Computers in Industry 50 (2003) 353366

Embedded fuzzy-control system for machining processes

Results of a case study
R.E. Habera,b,*, J.R. Aliquea, A. Aliquea, J. Hernandezc,
R. Uribe-Etxebarriac
b

a
Instituto de Automatica Industrial (CSIC), N-III km 22,800, La Poveda 28500, Madrid, Spain
Escuela Tecnica Superior de Informatica, Ciudad Universitaria de Cantoblanco Ctra, km 15, 28049 de Colmenar Viejo, Spain
c
Centro Tecnologico IDEKO, Arriaga kalea 2, Apartado de correos 80, E-20870 Elgoibar, Gipuzkoa, Spain

Received 8 September 2002; accepted 1 February 2003

Abstract
In this paper a fuzzy-control system has been designed, implemented and embedded in an open CNC. The integration process,
design steps and results of applying an embedded fuzzy-control system are shown through the example of real machining
operations. The controller uses internal CNC signals (i.e. spindle-motor current) that are gathered and mathematically processed
by means of an integrated application. The results show that, at least in rough milling operations, internal CNC signals can
double as an intelligent, sensorless control system. Actual industrial tests show a higher machining efficiency (i.e. in-process
time is reduced by 10% and total estimated savings the system would provide are about 78%).
# 2003 Elsevier Science B.V. All rights reserved.
Keywords: Fuzzy control; Nonlinear systems; Machining processes

1. Introduction
Todays manufacturing industry is characterized by
an increase in the demand for just-in-time production
and global manufacturing. Manufacturing has more
stringent productivity and profitability requirements
that can be satisfied only if production systems are
highly automated and extremely flexible. One of the
main activities the manufacturing industry has to deal
with is machining, a process that includes operations
that range from rough milling to finishing. There are a
*
Corresponding author. Tel.: 34-91-871-19-00;
fax: 34-91-871-70-50.
E-mail addresses: rhaber@iai.csic.es (R.E. Haber),
jralique@iai.csic.es (J.R. Alique), a.alique@iai.csic.es (A. Alique),
jhernandez@ideko.es (J. Hernandez), ruribe.etxebarria@ideko.es
(R. Uribe-Etxebarria).

number of angles from which to view the optimization

of the machining process, angles where minimum
production cost, maximum productivity and maximum profit are significant factors [1].
There are also various different ways of implementing machization. The implementation on which we
will focus here attains optimal goals via automatic
control of the machining process. The spectrum of
conventional methods (so named to distinguish
them from intelligent methods) available for designing
control systems is enormous. In the incessant pursuit
of better performance, newer approaches have been
also tested such as model reference adaptive control
(MRAC) [2], which incorporates an on-line estimation
scheme to tune controller parameters for time-varying
process dynamics. Another recently applied method
is robust control based on the quantitative feedback

0166-3615/03/$ see front matter # 2003 Elsevier Science B.V. All rights reserved.
doi:10.1016/S0166-3615(03)00022-8

354

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

theory (QFT) [3]. The results of these tests, however,

have not lived up to expectations, because all these
approaches have the indispensable design requisite of
an accurate (traditional) process model, e.g. differential equations, transfer functions and state equations.
Unfortunately, accurate models of this sort cannot yet
be attained for the machining process.
The complexity and uncertainty of processes like
the machining process are what make the realm known
as intelligent systems technology a feasible option to
classical control strategies. Indeed, artificial-intelligence techniques have roused considerable interest in
the scientific community and have been applied to
machining (e.g. [4,5]). Two interesting AI approaches
are the neural network and expert rule based on
adaptive-control constraints (ACC) [6] and evolutionary algorithms based on adaptive-control optimization
(ACO) [7].
However, the main disadvantage in the above-mentioned approaches is that neural-network and evolutionary algorithm-based computation requires time
and therefore limits the performance of the intelligent
control system. Nowadays, open CNC is powerful
enough to be built into an intelligent CNC. However,
there are constraints in terms of real-time signal
processing and the implementation of complex control
algorithms.
On the other hand, the majority of the work in
machining optimization is devoted to the issue of
adaptive techniques. Adaptive controllers are highly
expensive considering the time requirements, because
they estimate parameters on-line and adjust controller
gains accordingly. Also, adaptive-controller systems
must be carefully tuned, and they exhibit complex and
sometimes undesirable behavior.
In order to improve machining efficiency, the current
study focuses on the design and implementation of an
intelligent controller in an open CNC system. Fuzzy
logic (FL) is selected from all the available techniques
because it has proven useful in control and industrial
engineering as a highly practical optimizing tool. To
the best of our knowledge, the main advantage of the
present approach is that it includes: (i) an embedded
fuzzy controller in an open CNC to deal with a real-life
industrial process; (ii) a simple computational procedure to fulfill the time requirements; and (iii) no
restrictions in terms of sensor cost (it is a sensorless
application), wiring or synchronization with the CNC.

The results of the fuzzy-control strategy in actual

industrial tests show higher machining efficiency.
This paper is organized as follows. In Section 2 we
present a brief study of the machining process,
explaining why it is considered a complex process
and setting up the milling process as a case study. In
Section 3 we design the fuzzy controller to optimize
the milling process. Next, in Section 4 we describe
how the fuzzy controller can be embedded in open
CNCs, and we discuss the key design and programming stages. In Section 5 we share the experimental
results and explore some comparative studies. Finally
we give some concluding remarks.

2. The machining process

This section introduces the milling process, which
is characterized by both nonlinear characteristics and
varying process parameters. The task of controlling
this complex process is performed with the help of a
fuzzy controller that will be introduced in Section 3.
A panoramic view of a typical machining center is
given in Fig. 1.
The machining process, also known as the metalremoval process, is widely used in manufacturing. It
consists of four basic types of operations: turning,
drilling, milling, and grinding, performed by different
machine tools. One of the most complex of the four
operations is the milling process [8].
One important study of cutting-process monitoring
and control by means of current measurement shows

Fig. 1. Full view of a typical machining center.

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

the feasibility of motor-current measurement for adaptive control [9]. The direct relationship between the
current consumed and the cutting force is fair enough
for a real industrial implementation of the control
system to be made on the basis of a main spindles
drive current (e.g. [10]).
Nowadays the development of open control systems
in the NCK offers more facilities for using digital drive
signals without the need to install additional sensors.
Indeed, cutting processes show significant effects on
drive signals such as actual drive current and drive
torque. Digital drive signals have many limitations
for process monitoring alone because of the ratio
between process-related components of the signal
and non-process-related disturbances. However, main
spindle-drive current can be used to optimize cutting
speed. Signal behavior is complex because of specific
design considerations such as star-triangle switching
in the drive configuration. Nevertheless, correcting
the current offsets during the signal-processing stage
(before entering the fuzzy algorithm) solves the problem.
In terms of control-system design, the most important aspects are the variables, parameters and typical
measures of performance that we use to characterize
the system. After a preliminary study we selected:
 the spatial position of the cutting tool in terms of
Cartesian coordinate axes (x, y, z) (mm);
 spindle speed (s) (rpm);
 relative feed speed between tool and worktable
( f, feed rate) (mm/min);
 cutting power invested in removing metal chips
from the workpiece (Pc) (kW);
 current consumed at the main spindle during the
removal of metal chips (IS) (A);
 radial depth of cut (a, cutting depth) (mm); and
 cutting-tool diameter (d) (mm).
In order to evaluate system performance, we needed
to select certain suitable performance indices. The
milling process basically consists of two operations,
rough milling and finishing. The differences in these
operations objectives are what will decide which
performance index is useful in each operation. The
quality and geometric profile of the cutting surface
is paramount in finishing operations, whereas the
quantity of metal removed from the workpiece is
the main issue in rough-milling operations.

355

This work dealt essentially with rough milling, so

the main index was the metal-removal rate (MRR).

3. Fuzzy-logic controller for machining

processes
The fuzzy controller follows rather classic lines.
The controllers actual core performs on-line actions
to control the feed rate. The three basic tasks known
as fuzzification, decision-making and defuzzification
were implemented on-line. The main design steps for
this controller and their corresponding implementation were as follows.
1. Defining input and output variables, fuzzy partitioning and building the membership functions.
The input variables included in the error vector
e are the current-consumed error (DIS in A) and
the change in current-consumed error (D2IS in A).
The manipulated (action) variable we selected was
the feed-rate increment (Df in percentage of the
initial value programmed into the CNC), whereas
the spindle speed is considered constant and preset
by the operator.



u GC  Df 
eT KE  DIS ; KCE  D2 IS ;
(1)
where KE, KCE and GC are scaling factors for
inputs (error and change in error) and output
(change in feed rate), respectively.
The fuzzy partition of universes of discourse
and the creation of the rule base were drawn from
the criteria of skilled operators, although we did
have to apply a cut and trial procedure as well.
Fig. 2 shows the fuzzy partition thus obtained.
2. Constructing the rule base and generating crisp
output.
We will consider a set of rules consisting of
linguistic statements that link each antecedent
with its respective consequent, having, for instance, the following syntax if
DIS is positive and D2 IS is positive; then Df is
positive big
A matrix arrangement of the rule basis is shown
in Table 1.

356

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

Fig. 2. Fuzzy partitions and membership functions for (a) DIS, D2IS and (b) Df.

The controller output is inferred by means of the

compositional rule. The sup-product compositional
operator was selected for the compositional rule of
inference.
mDIS ; D2 IS ; Df
m n

S1 T2 mDISi DIS ; mD2 IS D2 IS ; mDfi Df 

i

i1

(2)

f
k f k  1 Df
k

where T2 represents the algebraic-product operation

and S1 represents the union operation (max), m
n 9. The crisp controller output, which is used to
change the machine-table feed rate, is obtained by
defuzzification employing the center-of-area (COA)
method [11] defined as
P

i mR Dfi  Dfi
(3)
Df P
i mR Dfi

Table 1
FLC rule base
D2IS

N
Z
P

where Df
is the crisp value of Dfi for a given crisp
input (DISi , D2 ISi ).
The crisp-control action (generated at each sampling instant) defines the final actions that will be
applied to the CNC set points. The strategy used to
compute f determines what type of fuzzy regulator is
to be used. In this case it can be a PI-FLC

DIS
N

NB
NS
ZE

NS
ZE
PS

ZE
PS
PB

(4)

or a PD-FLC
f
k f0 Df
k

(5)

Feed-rate values (f) were generated on-line by the

controller and fed in with the set point for the spindle
current (ISr) and measured value (IS) provided by
the CNC from the internal spindle-drive signal, as
explained in Section 4.
The static inputoutput mapping can be represented
by the nonlinear control surface shown in Fig. 3, considering that DIS 2 1:5; 1:5, D2 IS 2 1:5; 1:5 and
Df 2 10; 10.
These nonlinearities are essential in order to achieve
good performance. Recently it has been proved that
when trapezoidal membership functions are used,
the resulting system is the sum of a global nonlinear
controller (static part) and a local nonlinear PI controller (dynamically changing with regard to the input

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

357

Fig. 3. Fuzzy control surface.

space) [12]. Therefore, it is expected that this kind of

membership function can be relevant for dealing with
nonlinear process behavior.
3.1. Experiments for evaluating control-system
performance
The preliminary experiments were conducted using
a 5.8 kW, 4-axis milling machine, the ANAK-MATIC2000-CNC, equipped with a Fagor 8025 CNC, which
was interfaced with a personal computer over a 9600baud RS-232 communications link (see Fig. 4a).
The objective for this fuzzy-control system was to
manipulate the machine tools cutting conditions, with
some further possibilities in addition to those provided
by the CNC-PLC alone. In this new context, while the
CNC-PLC guided the sequence of tool positions and/or
the path of the tool during machining, the fuzzy controller determined the proper feed rate f. KE KCE 1
and GC 1 were used as scaling factors. We added a
personal computer to the control scheme in order to
measure the spindle-motor current IS on the basis of a
Sandvik Coromant TMCS-001 sensor and to perform
the generation of f
values. Other measurement devices
such as a PCL-718 acquisition card were also installed.
The set point ISr was fixed manually so that the
MRR would reach the maximum allowable value under
the rest of the technological constraints. The control

scheme was implemented on the basis of these definitions. We developed the real-time control system in
C/C programming language on a Pentium PC. Feedrate values (f) were generated on-line by the controller
and fed in with the set point ISr and measured value IS.
In order to evaluate the performance of the algorithm,
two tests were performed using a 2011 aluminum-alloy

Fig. 4. (a) Experimental setup and (b) cross-section of the

workpiece used in experiments.

358

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

Fig. 5. Real-time control dynamics for (a) test 1 and (b) test 2.

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

359

Table 2
Features and results of two laboratory tests

Nominal command feed rate f0 (mm/min)

Nominal spindle speed s0 (rpm)
Maximum cutting depth amax (mm)
Cutting tool
Diameter of the cutting tool (mm)
Number of teeth
Workpiece material
Setpoint ISr (A)
Overshoot Mpt (%)
Settling time ts in the last step (s)
IAE index
ISE index
ITAE index

workpiece. A cross-section of the workpiece used

is shown in Fig. 4b. The profile consisted of one ramp
and three step-shaped disturbances (12 mm maximum
depth of cut).
The spatial position of the cutting tool was set manually by the operator, and the tool was held at a constant
vertical position during the experiments. The machining was defined to proceed in one direction only. New
milling cutters were used in this experimental work.
The control systems behavior was evaluated by
assessing accuracy and control-loop stability. In order
to evaluate the accuracy of the control system, the
integral absolute error (IAE), the integral square error
(ISE) and the integral of time-weighted absolute error
(ITAE) were calculated. The transient response was
analyzed to evaluate control-loop stability as well.
Moreover, the overshoot Mpt was also computed
because it should be tracked as a warning of dangerous
situations and proximity to control-loop instability.
Fig. 5 shows the real-time control dynamics of two
different tests. First of all, the fuzzy-controlled milling
process is robustly stable under real-time conditions.
As clearly shown in Fig. 5, the spindle-motor current
remains constant during machining, as is most clearly
visible when the cutting depth is one-third to one-half
of tool diameter, regardless of disturbances and nonlinearities in the milling process. Only in the first step
(small depth of cut) is the set point not reached. Hence
the real-time experiment verifies the applicability of
fuzzy-control systems designed using the methods
herein introduced. All the experimental data including
performance indices are summarized in Table 2.

Test 1

Test 2

300
2000
10
F1205-10 Vallorbe
10
2
Aluminum alloy 2011
1.4
13.46
8.05
233.14
74.68
1105.8

200
1500
12
F1205-25 Vallorbe
25
2
Aluminum alloy 2011
2.2
17.59
9.45
511.54
484.41
1498.6

4. Open CNC and new control functions

The demand for open control systems is on the rise
as the result of the current need for enhanced control
functionality in industry. To date, only a small group
of machine-tool manufacturers has accepted adaptive
controllers. One main reason for their reluctance is the
need for an accurate model for calculating controller
parameters. Additionally, adaptive controllers cannot
be applied reliably to various combinations of machining processes, materials and tools without major
changes in the control algorithm and its parameters.
Some European CNC manufacturers do not include
any adaptive functions in their products [13], while
others offer an ac control whose performance is very
restricted [14]. The adaptive module is rarely used in a
flexible machining environment containing many types
and combinations of materials and tools, where rigid
CNCs would actually constitute a major handicap.
There are also some commercially available add-on
systems (e.g. OPTIMIL by OMAT control technologies) that can perform real-time adaptive control [15].
OPTIMIL is based on an expert system that combines
tool and material information. However, the underlying
technology has some limitations, such as the annoying
calibration procedure (a learning stage or teach-in is
required), plus synchronization problems with other
applications in open CNCs.
The way of embedding the new generation of
intelligent controllers in CNCs is now ready for
use. Nowadays open CNCs enable internal control
signals to be gathered and mathematically processed

360

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

by means of integrated applications such as additional

functions embedded right in the control core. By
this means control systems original features can be
improved upon, with the addition of machine-tool
manufacturers know-how and new control-system
developments incorporated in the CNC.
At present two levels of opening are available from
machine-tool and CNC manufacturers. The degree of
opening varies, however, depending on the CNC part
involved. There are two categories of opening/openness, differentiated by the importance of that feature
from the control-system viewpoint: opening of man/
machine communication (MMC) and opening of the
CNCs numerical control kernel (NCK). The first category is the one responsible for interaction with the user
(e.g. office applications can be developed using the
DDE dynamic data exchange protocol). Use of a bus
(e.g. a multi-point interface bus) enables communication with low-level data. The more important level
of opening from the control-system viewpoint can be
found in the NCK, where real-time critical tasks (e.g.
axis control) are scheduled and performed.
4.1. Embedding the fuzzy controller
This section explains how the fuzzy controller
designed in Section 3 is embedded in the open CNC.
Nowadays the integration of any control system in the
CNC is a complex task that requires the use of various

software utilities, technologies and development tools.

Three classical technologies can be used: a softwaredeveloping tool (C, Visual Basic), an open- and
real-time CNC and communications technology.
The general outline of the control system is depicted
in Fig. 6. First the fuzzy controller was programmed in
C/C and compiled, and as a result a dynamic-link
library (DLL) was generated. Other tools were used as
well. A Sun Workstation, the UNIX operating system
and C were used to program the NCK. A PC, the
WINDOWS 9X operating system and Visual Basic
were used for programming the MMC. Inter-module
communications between the MMC and the NCK
were established through DDE. Finally, and for the
sake of simplicity, the user interface was programmed
in Visual Basic.
The application was developed on the basis of a
Sinumerik 840D CNC. The process of integrating the
software application into the NCK involved a series of
steps. Development, including editing, compiling and
code linking, was run at a workstation. After that, the
file was transferred to the PC, where OEM software
ran debugging routines. Finally, the code was copied
on a PCMCIA card and inserted into the CNC [16].
An experimental internal data-acquisition system
was developed and used to record the control internal
spindle-drive signal. The system enables a selected
drive signal to be recorded and provides stored data on
the hard drive of the Sinumerik 840D user-interface

Fig. 6. Control system diagram.

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

PC (MMC). Internal data-acquisition software was

developed to obtain control information. The maximum sampling frequency could not be any longer than
500 Hz, defined by the servosystems control cycle.
Therefore the acquisition-signal software works at the
same time for which servosystems are configured
in the CNC (i.e. 2 ms). The software consists of a
data-acquisition module in the NCK that records the
selected data into an internal buffer and an MMC
background task that receives the completed measurement and stores it in the hard drive of the user-interface PC. Data transfer is performed by splitting the
recorded data into a number of fragments, due to
limitations in the Sinumerik 840Ds file systems.
The signals to be recorded (in our case the current
consumed at the spindle) are configured using specially added machine data. Recording can be started
and stopped either manually through an MMC application or under the NC programs control.

5. Industrial tests: evaluation

Tests were carried out in the SL400 SORALUCE
machining center, which was equipped with a SIEMENS open CNC Sinumerik 840D. The SL400 model
SORAMILL milling machine has a traveling column
and a fixed table configuration. It possesses highprecision slideways that allow all three axes to
reach a speed of up to 15 m/min. The machine has

361

high-rigidity, high-precision features, and the workpiece has no influence whatsoever on the moving part.
The cutting tool was a Sandvik D38 R218.192538.35-37.070HA (020/943) face mill 25 mm in diameter with four inserts of the SEKN 12 04 AZ (SEK
43A, GC-A p25) type. The workpiece material used
for testing was F114-quality steel. The maximum depth
of cut was 20 mm, the nominal spindle speed was S0
850 rpm, and the nominal feed rate, f0 300 mm/min.
The actual dimensions of the profile were 334 mm
486 mm. The profile is depicted in Fig. 7b.
We used an internal spindle-drive signal as our
control. The processing-signal analysis was done
on-line. Signal processing focused on correcting current offsets. Current offsets were corrected on the basis
of the model that relates spindle speed to current [17].
This model was created using linear regression. Moreover, further filtering was performed using the mean
value of the signal in the sampling period.
The set point was estimated according to a/the
constraint given by the power available at the spindle
motor, the material and the tool characteristics.
In order to compare the performance of the PI and
PD-FLC, a reference value of 18 A was set during the
experiments. In the second study, the set point used
was 22 A in order to make an appropriate comparison
with a commercial product.
The results of applying two fuzzy controllers (PI and
PD-FLC) to machine an irregular profile (see Fig. 7b)
are depicted in Fig. 8. Fig. 8a shows the behavior of

Fig. 7. (a) Tool used for industrial tests and (b) workpiece for industrial tests.

362

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

Fig. 8. (a) Behavior of the controlled variable and (b) action variable (feed rate) when machining an irregular workpiece.

the current consumed at the main spindle, and Fig. 8b

depicts the variation in the feed rate. The negative
effect of integral components is reflected in the transient response.

An illustrative example was inserted to uphold the

validity of the theoretical approaches examined in this
paper. The second study compared an embedded PDFLC, the OPTIMIL [15], and the CNC working alone.

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

363

Fig. 9. (a) Time responses of the three control systems and (b) behavior of the controlled variable for the three control systems analyzed.

The results are shown in Fig. 9. The reduction in

machining time was very similar for the OPTIMIL
and the embedded PD-FLC (10%), yet the overshoot
was bigger for the embedded PD-FLC (23.7%).
From the technological viewpoint, this overshoot is
allowable for rough milling operations. It is important
to note that in some approaches the goal of controller

design is to limit the overshoot percentage to 20%

[3].
The OPTIMIL, when used as external equipment,
requires synchronization steps and a tedious calibration procedure including a learning phase. On the
other hand, the embedded PD-FLC does not require
any calibration procedures or external sensors, and the

364

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

Fig. 10. The workpiece after eight operations using the PD-like FLC.

application can be supplied with just a CNC and no

other external equipment. The workpiece after the
rough milling operation is depicted in Fig. 10.
This study did not consider linear controllers for
this application, because the fuzzy controller has been
demonstrated to yield better results than linear control
loops [18]. The issue of stability is not dealt with here
either. Some strategies for checking the stability of
control systems of the type herein introduced are
suggested in [19].
5.1. Scientific and technological prospect
The embedded fuzzy-control system will have an
immediate application in milling machines and
machining centers. The methodology and algorithm
can, however, be applied in any cutting process such as
turning or drilling. Therefore, the results shown in this
paper will be applicable to a large sector of the
equipment-production industry. Certainly, in spite of
the fact that the software was developed for one kind
of open CNC, the architecture ensures flexibility and
allows the integration of a variety of control strategies
in order to embrace the largest possible number of
applications.
Furthermore, the solution is highly competitive,
since it uses the CNCs own resources and requires

no additional equipment such as sensors, acquisition

and signal-processing cards or other external devices.
This is a key issue if the proposed solution is to be
integrated in an industrial environment.
On the basis of the economic study provided by
SORALUCE S. Coop., this system may be expected to
boost machine-tool suppliers competitive capacity
considerably [20]. The economic benefits for
machine-tool manufacturers would come basically
from the increase in sales resulting from the availability of an easily installable, configurable, programmable solution that would lead to an increase in
minimum machine price and an immediate pay-off
of the manufacturers investment. Moreover, machinetool users could enhance their machine-tool use ratio,
machining-operation safety (due the automatic selection of cutting speeds) and the useful life of their
cutting tools (due to a reduction of improper cutting
conditions).
SORALUCE S. Coop. has run a major study to
evaluate the savings provided by installing this system. An approximate calculation was made for the
case of milling machines, considering the mean cost of
other commercial solutions. The estimates of cost and
savings are based on data on the standard machining
operations involved in milling. The estimate of cutting-tool costs, however, takes only rough-milling

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

operations into account. The total estimated savings

are about 78%.1

6. Conclusion
The results of transferring technology to a machinetool manufacturer through cooperation with a technological center show the effectiveness of a fuzzy-control system for dealing with the nonlinear behavior of
the machining process. The embedded fuzzy controller is able to work using only internal CNC signals.
Moreover, it can run in parallel with other applications
without posing any synchronization problems. However, future work is necessary in order to refine the
fuzzy controllers performance and so to improve the
transient response.
All this work, from hypothesis to implementation
and experimentation, including design and analysis,
is done following the classic patterns. Our focus at
all times lay on the practical implementation, so the
results we have presented in this paper are for a reallife industrial plant. The results show that internal
CNC signals can double as an intelligent, sensorless
control system. Actual industrial tests show a higher
machining efficiency; in-process time is reduced by
10% and the total estimated savings the system would
provide are about 78%.

Acknowledgements
This work has been partially supported by two
Spanish-funded projects, PETRI 95-0336-CT
COCOM and TAP99-1122-C02-01 MHECOS. The
authors would like to thank referees for their helpful
suggestions and comments. Moreover, the authors
express their deepest gratitude to all the people who
contributed to the implementation of the control system at SORALUCE S. Coop. and Ideko Technology
Center. Likewise, the authors gratefully acknowledge
the contributions of S. Ros, G. Arrate and J. Aranceta
in assisting with the implementation presented in this
work.
1

More details about costs and implementation cannot be

provided because they may include confidential data not releasable
under Mondrago n Corporacio n Cooperativa (MCC) policy.

365

References
[1] Y. Koren, Control of machine tools, Journal of Manufacturing
Science and Engineering 119 (1997) 749755.
[2] L.K. Lauderbaugh, A.G. Ulsoy, Model reference adaptive
force control in milling, ASME Journal Engineering of
Industry 111 (1989) 1321.
[3] S.J. Rober, Y.C. Shin, O.D.I. Nwokah, A digital robust
controller for cutting force control in the end milling process,
Journal of Dynamic Systems, Measurement and Control 119
(1997) 146152.
[4] R.E. Haber, R.H. Haber, S. Ros, A. Alique, J.R. Alique,
Dynamic model of the machining process on the basis
of neural networks: from simulation to real time application, Lecture Notes in Computer Science 2331 (2002) 574
583.
[5] L.-C. Lin, G.-Y. Lee, Hierarchical fuzzy control for C-axis of
CNC turning centers using genetic algorithms, Journal of
Intelligent and Robotic Systems: Theory and Applications 25
(3) (1999) 255275.
[6] Y. Liu, L. Zhuo, C. Wang, Intelligent adaptive control in
milling process, International Journal of Computer Integrated
Manufacturing 12 (5) (1999) 453460.
[7] Y. Liu, C. Wang, Neural networks based adaptive control and
optimization in milling process, International Journal of
Advanced Manufacturing Technology 15 (1999) 791795.
[8] H. Schulz, Hochgeschwindigkeitsbearbeitung Carl Hansen
Verlag Munchen Wien, Germany, 1999.
[9] Y. Altintas, Prediction of cutting forces and tool breakage in
milling from feed drive current measurements, ASME Journal
of Engineering of Industry 11 (1992) 386392.
[10] T. Kim, P. Kim, Adaptive cutting force control for a
machining centre by using indirect cutting force measurements, International Journal of Machine Tool Manufacturing
36 (8) (1996) 925937.
[11] R. Yager, D. Filev, Essentials of Fuzzy Modeling and Control,
Wiley, New York, 1994, pp. 313354.
[12] H. Ying, Analytical structure of a typical fuzzy controller
employing trapezoidal input fuzzy sets and nonlinear control
rules, Information Science 116 (1999) 177203.
[13] Fagor Automation, Needs of new numerical controls,
Produccio n Meca nica 3 (1999) 7483 (in Spanish).
[14] Siemens AG, Sinumerik 840D/FM-Nc., 1996, pp. 3640.
[15] A. Nosrat, Adaptive controls lead to peak efficiency,
American Machinist, April (1997) 7880.
[16] Siemens AG, Sinumerik 840D, OEM-package NCK, Software release 4, Users Manual, Siemens AG, 1999.
[17] M. Kaever, S. Platen, M. Weck, Control integrated process
monitoring for rough milling operationsinnovative logging
module solution, in: Proceedings of the 2nd International
Workshop on Intelligent Manufacturing Systems, Leuven,
Belgium, pp. 727736.
[18] R.E. Haber, R.H. Haber, A. Alique, S. Ros, Application of
knowledge-based systems for supervision and control of
machining processes, in: S.K. Chang (Ed.), Handbook of
Software Engineering and Knowledge Engineering, vol. 2,
World Scientific, Singapore, pp. 673710.

366

R.E. Haber et al. / Computers in Industry 50 (2003) 353366

[19] G. Schmitt-Braess, R.E. Haber, R.H. Haber, A. Alique,

Multivariable circle criterion: stable fuzzy control of a
milling process, in: Proceedings of the IEEE International
Conference on Control Applications, Glasgow, UK, 2002,
pp. 385390.
[20] SORALUCE S.C.L., Design and implementation of FLC to
regulate the cutting force and its application to machine tool
(CONBOMAQ), Internal Report, Spanish National Plan for
R&D (PROFIT), FIT-020200-2001-49, 2001.
R.E. Haber was born in Santiago the
Cuba, Cuba in 1969. He received the BE
degree with first class honors in automatic control engineering from the Universidad de Oriente (UO), in 1992. In
1995, he was the recipient of PhD grant
supported by the Spanish Office for
Scientific Cooperation (AECI) in Spain.
He received an excellent cum laude PhD
in industrial engineering from Technical
University of Madrid (UPM) in 1999.
He joined with the Spanish Council for Scientific Research
(CSIC) in 1999 working in several research and development
projects. In 1999, he also joined as Assistant Professor with the
Computer Engineering Department at the High Technical School
in the Universidad Autonoma de Madrid (UAM) teaching in
several graduates courses. He has published several technical
papers in specialized journals and chapters of books. His research
interests include control theory and applications, hardware
software solutions, soft-computing techniques, classical and
adaptive control, supervisory control, complex electromechanical
processes.
J.R. Alique received the BSc degree in
physics from the University Complutense of Madrid (UCM), Spain in 1969.
He received the MSc and PhD degrees in
physics sciences from International Institute of Philips, North Holland in 1971,
and University Complutense in 1973.
From 1969 to 1970, he was Assistant
Professor of physics and automation at
UCM. He has been working in automation since 1972 in the Spanish Council
for Scientific Research (CSIC). From 1991 to 1996, he was the
Director of the Technology Transfer Office in the CSIC. From 1996
to 2001, he was the vice director of R&D projects in the Ministry of
Science and Technology. He is currently researcher at Instituto de
Automatica Industrial (CSIC) and Head of the Computer Science
Department. His research interests include automation of machining

processes, numerical controls, robotics systems and intelligent

modeling and control.
A. Alique received the BSc and PhD
degrees in physics from University
Complutense of Madrid in 1972 and
1979, respectively. He is currently researcher at Instituto de Automatica
Industrial (CSIC). He has published
several technical papers on supervisory
fuzzy control and modeling of machining processes. He is currently the main
researcher of two projects funded by
CICYT. His research interests include
fuzzy modeling and control, robotics systems and intelligent
control.
J. Hernandez was born in 1969. He
received the degree of physics from the
Universidad de Valladolid in 1992. He
joined the Research Centre IKERLAN in
1993 being the recipient of PhD grant.
He received the PhD degree in physics in
1997. In 1998 he joined IDEKO Technology Centre. As a researcher in the
Control Engineering Department, he has
taken part in several European and
international projects (SIMON, COMPRO, IDMAP, IMS-SIMON) in the field of machining process
monitoring and control. His main fields of interest include:
monitoring and control of machining and forming processes,
monitoring of machine components, and development of control
strategies for improving machine-tool performance.
R. Uribe-Etxeberria was born in 1967.
In 1990 he got the degree of Industrial
Technical Engineering in the field of
electronics. In 1991 he received the MSc
degree in control engineering and information technology at UMIST (University
of Manchester Institute of Science and
Technology). He joined IDEKO in 1991
as a research engineer working in the
fields of control engineering, sensoring
and signal processing and their applications in the machine-tool world. Since 1996, he is the head of the
Control Engineering Department at IDEKO. His fields of interest are:
machine automation, monitoring and control of machining processes
and the application of information technology to machine tools.