A PROJECT REPORT ON

HYBRID APPROACH FOR PREDICTIVE MAINTENANCE

USING DIGITAL TWIN

THESIS
SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR
THE AWARD OF THE DEGREE OF

BACHELOR OF TECHNOLOGY
IN
MECHANICAL ENGINEERING

SUBMITTED BY
MEHR BILLAWARIA 17070125067
SHERON SISODIYA 17070125108
SNEH SONAM 17070125113
UTKARSH NATH 17070125122

Under the Guidance of

PROF. PRIYA JADHAV

Co-Guide- Dr. SATISH KUMAR

SYMBIOSIS INSTITUTE OF TECHNOLOGY

A CONSTITUENT OF SYMBIOSIS INTERNATIONAL (DEEMED UNIVERSITY)
Pune - 412115

YEAR- 2017-2021
i
 A PROJECT REPORT ON

HYBRID APPROACH FOR PREDICTIVE MAINTENANCE

USING DIGITAL TWIN

THESIS
SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENT FOR
THE AWARD OF THE DEGREE OF

BACHELOR OF TECHNOLOGY
IN
MECHANICAL ENGINEERING

SUBMITTED BY
MEHR BILLAWARIA 17070125067
SHERON SISODIYA 17070125108
SNEH SONAM 17070125113
UTKARSH NATH 17070125122

Under the Guidance of

PROF. PRIYA JADHAV

Co-Guide- Dr. SATISH KUMAR

SYMBIOSIS INSTITUTE OF TECHNOLOGY

A CONSTITUENT OF SYMBIOSIS INTERNATIONAL (DEEMED UNIVERSITY)
Pune - 412115

YEAR- 2017-2021

ii
 CERTIFICATE

The thesis entitled “Hybrid approach for Predictive Maintenance using Digital Twin”
submitted to the Symbiosis Institute of Technology, Pune for the award of B. Tech in Mechanical
Engineering is based on my original work carried out under the guidance of Prof. Priya Jadhav.
The thesis has not been submitted elsewhere for award of any degree.

The material borrowed from other sources and incorporated in the thesis has been duly
acknowledged and/or referenced.

I understand that I, myself could be held responsible and accountable for plagiarism, if any,
detected later on.

Date: 5 December 2020

Utkarsh Nath Sneh Sonam Sheron Sisodiya Mehr Billawaria

Co-Research Supervisor
Dr. Satish Kumar

Research supervisor HOD Director

Prof. Priya Jadhav Colonel (Dr.) Anand Pandey Dr. Ketan Kotecha

iii
CONTENTS

LIST OF FIGURES 5

LIST OF TABLES 5

LIST OF ABBREVIATION 6

ABSTRACT 7

CHAPTER I: INTRODUCTION 1

CHAPTER II: LITERATURE REVIEW 3

2.1 CNCMT 6

2.2 CURRENT PREDICTIVE MAINTENANCE METHODS

2.3 THE DEVELOPMENT OF DIGITAL TWIN

2.4 THE SYNTHETIC NATURE OF DIGITAL TWIN MODEL

2.5 FEED DRIVES

2.6 INTEGRATED MODEL OF SERVO FEED DRIVE AND CUTTING FORCE

2.7 LITERATURE GAP

CHAPTER III: AIM AND OBJECTIVE 8

3.1 AIM

3.2 OBJECTIVE

CHAPTER IV: METHODOLOGY 9

4.1 DESIGN OF DTMT DESCRIPTIVE MODEL 9

4.2 MULTI-DOMAIN MODEL IMPLEMENTATION 9

4.3 CUTTING TOOL DATA-DRIVEN MODEL REALISATION 9

iv
4.4 MODELLING AND SIMULATION METHOD OF CNCMT 10

4.5 MILLING OPERATION ON X AND Y AXIS

4.6 DESIGN OF EXPERIMENT

CHAPTER V: RESOURCES AND LIMITATIONS 15

CHAPTER VI: RESULTS AND DISCUSSIONS

CHAPTER VII: CONCLUSION

BIBLIOGRAPHY 17

APPENDICES

v
LIST OF FIGURES/TABLES

Fig. 1 Current predictive maintenance methods.

Fig. 2 Evolution of Digital Twin technology

Fig. 3 Characteristics, service, advantage and enabling factors defining tool system
digital twin model.

Fig. 4 Linear dynamics of a feed drive

Fig. 5 The architecture of an end-milling process and a servo feed drive for a CNC
machine tool system

Fig. 6 Flowchart of Modelling and Simulation

Fig. 7 CAD model of Milling Machine

Fig. 8 SIMULINK block representation of CAD model

Fig. 9 Block representation of different parts of the CAD model

Fig.10 (i) Transfer Fcn block and its parameter

Fig.10 (ii) Transfer Fcn block and its parameter

Fig. 11 Output flow of a integrator block

Fig. 12 Integrator block and its parameter

Fig. 13 To workspace block and block parameter

Fig. 14 Mechanics Explorer window for To workspace block

Fig. 15 Scope Block and its output window

Fig. 16 Saturation block and its block parameter

vi
Fig. 17 SIMULINK one mass model for Y-axis feed drive model

Fig. 18 SIMULINK one mass model for X-axis feed drive model

Fig. 19 Cutting force model for X-axis and Y-axis feed drive model

Fig. 20 Simscape model of mechanical elements of Y-axis feed drive of virtual CNC
portal machine tool

Fig. 21 Graphical representation of feed rate for the milling operation

Fig. 22 Graphical representation of cutting force for the milling process

Fig. 23 Influence of different values of position loop gain Kv on compliance of

position loop of feed drive

Fig. 24 Influence of different values of proportional gain of velocity controller Kp on

compliance of velocity loop of feed drive

Fig. 25 Angle V/S time for Simscape Multibody model of Y-drive

Fig. 26 Rotational velocity V/S time for Simscape Multibody model of Y-drive

Table 1 Parameters used in creating the SIMULINK model of Feed Drive System

vii
LIST OF ABBREVIATIONS

CNCMT Computer Numerical Control Machine Tool

MTM Multi-tasking machine

DT Digital Twin

GRU Gated Recurrent Unit

ARMA Auto Regressive Moving Average

PHM Prognostics and Health Monitoring

CPS Cyber Physical System

ANFIS Adaptive Neuro Fuzzy Inference System

IMT Intelligent Machine Tool

viii
ABSTRACT

Computer Numerical Control (CNC) Machines are an indispensable part of today’s

manufacturing industry. They control machining tools like mills, drills, lathes, etc. by the means
of an automated computer. The tools then perform certain actions on the given material that needs
to be processed, based on the instructions it is programmed to perform, to meet given
specifications. CNC systems reduce the need for manpower and increase the level of automation,
speed, accuracy, precision and quality of products produced.

CNC Machine Tools are sophisticated tools that perform certain actions like removal of
material, cutting, finishing, etc. to produce complicated parts with high level of accuracy to meet
industry standards. CNCMT is a typical manufacturing equipment and is therefore known as the
mother machine of industry. There are different kinds of machine tools present to perform different
kinds of tasks and functions. One of their key features is their high precision, speed and cost. This
makes them the superior choice for projects and activities that demand high levels of accuracy,
reliability and quality. So, any fault in CNC machine tool may decrease the precision, affecting
the production if time to time troubleshooting is not done .In case troubleshooting is not performed
regularly, the precision in the process might drop which will directly influence production, due to
a fault in the CNCMT. For this reason, there is a big significance of reliability of CNCMT.

This is where we introduce the concept of predictive maintenance. Predictive Maintenance is

a method of maintenance that uses various tools like data analysis, different methods to detect any
kind of irregularity in a given process or operation to be able to predict any future defects in a
given operation, process or piece of equipment. This then enables us to perform timely
maintenance and repair to prevent the failure or defect from occurring. The method of predictive
maintenance is an effective one because it helps in the prevention of errors, liabilities and
casualties. This kind of maintenance is gaining more attention in today’s industry since it can
reduce the cost of downtime and increase the availability of the manufacturing equipment, thus
resulting in higher output. Present techniques are unable to achieve timely, accurate and intelligent
assessments due the lack of consideration in variety of the status as well as consistency in the life
cycle of CNC machine tool.

ix
 For Predictive, efficient and precise management of CNCMT, a hybrid approach is studied
which is driven by a DT. It is based on the DT model and is a data driven hybrid. Research is done
on an algorithm of hybrid predictive maintenance that is focused on DT along with the DT data
and the proposed framework. Digital twin is nothing but a digital replication of the real life model.
It uses different kinds of sensors and data collection methods to predict the life of the machine and
possible occurrences of failures. With the help of hybrid approach driven by digital twin we can
come up with a more effective way for maintenance prediction to enable us to perform timely fixes
in order to prevent any kind of failure and extend the life of the CNCMT.

Matlab SIMULINK is the software used to virtually model the system. The effect on the
positioning accuracy & dynamic stiffness were simulated, evaluated, and checked by adjusting
and optimizing multiple feed drive variables such as position loop gain (Kv), relative gain of the
velocity controller (Kp), integral gain of velocity controller (Tn), electrical drive time constant
(Te), total moving mass m, sampling cycle (Ts), etc. The cutting force model was also considered
to forecast the tracking errors in the milling process.

At last, the tracking errors that had been estimated by the suggested model were compared to
the ones estimated by the traditional model without considering the transmission path.

x
CHAPTER 1

INTRODUCTION

Milling is a process conducted by a machine wherein the cutters rotate in the direction of the
angle with the tool axis to extract the substance (material) from the workpiece. With the aid of
milling machines, multiple operations as well as processes can be done, ranging from small to
massive objects. One of the most common manufacturing techniques used, to produce items and
parts of various kinds of sizes and shapes in machinery stores, factories and industries, is milling
machining.

Machines called (MTMs) that are also multifunctional and adept at performing tasks on
products that require both milling and turning, are also known as milling machines. The milling
machine has a cutter mounted on it that helps to extract the material from the workpiece's surface.
Once the material has cooled down, this is then taken off from the milling machine.

Milling machine are of two different types - the horizontal mill and the vertical mill. These are
talked about in further detail below:

Vertical mill: The spindle axis of the vertical mill is arranged vertically and rotates by remaining
on the same axis. The spindle may even be extended to perform certain operations like drilling and
cutting. Apart from this, there also exist two more kinds of vertical mill. The first being the turret
mill and the second, the bed mill.

The former comprises a table that in terms of cutting the material, travels perpendicular and
parallel to the spindle axis. However the spindle is stationary. With this by rotating the knee and
dropping or lifting the quill, two cutting techniques can be carried out.

The bed mill, on the hand, comprises a table that moves perpendicularly to the axis of the
spindle while the spindle itself travels parallel to the spindle axis.

Horizontal mill :A related cutter is indeed a horizontal mill, but their cutters are mounted on a
horizontal arbour. To help in processes that call for multi angle milling, there also exist horizontal

1
mills that consist of rotary tables. Such tables are known as universal tables. In addition to this the
horizontal mill will also use all the instruments that are found in a vertical mill.

Mass production has now evolved into customized production with the steady increase of smart
and automated manufacturing. As technology keeps developing, manufacturing develops with it.
What we know now, as digital networked manufacturing was earlier digital manufacturing and
will soon become the latest intelligent manufacturing by becoming more digitized, intelligent and
automated. Currently, the manufacturing process and verifying prototypes that are physical, to be
able to detect any weak points for optimization of the design is very intensive in terms of cost and
time consumption for machine tool builders to be able to afford. Therefore, to scale down the cost
and time consumption for testing the hardware and making iterative refinements to the physical
prototype, latest processes of machine tool design implement “virtual prototyping” technology.
This machine tool’s virtual prototype is nothing but a computerized simulated model of the actual,
physical item or product. This virtual model can then be studied and analyzed like the real product.
During the process of designing, iterative change of the machine tool’s simulated model and
exercising of design variations before the criteria for performance are met, significantly brings
down the cost and the time to develop the product. [12].

As the industry's mother machine, CNC Machine Tool(CNCMT) is one form of significant
production equipment [15].CNC machine tool’s performance and reliability is determined by the
quality of products it will process during its working. Feed drives are the most significant
components of each machine tool. The feed drives are very complex electro-mechanical systems
for CNC machine tools. In particular, good dynamics of control, precision of position, load
stiffness, capability of acceleration and process velocity must be displayed [13, 14].The non-
linearity in the process along with the noise is what causes degradation in the performance of the
Computer Numerical Control machine tool. As a result, an error occurs in the prognosis.

Errors that affect the multi-axis machine are described as the difference between the end
effector's actual and calculated position and orientation. These problems involve the efficiency of
the system to deteriorate, which in turn directly affects the requirements of the finished product.
Such errors must be observed and removed in order to ensure a consistent level of service for a
system. Two control priorities are maintained by machine tool control systems: To pursue as
accurately as possible a pre-specified trajectory, and to retain a pre-specified advance speed. These
2
systems split into small sections of the target trajectory, and move along these segments to form
the direction. Whenever the trajectories are multi-dimensional, the structures separate the
acceleration in various directions into parallel motions. To achieve the optimal direction, the
drivers of multiple axes have to travel synchronously with each other. The distance difference
between the actual path and the programmed path is known as the contour error. In order to
decrease contour error in linear movements it is important to provide relatively high gains and also
to equal the gains of the individual axis to decrease contour error in linear movements. The
disparity in open-loop axis gain values is the system's discrepancy, which produces a contour error.
Tracking errors are defined as the change in parameters like speed, displacement, that are used in
the feed drive system, with respect to time. Therefore, two main approaches to enhancing the
control efficiency of drive systems have arisen. The first method which is called tracking control,
focuses exclusively on minimizing the tracking error on each plane, resulting in an indirect
decrease in the contour error. The second method, called the contouring control, attempts to
measure in real time the contour error and to use this prediction and approximation in the feedback
control rule.

These changes have created challenges for fault detection and predictions which are considered
for new products, increasing the accuracy of prediction and product customization. Signal analysis
is typically the method used for the detection of fault and this is mostly dependent on the operator’s
experience and inspections. Usually statistical data are used to make predictions using the runtime
data. Digital twin is what we will use for fault prediction to optimize the dependability,
accessibility, and safety of the machine.

In the life cycle of industrial products, different forms of engineering applications and
digitalized machinery are widely adopted. As a result, immense data of different sorts is likewise
being delivered. These data, however are hysteretic and separated from one another, resulting in
poor productivity and limited use of such useful information. In its early planning stage, simulation
related to the theoretical and static models was a conventional and effective method for verifying,
validating, and integrating a system, but during system run-time little emphasis has been placed
on the simulation application. More data will be obtained with the advancement of new-generation
intelligence and digitalization technology, and it is an opportunity to figure a way to implement all
this data in detail. As a response, the idea of the digital twin has generated a great deal of concern

3
and is emerging very fast. In scholarly and industrial societies, conflicts and debates over the
principles, paradigms, structures, implementations and innovations of digital twins are all on the
increasing trend. 240 scholarly articles about digital twins are identified and categorized after a
complete review of multiple databases and appropriate examination according to the suggested
criteria.

There are numerous defects and reasons that cause hindrance and result in an inaccurate output.
Noise and the lack of comprehension of the various attributes of the machine are included in these.
In order to execute the hybrid predictive approach and overcome the defects, many scholars
proposed the idea of a Digital Twin. The principle of Digital Twin (DT) offers an optimal answer
for Implementation of a hybrid predictive maintenance approach . DT is an ancestor Cyber-
physical device enabling process (CPS) [25 - 27], which includes physical model, data from real-
time sensing and historical running data. DT provides a digital model of physical equipment based
on high fidelity. Physical laws that acquire data sensing from real-time during the process
Equipment operation and preservation of historical running data for future use. DT can,
meanwhile, provide accurate information through keen setting awareness and data mining, thus
achieving high levels of data mining. Fidelity and dynamic modeling, along with multi-domain
modeling Consistency management approach for model. DT, therefore, gives the Hybrid
predictive maintenance strategy option.

Currently, the reliability statistics method, the data driven method and the physical model
based method, are the three methods that are most commonly used for predictive maintenance of
complex equipment. Each approach would have its own features and fields that relate. In order to
realize the merging of models and the data, the hybrid approach involves multi-domain model
creation, data-driven model building, and development of hybrid algorithms.

Hybrid predictive maintenance can be implemented effectively with the help of DT concept
CPS are enabled by Digital Twins. Digital Twin allows the CPS method to be executed. It uses
laws of physics and thus results in having an extremely accurate digital model. This model can
then cache historical data used while running to be used further. It can also procure and sense data
in real time one the equipment is utilized. Many scholars have suggested the hybrid predictive
maintenance approach [23,24] to resolve the defects of the above-mentioned approaches. The
benefits of each process are augmented by the creation of a fusion technique. To understand the
4
hybrid predictive maintenance strategy, however it is important to develop precise mathematical
and physical Models, obtain accurate running data in real time, and render useful data.

The Chapter – 2 of the thesis contains the literature survey regarding the CNCMT and the concept
of DT. It also includes the literature on current predictive maintenance methods, i.e., reliability
statistics, the method based on physical models and the method based on data. We have discussed
the hybrid predictive maintenance method and the feed drives. The chapter forms the bases of the
current study as it includes all the information that will guide and help open new paths to the
research and enhance the present work.

The Chapter – 3 consists of the problem statement and the objective of the research work. The
problem statement is formulated on the basis of the literature survey and the literature gap.

The Chapter – 4 includes all the information of methods used in the work. The 3-D CAD model
of milling machine and its Simulink block representations are mentioned. Simulink is a
programming environment with graphical representation based on MATLAB to model, simulate
and analyze multi-domain dynamic structures. A graphical block diagramming tool and a
customized collection of block libraries are its main feature. Therefore, the chapter also contains
the Simulink one mass model for X and Y axis Feed drive. Lastly, a table which shows the
parameters used in creating the SIMULINK model of Feed Drive System is included.

The Chapter – 5 comprises all the simulation results and discussions. It shows 4 graphs : Feed
Rate, cutting force, position loop and velocity loop.

The Chapter - 6 It gives an idea about the limitations of using software like ANSYS and Matlab

5
CHAPTER 2
LITERATURE REVIEW

Production time studies in smaller businesses show that each operation's machining and
assembly activities constitute just 5% of the entire length of the operation. The other 95 percent
of the operational time mirrors the area of physical resources during arrangement periods during
the primer tasks or the holding up before the following activities. Most of the working time is
often spent excessively on transport and storage, etc. Nyhuis P et al., 2009 [1].

A change in view to 'virtual’ from 'real output , as a result has led to an expansion in the field of
research since the early 1990s. Simulating some of the operations of a physical production
system has become possible with the help of computers. The primary objective is to comprehend
and reenact the conduct of a particular creation framework on a machine preceding physical
manufacturing, thus minimizing the quantity of tests and the shop floor experiments.
Less material is lost by the use of a computer device and interruptions can be avoided in the ope
ration of an actual machine on the shop floor Radharamanan R, 2002 [2].

The method developed for the success of virtual output was focused on literature findings and
practical experience. The approach is referred to as PoVEIR (Production in Virtual Environment
Integrating Resources) which is explicitly designed to optimize processes of production in
smaller companies characterized by special small-scale and variable volumes of production.
So far, different modelling and simulation techniques have been introduced in the field of CNC
machining. They are important for the improvement and cost reduction of manufacturing.
Onosato M (1993), Mwinuka T.E (2015), Debevec M (2014), Čuš, F (2015), Hrelja M (2014)
and Todorovic P (2014) [3,4, 5, 6, 7, 8].

The target of current assembling advances is to deliver the initial segment accurately and in the
savviest path in the briefest time of time. The development and testing of actual models are
significant bottlenecks for the effective and financially profitable creation of present day machine

6
instruments, as the product complexities increment and the serious item life process durations are
diminished Weck, M et al., 2003 [9].

The making of present day programming tools that are used for simulation of properties of the
product have been significantly improved lately, initiated primarily by the automotive
and aircraft industries Weck M. et al., 2003 [10].

Before the first physical prototype is constructed, with the help of software that are advance
and hardware systems engineers test and optimize critical characteristics of the product
with virtual prototypes. For the various design phases of a machine tool, a large range of software
tools are available Weule, H et al., 2003 [11].

As of now, the machine device developers are unable to afford the cost centered manufacturing,
testing and the time consumed in actual models/prototypes to find weak
points and then enhance the design. Instead, modern machine tool design processes employ "virt
ual prototyping"technology to decrease the expense and time of equipment testing and iterative
actual model enhancement. A computer simulation model of a physical product that can be
interpreted and analysed like a real machine is the virtual prototype of a machine tool.
During the process
of the design, iterative change of a simulated model of the machine tool and exercise variations of
the design before the performance criteria are met reduces the entire
product development time and costs significantly Altintas Y et al., 2005 [12].

2.1 CNCMT (CNC Machine Tool)

As the industry's mother computer, the CNC machine tool (CNCMT) is one type of significant
production equipment. The achievement of CNCMT's consistency and reliability determines the
quality of the goods it has being stored. CNCMT, however can unpredictably incur defects, which
low accuracy, stagnation of output or even significant loss and loss can be caused. Casualties if
there is no prompt troubleshooting. Therefore, to prevent any fault and to make these faults more
reliable in CNCMT, we require predictive maintenance. T. Kjellberg et al., 2009 [13].

7
 Traditionally, the form of reliability statistics is generally used for Predict based on the st
atistical features of historical characteristics, data on faults. However the approach of reliability s
tatistics needs many checks, Historical & detailed data, that usually creates difficulty to apply
for complex and complex Large-scale , costly facilities, like airplanes, machine tools, electric
power stations, nuclear devices etc. Another is the actual model-based technique which by
characterizing the advanced device model, actual model for failure ( the failure model) & model
for fatigue life ( the fatigue life model), conducts fault predictions N. Cauch et al., 2017 [14]. It is
intensely reliant on the ability and information on specialists to make a viable numerical model for
complex electromechanical systems. The variety of status and quality of equipment during its life
cycle are not taken into account by these two types of approaches, resulting in unreliable, untimely
and unintelligent prediction outcomes.

As in today’s generation technologies such as Arduino based sensing, monitoring, simulation

in MATLAB and ANSYS, big data and AI are being developed. A lot of savvy strategies have
been applied to manufacturing, which give a proficient method for prescient support fueled by the
Data from manufacturing J. Wang et al., 2019 [15]. Based on the monitoring of historical
circumstances, Equipment info, characteristics reflecting the health status of equipment The data-
driven model for prediction and prediction can be extracted to construct the Diagnosis of
machinery. The performance of various calculations changes as per the qualities of the objective
framework for the data-driven process. As well as its information and inaccuracy occurs because
of the lack of understanding The physical features, noise and uncertainty of equipment Duan L et
al., 2016 [16].

Many scholars have suggested the hybrid predictive maintenance approach to resolve the
defects of the above-mentioned approaches H. Hanach (2019) and L. Liao (2016) [17,18]. The
benefits of each process are augmented by the creation of a fusion technique. To understand
predictive maintenance with a hybrid approach, however, it is important to develop precise models
that are physically and mathematically based, in order to obtain accurate running data in real time
and render useful data.

8
2.2 Current Predictive Maintenance Methods

Currently, there are 3 strategies that are mostly utilized for predictive maintenance of complex
hardware, in particular the strategy for unwavering quality measurements, the technique dependent
on actual models and the strategy dependent on information/data, refer Fig. 1.all the methods have
their own features and fields that apply Mobley R K, 2002[19].

Fig. 1 predictive maintenance methods used currently.

(1) Reliability statistics method: Statistical features of authentic shortcoming information are
utilized for deficiency forecast in the reliability statistics method, ref Fig. (a), wherein a lot of
detailed information is not required and no specific data or model that is mathematically based , is
required Shimada J (2016) and C. Li (2012) [20, 21]. A progression of different probability density

9
functions contain the information required for prediction. Weibull distribution, Bayesian method,
and fuzzy logic, etc., include fault prediction method based on reliability statistics.

These techniques are predominantly used for large batches of commercial products. It is
unsuitable for systems that contain millions of components and parts and are thus complex, or for
any other costly and complex equipment. Examples of these are aircrafts, spacecrafts and
CNCMTs. In addition, this method results in low accuracy of prediction since it does not take
factors like, the complicated working atmosphere, output degradation of the tool, etc., into
consideration. Thus further resulting in low levels of confidence in this method.

(2) Physical model-based method: The physical mathematical model reflecting the execution
deterioration of the framework is set up as indicated by the inner working component of the target
system in the physical model-based method Ortiz J et al., 2016[22], as shown in Fig. (b) from Y.
Lei (2016) and B. Iung (2008) [23,24]. The real entity of the actual object can be clearly defined
by a numerical model from which the deterioration pattern of the item can be predicted to achieve
the precise outputs of the fault prediction.

In spite of the fact that the actual model-based strategy can uncover the framework's issue
rationale without gathering a ton of information, it needs master backing to plan and build up the
model. The greater part of the hardware is commonly mind boggling mechanical and electrical
frameworks, the relating corruption models of which are hard to precisely decide because of the
absence of a mechanism of degradation.

(3) Data-driven method: The data can be collected from running devices in the data-driven
predictive maintenance method, and no exact flaw advancement model or execution debasement
measure is required M. Baptista (2018), Daily J, Canizo M (2017) and J. Wang (2019) [25 - 27].
Autoregressive models, artificial neural network support vector machines, correlation vector
machines, Gauss regression system etc. are widely used for the study of Big data. Wang et al.,
2019 [28],. As shown in Fig. (c), highlights are separated from recorded information and afterward
changed over into framework information via data collection and analysis, system information on
health status and deterioration that is concealed in data can be mined.

Numerous sensors are generally introduced to gather information to get the operational status
and data of equipment, but some key components are unable to install sensors, resulting in data
10
acquisition and further data analysis difficulties J. Wang (2017) and J. Xie (2017) [29, 30]. What's
more, in different systems, different calculations have discrepant outcomes. Therefore to achieve
the precise algorithm & parameters corresponding to it, heavy work is required for the algorithm
experiments.

In summary, there are different flaws in the existing predictive maintenance approach based on
a single technique and do not satisfy the criteria of greater precision and reliability for fault
prediction. With the advent of new technologies, as shown in Fig., the hybrid approach (d) has
been a hotspot for research L. Liao (2016), J. Wang (2015) and Y. Liang (2020) [31 - 35]. In a
hybrid approach, it is possible to follow the benefits of each method, essentially eliminating the
disadvantages of each single method.

2.3 DT Development

With the evolution of production towards intelligent mode, automatic technology,

convergence, the concern has increased about DT along with its growth, ref Fig. 2. Initially, the
idea of "twin" was initially derived from the Apollo Project of NASA, when the aircraft was ``A
real physical system is a twin body Hochhalter J et al., 2014 [36]. Nowadays a twin body model
can assist astronauts and personnel make decisions in the context of circumstances of emergency.
Further, for this Digital Twin idea, a product lifecycle is proposed for the formation of The 2002
Management Center Rosen R et al., 2015 [37]. Enabling the IoT to help new intelligent services,
the Digital Twin model connects and interacts with a physical object Wang, P and Zuehlke D.,
2015 [38, 39]. These ones, these capabilities permit makers to perceive and drive new plans of
action, including real-time simulation and simulation issue forecast Tao F et al., 2014 [40]

11
 Fig. 2. Evolution of Digital Twin technology

There are a few organizations imagining and executing genuine Digital Twin applications, for
example, General Electric, PTC, NASA, Siemens, etc.

Currently, several studies have been conducted on digital twins Rosen R(2015), W. Luo(2018),
Guo(2018), Zhang(2017), QiangLiu(2018) and Jiewu Leng(2018) [41-46]. From the above
research and studies, it can be found that most of the work has been done in the development of
digital twin-based computer tools involving technologies such as big data processing, the Internet
of Things, cyber-physical systems and cloud computing. The advantages of the advanced twin as
far as finding and anticipation of shortcoming are that it is conceivable to deal with data from
assorted or various sources and to follow the genuine working conditions, and that the prescient
outcomes might be placed in the computerized model to exhibit the accuracy of the slicing amount
or to all the more likely feed the tool systems Jinjiang Wang et al.[47].

With the advancement of technology, approaches to tool system fault prognosis are also
strengthened. There are several works on the approach of data driven method to prognosis at the
moment. New system to track the state of motor bearings and diagnosis of faults is used for under
sampling vibration signals Siliang Lu et al. [48]. The people group strategy for information dealing
with (GMDH) proposes the technique for foreseeing and identifying cutting instrument failure in
a cycle gadget Uematsu T et al. [49]. Such methods of diagnosis and prognosis based on statistical
processes have been suggested by El Wardany T, Lamiaa M, Kang J (2007) and Park S [50-53].
The above model and methods are all about the model-based approach to prognosis.

12
 There are some special benefits of current methods of fault prognosis. However there are some
drawbacks: 1) the system is always too complex to model for some, 2) the approach relies heavily
on expert expertise and requires a large number of rules, 3) the model lacks a plan in order to revise
and upgrade it further using latest test data.

Therefore, investigations are performed on Digital Twin models that are used as reference in to
help for the analysis and evaluation of faults. Big data is used for sensor information interpretation
and processing, the frequency bands are recognized by a computer control unit and directs the
actual devices by CPS to the succeeding stages. The model algorithm for the prediction of the fault
is given by deep learning(IoT) . This methodology is paired with a data-driven and based on model
service. A viable virtual and genuine model of connection can be made, which can really speak to
the kind of development and the material structure of actual items in the world that is virtual. In
addition, the projected effects are simple to make visible in 3D. This may allow operators to
appreciate, outcomes of the prognosis.

2.4 Digital Twin model- The Synthetic Nature of DT

In essence, DT is an exceptional living actual gadget representation with support for

empowering advances, for example, multi-physical simulation, AI, AR/VR & cloud service, and
so forth. It permits uninterrupted adaptation to environmental or organizational variation and
produces supreme industrial performance. It proposes tremendous potential for improving
operative functions and maintenance, also speeds up the pace of the procedure generated by the
new product.

Simply in Fig.3, we present another philosophy zeroed in on digital twin innovation

information from current writing and investigation, demonstrating a computerized twin model
instrument framework as far as six innovation highlights, three empowering factors, four
preferences, four service implementations are used to operate the improvement of the
computerized model of the twin apparatus framework.

13
Fig. 3 Characteristics, service, advantage and enabling factors defining tool system digital twin
model.

Different facets of technology (cyber-physical structures, IoT/IoS, AI, machine utilization

(energy productivity ) and operations (cloud development, multiple intelligence interaction &
virtual- physical interlinkage ) constitute the characteristics of the digital twin model. in today's
world of machines with computer and physical facilities, cyber-physical systems will create a
strong linkage between people & the world of digital information Baheti R et al., 2011[54]. The
Internet of Things (IoT/IoS) utilizes different sensors, (for example, radio recurrence
acknowledgment, infrared acceptance and worldwide situating, and so forth to deal with huge
volumes of information of various kinds to understand the elements of danger ID, shortcoming
area and wellbeing the board Kelly S et al. [55]. From a security point of view, there are several
resources (monitoring of the process, prognosis as well as the diagnosis of the faults , management
of the alarms , Shorten MTBF) in the latest put forward model of DT of the tool system.

DT accomplishes human and the machine cooperation that convey progressively to the world
and actual elements. It has a one of a kind identifier that is connected uniquely to the device of the
14
instrument. The technology is incorporated in the digital twin model with Artificial Intelligence
(ontology, ML, IoT & deep learning, etc.). Due to the feedback mechanism of the DT model, it
can progressively assist to rectify faults in the designing process of the tool system. Digital Twin
allows the model, motion, and fault visibility of the tool system, and can reliably determine
regardless of the feed framework or the fault in the device.

The enabling two factors are defined as education, knowledge & rules of conduct. Education
that is creative and innovative, provides humans with more criteria for smart production and gives
a wider view, ref Fig. 3. More specific technology for data processing needs to accommodate a
large volume of data processing. When it is designed, the code of behavior guarantees the
consistency and continuity of the digital twin model.

2.5 Feed Drives

Feed drives are the most significant components of each machine tool. The feed drives are very
complex electro-mechanical systems for CNC machine tools. In particular a good control
dynamics, capability of the acceleration , precision of the position, the load stiffness and velocity
of the process must be displayed Gross H (2006) and Pandilov Z (2014) [56, 57].

The various mistakes of each feed drive pivot for machine devices should be explored Ramesh
R et al., 2005 [58] to achieve high precision machining. A machine instrument comprises of servo
engines, spindles, structures, etc., and a complex interaction of a lot of characteristics like servo
dynamics, friction, structural vibration, the thermal effects,& the cutting dynamics is its dynamic
response Ramesh R (2000) and Muhammad BB [59, 60].

It is conceivable the traditional feed drive has simplistic linear dynamics, ref Fig.4 Here, the
control signal u '(V) is applied to the current amplifier input, denoted by the Ka (A/V) gain.
Amplifier then induces the current (A) in the armature of the motor, which gives us the output in
the Tm (Nm) motor torque, that is linearly proportional to I wherein the Kt (Nm/A) motor torque
constant. The engine shaft is additionally exposed to a disturbance force Td (Nm) notwithstanding
the current produced force going through armature, which remembers the impact of grating for the
guideways, ball screws & bearings, just similar to the cutting forces Y. Altintas et al., 2005 [61].
15
 Fig. 4 Linear dynamics of a feed drive

Various theories have examined procedures to minimize the errors that are dynamic & static,
like suppression of structural vibration, enhancement of servo rigidity, compensation for backlash
or friction, avoidance of chatter, and compensation for thermal errors. The dynamics of the
configuration of a machine tool can be calculated using the process of receptance synthesis Zhang
GP et al., 2003 [62]. The simulation of association across the structure and cutting elements of a
machine apparatus, a powerful position-subordinate multibody dynamic model of a machine
device was created Law M et al., 2013 [63]. To enhance the servo response a feedforward
controller was applied Wu J et al., 2017 [64]. To anticipate the following mistake, a strong
following servo gadget with the feedforward regulator utilizing coprime factorization and
unsettling influence spectator was implemented Ohishi K et al., 2006 [65]. The servo output of the
system table throughout pre-sliding to sliding can be strongly influenced by friction and backlash.
A variety of backlash & friction models were proposed to improve its efficiency and the related
compensation methods were investigated Erkorkmaz K (2001), Chen Y (2002) , Olsson H (1998)
and Nguyen BD (2007) [66-69]. Hence the previously mentioned writing shows that the relations
between servo elements and the technique for cutting have infrequently been tended to.

2.6 Integrated models of servo feed drive and cutting force

Milling is a cutting method that separates material from a workpiece using a milling cutter. The
design of a CNC machine instrument, which is electromechanical, incorporates a CNC regulator,
a servo drive framework, mechanical framework, etc., as introduced in Fig. The CNC deciphers

16
G-codes, increasing speed and retardation plans , and afterward dispatches the servo position
orders & axle speed orders to the servo and axle drive frameworks, separately, at various testing
times. To give better situating exactness of the machine table, the servo drive framework for the
most part embraces a full-shut circle position control by utilizing a direct scale to input the current
situation of the table. In the current segment, the model of the end-processing cutting power also
the model of servo drive are introduced independently and afterward consolidated as an
incorporated framework model Chen-Jung L et al., (2019)[70]

Fig. 5 The architecture of an end-milling process and a servo feed drive for a CNC machine tool system

17
2.7 LITERATURE GAP

The fact that the literature mainly comprises conceptual studies rather than studies of particular
cases are indicative of how DT is still in its very early phases of development. Although lower
level studies of assimilation like DS and DM do exist. As demonstrated, production planning and
control is one of our key focuses based on the latest research about DT in production and
manufacturing due to the fact that it is a chief data sink in a system of production that binds each
thing. Therefore, a lot of times simulations are used to make the most of the models at their best
because it has a middle level time horizon. Although, areas that have higher frequencies of time
can also use DT.

An example of this is the utilization of maintenance that is based on the control of the process
and also its conditions without the usage of simulations that are time intense; Rather, information
and data driven methods are used. To be able to understand the benefits that can be attained with
the help of DT, we need to research various case studies in different environments of the industry.

Most of the researchers have excluded the cutting force model in the design. We have included
it in our design because the primary execution is the machining nature of the workpiece.

18
CHAPTER – 3

AIM AND OBJECTIVES

In the manufacturing industry, the computer numerical control machine tool (CNCMT) is
considered to be the one of the most important machines of the industry. The machining process
we selected is cutting and the component we chose is the cutting tool due the major role it plays in
the process. The surface finish, and accuracy of dimensioning are some of the few important things
that make this tool very important. The key aim and objectives of our paper are as follows:

3.1 Aim

To create a multi-domain model of hybrid predictive maintenance for a Computer Numerical

Control Milling Machine (CNCMT) using MATLAB Simulink

3.2 Objectives

i.) To find the reliability of CNCMT to troubleshoot the problems in the tool to avoid negative
effects on production like loss of precision, etc.

ii.) To make a virtual model of CNC machine feed drive system using MATLAB Simulink, and
then simulate the results.

iii.) To make the cutting force model to find the tracking error in the milling process.

iv.) To compare the tracking errors predicted with the help of the presented model and the
traditional model.

19
CHAPTER 4

METHODOLOGY (WORK PROGRESS TILL DATE)

4.1 Design of DTMT descriptive model

CNCMT is an intricate mechatronic system that contains many objects with many properties.
We closely assimilate various stages of the lifecycle of a machine tool with a multi-domain
consolidated model so as to satisfy the need of DTMT. In order to recognize the self-adaptation,
self-prediction and self-evaluation of DTMT, we consider the general comprehensive performance
of the model along with how different parts are coupled. The strategies of design and simulation
of multi-domain machines include a software program interface method. We design a 3-D CAD
model of the CNC portal milling machine using design software, Solidworks as per our design
specifications. Solidworks allows for creating a high precision, more accurate and a simple to
understand designs. The model is designed using a direct lead screw feed drive for the X, and Y
axis. Feed drive is used to synchronize the movement of the X and Y axis to generate the required
machine surface. It consists of actuators and control loops.

4.2 Multi-domain model implementation

While designing the CAD model, we take into account various areas and domains of physics
like thermodynamics, hydraulics, mechanics, electronics, etc. Solidworks, Matlab, Simulink are
all software that support multi-faceted modelling. So, with the aid of these software we can create
models of objects like the cutting tool, feed drive, form subsystems and then integrate these into a
consolidated multi-domain system.

4.3 Realization of the Cutting tool data- driven model

The records accumulated from sensor-acceleration, reducing pressure (cutting force) & AE
statistics. The raw data tends to consist of trend items because of drift in sensors from variation in
temperature. To provide better accuracy in the positioning of the machine table, the model of the
feed drive comprises of a full-closed loop location control, it uses linear scale to feed the table's
current position. For the cutting force model, we assume that, he end-milling cutter also the
20
workpiece is stiff as well as the cutting force becomes proportional to the chip's cross-sectional
area.. Apart from this, the burrs in the sign waveform end up including noise signs like random
interference signal, power frequency sign and periodic interference due to the superimposition of
the noise sign with the raw facts that are sampled. Smoothing of raw information is done for the
purpose of reduction of the effect of the signal interface & improvement of the curve smoothness
of the vibration

4.4 Modelling and simulation method of CNCMT

Figure 6. Flowchart of Modelling and Simulation

1. The first step is to create a virtual three dimensional CAD model of one CNC portal milling
machine. it should have a direct lead screw feed drive for all the axes. A vertical milling machine
21
is designed in Solidwork. The PX 10 series of vertical machining centers has been designed, it
offers high cutting rigidity and low vibration for heavy cutting loads and provides for an efficient
chip flow. The tool progresses perpendicularly to the workpiece on a dedicated axis to conduct
multiple special, tiny cuts in the shape of chips to extract material from a surface.

Figure 7. CAD model of Milling Machine

2. On completion of the virtual model in Solidworks, we export the solid model into Simulink
using Simscape Multibody link. Simscape Multibody or SimMechanics enables us to work with
three dimensional mechanical systems, like robots, suspensions of the vehicles by providing a
multibody simulation environment. It enables us to model multibody systems using blocks which
represent bodies, joints, constraints, force elements, and sensors. We can also import complete
CAD assemblies from solidworks using Simscape Multibody which includes information about
mass, inertia, geometry, into the model. We are also able to generate animation through the blocks
imported to visualize the dynamics of our system.

22
 Figure 8. SIMULINK block representation of CAD model

World block- Provides entry to a special motionless, orthogonal, right-handed reference system
pre set in every mechanical model, to the universe or ground frame. In a mechanical paradigm, the
World Frame is the foundation of all frame networks. The Solver Configuration block specifies
the solver variables that are required by the model before simulation can be started. To connect
to a block diagram, it precisely needs one solver block only.

A fixed three dimensional rigid transition across two frames is defined by the Transform Block.
The translational and rotational aspects of the transformation are separately determined by two
elements. Various translations along with the rotations can be freely combined.

3. Different parts of the solid model are imported into Simulink as blocks are connected to each
other. The blocks offer very detailed information about the solid part. On exporting the 3-D CAD
model of CNC portal vertical milling machine from Solidworks into Simulink, every part of the
model gets converted into Simulink blocks with certain parameters. Blocks are the elements which
are used to create a model in Simulink, the blocks represent certain equation related to cutting
force and modelling components related to the CAD model. Different blocks for various machine
parts such as work table, saddle, chuck, handle, hand wheel, etc. are generated. The parameters
associated with the blocks can be modified inside Simulink itself as per the requirements of the
experiments.

23
 Figure 9. Block representation of different parts of the CAD model

4. The next step in modelling is to create a Feed Drive System for the X and Y axis. Every axis
requires a separate feed drive system. We create a SIMULINK one mass model for all the axes to
check the influence of changing several parameters related to the drive system and the compliance
of the drive system in the frequency domain.

Simulink blocks are imported from the block library with certain in-built parameters which
need to be changed as per the process requirements. The steps associated with importing and
editing the parameters related to the project are given below.

I. Inserting/Importing Simulink blocks

A. Open a blank model. Go to the library browser.
B. Select the block required, drag and drop it on the blank model screen.
C. Alternatively, single or double click on the screen, type the name of the block
needed into the search bar and the block gets imported into the model.
II. Editing the block parameter

Double click on the block you want to edit. A parameter edit box will open up. Enter values
according to the data available.

24
 A. Transfer Fcn Block

Fig. 10 (i) Transfer Fcn block and its parameter

Fig. 10 (ii) Transfer Fcn block and its parameter

The first value in the square bracket is always accompanied by ‘s’. S is a Laplace operator with
self-assigning values. The order for the denominator ought to be more prominent than/ can also be
equivalent to the order for the numerator.

25
 B. Integrator block

Fig. 11 Output flow of a integrator block

Fig. 12 Integrator block and its parameter

The Integrator block yields the estimation of the necessary of its info signal as for time.

26
 C. To workspace block

Fig. 13 To workspace block and block parameter

Fig. 14 Mechanics Explorer window for To workspace block.

The To Workspace block composes input signal information to a workspace. During

simulation, the block composes information to an interior support. At the point when you stop the

27
simulation or the simulation finishes, that information is kept in touch with the workspace.
Information isn't accessible until the simulation delays or stops.

D. Scope block

Fig. 15 Scope Block and its output window.

Scope block display signals generated during simulation. Usually the axes are set to auto-scale.
Go to Tools- Axes scaling, select auto-scale.

E. Saturation block

Fig. 16 Saturation block and its block parameter

28
 The saturation block restricts to the top and bottom saturation values of the input signal. By
default the upper and lower limit is set to +0.5 and -0.5 respectively which can be changed as per
the requirement.

The Simulink model for the axes provides a detailed description about the process of milling
on a vertical milling machine at a particular axis. Several parameters such as proportional gain,
velocity gain, disturbance/cutting force, torque constant, time constant, spindle speed, etc. are
considered.

Fig. 17 Simulink one mass model for Y-axis feed drive model

Step or gap is provided by the step block, between two values at the time specified. For
instance, simulation time is higher than the step size, the resulting outcome will be influenced by
the initial conditions of the process, whereas if the simulation time is more the output is the final
value of the parameter value.

29
 Fig. 18 Simulink one mass model for X-axis feed drive model

The model is designed to show the entire princess of milling carried out along the X and Y
axis using mathematical relations. For this parameter such as specific cutting force, load inertia,
Laplace operator/sampling period, etc. are considered. Transfer blocks are used to formulate the
mathematical formulas for the process. The transfer block models a linear system by using transfer
function of Laplace domain variables s. It has a preset value of zero.

Fig.19 Multibody Simscape model of Y-axis feed drive of virtual CNC portal milling machine

30
Fig.20 Simscape model of mechanical elements of Y-axis feed drive of virtual CNC portal
machine tool

4.5 Milling operation on X and Y axis

The milling operation on the X axis takes place when the mill axis moves from left to right
(when the operator is facing the machine from front of the machine) in the X-axis of the milling
machine. To easily determine the direction of the movement we can check for the positive cutter
motion of the vertical mill. If the cutter motion directions are to the right of the operator the
movement of the tool is along the X-axis, if the motion is away from the operator, then the tool
moves along the Y-axis. In a CNC milling machine, the controller always keeps track of the
position of the cutting tool with reference to a certain origin. The origin or the ‘Zero point’ is the
point where all the axes have zero value.

4.6 Design of Experiment

The Simulink model for the milling operation on the X and Y axis is designed taking into
considerations the degree of freedom around the axes and their interactions with the parameters.

31
32
 Table 1. Parameters used in creating the SIMULINK model of Feed Drive System

CNCMT is an intricate mechatronic system that contains many objects with many properties.
We closely assimilate various stages of the lifecycle of a machine tool with a multi-domain
consolidated model so as to satisfy the need of DTMT. In order to recognize the self-adaptation,
self-prediction and self-evaluation of DTMT, we consider the general comprehensive performance
of the model along with how different parts are coupled. The strategies of design and simulation
of multi-domain machines include a software program interface method.

33
CHAPTER 5

RESULTS AND DISCUSSIONS

Graphical values of feed rate, cutting force, velocity loop and proportional loop are obtained
from the Simulink model.

Fig. 21 Graphical representation of feed rate for the milling operation

Fig. 21 shows that the feed rate for the milling operation remains constant for some time. Then
it increases drastically for some time and after reaching a saturation point it decreases for a very
short time. In the end, the feed rate again becomes constant.

34
 Fig. 22 Graphical representation of cutting force for the milling process

Cutting force also follows the same path as the feed rate for the milling process. ref Fig 22.

35
 Fig. 23 and 24 show that the dynamic stiffness (inverse of compliance) . It increases with the
increase in the Kv, position loop gains as well as the proportional gain of the Kp, velocity
controller. therefore, it is directly proportional to Kv and Kp.

36
Fig. 25 Angle V/S time for Simscape Multibody model of Y-drive

Fig. 26 Rotational velocity V/S time for Simscape Multibody model of Y-drive

37
CHAPTER 6

RESOURCES AND LIMITATIONS

Resources:

ANSYS, MATLAB/ Simulink are examples of commercial software interfaces used in

interdisciplinary fields corresponding to a modelling method called software interface. The
software is quite mature because of the build-up of relevant skilled information, however
developmental difficulties were redoubled due to the compatibility of interfaces between
completely dissimilar computer coding programs.

Limitations:

➢ Technology's progress is based on the connectivity of the internet .

➢ The definition of digital twins is based on 3D CAD models and not on drawings in 2D.
➢ Fine models for simulation may be expensive. Simulation also takes a tremendous amount
of time for computation and is therefore costly.
➢ Simulation provides a means for solutions to be tested but does not itself produce solutions.

38
CONCLUSION
Our paper proposes a method of maintenance where a DT drives a hybrid predictive
maintenance method for a Computer Numerical Control Machine Tool (CNCMT). Our paper uses
an approach to create a hybrid algorithm using software such as MATLAB and SIMULINK for
fault detection and maintenance.

In the given method, a multi-domain DT model is created using physical, electrical, hydraulic
and thermodynamic information to reflect the working conditions in their actual form. After this,
we use the different values and amalgamate them using our algorithm in order to give us results
that are more accurate than existing methods or methods that use a single strategy for prediction.
When compared with the predictions of the traditional model, we found through experimental
outcomes that the tracking errors estimated by the suggested model are more accurate when the
transmission path was not taken into account.

39
BIBLIOGRAPHY

1. Nyhuis, P., & Wiendahl, H. P. (2008). Fundamentals of production logistics:

theory, tools and applications. Springer Science & Business Media.

2. Radharamanan, R. (2002). Virtual manufacturing: an emerging technology. age, 7,

1.

3. Onosato, M., & Iwata, K. (1993). Development of a virtual manufacturing system

by integrating product models and factory models. CIRP annals, 42(1), 475-478.

4. Mwinuka, T. E., & Mgwatu, M. I. (2015). Tool selection for rough and finish CNC
milling operations based on tool-path generation and machining
optimisation. Advances in Production Engineering & Management, 10(1).

5. Debevec, M., Simic, M., & Herakovic, N. (2014). Virtual factory as an advanced
approach for production process optimization. International journal of simulation
modelling, 13(1), 66-78.

6. Cus, F., & Zuperl, U. (2015). Surface roughness control simulation of turning
processes/Simulator sistema vodenja hrapavosti povrsine pri procesu
struzenja. Strojniski vestnik-Journal of Mechanical Engineering, 61(4), 245-255.

7. Hrelja, M., Klancnik, S., Irgolic, T., Paulic, M., Jurkovic, Z., Balic, J., & Brezocnik,
M. (2014). Particle swarm optimization approach for modelling a turning
process. Advances in Production Engineering & Management, 9(1), 21.

8. Todorović, P., Vukelic, D., Tadić, B., Veljkovic, D., Budak, I., Macuzić, I., & Lalic,
B. (2014). Modelling of dynamic compliance of fixture/workpiece
interface. International Journal of Simulation Modelling.

9. Weck, M., Queins, M., & Witt, S. (2003). Effektive Entwicklung von
Werkzeugmaschinen. VDI-Z, 145(10), 32-36.

40
10. Altintas, Y., Brecher, C., Weck, M., & Witt, S. (2005). Virtual machine tool. CIRP
annals, 54(2), 115-138.

11. Weule, H., Fleischer, J., Neithardt, W., Emmrich, D., & Just, D. (2003, June).
Structural Optimization of Machine Tools including the static and dynamic
Workspace Behavior. In The 36th cirp-international seminar on manufacturing
systems (Vol. 12, pp. 56-60).

12. Brecher, C., Esser, M., & Witt, S. (2009). Interaction of manufacturing process and
machine tool. CIRP annals, 58(2), 588-607.

13. Kjellberg, T., von Euler-Chelpin, A., Hedlind, M., Lundgren, M., Sivard, G., &
Chen, D. (2009). The machine tool model—A core part of the digital factory. CIRP
annals, 58(1), 425-428.

14. Li, C. J., Tseng, H. C., Tsai, M. S., & Cheng, C. C. (2020). Novel servo-feed-drive
model considering cutting force and structural effects in milling to predict servo
dynamic behaviors. The International Journal of Advanced Manufacturing
Technology, 106(3), 1441-1451.

15. Wang, J., Yan, J., Li, C., Gao, R. X., & Zhao, R. (2019). Deep heterogeneous GRU
model for predictive analytics in smart manufacturing: application to tool wear
prediction. Computers in Industry, 111, 1-14.

16. Duan, L., Xie, M., Bai, T., & Wang, J. (2016, June). Support vector data description
for machinery multi-fault classification with unbalanced datasets In 2016 IEEE
International Conference on Prognostics and Health Management (ICPHM) (pp.
1-6). IEEE.

17. Hanachi, H., Yu, W., Kim, I. Y., Liu, J., & Mechefske, C. K. (2019). Hybrid data-
driven physics-based model fusion framework for tool wear prediction. The
International Journal of Advanced Manufacturing Technology, 101(9-12), 2861-
2872.

41
18. Liao, L., & Köttig, F. (2016). A hybrid framework combining data-driven and
model-based methods for system remaining useful life prediction. Applied Soft
Computing, 44, 191-199.

19. Mobley, R. K. (2002). An introduction to predictive maintenance. Elsevier.

20. Shimada, J., & Sakajo, S. (2016, July). A statistical approach to reduce failure
facilities based on predictive maintenance. In 2016 International Joint Conference
on Neural Networks (IJCNN) (pp. 5156-5160). IEEE.

21. Li, C., Zhang, Y., & Xu, M. (2012). Reliability-based maintenance optimization
under imperfect predictive maintenance. Chinese Journal of Mechanical
Engineering, 25(1), 160-165.

22. Ortiz, J., & Carrasco, R. A. (2016, July). Model-based fault detection and diagnosis
in ALMA subsystems. In Observatory Operations: Strategies, Processes, and
Systems VI (Vol. 9910, p. 99102S). International Society for Optics and Photonics.

23. Lei, Y., Li, N., Gontarz, S., Lin, J., Radkowski, S., & Dybala, J. (2016). A model-
based method for remaining useful life prediction of machinery. IEEE Transactions
on Reliability, 65(3), 1314-1326.

24. Iung, B., Monnin, M., Voisin, A., Cocheteux, P., & Levrat, E. (2008). Degradation
state model-based prognosis for proactively maintaining product
performance. CIRP annals, 57(1), 49-52.

25. Baptista, M., Sankararaman, S., de Medeiros, I. P., Nascimento Jr, C., Prendinger,
H., & Henriques, E. M. (2018). Forecasting fault events for predictive maintenance
using data-driven techniques and ARMA modeling. Computers & Industrial
Engineering, 115, 41-53.

26. Daily, J., & Peterson, J. (2017). Predictive maintenance: How big data analysis can
improve maintenance. In Supply Chain Integration Challenges in Commercial
Aerospace (pp. 267-278). Springer, Cham.

42
27. Canizo, M., Onieva, E., Conde, A., Charramendieta, S., & Trujillo, S. (2017, June).
Real-time predictive maintenance for wind turbines using Big Data frameworks.
In 2017 IEEE International Conference on Prognostics and Health Management
(ICPHM) (pp. 70-77). IEEE.

28. Wang, J., Fu, P., Zhang, L., Gao, R. X., & Zhao, R. (2019). Multilevel information
fusion for induction motor fault diagnosis. IEEE/ASME Transactions on
Mechatronics, 24(5), 2139-2150.

30. Wang, J., Yan, J., Li, C., Gao, R. X., & Zhao, R. (2019). Deep heterogeneous GRU
model for predictive analytics in smart manufacturing: application to tool wear
prediction. Computers in Industry, 111, 1-14.

31. Wang, J., Zheng, Y., Wang, P., & Gao, R. X. (2017). A virtual sensing based
augmented particle filter for tool condition prognosis. Journal of Manufacturing
Processes, 28, 472-478.

32. J. Wang, J. Xie, R. Zhao, et al., Multisensory fusion based virtual tool wear sensing for
ubiquitous manufacturing, Robot. Comput. Integr. Manuf. 45 (2017) 47–58.

33. Liao, L., & Köttig, F. (2016). A hybrid framework combining data-driven and
model-based methods for system remaining useful life prediction. Applied Soft
Computing, 44, 191-199.

34. Wang, J., Wang, P., & Gao, R. X. (2015). Enhanced particle filter for tool wear
prediction. Journal of Manufacturing Systems, 36, 35-45.

35. Wang, J., Liang, Y., Zheng, Y., Gao, R. X., & Zhang, F. (2020). An integrated fault
diagnosis and prognosis approach for predictive maintenance of wind turbine
bearing with limited samples. Renewable Energy, 145, 642-650.

36. Hochhalter, J., Leser, W. P., Newman, J. A., Gupta, V. K., Yamakov, V., Cornell,
S. R., ... & Heber, G. (2014). Coupling Damage-Sensing Particles to the Digital
Twin Concept.

43
37. Rosen, R., Von Wichert, G., Lo, G., & Bettenhausen, K. D. (2015). About the
importance of autonomy and digital twins for the future of manufacturing. IFAC-
PapersOnLine, 48(3), 567-572.

38. Wang, P., Gao, R. X., & Fan, Z. (2015). Cloud computing for cloud manufacturing:
benefits and limitations. Journal of Manufacturing Science and
Engineering, 137(4).

39. Zuehlke, D. (2010). SmartFactory—Towards a factory-of-things. Annual reviews in

control, 34(1), 129-138.

40. Tao, F., Cheng, Y., Da Xu, L., Zhang, L., & Li, B. H. (2014). CCIoT-CMfg: cloud
computing and internet of things-based cloud manufacturing service system. IEEE
transactions on industrial informatics, 10(2), 1435-1442.

41. Rosen, R., Von Wichert, G., Lo, G., & Bettenhausen, K. D. (2015). About the
importance of autonomy and digital twins for the future of manufacturing. IFAC-
PapersOnLine, 48(3), 567-572.

42. Luo, W., Hu, T., Zhu, W., & Tao, F. (2018, March). Digital twin modeling method
for CNC machine tool. In 2018 IEEE 15th International Conference on Networking,
Sensing and Control (ICNSC) (pp. 1-4). IEEE.

43. Guo, J., Zhao, N., Sun, L., & Zhang, S. (2019). Modular based flexible digital twin
for factory design. Journal of Ambient Intelligence and Humanized
Computing, 10(3), 1189-1200.

44. Zhang, H., Liu, Q., Chen, X., Zhang, D., & Leng, J. (2017). A digital twin-based
approach for designing and decoupling hollow glass production lines. IEEE
Access, 5, 26901-26911.

45. Liu, Q., Zhang, H., Leng, J., & Chen, X. (2019). Digital twin-driven rapid
individualised designing of automated flow-shop manufacturing
systems. International Journal of Production Research, 57(12), 3903-3919.

44
46. Ruiz, E., Osmani, V., Sucar, L. E., & Mayora, O. (2019). Detecting dressing failures
using temporal–relational visual grammars. Journal of Ambient Intelligence and
Humanized Computing, 10(7), 2757-2770.

47. Wang, J., Ma, Y., Zhang, L., Gao, R. X., & Wu, D. (2018). Deep learning for smart
manufacturing: Methods and applications. Journal of Manufacturing Systems, 48,
144-156.

48. Lu, S., Zhou, P., Wang, X., Liu, Y., Liu, F., & Zhao, J. (2018). Condition monitoring
and fault diagnosis of motor bearings using undersampled vibration signals from a
wireless sensor network. Journal of Sound and Vibration, 414, 81-96.

49. Uematsu, T., & Mohri, N. (1986). Prediction and detection of cutting tool failure by
modified group method of data handling. International Journal of Machine Tool
Design and Research, 26(1), 69-80.

50. El Wardany, T. I., & Elbestawi, M. A. (1997). Prediction of tool failure rate in
turning hardened steels. The International Journal of Advanced Manufacturing
Technology, 13(1), 1-16.

51. Elshenawy, L. M., & Mahmoud, T. A. (2018). Fault diagnosis of time-varying

processes using modified reconstruction-based contributions. Journal of Process
Control, 70, 12-23.

52. Kang, J., Kang, N., Feng, C. J., & Hu, H. P. (2007, November). Research on tool
failure prediction and wear monitoring based hmm pattern recognition theory.
In 2007 International Conference on Wavelet Analysis and Pattern
Recognition (Vol. 3, pp. 1167-1172). IEEE.

53. Park, S., Kapoor, S. G., & DeVor, R. E. (2006). Microstructure-level model for the
prediction of tool failure in WC-Co cutting tool materials.

45
54. Sanislav, T., & Miclea, L. (2012). Cyber-physical systems-concept, challenges and
research areas. Journal of Control Engineering and Applied Informatics, 14(2), 28-
33.

55. Kelly, S. D. T., Suryadevara, N. K., & Mukhopadhyay, S. C. (2013). Towards the
implementation of IoT for environmental condition monitoring in homes. IEEE
sensors journal, 13(10), 3846-3853.

56. Pandilov, Z., Milecki, A., Nowak, A., Górski, F., Grajewski, D., Ciglar, D., ... &
Mulc, T. (2015). Virtual modelling and simulation of a CNC machine feed drive
system. Transactions of FAMENA, 39(4), 37-54.

57. Pandilov, Z., & Dukovski, V. (2014). Analytical Determination of the CNC
Machines High-Speed Feed Drives Position Loop Gain. In Applied Mechanics and
Materials (Vol. 555, pp. 505-510). Trans Tech Publications Ltd.

58. Ramesh, R., Mannan, M. A., & Poo, A. N. (2005). Tracking and contour error
control in CNC servo systems. International Journal of Machine Tools and
Manufacture, 45(3), 301-326.

59. Ruijun, L., Wenhua, Y., Zhang, H. H., & Qifan, Y. (2012). The thermal error
optimization models for CNC machine tools. The International Journal of
Advanced Manufacturing Technology, 63(9-12), 1167-1176.

60. Muhammad, B. B., Wan, M., Feng, J., & Zhang, W. H. (2017). Dynamic damping
of machining vibration: a review. The International Journal of Advanced
Manufacturing Technology, 89(9-12), 2935-2952.

61. Altintas, Y., & Ber, A. A. (2001). Manufacturing automation: metal cutting
mechanics, machine tool vibrations, and CNC design. Appl. Mech. Rev., 54(5),
B84-B84.

62. Zhang, G. P., Huang, Y. M., Shi, W. H., & Fu, W. P. (2003). Predicting dynamic
behaviours of a whole machine tool structure based on computer-aided

46
 engineering. International Journal of Machine Tools and Manufacture, 43(7), 699-
706.

63. Law, M., Altintas, Y., & Phani, A. S. (2013). Rapid evaluation and optimization of
machine tools with position-dependent stability. International Journal of Machine
Tools and Manufacture, 68, 81-90.

64. Wu, J., Han, Y., Xiong, Z., & Ding, H. (2017). Servo performance improvement
through iterative tuning feedforward controller with disturbance
compensator. International Journal of Machine Tools and Manufacture, 117, 1-10.

65. Ohishi, K., Miyazaki, T., Inomata, K., Yanagisawa, H., Koide, D., & Tokumaru, H.
(2006). Robust tracking servo system considering force disturbance for the optical
disk recording system. IEEE Transactions on Industrial electronics, 53(3), 838-
847.

66. Erkorkmaz, K., & Altintas, Y. (2001). High speed CNC system design. Part II:
modeling and identification of feed drives. International Journal of Machine Tools
and Manufacture, 41(10), 1487-1509.

67. Chen, Y. Y., Huang, P. Y., & Yen, J. Y. (2002). Frequency-domain identification
algorithms for servo systems with friction. IEEE transactions on control systems
technology, 10(5), 654-665.

68. Olsson, H., Åström, K. J., De Wit, C. C., Gäfvert, M., & Lischinsky, P. (1998).
Friction models and friction compensation. Eur. J. Control, 4(3), 176-195.

69. Do, N. B., Ferri, A. A., & Bauchau, O. A. (2007). Efficient simulation of a dynamic
system with LuGre friction.

70. Li, C. J., Tseng, H. C., Tsai, M. S., & Cheng, C. C. (2020). Novel servo-feed-drive
model considering cutting force and structural effects in milling to predict servo
dynamic behaviors. The International Journal of Advanced Manufacturing
Technology, 106(3), 1441-1451.

47
APPENDIX-1

48