Available online at www.sciencedirect.

com

Procedia CIRP 1 (2012) 132 137

5th CIRP Conference on High Performance Cutting 2012

Using NC-path deformation for compensating tool deflections

in micromilling of hardened steel
D. Biermann a, E. Krebsa*, A. Sacharowa, P. Kerstinga
a
Institute of Machining Technology, TU Dortmund University, Germany
* Corresponding author. Tel.: +49 231/755-5820; fax: +49 231 755-5141.E-mail address: krebs@isf.de.

Abstract

During the micromachining of hardened materials, the low stiffness of the milling tool results in an increased tool deflection which
has a great influence on the shape and dimensional accuracy of the machined components. In order to compensate these deflections,
an optimization method is presented in this paper. Based on measured form errors of the machined workpieces, the NC programs
are optimized iteratively to reduce the shape deviations. To verify this method, experimental investigations were carried out by
milling pockets in hardened steel. The results show a significant reduction of the tool deflection after the optimization.

2012
2012 The Authors.
Published byPublished by Elsevier
Elsevier BV. B.V.and/or
Selection Selection and/or peer-review
peer-review under responsibility
under responsibility of Professor
of Prof. Dr. Konrad Wegener
Konrad Wegener

Keywords: Milling, Micromachining, Analysis, Compensation

1. Introduction leading to ploughing instead of cutting exists [7]. This

For the manufacturing of complex lightweight
components with integrated functional elements, the
sheet-bulk metal forming process, which combines the
advantages of sheet and bulk metal forming, gives the
opportunity to produce these parts economically. For this
purpose, complex dies, which are usually made of
materials with a high hardness (e. g., high-speed steel
S6-5-2, 63 HRC), with small structures and functional
elements are required. Due to the hardness of the
machined material, the precise manufacturing of these
elements by micromilling is a challenge [1, 2]. In
previous studies, the machinability of steels with
hardness above 60 HRC using commercially available
micro end-milling cutters was shown [3]. This is
remarkable since the milling tools consist of cemented
carbide. However, the stiffness of the milling tools plays
a decisive role since the tool deflections during the
machining of hardened steel result in an over- or
undersize of the machined workpiece. This effect is not
negligible for a precise micromilling of these structures
[4, 5].
A commonly used approach to reduce tool deflections
is the adjustment of cutting parameters in order to
minimize the process forces [6]. However, in
micromachining a risk to achieve chip thicknesses
has a negative effect on the surface quality and tool life,
especially during the manufacturing of difficult to
machine materials. In addition, the increased milling
time based on a low chip removal rate leads to higher
manufacturing costs.
The manipulation of NC paths represents an
alternative method. For simple manufacturing tasks, the
shape correction can be realized directly at the machine
control by an over-positioning of the tool. For
geometrically more complex parts, the use of a CAM
system is necessary. However, in most conventional
programs such correction functions are not offered.
Therefore, a new CAD model and NC program have to
be constructed. This can be very complicated and is
often not realizable without computer assistance [8].
In this article a practical solution to this problem is
presented. In order to minimize the form errors, an
analysis and optimization approach is applied which
modifies NC programs by taking tool deflections into
account and requires no reconstruction in CAD or CAM
systems. The verification of the compensation approach
is performed for pocket micromilling in hardened high-
speed steel S6-5-2.

2212-8271 2012 The Authors. Published by Elsevier B.V. Selection and/or peer-review under responsibility of Professor Konrad Wegener
http://dx.doi.org/10.1016/j.procir.2012.04.022
 D. Biermann et al. / Procedia CIRP 1 (2012) 132 137
2. Shape-deviation compensation
For the compensation of form errors, which can occur
during the milling process due to tool deflections, an
approach is applied that deforms the original NC
programs of the workpiece with respect to the
corresponding shape deviations [9]. In Fig. 1, a
schematic diagram of the extended process chain is
shown. The advantage of this procedure is
implementation of the shape modifications without the
process of surface reconstruction, which may be very
time-consuming and costly for complex, free-form
geometries [8].
The direct approach for form-error compensation
consists of two steps. The first one is the analysis of the
workpiece deformations. At this stage, the measured
data is compared to the reference geometry as
represented by the CAD model. The result of the
analysis is a discrete deformation field describing the
shape deviations of the manufactured workpiece (see
Fig. 2, left). In the second step, a continuous
deformation function is calculated with respect to the
determined deformation field. The NC programs for the
milling process are modified using this deformation
function. Afterwards, the new NC programs can be
directly used for milling on the machine. This
compensation process can be iterated until the shape-
deviation measure e, e.g. maximal deviation with respect
to the deformation field, is within the range of tolerance
Regular process chain
a the workpiece, however, the matching calculated by
CAD
CAM Compensation of
the tool deflection
b methods allow slight deformations of data during the
Form error
Machine
compensation
e>t
c
Measurement Analysis tool
e<t
Machined workpiece
a Reference geometry (CAD model)
b New NC program
c Measured data
e Shape-deviation measurement
t User-defined threshold
Fig. 1. Schematic diagram of the new process chain
133
t, which is defined by the user. In the following, these
two steps of the form-error compensation approach are
explained in more detail.
2.1. Shape-deviation analysis
In general, registration methods are used for the
comparison of two 3D surfaces. The aim of registration
is the adjustment of these shapes by minimizing a
specified error measure, e.g., maximal or least-squares
distance between the shapes. During this process one
shape (data) is transformed in order to fit the other shape
(model) in the best possible way. A well-known
registration method is ICP (Iterative Closest Point)
[10, 11]. It adjusts the data into the model by calculating
a rigid transformation, which minimizes the least-
squares distance between the corresponding points. An
important part of the registration procedure is the
determination of the corresponding pairs. For every data
point, a corresponding model point with the shortest
distance is computed by nearest-neighbor search. In case
the model is represented as a point cloud, appropriate
search structures like kd-trees [12] are used for matching
data points to model points. Since the model is given by
a mesh or a set of parametric surfaces, point projection
methods [10, 13] are used for finding the corresponding
points.
The advantage of rigid registration methods like ICP
is the efficiency and high applicability. Moreover, the
corresponding points establish a field of deformation
vectors and, thus, describe the shape deviation between
the data and the model. In case of strong deformations of
point projection or nearest-neighbor search may be
error-prone and should thus not be used for the analysis
and compensation of the form errors. In this case, non-
rigid registration methods should be applied [14]. These
registration in order to obtain an accurate adjustment.
2.2. NC-path deformation
The idea for compensating form errors, which are
caused by deflections of the milling tool, is based on the
displacement of the original positions of the tool with
respect to the shape deviations determined in the
analysis phase. For example, in case of an oversize, the
positions of the milling tool should be displaced closer
to the workpiece surface in order to remove more
material. Biermann et al. [9] already presented an
approach for the manipulation of NC programs using a
B-spline based free-form deformation. This method
allows slight shape modifications of a workpiece using a
direct displacement of the NC paths. For example,
134 D. Biermann et al. / Procedia CIRP 1 (2012) 132 137
forming tools in sheet metal forming have been modified
in order to compensate the shape deviations caused by
springback. This procedure can be applied for the
compensation of form errors in micromilling caused by
tool deflections in the same manner.
A great challenge of this approach is to find a smooth
deformation function, which approximates the discrete
deformation field and can be used for adjusting the tool
positions. Due to the noise in the measured data, it is
important to define a deformation function including
smoothing properties. A well-known technique, which
can be used for this purpose, is a B-spline based free-
form deformation [9, 15]. It pointwisely deforms the
space and all the 3D objects inside the space by applying
a continuous function F: 3 to all points of the
objects. The new position of an arbitrary point p is
defined as a trivariate weighted sum of control points
Pi,j,k:
l ,m,n
p ' = F (u , v, w) = B (u) B (v) B (w) P
i , j , k =1
i j k i , j ,k
where the weights Bi(u), Bj(v), and Bk(w) are B-spline
basis functions defined on the parameter space and
(,,) are the local parameters of the point p [13].
The space deformation results from a displacement of
the control points, which are initially located on a
regular (lmn)-lattice (see Fig. 2). In order to
approximate a discrete vector field, the required
displacements of the control points can be calculated
using a least-squares method. Thereby, the quality of the
approximation depends on the number of control points.
In general, it is difficult to find a function F, which
fulfils defined precision requirements. In order to assure
a desired shape accuracy of the workpiece after
deformation, an adaptive approach for the calculation of
a B-spline based free-form deformation was presented in
[16].
This technique can be used for pointwise deformation
of 3D objects like meshes or even NC programs. Thus,
Fig. 2. Example for the deformation of data by displacing one control
point of a B-spline volume in 2D case
in order to compensate the form errors caused by tool
deflection, the tool positions can be directly displaced by
the function F with respect to a set of deformation
vectors. Using this approach, the NC programs with
several thousands of points can be optimized in a few
minutes.
3. Experimental setup
In order to verify the presented approach, small
pockets of dimensions 5 mm x 5 mm with a corner
radius of 1.5 mm were machined. In order to analyze the
effect on the resulting shape deviation, the width of cut
ae was varied as follows. Experiment A is conducted
using a width of cut ae = 200 m, experiment B
ae = 400 m, and experiment C ae = 600 m. Each
experiment was performed for three times. The
remaining cutting parameters are set according to
fundamental investigations of Biermann et al. [3]. Fig 3
gives an overview of the parameters.
(1)
Fig. 3. Micro end mill cutter and picture of the process
These experiments were carried out on a CNC micro
milling machine KERN HSPC 2522. To achieve high
cutting speeds, the spindle Precise SC 1060-OA, with an
operating range from 60,000 min-1 to 160,000 min-1, was
used. The machine manufacturer specifies a production
accuracy of one micron.
The tools are cemented carbide end mill cutters with a
diameter of d = 1 mm. They have a long cutting edge
l2 = 8.5 mm. The radial run-out is specified as ro = 5 m.
Fig. 3 shows the tool geometry and a picture of the
process.
The workpiece material is a high-speed steel, S6-5-2
(1.3343), with a hardness of approx. 63 HRC. The tool
paths were programmed in the CAM system
CimatronE 9.0. The measurements were carried out on a
confocal white-light microscope (NanoFocus surf).
This equipment allows accurate measurements to be
achieved and to be saved as a point cloud, which can be
directly used for the analysis and compensation
procedure. The surface was measured with the horizontal
 D. Biermann et al. / Procedia CIRP 1 (2012) 132 137 135

resolution of three microns and the vertical resolution of

0.2 microns.

4. Results and Discussion

First, during the evaluation of the experiments A, B,

and C the length of the milled pockets is measured and
compared to the reference geometry, as shown in Fig. 4.
Therefore, three measurements were carried out in each
direction. The tool deflection is then deduced from the
shape deviation. The averaged values are shown in Fig.
5. The deflection for a tool with a diameter d = 1 mm is
unexpectedly high. In experiment C the maximum Fig. 4. Drawing of the shape deviation measurement
values are far beyond 100 microns. In the literature, the
typical deflection of 30 microns is indicated [4, 17]. In contrast to the measurement results, shown in
Furthermore, a progressivity of the tool deflection can Fig. 5, where only the length of the pockets was
be recognized in the graph for increasing width of cut ae. considered, the shape deviations used during the
Thereby, the gradient between 400 and 600 microns of compensation are based on the deformation field from
the width of cut is higher than between 200 and 400 the analysis step and include the whole measurement
microns. Thus, a linear behavior cannot be determined. data.
A possible reason for the increased gradient can be
found in the change of the engagement situation. When
the width of cut is higher than the tool radius, a varied
down- and up-milling is occurred in the process, which
changes the curve of the force. Overall, the influence of
the width of cut on the tool deflection is quite high.
Published investigation results show that the width of cut
has a rather low influence on the process forces [3]. This
will be analyzed in future works.
For the compensation of the tool deflection,
experiment B was selected with the knowledge that a
higher radial feed will still ensure the process reliability.
The deformation of the corresponding NC path is
computed based on the determined deviation between Fig. 5. Average tool deflections from the distance measurements
the measured data and the reference geometry. The
results after the first iteration step are shown in Fig. 6. It In general, the shape deviation can be reduced with
is a top view on the pocket shape in the used CAE each iteration step, but is limited by the precision of the
environment. To make it more obvious, the negative experimental procedure. In addition to inaccuracies in
shape of the measured pocket is depicted. The the milling process, which may arise through the
deformation field presents the distance between the outer machine (1 m) and the tool radial run-out (5 m),
points and the boundary (model). Despite noise in data measurement errors have to be taken into account. Due
and discretization errors, the deformed NC paths are to some surface defects during the milling process, the
smooth. measurement accuracy in the current experiments is
specified by six microns. An additional source of error is
the discretization of the measurement data. Due to the
large size of the dataset (4,915,089 data points), a data
reduction of 66 % was carried out to improve the
handling. So an additional inaccuracy of nine microns
has to be taken into account with respect to the
measurement resolution. Overall, in the presented
compensation process an accuracy of 21 microns can be
assumed. Thus, the range of tolerance t was set to 21
microns.
136 D. Biermann et al. / Procedia CIRP 1 (2012) 132 137
Fig. 6. NC-path deformation on the example of pocket milling
For the compensation of the tool deflection, the
analysis tool and the form-error compensation were
carried out iteratively (see Fig. 1.). As the results show
in Fig. 7, the maximum distance e has a value of 17
microns after two iterations, which is within the defined
tolerance range t. Accordingly, two iterations were
sufficient. Due to the larger area of contact, the tool
deflection is higher at the radii, so the value of the
average distance increases compared to Fig 4. In
conclusion, both, the average and maximum distance
from the reference geometry were reduced significantly
using the presented optimizing approach.
5. Conclusion
In micromachining, deviations from the ideal
geometry cannot be avoided completely. An enormous
tool deflection of about 100 microns was measured,
particularly, when using tools with a cantilever length to
machine deep cavities in hardened steel.
In order to compensate deflections of the milling tool
and to achieve a high shape accuracy, a new method was
presented in this paper. Based on measured data, the
shape deviations from the reference shape are
determined. The corresponding manipulation of NC
paths is performed using a free-form deformation
technique. The results show that this method can be
applied for the compensation of tool deflections during
the micromilling process of hardened high-speed steel
(63 HRC). In the experiments, a huge improvement of
the shape and of the dimensional accuracy was achieved
after only two iteration steps.
Fig. 7. Shape deviation after the NC-path deformation
The next step in this research field could be a further
extension of the process chain. A milling simulation,
which is suitable for the computation of tool deflections
during the micromachining process, would save costs
and time in order to carry out the iteration steps of the
compensation approach virtually.
Acknowledgements
This work was supported by the German Research
Foundation (DFG) within the scope of the Transregional
Collaborative Research Centre on sheet-bulk metal
forming (SFB/TR73) in the subproject B2 and the
SFB708 in the subproject C4.
References
[1] Bissacco G, Hansen H, De Chiffre L. Micromilling of hardened
tool steel for mould making applications. Journal of Materials
Processing Technology 2005; 167: 201-7.
[2] Li P, Hoogstrate A M, Oosterling J A J, Langen H H, Munnig
Schmidt R. Experimental validation of micro endmill design for
hard milling application. Proceedings of the 3rd International
CIRP High Performance Cutting Conference. June 12th-13th
2008, Dublin, 69-78.
[3] Biermann D, Baschin A, Krebs E, Schlenker J. Manufacturing of
dies from hardened tool steels by 3-axis micromilling. Production
Engineering 2011; 2: 209-17.
[4] Kahnis P. Analyse von Greneinflssen bei einer Herabskalierung
des Frsprozesses in den Mikrobereich. Dissertation. Essen,
Germany: Vulkan Verlag; 2008.
[5] Uriarte L, Herrero A, Zatarain M, Santiso G, Lacalle L N Lopez
de, Lamikiz A, Albizuri J. Error budget and stiffness chain
assessment in a micromilling machine equipped with tools less
 D. Biermann et al. / Procedia CIRP 1 (2012) 132 137 137

than 0.3 mm in diameter. In: Precision Engineering 2007;

31(1):112.
[6] Rall K, Tauchen M, Findeklee J. Reduktion der
Werkzeugabdrngung beim HSC-Schlichtfrsen. VDI-Z
Integrierte Produktion 2000; 142 (3): 32-4.
[7] Hui T Y, Segon H, Min K L, Byung-Kwon M, Sang J L. Ploughing
detection in micromilling processes using the cutting force signal.
In: International Journal of Machine Tools and Manufacture
2011;51(5):37782.
[8] Varady T, Martin RR, Cox J. Reverse engineering of geometric
models an introduction. Computer-Aided Design
1997;29(4):255-68
[9] Biermann D, Sacharow A, Surmann T, Wagner T. Direct free-form
deformation of NC programs for surface reconstruction and form-
error compensation. Production Engineering Research and
Development 2010;4(5):501-7.
[10] Besl P, McKay N. A method for registration of 3-d shapes. IEEE
Transactions on Pattern Analysis and Machine Intelligence
1992;14(2):23956.
[11] Chen Y, Medioni G. Object modeling by registration of multiple
range images. IEEE Conference on Robotics and Automation
1991;3:272429.
[12] Bentley JL. K-d trees for semidynamic point sets. Proceedings of
the sixth annual symposium on Computational geometry (SCG
'90). 1990;187-97.
[13] Piegl L, Tiller W. The NURBS book (2nd ed.). New York,
NY,USA: Springer-Verlag, 1997.
[14] Sacharow A, Balzer J, Biermann D, Surmann T. Non-rigid
isometric ICP: A practical registration method for the analysis and
compensation of form errors in production engineering.
Computer-Aided Design 2011;43(12):1758-68.
[15] Sarraga R. Modifying CAD/CAM surfaces according to
displacements prescribed at a finite set of points. Computer-Aided
Design 2004;36(4):343-9.
[16] Sacharow A, Surmann T, Biermann D. Adaptive free-form
deformation for the modification of CAD/CAM data. ADVCOMP
2011; 27-31.
[17] Schwietering C. Technologische Aspekte der mikromechanischen
Frsbearbeitung mit Schaftwerkzeugen. Dissertation. Essen,
Germany: Vulkan-Verlag; 2003.