Challenges for Industrial Robots Towards Milling Applications

Abstract

Robotic machining is proposed to be a manufacturing technology alternative to CNC machining to lower the overall costs, especially in large scale part manufacturing industries such as nuclear, aerospace and partially automotive. Opposed to several advantages, industrial robots introduce several major technical challenges. In this paper, the industrial robots are compared to CNC machine tools towards the purpose of machining operations in terms of technical challenges, potential cost savings and the required research to increase the manufacturing readiness level of robotic machining. The effect of industrial robots on machining dynamics is discussed through case studies; the inaccuracies for industrial robots on tool path contouring and potential cost reduction are discussed.

Full text

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
1
CHALLENGES FOR INDUSTRIAL ROBOTS
TOWARDS MILLING APPLICATIONS
Lutfi Taner Tunca,b , Omer Faruk Sapmaza
a, Sabanci University Faculty of Engineering and Natural Sciences, Composite Technologies Centre of Excellence,
Istanbul/TURKEY, ttunc@sabanciuniv.edu
b, Nuclear AMRC, The University of Sheffield, Sheffield/United Kingdom
Abstract
Robotic machining is proposed to be a manufacturing technology alternative to CNC machining to lower the
overall costs, especially in large scale part manufacturing industries such as nuclear, aerospace and partially
automotive. Opposed to several advantages, industrial robots introduce several major technical challenges. In
this paper, the industrial robots are compared to CNC machine tools towards the purpose of machining
operations in terms of technical challenges, potential cost savings and the required research to increase the
manufacturing readiness level of robotic machining. The effect of industrial robots on machining dynamics is
discussed through case studies; the inaccuracies for industrial robots on tool path contouring and potential cost
reduction are discussed.
Keywords: Robotic machining, Economical analysis, Milling dynamics, Accuracy
1 Introduction
Large-scale part manufacturing requires large-scale machine tools and as well as heavy-duty part moving
between manufacturing units. Thus, the capital investment and operational costs drastically increases [1].
Involvement of low cost industrial robots in the mobile machining context has been proposed for the last two
decades to employ “process-to-part” approach. Industrial robots were initially utilized for tasks requiring
millimeter level accuracy and repeatability, where the end effector was under almost static loads as in handling,
assembly and welding. For the last two decades, industrial robots are proposed to be candidates for machining
applications, specifically drilling and milling. However, the flexible links, primitive control systems and
programming capabilities are the technical barriers towards the use of industrial robots for milling operations
[2].
Considerable amount of research has been carried out to analyze, model and improve both positional accuracy
and dynamic stiffness of robots used in mobile machining applications. Most of the literature covers robots in
serial arm, where milling stability, kinematic, control, programming and process development are focused.
The compliance caused by the bearings and gears are identified to be up to 50% and for large robots even up
to 75% [3]. One of the first models to predict the robot deflection, under milling forces, was proposed by Abele
et al. [3]. Later, Slavkovic et al. [4] used off-line tool path compensation method to improve positional accuracy
in robotic milling. They used cutting force simulations to predict the robot deflection, then used the deflection
amount for off-line compensation purposes. It was shown that the robot positional accuracy can be improved
through compensation techniques. Dumas et al. [5] introduced a robust and fast procedure to identify the joint
stiffness values of a six-revolute serial robot. They considered both translational and rotational displacements
of the robot end-effector under machining forces.

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
2
Dynamics of robotic milling been investigated by several researchers. In one of the early studies, Pan et al. [6]
emphasized the differences between the CNC machine tools and a vertical articulated robot in terms of
machining dynamics and pointed out the low frequency modes, asymmetrical dynamics and vibrations along
the tool axis. Zaghbani et al. Error! Reference source not found. applied variable spindle speed strategy for
robotic machining system, where vibration and cutting forces are used as the stability criterion.
In the literature it is emphasized that the introduction of industrial robots into the milling applications brings
back several challenges, which may be mostly ignored or dealt with simplifications in CNC machining
applications. In this paper, such challenges in robotic milling are highlighted as opposed to CNC machine
tools. The issues include effect of robotic platforms on milling dynamics as a summary of the previous work
based on Tunc and Shaw [9][10] in Section 2 and Section 3. This is followed by the issues on programming
and dynamic tool path contouring as discussed in Section 4. Then, the paper is finalized with conclusions.
2 Effects of Robotic Platform on Milling Dynamics
Dynamics and stability of milling operations with industrial robots involves several special cases compared to
CNC machine tools due to the significant differences in terms of structural build up and configuration. CNC
machine tools are generally configured in axis on axis principal, where massive and rigid structural blocks
carry the corresponding axis. However, in robotic milling applications, the milling spindle is attached on to
the end effector of the robot, which may be in parallel kinematic or serial arm configuration. In this study, a
hexapod type industrial robot is considered as shown in Figure 1. In this section, effects of the hexapod robotic
platform on the dynamics and stability of milling is discussed through experimental observations, simulations
and cutting tests. Such discussions help the major issues to be identified and addressed in robotic milling. The
conclusions include the below issues;
(i) Asymmetrical tool tip FRF (Frequency Response Function) and feed direction dependent stability
due to robotic platform
(ii) Position dependent tool tip dynamics and stability
(iii) Dynamic absorber effect of the low frequency modes introduced by the robot
(iv) Increased effect of cross FRFs
Some of the above issues can be generalized for most of the robotic milling cases, as explained in the rest of
this section.
Figure 1: Comparison of CNC milling and robotic milling setup [10].
2.1 Experimental observations on dynamics of robotic milling
In the robotic milling configuration studied in this paper, the translational and rotational motion are provided
by the joint movements of the six struts fixed to the ground. The spindle axis is horizontally mounted and
hence the struts are not subjected to bending moment due to the weight of the spindle assembly. In such a
configuration, the robot introduces flexibility along the tool axis and one of the transversal directions of the
cutting tool, where the other transversal direction is in line with the axis of the struts.

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
3
In this paper, the frequency response functions at the tool tip are measured through impact-hammer tests, then
the results are discussed. The low frequency modes contributed by the robotic platform are compared with the
modes contributed by CNC machine tools and cutting tools as shown in Figure 2. Tool 1 is an inserted 25 mm
diameter end mill with two-flutes at 160 mm length. Tool 2 is an inserted flat end mill having integral holder,
16 mm diameter at 141 mm length, whereas Tool 3 is a 65 mm diameter face mill with 140 mm overhang
length.
(a) Comparison with CNC machine tool modes
(b) Comparison with cutting tool modes.
(c) Asymmetrical tool tip FRF with cross FRFs
Figure 2: Low frequency modes contributed by the robot [10].
The comparison of the low frequency modes due to the robotic platform with the modes contributed by the
CNC machine tools are given in Figure 2a. It is seen that the modes around 20-40 Hz are at least one order of
magnitude more flexible than that of the CNCs. Comparison with the cutting tool flexibility (see Figure 2b)
shows that the modes contributed by the robotic platforms may be as high as the cutting tool modes and in
some cases, such as face milling, the tool may be even much stiffer than the robot. As a result, the robotic
milling system has multiple modes at comparable amplitudes, both at low and high frequency ranges. This is
an unlikely case for most of the conventional CNC milling systems.
In most of the analytical stability analysis [8] the cross FRFs are mostly ignored and even in some of them the
tool tip dynamics may be assumed as symmetrical [8]. However, this may not be a valid assumption for robotic
milling as shown in Figure 2c, where the significance of cross FRF, i.e. Gxy, can be clearly observed. Around
the robot modes the magnitude of Gxy is much higher than that of Gyy, and half of Gxx. As the tool modes
are observed, Gxy is almost half of Gxx. The FRFs along X and Y directions show significantly asymmetrical
behavior, as well. The effect of such cases on stability of milling will be further discussed.
Considering that the tool tip FRF is a couple response of robot, spindle, tool holder and the tool, it can be
expected that if the dynamics introduced by the robot changes with position, the coupled response would
change, as well. Such a position dependent tool tip dynamic may be observed depending on the ratio between
the robot modes, which is changing, and the tool modes. In Figure 3a, the tool modes measured at the tool tip
show significant variation in terms of both the natural frequency and amplitude, among 7 random positions of
the robot. In this tooling case, i.e. Tool 1, the modes contributed by the tool are in the range of 450-500 Hz,

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
4
which is relatively close to the robot modes. However, in the second case, i.e. Tool 2, the tool modes are at
relatively very high frequencies. It is clearly seen that the tool modes do not change significantly. Such a
conclusion clearly demonstrates the importance of tooling selection for robotic milling applications.
(a) Variation of Gxx - Tool 1. (b) Variation of Gxx - Tool 2.
Figure 3: Position dependent tool tip FRF based on tooling [10].
2.2 FEM Simulation of robot dynamics
Following the experimental observations about the effect of robot modes on the tool tip dynamics, the structural
modes of the robotic platform are simulated using a finite elements model as described in this section. The
components included in the solid model are shown in Figure 4.
(a) Assembly
(c) Yoke 1
(d) Top Plate
(e) Base Plate
(b) Upper Rod (f) Lower rod (g) Yoke 2 (h) Support 1
Figure 4: Solid models of the components.
In FEM simulations, four variants of the robotic platform are considered to mimic various real case scenarios,
as shown in Figure 5, which are explained as follows. In the first model, the hexapod legs are represented as
piston-cylinders, where the upper rod is sliding inside the lower rod. In the second scenario, a mass is attached
onto the top plate. In the third model, the legs are represented as lead screw geometry, where the lower rod is
sliding inside the upper rod. Finally, in the fourth case, the robot is mounted on a base plate rather than the
earth. For each model, the robot’s vibration modes are simulated at four different positions and the variation
of the natural frequencies are observed.
(a) Model
(b) Model 2
(c) Model 3
(d) Model 4
Figure 5: Model variants of the robotic assembly.

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
5
2.2.1 Simulation results and verification
The simulations are performed for the first 6 modes of the robotic platform, counted in X, Y and Z directions.
Then, the directions are decided by observing the mode shapes. The variation of the modal frequencies are
plotted in Figure 6, for the minimum height, i.e. the stiffest position, of the robot.
(a) Model 1 (b) Model 2
(c) Model 3 (d) Model 4
(e) Experimental measurement of the 1st mode.
Figure 6: Simulated Low frequency modes contributed by the robotic platform.
In Figure 6a, it is seen that the first 2 modes, which are due to the X and Z directions are around 142 and 149
Hz. Then, the torsional mode is observed around 248 Hz. A similar distribution is observed when a heavy mass
is added onto the robot top plate, where the modal frequencies decreased significantly with the effect of the
attached mass. Such a comparison clearly emphasizes the importance of the interaction between the robotic
platform and the other components such as spindle. In Model 3 and Model 4, the robot modes are simulated
for the longest and shortest reach of the hexapod robot in Y direction. Comparison of Figure 6c and Figure 6d,
shows that the modal frequencies significantly decrease with the increasing height of the robot. At the shortest
and longest reach of the robot, the bending mode is around 35 Hz, and 25 Hz, respectively. Similar trend is
seen in the measurement of the first mode along X direction, which changes from 32 Hz to 22 Hz.
0
50
100
150
200
250
1 2 3 4 5 6
142 149
248 637 639 671
Natural Freq (HZ)
Mode Number
0
200
400
1 2 3 4 5 6
49 49 129 205 216 224
Natural Freq (Hz)
Mode number
0
200
1 2 3 4 5 6
35 37
107 135 146 153
Natural Freq. (Hz)
Mode Numer
0
200
1 2 3 4 5 6
23 26
87 109 116 117
Natural Freq. (Hz)
Mode Number

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
6
3 Stability of Robotic Milling
In the previous section, it was well demonstrated that the structural modes of the robot may significantly affect
the tool tip in various aspects. Such an interaction would obviously change the stability behavior of the milling
process. In this section, such affects are discussed through stability simulations and some experimental
verifications. The cases for position dependent stability, asymmetrical tool tip FRF and hence the feed direction
effect and the variation of stability limits with the inclusion of cross FRFs are demonstrated.
Tool
No
Milling
Mode
Radial
Immersion
Case 1
Tool 1
Down
40%
Case 2
Tool 1
Down
30%
Case 3
Tool 2
Down
50%
Table 1: Simulation cases to investigate position dependant stability.
3.1 Tooling and Position dependent stability
The interaction between the cutting tool modes and the robot modes is important to identify the component
governing chatter at a given spindle speed, which changes the excitation frequency and hence the vibrating
component may vary. For a representative, half immersion, down milling case, the variation of the chatter
frequencies with spindle speed, for three cutting tool configurations, is shown in Figure 7a. As Tool 1 is
considered, the chatter mode shifts to robot mode for a very short range of spindle speeds around 5000 rpm.
For Tool 2, the chatter frequency is always around the tool mode. The chatter frequencies simulated for Tool
3 is always around the robot or spindle modes, i.e. low frequency modes. Thus, it can be concluded that tooling
is critically important for robotic milling applications.
(a) Chatter mode variation (b) Tool 1
(c) Tool 2 (d) Tool 3
Figure 7: Position and tool dependent stability diagrams in robotic milling [9].
The position dependent stability for these specific tools are also plotted in Figure 7b, Figure 7c and Figure 7d,
respectively. In general, it is observed that as the tool dynamics is affected by the interaction with the robotic
platform, it becomes more position dependent. For instance, the stability diagram of Tool 3 shows significant
variations both in terms of the spindle speed and stability limits. However, for the case of Tool 1 and Tool 2,
the variation is not that drastic, i.e. at least come common stability regions can be identified.

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
7
3.2 Feed direction effects on stability diagrams
The robotic platform introduces asymmetric compliance in x and y directions of the cutting tool, which causes
the tool tip FRF to become asymmetrical. In milling stability analysis [8], the feed direction is considered
along with the x direction and the cross feed direction is in y direction. Considering that the stability limits
depend on the FRFs in both directions, it would be expected, for asymmetrical systems, that the feed rate
direction to affect the stability limits and pockets. For a representative half immersion case, the effect of feed
direction on the stability diagrams is demonstrated in Figure 8a, where the feed direction is varied at 0, 30, 45,
75 and 90 degrees. The change of stability limits with respect to spindle speed is clearly seen. This case is
specific to robotic milling as the tool tip FRF is mostly symmetrical for CNC machine tools as compared in
Figure 8b.
(a) Effect of feed direction (b) Comparison of tool tip FRF
Figure 8: Feed direction dependant stability diagrams in robotic milling [9].
In the robotic milling case, the tool tip FRF measured in x and y directions are asymmetric in terms of natural
frequency, number of modes and amplitude. However, FRF in x and y directions are almost the same when the
same tool is attached on a CNC machine tool. Such an observation is important for utilization of robotic milling
as it is clear that the chatter-free material removal will depend on the feed direction and hence on the selected
tool path pattern, which is not the case in CNC machine tools.
3.3 Effect of amplified Cross FRFs
In most of the stability analysis, the cross FRFs are ignored for simplicity and as they are usually insignificant
[8]. However, the flexible kinematic chain of industrial robots may introduce cross FRFs under some
conditions, which may affect stability diagrams. In this section, the effect of significant cross FRFs on the
stability diagrams is demonstrated through simulations. For a face milling type of operation, the direct and
cross FRFs are plotted in Figure 9a. The amplitude of the cross FRF is about 50% of Gxx and 25% of Gyy.
The effect of the cross FRF on milling stability for a slotting case is demonstrated in Figure 9b. The
corresponding stability diagram shows that when the cross-transfer function is considered in the stability
solution, substantial increase is observed in the stability limits.
(a) Tool FRF (b) Face milling (Slot)
Figure 9: Effect of cross transfer function on stability diagrams [9].

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
8
3.4 Representative Experiments
The observations performed in dynamics and stability of robotic milling compared to the CNC machining, is
experimentally discussed for representative cases in this section, the experiment conditions are given in
Table 1. The cutting tests are conducted on AL7075-T6, having coefficients are Ktc=800 MPa, Krc=300 MPa.
Test 1 and Test 2 are conducted to assess the effect of machining location and feed direction on stability. Test
3 is done to observe the effect of tooling on stability of robotic milling, where in Test 4 the effect of cross FRF
is demonstrated. In the tests, vibration and sound during the tests are measured to assess stability.
Tool
No
Radial
Depth
Cut
Type
Feed
Direction
Test 1
Tool 1
Half
Down
x
Test 2
Tool 1
Half
Down
y
Test 3
Tool 2
Half
Down
x
Test 4
Tool 1
Quarter
Up
x
Table 1: Experiment cases.
In Test 1, Test 2 and Test3, the stability diagrams are simulated by CutPro ©, where cross FRF is ignored. In
Test 4, the stability diagrams are simulated by introducing CTF to the stability solution [8] and eigenvalue is
solved numerically. The simulated stability diagrams and experimental results are plotted in Figure 10.
(a) Test 1 (b) Test 2
(c) Test 3 (d) Test 4
Figure 10: Stability tests [9].
In Figure 10a, it is observed that the stability diagrams do not vary significantly with position when the feed
is in the x direction. However, in Test 2, where the feed direction is along y direction, the stability diagrams
are changing with robot position, significantly as plotted in Figure 10b. The absolute stability limit and stability
lobes shift according to the machining location and feed direction. Such observations are verified through
experiments, where a good agreement is observed.
The stability diagrams for Test 3 are calculated using the FRFs measured at four machining locations for feed
direction x and y (see Figure 10c). It is can be seen that the stability diagrams do not vary much even the feed

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
9
direction is in y, contrary to Test 2. Besides, the absolute stability limit is significantly higher than that of Tool
1. So, it can be said that milling with Tool 2 in feed y direction would lead to higher chatter-free cutting depth
with less variation according to the feed direction and machining location. This observation significantly
emphasizes the importance of tooling in robotic milling applications. In Test 4, the stability diagrams are
calculated with and without the cross FRFs. As seen in Figure 10d, there is a significant difference in stability
diagrams, which is experimentally verified.
4 Toolpath Contouring Accuracy and Programming
Industrial robots are known with their low positioning accuracies. However, in most of the cases, robot
manufacturers provide the point-to-point motion repeatability of robots rather than the dynamic tool path
contouring accuracy. Thus, in general, the contouring accuracy of industrial robots are not well-known. It is
also important to keep a level of accuracy in the actual feed rate. Considering that the cutting forces, chip
formation and other mechanical effects strictly depend on the feed motion, the fluctuations in the feed rate
throughout the tool path will cause several deficiencies such as random cutting forces, uneven chip formation,
unpredictable deflections etc. (see Figure 11). Thus, it is important to assess both dynamic tool path contouring
and feed rate accuracy of industrial robots. In this section, such issues are experimentally observed together
with the programming capabilities of the robotic platform.
Figure 11: Effect of feed rate fluctuations on chip formation and cutting forces.
4.1 Experimental setup and test conditions
The dynamic measurement of the end effector is measured by attaching a spherical measurement reflector
(SMR) on the tip of the robot. Thus, the measurements are performed in air cutting conditions, i.e. there is not
any cutting forces acting on the robot. Leica AT901 laser tracker (see Figure 12a) is used for position
measurements, which can provide data out at a rate of 1000 Hz, which is enough to dynamically track the robot
position throughout a tool path. The tool path types are designed according to the ISO 9283:1998 Preview
Manipulating industrial robots -- Performance criteria and related test methods [11]. According to the named
standards, a prismatic working envelope is defined considering the reach of the robot (see Figure 12c and
Figure 12c). Then, circular, linear and spline type of paths are defined in the diagonal plane of the prismatic
envelope. The circular tool paths are performed at different radii values to assess the control bandwidth.

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
10
(a) Leica AT901 laser tracker (b) Robot with SMR attached
(c) ISO 9283 standard (d) the tool paths designed for the test
Figure 12: Experimental setup and test tool path types.
4.2 Programming Issues
Industrial robots do not provide as sophisticated programming functions as CNC machine tools. In general,
two motion times can be defined, i.e. Linear and Circular, as shown in Figure 13. In linear motions, there is a
look ahead function for only four consecutive points, this limits the contouring capability of the robot
especially for densely created tool paths. The motion accuracy for linear motion segments are set by FINE and
CNT options. When FINE option is set, the robot stops at every tool path point (see Figure 13a), which is not
suitable for machining applications. When CNT option is used, the number between 0 and 100 following CNT
sets the accuracy level. Similarly, CNT 0 setting stops every tool path point and as it is increased towards CNT
100, the tool path contouring accuracy is relaxed. However, the effect of such a setting for a densely packed
point cloud along a tool path is not known. The experimental analysis aims to assess the trade-off between
CNT setting and the tool path contouring performance specifically in milling tool paths. In circular tool path
commands, the robot cannot complete a full circle in one command, thus the circular path motion commands
are divided into two equal segments. The robot control does not provide any B-Spline interpolator.
Figure 13: Motion types and settings for the robot [12].
4.3 Contouring accuracy
The circular and B-spline type paths are shown in Figure 14a and Figure 14b. The circular tool paths have 3
segments, linear (from P0 to P1), circle half-1 (P1 to P2) and circle half-2 (P2 to P1). A linear segment is added
in front of the circular path to assess the feed rate response and hence the control bandwidth of the robot, as
well. This further explained in the next section. B-Spline path is designed in a such a way that it has varying
radius of curvature along the tool path. The tests are performed at gradually increasing feed rate values, and at
different accuracy settings for the B-Spline path.

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
11
(a) Circular path sections (b) B-spline path
Figure 14: The circular and B-spline paths designed for the tests.
The tool path contouring error for 75mm diameter and 30 mm diameter circles are shown in Figure 15a and
Figure 15b, respectively. It is seen that the contouring accuracy, while running around a 75-mm circle, is
mostly less than 0.1mm when the feed rate is 3000mm/min, which increases to 0.2 and 0.7 at feed rates of
6000 mm/min and 12000 mm/min, respectively. As the diameter is decreased down to 30 mm, the contouring
error increases significantly from 0.5 mm to 2 mm at increasing feed rate values. Such a result shows that the
robot’s relatively low control bandwidth limits the path contouring accuracy at lower circle diameters, which
is further discussed in the next section. Another observation in Figure 15b, is that the contouring accuracy is
not the same while the robot is contouring from P1 to P2 and from P2 to P1. The fundamental reason for such
a difference is thought to be due to the gravitational effect. From P1 to P2, the robot is acting with the
gravitational force, however from P2 back to P1, the robot is moving against the gravitational force, which
may be a factor to increase the accuracy by decreasing the inertial overshoots.
(a) 75 mm diameter circular path (b) 30 mm diameter circular path
(c) Effect of accuracy setting on B-Spline path (d) Effect of feed rate on B-Spline path
Figure 15: Tool path contouring accuracy.
The tool path contouring error for B-spline type of path at varying feed rate values and accuracy settings, is
shown in Figure 15c and Figure 15d, respectively. In Figure 15c, it is seen that, the accuracy setting does not

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
12
have any significant effect on the tool path contouring error along a densely packed tool path. Thus, at the
programming phase, the most relaxed setting, i.e. CNT 100, would be more beneficial in terms of the dynamic
feed rate response of the system, as briefly emphasized in the next section. As the feed rate is increase gradually
from 1200 mm/min to 4800 mm/min the tool path contouring error increases significantly. Considering the
level of feed rate at which there is jump in the tool path error, it can be said that the robot is not suitable for
high speed machining strategies, where relatively high feed rate values are utilized.
4.4 Feed rate response
The accuracy of the attained feed rate is another important factor to properly perform the desired milling
operation. It is also good to know the dynamic feed rate response and the bandwidth of the robot for proper
programming purposes. Assuming that the robot is able to turn around a circle at a maximum angular
velocity, , the maximum attainable linear feed rate, F, will change with the radius of curvature, r, on the
tool path as express below; (also see
(a) Angular velocity (b) Programmed – attained feed (c) Circle radius – attained feed fit
(d) Interpolation type in circular path (e) Accuracy setting vs feed profile
Figure 16a).
feed =  (rad) x r (radius)
(1)
In this study, the feed rate response of the robotic platform is assessed by running the same tool path at
gradually increasing feed rate values and observing the attained feed rate. Then, the programmed feed rate is
plotted against the attained feed rate. At each circle radius, the feed rate value at which the attained feed rate
diverges from the programmed feed rate is identified as shown in
(a) Angular velocity (b) Programmed – attained feed (c) Circle radius – attained feed fit

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
13
(d) Interpolation type in circular path (e) Accuracy setting vs feed profile
Figure 16b. Then, the variation of the attainable feed rate is plotted against the circle radius to identify the
control bandwidth for milling applications. It is given in
(a) Angular velocity (b) Programmed – attained feed (c) Circle radius – attained feed fit
(d) Interpolation type in circular path (e) Accuracy setting vs feed profile
Figure 16c that, the linear regression line, of the circle radius and the attained feed rate, has a slope of 4.3,
which can be used to calculate the maximum linear feed rate to be set to run a circular path.
(a) Angular velocity (b) Programmed – attained feed (c) Circle radius – attained feed fit

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
14
(d) Interpolation type in circular path (e) Accuracy setting vs feed profile
Figure 16: Analysis of the dynamic feed rate response of the robotic platform.
Similar to CNC machine tools, circular tool path can be contoured by either linera interpolation or circular
interpolation. It is also important to know which type to use. The effect of interpolation type on contouring a
circular path of 75 mm diamater is demonstrated in
(a) Angular velocity (b) Programmed – attained feed (c) Circle radius – attained feed fit
(d) Interpolation type in circular path (e) Accuracy setting vs feed profile
Figure 16d, where it is seen that, when linear interpolation is used to contour the circle, the maximum
attainable feed rate is limited to 360 m/min. However, when circular interpolation is used, even 2400
mm/min of feed rate can be attained compatible with
(a) Angular velocity (b) Programmed – attained feed (c) Circle radius – attained feed fit

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
15
(d) Interpolation type in circular path (e) Accuracy setting vs feed profile
Figure 16c.
In Section 4.3, it was mentioned that the accuracy setting, i.e. CNT option, does not affect the tool path
contouring accuracy. However, as its effect on the feed profile a significant difference can be observed as
shown in
(a) Angular velocity (b) Programmed – attained feed (c) Circle radius – attained feed fit
(d) Interpolation type in circular path (e) Accuracy setting vs feed profile
Figure 16e. When CNT 0 is set as the accuracy level, the feed profile is bouncing back and forth between the
set feed and 0. As the accuracy setting is relaxed towards CNT 50 and CNT 100, the feed profile becomes even
stable. Thus, considering both path contouring accuracy and feed rate issues, the linear interpolation of B-
spline type paths should be performed using the CNT 100 option.
5 Conclusions
In this paper, the challenges introduced by the flexible structure and simple control system of the robotic
platforms, introduced to robotic milling applications, have been summarized as opposed to CNC machine
tools. The main issues include i.e. specific issues in process dynamics, effects on milling stability, tool path
contouring and feed rate response. It is discussed that most of the simplifications, assumptions and ignore
points in CNC machining become valid when robotic milling is considered. The effect of low frequency modes
on the governing modes of stability are highlighted. Depending on the interaction between the low frequency

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
16
modes and the tool modes, the tool tip FRF may become position dependent and asymmetrical. Apart from
this, the cross FRFs become significant due to the flexible and complex kinematic chain of the robots, which
affects stability issues significantly. Selection of the cutting tool is also another issue for increased stability in
robotic milling as summarized in this paper. It is shown that the position dependency may be decreased based
on the selected tool.
The programming issues have been discussed for circular and B-Spline type of tool paths. The effects of
interpolation type and accuracy settings on robot motion are discussed and preferable conditions have been
highlighted. It was found out that the accuracy setting, i.e. CNT option, does not affect the tool path contouring
accuracy for densely discretized B-spline paths but significantly affects the dynamic feed rate response. Use
of the CNT 100 option leads to much stable feed rate response, which is very important for proper milling
process. In some cases, the gravitational forces may hinder the tool path contouring accuracy, which should
be considered in feed rate planning for robotic milling.
6 Acknowledgements
The authors gratefully acknowledge the support of ESPRC, The University of Manchester and The University
of Sheffield under NNUMAN Programme with grant number EP/J021172/1. The authors acknowledge the
support of the Erasmus Internship Mobility Programme, as well.
7 References
[1] Uriarte L., Zatarain M., Axinte D., Yague-Fabra J., Ihlenfeldt S., Eguia J., Olarra A. 2013. “Machine
Tools for Large Parts”, CIRP Annals, Cilt:62, pp.731-750.
[2] Zhang, H., Wang, J., Zhang, G., Gan, Z., Pan, Z., Cui, H., & Zhu, Z. 2005. “Machining with flexible
manipulator: toward improving robotic machining performance”, Advanced Intelligent Mechatronics.
Proceedings, pp. 1127-1132.
[3] Abele E., Weigold M., RothenBücher S., 2007. “Modelling and Identification of an Industrial Robot
for Machining Applications”, CIRP Annals, Cilt:56, pp.387-390.
[4] Slavkovic, N. R., Milutinovic, D.S., Glavonjic, M.M., 2014. “A method for off-line compensation of
cutting force-induced errors in robotic machining by tool path modification”, International Journal of
Advanced Manufacturing Technology, Cilt:70, pp.2083-2096.
[5] Dumas C., Caro, S., Garnier, S., and Furet, B., 2011, Joint Stiffness identification of six-revolute
industrial serial robots, Robotics and Computer-Integrated Manufacturing, Vol: 27, pp.881-888.
[6] Pan Z., Zhang H., Zhu Z., Wang J., 2006. “Chatter Analysis of Robotic Machining Process, Journal of
Materials Processing Technology”, Cilt:173, pp.301-309.
[7] Zaghbani I., Songmene V., Bonev I., 2013. “An Experimental Study on the Vibration Response of a
Robotic Machining System”, Proceedings of the Institution of Mechanical Engineers, Part B: Journal of
Engineering Manufacture, Cilt:227, Sayı:6, pp.866-880.
[8] Budak E., 2006. “Analytical Models for High Performance Milling, Part II: Process Dynamics and
Stability”, International Journal of Machine Tool & Manufacture, p.1489-1499.
[9] Tunc, L. T., & Shaw, J., 2016. “Investigation of the effects of Stewart platform-type industrial robot
on stability of robotic milling” The International Journal of Advanced Manufacturing Technology, Cilt:87
pp.189-199.
[10] Tunc, L. T., & Shaw, J., 2016. Experimental study on investigation of dynamics of hexapod robot for
mobile machining” The International Journal of Advanced Manufacturing Technology, Cilt:84, pp.817-830.
[11] ISO 9283:1998 Preview Manipulating industrial robots -- Performance criteria and related test
methods.

2-4 NOVEMBER 2017 | 8th INTERNATIONAL SYMPOSIUM ON MACHINING
17
[12] FANUC, 2002, FANUC Robot series R-30iA Handling Tool, Operator’s Manual, User Manual,
FANUC LTD.