A Hybrid Cutting Force Model for High-speed Milling of Titanium Alloys

Z.G. Wang, M. Rahman (I), Y.S. Wong and X.P. Li Department of Mechanical Engineering, National University of Singapore, 10 Kent Ridge Crescent, Singapore, 119260
Abstract In this paper, the Johnson-Cook (JC) strength model is used to describe the flow stress of Ti6A14V and to estimate two important parameters in Oxley's model: the strain-rate constant and the angle made by the resultant force and the shear plane. The JC model is also incorporated into a finite element method (FEM) simulation for the deformation process of Ti6A14V. Finally, a hybrid cutting force model based on the FEM simulation and Oxley's theory is proposed to predict cutting forces when machining Ti6A14V. Experimental results are found to substantiate the developed model. Keywords: Cutting force model, finite element method, flow stress

1 INTRODUCTION High-speed milling (HSM) of difficult-to-cut materials, such as titanium alloys, has been facilitated by the development of advanced tool materials such as CBN, PCD and binderless CBN. However, the cost-effective application of HSM requires a fundamental understanding of its machining mechanism. It is costly to carry out large amount of experiments to investigate the cutting performance of HSM of titanium alloys, because the cutting tools and titanium alloys are very expensive. Modeling and simulation are indispensable for better understanding of HSM of titanium alloys. Many models have been developed to model and predict cutting forces in machining operations. Oxley [ I ] developed a more effective model, which considered the variation of flow stress properties in terms of the strain, strain-rate and temperature. This model assumes a thin shear zone, chip equilibrium and uniform shear stress in the secondary deformation zone at the tool-chip interface. Oxley's cutting force model is widely used to predict a comprehensive range of the machining characteristic factors. One major assumption in Oxley's model is that the normal stress at the tool-chip interface is assumed to be distributed uniformly. This assumption is ambiguous for high-speed machining. In this study, the axial depth of cut is smaller than the nose radius of the cutting tool. Therefore, Oxely's cutting force model requires some modification before it is applied. In numerical modeling, finite element method (FEM) techniques are the most commonly used tool [2] to predict chip flow, cutting forces and, especially the distribution of tool temperatures and stresses for various conditions. In this approach, the solution region is first divided into many smaller elements, so that various tool geometry, cutting conditions, and more sophisticated material and friction models can be incorporated [3]. Ozel and Altan [4, 51 have done much definitive works in this field. They developed a predictive model for HSM based on FEM simulations. Using their model, the resultant cutting forces, tool stresses and temperatures in turning and flat end milling were primarily predicted. More importantly, with fewer numbers of experiments, this method was able to estimate the variations of flow stress and friction conditions of HSM. In their model, the tooth-path was assumed to be circular. However, this approximation will cause some error for HSM. In this paper, a predictive force model for HSM is presented. In the proposed model, an analytical solution is employed to get an accurate uncut chip thickness in milling. Then, a 2-D equivalent element, which is more suitable for FEM simulation, is used to represent the nonuniform uncut chip area in 3-D milling. For the force

prediction, Oxley's cutting force model is used with two modifications: (1) The Johnson-Cook (JC) strength model of flow stress is used to describe the deformation behavior of Ti6A14V; (2) The value of the strain-rate constant C' is determined based on FEM simulation results. Thus the assumption about the uniform normal stress at the toolchip interface is not needed. 2 GEOMETRICAL MODELING OF MILLING PROCESS In milling, the undeformed chip thickness is determined by the distance between two consecutive cut surfaces. This distance varies with the instantaneous angle of immersion during the chip formation of a cutting tooth (insert) and is measured in the direction perpendicular to the preceding cut surface. In this study, only one flat end milling insert is used. For the convenience of analysis, the point on the tip of the insert is selected to describe the true tooth trajectory. The true tooth trajectories of two consecutive cut are the two solid curves shown in Figure 1, and the undeformed chip thickness is decided by these two curves. The position of the center of the cutter is located at points 0 and Of, when these two consecutive cuts start. When the tooth tip is at point B for the second cut, the angular position of the tooth is p. At this time, the undeformed chip thickness depends on the locations of points B and C, and the position of point C is determined by the tooth tip with the angular position 90for the first cut. When 9 = 90, the undeformed chip thickness is equal to the value of feed rate f. And when 0 < 9 < 90, let S=@-q, so @=q+S. Using the sine law for AACD, we get:
-

( 2 z - 6 )f l ( 2 z )

where R is the radius of the cutter. After simplification, the following relationship is obtained:

2zsin(b)+a6-2m = O

(2)

where a = f cos(p)/R After solving Eq.(2), the explicit expression of 6is given as:
a3 3+8z2 a4 +o(a5) 12z2 24z3 Then, the length of CD can be expressed as: -

6 = a - - - a1

2z

+- 3 + 2 z 2

(3)

CD = R C O S+ (~)/cos(~)

(4)

Finally, the instantaneous chip thickness at the position defined by angle p i s given by:

Z = R - ~ = R [ c o s ( ~ ) - c o s ( ~ + ~ ) ] / c o s( 5 () ~,)
Similarly, the undeformed chip thickness can be derived when 90' < p < 180'.

the cutting forces can be predicted if the shear angle #, shear stress and the strain rate constant C' are known. The procedure to estimate these three parameters is given as follows.

Workpiece

Figure 1: Geometry of chip thickness of the milling process. In this study, the nose radius of the indexable tooth is larger than the axial depth of cut so that different parts of the nose involved in cutting are under different cutting load. In order to investigate the effects of the nose geometry, it is necessary to simulate the cutting process around the tooth tip. For this purpose, there is a need to establish a model of the 3-D milling process, which requires longer computation time even for computers with faster computation speed. An alternative approach is to represent the 3-D milling with 2-D metal deformation process, which has been used by others [4, 51. Ozel [4, 51 transformed the uneven intersection surface of the undeformed chip to an equivalent one that is suitable for axisymmetric deformation simulation. As discussed in [ I ] , when the chip thickness is less than 0.05mm, the size effect is very obvious, and special care has to be taken to ensure that the cutting edge radius is much smaller than the value of undeformed chip thickness. In this study, the cutting edge radius is 0.015 mm; so it is difficult to claim that it is much smaller than the undeformed chip thickness, whose maximum value is 0.025 mm. Thus, Ozel's transformation method is not suitable in this study. Based on the observation of the geometry of uncut chip area, a new idea about equivalent element representation is used, as shown in Figure 2, where r is the nose radius and a is the axial depth of cut. The area of region I is equal to that of region II, so the uncut area can be represented with the equivalent element of Figure 2 (b). Then, the equivalent area is further simplified into a parallelogram (A'BCD') as shown in Figure 2 (c), where a, is equal to the uncut chip area divided by instantaneous chip thickness h(rp), which is calculated from Eq.(5). Thus, the undeformed chip around the nose is represented by the equivalent chip of uniform thickness. With this equivalent chip representation, the milling process can be assumed to be a facing process in turning with depth of cut a, and the changing feed-rate h(rp). Finally, the plain deformation of the turning process can be simulated with FEM.

Figure 3: Deformation zones of FEM simulation in machining of Ti6A14V. 3.1 Modeling of flow stress properties The yield stress of a metal under uniaxial conditions is defined as the flow stress or effective stress. The metal starts deforming plastically when the applied stress reaches the values of flow stress [4]. The flow stress is mostly influenced by temperature, strain, strain-rate and material properties. Accurate and reliable flow stress models are very important for describing the deformation behavior of the work material during practical machining processes. The widely used constitutive model of flow stress is the JC strength model, which was proposed by Johnson and Cook [6]. The JC model represents the flow stress 0 of a material as the product of strain, strain-rate and temperature, shown as the following equation: O = [A+B(~)n][l+Cln(r)][l-(& T-Ty )ml (6) SO Tm - T y The parameter A is the initial yield strength of the material at room temperature and at a strain-rate of 1 s-'; 2 is the equivalent plastic strain; is the strain-rate normalized by a reference strain-rate E0 . The temperature term is valid within the range from room temperature (TJ to melting temperature of the work material (T,,,). The parameters B, C, m and n are fitted to the experimental results obtained from the corresponding compression and tension tests. This model can be calibrated more easily. Therefore, some researchers chose the JC model as constitutive equation for deformation behavior of metals at higher strain rate and high temperature. Lee and Lin [7] investigated the deformation behavior of Ti6A14V using the split Hopkinson bar (SHPB). They fitted SHPB test results into the JC model at the strain-rate of 2x103 s-'. Meyer Jr. and Kleponis [8] also studied high strain rate behavior of Ti6A14V and low-cost titanium. Based on the published data listed in [7-101, the GaussNewton algorithm was used to find the parameter estimates for the JC model. The estimated parameters are listed in Table 1, and those parameters for the JC model found in [7, 81 are also shown in Table 1 for reference. In Oxley's model [I], flow stress in the shear plane zone
IQB

L ! .

Figure 2: Equivalent chip element with uniform chip thickness.

3 PREDICTION OF CUTTING FORCES A typical FEM simulation of the HSM of Ti6A14V is shown in Figure 3, where VC is the cutting speed, # is the shear angle. The area between the boundary CD and E f is the chip formation zone or the shear zone. In Oxley's theory,

can be calculated as kAB = c l E A B n

/J5.

This is

replaced by: kAB = O / & , where

o is

the effective flow

stress along AB, which can be calculated using the constitutive equation Eq. (6). Table 1 Parameters of JC constitutive model for Ti6A14V. Reference A B n C m Test Lee and Lin [7] 782.7 498.4 0.28 0.028 1.0 SHPB Meyer et al. [8] 862.5 331.2 0.34 0.0120 0.8 SHPB This approach 1 1165.51236.61 0.29 10.035510.42 -

In this approach, according to the chip formation model from [ I ] and the constitutive equation of flow stress for Ti6A14V, the value of C' is derived as follows. The change rate of flow stress (dWds2) normal to AB can be assumed to be only related to the actual strain-rate. Therefore, dklds2 can be derived using the following equation [I]: dk dk dy dk dy dt --(7) ds, dy ds, dy dt ds, where t is time The first term on the right-hand side of Eq. (7) can be obtained as:

workpiece combination with a rigid-viscoplastic material behavior. Kobayashi et al. [ I l l found that the governing equations for viscoplastic deformation are formally identical to those of plastic deformation, except that the effective stress is a function of strain, strain-rate and temperature. For the rigid-plastic materials, the deformation process is a boundary-value problem. Solutions to the boundary-value problem are the velocity distribution that satisfies the governing equations and the boundary conditions. For the boundary conditions, the velocity vector u is prescribed on a part of surface S, together with traction f on the remainder of the surface SF. The governing equations can be expressed mathematically by the following two equations [ I I ] :

n=
&=

S F
Sv

- SZu,dS

(16)

0 d V - F,G,dS+K $v&vdV
SV

S,

The second term on the right-hand side of Eq. (7) is the strain-rate. The last term is the reciprocal of the cutting speed normal to AB, which can be presented as: dt I dr, = 1l ( V c sin(#)) (9) By substituting for Eq. (7) with Eqs. (8), (9), and strain-rate equation in [ I ] , the following relation is obtained: (10) /[A+B(FAB)"] where I is the length of AB. According to the stress equilibrium equation along AB from [ I ] , the following relation exists:
ds2
--

is the where 0 is the effective stress in the region V; effective strain-rate; f , represents the surface tractions applied over the surface S; u, is admissible velocity; is the volumetric strain-rate; and K is a penalty constant For the analysis of heat transfer effect on deforming material, the energy balance equation can be expressed by [ I I ] :

sv

klT,,,6TdV - pT6TdV
SV SV

ko,i.,cTdV
SV

=0

dk

2kABC'nB(EAB)"

By applying the equation along AB, substituting for dklds2 from Eq. (10), the next equation is given:

where pa and p~ are the hydrostatic stresses at points A and B, respectively. Finally the value of C' is given by: (13) 2kABnB(E,B 1" In Oxley's theory, the angle 8 made by the resultant force R with the shear plane AB is expressed as: n A k l tane = 1+2(-- #) 4 2 k a~ s2 ~ By substituting for Eq. (14) with Eq. (lo), the following equation is obtained:

C ' = (PA -PB)[A+B(FAB)nl

tan 8 = 1 + 2(4

- #) -

C'n

BEAB" A + BEAB"

where kl is the thermal conductivity; T,,, is used for T,,,with the comma denoting differentiation and repeated subscript meaning summation (Laplace differential operator applied to temperature T ) ; p is the density; c is the specific heat; k is the heat generation efficiency. Equations (16), (17) and (18) are the basic equations for the finite-element discretization. FEM is used to obtain the closed form solution to the velocity distribution. Once the solution for the velocity field is obtained, then the corresponding stress can be calculated using the flow rule and the known mean stress distribution. One main problem for FEM simulation is the determination of the boundary condition. The boundary conditions along the cutting toollchip interface are very complicated. It is extremely difficult to determine the frictional stress at the toollworkpiece interface. Numerous works have been done to investigate the friction mechanism at the toollchip interface. Ozel [4] found that variable friction coefficients as a function of normal pressure at the tool rake surface can ensure more accurate results in FEM simulations. At the toollchip interface, two regions are considered on the tool rake face: sticking region and sliding region. In the sticking region, frictional stress is known to be equal to the local shear stress (kch,p). In the sliding region, shear stress decreases on the rake face, Coulomb's friction law can be applied to calculate the frictional stress. This friction model is also used is this approach. From FEM simulation, the value of C'can be calculated with Eq. (13), and not based on the assumption of uniform normal stress at the tool-chip interface used in Oxley's method. The following procedure is employed to estimate the cutting forces. Firstly, based on FEM simulation, the values of # and k a ~ are found at different instantaneous chip thickness. The required value of C'is then determined from Eq. (13), and the value of angle 8 is obtained from Eq. (15). Finally, the cutting forces are obtained based on Oxley's theory. With this procedure, the cutting forces can be calculated accurately and realistically.

3.2 FEM simulation of machining process In this study, FEM is used to simulate the cutting process, which is an extension of FEM application in the analysis of metal-forming. During practical machining, cutting heat is generated. Thus, the consideration of temperature effects in the analysis of plastic deformation during machining process is very important. So in cutting process, work materials deform viscoplastically under the cutting load. Although there still exists elastic deformation except viscoplastic one, viscoplastic strains outweigh elastic strains. So it is reasonable to assume the cutting tool and

4 VERIFICATION OF THE CUTTING FORCE MODEL In this study, a new tool material, which is binderless CBN, has been used to mill Ti6A14V at high cutting speed. The details of the tool material, cutting conditions and experimental setup are given in [12]. Figures 4-6 show the estimated and experimental cutting forces. Unlike the experimental cutting forces, there is much fluctuation in the estimated cutting forces. The possible reason for this may be the frequent remeshing and the discretization of the cutting area. Similar conclusions have also been drawn by Ozel [4]. After fitting the trends of the fluctuating estimated cutting forces, their mean values are obtained (the solid, dashed and dot lines in Figures. 4-6).

difference is not so large, so that the cutting forces can be predicted with good accuracy for all three directions. 5 CONCLUSIONS In this paper, a predictive force model for HSM is presented. In this model, an analytical solution has been derived to obtain accurate undeformed chip thickness in milling. Because of the effects of nose radius, the uncut chip area is non-uniform. Thus, the non-uniform uncut chip area in 3-D milling is represented with a 2-D equivalent element for more efficient FEM simulation. With this equivalent element representation, HSM of Ti6A14V is simulated with FEM based on the constitutive model for Ti6A14V. Based on FEM simulation results and Oxley's theory, a new hybrid cutting force model is proposed. In the proposed model, a more sophisticated friction model at tool-chip interface is incorporated, and the assumption of uniform distribution of the normal stress at the tool-chip interface used in Oxley's model can be eliminated. Finally, experimental milling tests have been conducted to verify the developed force model. The force model has been found to predict the cutting forces with good accuracy.
REFERENCES Oxley, P. L. B., 1989, The mechanics of machining: an analytical approach to assessing machinability, E. Horwood, Chichester, England. van Luttervelt, C. A,, Childs, T. H. C., Jawahir, I. S., Klocke, F., Venuvinod, P. K., 1998, Present situation and future trends in modelling of machining operations. ClRP Annals, 47/2: 587-626. Altintas, Y., 2000, Modeling Approaches and Software for Predicting the Performance of Milling Operations, at MAL-UBC, ClRP 2000 machining workshop, University of New South Wales, Sydney. Ozel, T., 1998, Investigation of high speed flat end milling process: prediction of chip formation, cutting forces, tool stresses and temperatures, Ph.D Thesis, The Ohio State University. Ozel, T., Altan, T., 2000, Process simulation using finite element method - prediction of cutting forces, tool stresses and temperatures in high-speed flat end milling, Int J Mach Tools Manuf, 40/5: 713-738. Johnson, G. R., Cook, W. H., 1983, A constitutive model and data for metals subjected to large strains, high strain rates and high temperatures, Proc. 7th lntl Symp. on Ballistics, the Netherlands, pp.541-547. Lee, W. S., Lin, C. F., 1998, High-temperature deformation behaviour of Ti6A14V alloy evaluated by high strain-rate compression tests, J Mater Process Technol, 75(1-3): 127-136. Meyer Jr., H. W., Kleponis, D. S., 2001, Modeling the high strain rate behavior of titanium undergoing ballistic impact and penetration, Int J Impact Eng, 26( 1-10): 509-521. Majorell, A,, Srivatsa, S., Picu, R.C., 2002, Mechanical behavior of Ti-6AI-4V at high and moderate temperatures-Part I: Experimental results, Mater Sci Eng A, 326/2: 297-305 Lee, W. S., and Lin, C. F., 1998, Plastic deformation and fracture behaviour of Ti-6AI-4V alloy loaded with high strain rate under various temperatures, Mater Sci Eng A, 241(1-2):48-59. Kobayashi, S., Oh, S.I., Altan,T, 1989, Metal forming and the finite-element method. New York: Oxford University Press. Wang, Z. G., Wong, Y. S., Rahman, M., 2005, Highspeed milling of titanium alloys using binderless CBN tools, Int J Mach Tools Manuf, 45/1: 105-114.

Figure 5: Experimental and estimated cutting forces at a = 0.075mm, f = 0.075mm/r, and v = 350m/min.

Figure 6: Experimental and estimated cutting forces at a = 0.075mm, f = 0.075mm/r, and v = 400m/min. The estimated cutting force values of f , and f yare similar to those obtained experimentally. This indicates that the cutting forces can be predicted with good accuracy in these two directions. But for all three cases, the estimated cutting forces of f zare slightly smaller than experimental ones. The difference may have resulted from the plowing forces in milling, which is not considered in the cutting force model proposed in this paper. However, the