53(16):195–203 ing machines based on thermal equilibrium and empirical mod-
Chen Y, He F, Wu Y, Hou N (2017) A local start search algorithm eling. J Clean Prod 52:113–121
to compute exact Hausdorff distance for arbitrary point sets. Li J, Tao F, Cheng Y, Zhao L (2015a) Big data in product lifecycle
Pattern Recognit 67:139–148 management. Int J Adv Manuf Technol 81(1):667–684
Choi Y, Paul X (2014) Production planning in highly automated manu- Li X, Li W, Cai X, He F (2015b) A hybrid optimization approach
facturing system considering multiple process plans with different for sustainable process planning and scheduling. Integr Comput
energy requirements. Int J Adv Manuf Technol 70(9):853–867 Aided Eng 22(4):311–326
Dai M, Tang D, Giret A, Salido M, L i W (2013) Energy-efficient Li C, Chen X, Tang Y, Li L (2016) Selection of optimum parameters
scheduling for a flexible flow shop using an improved genetic- in multipass face milling for maximum energy efficiency and
simulated annealing algorithm. Robot Comp Integr Manuf minimum production cost. J Clean Prod 140:1805–1818
29(5):418–429 Li K, He F, Yu H, Chen X (2017a) A correlative classifiers approach
Dai M, Tang D, Xu Y, Li W (2015) Energy-aware integrated process based on particle filter and sample set for tracking occluded
planning and scheduling for job shops. Proc Inst Mech Eng Part target. Appl Math A J Chin Univ 32(3):294–312
B J Eng Manuf 229(1):13–26 Li L, Li Y, Tang Y (2017b) Influence factors and operational strate-
Diaz N, Redelsheimer E, Dornfeld D (2011) Energy consumption gies for energy efficiency improvement of CNC machining. J
characterization and reduction strategies for milling machine Clean Prod 161:220–238
tool use. In: Proceedings of the 18th CIRP international con- Li K, He F, Yu H, Chen X (2018a) A parallel and robust object
ference on life cycle engineering globalized solutions for tracking approach synthesizing adaptive bayesian learning and
sustainability in manufacturing, pp 263–267. https ​ : //doi. improved incremental subspace learning. Front Comput Sci.
org/10.1007/978-3-642-19692​-8_46 https​://doi.org/10.1007/s1170​4-018-6442-4
Fang K, Lin B (2013) Parallel-machine scheduling to minimize tardi- Li K, He F, Yu H (2018b) Robust visual tracking based on convo-
ness penalty and power cost. Comput Ind Eng 64(1):224–234 lutional features with illumination and occlusion handling. J
Fratila D, Caiza C (2011) Application of Taguchi method to selection Comput Sci Technol 33(1):223–236
of optimal lubrication and cutting conditions in face milling of Li X, Li W, He F (2018c) A multi-granularity NC program optimiza-
AlMg3. J Clean Prod 19(6):640–645 tion approach for energy efficient machining. Adv Eng Softw
115:75–86

13
A cloud-terminal-based cyber-physical system architecture for energy efficient machining…

Lin W, Yu D, Zhang C, Liu X, Zhang S, Tian Y, Liu S, Xie Z (2015) large-scale wireless sensor networks. J Ambient Intell Humaniz
A multi-objective teaching learning-based optimization algo- Comput. https​://doi.org/10.1007/s1265​2-017-0619-9
rithm to scheduling in turning processes for minimizing makes- Sun J, He F, Chen Y, Chen X (2016) A multiple template approach for
pan and carbon footprint. J Clean Prod 101:337–347 robust tracking of fast motion target. Appl Math A J Chin Univ
Lin W, Yu D, Zhang C, Tian Y, Liu S, Luo M (2016) Multi-objec- 31(2):177–197
tive optimization of machining parameters in multi-pass turn- Tang D, Dai M, Salido M, Giret A (2016) Energy-efficient dynamic
ing operations for low-carbon manufacturing. J Eng Manuf scheduling for a flexible flow shop using an improved particle
231(13):2372–2383. https​://doi.org/10.1177/09544​05416​62909​ swarm optimization. Comput Ind 81(3):82–95
8 Tao F, Qi Q (2017) New IT driven service-oriented smart manufactur-
Liu Y, Dong H, Lohse N, Petrovic S, Gindy N (2014) An investiga- ing: framework and characteristics. IEEE Trans Syst Man Cybern
tion into minimizing total energy consumption and total weighted Syst. https​://doi.org/10.1109/TSMC.2017.27237​64
tardiness in job shops. J Clean Prod 65:87–96 Tao F, Bi L, Zuo Y, Nee A (2017) A cooperative co-evolutionary algo-
Liu Y, Dong H, Lphse N (2016) A multi-objective genetic algorithm rithm for large-scale process planning with energy consideration. J
for optimization of energy consumption and shop floor production Manuf Sci Eng 139(6):061016. https​://doi.org/10.1115/1.40359​6
performance. Int JProd Econ 179:259–272 Tao F, Cheng J, Qi Q, Zhang M, Zhang H, Sui F (2018a) Digital twin
Lv X, He F, Cai W, Cheng Y (2017) A string-wise CRDT algorithm driven product design, manufacturing and service with big data.
for smart and large-scale collaborative editing systems. Adv Eng Int J Adv Manuf Technol 94(9):3563–3576
33:397–409 Tao F, Qi Q, Liu A, Kusiak A (2018b) Data-driven smart manufac-
Lv X, He F, Cai W, Cheng Y (2018) Supporting selective undo of turing. J Manuf Syst. https​://doi.org/10.1016/j.jmsy.2018.01.006
string-wise operations for collaborative editing systems. Future Vijayaraghavan A, Dornfeld D (2010) Automated energy monitoring of
Gener Comput Syst 82:41–62 machine tools. CIRP Ann Manuf Technol 59(1):21–24
Mohamed F, Abderrahlm B, Mostafa S (2017) Energy efficient and Wang L (2015) An overview of function block enabled adaptive pro-
fault tolerant distributed algorithm for data aggregation in wireless cess planning for machining. J Manuf Syst 35:10–25
sensor networks. J Ambient Intell Humaniz Comput. https​://doi. Wang Y, Lin L (2014) Joint production and energy modeling of sus-
org/10.1007/s1265​2-018-0704-8 tainable manufacturing systems: challenges and methods. In: the
Moon J, Shin K, Park J (2013) Optimization of production schedul- ASME 2014 International Manufacturing Science and Engineer-
ing with time dependent and machine-dependent electricity cost ing Conference (MSEC’14). https​://doi.org/10.1115/MSEC2​
for industrial energy efficiency. Int J Adv Manuf 68(61):523–535 014-4068
Mouzon G, Yildirim M (2008) A framework to minimise total energy Wang L, Song Y, Gao Q (2009) Designing function blocks for distrib-
consumption and total tardiness on a single machine. Int J Sustain uted process planning and adaptive control. 22: 1127–1138
Eng 1(2):105–116 Wang J, Wu X, Fang X (2015a) A two-stage ant colony optimization
Newman S, Nassehi A, Imani-Asrai R, Dhokia V (2012) Energy effi- approach based on a directed graph for process planning. Int J Adv
cient process planning for CNC machining. CIRP J Manuf Sci Manuf Technol 80(5):839–850
Technol 5(2):127–136 Wang S, Lu X, Li X, Li W (2015b) A systematic approach of process
Ni B, He F, Pan Y, Yuan Z (2016) Using shapes correlation for active planning and scheduling optimization for sustainable machining.
contour segmentation of uterine fibroid ultrasound images in J Clean Prod 87:914–929
computer-aided therapy. Appl Math A J Chin Univ 31(1):37–52 Wang L, Guo C, Li Y, Du B, Guo S (2017) An outsourcing service
Pavanaskar S, Pande S, Kwon Y, Hu Z, Sheffer A, McMains S (2015) selection method using ANN and SFLA algorithms for cement
Energy efficient vector field based toolpaths for CNC pocket equipment manufacturing enterprises in cloud manufacturing. J
machining. J Manuf Process 20:314–320 Ambient Intell Humaniz Comput. https​://doi.org/10.1007/s1265​
Petrovic M, Mitic M, Vukovic N, Miljkovic Z (2016) Chaotic particle 2-017-0612-3
swarm optimization algorithm for flexible process planning. Int J Wen X, Li X, Gao L, Sang H (2014) Honey bees mating optimiza-
Adv Manuf Technol 85(9):2535–2555 tion algorithm for process planning problem. J Intell Manuf
Qian B, Wang Q, Hu R, Zhou Z, Yu C, Zhou Z (2017) An effective 25:459–472
soft computing technology based on belief-rule-base and particle Wu Y, He F, Zhang D, Li X (2015) Service-oriented feature-based data
swarm optimization for tipping paper permeability measurement. exchange for cloud-based design and manufacturing. IEEE Trans
J Ambient Intell Humaniz Comput. https:​ //doi.org/10.1007/s1265​ Serv Comput Online. https​://doi.org/10.1109/TSC.2015.25019​81
2-017-0667-1 Wu M, Li X, Liu C, Liu M, Zhao N, Wang J, Wan X, Rao Z, Zhu L
Rangarajan A, Dornfeld D (2004) Efficient tool paths and part orien- (2017) Robust global motion estimation for video security based
tation for face milling. CIRP Ann Manuf Technol 53(1):73–76 on improved k means clustering. J Ambient Intell Humaniz Com-
Salido M, Escamilla J, Giret A, Barber F (2016) A genetic algorithm put. https​://doi.org/10.1007/s1265​2-017-0660-8
for energy efficiency in job-shop scheduling. Int J Adv Manuf Yan J, Li L (2013) Multi-objective optimization of milling param-
Technol 85(5):1303–1314 eters—the trade-offs between energy, production rate and cutting
Salonitis K, Ball P (2013) Energy efficient manufacturing from quality. J Clean Prod 52:462–471
machine tools to manufacturing systems. In: 46th CIRP Confer- Yan J, Li L, Zhao F, Zhang F, Zhao Q (2016) A multi-level optimiza-
ence on Manufacturing Systems (CIRP CMS 2013) vol, 7, pp tion approach for energy-efficient flexible flow shop scheduling.
634–639 J Clean Prod 137:1543–1552
Sarıkaya M, Yılmaz V, Güllü A (2016) Analysis of cutting param- Yan X, He F, Chen Y (2017) A Novel hardware/software partition-
eters and cooling/lubrication methods for sustainable machining ing method based on position disturbed particle swarm optimi-
in turning of Haynes 25 super alloy. J Clean Prod 133:172–181 zation with invasive weed optimization. J Comput Sci Technol
Sharma D, Dhurandher S, Agarwal D (2018) kROp: k‑Means cluster- 32(2):340–355
ing based routing protocol for opportunistic networks. J Ambi- Yan X, He F, Hou N (2018) An efficient particle swarm optimization
ent Intell Humaniz Comput. https ​ : //doi.org/10.1007/s1265​ for large scale hardware/software co-design system. Int J Coop Inf
2-018-0697-3 System 27(1):1741001–1741028
Su S, Zhao S (2017) A hierarchical hybrid of genetic algorithm
and particle swarm optimization for distributed clustering in

13
 X. X. Li et al.

Yi Q, Li C, Tang Y, Chen X (2015) Multi-objective parameter optimi- Zhang C, Sheng B, Yin X, Zhao F, Shu Y (2017a) Research and devel-
zation of CNC machining for low carbon manufacturing. J Clean opment of off-line services for the 3D automatic printing machine
Prod 95:256–264 based on cloud manufacturing. J Ambient Intell Humaniz Comput.
Yu H, He F, Pan Y (2018) A novel region-based active contour model https​://doi.org/10.1007/s1265​2-017-0615-0
via local patch similarity measure for image segmentation. Mul- Zhang D, He F, Han S, Zou L, Wu Y, Chen Y (2017b) An efficient
timed Tools Appl. https​://doi.org/10.1007/s1104​2-018-5697-y approach to directly compute the exact Hausdorff distance for 3D
Zhang Y, Ge L (2015a) Method for process planning optimization point sets. Integr Comput Aided Eng 24(3):261–277
with energy efficiency consideration. Int J Adv Manuf Technol Zhou Y, He F, Qiu Y (2017) Dynamic strategy based parallel ant colony
22(9):2197–2207 optimization on GPUs for TSPs. Sci China Inf Sci 60(6):068102
Zhang H, Zhao F, Fang K, Sutherland J (2014) Energy-conscious flow Zhou Y, Fazhi He F, Hou N, Qiu Y (2018) Parallel ant colony opti-
shop scheduling under time-of-use electricity tariffs. CIRP Ann mization on multi-core SIMD CPUs. Future Gener Comput Syst
Manuf Technol 63(1):37–40 79(2):473–487
Zhang Y, Zou P, Li B, Liang S (2015b) Study on optimized principles
of process parameters for environmentally friendly machining Publisher’s Note Springer Nature remains neutral with regard to
austenitic stainless steel with high efficiency and little energy jurisdictional claims in published maps and institutional affiliations.
consumption. Int J Adv Manuf Technol 79(1):89–99
Zhang D, He F, Han S, Li X (2016a) Quantitative optimization of inter-
operability during feature-based data exchange. Integr Comput
Aided Eng 23(1):31–51
Zhang Z, Tang R, Peng T, Tao L, Jia S (2016b) A method for minimiz-
ing the energy consumption of machining system: integration of
process planning and scheduling. J Clean Prod 137:1647–1662

13