Dynamic analysis of runout correction in milling

E. Diez Cimentes3'*, H. Pérez García5, M. Guzmán Villaseñor3, A. Vizán Idoipe'

a
b
Departmentof Mechanical Engineering, University of La Frontera, Temuco, Chile
Departmentof Mechanical Engineering, University of León, 2407Í León, Spain
c
Department of Mechanical and Manufacturing Engineering, Technical University of Madrid, 28006 Madrid, Spain

ABSTRACT

Tool runout and its effects is an important area of research within modelling, simulation, and control of
milling forces. Tool runout causes tool cutting edges to experience uneven forces during milling. This
fact also affects tool life and deteriorates workpiece surface quality. In this article a procedure, in order
to diminish the effects of tool runout, is presented. The procedure is based on chip thickness
modification by means of the fast correction of the tool feed rate. Dynamic feed rate modification is
provided by superposing our own design of a fast feed system driven by a piezoelectric actuator to the
Keywords: conventional feed drive of the CNC machine tool. Previously, a model of the dynamic behaviour of the
Milling system was developed to analyze the influence of fast feed rate modification on cutting forces.
Tool runout The model incorporates the piezoelectric actuator response as well as the structural dynamics of the
Cutting force control tool and the designed Fast Feed Drive System (FFDS). Simulated and experimental results presented in
Piezoelectric actuator
this paper show the effectiveness and benefits of this new tool runout correction procedure.

1. Introduction on the identification of radial runout on each cutting edge of the

milling cutter.
Milling is one of the most extended cutting processes due to its Tool runout in milling processes can come from many sources.
large field of application within automotive, aerospace, mould and According to Bao and Tansel [7], tool runout mainly depends on
die, and other industries. For more than one hundred years, spindle and toolholder characteristics. On the other hand, Schmitz
modelling, simulation and control of cutting forces have been et al. [8] assigns more importance to errors depending on the
important research issues for researchers with the objective of radial position of the teeth on the cutter. According to Hekman
improving aspects like workpiece surface quality, tool wear and Liang [9], precise measurement or control of the static and
resistance or process stability. Nevertheless, there still exist dynamic tool runout is rather complex. While static runout may
interesting aspects in milling that need to be explored in depth come from spindle errors, thermal deformation, insert settings
to achieve a better understanding of the process. and tool dimensional errors, dynamic tool runout may arise from
Unlike other machining processes, milling is affected by other sources such as cutting force variation, spindle and tool
several variables that make its modelling especially complex. imbalance and non-uniform progression of tool wear. Important
Tool runout (radial and axial), uneven tooth positions, deforma- effects of tool runout in milling are the decrease of the milled
tions and tool wear are characteristics affecting milling operations surface quality and the increase in the radial and axial force
[1]. Among these characteristics, radial tool runout in end milling variation [5], which may lead to shorter tool life under uneven
has received special attention from researchers, starting with the wear conditions for each cutting edge.
work published by Kline and DeVor [2], which focused on the More recently, research has been focused on the incorporation
effects of radial tool runout on cutting forces in end milling. Other of tool runout in the analysis of the dynamic stability of the
researchers have studied how tool runout affects cutting process, the quality of machined surfaces and the milling of
geometry in end milling [3] proposing equations to calculate curved surfaces. Insperger et al. [10] and Wan et al. [11] showed
chip thickness in the case of tool runout. The identification of that the predicted stability diagram in absence of tool runout
simple eccentricity, produced when tool has a different rotation differs considerably from the diagram obtained when tool runout
axis than the geometric axis, was studied by Liang and Wang, and is considered. These investigations also claim that the mathema-
Wang and Zheng [4,5] while Seethaler and Yellowley [6] focused tical background necessary to address the solution of regenerative
milling equations is more complex when tool runout is con-
sidered. Arizmendi et al. [12] proposed a model for the prediction
* Corresponding author. of the surface topography machined by peripheral milling in
E-mail address: ediez@etsii.upm.es (E. Diez Cifuentes). presence of tool runout. This research concluded that tool runout
 Nomenclature K0 Constant gain.
[K\ Stiffness coefficients matrix of the system.
ap Axial d e p t h of cut. mtabie^ ctabie^ ¡¿able p eec j direction modal mass, damping and
a0 Coefficient of the zero order term in the actuator stiffness coefficients of the worktable.
transfer function, m" 0 ', 4 00 ', /400' X-axis modal mass, damping and stiffness
eh Coefficient of the first order term in the actuator coefficients of the cutting tool.
transfer function. my00l} cy00l} ky001 Y-axis modal mass, damping and stiffness
a2 Coefficient of the second order term in the actuator coefficients of the cutting tool.
transfer function. [M] Mass matrix of the system.
[C] Damping coefficients matrix of the system. N Spindle speed.
D Nominal tool diameter. R Nominal tool radius.
ft Nominal feed per tooth. Rj Tool radius at tooth j .
Fsj Tangential, radial and axial cutting forces, for s = t,r,a. 7\, T2 Time constants.
Fs X, Y and Z c o m p o n e n t s of cutting forces, for s=X,Y,Z. Xcom Position command to the actuator.
{F} Cutting force vector of the system. xpa Position response of the actuator.
x
{F} Vector that represents piezoactuator action on the tabie> Xmbiet xtabie X-axis displacement, velocity and acceleration
worktable. of the table,
Gpa(s) Transfer function of the actuator-drive-control sys- "tool. Xtoou Xtooi X-axis displacement, velocity and acceleration
tem. of the tool.
hj(<p) Uncut chip thickness without dynamic effects for y toou y toob y tool Y-axis displacement, velocity and acceleration
tooth j . of the tool.
\iro Chip thickness deviation due to cutter radial runout {X},{X},{X} Position, velocity and acceleration vectors describ-
for cutting edge j . ing the movement of the system.
hd(f) Uncut chip thickness including dynamic effects. <p Cutter rotation angle.
ks Tangential, radial and axial cutting force coefficients, <pj" Entry angle of cutting edge j .
for s=t,r,a. <pjX Exit angle of cutting edge j .
feto,m Cutting parameters depending on workpiece material Pi Runout of cutting edge j .
and cutting tool for tangential direction. T(t) Time dependent delay of the milling process.

and spindle tilting have a strong influence on the heterogeneity correction. This correction is carried out by means of the use of
bands produced on the milled surface. On the other hand, Desai piezoelectric actuators working in parallel with the feed drive
et al. [13] showed that when milling curved surfaces, tool runout system of a conventional CNC machine tool. The presented
plays an important role because it changes the interaction analysis emphasizes the study of tool and feed system dynamics,
between the trajectories of the cutting teeth. and how these elements affect the behaviour of force control.
One way to reduce the unwanted effects on milling processes
would be through the use of systems allowing the correction of the
cause for tool runout. In this sense, some papers have been 2. Cutting force control in milling tools with runout
published dealing with the modification of uncut chip thickness to
improve milling processes. Liang and Perry [14] proposed chip load Tool runout causes chip thickness to vary over the tool rotation
compensation for the elimination of cutting force oscillations due and therefore the tool experiences uneven forces during each
to cutter runout. This correction was based on the control of the cutting tooth period. The objective of runout correction is to
spindle frequency force component, since this component includes diminish this effect by means of feed rate modification, making
all the effects of runout on cutting forces, as various researchers each cutting edge carry the same chip load.
have pointed out in previous articles [4,5]. This method is very In this study, feed rate modification is carried out using a Fast
effective, but the control time may be a limit at higher spindle Feed Drive System (FFDS) driven by a piezoelectric actuator. The
speeds. This limitation could be solved if the evolution of milling FFDS is superposed on the conventional machine tool feed drive
forces, when the piezoelectric actuator is varying the chip load, is system. This combined feed drive system scheme is similar to that
better understood. As a result, with an adequate command signal presented by Elfizi et al. [17] and Perez et al. [18]. It is well known
to the piezoelectric actuator, it is possible to deal with higher that piezoelectric based FFDSs have higher dynamic response and
spindle speeds. Another approach to compensate radial runout in better resolution than conventional feed drive systems. However,
face milling was presented by Sastry et al. [15]. The compensation its application on machine tools has been restricted due to its
was based on the continuous variation of the spindle speed in a limited range for linear displacements. Combined feed drive
sinusoidal pattern. It was shown that this methodology reduced systems allow micrometric changes in the working table position
the effect of tool runout on the cutting force pattern considerably at high frequency over a large field of displacement.
and the surface profile was improved. However, using a sinusoidal The controlled movement of the FFDS to modify feed rate is
pattern to vary spindle speed would be the solution when runout determined directly from measured forces based on the relation-
comes only from tool eccentricity. In the field of simulation, a ship between uncut chip thickness and cutting forces according to
dynamic model in order to simulate a process controller for end the cutting force model presented in the next section of this article.
milling was presented by Kolartis and DeVries [16]. The model is
able to predict cutting forces and surface profile error under 2.2. The cutting force model
changes of feed rate and other cutting conditions.
This article focuses on the procedures developed to diminish Several papers have been published dealing with cutting force
the runout effect on milling processes based on dynamic modelling including the effects of tool runout [2,7,16,19-21]
where other issues that may affect the simulation process have dimensional errors of the cutting edges and spindle tilting. Due
also been considered such as the real tool trajectory, the use of to the difficulty of measuring tool runout precisely, in this paper
microtools and the modification of the exit and entry angles. The \iro is calculated indirectly from the measured cutting forces
main contribution of the present paper is to analyze the effect of acting on each flute.
including a piezoactuator on cutting force control. In order to Since tool runout is significantly affected by the eccentricity of
simulate the cutting forces in a milling process, a mechanistic tool-tool holder-spindle assembly, researchers usually relate tool
cutting force model was employed. Cutting forces were calculated runout to eccentricity, using for the eccentricity definition two
based on the cutting geometry engaged into the workpiece and parameters, its magnitude and its angular position referring to a
the empirical equations that relate the cutting forces with the reference flute. In this case, the effective radius varies along the
uncut chip thickness. The trajectory of the cutting tool was cutting edge. In addition, this definition excludes other possible
approximated by a circular path, thus the chip thickness including causes for error.
tool runout can be calculated as In this study, where small values of depth of cut and helix
angles were considered, the variation of the effective radius along
hj(<p)=ftsm(<p) + h{0 0) the cutting edge may be neglected and then the tool runout can be
adequately defined by means of the difference between the actual
where ft^is the variation of the chip thickness due to runout in
and the nominal trajectory for each flute.
cutting edge j , and can be calculated as follows:
Considering the specific cutting force as a potential function of
hi ••Pi-Pi-i (2) chip thickness, tangential, radial and axial cutting forces along the
cutting edge can be calculated using Eq. (3).
The relationship between variables involved in calculating the
static chip thickness for a tool considering runout can be seen in Fsj((f>) = ks(hj)h0)ap, s = t,r,a (3)
Fig. 1. The runout of cutting edge, p¡, represents the variation of Cutting force coefficients in tangential, radial and axial
the effective radius of each flute which results in a variation directions are expressed as a potential function of the instanta-
on cutting forces. The effective radius is the radius of the neous chip thickness. For tangential direction
actual trajectory of the flute, which differs from its nominal
kt(hj) = k[0(hjrm (4)
radius due to runout. This parameter takes into account errors
coming from different sources such as tool eccentricity, Thus for tangential direction, Eq. (3) can be written as
Ft,0) = kto(h}y-map (5)
In a similar way, equations can be obtained for the other two
directions.
Tangential, radial and axial forces are projected in X, Y and Z
directions, according to the following transformation:
Xj#)l ' -cos(<p) -sin(^) 0" \Ftj(.<f>)
Vj#) = sin(<p) -cos(<p) 0 FrM (6)
\0) 0 0 1 Faj(.<f>)

The result of the contribution of all cutting edges in the total

force acting on the tool in X, Y and Z axes is given in Eq. 7.
N
Fs(<P)-- =x,y,z (7)

2.2. The dynamic model and force control in milling

A regenerative milling model was used to describe the

dynamic behaviour of the milling process [22-28]. A dynamic
milling model is proposed that includes a system with two
degrees of freedom for the tool and a system with one degree of
freedom for the FFDS, which coincides with feed direction. The
effect of piezoelectric actuator movement on the milling forces is
calculated by means of the regenerative cutting force model,
where the uncut chip thickness depends on tool and workpiece
vibration, which is caused mainly by the controlled oscillations of
the piezoelectric actuator. Fig. 2 shows the FFDS designed to
correct tool runout effects and the associated model including
actuator system dynamics. From this model, the force diagrams
and the signals acting on each element were determined in order
to obtain the equations that model the problem. These diagrams
can be seen in Fig. 3.
Feed The resulting equations that model the system are shown in
°M °J direction Eq.(8)
[M]{X} + [C]{X} + [K]{X} = {F} + {F} (8)
Fig. 1. Geometry for the cutting forces calculation: (a) the smaller effective radius
is cutting and (b) the bigger effective radius is cutting. Segmented line represents where [M], [C], y [K\ are diagonal matrices, {F}is the cutting
tool trajectory without runout. force vector and {F} represents the action of the piezoelectric
 9 x
tool
^ V pif,^" [C] =
y tool
; m = { table
x
(10)
x
pa
a.

x
tool
1
[K] =
kf y mol
uable ;{X}={ x (H)
table

a0 •pa

CM
{F}={ (12)
Fx
0

(0 i° ] i°0
0 0
{F} = [Q{ \+W\ >+•0 (13)
Xpa Xpa
o lo J , Xcom

Fx and Fy are defined by Eq. (7). The dynamic structural

parameters of the cutting tools and the FFDS were determined
Fig. 2. (a) Fast feed drive system designed to control tool runout and (b) the by experimental modal analysis. The dynamic parameters of the
corresponding dynamic model. response of the piezoelectric actuator, the drive and controller
were obtained by means of a procedure for system identification.
These procedures are described in the next section of this article.
Considering tool flexibility and the feed rate variation due to the
piezoelectric actuator, regenerative action causes Eq. (1) to be
kn ,oolv c'°°'y,
V Jit dependent on tool and workpiece vibration. This situation is
modelled using delayed differential equations. If an ideal case is
presented where the tool has equally spaced flutes and the runout
is negligible, the delay present in the differential equations will be
constant. However, when tool runout is considered, the delay
becomes time dependent, due to the small variation produced at
the entry and exit angles for each cutting edge [10,11,29,30]. In
addition, it was considered that tool and workpiece vibration
levels do not affect the delay. Information related to solving these
equations can be found in Ref. [31]. In the present paper the
following aspects were taken into account relating to this matter:

C (X!abk X
pJ
• All simulations and experiments were carried out using 2 flute/
insert milling tools. The model is oriented to small diameter
tools where two flutes is more common. For the case of small
*T*«*-JU width of cut, the number of flutes simultaneously cutting is
less than the number of cutting edges of the tool.
• Both static and dynamic runout were lesser than the nominal
m' feed rate per tooth.
• The variation of entry and exit angles due to runout was
considered in simulation.
1 y
pa " com
C (si ~
re* ' a s2+ a s + a Previous assumptions allow us to be sure that there will be a
2 1 0
maximum of two delays in each tool revolution. In addition, based
Fig. 3. Forces diagram acting on (a) the tool and (b) the table, and (c) diagram of on these assumptions, it can be observed that for the main part of
signals acting on the system actuator-drive-control. the cutting zone, the delay that defines the regenerative effect of
the milling operation will be equal to half of the tool rotation
period. Moreover, at the entry and exit of the cutting edge with a
actuator on t h e table bigger effective radius, there will be a tiny cutting zone where the
delay will be equal to the cutting tool rotation period. This means
that for the flute/insert with the bigger effective radius there will
be zones at entry and exit where the cutting edge will remove
y mot
[M] = {X\=< x
(9) material left by the cutting edge itself during the previous tool
table
revolution (delay equal to the tool rotation period) but there will
a2 pa be another zone where the cutting edge will remove material left
by the previous flute/insert (delay equal to half of the tool rotation fitted frequency response functions are shown in Fig. 4, while the
period). modal parameters extracted are shown in Table 1.
The inclusion of all effects previously mentioned causes chip
thickness in Eq. 1 to be time dependent. Then the expression for 2.3.2. Piezoelectric actuator, drive and position control loop
the chip thickness, which includes tool runout and dynamic The piezoelectric actuator employed to drive the FFDS was a
effects is as shown in Eq. 14. P-212.4S from Physik Instrumente. This model is able to provide
hd(t) = \ft-xtooi(t)+xt00,(t-z(t))-xtable(t)+xtable(t-x(t))]sm[<p(t)] a push/pull force of 2000/300 N and a maximum displacement
+ [-yiool(t) +y t ool(t-T(t))]COS[0(t)] + hro(t) (14) of 60 um. The static stiffness of the actuator is 34 N/um. The
dynamic performance of the actuator is strongly influenced by
both the available power of the drive used and the control
strategy employed to control its position [32]. The drive used to
2.3. Characterization of the milling force control system feed the actuator was the E-471 high power amplifier from Physik
Instrumente. In order to control its position, the actuator was
In this part of the article, the characterization of the proposed operated in closed loop mode. In this way a higher dynamic
system is presented. Firstly, some design details of the FFDS and stiffness is achieved as well as a better dynamic response. At the
the experimental determination of its dynamic parameters are same time, hysteresis is minimized and can be overlooked in the
pointed out. Secondly, technical data of the piezoelectric actuator movement range used for the purposes for this study. Although
used are presented and the system identification procedure for piezoelectric actuator system behaviour is complex and relates
the piezoelectric actuator, the drive and the controller is voltage/current inputs with position/force outputs, some simpli-
explained. Finally, the identification of tool characteristics is fications may be assumed depending on the actuator setup
described. (preload) and the operating conditions. In this case, the actuator is
operated at a load level significantly lower than the maximum
load level that it can bear, hence a linearity condition between the
2.3. Í. The fast feed drive system command signal and the actuator response can be supposed
The FFDS designed to carry out runout correction is made up of regardless of the dynamic forces acting on it.
a worktable, a pair of flexible couplings, a flexible structure, and a The parameters governing the dynamic behaviour of the
piezoelectric actuator which provides the rapid movement to the actuator were estimated from the experimental identification of
table, as is shown in Fig. 2. In order to obtain an accurate the transfer function of the actuator, its drive and its position
performance of the FFDS, some important issues were taken into control loop. The input signal considered for the identification
account in the design of the mechanical system, as described process was a position command, a step signal with 6 urn in
below: amplitude. The output of the system was the position of the
actuator measured with its own strain gauge system. The transfer
• The use of flexure joints to eliminate friction and to minimize function of the system was fitted to a transfer function with two
backlash. real poles. The general form of the transfer function for this
• High stiffness of the base structure to avoid undesirable system is shown in Eq. 15.
deformations.
• The use of appropriate flexible couplings to avoid tilting and Gpa(s) --
-XDI K0 (15)
shearing forces, which may damage the actuator. (1-T,s)(l-T 2 s)
Xcom(S)
The estimated parameters characterizing the dynamic beha-
Simulations based on the Finite Element Method, which are not viour of the actuator, its drive and controller are shown in Table 2.
presented here, were performed in order to check the mechanical From the data presented in Table 2, the coefficients a0, a^ and
behaviour of the system designed. The FFDS structure, including a2 modelling the dynamics of the actuator (shown in Fig. 3c) were
worktable, couplings and the actuator structure, was modelled as calculated and the obtained result is
a one degree of freedom system. The structural dynamic response 1
was determined by means of experimental modal analysis. The a
°=K
system response was measured with a modal accelerometer
attached to the worktable of the FFDS. Both the experimental and T,+T2
a-i (16)

140
Table 1
120 Modal parameters adjusted for the FFDS structure response.
z
"« 100 [kg] ' [N/m] é°"1' [Ns/m]
80
i Experimental
0.194 1.7E7 77.3

60
Simulation
40

20

0
500
^J W 1000 1500 2000
Table 2
Estimated parameters modelling the dynamic behaviour of the actuator, its drive
and controller.

Ko Ti
Frequency [Hz]
0.99536 0.0011303 0.0011376
Fig. 4. Amplitude of the FFDS frequency response function in the feed direction.
 T,T2 3. Experimental and simulation results
a2: K
In order to determine milling parameters, a procedure similar
to that presented in [33] was followed. Kt and kr were considered
to be potential functions of chip thickness along the cutting zone.
2.3.3. Cutting tool The material used in all the cutting tests was AL-7040. The
The dynamic response of the tool can be studied when the potential fitted functions for specific cutting forces are shown
frequency response function, measured at the tool tip, is known. in Fig. 6 (for 8 mm diameter, ft=0.05 mm/tooth, ap = \ mm,
In this study, frequency response function was obtained by means N=1200rpm).
of experimental modal analysis. The excitation was achieved The control strategy was implemented in LabVIEW 8.2.
using a modal impact hammer, while the response was measured Command signal generation, control and data acquisition tasks
with a small size accelerometer. Force and acceleration signals were carried out using the NI PCI 6251 high speed data acquisition
were measured using a data acquisition system and later device. Fig. 7 shows the layout of the experimental setup. The
processed with our own modal analysis software to obtain the command signal to the actuator xcom was synchronized with tool
frequency response functions and the modal parameters of the rotation. In this way it is possible to control the entry of each flute
predominant mode. As shown in Fig. 5, frequency response into the cutting zone. The reference pulse for the control was
functions for each axis of the tool are very similar, thus the same generated in each tool revolution and it was measured using an
fitting was used for both directions. The adjusted values for the optic sensor. From the reference pulse, a controllable delay pulse
natural frequency, around 900 Hz, are shown in Table 3. was generated, which coincided with the entry in the cutting
zone. The command signal xcom had the same frequency as tool
rotation and its amplitude was determined by the difference
between the cutting forces by estimating \iro for each flute. The
FFDS was placed on a dynamometric table to measure cutting
forces on X, Y and Z axes. Additionally, the table position was
measured directly by using a laser interferometer to verify table
movement xtable.
In order to evaluate the effectiveness of runout control, several
milling tests were conducted with various tools ranging from
25 to 8 mm in diameter. Fig. 8a shows measured and simulated
cutting forces for the 25 mm tool used in the study. The
amplitude variation in Fy for each cutting edge, due to tool
runout, can be seen in both simulation results and experimental
measurements. Fig. 8b shows experimental and simulation
results when the runout control strategy was operating. The
control action can be seen to modify the cutting force pattern,
500 1000 1500 2000 causing the Fy amplitude to become the same for each cutting
Frequency [Hz]
edge. According to experimental results obtained in this research
Fig. 5. Magnitude of the frequency response functions at tool tip for X and Y axes. and other works cited previously, runout in large diameter
milling tools might not be an important issue because its
magnitude is much lower than feed per tooth. However, runout
is a serious drawback for milling tools with a smaller diameter
Table 3 because tool fabrication errors and clamping errors are more
Dynamic parameters fitted for one of the tools used in the experiments. significant in this case.
D—25 mm.
The effectiveness of runout control and its effect on cutting
Diameter [mm] LjD Type Inserts m[kg] k [N/m] c [Ns/m] forces was evaluated using an 8 mm tool. In this case the static
tool runout was measured in the direction of feed rate, reaching a
25 3.8 Insert 2 0.3641 9.67E+06 244
value of 20 um in diameter. Fig. 9a shows the measured and
simulated cutting forces for the tool with runout. A difference in
amplitude of 38% between the cutting forces for each flute, due to
runout, can be observed. The corrected cutting forces can be seen
3000 in Fig. 9b. It can be observed that, despite the large amount of
runout, the control is able to correct the undesirable effect of
2500 runout on cutting forces.
Fig. 10 is useful in evaluating the FFDS performance to control
1/V\ k, = 294.2 h-°M
2000 runout effects in milling. Experimental (Fig. 10b) and simulation
\ ^^A R2 • 0 9462
(Fig. 10a) results are presented for the command position signal,
1500 actuator response, and worktable position measured with a laser
sensor. Despite the influence of the cutting forces disturbing the
k = 63.21 h "°,e1 >S»A
1000 dynamic system formed by the FFDS, this effect is almost
R2 = 0.9764 ^^&l
unnoticeable on the control performance, as shown in Fig. 9.
500 In Fig. 10, the required command signal to correct tool runout
was 30 um in amplitude. On the other hand, the fitted value of tfro
0 obtained from measured cutting forces was 26 um. Accordingly,
0.02 0.04 0.06
Chip thickness [mm]
the prediction of tool runout was precise enough to allow good
cutting force estimation. In addition, if these values of tool runout
Fig. 6. Fitted values for kt and kr. are compared to measured static tool runout, a larger difference is
 FAST FEED DRIVE SYSTEM (FFDS)

PC BASED
CONTROL

3Sk
LabVIEW 8.2
4 ANALOG 2 DIGITAL
JUUL

CONVENTIONAL FEED DRIVE SYSTEM

Fig. 7. Experimental setup.

observed. This difference, between static and dynamic tool experiments were carried out at a spindle speed of 1200 rpm.
runout, can be attributed to the non-inclusion, in static measure- The use of an elastic structure to provide table displacement
ments, of several aspects affecting tool runout, such as uneven and a command signal at the spindle rotation frequency allowed
tooth spacing or dynamic effects. The same observation was the control at higher frequencies rather than using other
reported by Wang and Zheng [5]. approaches.
In Fig. 10 both the actuator and the FFDS follow the command
signal xcom showing deviations lower than 10% for simulation, and
even smaller for experimental results. For slot milling experi-
ments, the differences between the command signal and table 4. Conclusions
position were greater at the entry and exit of the tool. The
influence of these deviations on control effectiveness is not In this article, a cutting force control in order to correct tool
significant because chip thickness is relatively small at the entry runout has been presented. The main conclusions of this work can
and exit angles. Simulation and experimental results show that be summarized as follows:
for the frequency range considered in this research, a second
order model fits the piezoelectric actuator and its dynamic 1. The cutting force model with fast feed rate modification:
response adequately. A cutting force model to investigate the influence of feed
Regarding the control procedure, the spindle speed limit was rate modification on cutting forces in milling, using delayed
imposed mainly by the dynamic response of the piezoelectric differential equations, was developed. Mechanistic cutting
actuator and its position control loop. In this article, the force modelling was used to predict cutting forces considering
 100 100

£• o

-100 -100
0.05 0.1 0.15 0.2 0.25 0.05 0.1 0.15 0.2 0.25
Time [s] Time [s]
100 100

£• °

•100
0.05 0.1 0.15 0.2 0.25 -100
0.05 0.1 0.15 0.2 0.25
Time [s]
Time [s]
Cutting conditions: Cutting conditions:
D=25 mm ft=0.05 mm/tooth D=8mm ft=0.05 mm/tooth
N=1200 rpm h r o =4/.jm N=1200 rpm h ro =26 //m
8n = 1 mm ap=1 mm

Fig. 8. Cutting forces for the 25 mm tool: (a) without correction and (b) with Fig. 9. Cutting forces for the 8 mm tool: (a) without correction and (b) with
correction. correction.

the presence of tool runout. Tool runout, calculated indirectly for each flute. In this sense, experimental results performed
from cutting forces, is characterized by means of a single with 3 mm tools, which are not shown in this paper,
parameter which represents the difference between the revealed that the application of this correction procedure is
chip thicknesses on each tool flute. This parameter was found possible.
to be effective in the correction of tool runout when reduced Experimental verification: In order to vary the fast feed rate,
depths of cut and small helix angles are used. The dynamic a feed drive system with a time response, comparable to the
character of the process was taken into account by consider- tool rotation frequency, was needed. Accordingly, a fast feed
ing the dynamic response of both the tool and the system system driven by a piezoelectric actuator was designed and
designed to modify the feed rate. Simulating the process, the constructed. Although the piezoelectric actuator has a high
controlled feed rate variation was found not to produce a dynamic response, its load capacity is restricted. This limita-
decline in the efficiency of the process and that the tool tion restricts the use of these kinds of actuators.
responded correctly to fast changes in feed rate. Simulation The required feed rate modifications involve micrometric
permitted, prior to FFDS construction, the possible correction actuator displacements, thus it is necessary to design a feed
of cutting forces through feed rate variation proportional to drive system with micrometric resolution. This makes the use
tool runout. of an elastic-type transmission, to provide the modification of
Force correction methodology: The aim of this force control is fast tool feed rate, necessary.
to vary cutting forces so that each flute cuts the same amount It was shown that the designed and constructed FFDS is able to
of material. This strategy is based on feed rate variation in correct the effect of tool runout on cutting forces, as the
order to modify chip thickness. Using numerical simulation, measured cutting forces in Section 3 of the paper show. By
the results proved to be very good. The dynamic character- means of actuator displacement and worktable displace-
istics of the tool, stiffness and damping, do not affect the ment measurements, it was shown that deviations from the
cutting capacity of each flute negatively. For tools of low command signal to the actuator were higher at the entry
stiffness, it is possible not to reach the desired correction and exit angles, where the influence on cutting forces is
of chip thickness in order to obtain the same cutting capacity smaller.
 [6 R.J. Seethaler, I. Yellowley, The identification of radial runout in milling
operations, Transactions of the ASME Journal of Manufacturing Science and

nnn r
Engineering 121 (1999) 524-531.
30 W.Y. Bao, I.N. Tansel, Modeling micro-end-milling operations. Part II: tool
run-out, International Journal of Machine Tools and Manufacture 40 (2000)
2175-2192.
2 T.L. Schmitz, J. Couey, E. Marsh, N. Mauntler, D. Hughes, Runout effects in
.2 20
E. milling: surface finish, surface location error, and stability, International
Journal of Machine Tools and Manufacture 47 (2007) 841-851.
O K.A. Hekman, S.Y. Liang, In-process monitoring of end milling cutter runout,
& 10 Mechatronics 7 (1997) 1-10.

y U UU
V)
o T. Insperger, B.P. Mann, T. Surmann, G. Stépán, On the chatter frequencies of
D. milling processes with runout, International Journal of Machine Tools and
Manufacture 48 (2008) 1081-1089.
M. Wan, W.H. Zhang, J.W. Dang, Y. Yang, A unified stability prediction
method for milling process with multiple delays, International Journal of
0.05 0.1 0.15 0.2 0,25 Machine Tools and Manufacture 50 (2010) 29-41.
M. Arizmendi, J. Fernández, A. Gil, F. Veiga, Model for the prediction of
Time [s] heterogeneity bands in the topography of surfaces machined by peripheral
milling considering tool runout, International Journal of Machine Tools and
Manufacture 50 (2010) 51-64.
K.A. Desai, P.K. Agarwal, P.V.M. Rao, Process geometry modeling with cutter
runout for milling of curved surfaces, International Journal of Machine Tools
and Manufacture 49 (2009) 1015-1028.
S.Y. Liang, S.A. Perry, In-process compensation for milling cutter runout via
chip load manipulation, Transactions of the ASME Journal of Engineering for
Industry 116 (1994) 153-160.
S. Sastry, S.G. Kapoor, R.E. DeVor, Compensation of progressive radial run-out
in face-milling by spindle speed variation, International Journal of Machine
Tools and Manufacture 40 (2000) 1121-1139.
F.M. Kolartis, W.R. DeVries, A mechanistic dynamic model of end milling for
process controller simulation, Transactions of the ASME Journal of Engineer-
ing for Industry 113 (1991) 176-183.
A.T. Elfizy, G.M. Bone, M.A. Elbestawi, Design and control of a dual-stage feed
drive, International Journal of Machine Tools and Manufacture 45 (2005)
153-165.
0.05 0.1 0.15 0.2 0.25 H. Pérez, J. Rios, E. Diez, A. Vizán, Increase of material removal rate in
Time [s] peripheral milling by varying feed rate, Journal of Materials Processing
Technology 201 (2008) 486-490.
Cutting conditions: M. Wan, W.H. Zhang, Systematic study on cutting force modelling methods
for peripheral milling, International Journal of Machine Tools and Manu-
D=8 mm f t =0.05 mm/tooth facture 49 (2009) 424-432.
H.Z. Li, X.P. Li, A numerical study of the effects of cutter runout on milling
N=1200rpm hro=26/jm process geometry based on true tooth trajectory, The International Journal of
an = 1 mm Advanced Manufacturing Technology 25 (2004) 435-443.
E.J. Armarego, N.P. Deshpande, Computerized end-milling force predictions
Fig. 10. Worktable performance: (a) simulation and (b) experiment. with cutting models allowing for eccentricity and cutter deflections, CIRP
Annals—Manufacturing Technology 40 (1991) 25-29.
S. Tobias, Vibraciones en máquinas-herramientas, Ediciones URMO, Bilbao,
1972.
Acknowledgements S. Smith, J. Tlusty, Efficient simulation programs for chatter in milling, CIRP
Annals—Manufacturing Technology 42 (1993) 463-466.
Y. Altinta?, in: Manufacturing automation: Metal cutting mechanics, Machine
The authors would like to thank the financial support received Tool Vibrations, and CNC, Cambridge University Press, Cambridge, 2000.
from the Spanish Ministry of Education and Science to this Y. Altinta?, E. Budak, Analytical prediction of stability lobes in milling, CIRP
Annals—Manufacturing Technology 44 (1995) 357-362.
research project, CICYT DPI 2005-08932. Mr. Diez is supported by T. Insperger, J. Gradisek, M. Kalveram, G. Stépán, K. Weinert, E. Govekar,
a Doctoral Scholarship from the Chilean Republic and a Doctoral Machine tool chatter and surface quality in milling processes, in: Proceedings
Scholarship from Fundación Carolina in Spain. of IMECE'04, November 13-19, 2004.
J. Gradisek, M. Kalveram, T. Insperger, K. Weinert, G. Stépán, E. Govekar,
I. Grabec, On stability prediction for milling, International Journal of Machine
References Tools and Manufacture 45 (2005) 769-781.
E. Budak, Analytical models for high performance milling. Part II: process
dynamics and stability, International Journal of Machine Tools and Manu-
[1] S. Spiewak, An improved model of the chip thickness in milling, CIRP facture 46 (2006)1489-1499.
Annals—Manufacturing Technology 44 (1995) 39-42. A.M. Gouskov, S.A. Voronov, H. Paris, S.A. Batzer, Nonlinear dynamics of a
[2] W.A. Kline, R.E. DeVor, The effect of runout on cutting geometry and forces in machining system with two interdependent delays, Communications in
end milling, International Journal of Machine Tool Design and Research 23 Nonlinear Science and Numerical Simulation 7 (2002) 207-221.
(1983) 123-140. X. Long, B. Balachandran, B.P. Mann, Dynamics of milling processes with
[3] J.-J. Wang, S.Y. Liang, Chip load kinematics in milling with radial cutter variable time delays, Nonlinear Dynamics 47 (2007) 49-63.
runout, Transactions of the ASME Journal of Engineering for Industry 118 L.F. Shampine, Solving delay differential equations with dde23, 2002.
(1996) 111-116. Available from: <http://www.mathworks.com/matlabcentral/fileexchange/
[4] S.Y. Liang, J.J. Wang, Milling force convolution modeling for identification of 3899 >.
cutter axis offset, International Journal of Machine Tools and Manufacture 34 M. Pozzi, T. King, Piezoelectric actuators in micropositioning, The IEE
(1994) 1177-1190. Engineering Science and Educational Journal 10 (2001) 31-36.
[5] J.-J.J. Wang, CM. Zheng, Identification of cutter offset in end milling without a H. Pérez, A. Vizán, J.C. Hernandez, M. Guzman, Estimation of cutting forces in
prior knowledge of cutting coefficients, International Journal of Machine micromilling through the determination of specific cutting pressures, Journal
Tools and Manufacture 43 (2003) 687-697. of Materials Processing Technology 190 (2007) 18-22.