International Journal of Machine Tools & Manufacture 94 (2015) 1–14

Contents lists available at ScienceDirect

International Journal of Machine Tools & Manufacture

journal homepage: www.elsevier.com/locate/ijmactool

Feasibility study of in-process compensation of deformations in flex-

ible milling
Eduardo Diez a,n, Hilde Perez b, Juan Marquez c, Antonio Vizan c
a
Department of Mechanical Engineering, Universidad de La Frontera, Temuco, Chile
b
Department of Mechanical, Informatics and Aerospace Engineering, Universidad de Leon, Leon, Spain
c
Department of Mechanical and Manufacturing Engineering, Universidad Politecnica de Madrid, Madrid, Spain

art ic l e i nf o a b s t r a c t

Article history: During the machining of thin-walled parts, deformation can occur resulting in dimensional errors. These
Received 11 February 2015 dimensional errors cause a variation on cutting forces. From the actual measured cutting forces and the
Received in revised form estimated forces resultant from rigid machining, it is possible to determine the value of this deformation.
26 March 2015
Based on this, an on-line system for compensating workpiece errors, has been developed. The system is
Accepted 27 March 2015
based on correcting the relative position of the tool-workpiece during machining by means of a piezo-
Available online 21 April 2015
electric actuator. The objective is achieved in real time to compensate for the part deformations from the
Keywords: measurement of the cutting forces, without the programming of the tool path trajectories in the machine
Milling tool being affected.
Low rigidity parts
& 2015 Elsevier Ltd. All rights reserved.
In-process compensation
Piezoelectric actuator

1. Introduction occurs in micro-electro-mechanical systems (MEMS) components,

At present, machining of low rigidity parts is a common man-
ufacturing operation in many industrial sectors. There are ex-
amples in areas as diverse as the aeronautics and aerospace in-
dustry, the manufacture of molds and dies in the plastics industry,
or the manufacture of micro-electro-mechanical systems (MEMS).
In the machining of thin-walled parts, one of the most significant
problems is the deformation the workpiece suffers as a result of
the forces that occur during cutting. In order to reduce the errors
produced by deformations, the common practice is to select con-
servative cutting conditions, such as lower feed rates or depths of
cut (axial or radial), or to increase the number of tooth passes, thus
raising machining times and costs.
In particular, in the aeronautical sector, the design of thin-
walled components continues to be of great interest because they
make possible the production of lightweight resistant structural
parts, which provide airplanes with greater energy efficiency [1].
In this sector, high-speed machining has improved the capacity to
produce thin-walled components because the forces produced are
greatly reduced. In spite of this, better control of these deforma-
tions would enable more productive processes. This problem also
n the machining geometry, stability of the process, indirect estima-
Corresponding author.
E-mail addresses: eduardo.diez@ufrontera.cl (E. Diez),
hilde.perez@unileon.es (H. Perez), avizan@etsii.upm.es (A. Vizan).
which are desired for machining processes. In micro-milling op-
erations, although the tool can have a greatly reduced diameter,
the thickness of the walls can also be reduced. In addition, it can
occur when machining worpkpieces without thin walls that their
morphology may encourage deformation during machining.
2. State of the art
For decades, one of the most important aspects covered in the
study of machining processes has been the estimation of the
cutting forces. The interest in estimating these forces using reliable
models lies in the need, among others, to obtain a more precise
knowledge of the process, in having models that allow the ne-
cessary machining power to be estimated, in the possibility of
accurately defining the elements of the machine-tool and in the
use of the information of the cutting forces in process control al-
gorithms that optimize the machining process. Through the cut-
ting forces it is possible to obtain large amount of data on the
material removal process, so that any variation in the values of the
forces will indicate changes in the machining conditions, as for
example a change in the properties of the workpiece, changes in
tion of the surface finish, etc.
In milling of low rigidity parts, limitations of the process are

http://dx.doi.org/10.1016/j.ijmachtools.2015.03.008
0890-6955/& 2015 Elsevier Ltd. All rights reserved.
2 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14

Nomenclature ka [N/mm] specific cutting pressure in axial direction

n [rpm] spindle speed
ap [mm] nominal axial depth of cut N [-] tool flute number
apa [mm] actual axial depth of cut R [mm] tool radius
ae [mm] radial depth of cut (width of cut) φ [rad] tool position angle
A0 [mm2] mean chip section φaver [rad] angle of the engagement arc where cutting force is
d [mm] tool diameter applied
fz [mm/tooth] feed per tooth φen [rad] Entry angle
f [mm/min] linear feed φe [rad] engagement angle
Ft [N] tangential cutting force φena [rad] actual entry angle
Fr [N] radial cutting force φex [rad] Exit angle
Fa [N] axial cutting force φj [rad] angle of the tip of the flute j
Fx [N] cutting force in X direction φm [rad] approximated angle of the engagement arc where
Fy [N] cutting force in Y direction cutting force is applied
Fz [N] cutting force in Z direction φpr [rad] projected angle of the cutting edge
h [mm] mean chip thickness δ [mm] workpiece deviation due to cutting force
kt [N/mm] specific cutting pressure in tangential direction λs [rad] tool helix angle
kr [N/mm] specific cutting pressure in radial direction xc [V] actuator command signal

caused by the flexibility of the workpiece, which contributes to

both the instability of the process (high vibrations during the
process, chatter) and static deformations of the workpiece during
machining. Both issues directly affect the production times and the
part quality. For the case of difficult-to-cut materials, like titanium
alloys or nickel based alloys, deformation is produced due to high
cutting forces needed to cut these materials.
In order to improve the machining operations of flexible
workpieces, one of the most frequently studied aspects for several
years has been the stability of the process and dynamic related
issues, with the goal being to avoid chatter during machining.
These studies have concentrated mainly on determining the op-
Fig. 1. Basic diagram of the system to compensate for deformations.
erating conditions that maximize material removal, avoiding the
instability of the process associated with one or more of the ei-
genmodes of the structure of either the spindle-tool system or the
workpiece. An extensive amount of literature on this subject has
been published. Since the topic of this article is related to com-
pensation of the errors due to static deformations of the ma-
chining system under stability conditions, literature review in this
article do not cover dynamic related issues.
In the case of static deformations during the machining of low
rigidity components, studies mainly concentrate on quantifying
the errors by means of direct measurement or numerical simula-
tion. In the case of simulation, the deformations and the cutting
forces must be calculated simultaneously. These calculations have
been made using finite element simulation [2] and mechanistic
models of the cutting process [3]. In any case, once the errors have
been quantified, an offline correction is made by reprogramming
the CNC code or by creating machining strategies that include the
deformations the workpiece will undergo. Ratchev et al. [4,5]
proposed a methodology to evaluate and correct the deformations
that a flexible workpiece suffers during machining. The approach
is based on identifying and modeling key characteristics of the
process that affect the deformation of the workpiece, modeling the
cutting process numerically (calculation of forces), predicting the
deformations using the finite element method, and finally cor-
recting the deformation by programming an improved tool tra-
jectory. A variant of this methodology, in which the forces are
calculated by means of force and deformation models, was de-
veloped by Ratchev et al. [6]. A similar approach, applied to ma-
chining of tubular workpieces, was developed by Bera et al. [7].
Such methodologies base their effectiveness on the hypothesis
that the machining operation has high repeatability. Another ap-
proach to correcting machining errors due to deformations is by Fig. 2. Cutting geometry for milling of low-rigidity parts.
 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14 3

Fig. 3. Cutting situations depending on the angular position of the tool: (a) cutting zone A, (b) cutting zone B and (c) cutting zone C.

Table 1
Cutting conditions used in the tests for determining the coefficients of the specific
cutting pressure.

Test N d n fz ap ae Material Tool

– – mm rpm mm mm mm – –

1 1 8 1200 0.020 7 0.5 AA7075 Garant 191000

2 1 8 1200 0.040 7 0.5 AA7075 Garant 191000
3 1 8 1200 0.060 7 0.5 AA7075 Garant 191000
4 1 8 1200 0.080 7 0.5 AA7075 Garant 191000
5 1 8 1200 0.100 7 0.5 AA7075 Garant 191000
6 1 8 1200 0.120 7 0.5 AA7075 Garant 191000
7 1 8 1200 0.010 8 1 AA2011 Izar 4412
8 1 8 1200 0.015 8 1 AA2011 Izar 4412
9 1 8 1200 0.020 8 1 AA2011 Izar 4412
10 1 8 1200 0.025 8 1 AA2011 Izar 4412
11 1 8 1200 0.030 8 1 AA2011 Izar 4412
12 1 8 1200 0.035 8 1 AA2011 Izar 4412
13 1 8 1200 0.040 8 1 AA2011 Izar 4412
14 1 8 1200 0.045 8 1 AA2011 Izar 4412
15 1 8 1200 0.050 8 1 AA2011 Izar 4412
16 1 8 1200 0.055 8 1 AA2011 Izar 4412
17 1 8 1200 0.060 8 1 AA2011 Izar 4412
18 1 8 1200 0.065 8 1 AA2011 Izar 4412
19 1 8 1200 0.070 8 1 AA2011 Izar 4412
20 1 8 1200 0.075 8 1 AA2011 Izar 4412
21 1 8 1200 0.080 8 1 AA2011 Izar 4412

Fig. 4. Evolution of the actual axial depth of cut, the average chip thickness and the
cutting forces for the different cutting zones.

integrating on-machine measuring tools and artificial intelligence.

In this approach, the corrected trajectory is calculated by a neural
network model via an iterative learning process through error data
obtained using on-machine measurement [8].
Another way to approach the problem of static deformations
during machining low rigidity parts, and that has been less fre-
quently studied to date, is the in-process measurement and
compensation of the deformations. This option has the advantage
of eliminating expensive computer simulation hours when offline Fig. 5. Evolution of the specific cutting pressure for aluminum alloys and tested
compensation is required. In addition, in many cases the tools.
4 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14
Table 2
Cutting coefficients for aluminum alloys and tested tools.
AA7075 kt ¼ 19,477 h  0.472 kr ¼ 33,926 h  0.764
AA2011 kt ¼ 40,057 h  0.36 kr ¼ 7343 h  0.669
Fig. 6. Equivalent assembly to maintain constant rigidity.
machining process is not sufficiently repeatable to ensure the
success of an offline correction of the tool trajectory. Another ad-
vantage of in-process compensation is that a higher grade of
workpiece accuracy can be achieved in a relatively lower grade
machine tool through the use of error compensation techniques.
This is desirable in the industry today as it involves a combination
of high accuracy, low cost and high production rate [9]. Online
compensation of the static deformation has received less attention
from researchers because traditionally there has been a resistance
to incorporating elements attached to the conventional machine
tool environment (sensors, actuators and tools). However, the
current machine-tool's own systems, feed drives and main drive
are incapable of providing a rapid response to execute such cor-
rections in real-time, from the direct measurement of the de-
formation and this would justify the incorporation of mechatronic
systems to meet such requirements.
The development of mechatronic systems has enabled new
technological options to approach the problem from a mecha-
tronic point of view [10]. One option is the integration of
Fig. 7. Setup scheme to study cutting forces and part deformation in flexible machining.
piezoelectric actuators into the manufacturing environment to
improve some machining aspects [11]. Piezoelectric actuators are
devices able to provide highly precise movements (in the nano-
metric range) with a very low response time. In addition, their
stiffness is high, which makes them especially appropriate for
precision machining applications [12,13]. Although the applic-
ability of piezoelectric actuators as autonomous feed drive systems
is restricted in the field of macro-manufacturing due to their short
range of motion [14], they are widely applied in combined feed
drive systems, as the review by Park et al. shows [11]. One im-
portant characteristic of piezoelectric actuators is that they can be
easily incorporated into the machine tool; this means that piezo-
electric actuators can be applied in any production process where
sub-micrometric positioning precision or fast positioning response
is required. The idea of using piezoelectric actuators to improve
manufacturing tasks was introduced some years ago when a pie-
zoelectric actuator was used to position a high-precision turning
tool quickly and accurately [15,16]. The researchers designed and
built a fast servo system for precise and fast tool positioning (Fast
Tool Servo, FTS) for a conventional CNC lathe. The servo system
had one degree of freedom and was driven by a piezoelectric ac-
tuator. The servo system is operated in a closed loop (using a ca-
pacitive displacement sensor) with a control loop that provides a
bandwidth of 200 Hz and a positioning resolution of 20 nm. A si-
milar concept applied to milling was developed by Elfizi et al. [17]
with the aim of designing and building precision drive systems to
improve milling operations. For the single-axis system they de-
veloped, the tracking error was reduced in a milling operation
from 100 μm to 10 μm, compared to a linear motor drive system.
For the dual-axis system they developed, the combined feed drive
system reduced the tracking error by 38%.
Piezoelectric actuators have also been applied to improve the
positioning precision of the tool in several degrees of freedom, as
researchers Drossel and Wittstock [18] and Neugebauer et al. [19]
propose. In their work, a support is presented to control the or-
ientation of the spindle to improve machining operations, called
adaptive spindle. The adaptive spindle is a component that allows
precision positioning during machining operations using a hex-
apod structure that is driven by piezoelectric actuators and that
supports the machine spindle.
 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14
Fig. 8. Deformation and cutting forces in stable machining, fz ¼0.04 mm, n¼ 1200 rpm, d ¼ 12 mm, ap ¼16 mm, ae ¼ 1.07 mm. Material AA2011.
This paper explores the application of a piezoactuator based
strategy to compensate for deformations of low rigidity parts
during peripheral milling. The proposed in-process compensation
strategy is based on the comparison of measured cutting forces
and nominal cutting forces calculated for rigid machining under
the same cutting conditions. Compared data is used to calculate
and generate the motion command signal to the piezoelectric
actuator, in order to produce small variations in the width of cut
that will compensate for deviations on the nominal tool trajectory
caused by workpiece deformations.
The calculation of the nominal cutting forces is realized based
on a reference model which is executed while machining. The
implementation of both the reference model and the data acqui-
sition processes are performed simultaneously (Fig. 1).
This paper presents, first and foremost, the reference model
that allows the calculation of cutting forces considering the de-
formation of the workpiece at an execution speed consistent with
the actual cutting process. Secondly, the experimental verification
of the model is shown with and without any workpiece de-
formation. The experimental equipment and the use of equivalent
test specimens are described, with which it is possible to observe
the influence of workpiece deformation on the cutting forces.
Subsequently, the compensation strategy for the part deformation
with a piezoelectric actuator, based solely on the measurement of
the cutting forces is described. Finally the on-line experimental
equipment used during the tests is described and the performance
5
of the compensation system is tested.
3. Reference model for the estimation of cutting forces when
milling low-rigidity parts
In order to compensate for workpiece deformation, a model
which allows accurate and fast calculations of cutting forces is
required. Small deviations from the actual cutting conditions re-
garding the nominal ones will be detected.
For this, a mechanistic model based on the average chip
thickness developed by the authors in previous works [20] is used
as the reference model. As this model is based on average chip
thickness, it is a fast method to estimate cutting force in real time
control environments. Unlike other models used in the research
community, it does not require iterations to calculate the cutting
forces along the edge. In this section the influence of the part
deformation on cutting forces when machining low-rigidity parts
with high strength tools is analyzed. This influence results in
variations on the cutting force values. These changes will be the
basis to compensate for part deformations.
In finish milling, the radial depth of cut is small in comparison
with the axial depth of cut in order to avoid large deformations
during machining. Therefore, the arc of engagement is small, as
shown in Fig. 2. The arc of the projected cutting edge (Eq. (1)),
which depends on the tool diameter, the helix angle and the axial
6 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14

depth of cut, is greater than the arc of engagement of the cut. performed for each angular position of the tool. Unlike other
methods for cutting force estimation [22–25], the model based on
2⋅ tan λs ⋅ap
φpr = average chip thickness can predict the cutting forces for every
D (1)
position of the tool in one step, without iteration along the edge
Because the arc of engagement is limited, only a portion of the [20]. The cutting forces in the tangential, radial and axial direction
flute is in the cut at any time. These conditions are typical for are determined by the following equations:
milling low stiffness parts, resulting in three cutting cases which
are identified by the entry and exit angles of the cutting flute. At Ft (φ) = kt (φ)⋅apa ⋅h¯ (φ) (5)
any time, the cutting geometry in such operations can be classified
into three different cases illustrated in Fig. 3. Fr (φ) = kr (φ)⋅apa ⋅h¯ (φ) (6)
Case A: The flute enters the cutting zone.
Case B: The flute exits the cutting zone but has not finished Fa (φ) = ka (φ)⋅apa ⋅h¯ (φ) (7)
exiting.
Case C: The flute exits the cutting zone. The axial depth of cut varies from the entry of the cutting edge
Depending on the cutting conditions, the milling forces may into the workpiece, increasing to a maximum in the middle cutting
lead to deformations of the workpiece and the tool. From the point zone (Zone B). The evolution of the axial depth of cut varies dif-
of view of this work, which focuses on the behavior of low rigidity ferently depending on the cutting area where the flute is. Eq. (8)
parts, the deformations that occur during machining are mainly provides the value of apa for these cutting zones.
due to the deformation of the workpiece. These conditions occur
⎧ φj − φen
when the ratio between the tool diameter and the wall thickness ⎪ ap φen < φj ≤ φex Zone A
of the workpiece to be machined is high. The workpiece de- ⎪ φpr
formation results in a variation of the radial depth of cut which ⎪
⎪ (φex − φen )
will affect the chip thickness and the entry angle of the cutting apa (φ) = ⎨ ap φex < φj ≤ φen + φpr Zone B
⎪ φpr
edge as well, making the cutting forces vary.
⎪
Therefore, it is necessary to ensure that, in addition to the ex- ⎪ a + (φex − φj ) a φ < φ ≤ φ + φ
ecution speed, the reference model is sensitive enough to reveal ⎪ p p ex j ex pr Zone C
⎩ φpr (8)
changes in real forces due to any small variations in the radial
depth of cut, normally less than 200 μm. The evolution of the average chip thickness, the variable axial
In the model, the estimated cutting forces are determined from depth of cut and the cutting forces is depicted in Fig. 4. A constant
the average thickness corresponding to the active portion of the range for all these variables is shown, as a consequence of the high
engaged flute. The average chip thickness for the different cutting ratio between the axial depth of cut and the width of cut, typically
situations will depend on the position of the cutting flute (Fig. 3): encountered when milling low-rigidity parts.
The force application point on the active cutting edge is

⎧ 1
⎪ fz ⎡⎣cos(φen ) − cos(φj ) ⎤⎦ φen < φj ≤ φex Zone A
⎪ j φen
φ −
⎪
⎪ 1
h¯ (φ) = ⎨ fz ⎡⎣cos(φen ) − cos(φex ) ⎤⎦ φex < φj ≤ φen + φpr Zone B
⎪ φex − φen
⎪ 1
⎪ fz ⎡⎣cos(φj − φpr ) − cos(φex ) ⎤⎦ φex < φj ≤ φex + φpr Zone C
⎪
⎩ φ − ( φ − φ )
ex j pr (2)

considered to be located in the angle φaver (Eq. (9)).

In order to obtain a proper evaluation of the average chip
thickness, a precise determination of the entry and exit angles of ⎛ h¯ (φ) ⎞
φaver = sin−1⎜⎜ ⎟⎟
the cutting edge is required. In the calculation of the entry and exit ⎝ fz ⎠ (9)
angles, the actual engagement is related to the nominal radial
depth of cut and the part deformation [21], according to Eq. (3). This expression may be approximated in terms of the angles
⎛ which determine the active part of the cutting edge, according to
aea ) ⎞
φena (φ) = π − arccos ⎜1 − ⎟ the equation:
⎝ R ⎠ (3)

where aea is the actual radial depth of cut, defined by Eq. (4).
aea = ae − δ (φ) (4)
φm (φ)
where δ(φ) is the workpiece deviation produced as a consequence
⎧ φj − φen
of the workpiece deformation due to the action of cutting force on ⎪ φj − φen < φj ≤ φex Zone A
Y direction. Both the force and the deviation are calculated by the ⎪ 2
model in an iterative fashion. In this calculation it is considered ⎪
⎪ (φex − φen )
the well known proportional relationship between displacement = ⎨ φex − φex < φj ≤ φen + φpr Zone B
⎪ 2
(deviation) and force according to the static rigidity in Y direction. ⎪
The value of the rigidity has been established experimentally. ⎪ φ − ex (
φ − (φj − φpr ) )
⎪
⎩ ex
φex < φj ≤ φex + φpr Zone C
A calculation of the cutting forces from chip thickness, is 2 (10)
 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14 7

As mentioned above, the cutting forces acting on a cutting edge

in the X-, Y- and Z-directions are defined by Eqs. (11)–(13).
Fx (φ) = Ft (φ)⋅ cos φm + Fr (φ)⋅ sin φm (11)

Fy (φ) = − Ft (φ)⋅ sin φm + Fr (φ)⋅ cos φm (12)

Fz (φ) = Fa (φ) (13)

4. Verification of the reference model

4.1. Determination of the cutting force coefficients

The determination of the cutting force coefficients must be

performed under ideal conditions such as absence of tool or
workpiece deformation and absence of runout. Therefore, the es-
timation of the cutting force coefficients must be done under rigid
workpiece conditions.
Therefore, when machining two pieces, a rigid one and flexible
one, using the same tool and the same cutting conditions, the
coefficients of the specific pressure will be the same. Conse-
quently, any change in the cutting forces will be attributable only
to changes in the stiffness of the part. According to the above, to
ensure the correct estimation of the cutting forces, it is necessary
to carry out an adequate calibration of the cutting force coeffi-
cients for a specific tool-workpiece under rigid workpiece
conditions.
There are several methods for the calibration of cutting force
coefficients. In this paper, the specific cutting pressure is con-
sidered to be a potential function of chip thickness, unlike other
approaches that consider a linear relationship [25]. Owing to the
cutting geometry in finish milling of thin-walled parts with ap»ae,
the accuracy of the estimated cutting force coefficients from re-
gistered cutting forces for one revolution, may be affected. In these
conditions, the average chip thickness in a finish operation is
constant in the middle zone (Zone B Fig. 4) and variable in the
transients at entry and exit (Zones A and C in Fig. 4 respectively).
The conditions under which these variations of the average chip
thickness are produced at the entry and exit of the flute, affect the
accuracy of the estimation, make it necessary to register the cut-
ting forces from several cutting experiments carried out with
different feeds.
The experiments for the verification of the proposed compen-
sation system were performed with two tools and two different
materials. The cutting force coefficients were obtained for the
workpiece material AA2011 with a Izar 4412 cutting tool and for
the workpiece material AA7075 with a Garant 191000 cutting tool.
In the experiments carried out for the calibration of the force
coefficients, all the flutes in the tool were eliminated except for
one in order to avoid the runout effect. Table 1 shows the ma-
chining conditions used in the experimental tests for the calibra-
tion of the force coefficients for a specific combination tool-
workpiece.
The cutting force coefficients were obtained in each case by
interpolating the values from Eqs. (14) and (15). The cutting forces
used for each case correspond to cutting zone B (Fig. 4).
Ft
kt =
A0 (14)

Fig. 9. Comparison of simulation and experimental results. Cutting conditions: Fr

kr =
d ¼8 mm, ap ¼ 8 mm, ae ¼ 1 mm, material: AA2011. Support rigidity: 1300 N/mm. A0 (15)

Fig. 5 shows the specific pressure in accordance with the data

obtained in the tests. In the X and Y directions, the least squares
8 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14
Fig. 10. System to compensate for workpiece deformation.
Fig. 11. Variation of the maximum cutting force as a function of radial depth of cut.
adjustment results in the cutting force coefficients showed in
Table 2.
4.2. Experimental verification of the cutting force model for low-ri-
gidity parts
Taking into account that the estimation of the cutting forces
with the reference model is precise, a suitable adjustment of
workpiece deformation is possible. The reference model proposed
implemented in the system is verified for a rigid workpiece, with
no deformation and for a flexible workpiece, with low rigidity. In
order to verify the correlation between the measured and esti-
mated cutting forces in both conditions, several cutting experi-
ments were carried out for different cutting widths, depths of cut
and workpiece stiffness.
To perform the cutting experiments efficiently and in order to
avoid the machining of different test specimens for each test, an
elastic support is used on which a solid workpiece is mounted
(Fig. 6). The stiffness of this support is equivalent to the stiffness of
the thin-walled workpiece to be machined. Thus, owing to the
successive trajectories of the cutting tool during the experiments,
the machining and preparation of a new test specimen is not re-
quired. On the other hand, it ensures that the equivalent stiffness
remains constant throughout the test.
Fig. 7 shows a diagram of the experimental equipment. An
elastic support is mounted on a Kistler 9257BA table dynam-
ometer, ensuring that the stiffness is equivalent to the thin-walled
part to be machined. Thus, the workpiece displacement due to the
cutting force action, will be of the same order than the deforma-
tion of the thin-walled workpiece. The deformation measurement
is performed by a MTI Microtrak II SA laser triangulation sensor,
with a resolution of 0.25 μm and a frequency response of 20 kHz.
With this sensor the deformation of the assembly is measured and
thus the equivalent deformation of the workpiece.
As a result of the low stiffness of the workpiece, the vibration
level present in the cutting tests is high. In order to facilitate the
comparison of the static components of the rigid model with the
flexible one, signal filtering is performed with a low-pass filter
with a cut-off frequency of 100 Hz. Therefore, the results shown in
the force and deformation diagrams mostly reflect the static
component of the cutting force, which in fact causes the static
deformation.
Fig. 8 shows the experimental results performed in stable
machining where the influence of the workpiece deformation on
cutting forces is noticeable. As shown in these graphs, the ampli-
tude of the cutting forces is lower when using flexible machining
(Fig. 8b) rather than rigid machining (Fig. 8a), according to the
theoretical basis presented in the previous section. For this
workpiece, these results show that, a workpiece deformation of
45 mm corresponds to a reduction of 10–20% on the cutting forces,
depending upon the components.
A comparison between the estimated cutting forces with the
reference model and measured cutting forces is depicted in Fig. 9.
In Fig. 9a, the cutting forces for rigid machining are compared.
Fig. 9b corresponds to the measured and estimated cutting forces
in flexible machining. The resultant deformations are shown in
Fig. 9c. Fig. 9 shows that the results from the reference model
agree with the measurements. It therefore follows that it can be
used reliably to determine changes in cutting conditions when
machining flexible workpieces.
5. Compensation for part deformation in flexible machining
The deformation of the workpiece will result in dimensional
 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14 9

Fig. 12. Compensation system for part deformations: (a) operating diagram and (b) experimental setup.

cutting forces reaches the rigid machining values (nominal cutting

Table 3 conditions). The cutting force calculation for rigid machining is
Cutting conditions for the compensation test. Material Aluminum 7075.
achieved through the reference model. The compensation move-
N d n fz ap ae ment applied on the workpiece, so as to reduce the deformation, is
– mm rpm mm mm mm performed by a piezoelectric actuator.
4 8 800 0.06 8 0.3
5.1. Compensation strategy

errors. In order to avoid these errors, a compensation system for

The in-process calculation of the required compensation is
part deformation was developed. This system is based on the carried out according to the block diagram in Fig. 10. For the im-
variation of the radial depth of cut, ae, until the amplitude of the plementation of this strategy no variations in the workpiece
10 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14

Fig. 13. Simulated results for cutting forces and part deformation for rigid and flexible machining. d ¼ 8 mm, ap ¼ 8 mm, ae ¼ 0.3 mm. Material AA7075. Support rigidity:
900 N/mm.

geometry exist that may affect the axial depth of cut, the radial
depth of cut or the workpiece stiffness.
The strategy starts with the comparison of the measured and
the estimated cutting forces from the reference model. The re-
quired set point to move the piezoelectric actuator is determined
from the reference model, by comparing the measured and esti-
mated cutting forces perpendicular to the feed direction (Fy). The
variation of the maximum cutting force is linearly dependent on
the axial depth of cut. Fig. 11 shows the variation of the maximum
cutting force values with the axial depth of cut for several ma-
chining conditions.
The workpiece deformation results in a decrease of the actual
cutting forces as a consequence of the reduction of the radial
depth of cut. Therefore, the difference between the value of the
actual cutting force and the value of the reference force is used to
determine the radial depth of cut. The difference between the
nominal depth of cut and the actual depth of cut corresponds with
the workpiece deformation to be compensated for by the piezo-
electric actuator.

5.2. Experimental verification of the compensation system

The compensation system proposed is based on a piezoelectric

Fig. 14. Simulated cutting forces in flexible machining with compensation.
d ¼8 mm, ap ¼ 8 mm, ae ¼0.3 mm. Material AA7075. Support rigidity: 900 N/mm.
actuator (1) displacing the table (2) which serves as a support for
the workpiece (3), as shown in Fig. 12. The table should move in a
narrow range with a very high resolution. In order to perform this
movement, a mechanism with flexure hinges is selected. Its dy-
namic response is good because there is no friction or gaps since
there are no mechanical transmission elements in the assembly.
The actuator can generate a force of up to 2000 N and it is pre-
loaded with 300 N. The structure (4), which provides the basis for
 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14
Fig. 15. Cutting forces for flexible machining/compensation. d ¼ 8 mm, ap ¼ 8 mm, ae ¼ 0.3 mm, material: AA7075.
these elements, is mounted directly on a dynamometric platform
(5). For the measurement of the actuator displacement, strain
gauges located within the actuator are used. The deformation of
the workpiece is measured by an external laser sensor (6) of high
frequency and resolution. In a previous work by the authors [26],
this precision positioning system is detailed.
The compensation strategy was implemented using the pro-
gram LabVIEW 2011. The signals measured by the sensors were
digitized using an NI PCI 6251 card, where the analog setpoint for
the piezoelectric displacement was also generated.
The compensation strategy was verified for different machining
conditions, demonstrating the feasibility of correcting the part
deformations on-line. The simulation and experimental verifica-
tion presented below were carried out for the machining condi-
tions shown in Table 3.
First and foremost, in order to validate the system, a simulation
for a workpiece rigidity of 900 N/mm is performed. The calculated
cutting forces with the reference model are shown in Fig. 13a. As
the workpiece is deformed (Fig. 13c), a noticeable reduction on the
cutting forces occur (Fig. 13b).
From the difference between the cutting forces for flexible and
rigid machining, the simulation model was used to determine the
compensation value which will increase the depth of cut, resulting
in an increase on cutting forces. For this case the radial depth of
cut that compensated deformation of flexible part was 0.34 mm.
11
The cutting forces in flexible machining with compensation are
presented in Fig. 14. As shown, it has been possible to establish the
conditions of rigid machining, which means that the part de-
formation has been compensated for.
Once the system has been verified by simulation, the experi-
mental verification is carried out in real time and online. A support
for the workpiece is built with an equivalent stiffness of k ¼900 N/
mm. The experiment starts without the corrective action of the
system (range without correction). Therefore, the cutting forces
shown in Fig. 15b correspond to flexible machining, where the
workpiece (support) is deformed and the cutting forces vary as a
consequence. After 3.7 s the compensation system is externally
activated, starting the compensation strategy. The transition from
one state to another is depicted in Fig. 15c. The system, based on
the reference cutting forces and the measured cutting forces, cal-
culates the reduction of depth of cut due to part deformation. This
reduction is the correction which acts on the piezoelectric ac-
tuator. When the system has stabilized, it is found that the forces
are increased (Fig. 15d). This increase represents the highest ma-
terial removal and compensates for the dimensional error caused
by deformation.
Fig. 16 shows the measured cutting forces for rigid machining.
These forces are substantially similar to those that the system
established as a reference, which are reached once the strategy is
initiated (Fig. 15d). During flexible machining it is also observed
12 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14
Fig. 16. Cutting forces for rigid machining. d¼ 8 mm, ap ¼8 mm, ae ¼0.3 mm.
that the effect of tool runout on cutting forces is reduced com-
pared to rigid machining, which has already been studied in sci-
entific literature by researchers like Sutherland and DeVor [3].
The evolution of deformations and displacements for the same
cutting test are depicted in Fig. 17. These graphs correspond to one
tool trajectory on the workpiece.
Fig. 17 shows the movements of the actuator and of the support
with compensation and without compensation. In the case with-
out compensation, the measured displacement signals are
equivalent to the deviation experienced by the support (the
workpiece). In the case with compensation, the measured
displacement signals encompass the support deviation and the
compensation displacement provided by the actuator. The latter
corresponds to the cutting width variation compensating for the
cutting forces, and therefore corrects the errors associated with
the deformation of the workpiece. In Fig. 17, we see that the sys-
tem compensated for deformations by the actuator performing a
linear displacement of 39 μm amplitude. Owing to this action, the
measured cutting forces in flexible milling are compensated for,
reaching the same amplitude of the nominal forces, as shown in
Fig. 16, which correspond to rigid machining.
The dimensional verification was measured externally using a
probe installed in the machine (Heidenhain TS631). To begin with,
the initial coordinates of the piece are checked at two points in
order to consider possible errors of straightness (Fig. 18). At the
end of the tool path trajectory, the coordinates in the range
without compensation and in the range with compensation are
measured.
From the results presented in Fig. 18, it can be seen that the
actual depth of cut obtained without the system performance is
264 mm, producing an error of 36 mm. Once the compensation
system has performed, the resultant depth of cut is 309 mm and
therefore, the error due to part deformation has been corrected.
The small difference between the deformation value and the set
point value of the piezoelectric actuator may be due to deforma-
tions of the elements in the transmission chain.
6. Conclusions
In this paper a compensation system for static deformations
experienced in machining of low-rigidity parts has been pre-
sented. The strategy used in the compensation system is based on
the indirect calculation of part deformation from the measure-
ment of cutting forces. From the results, the following conclusions
can be drawn:
 The use of a control system based on a reference model that
runs in real time, is a good procedure to compensate for di-
mensional errors due to part deformation in finish milling.
 The reference model used for cutting force estimation is of a
mechanistic type, based on average chip thickness. In this way,
a high execution speed is achieved according to the actual
development of the cutting process, making the iteration along
the cutting flute unnecessary, a method which is commonly
used in research.
 It has been proven experimentally that the estimated forces for
different machining conditions have to be accurate in order to
ensure a good comparison with the actual forces so as to de-
termine part deformation.
 The indirect measurement of part deformation from the change
in the width of cut, which in turn is obtained from the com-
parison between the actual measured cutting force and the
estimated ones, allows online correction of the deformation.
 The experimental verification has been shown to be a good
method of correcting the deformation, thereby improving the
machining of flexible parts. In this way, the direct measure-
ment of the deformation of the workpiece is not required.
 The system has been developed based on the use of a piezo-
electric actuator which modifies the relative position between
the tool and workpiece in order to compensate for the variation
in the width of cut due to part deformation. The system is in-
dependent of the machine tool control system and does not
affect the programming of tool path trajectories.
 The experimental development and the use of a flexible sup-
port have enabled efficient running of cutting tests with good
reproducibility. It has also been shown that the measurement
 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14
Fig. 17. Displacement in flexible machining/compensation. Response signals from the actuator movement and the displacement of the support. d¼ 8 mm, ap ¼ 8 mm,
ae ¼0.3 mm, material: AA7075.
Fig. 18. Dimensional results from the measuring probe.
of the piezoelectric actuator displacements and the part de-
formations using a laser sensor correlates well, although it is
dependent on the geometry of the overall mechanism.
Therefore it is considered that the on-line correction of part
deformations is possible using only the measurement of cutting
forces.
13
Acknowledgments
The authors would like to thank the financial support from
CICYT-Spain (DPI2009-14535) and CONICYT-Chile (FONDECYT
11130721). The authors would also like to express their thanks to
Mr. Ramón San Miguel and Mr. Javier Tena for their technical as-
sistance in running the cutting experiments.
Appendix A. Supplementary material
Supplementary data associated with this article can be found in
the online version at http://dx.doi.org/10.1016/j.ijmachtools.2015.
03.008
References
[1] Hogan B., 2011. Aerospace materials: what will builders use?. Aerospace and
Defense 2011. Supplement to Manufacturing Engineering Magazine. Society of
Manufacturing Engineers.
[2] L. Gang, Study on deformation of titanium thin-walled part in milling process,
J. Mater. Process. Technol. 209 (2007) 2788–2793.
[3] J.W. Sutherland, R.E. DeVor, An improved method for cutting force and surface
error prediction in flexible end milling systems, J. Eng. Ind. 108 (1986)
269–279.
[4] S. Ratchev, E. Govender, S. Nikov, K. Phuah, G. Tsiklos, Force and deflection
modeling in milling of low-rigidity complex parts, J. Mater. Process. Technol.
14 E. Diez et al. / International Journal of Machine Tools & Manufacture 94 (2015) 1–14
143–144 (2010) 796–801.
[5] S. Ratchev, S. Liu, W. Huang, A.A. Becker, An advanced FEA based force induced
error compensation strategy in milling, Int. J. Mach. Tools Manuf. 46 (2006)
542–551.
[6] S. Ratchev, S. Liu, A.A. Becker, Error compensation strategy in milling flexible
thin-wall parts, J. Mater. Process. Technol. 162–163 (2005) 673–681.
[7] T.C. Bera, K.A. Desai, P.V.M. Rao, Error compensation in flexible end milling of
tubular geometries, J. Mater. Process. Technol. 211 (2011) 24–34.
[8] M.-W. Cho, G.-H. Kim, T.-L. Seo, Y.-C. Hong, H. Cheng, Integrated machining
error compensation method using OMM data and modified PNN algorithm,
Int. J. Mach. Tools Manuf. 46 (2006) 1417–1427.
[9] R. Ramesh, M.A. Mannan, A.N. Poo, Error compensation in machine tools — a
review Part I: geometric, cutting-force induced and fixture dependent errors,
Int. J. Mach. Tools Manuf. 40 (2000) 1235–1256.
[10] R. Neugebauer, B. Denkena, K. Wegener, Mechatronic systems for machine
tools, CIRP Ann.-Manuf. Technol. 56 (2007) 657–686.
[11] G. Park, M.T. Bement, D.A. Hartman, R.E. Smith, C.R. Farrar, The use of active
materials for machining processes: a review, Int. J. Mach. Tools Manuf. 47
(2007) 2189–2206.
[12] Electronic resource, Physik instrumente, 〈www.pi.ws〉, (accessed 2014).
[13] Electronic resource, Piezomechanik, 〈www.piezomechanik.com〉, (accessed
2014).
[14] M. Pozzi, T. King, Piezoelectric actuators in micropositioning, Eng. Sci. Educ. J.
10 (2001) 31–36.
[15] A. Woronko, J. Huang, Y. Altintas, Piezoelectric tool actuator for precision
machining on conventional CNC turning centers, Precis. Eng. 27 (2003)
335–345.
[16] Y. Altintas, A. Woronko, A Piezo tool actuator for precision turning of hardened
shafts, CIRP Ann.-Manuf. Technol. 51 (2002) 303–306.
[17] A.T. Elfizy, G.M. Bone, M.A. Elbestawi, Design and control of a dual-stage feed
drive, Int. J. Mach. Tools Manuf. 45 (2005) 153–165.
[18] W. Drossel, V. Wittstock, Adaptive spindle support for improving machining
operations, CIRP Ann.-Manuf. Technol. 57 (2008) 395–398.
[19] R. Neugebauer, W. Drossel, A. Bucht, B. Kranz, K. Pagel, Control design and
experimental validation of an adaptive spindle support for enhanced cutting
processes, CIRP Ann.-Manuf. Technol. 59 (2010) 373–376.
[20] H. Perez, E. Diez, J.J. Marquez, A. Vizan, An enhanced method for cutting force
estimation in peripheral milling, Int. J. Adv. Manuf. Technol. 69 (2013)
1731–1741.
[21] E. Budak, Y. Altintas, Modeling and avoidance of static form errors in per-
ipheral milling of plates, Int. J. Mach. Tools Manuf. 35 (1995) 459–476.
[22] W.A. Kline, R.E. DeVor, The effect of runout on cutting geometry and forces in
end milling, Int. J. Mach. Tool Des. Res. 23 (1983) 123–140.
[23] M. Wan, W.H. Zhang, Calculations of chip thickness and cutting forces in
flexible end milling, Int. J. Adv. Manuf. Technol. 29 (2006) 637–647.
[24] M. Wan, W. Zhang, Systematic study on cutting force modeling methods for
peripheral milling, Int. J. Mach. Tools Manuf. 49 (2009) 424–432.
[25] M. Wang, L. Gao, Y. Zheng, An examination of the fundamental mechanics of
cutting force coefficients, Int. J. Mach. Tools Manuf. 78 (2014) 1–7.
[26] E. Diez Cifuentes, H. Pérez García, M. Guzmán Villaseñor, A. Vizán Idoipe,
Dynamic analysis of runout correction in milling, Int. J. Mach. Tools Manuf. 50
(2010) 709–717.