Professional Documents
Culture Documents
a r t i c l e in fo abstract
Article history: Experimental measurements of cutting and thrust forces and chip shear plane angles have been carried
Received 4 September 2008 out for six steels turned at a feed of 0.25 mm/rev at speeds from 50 to 250 m/min. The results have been
Received in revised form successfully reproduced by finite element simulations. For this it has been necessary to include an
25 March 2009
initial yield drop from an upper to a lower yield stress in the description of the steels’ plastic strain
Accepted 6 April 2009
Available online 17 April 2009
hardening. It has also been necessary to assume flow stress to reduce non-linearly with increasing
temperature in the manner proposed by Zerilli and Armstrong, up to a temperature E900 1C above
Keywords: which rapid softening takes place. A comparison is made between the present work and the earlier work
FE analysis of Oxley and his group.
Plasticity
& 2009 Elsevier Ltd. All rights reserved.
Yield delay
Thermal softening
Metal machining
1. Introduction after the manner introduced by Zerilli and Armstrong [3]. Their
work and developments from it used in this paper are reviewed
Part I of this paper [1] showed that finite element simulations in Section 2.
of the machining of a softened carbon steel could be brought into Agreements between experiments and simulations are
improved agreement with experiments by including a description checked in three different ways. Most directly, the predicted and
of an upper yield point in the steel’s constitutive equation. This experimental quantities FC, FT and f are compared. These are,
was presented as a general result, applicable to all such steels. It respectively, the cutting and thrust forces per unit cutting edge
was not attempted to predict the behaviour of any particular steel. engagement length and shear plane angle. f is obtained (Eq. (1))
That would have required a consideration of its flow stress depen- from the chip thickness ratio t/ac, where t is the chip thickness and
dence on temperature and strain rate too. ac the undeformed chip thickness, and the tool rake angle g. The
In this, Part II, paper, attention is turned to simulating the predicted and experimental dependences of (fg) on l are also
behaviour of six particular steels. The paper includes experi- compared. l is the direction between the resultant force and
mental measurements of the steels’ flow stress dependences on the normal to the rake face i.e. the friction angle that may be
strain, strain rate and temperature. Flow stress at high strain rate calculated from Eq. (2). This method of comparison is helpful
reducing linearly with temperature was observed, as is frequently because it gives guidance as to reasons for differences between
assumed in machining simulations and was first reported by simulated and experimental values of FC, FT and f, should there be
Johnson and Cook [2]. The data were at first fitted to the power- any [4]. Finally, comparisons are made between predicted and
law (PL) form of constitutive equation (Eqs. (3) and (4) of Part I). experimental shear stresses k calculated to act on the primary
Simulations with this constitutive equation were not in agree- shear plane (Eq. (3)), also because of insights that this gives.
ment with experiments. It was also observed that predicted
t=ac ¼ cosðf gÞ= sin f (1)
temperatures were higher than those used in the flow stress
determinations. Excellent agreement with experiments could be
obtained by both including the upper yield point and modifying l ¼ tan1 ðF T =F C Þ þ g (2)
the high temperature dependence of flow stress on temperature
k ¼ ½ðF C cos f F T sin fÞ sin f=ac (3)
Corresponding author. Tel.: +44 113 34 32165; fax: +44 113 34 32150. It might be thought that there is no longer a need for research
E-mail address: t.h.c.childs@leeds.ac.uk (T.H.C. Childs).
into the prediction of chip form and forces in the orthogonal
1
Current address: Department of Manufacturing and Industry, Universiti Tun machining of carbon steels. However, in much previous finite
Hussein Onn, Malaysia. element work, for example [5–8], agreement between simulation
0020-7403/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmecsci.2009.04.001
ARTICLE IN PRESS
466 T.H.C. Childs, R. Rahmad / International Journal of Mechanical Sciences 51 (2009) 465–472
and experiment has been reported in only up to two out of three s̄ ¼ C 1 þ C 5 ¯ c1=n þ C 2 exp C 3 þ C 4 ln _¯ T abs ; ¯ 4c (4d)
of cutting and thrust force and shear plane angle; or it has not
been attempted to obtain all three. It is common that if the
s̄ ¼ su ; ¯ pu (4e)
(active) cutting force is correctly predicted, the (passive) thrust
force is not. This paper’s contribution is to present simulations in There is published work on typical values of the coefficients
which all three are obtained in near-agreement with experiment. C1–C5, for pure iron and carbon steels. Values of C3 and C4 that
It also proposes a practical way to obtain the constitutive equa- control temperature and strain-rate dependence are shown
tions of other steels that have not been studied here that is the in Table 1. In this paper, simulations are presented using
necessary input to the simulations. Zerilli–Armstrong (ZA) constitutive equations. C3 and C4 have
There is an earlier analytical machining literature, associated been assumed to be the same for all the steels, namely mean
with the name of Oxley [9], with particular researches jointly with values from Table 1, 5.6 103 and 2.7 104, respectively. The
Hastings et al. [10] particularly relevant to this paper. The present other coefficients have been obtained experimentally. Simulations
paper’s results in relation to these earlier studies are discussed. are also presented using the power-law equations (3) and (4) from
Part I. Then, in addition to those with ‘as measured’ constitutive
equation coefficients, thermal softening coefficients have been
2. Temperature and strain-rate modelling chosen to mimic Zerilli–Armstrong temperature dependence.
Table 2
Work material compositions and thermo-physical properties at 20 1C.
C Si Mn P S Cr Mo Ni Pb
070M20 0.20 0.25 0.70 0.05 0.05 0.30 0.15 0.40 – 52 490
080M40 0.40 0.25 0.80 0.05 0.05 0.30 0.15 0.40 – 48 480
080M46 0.47 0.25 0.80 0.05 0.05 0.30 0.15 0.40 – 48 480
9SMn36 0.09 0.01 1.01 0.06 0.33 0.03 0.01 0.04 – 55 470
9SMnPb36 0.08 0.04 1.49 0.07 0.42 0.02 0.00 0.02 0.27 55 470
SAE1144 0.44 0.13 1.45 0.01 0.30 0.15 0.00 0.02 – 47 490
ARTICLE IN PRESS
T.H.C. Childs, R. Rahmad / International Journal of Mechanical Sciences 51 (2009) 465–472 467
respectively, from the measured circumferential force and by 4.1. Physical experiments
resolving the measured axial and radial components normal to the
cutting edge. Mechanical testing: Figs. 1a, and b present the combined tensile
Mechanical testing: Tensile test pieces were made from all six testing and PSC results for the plain and free-machining carbon
steels, to obtain their lower yield stress and initial strain- steels. Fig. 1b in addition presents the HB data for the free-
hardening behaviour at low strain rate and room temperature. machining steels. Fig. 2 shows results from the elevated temper-
Plane strain compression (PSC) samples were also made, to extend ature HB tests, with two further sets of data from previously
equivalent stress, equivalent strain behaviour up to a strain E1. published work [20,21]. It plots the flow stress at temperature,
These tests were carried out incrementally, with lubrication relative to the flow stress at 20 1C, for ¯ ¼ 1 and _¯ ¼ 103 =s. The
between increments, and changing platens to keep platen width dashed line shows the linear reduction to zero as temperature
to sample thickness in the range 2–4, as is recommended [17]. increases from ambient to the melting temperature, as was pro-
Hopkinson bar (HB) test results were available for the free- posed, based on experiments, by Johnson and Cook [2].
machining steels, from a previous project. These had been carried The values of the coefficients in the power-law constitutive
out at room temperature, at a strain rate of 103/s, incrementally, model (Part I, Eqs. (3) and (4)), based on Figs. 1 and 2, are
in compressive strain steps of 0.05, up to a final strain of 1.0, with presented in Table 3. Coefficient values C1, C2, C5, n and ¯ c for the
lubrication between steps. They had also been carried out at tem- Zerilli–Armstrong constitutive model (Eqs. (4a) and (4d)) are
peratures from room temperature to 700 1C, with rapid heating entered in Table 4 (C3 and C4, not included in the table, are the
and cooling before and after each step, as described in [18]. mean values from Table 1). How the coefficient values were
obtained is described in Appendix A1.
3.2. Simulations Machining tests: Representative experimentally determined
cutting and thrust forces and shear plane angles are distributed
Simulations were performed at the three cutting speeds 50, among Figs. 3, 6 and 7, with other data in Appendix A2.
150, 250 m/min and ac ¼ 0.25 mm. The work material was given
nominal room temperature thermal properties (K ¼ 52 W/m K,
C ¼ 490 J/kg K, Table 2) and a density of 7860 kg/m3. The cutting
1000
tool was given a rake angle of 61 and a clearance angle of 101. Its
cutting edge radius was 20 mm. The software did not allow free
choice of the tool’s thermal conductivity (as explained in Part I).
An option was chosen with K estimated to be 75 W/mK, as close 800
Equivalent stress (MPa)
Experiment
1.0 Simulation PL1 simulation ZA1
1400 Simulation PL4 Simulation ZA2 28
9SMn36
0.8 9SMnPb36 1200 24
SAE1144
FC, FT (N/mm)
708M40 [21] φ
800 16
0.4
600 FC 12
0.0 200 4
0 200 400 600 800 1000 1200 1400 1600
Temperature (°C) 0 0
0 50 100 150 200 250
Fig. 2. Flow stresses relative to ambient values, ¯ ¼ 1:0 and _¯ ¼ 103 s1 . Cutting speed (m/min)
Table 3
Power-law (PL) coefficient values from mechanical testing. 600
15
4.2. Simulations
Experiment
The simulations of 070M20 machining are summarised in Fig. 3.
10 Simulation PL1
Fig. 3a shows that the simulations PL1 and ZA1 (with materials
Simulation PL2 Simulation ZA1
data as in Tables 3 and 4) give rise, respectively, to shear plane
Simulation PL3 Simulation ZA2
angles larger than and smaller than experimental. Their cutting
Simulation PL4 Simulation ZA3
forces are, respectively, smaller than and larger than experimental. 5
The thrust force for PL1 is much less than experimental, whereas 15 20 25 30 35 40 45
that for ZA1 is in agreement with experiment. λ (°)
Fig. 3b shows that both simulations give primary plane shear
stresses in agreement with experiment. Fig. 3c clearly shows differ- Fig. 3. 070M20 experimental and simulated dependencies of (a) forces and shear
plane angle and (b) primary plane shear stress on cutting speed; (c) (fg) on l.
ences between the two and from experiment. PL1 shows both lower
values and a wider range of friction angles l than ZA1. ZA1’s l
values are approximately in agreement with experiment. Inspection
showed that the differences between PL1 and ZA1 could largely be Fig. 3c shows simulation PL2 to give almost identical results to
attributed to their different thermal softening behaviours. ZA1. Both have shear plane angles lower than experiment.
The simulation PL2 was therefore created with thermal soften- Therefore (Table 5) simulations PL3, 4 and ZA2 were created,
ing coefficients modified to bring its behaviour closer to ZA1. with upper yield points su ¼ 800 (PL3) and 1000 (PL4, ZA2) MPa
As shown in Table 5, c2 was given a non-zero value, also Tc, Tm (eu ¼ 0.05 in all cases). Fig. 3c shows these to give results closer to
were specified (see also Fig. 4 for a graphical interpretation). experiments than do PL2, ZA1. Fig. 3b shows their primary plane
ARTICLE IN PRESS
T.H.C. Childs, R. Rahmad / International Journal of Mechanical Sciences 51 (2009) 465–472 469
Table 5
Modified PL and ZA (Eqs. (4b) and (4c)) coefficient values for 070M20 simulations.
1000 a
Simulation c2 m _¯ 0 su (MPa) eu Tc (1C) Tm (1C) ET (1C)/Eref c
that of PL1, though its (fg) values are higher (due to its upper
yield stress value).
600 d
Insight into the thermal softening behaviours of the models
PL4, ZA2 and ZA3 is gained from Fig. 4. It shows, for tests at a b
cutting speed of 250 m/min., equivalent stress dependence on
400
temperature extracted from Gauss points that are plastically
deforming (as judged by a strain rate 410/s). In all there is a
low temperature part ‘a’ for which the equivalent stress is the
200
upper yield stress. There is a section ‘bc’ over which flow stress
increases with temperature, due to strain-hardening (the plastic
work causes the temperature rise). Then there is a section ‘cd’ in
0
which thermal and strain-rate effects dominate. In Fig. 4b, the
0 200 400 600 800 1000 1200 1400 1600
non-linearity of ‘cd’ is the natural result of the exponential term
Temperature (°C)
in the ZA equation. In Fig. 4a, the same characteristics have been
created by the choice of coefficients c1 and c2. In Fig. 4c, the influ-
ence of elastic modulus softening has made the equivalent stress
1000 a
thermal softening become almost linear with increasing tem-
perature. This is similar to the behaviour associated with the PL1 c
simulations.
800
The scatter in Fig. 4 results comes from the range of strain rates
Equivalent stress (MPa)
1000 a
Experiment
1200 36
Simulation PL4
c
Simulation ZA2
800 1000 30
Equivalent stress (MPa)
FC, FT (N/mm)
600 b A e
600 FC 18
400
400 12
FT
200
Simulation ZA2 200 6
from Hastings, Oxley, Stevenson [10]
0 0 0
200 400 600 800 1000 0 50 100 150 200 250
Velocity modified temperature, Tmod (K) Cutting speed (m/min)
Fig. 5. Data of Fig. 4b re-plotted against Tmod and compared with previous work.
Experiment
too low primary zone shear stresses rather than with discre- Simulation PL4
750
800 24
Appendix A1. Derivation of Tables 3 and 4 coefficients from
600 18
Table 3: s0 values were obtained from the tensile tests’ lower
FC yield stress and n from the slope of PSC log (stress)/log (strain)
400 12 plots. ¯ 0 values were then chosen to give the correct PSC stress
level at a strain of 1.0. ¯ c was taken to be 1.0. Strain-rate constants
m (m ¼ m1 ¼ m2) and _¯ 0 were obtained by comparing PSC and HB
200 FT 6
results (Fig. 1b). Inspection shows stress levels in HB tests
(_¯ ¼ 103 =s) to be E1.3 times those in the PSC tests (_¯ o102 =s).
0 0 Then, from the form of the PL’s strain-rate term, ð103 =_¯ 0 Þ1=m 1:3.
0 50 100 150 200 250 Possible (m, _¯ 0 s1 ) combinations range from (25, 1), through
(35, 101), to (45, 102). m ¼ 25 has been found to give too large a
Cutting speed (m/min)
strain-rate dependence in the present machining tests, where
Fig. 7. Experimental and simulated dependencies of forces and shear plane angle strain rates 4104/s have been generated. For this reason, (35, 101)
on cutting speed for steel 9SMn36. and (45, 102) have been considered, with (45, 102) shown in
Table 3, as explained next.
The actual value, 900 1C, depends on the accuracy of the simu- _¯ 0 has the physical significance that flow stress dependence on
lations’ temperature calculations. In that respect, it should be strain rate is much reduced at strain rates less than it. Although
remarked that it has been assumed that, at plastic strains 41, there is evidence that _¯ 0 101 at room temperature [12], the
100% of plastic work is converted to heat. This is believed to be combination of (m, _¯ 0 ) with _¯ 0 ¼ 102 gave slightly better agree-
more realistic than the common, alternative, view that E90% is ments with experiment in simulations of 070M20 machining
converted to heat, whatever the strain. However, the accurate (as is shown in Fig. 3b). (45, 102) is therefore entered in Table 3.
calculation of temperature and its experimental verification It is assumed to be the same for all the steels. Finally, Table 3’s
remains an open issue. thermal softening constants c0–c5 give the linear softening shown
Good agreements between simulations and experiments have by Fig. 2’s dashed line. In this case, there is no need to specify the
been obtained with both ZA- and PL-based constitutive equations cut-off temperature Tc.
although the ZA-based simulations have been marginally better Table 4: For all the steels, C5 is the PSC flow stress increase from
than the PL ones. However, the ZA form of constitutive equation ¯ ¼ 0 to 1 and n comes from the slope of the log(stress)/log(strain)
(for b.c.c. metals), adding strain hardening and temperature/ plot. (The n values differ from those of the PL descriptions because
strain-rate terms, makes the derivation of its coefficients more of the different form of the strain-hardening term.) For the free-
difficult than for the PL type of equation. From a practical point machining steels, C2 is obtained from the difference in stress
of view, it is more straightforward to use a PL-type equation, levels between the PSC and HB tests, after calculating the room
requiring only a steel’s room temperature and low strain-rate temperature values of the exponential term at strain rates of 103
strain hardening to be measured, assuming its temperature and and 102/s. Then C1 is determined such that at zero strain the
strain-rate dependencies to be the same as for every other steel. tensile test lower yield stress value is obtained. For the plain
With this assumption, the largest difference between measured carbon steels, in the absence of any high strain-rate test data, it
and simulated shear stresses on the primary shear plane has has been assumed that C2 for 070M20 is the same as for 9SMn36
occurred for the 080M40 steel (Fig. 6b), but this difference has (because of the similarity between their PSC behaviours) and for
only been E10%. The assumption would seem to be good to that 080M40 and M46 it is the same as for SAE1144. Then C1 has been
level of accuracy. calculated as for the free-machining steels.
Finally, the present work’s adoption of non-linear thermal
softening may be compared with the earlier (pre-finite-element)
work of Oxley and his colleagues. In [10] they consider, for a steel Appendix A2. Machining data (to 2 significant figures)
of similar carbon content to 070M20, flow in the high tempera- not presented in main text
ture conditions next to the rake face, which determines chip/tool
friction stresses. The dashed line in Fig. 5 is their flow stress at Experimental data and simulation data for 080M46, 9SMnPb36
unit strain determined from high-speed compression tests. The and SAE1144 steels are represented in Tables A1 and A2, respectively.
portion AB was attributed to blue-brittleness but the more strik-
ing feature is the overlapping of the present Zerilli–Armstrong and
References
[10] curves over the Tmod range 200–700 K. There are however
large differences at higher values of Tmod. It draws attention to
[1] Childs THC. Modelling orthogonal machining of carbon steels. Part I: strain
the need for more understanding of flow stress and friction in this hardening and yield delay effects. International Journal of Mechanical Sciences,
highest temperature range. in press, doi:10.1016/j.ijmecsci.2009.03.007.
[2] Johnson GR, Cook WH. A constitutive model and data for metals subjected to
large strains, high strain rates and high temperatures. In: Proceedings of the
seventh international symposium on ballistics, The Hague, Netherlands, 1983.
Acknowledgements p. 541–7.
[3] Zerilli FJ, Armstrong RW. Dislocation-mechanics-based constitutive relations for
The free-machining steels were from stock procured as part material dynamics calculations. Journal of Applied Physics 1987;61:1816–25.
[4] Childs THC. Numerical experiments on the influence of material and other
of the ECSC agreement No. 7210.MA/821. The Hopkinson bar tests variables on plane strain continuous chip formation in metal machining.
reported on these were carried out as part of that project, under International Journal of Mechanical Sciences 2006;48:307–22.
ARTICLE IN PRESS
472 T.H.C. Childs, R. Rahmad / International Journal of Mechanical Sciences 51 (2009) 465–472
Table A1
Experimental data for 080M46, 9SMnPb36 and SAE1144 steels.
Table A2
PL4 and ZA2 simulation data for 080M46, 9SMnPb36 and SAE1144 steels.
[5] Mamalis AG, Horvath M, Branis AS, Manolakas DE. Finite element simulation [14] Jaspers SPFC, Dautzenberg JH. Material behaviour in conditions similar to
of chip formation in orthogonal metal cutting. Journal of Materials Processing metal cutting: flow stress in the primary shear zone. Journal of Materials
Technology 2004;110:19–27. Processing Technology 2002;122:322–30.
[6] Shi G, Deng X, Shet C. A finite element study of the effect of friction in ortho- [15] Woolman J, Mottram RA. The mechanical and physical properties of the
gonal metal cutting. Finite Elements in Analysis and Design 2002;38:863–83. British Standard EN steels B.S. 970–1955, vol. 1. Oxford: Pergamon Press; 1964
[7] Ozel T. The influence of friction models on finite element simulations of En1–20.
machining. International Journal of Machine Tools and Manufacture 2006; [16] Brookes KJA. World directory and handbook of hardmetals and hard materials.
46:518–30. 5th ed. East Barnet: International Carbide Data; 1992.
[8] Sartkulvanich P, Altan T. Effect of flow stress and friction models in finite [17] Johnson W, Mellor PB. Engineering plasticity. Chichester: Ellis Horwood;
element simulation of orthogonal cutting – a sensitivity analysis. International 1983. p. 114–5.
Journal of Machining Science and Technology 2005;9:1–26. [18] Shirakashi T, Maekawa K, Usui E. Flow stress of low carbon steel at high
[9] Oxley PLB. Mechanics of machining. Chichester: Ellis Horwood; 1989. temperature and strain rate (Part 1) – propriety of incremental strain method
[10] Hastings WF, Oxley PLB, Stevenson MG. Predicting a material’s machining in impact compression test with rapid heating and cooling system. Bulletin of
characteristics using flow stress properties obtained from high-speed com- Japanese Society of Precision Engineering 1983;17:161–6.
pression tests. Proceedings of the Institution of Mechanical Engineers 1974; [19] Childs THC. Friction modelling in metal cutting. Wear 2006;260:310–8.
188:245–52. [20] Maekawa K, Kitagawa T, Childs THC. Effects of flow stress and friction char-
[11] Goldthorpe BD, Butler AL, Church P. A wide range constitutive equation for acteristics on the machinability of free cutting steels. In: Proceedings of the
medium and high strength steels. Journal de Physique IV 1994;4(C8):471–6. second international conference on the behaviour of materials in machining,
[12] Campbell JD, Ferguson WG. The temperature and strain-rate dependence of Institute of Metals (London) Book, vol. 543, 1991. p. 132–45.
the shear strength of mild steel. Philosophical Magazine 1970;21:63–82. [21] Childs THC, Maekawa K. Computer aided simulation and experimental
[13] Church P, Macmahon J. A comparison of 3D simulations against experiments studies of chip flow and tool wear in turning low alloy steels by cemented
for the torsion test. Journal de Physique IV 1997;7(C3):747–52. carbide tools. Wear 1990;139:235–50.