ELEKTROTEHNI ˇ SKI VESTNIK 92(4): 181–190, 2025
ORIGINAL SCIENTIFIC PAPER
A Control Framework for the Machining Center with
Multiple Spindles
Karlo Gripari´ c
1,† , Diego Suˇ sanj
1
, Vilijan Matoˇ sevi´ c
1
1
Juraj Dobrila University of Pula, Faculty of Engineering,
Alda Negrija 6, 52100 Pula
† E-mail: karlo.griparic@unipu.hr
Abstract. The paper addresses the problem of minimizing the tool exchange time in computer numerically
controlled (CNC) machining centres. The increasing variety of products and the demand for smaller quantities
have placed a significant pressure on the CNC systems to enhance the adaptability, improve her flexibility,
and shorten the preparation and operating times. In milling machines, a certain amount of the operating time
is consumed during the tool exchange process. To reduce the exchange time, the paper proposes a solution
that introduces a second spindle positioned physically close to the main spindle, controlled by a pneumatic
system. For this purpose a novel control framework is developed that reduces the operation time, facilitates the
operator handling, and improves the efficiency. A combinatorial optimization approach is then used to solve
the tool scheduling problem, in order to minimize the total exchange time of the used tools. The presented
novel approach to the machine is based on the initial and final positions of each tool and on the magazine
holders positions. The experimental results show that by implementing a control framework to manipulate
the multiple spindles, the reduction in the exchange time is significant and dependent on the number of tools
needed for a certain machining process. Moreover, the tool scheduling algorithm optimizes the tools assignment
at magazine locations and efficiently reduces the total execution time. This can be important for items with a
short operational time and a large number of the used tools.
Keywords: Computer Numeric Control (CNC), Automatic Tool Exchanger (ATC), Multiple Spindles,
Combinatorial optimization, Automata
Okvir za vodenje obdelovalnega centra z veˇ c vreteni
ˇ
Clanek obravnava problem zmanjˇ sanja ˇ casa menjave orodij v
raˇ cunalniˇ sko krmiljenih (CNC) obdelovalnih centrih. Zaradi
vse veˇ cje raznolikosti izdelkov in manjˇ sih serij se na CNC-
sisteme vrˇ si pritisk po veˇ cji prilagodljivosti, krajˇ sem ˇ casu
priprave in uˇ cinkovitejˇ sem delovanju. Pri rezkalnih strojih del
obratovalnega ˇ casa odpade na menjavo orodij, zato ˇ clanek
predlaga reˇ sitev z uvedbo drugega vretena, nameˇ sˇ cenega v
bliˇ zini glavnega in krmiljenega s pnevmatskim sistemom. Za
ta namen je bil razvit nov okvir za vodenje, ki zmanjˇ suje ˇ cas
delovanja, olajˇ sa upravljanje in poveˇ ca uˇ cinkovitost. Za raz-
porejanje orodij je uporabljena kombinatoriˇ cna optimizacija,
ki minimizira skupni ˇ cas menjav glede na zaˇ cetne in konˇ cne
poloˇ zaje orodij ter poloˇ zaje drˇ zal v magazinu. Eksperimen-
talni rezultati kaˇ zejo, da uporaba veˇ c vreten in predlaganega
krmilnega okvira znatno skrajˇ sa ˇ cas menjave, pri ˇ cemer je
prihranek odvisen od ˇ stevila uporabljenih orodij. Algoritem za
razporejanje orodij dodatno optimizira njihovo razporeditev v
magazinu in uˇ cinkovito zmanjˇ suje skupni ˇ cas obdelave, kar je
posebej pomembno pri izdelkih s kratkim obratovalnim ˇ casom
in velikim ˇ stevilom uporabljenih orodij.
Received 17 April 2025
Accepted 31 July 2025
Copyright: © 2025 by the authors.
Creative Commons Attribution 4.0
International License
1 INTRODUCTION
In the last decades, the increasing demands for flexibil-
ity, product diversity, and speed of production in industry
have drawn a tremendous interest in developing high-
performance CNC machine tools. Modern manufactur-
ing processes should be fully digitalized to provide an
efficient production of complex mechanical parts. This
is one of the main goals of the manufacturing systems
developed under the fourth industrial revolution known
as Industry 4.0, comprised of cyber-physical systems
[1], Internet of Things (IoT) [2], and digital twins [3].
The important requirements in the production industry
that uses machine tools are also the repeatability and
accuracy. Therefore, the development of such machine
faces many challenges in the multidisciplinary field
of mechanical, control, and computer engineering [4],
[5]. In the process of designing a CNC machine tool,
achieving an optimal balance between automation and
profitability is crucial for success.
On the other hand, complex parts require several
different operations, such as roughing, drilling, tapping,
milling, etc., where for each operation a different tool
is used. To increase the productivity of a machining

182 GRIPARI
´
C, SU
ˇ
SANJ, MATO
ˇ
SEVI
´
C
process, manufactures have designed an Automatic Tool
Exchange (ATC) system. Its most widespread design
concepts used in practice include tool change systems
with an indirect tool change, system with a tool mag-
azine positioned close to the spindle, machines with
multiple spindles, and other innovations. A modern CNC
machine tool is unthinkable without some type of ATC
that has been proven as a powerful tool in improving the
productivity and accuracy, and decreasing the possibility
of a failure. In [6] and [7], the authors develop an
ATC for two types of milling machines: horizontal and
vertical machine tools. [8] presents pioneer work for
operation scheduling using multiple spindles in CNC
machines. [9] highlights the importance of integrating
hierarchical and collaborative control structures into an
intelligent manufacturing system. [10] presents several
approaches to increase the energy efficiency of the
CNC machining process, including redesigned machines
and controllers, and additional external energy saving
devices. [11] combines two independent tools, i.e., a
laser and a micro-milling tool on a five-axis machine
tool. An extensive work has been done in developing
a hardware solution that incorporates a laser module
in an off-the-shell numeric controller. However, there
has been no algorithm presented to control the change
of the tools, and no coordinate system offset caused
by a tool change has been taken into account. The
developed solution which uses only two different tools
limits the flexibility. [12] describes an open control
architecture built in the Microsoft Visual C for a micro-
milling machine with three axes and a single spindle.
However, control system provides no novelties for the
machine control system compared to the commercial
CNC systems. The control system exhibits no flexibility
and scalability to implement its principles in different
types of machine tools.
[13] proposes an algorithm to control the automatic
tool changer for a vertical machining center. It searches
for a programmed tool located in a turret head using
the shortest path. The solution uses the well known
ladder diagram and combines it with an existing machine
numeric controller. [14] recently introduces a concept
of an invisible numerical control with a distributed
and networked architecture to achieve a high recon-
figurability, scalability and functional reusability. The
limitations of the traditional CNC controllers are defined
and a novel numeric control paradigm improves the
performances of flexible manufacturing systems. [15],
[16], [17] introduce well known concept of Industry 4.0,
a networked system of multiple objects such as humans,
resources and machines to create an intelligent factory.
Though the traditional CNC systems have been used
for decades, manufacturing demands of the modern
industry have raised many new challenges that machine
tools should be able to cope with, including the part
complexity, speed, product diversity, short product life-
time and etc. [18] provides a comprehensive overview of
the limitations of the CNC system. A possible approach
to cope with those limitations includes new standards
[19] to enhance the adaptability and interoperability
of the CNC machine tools. The paper emphasizes the
potential of such technologies in replacing traditional
CNC programming methods for being a driving inno-
vation in manufacturing systems. [20] presents several
solutions to improve the manufacturing productivity
and efficiency by using the advantages of high-speed
industrial communication networks. [21] studies some
similar approaches to minimize the total non-machining
time on machines with cutting tools located in the turret
magazine. In such magazine, the rotation time needed
to turn the turret for a single position is considered
constant. Therefore, the optimization problem comes
down to the problem of minimization of the total turret
rotations. [22] implements an interesting solution to
the allocation tool problem in the turret magazine. The
Dijkstra shortest path algorithm is proposed to minimize
the indexing time of the turret magazine. [23] introduces
the shortest path search on a graph created by a genetic
algorithm to generate a graph of all routing alternatives.
Since the milling machine centers are the most
widespread type of the machine in industry, the paper
proposes a solution to increase the efficiency of the ma-
chine center by adding an additional spindle. A detailed
analysis is made of the time taken to switch the tools
between the machining operations. The aim of the anal-
ysis is to provide a basis for developing a benchmarking
model to evaluate the effectiveness of the proposed
optimization algorithm and control strategy. Our main
objective is to reduce the time taken to exchange the
tool in order to increase the effectiveness. The machine
is therefore mechanically redesigned to provide the basis
for a novel control framework to improve the efficiency
and energy consumption as well as but also the accuracy
machining. A combinatorial optimization problem to
solve the tool scheduling problem is proposed. The
optimization problem takes into account the magazine
holders positions, and the initial and final positions of
each operation.
The rest of the paper is organized as follows. Sec-
tion 2 describes the CNC milling center, explains the
machine elements of the automatic tool change system
and presents the modifications implemented to install
the second spindle. Section 3 presents the CNC control
architecture and the proposed controller based on the
Moore finite-state machine for multiple spindles. Section
4 formulates the tool scheduling in a form of a combi-
natorial optimization problem. Sections 5 and 6 provide
experimental results and draw conclusions of the paper.

A CONTROL FRAMEWORK FOR A MACHINING CENTER WITH MULTIPLE SPINDLES 183
Figure 1.: Machining center Biesse Rover 336 ATC with four axes, two spindles and two tool magazines located
at different sides of the working space.
(a)
Spindle
SA
Spindle
SB
Working table
(b)
Figure 2.: Machine spindles real (a) and graphical representation (b). The left spindle is labeled with SA and the
right with SB.
2 SYSTEM DESCRIPTION
The used machine tool is a CNC milling center Biesse
Rover 336 ATC presented in Fig. 1. It has undergone
mechanical upgrades and electrical retrofitting. It fea-
tures four axes (three linear and one rotational), two
vertically positioned spindles and two tool magazines.
The spindles moves within the Cartesian coordinated
system, where one of the spindles rotates its head around
axis Z. The rotational axis, referred to as axis C, does
not affect the automatic tool exchange system. The axis
displacement is implemented using three linear and one
rotational controlled servomotors, where the maximum
axis stroke is 3200 mm for the X axis, 1250 mm for
the Y axis and 185 mm for the Z axis. The machine is
equipped with two tool magazines, each containing five
tool holders located on different sides of the working
area.
The milling center is designed, but not limited, for
advanced woodworking tasks, therefore the rotational
speed of the used spindles is up to 24 000 rpm with a
maximal power of 6.8 kW. Basically, each spindle is an
independent AC motor on which shaft is a standardized

184 GRIPARI
´
C, SU
ˇ
SANJ, MATO
ˇ
SEVI
´
C
ISO 30 conical slot for the tool holder. An automatic tool
change is enabled through a pneumatic system located
in the motor housing and controlled by electromag-
netic valves. The rotational speed is adjustable using
a frequency drive, where the driver is connected in the
manner that it can be switched between two spindles.
This means that only one spindle can be in operation
at a time, thus significantly reducing the overall cost of
the system because only one driver is used to power
both spindles. As shown in Fig. 2a, the machine head
contains two spindles labeled SA and SB, starting from
the spindle on the left. Spindle SB is equipped with
the rotational C axis, and is used for operation where
tool orientation is important, like a circular saw blade
or milling cutter whose rotational axis is not collinear
with the Z axis. As seen, the spindles are located at a
certain distance from each other, causing an offset in the
machine coordinate system. The offset varies according
to the active spindle. The offset of the coordinate sys-
tem and the procedure for switching the single driver
between the two spindles are integrated into the control
algorithm presented in Section 3.
As mentioned above, the machine uses a pneumatic
mechanical solution to facilitate a quick and efficient
tool exchange. Each spindle is mounted on a mechanical
support that can be translated along the Z axis for
a certain distance by using a pneumatically-controlled
cylinder. The inactive spindle is positioned farther from
the working table, eliminating the need to unload and
store its tool in the magazine. The advantage of the
solution is that one spindle can work while the other is
in a position that does not interfere with the machining
operation. Hence, two combinations are possible: the
first is when the SA spindle is in operation and the
SB spindle is deactivated. The second is a situation
when the SB spindle is in operation. When the SA
spindle is active, as depicted in Fig. 2b, the pneumatic
support is repositioned to the lowest possible position
referred from the working table, while the SB spindle
is inactive and is moved to the upper position. When
a command from the machine controller is received
the active spindle stops its rotation if running. Then
one pneumatic cylinder is activated and the other one
deactivated, causing the first spindle to be translated to
the working position and the other one to be moved to
a standby position. The pneumatic support is equipped
with electromagnetic sensors confirming the spindles
have reached their requested positions. If the support
is not adequately positioned the system issues an alarm
to prevent a collision of the tool loaded in the nonactive
spindle with the workpiece in the working area.
The tool magazines are located inside the machine
body on each side of the working plane. The magazine
is closed with a pneumatically controlled cover that
prevents material chips and residuals from entering the
Figure 3.: CNC control architecture comprising a user
interface, G-code interpreter and numerical and machine
controllers. The arrows show the coordination between
the components.
magazine. It is equipped with five tool slots located in a
fixed place inside the working plane and mounted on a
pneumatic cylinder that comes out during the exchange
phase. The procedure for changing a tool depends on
the requested action. It can either be returning or taking
the tool to/from the magazine. The following steps are
taken to remove the tool from the spindle and to load a
one tool into it:
1) Opening the tool magazine cover,
2) Moving the spindle at the slot location where the
active tool should be returned,
3) Rising up the relevant slot using the pneumatic
cylinder,
4) Moving the spindle in the position above the tool
holder enters the slot,
5) Unloading the tool and setting down the slot,
6) Moving the spindle at the slot location where the
new tool is located,
7) Rising up the relevant slot using the pneumatic
cylinder,
8) Moving the spindle in the position when the tool
holder enters the spindle and
9) Loading the tool and set down the slot.
3 THE PROPOSED CONTROLLER
The CNC controller architecture (see Fig. 3 is hierar-
chically divided into two levels: the numerical and the
machine controller. The main task of the CNC system,
embodied in the numerical controller, is to guarantee the
movement of the axes of the machine within its working
space while taking into account the acceleration, ve-
locity, and positional accuracy of each axis. The aim

A CONTROL FRAMEWORK FOR A MACHINING CENTER WITH MULTIPLE SPINDLES 185
is to synchronize the movement of the axes according
to the static and dynamic requirements. Typically, the
movement of the axis is implemented inside a cascade
closed control loop. The commonly used components
of the machine tools for positioning include an AC
servomotor, ball screws, and an encoder. An effective
control of various machine tool components, including
hydraulic systems, pneumatic cylinders, tool magazines,
safety barriers, etc., is essential for an overall control
of the machine, and is typically implemented in the
machine controller. This is mainly achieved with a
programmable logic controller (PLC) adapted for the use
in machine tool applications. It continuously monitors
the input signals (in most cases the digital input signals),
executes decisions based on a predefined logic, and
sets the state of the digital output signals. The numeric
and machine controllers are both equipped with internal
communication interfaces that facilitate their synchro-
nization throughout the machining process. Finally, the
entire system is supervised via a user interface and G-
code interpreter, which is a part of the highest level of
the CNC controller architecture.
The G-code is a widely used standardized program-
ming language for machine tools [24]. Besides the axis
movements, it supports a vast range of other commands
needed in the machining process. The tool exchange is
one of the auxiliary actions that are typically initiated
by a command in the G-code, i.e., when the tool
identification number and the exchange request (M6)
are issued. The machine controller should then executes
the necessary steps to replace the old tool with a new
one. During the exchange process, the numeric controller
waits with the program execution until it receives a
confirmation from the machine controller that the new
tool is ready for operation.
A concise overview of the control framework to
handle ATC for the machining center with multiple
spindles and tool magazines is given below. Based
on the requirements for a single spindle, the steps
needed to control the tool exchange process between
the spindle and the magazine are identified. To design
an effective spindle controller, we propose a finite-state
machine (FSM) for each spindle. The approach provides
a scalable and reusable control method, allowing an easy
expansion by adding numerous spindles. A type of FSM
is the Moore state machine, in which the output depends
on the current state. The digital input signals, which can
be periodically read by the machine controller, affect the
transitions between the states. In other words, the next
state of the Moore state machine is determined by the
current state and input signals. The Moore FSM can be
defined by:
M = (I,O,S,ρ,λ,s
0
), (1)
where I is the finite set of inputs, O is the finite set of
outputs, S is the finite set of the states, ρ :S×I→S
is the state transition function, λ :S→O is the output
function, and s
0
is the initial state of the machine. A
comprehensive FSM overview is given in [25].
The Moore FSM to control the tool exchange task
of a single spindle in the machining center is designed
by identifying every state that can occur in a change
maneuver. The control framework is designed based on
the following requirements:
• The spindle can be either in an active or inactive
state. In the active state, the spindle performs the
machining and vice versa.
• The spindle can be in an inactive state without a
loaded tool.
• The exchange process does not depend on the
tool location inside the magazine. Thus, when a
machine tool has multiple magazines, the spindle
exchanges the tools with each magazine.
• The manual operation mode is allowed when the
operator can sets the tool manually.
• At an unexpected behavior, an error is issued.
State Description
S
1
The spindle is empty
S
2
The tool exchange is in progress
S
3
The spindle is loaded and active
S
4
The spindle is loaded and inactive
S
5
The spindle error
Table 1.: The description of the states of the Moore
finite-state machine.
Figure 4.: State transition diagram of a spindle where
states S
i
are represented with circles and transitions I
i
are indicated with arrows.
A state table for the spindle tool exchange control
framework to list the possible states is developed. Table
1 identifies the states that satisfy the requirements. The
corresponding state transition diagram visualizing the
states and transitions is shown in Fig. 4. The states
are represented with a circle, while transitions between
states, associated with input signals I
i
, are indicated
with arrows. One of the benefits of the proposed FSM

186 GRIPARI
´
C, SU
ˇ
SANJ, MATO
ˇ
SEVI
´
C
Figure 5.: Complete control architecture of the ATC for a machining center with multiple spindles.
is its simplicity to determine the input signals neces-
sary for the state transitions. The implemented input
signals are read from numerous subsystems, such as
the numerical controller, digital sensors, error flags,
and user commands (see Table 2). The signal of the
numerical controller represents the commands written
in the G-code executed during the machining program.
The digital signals are generated from inductive sensors
connected to the general purpose digital inputs. The
error signals monitor any forbidden condition in case
of its appearance, along with eventual time delays, the
machine controller issues an alarm. If the alarm occurs
during the execution of the user program, the process is
stopped immediately.
Input Description
I
1
Tool exchange request (M6 command)
I
2
Tool exchange completed
I
3
Spindle inactive command
I
4
Spindle active command
I
5
Tool exchange request (M6 command)
I
6
Spindle empty command
I
6
... I
10
Error signal
I
11
Reset error command
Table 2.: The description of the input signals.
In the final step, designing the control system to
manage two spindles and two magazines is described.
The NC issues a tool exchange request which is then
handled by the machine controller. The used tools are
scheduled in the magazine holders during the planning
and programming stage of certain machining operations.
The task of the machine controller is to coordinate
the execution between two finite-state machines, handle
the necessary steps for thr tool replacement, and com-
municate with the upper control level. Once the tool
change is completed, the machine controller sends an
acknowledgment to NC, confirming that the operation
has been successfully executed. Fig. 5 presents a com-
plete ATC control architecture for a machining center
with multiple spindles. The machine controller manages
the tool exchange process using two identical FSMs, one
controls the SA spindle and the other the SB spindle.
4 THE TOOL SCHEDULING PROBLEM
4.1 Problem statement
In general, the problem of scheduling a set of N
machining tools T ∈{T
1
,T
2
,...,T
N
} of a CNC ma-
chine is assumed with single and multiple spindles
and the M ∈ R tool magazines placed at different
locations within machine working space W . In case of
multiple spindles, they assumed to be located on the
same head, i.e. they move simultaneously along with
the axis movement. S is the number of spindles and m
denote the number of tool magazines, where s> 1 and
m > 1. In the machining center, each tool holder is
located at different positions inside machine workspace
W . Set P = [p
1
,...,p
j
,...,p
P
] defines the location of
tool holder j within workspace W where p
j
= (x
j
,y
j
)
are the coordinates of tool holderj from the basic origin
of machine O. The initial working position of tool T
i
is denoted by S
i
(x
s
i
,y
s
i
). The final position is given by
F
i
(x
f
i
,y
f
i
). Hence, the distance between tool holder p
j
and the initial or final position of tool T
i
is defined as
d
s
ij
and d
f
ij
, respectively. A schematic overview of the
machine tool scheduling problem is depicted in Fig. 6.
Initial S
i
and final F
i
point of tool i are assumed to be
available prior to the program execution. The distance
between each tool position and holder is then defined
by the following expression:
d
s
ij
=
q
(x
s
i
−x
j
)
2
+ (y
s
i
−y
j
)
2
,
d
f
ij
=
q
(x
f
i
−x
j
)
2
+ (y
f
i
−y
j
)
2
,
(2)

A CONTROL FRAMEWORK FOR A MACHINING CENTER WITH MULTIPLE SPINDLES 187
p
1
p
2
p
4
p
3
p
5
p
6
p
7
p
9
p
8
Magazine 1 Magazine 2
S
i
(x
i
s
 , y
i
s
 )
F
i
(x
i
s
 , y
i
s
 )
d
ij
s
d
ij
f
Machine working space
p
10
Figure 6.: Schematic overview of the machine tools scheduling problem.
where i∈ [1, 2,...,N] and j∈ [1, 2,...,P ].
Our goal is to design an optimization algorithm
to improve the tool scheduling approach of the CNC
machining centers. When the tool scheduling strategy
is not applied, the tools are usually assigned by the
order of the appearance in the machining operation plan.
The presented optimization algorithm eliminates a prior
allocation of the tools inside the magazine holders by en-
suring that the total exchange time is minimal. Moreover,
the algorithm enhances the scalability by taking into
account the optimization problem of both the exchange
time between the spindle and the magazine position, as
well as the initial and final working position of each
tool. It should be noted that the type of the machine is
important when applying an adequate tool assignment
optimization strategy. For instance, if a machine has
a turret type magazine located on the machine head
and it moves along with the spindle, the start and the
finish positions of each operation do not affect the tool
exchange time.
In the machining operation strategy, the operation
sequence and the tools paths are written in the G code
and executed online by the CNC system. Therefore, a
new tool is set when the numeric controller executes
an adequate command. After a tool request command
is executed, the active tool defined by the machine
controller physically sets inside the spindle and the
spindle support is moved to an appropriate position,
i.e. a position in which machining is possible. It is
assumed that a set of the tools needed to execute a
certain machining task, is available in one of the M
magazines whose tool parameters and each tool length
and diameter are stored in the controller memory prior
to a program execution.
Generally a tool exchange operation using ATC on a
CNC milling machine is performed in three steps: 1)
removing the used tool from the spindle and putting
it in an adequate magazine position, 2) positioning the
magazine to receive a new cutting tool, and 3) inserting
the new cutting tool into the spindle. The total tool
change time can generally be defined as the duration
of the three steps.
4.2 Optimization
The aim of our optimization problem is to minimize
the time taken during the tool exchange process on
machines that require multiple operations using different
tools. We seek the optimal value from a large number
of finite data. Our selection is based on the objective
function that evaluates each candidate’s solution. In the
context of the decision problems, the results are binary,
with only two possible solutions: toolT
i
is not assigned
to holder H
j
, or tool T
i
is assigned to holder H
j
.
The combinatorial optimization problems, such as the
knapsack problem [26], the traveling salesman problem,
the job scheduling [27], and the routing problems, can
be effectively addressed using the linear assignment
problem [28].
The overall machining time for a specific part includes
both the time taken to complete each operation and the
time taken for tool exchanges between these operations.
In the initial experiment, our goal is to optimize the
arrangement of the tools so as to reduce the total time
taken for all tool exchanges, considering the distance
between each holder and the initial and final points for
each operation. Here, decision variablev
ij
indicates the
assignment of tool T
i
to holder H
j
. The problem is

188 GRIPARI
´
C, SU
ˇ
SANJ, MATO
ˇ
SEVI
´
C
Number of
exchanges 2 3 4 5 6 7 8 9 10
T
1
[s] 30 60 90 120 150 180 210 240 270
T
2
[s] 4 38 52 86 100 134 148 182 196
∆t [s] 26 22 38 34 50 46 62 58 74
∆t [%] -86.6 -36.6 -42.2 -28.3 -33.3 -25.5 -29.5 -24.2 -27.4
Table 3.: The time taken for an exchange in two scenarios. T
1
is the total time taken to conduct an exchange of n
tools in the first scenario where a single spindle is used. T
2
is the total exchange time when using two spindles.
modeled as a combinatorial optimization problem:
min
N
X
i=1
P
X
j=1
c
ij
v
ij
s.t.
N
X
i=1
v
ij
= 1, ∀j∈P,
P
X
j=1
v
ij
= 1, ∀i∈N,
v
ij
∈{0, 1}.
(3)
Cost coefficient c
ij
∈R for tool i and holder j is
determined as a sum of the distances between the holder
and the initial point (S
i
) of operationd
s
ij
and the distance
from the final point (F
i
) to the holder of tool d
f
ij
.
c
ij
=d
s
ij
+d
f
ij
. (4)
The investigated optimization problem belongs to the
class of integer optimization problems that are NP-
complete, while the algorithms used to solve them have
no polynomial convergence time [29]. In some cases,
these problems can be resolved as a linear sum assign-
ment problem if the linear relaxation yields a solution.
The optimization algorithm should determine one tool
holder for each tool only once, thus minimizing the sum
of the selected elements.
5 EXPERIMENTAL VERIFICATION
To evaluate the benefits of introducing the proposed
modification and the novel control strategy for the
described machining center with two spindles and two
magazines, several experiments on the base model of the
machine are performed. Our benchmarking experiment
is made using a single spindle. Generally speaking, the
total time taken to exchange tools the between spindle
and the magazine in a single working cycle is related to
the number of the tools used to complete the machining
operation plan. The experimental results given in Table 3
show the total execution times obtained by varying the
number of the tools from two to ten in two possible
cases. In the first scenario, the total exchange time
involves using n tools with a single spindle. In the sec-
ond scenario, the time is determined when two closely
mounted spindles are used. A graphical representation
is given in Fig. 7.
1 2 3 4 5 6 7 8 9 10
Number of tools
0
50
100
150
200
250
300
t [s]
1 spindle
2 spindles
Figure 7.: Graphical comparison of the total exchange
time in conducted experiments with one and two spin-
dles.
An evaluation program is used for the numeric con-
troller. The elapsed time taken is sampled from the
machine controller. Given that the single spindle setup
is the most common configuration for milling machines,
the time difference, denoted as ∆t, is calculated, in
the duration of the tool exchange process between a
single and a double spindle machine configuration. The
difference time is ∆t = T
1
−T
2
. The reduction in the
run time for a double spindle machine configuration is
also calculated in the percentages denoted as ∆t[%]. The
results show that when two tools are utilized, the time is
26 seconds shorter, representing an 86.6% reduction. In
the scenario using ten tools the time taken for the tool
exchange is 74 seconds shorter, representing a 27.4%
decrease. In our experiments, the initial position of the
tool before the exchange is assumed at the center of the
machine working area.
5.1 Tool scheduling strategy
To validate the proposed tool scheduling strategy,
solved by using a combinatorial optimization algorithm
implemented in Python using the SciPy library, exper-
iments are conducted to evaluate the performance of
the tool allocation in magazine slots. In the machines

A CONTROL FRAMEWORK FOR A MACHINING CENTER WITH MULTIPLE SPINDLES 189
Tool ID 1 2 3 4 5 6 7 8 9 10
x
s
[mm] 2718 1079 2363 379 1768 2482 1983 1686 2092 2363
y
s
[mm] 62 612 139 505 1196 441 211 564 236 782
x
f
[mm] 468 907 557 2421 2398 2563 2255 2148 381 2856
y
f
[mm] 1142 684 1214 177 1200 1168 490 39 1030 43
Table 4.: The start and finish positions in millimeters for each tool.
Slot no. 1 2 3 4 5 6 7 8 9 10
Ex. 1 T2 T1 T4 T3 T9 T5 T10 T6 T8 T7
Ex. 2 T2 T4 T3 0 0 T5 0 0 T1 0
Table 5.: Experimental results of the tool scheduling.
Experiment 1 Experiment 2
10 tools 5 tools
T
B
[s] 419.72 216.82
T
O
[s] 405.60 207.71
∆t [s] 14.12 9.11
∆t [%] 3.36 4.20
Table 6.: The total exchange time for the tools ordered
in an ascending order (T
B
) and scheduled by the opti-
mization algorithm (T
O
).
with a large working space, the travel time from the
tool slot in the magazine can be divided into two
groups. The travel time from the magazine slot to the
first point (starting point) in the working area, where
the tool begins its operation at a programmed working
velocity, to the finish point where it can again restart the
movement at a maximal velocity. In our experiments the
starting (x
s
,y
s
) and end points (x
f
,y
f
) of each tool are
determined randomly (see Table 4). The total exchange
time (T
B
) is measured for the scenario in which the tools
arranged in holders are in an ascending order, unlike in
the case where the tools are assigned to holders on the
basis of their starting and finishing positions within the
working area and holder location. The tool allocation is
determined out prior to the execution of the machining
program. The maximum feed rate for a rapid movement
is set to 15 m/min.
The objective of our experiments is to schedule
the machining tools inside the magazines in order to
minimize the total exchange time. Since the sequence
of the tools is defined by the machining technology,
their order of appearance can not be changed. Two
machining experiments of a differed number of the
cutting tools are performed. The first experiment (Ex.1)
evaluates the performance of the optimization algorithm
when all magazine slots are used, specifically the tools.
The second experiment (Ex.2) utilizes five tools. The
obtained results are given in Table 5, where tool Tx is
placed in a slot numbered from 1 to 10.
Table 6 shows the total exchange time for the tools or-
dered in an ascending order (T
B
), the exchange time for
the tools scheduled by the optimization algorithm (T
O
),
the difference between the duration of the exchange
without and with applying the scheduling approach
where ∆t =T
B
−T
O
, and the difference expressed in
percentages. Compared to the benchmarking experiment,
our approach, reduces the tool exchange time by 3.36%
in case of ten tools, and by 4.20% in case of five tools.
6 CONCLUSION
The paper presents results of an attempt to improve
the machining center by mechanically upgrading it with
an additional spindle. It proposes a control framework
embodied in the machine computer numeric controller
based on the Moore finite-state machine. The adaptabil-
ity of the approach enables an effective handling of
a second tool placed on the machine head. As to the
programming perspective in the G-code, the second spin-
dle is accessed in the same way as the master spindle,
thus, avoiding any additional commands in the program
and being one of the main advantages of the proposed
approach to control multiple spindles by using a finite-
state machine. The experimental verification performed
on a large milling machine showns that the proposed
method significantly reduces the tool exchange time.
A combinatorial optimization algorithm to schedule the
tools inside the magazine slots is also proposed. The
approach minimizes the total travel time between the
magazine slots and tool positions in the working area.
The experimental results are obtained for the machine
with two specially distributed magazines.
While the results presented in the paper are obtained
for a machining center in which ATC combines multiple
spindles and multiple tool magazines, the methodology
used is not limited to above application. It can be used
to control switching between tools in cases where an
additional task is required or where an external han-
dling system, like a robot, is introduced. Reducing the
exchange time is important in situations when machining
an item does not take much time but requires a large
number of tools to complete the operation plan. The

190 GRIPARI
´
C, SU
ˇ
SANJ, MATO
ˇ
SEVI
´
C
proposed methodology thus significantly improves both
the productivity and flexibility of a machine tool.
REFERENCES
[1] Kunpeng Zhu and Yu Zhang. A cyber-physical production
system framework of smart cnc machining monitoring system.
IEEE/ASME Transactions on Mechatronics, 23(6):2579–2586,
December 2018.
[2] Mirco Piron, Junzhang Wu, Andrea Fedele, and Alessandro
Manzardo. Industry 4.0 and life cycle assessment: Evaluation
of the technology applications as an asset for the life cycle
inventory. Science of The Total Environment, 916, March 2024.
[3] Daniel Klic, Martin Minarcik, Ales Polzer, Jiri Tuma, Jiri Marek,
and Michal Holub. Digital twin for cnc machine tools design. In
2024 21st International Conference on Mechatronics - Mecha-
tronika (ME). IEEE, December 2024.
[4] Suk-Hwan Suh, Seong Kyoon Kang, Dae-Hyuk Chung, and Ian
Stroud. Theory and Design of CNC Systems. Springer London,
2008.
[5] Alessandro Fortunato and Alessandro Ascari. The virtual design
of machining centers for hsm: Towards new integrated tools.
Mechatronics, 23(3), April 2013.
[6] Selcuk Avci and M.Selim Akturk. Tool magazine arrangement
and operations sequencing on cnc machines. Computers &
Operations Research, 23(11), November 1996.
[7] J. P. Rogelio and R. G. Baldovino. Development of an automatic
tool changer (atc) system for the 3-axis computer numerically-
controlled (cnc) router machine. In 2014 International Con-
ference on Humanoid, Nanotechnology, Information Technology,
Communication and Control, Environment and Management
(HNICEM), pages 1–5, 2014.
[8] B. A. Norman and J. C. Bean. Multiple spindle cnc machines.
In Proceedings of 1997 IEEE International Conference on Evo-
lutionary Computation (ICEC ’97), pages 425–429, 1997.
[9] Ming Lei, Xiaohong Yang, Mitchell M. Tseng, and Shuzi Yang.
Design an intelligent machine center—strategy and practice.
Mechatronics, 8(3):271–285, April 1998.
[10] S.T. Newman, A. Nassehi, R. Imani-Asrai, and V . Dhokia.
Energy efficient process planning for CNC machining. CIRP
Journal of Manufacturing Science and Technology, 5(2):127–
136, January 2012.
[11] Zhenyuan Jia, Yuqin Chen, Wei Liu, Kai Zhao, Ling Wang,
and Xianming Tu. The control system for precise five-axis
laser/micro-milling machine tool. International Journal of In-
dustrial and Systems Engineering, 22(2):145, 2016.
[12] Xinxin Wang, Xiaohong Lu, Zhenyuan Jia, Wentao Wang, and
Fengchen Wang. Control system for micro milling machine.
International Journal of Industrial and Systems Engineering,
21(1):83, 2015.
[13] M. Mikhov and M. Zhilevski. Control device for tool magazine
drives of vertical machining centers with cnc. In 2020 Interna-
tional Congress on Human-Computer Interaction, Optimization
and Robotic Applications (HORA), pages 1–5, 2020.
[14] Yingxue Yao, Miao Liu, Jianjun Du, and Liang Zhou. Design
of a machine tool control system for function reconfiguration
and reuse in network environment. Robotics and Computer-
Integrated Manufacturing, 56:117–126, April 2019.
[15] L. Monostori, B. K´ ad´ ar, T. Bauernhansl, S. Kondoh, S. Kumara,
G. Reinhart, O. Sauer, G. Schuh, W. Sihn, and K. Ueda. Cyber-
physical systems in manufacturing. CIRP Annals, 65(2):621–
641, 2016.
[16] W. Chang and S. Wu. Investigated information data of cnc
machine tool for established productivity of industry 4.0. In
2016 5th IIAI International Congress on Advanced Applied
Informatics (IIAI-AAI), pages 1088–1092, 2016.
[17] B. Kada, A. Alzubairi, and A. Tameem. Industrial communica-
tion networks and the future of industrial automation. In 2019
Industrial Systems Engineering Conference (ISEC), pages 1–5,
2019.
[18] X.W. Xu and S.T. Newman. Making CNC machine tools more
open, interoperable and intelligent—a review of the technologies.
Computers in Industry, 57(2):141–152, February 2006.
[19] X.W. Xu, Lihui Wang, and Yiming Rong. Step-nc and function
blocks for interoperable manufacturing. IEEE Transactions on
Automation Science and Engineering, 3(3):297–308, July 2006.
[20] Dong Yu, Yi Hu, X.W. Xu, Yan Huang, and Shaohua Du.
An open CNC system based on component technology. IEEE
Transactions on Automation Science and Engineering, 6(2):302–
310, April 2009.
[21] Kaveh Amouzgar, Amir Nourmohammadi, and Amos H.C.
Ng. Multi-objective optimisation of tool indexing problem:
a mathematical model and a modified genetic algorithm. In-
ternational Journal of Production Research, 59(12):3572–3590,
March 2021.
[22] Adil Baykasoglu and Fehmi Burcin Ozsoydan. An improved
approach for determination of index positions on cnc magazines
with cutting tool duplications by integrating shortest path algo-
rithm. International Journal of Production Research, 54(3):742–
760, June 2015.
[23] T¨ urkay Dereli and A. H¨ useyin Filiz. Allocating optimal index
positions on tool magazines using genetic algorithms. Robotics
and Autonomous Systems, 33(2-3):155–167, November 2000.
[24] Alan Overby. CNC Machining Handbook: Building, Program-
ming, and Implementation. McGraw-Hill, Inc., USA, 1 edition,
2010.
[25] Arthur Gill. Introduction to theory of finite-state machines.
McGraw-Hill, New York, NY , December 1962.
[26] N. VanHop. The tool-switching problem with magazine capacity
and tool size constraints. IEEE Transactions on Systems, Man,
and Cybernetics - Part A: Systems and Humans, 35(5):617–628,
September 2005.
[27] Min Dai, Dunbing Tang, Adriana Giret, and Miguel A.
Salido. Multi-objective optimization for energy-efficient flexible
job shop scheduling problem with transportation constraints.
Robotics and Computer-Integrated Manufacturing, 59:143–157,
October 2019.
[28] Majid Soolaki and Naeme Zarrinpoor. A new 0-1 linear program-
ming approach and genetic algorithm for solving assignment
problem in flexible manufacturing system. The International
Journal of Advanced Manufacturing Technology, 75(1-4):385–
394, July 2014.
[29] Egon Balas. Integer programming. Encyclopedia of optimization,
pages 1617–1624, 2009.
Karlo Gripari´ c received his Ph.D. degree in electrical engineering
from the University of Zagreb, Faculty of Electrical Engineering and
Computing in 2018. Currently he is an Assistant Professor and Head
of Department for Computing at the Juraj Dobrila University of Pula,
Faculty of Engineering. His research interests include decentralized
control, multi-agent systems, robotics and industrial automation.
Diego Suˇ sanj is an Associate Professor and Head of the Laboratory
for Intelligent Systems and Computer Perception at the Juraj Dobrila
University of Pula, Faculty of Engineering. He received his PhD in
Computer Science in 2021 from the University of Rijeka, Faculty
of Engineering. His research focuses on computer vision, machine
learning, image processing, and embedded systems.
Vilijan Matoˇ sevi´ c is an electrical engineer specializing in the de-
sign and development of electric machines. He received his PhD in
Electrical Engineering from the University of Zagreb, Croatia. His
research interests include electric machine systems, electromagnetic
and multiphysics modeling, and automatic control.