CIRP Annals - Manufacturing Technology: E. Budak (1), E. Ozturk

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

CIRP Annals - Manufacturing Technology 60 (2011) 383–386

Contents lists available at ScienceDirect

CIRP Annals - Manufacturing Technology


jou rnal homep age : ht t p: // ees .e lse vi er. com/ci rp/ def a ult . asp

Dynamics and stability of parallel turning operations


E. Budak (1)*, E. Ozturk
Manufacturing Research Laboratory, Sabanci University, Tuzla, Istanbul, Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: Parallel turning offers increased productivity due to multiple cutting tools in operation. The dynamic
Chatter
interaction between the tools needs to be analyzed as it affects the stability of the process. In this study,
Stability
dynamics and stability of parallel turning processes are modelled. The results of the developed stability
Parallel turning
models in frequency and time domains show reasonable agreement. One of the interesting outcomes is
that the stability could be increased due to dynamic interaction between the tools creating an absorber
effect on each other. The predicted stability limits are compared with experimental results where
reasonable agreement is demonstrated.
ß 2011 CIRP.

1. Introduction coupling decreased the stability limits. Later, Ozdoganlar and


Endres [7] developed a parallel turning process on a modified
Parallel turning refers to turning operations with multiple vertical milling machine where they cut different surfaces. They
cutting tools. Use of multiple tools offers potential for increased employed a flexible plate to achieve dynamic interaction between
material removal rate provided that appropriate cutting conditions the tools. They developed an analytical solution for dynamically
are used. Selection of stable conditions to avoid chatter vibrations symmetric systems where the cutting tools have the same transfer
is an important part of the process planning. In order to understand functions, and validated the developed solution through experi-
the dynamics of parallel turning processes the interaction between mental results. Tang et al. [8] developed an optimization tool that
the cutting tools need to be studied. When parallel tools cut the considers constraints such as cutting forces, spindle power and tool
same surface simultaneously they share the static chip thickness, life in parallel turning. They predicted the cutting forces, spindle
i.e. feed/revolution, similar to milling cutters. Hence, each tool power and tool life using empirical equations but they did not take
experiences less cutting force. Moreover, the dynamical interac- the stability limits into account. An optimization tool that can also
tion between the tools may decrease or increase the stability of the take the stability as an additional constraint can be quite useful for
process. In some cases, parallel tools may be more stable compared optimization of parallel turning processes.
to single tool turning processes. As a result, a stability model for In this paper, the dynamics of parallel turning with two tools
parallel turning is required for detailed understanding and cutting the same surface is formulated. A frequency domain model
planning of these operations. to determine the stability diagrams for parallel processes is
There have been considerable amount of work for modelling presented in the next section. A time domain model is described in
dynamics of single tool turning processes. Tobias and Fishwick [1], Section 3 to verify the predictions of the frequency domain model.
and Tlusty and Polacek [2] are the pioneering researchers who Finally, the model’s predictions are compared with experimental
modelled the orthogonal cutting stability by identifying the results.
regenerative effects causing chatter vibrations. Later, Tlusty and
Ismail [3] developed time domain stability models for more 2. Frequency domain model
accurate prediction of stability limits. Rao and Shin formulated [4]
the stability of oblique turning operations by considering the 2.1. Stability formulation
multi-dimensional dynamics in both radial and axial directions. In
another study, Budak and Ozlu [5] showed that multi-dimensional A parallel turning process with two tools that cut the same
stability solution should be employed when inclination angle or surface is illustrated in Fig. 1. The tools perform outer diameter
nose radius exists on the tool. turning operation in parallel in this case. Since they are cutting the
There have been only a few studies on the stability of parallel same surface, feed velocity of both of the tools should be set equal
turning processes. Lazoglu et al. [6] developed a time domain whereas their cutting depths can be different. According to the
model for parallel turning where each tool cuts a different surface. notation used in the model, the tool with a higher cutting depth is
There is no direct interaction between the tools in the presented named as the second tool and the other one is the first tool (Fig. 2).
case; however a dynamic coupling between the tools occurs The cutting depths of tool 1 and tool 2 are represented by a1 and a2,
through the flexible workpiece. They showed that the dynamic respectively.
The stability formulation in the paper is kept in one dimension
[5] since the aim of the paper is to focus on the effects of the
* Corresponding author. parallel turning. For that reason, turning tools without side cutting

0007-8506/$ – see front matter ß 2011 CIRP.


doi:10.1016/j.cirp.2011.03.028
[()TD$FIG]
384 E. Budak, E. Ozturk / CIRP Annals - Manufacturing Technology 60 (2011) 383–386

for region a1 as h21 and for the region (a2  a1) as h22 as follows:
  "  t #
h21 ðtÞ z2 ðtÞ þ z1 t 
½h2 ðtÞ ¼ ¼ 2 (2)
h22 ðtÞ z ðtÞ þ z ðt  t Þ
2 2

After dynamic chip thicknesses on tools are defined, dynamic


cutting forces on them can be determined:
   
F 1 ðtÞ a h ðtÞ
¼Kf 1 1 (3)
F 2 ðtÞ a1 h21 ðtÞ þ ða2  a1 Þh22 ðtÞ

where Kf is the cutting force coefficient in the feed direction. Edge


forces are neglected in the stability formulation since they do not
contribute to the regenerative process.
Fig. 1. Parallel turning process. Dynamic displacements (z1 and z2) can be calculated in terms of
cutting forces and transfer functions (G11, G12, G21 and G22) as
edge and oblique angles are considered. In this case, only the follows:
    
displacements of the tools in the feed direction affect the z1 ðtÞ G11 G12 F 1 ðtÞ
regeneration mechanism and one-dimensional stability formula- ¼ (4)
z2 ðtÞ G21 G22 F 2 ðtÞ
tion can accurately represent the system. However, the formula-
tion can easily be extended to three-dimensional cases. The dynamic displacements and cutting forces when the system
The tools can be modelled as being attached to a rigid surface of is marginally stable can be expressed as
the machine with springs (k1, k2) and damping elements (b1, b2) as        
z1 ðtÞ z F 1 ðtÞ F 1 ivc t
shown in Fig. 2. The cross interaction between the turning tools is ¼ 1 eivc t ; ¼ e (5)
z2 ðtÞ z2 F 2 ðtÞ F2
also possible but they are not shown in Fig. 2 in order not to
complicate the figure. The dynamics of the workpiece can simply Marginal stability refers to the transition phase between the
be included in the formulation. However, it is neglected here since stable region and the unstable region. Since two tools are
the workpiece is considerably rigid with respect to the cutting interacting with each other dynamically, they vibrate with the
tools along its axial direction, i.e. the Z-axis. same chatter frequency vc. Additionally, the dynamic displace-
There can be some offset in the axial direction between the ment values in the previous rotation at the limit of stability can be
cutting tools and it should be kept at minimum. The tools cut the written as shown in Eq. (6). Moreover, the displacements at one-
same surface (i.e. a tool removes the surface left by the other one) half rotation period before can be calculated in a similar way:
only if the offset is less than half of the feed per rev.    
z1 ðt  t Þ z ðtÞ ivc t
If cutting depths of the tool are not equal, there are two different ¼ 1 e (6)
z2 ðt  t Þ z2 ðtÞ
regions with different dynamic characteristics in the process. The
region with depth of a1 is removed by both of the tools as shown in After the presented formulations are substituted into Eq. (3)
Fig. 2. In this region, dynamic chip thickness on a tool is affected by and re-arranged, the cutting forces at the limit of stability can be
the displacement of the tool at present time and the displacement written as follows:
of the other tool at a half rotation period (t/2) before. The feed per    
F 1 ivc t F
revolution is shared between the tools in this region as the static e ¼ B 1 eivc t (7)
F2 F2
chip thickness. On the other hand, the region with a depth of
a2  a1 is only removed by the second tool. Hence, the dynamic The closed form of B matrix is a lengthy expression that depends on
chip thickness depends on the dynamic displacement of the second Kf, a1, a2, G11, G12, G21, G22, vc and t.
tool at present time and at one rotational period (t) before. The
static chip thickness on the second tool is equal to the feed per 2.2. Calculation of stability limits
revolution in this region. Since the static chip thicknesses on the
tools do not affect the regeneration mechanism, they can be The relations developed for dynamic cutting forces in Eq. (7)
removed from the stability formulation. Hence, dynamic chip take the following form after some arrangements:
thickness on the first tool can be represented as follows:  
F
h  ½I  B 1 ¼ 0 (8)
t i F2
½h1 ðtÞ ¼ z1 ðtÞ þ z2 t  (1)
2
where I is the 2  2 identity matrix. In order to have non-trivial
where z1 and z2 represent the displacements of the first and the solutions of the equation, the determinant of [I  B] matrix should
second tool, respectively, and t stands for the present time. be zero. The determinant results in a complex equation with
Moreover, the dynamic chip thickness on the second tool is defined variables a1, a2, vc and t. When the real and imaginary parts of the
[()TD$FIG] equation are equated to zero, two independent equations are
obtained (Eq. (9)) [9]. Since the resulting equations are lengthy,
they are presented symbolically as follows:
Realðdet½I  BÞ ¼ 0
(9)
Imagðdet½I  BÞ ¼ 0

One of the cutting depths should be set before calculation of the


stability diagram for the other tool. In this study, the stability
diagram for a1 is generated for a given a2. But it should be
remembered that a2 is assumed to be higher than a1 in this
formulation. Hence, only the stability limit values for a1 which are
less than a2 should be considered as solution. Since a2 is now set,
there are three unknowns, namely a1, vc and t, in the formulation
whereas there are only two independent equations at hand. For
that reason, one of the variables need to be given in a range in order
Fig. 2. Notations in parallel turning. to solve for the other two unknowns.
E. Budak, E. Ozturk / CIRP Annals - Manufacturing Technology 60 (2011) 383–386 385

Cutting depth a1 is solved in terms of vc and t using the real part Table 1
Modal data of the tools.
of the complex equation in Eq. (9), and this relation is substituted
into the imaginary part of the complex equation. Hence, a1 term is Mode fn (Hz) z (%) k (N/m)
eliminated, and the imaginary part of the complex equation is G11 1 1688.1 3.85 1.495  107
obtained with 2 parameters, vc and t, only. The resulting equation 2 2060.2 0.87 2.483  108
includes many trigonometric functions, and thus a closed form G22 1 1922.1 4.72 6.429  106
analytical solution for vc or t is not possible to obtain. Hence, a
search algorithm, named as golden section search [10], is used to
solve t for a given vc. Firstly, a chatter frequency range and a 1050 steel. The frequency response functions of the turning tools
frequency increment (Dv) are selected. Since chatter frequencies are measured by tap testing as shown in Fig. 3.
(vc) are expected to be close to the natural frequency of the tools, The edge and cutting force coefficients in the feed direction are
the selected range should contain all significant natural frequen- calibrated mechanistically using the linear-edge force model [11]
cies of the system. Then, the spindle speeds are swept with Dn as 86.5 N/mm and 1100 MPa, respectively. The calibration was
increments for a given chatter frequency (vc). Each spindle speed n performed for 200 m/min cutting speed and feed range of 0.04–
corresponds to a rotational period t by n = 60/t. For each vc and t 0.13 mm. In some tests, the exact cutting speeds on the tools are
pair, the imaginary part of the complex equation in Eq. (9) is not equal since their cutting depths are different. However, since
calculated. If there is a sign change between consecutive t values, it the ratio of radius of the workpiece to cutting depths is usually very
means that a root of the equation is bracketed. Then, using the large, the cutting speeds can be assumed equal and the same force
golden section search [10], the spindle speed value that satisfies coefficients can be used in the simulations for both tools.
the equation is identified with a preset tolerance. The direct transfer functions of both tools are measured. There
For each chatter frequency, more than one spindle speed is was no coupling between the tools since they are located on
determined corresponding to different lobe numbers in the different turrets, and thus cross FRFs were neglected. Modal data fit
stability diagram. Using the calculated rotational periods and to the measured transfer functions is tabulated in Table 1.
given chatter frequencies, a1 values are calculated by the real part The stability diagrams for both of the tools when each of the
of Eq. (9). vc and t pairs resulting in negative a1 values are tool cuts alone (in single mode) are calculated using the available
eliminated from the solutions. Finally, the stability diagram can be orthogonal stability model [9]. For turning processes with single
obtained by plotting a1 with respect to the spindle speed. tool, the absolute stability of the first and second tool are 1.1 mm
Since a search algorithm is employed to obtain the stability and 0.6 mm, respectively. For parallel turning operations, where
diagrams, increments in the frequency and spindle speed ranges, the two tool cut the same surface, stability diagrams for a1 are
which are represented by Dv and Dn, have considerable effects on calculated for different values of a2. It is observed that when a2 is
the stability diagrams. Hence, they should be selected small selected higher than second tool’s absolute stability limit for the
enough until a convergence in the solution is obtained. single tool operation (i.e. when a2 value is selected from the
unstable region for the single tool process), two limits are observed
in the stability diagram of a1 instead of just one limit. This is
3. Time domain model
presented in Fig. 4. Between these two limits, parallel process is
stable otherwise it is unstable. The predictions of the frequency
The cutting forces on the turning tools result in displacements
domain model are also compared with time-domain model. Three
of the tools. These displacements result in changes in chip
points are selected in Fig. 4 to represent the two limit case and time
thickness values of the tools. Since cutting forces depend on chip
domain model’s results are presented for these three points in
thickness, cutting forces are also affected. The time domain model
Fig. 5. Variation of the displacement of the second tool is presented
working in Matlab/Simulink environment simulates this closed
for certain number of revolutions in Fig. 5. Time domain model’s
loop at discrete time intervals for about 50 revolutions. It employs
the equations for chip thickness, cutting forces and tool displace- [()TD$FIG]
ments presented in Section 2. Although they do not affect the
stability limits, edge forces and static chip thickness are also
included in the time-domain simulations. It is observed that
sampling rate should be selected such that there are at least 100
simulation points in one chatter wave. Otherwise, the model may
not accurately predict the dynamics of the process. For solution of
the differential equations Runge-Kutta method [10] is employed.

4. Experiments and simulations


Fig. 4. Stability diagram for a1 when a2 is 1.5 mm (points: (f) a1 = 1.5 mm, (e)
An Index ABC machine tool is used in the tests. The turning a1 = 1 mm, (d) a1 = 0.5 mm).
inserts are selected as Taegutec TPGN 110304 TT8115. The
[()TD$FIG]
[()TD$FIG]
workpiece material is a 32 mm diameter cylinder made out of

Fig. 5. Time domain solution of the variation of the 2nd tool’s displacement (z2) at 3
Fig. 3. Measurement of FRFs of the first turning tool. different a1 values when a2 is 1.5 mm.
[()TD$FIG]
386 E. Budak, E. Ozturk / CIRP Annals - Manufacturing Technology 60 (2011) 383–386

shown that tests with a1 = 1.5 mm and a1 = 0 mm result in poor


surface quality with chatter marks, while the test with a1 = 0.5 mm
creates a chatter free surface. The chatter frequency around
2100 Hz is easily visible in the sound spectrum data presented in
Fig. 7 for the unstable tests.

5. Conclusion

A frequency domain model developed for calculation of


stability diagrams in parallel turning is presented. A time domain
model is also developed to verify the stability limit predictions of
the frequency domain model. The results of both models agree
with each other. It is demonstrated that dynamical interaction
between the tools can work in favour of increasing the stability
limits compared to turning operations with single tool. This effect
Fig. 6. Stability limits for a1 predicted by frequency domain model at 2100 rpm for is especially seen when the natural frequencies of the turning tools
given a2 values. Experimental results are shown by markers.
are close to each other. The models’ predictions are compared with
[()TD$FIG] experimental results. Although there are some discrepancies, it can
be concluded that the general agreement is reasonable considering
the sensitivity of the predictions to the measurement errors in
transfer functions and cutting force coefficients. Stability diagram
with two stability limits which is not an intuitive phenomenon is
demonstrated by both simulations and experiments. As a next step,
the presented stability formulation is to be extended by including
the dynamics in the radial direction to simulate more general
parallel turning processes.

Acknowledgements
Fig. 7. Surface photos and spectrum of sound data for 3 different a1 values at
2100 rpm when a2 is 1 mm. The authors thank to TUBITAK (grant 108M340) and Pratt&-
Whitney Canada for their support.
results also confirm the existence of two limits in the stability
diagram. As can be seen from Fig. 5, point d (a1 = 0.5 mm) and
f (a1 = 1.5 mm) are unstable, while point e (a1 = 1 mm) is stable. References
Frequency domain model was run for different a2 values and
[1] Tobias SA, Fishwick W (1958) The Chatter of Lathe Tools under Orthogonal
predicted lower and higher stability limits at 2100 rpm are
Cutting Conditions. Transactions on ASME 80:1079–1088.
presented in Fig. 6. When a2 is less than 0.6 mm which is the [2] Tlusty J, Polacek M (1963) The Stability of Machine Tools Against Self Excited
predicted stability limit for the single turning operation with tool Vibrations in Machining. International Research in Production Engineering ASME
2, the lower limit becomes zero. As a2 increases lower and higher 465–474.
[3] Tlusty J, Ismail F (1981) Basic Non-linearity in Machining Chatter. Annals of the
limits converge each other and they become equal around 2 mm CIRP 30:21–25.
when a2 is 3.1 mm. For higher values of a2, the process becomes [4] Rao BC, Shin YC (1999) A Comprehensive Dynamic Cutting Force Model for
unstable regardless of the value of a1. Moreover, it should be noted Chatter Prediction in Turning. International Journal of Machine Tools & Man-
ufacture 39:1631–1654.
that predicted stability limits for a1 are not real solutions if they are [5] Ozlu E, Budak E (2007) Analytical Modeling of Chatter Stability in Turning and
higher than a2 since a2 is assumed to be equal or greater than a1 in Boring Operations. Part I. Model Development. Journal of Manufacturing Science
the formulation. For that reason, the stability limits above a1 = a2 and Engineering 129:726.
[6] Lazoglu I, Vogler M, Kapoor SG, DeVor RE (1998) Dynamics of the Simultaneous
line, which is also presented in Fig. 6, should not be taken into Turning Process. Transactions of the North American Manufacturing Research
consideration. Conference, NAMRC XXVI, 135–140.
Experimental cuts have been performed at 2100 rpm for [7] Ozdoganlar OB, Endres WJ (1999) Parallel-process (Simultaneous) Machining
and its Stability, Presented at ASME IMECE’99, Nashville, TN. Proc., Symp. on
different a1 and a2 values. By analyzing the measured sound data
Mach. Sci. and Tech., MED-10, 361–368.
and resulting surface quality, the tests are classified as chatter, [8] Tang L, Landers RG, Balakrishnan SN (2008) Parallel Turning Process Parameter
marginal and stable. Experimental results are also demonstrated in Optimization Based on a Novel Heuristic Approach. Journal of Manufacturing
Science and Engineering 130:031002–31011.
Fig. 6. Although there are some discrepancies between experi-
[9] Altintas Y (2000) Manufacturing Automation: Metal Cutting Mechanics, Machine
ments and frequency domain model, the agreement is acceptable. Tool Vibrations and CNC Design. Cambridge University Press.
Moreover, as a result of the experimental tests, it is confirmed that [10] Press WH, Teukolsky SA, Vetterling WT, Flannery BP (2002) Numerical Recipes
stability diagrams with two stability limits exist for certain a2 in C++: The Art of Scientific Computing. Cambridge University Press.
[11] Budak E, Altintas Y, Armarego EJA (1996) Prediction of Milling Force Coeffi-
values. To be representative, surface photos and sound spectrum of cients from Orthogonal Cutting Data. Transactions of the ASME Journal of
three cutting tests are presented in Fig. 7. When a2 is 1 mm, it is Manufacturing Science and Engineering 118:216–224.

You might also like