Вы находитесь на странице: 1из 8

Journal of Materials Processing Technology 173 (2006) 252259

Constitutive model of the alloy 2117-T4 at low strain rates and temperatures
H.V. Mart nez a, , D. Coupard b , F. Girot b
b

GINUMA, Ponticia Bolivariana University, A.A. 56006 Medell n, Colombia LAMEFIP-ENSAM, Esplanade des Arts et M etiers, 33405 Talence Cedex, France

Received 16 July 2004; received in revised form 16 July 2004; accepted 26 May 2005

Abstract This paper relates the cold forming of the alloy 2117-T4 and the description of a constitutive model for low temperatures and plastic strain rates. The characterization of the alloy was done by cylinder upsetting aided by a nite element code (FORGE2 ) and following an inverse approach. Taking into account the assumptions associated to the software, the parameters of the HanselSpittel and hyperboloid-type models were identied assuming a von Mises material. 2005 Elsevier B.V. All rights reserved.
Keywords: 2117-T4 aluminum alloy; Cylinder upsetting; Viscoplasticity; HanselSpittel constitutive model

1. Introduction Actual trends in metals forming processes are characterized by increasing demands for complex shapes and close tolerances. Numerical simulation of forming operations can be helpful in understanding the effect of process parameters on the product quality. It can also be used to validate a tool design. The quality of the numerical results is strongly related to the material properties and friction conditions at the die/workpiece interface. Numerous models have been developed to take into account both the material ow stress and friction conditions. Such models often comprise several parameters to estimate from simple mechanical tests (tensile, compression, torsion, ring test, . . .). This paper relates the constitutive model for low temperatures and plastic strain rates of the 2117-T4 aluminium alloy. This material is commonly used in aeronautical applications [1], i.e. for riveting operations. Taking into account the initial geometry of the rivet, i.e. a length of 7.14 mm and a diameter of 4.76 mm, a cylinder upsetting test has been chosen instead of a classical tensile

Corresponding author. Tel.: +574 415 9195; fax: +574 411 2372. E-mail address: hadervm@upb.edu.co (H.V. Mart nez).

test. Nevertheless, compression testing does not directly give access to the material ow characteristics. For example, the boundary conditions give rise to a multiaxial stress state preventing a simple and direct interpretation of the results. The stress state is often triaxial in the sample volume due to barreling and biaxial on the free surfaces [2]. Interpretation of the upsetting test by analytical methods such as those used for tensile testing can thus exceptionally be considered when friction at the die/workpiece interface is very restricted. However, during a lubricated cylinder upsetting of a sample with a height to diameter ratio greater or equal to 1.5, the force is mainly inuenced by the material rheology and not much by the friction coefcient, whereas the barreling of the deformed and relaxed cylinder is sensitive to friction but not much to the material rheology [3]. With this remark, rheology and tribology can thus be identied separately. The force is used to determine the rheological parameters for any kind of friction condition and the geometry enables the evaluation of the friction coefcient with the material parameters provided by the rst identication. In this study, a thin MoS2 solid lubricant layer was sprayed at the die/workpiece interface in order to (a) minimize the friction coefcient and (b) approach as much as possible a homogeneous upsetting test for which friction has a little

0924-0136/$ see front matter 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jmatprotec.2005.05.056

H.V. Mart nez et al. / Journal of Materials Processing Technology 173 (2006) 252259

253

inuence on the material ow. The rheology has been identied through an inverse method with the aid of the nite element forming software FORGE2 v3.0.

Cook [9], Zerilli and Armstrong [10], Norton and Hoff [11], Hamouda and Hashmi [12] and Hansel and Spittel [13] mod and T. els depends on , 2.3. Description of the nite element code FORGE2

2. Theory 2.1. Contact laws The die/workpiece contact is generally related to many parameters (rheology of involved materials, surface roughness of both the sample and the platens and the rheology of the lubricant in case of lubricated surfaces). A friction model often includes some fundamental parameters and the unknown variables are taken into account through the friction coefcient. Two models known as the Coulomb and Tresca friction model are commonly used [4]. In the Coulomb friction model, the tangential stress , is directly proportional to the normal stress n . The Coulomb friction coefcient , is an average isotropic coefcient over the total contact zone. It is generally valid in the case of contact with a moderate normal stress. In the Tresca friction model, the tangential stress is connected to the shear yield stress k by the Tresca friction coefcient m , which is also an average friction coefcient. This model is rather used for contact congurations with a high normal stress. If the material obeys the von Mises criterion, k = o / 3 while k = o /2 for a Tresca material, with o being the ow stress. A third model used by Duan and Sheppard in simulations with FORGE3 [5], seems interesting to take into account the normal stress evolution during a thermomechanical process. This Coulomb limited to Tresca friction model, is described in (1) for a material obeying the von Mises criterion: m o n , if n < 3 = (1) m o o / 3, if n m 3 2.2. Constitutive models for metal forming Modeling of a mechanical forming process always needs to dene constitutive models for the material involved. As for many metallic materials, the mechanical behavior of aluminium alloys at ambient temperature is often little or not sensitive to the strain rate, but as the temperature raises, the ow stress drops and becomes sensitive to the strain rate [6]. Many models are available for computing the changes in ow stress depending on the deformation conditions, temperature and strain rates. The pure ow stress models, = (T, ), are primarily used for cold forming, for example, the power law, the Hollomon equation or the models of Ludwik, Swift-Krupkowski and Voce [7]. The pure viscous models ), such as the Norton model [8], only take into = (T, account a strain rate and temperature dependences. The vis ), such as the Johnson and coplastic models, = (T, , FORGE2 , is a commercial package for nite element analysis of hot and cold metal forming. The key phenomena of material viscoplastic behavior, die/workpiece friction, heat transfer and thermomechanical coupling are incorporated into the nite element solver. FORGE2 also features automatic remeshing, which avoids excessive element distortion, when analyzing large-scale deformations. The Eqs. (2)(4) [14], indicate, respectively, the mechanical equilibrium, the plastic incompressibility and the energy balance used in FORGE2 .

where [ ] is the stress tensor, v the velocity eld, [ pl ] the plastic strain rate tensor , C the heat capacity, the ther =: the heat part dissipated through mal conductivity, W plastic deformation and (t) is the sample domain at the time t (Fig. 1). In case of unilateral for which a die/workpiece separation is allowed, the boundary conditions are:

where vdie is the die velocity, n the normal stress vector, c the die/workpiece interface and n is the specimen free surface (Fig. 1). The thermal boundary conditions are: (i) Flow condition on : T n = (8)

Fig. 1. Free and contact surfaces during cylindrical upsetting.

254

H.V. Mart nez et al. / Journal of Materials Processing Technology 173 (2006) 252259

with = h(Tp Text ) = bi b1 + b 2 in n (v vdie ) in c (9) (10)

where Tp is the sample surface temperature, Text the surrounding temperature, h a global heat transfer coefcient (convection and radiation), is indicated by (1), b1 and b2 are, respectively, the sample and die effusivities, bi = b1 or b2 depending whether the heat ow is dissipated inside the specimen or the die. The effusivities can be calculated as [15]: b= C (11)

(ii) Temperature condition on c : Tp = Tdie (12)

The tribology is represented by the Coulomb limited to Tresca model. Two additional friction conditions can also be retained: a sliding contact for which and m = 0 or a sticking contact for which sliding is not allowed at the interface. As a fundamental assumption, FORGE2 assumes that the material is homogeneous, isotropic and obeys the Von Mises criterion. Considering the principal stresses 1 , 2 and 3 , the equivalent stress is then: 1 1/2 eq = [(1 2 )2 + (2 3 )2 + (3 1 )2 ] 2 (13) The plastic ow stress o can be expressed by a HanselSpittel law (14) or a point-to-point law (15): o = A em1 T T m9 m2 em4 (1 + )m5 T em7 ( )m3 ( )m8 T eq , T ) eq = f (eq , (14) (15)
Fig. 2. Experimental upsetting test results for two temperature and die velocities conditions.

3. Experimentation and discussion The cylindrical upsetting tests were carried out for two levels of platen velocity (vb = 1 mm/min, vh = 60 mm/min) and temperature (Tb = 20 C, Th = 40 C) in an Adamel Lhomargy testing machine. The load cell is a 100 kN full scale with a class of precision of 0.5. The relative displacement of the platen is measured by an inductive position sensor with a full range of 4 mm and a class of precision of 0.5. The compression dies are made of a hard stainless steel alloy X19CrNi17.2 (AISI 431). In this study, the samples were machined in order to obtain a height to diameter ratio (Lo/) of 1.5. The diameter and height of the cylindrical samples are, respectively, 4.76 0.01 mm and 7.14 0.01 mm. In cylindrical upsetting, a ratio, ranging between 1.5 and 2 is recommended to characterize the compressive strength and avoid the risk

of buckling [16]. Massoni et al. [3] reports that the ratio Lo/ = 1.5 is more sensitive to the rheology than 0.75 for which friction has an important inuence in the material ow. On the other hand, there is no way to have a die/workpiece frictionless interface during upsetting. It is thus common either to limit it by interposing a solid or oil-type lubricant layer, or to maximize it such as to create a sticking contact. The minimization of the friction coefcient has been selected in this study. This makes possible to have a relatively homogeneous compression test (not much triaxial) in which friction has a rather weak inuence on the material ow. The standard ASTM E9 suggests the use of PTFE or MoS2 . In this study, a MoS2 solid lubricant Molykote [17,18] was sprayed in both faces of the die. 3.1. Identication and inuence of the tribological conditions The identication is based on the comparison of the experimental and numerical results in terms of the deformed sample geometry after upsetting. The inuence of the tribological conditions is studied through the effect on the load/displacement curve. The rst objective was to evaluate the magnitude of the friction coefcient at the die/workpiece interface, and then, to estimate, for the order of magnitude found, its inuence on the force/displacement curve. A typical experimental load/displacement curve is shown in Fig. 2. Whatever the temperature or die velocity condi-

H.V. Mart nez et al. / Journal of Materials Processing Technology 173 (2006) 252259

255

Fig. 3. Experimental () and simulated ( ), ( ting of a vb Tb sample.

), ( ) geometry after upset-

tions, the load/displacement curves are similar, taking into account the load cell precision that is showed by the error bars in Fig. 2. The material is thus insensitive to the strain rate and temperature under the studied range, which is between 20 C and 40 C and, for a strain rate between 103 s1 and 101 s1 . In order to simulate the upsetting test, the following parameters of a simplied and unoptimized Spittel law have been selected: A = 670 MPa, m2 = 0.24 and m7 = 0.12. The study is limited to a maximum friction coefcient of 0.1, the use of Molykote should prevent friction coefcient to be higher than this value [19]. The Tresca friction coefcient was m = 1, which means that one considers tmax = o / 3. Fig. 3 compares the dimensional characteristics of experimentally upset vb Tb samples with those obtained after simulation with FORGE2 . The dimensional characteristics measured are: (a) the upper diameter, i.e. the diameter of the sample in contact with the upper platen, (b) the lower diameter, i.e. the diameter of the sample in contact with the lower platen and (c) the maximum diameter which is usually located close to the mid-height of the sample when friction are similar in both platens. The relaxed sample height is given by the x-coordinate with x = 0 corresponding to the position of the sample in contact with the lower platen (diameter, dx = 0 ). In a same way, xmax is the sample position in contact with the upper platen (diameter, dxmax ). The experimental values of dx = 0 and dxmax are different because of non equal friction conditions on both platens. This small mismatch (around 0.15 mm) can be explained by a difference in the sample/platen roughness and lubricating conditions. Simulated values of dx = 0 and dxmax are also different because of a mismatch in the chosen values of in both platens ((lp) represents friction over the lower platen and (up) , friction over the upper platen). It is shown that couples A ((lp) = 0.01, (up) = 0.03) and especially C ((lp) = 0.02,

Fig. 4. Upset samples (a) without lubricant, (b) with lubricant Molykote and (c) simulated with (lp) = 0.02 and (up) = 0.03.

(up) = 0.03) correlate the experimental results rather well. The couple B ((lp) = 0.06, (up) = 0.1) presenting rather low friction coefcients, but higher than A and B, clearly shows rather bad correlation with respect to the experimental results. The magnitude order for the platen/sample Coulomb friction coefcient was systematically found around 0.020.03 in all experiments. Fig. 4 shows two deformed samples after upsetting with two friction conditions in which one of them was simulated. The sample shown in Fig. 4a has been deformed without lubricant, whereas the sample in Fig. 4b was lubricated with Molykote . In Fig. 4a, it is possible to notice a pronounced barreling effect in comparison with Fig. 4b. Fig. 4c is an upset specimen simulated with (lp) = 0.02 and (up) = 0.03. It can be noticed that there is a great similarity between Fig. 4b and c. Fig. 5 shows the experimental load/displacement curves of three vb Tb specimens and three simulations with different Coulomb friction conditions. Over the studied range ( < 0.1), the friction value has a little inuence on the simulated load/displacement curves except at high displacements

256

H.V. Mart nez et al. / Journal of Materials Processing Technology 173 (2006) 252259

Fig. 6. Geometrical congurations used for the platens and the specimen.

Fig. 5. Experimental and simulated ( ), ( for three vb Tb specimens.

), ( ) load/displacement curves

(>3 mm that is a true strain > 55%). For values between 0 and 0.04, corresponding to the experimental friction range, the load/displacement curve mismatch is less than the load cell precision. Therefore, taking in account a little variation of the friction coefcient from an experiment to another, it is possible to conrm that tribology does not inuence rheology. Tribology can thus be identied rst and the ow stress is then optimized considering the values of the initial identication. 3.2. Inuence of the parallelism defect of the specimen/sample system One of the main problems of the upsetting test is related to the setting of the specimen/platen parallelism. In order

to quantify precisely the inuence of this defect on the load/displacement curve, a 3D simulation should be done by considering either a parallelism specimen defect, either a parallelism platen defect, or both. An order of magnitude of this defect has been investigated in this study with the aid of FORGE2 . Fig. 6 shows the geometrical congurations used for both the platens and the sample. Congurations in Fig. 6a and b are, respectively, supposed to approach the effect of a mismatch parallelism between the platens and the specimen. The parameters defpar and defep quantify, respectively, the magnitude of the platen and specimen mismatch parallelism. The experimental load/displacement curves of six specimens deformed under the conditions vb Tb and vh Tb are given in Fig. 7. Fig. 7 also describes the simulated results for three platen/specimen congurations. The condition defpar = 0 corresponds to a perfect platen/specimen conguration, defpar = 0.05 is a mismatch of 0.05 mm (Fig. 6a) and the condition defep = 0.03 is a mismatch of 0.03 mm (Fig. 6b). The study is conducted with the same rheology and tribology as in Section 3.1. According to Fig. 7, it is clear that the platen/specimen mismatch only affects the experimental load/displacement curve at the beginning of the upsetting test until around 0.1 mm, which is below a mean true strain close to 1.5%. At high strains, the simulated results correlate the experimental results whatever the platen/specimen mismatch. The parallelism defect gives rise to a decrease in the slope of the load/displacement curve at the beginning of the upsetting test. This can be explained by considering that the load is not applied at the beginning of the upsetting test over the entire cross section of the sample. The magnitude order of the parallelism defect giving rise to a fairly good correlation between experimental and simulated results should be around 0.030.05 mm. The parallelism defect will thus induce rather bad precision with respect to the determination of the Young modulus and yield stress. However, it will not prevent a correct estimation of the plastic law.

H.V. Mart nez et al. / Journal of Materials Processing Technology 173 (2006) 252259

257

Fig. 7. Inuence of a platen/specimen mismatch parallelism on the load displacement curve.

Fig. 8. Experimental (IVI) and simulated (Spittel ) load/displacement curves. Table 1 Optimized parameters of the Spittel law

3.3. Rheological identication As said previously, the alloy is insensitive to the strain rate and temperature over the investigated range of both parameters. The ow stress model to determine must thus only be strain dependant. Two kinds of models have been selected, a simplied HanselSpittel law given by Eq. (16): 0 = A( + 0 )m2 exp m4 + 0 exp(m7 ( + 0 )) (16)

Parameters

HanselSpittel law Spittel inf Spittel ave 619.8 0.0715 0.047 0.0435 0.052 Spittel sup 668 0.069 0.048 0.0785 0.058

A (MPa) m2 m4 m7 0

577.1 0.083 0.045 0.015 0.045

And a hyperboloid-type model expressed as follow: 0 = A exp m2 1 + m3 m4 1 + m5 m6 (17)

until an asymptotic value given by the parameter A. The HanselSpittel model increases until a maximum value of 583 MPa for an equivalent strain around 2.1% and then decreases until a null asymptotic value. It is thus clear that the HanselSpittel model can not physically account for the material behavior for eq > 2.1%. Below eq = 2.1%, both

Taking into account the load cell precision, three different HanselSpittel and hyperboloid models were identied. The Spittel sup law and Spittel inf law shown in Fig. 8 are, respectively, those running through the upper and lower error bars. The Spittel ave law is dened as an average law between both previous laws. The identication of parameters of these laws was conducted by an inverse approach based on FORGE2 . Table 1 gives the values of the optimized HanselSpittel parameters and Table 2 the optimized values for the hyperboloid model. Fig. 9 shows the ow stress versus equivalent deformation for both models. The hyperboloid-type model increases

Table 2 Optimized parameters of the hyperboloid law Parameters Hyperboloid law Hyperboloid inf A (MPa) m2 m3 m4 m5 m6 625.5 1.367 0.121 1.093 9.926 0.83 Hyperboloid ave 655.7 1.218 0.239 0.47 10.316 0.822 Hyperboloid sup 651.1 1.013 0.318 1.584 11.958 0.927

258

H.V. Mart nez et al. / Journal of Materials Processing Technology 173 (2006) 252259

Fig. 9. Comparison of the HanselSpittel and hyperboloid-type models: (a) global curve and (b) zoom at the beginning.

models are equivalent in terms of correlation to the experimental results as shown by Fig. 10. The HanselSpittel model only needs the evaluation of ve parameters, which can be viewed as an advantage over the hyperboloidtype model which require the determination of six parameters.

3.4. Limit of the constitutive law These laws are only valid in rst approximation until an equivalent strain of 90% and extrapolation above this value is not recommended. In order to quantify more precisely the limit of validity of both constitutive laws, some samples were deformed at various maximum strain levels (40%, 50% and 90%) and then optically examined in different areas after polishing to determine the occurrence of cracks. Three areas were observed: (a) the free surface of the cylindrical sample close to the mid-height, (b) both faces in contact with the platens and (c) the bulk. The uniaxial compression tests were done at low strain rate, at room temperature and 40 C. As observed in Fig. 11a for a maximum strain of 90%, some small cracks less than 10 m appears at the free surface close to the mid-height of the sample where the greatest strains are encountered due to the barreling effect. These zones are not affected however when the maximum strain is around 40% or 50%. In addition, few short cracks less than 10 m were found in the periphery of the platen/sample contact zone in case of maximum strain around 50%. The bulk of tested samples does not show any crack even for maximum strain around 90% (Fig. 11b). According to these results, the constitutive laws found for the 2117-T4 alloy is valid until 50% of strain. Between 50% and 90%, the constitutive laws are not so valid due to few small cracks observed in two areas, but their size, being less than 10 m for maximum strain around 90%, should not greatly affect the identied constitutive law.

Fig. 10. Experimental (IVI) and simulated (Spitel-ave law and hyperboloidave law) load/displacement curves.

H.V. Mart nez et al. / Journal of Materials Processing Technology 173 (2006) 252259

259

Fig. 11. Microscopically view of: (a) free surface and (b) bulk of samples after uniaxial compression tests.

4. Conclusions The material is insensitive to the strain rate and temperature under the studied range, which is between 20 C and 40 C, and, for a mean strain rate between 103 s1 and 101 s1 , which are the range of parameter variations during slug forming or riveting. The order of magnitude for the platen/sample Coulomb friction coefcient was systematically found around 0.020.03 in all experiments. For values lower than 0.04 and a sample height to diameter ratio of 1.5, the platen/sample Coulomb friction coefcient has a little inuence (less than the load/cell precision) on the simulated load/displacement curves. Tribology can thus be identied rst and the ow stress is then optimized considering the values of the initial identication. The parallelism defect will induce rather bad precision with respect to the determination of the Young modulus and yield stress. However, it will not prevent a correct estimation of the plastic law. Two ow stress models were determined: (a) a simplied HanselSpittel law, and (b) a hyperboloid-type law. The identied constitutive laws are valid until 50%. Between 50% and 90%, the validity is not so obvious due to few small cracks observed in two areas, even if their size, being less than 10 m for maximum strain around 90%, should not greatly affect the identied constitutive law.

References
[1] P.L. Mu noz, Los composites y materiales avanzados en los aviones de nueva generaci on, in: Proceedings of the V Congreso Nacional de Materiales Compuestos, AEMAC, Zaragoza, 2003. [2] J.P. Wang, An investigation into friction in dynamic plane upsetting, J. Mater. Process. Technol. 123 (2002) 323328. [3] E. Massoni, B. Boyer, R. Forestier, Inverse analysis of thermomechanical upsetting tests using gradient method with semi-analytical derivatives, Int. J. Therm. Sci. 41 (2002) 557563. [4] E. Felder, Lubrication des surfaces lors de la mise en forme, Techniques de lIng enieur, Trait e Mat eriaux M etalliques 597 (1992) 230.

[5] X. Duan, T. Sheppard, Three dimensional thermal mechanical coupled simulation during hot rolling of aluminium alloy 3003, Int. J. Mech. Sci. 44 (10) (2002) 21552172. [6] F. Montheillet, M etallurgie en mise en forme, Techniques de lIng enieur, Trait e Mat eriaux M etalliques M600 (1996) 216. [7] Z. Gronostajski, The constitutive equations for FEM analysis, J. Mater. Process. Technol. 106 (2000) 4044. [8] M. Rappaz, M. Bellet, M. Deville, Trait e des mat eriaux, Mod elisation num erique en science et g enie des mat eriaux, vol. 10, Presses Polytechniques et Universitaires Romandes, 1998, pp. 291301. [9] G.J. Johnson, W.H. Cook, A constitutive model and data for metals subjected to large strains, high strain rates and high temperaturesdkjdot, in: Proceedings of the Seventh International Symposium on Ballistics, The Hague, 1983, pp. 541547. [10] F.J. Zerilli, R.W. Armstrong, Dislocation-mechanics-based constitutive relations for material dynamics calculations, J. Appl. Phys. 61 (5) (1987) 18161825. [11] P. Montmitonnet, J.L. Chenot, Introduction of anisotropy in viscoplastic 2D and 3D nite-element simulations of hot forging, J. Mater. Process. Technol. 53 (1995) 662683. [12] A.M.S. Hamouda, Effect of energy losses during an impact event on the dynamic ow stress, J. Mater. Process. Technol. 124 (2002) 209215. [13] X. Duan, T. Sheppard, Computation of substructural strengthening by the integration of metallurgical models into the nite element code, Comput. Mater. Sci. 27 (2003) 250258. [14] R. Forestier, Y. Chastel, E. Massoni, Development of an inverse module and of a semi-analytical sensitivity analysis for thermomechanical parameter identication, in: IPES 2003 Inverse Problems in Engineering Symposium, The University of Alabama, Tuscaloosa, AL, June 910, 2003. [15] Guy Snape, Sally Clift, Alan Bramley, Parametric sensitivity analyses for FEA of hot steel forging, J. Mater. Process. Technol. 125126 (2002) 353360. [16] ASTM, Designation E9-87, Standard Test Methods of Compression Testing of Metallic Materials at Room Temperature, pp. 152 158. [17] T. Xincai, Comparisons of friction models in bulk metal forming, Tribol. Int. 35 (2002) 385393. [18] S.P.F.C. Jaspers, J.H. Dautzenberg, Material behavior in conditions similar to metal cutting: ow stress in the Primary Shear Zone, J. Mater. Process. Technol. 122 (2002) 322330. [19] H. Han, The validity of mathematical models evaluated by two-specimen method under the unknown coefcient of friction and ow stress, J. Mater. Process. Technol. 122 (2002) 386 396.

Вам также может понравиться