Jtam.dvi JOURNAL OF THEORETICAL and APPLIED MECHANICS 45, 2, pp. 203-223, Warsaw 2007 ADIABATIC HEAT EVALUATION FOR DYNAMIC PLASTIC LOCALIZATION Patrice Longère Laboratoire de Génie Mécanique et Matériaux, Université de Bretagne Sud, Lorient Cedex, France e-mail: patrice.longere@univ-ubs.fr André Dragon Laboratoire de Mécanique et de Physique des Matériaux, CNRS-ENSMA, Futuroscope-Chasseneuil Cedex, France In most cases, adiabatic plastic heating is evaluated using the inelastic heat fraction coefficient, also knownasTaylor-Quinneycoefficient,which is usually assumed to be constant. From the thermodynamic viewpoint, this method consists in neglecting (or including some part of) various thermomechanical couplings in the heat equation. Nevertheless this co- arsemethod can lead to over-estimation of the temperature rise. In this paper, adiabatic shear banding incipience is considered in the context of thermal instability. It is shown that the accuracy in the prediction of favourable conditions for the onset of this plastic localization is strongly dependent on the technique retained for evaluating the plasticity indu- ced heating. This paper aims at showing the influence of various levels of simplification of the thermal equation on the critical conditions at localization onset, the latter being obtained from a criterion based on the linear perturbationmethod. Key words: adiabatic shear banding, inelastic heat fraction, non-linear mechanics, viscoplasticity 1. Introduction Dynamic plastic localization in the form ofAdiabatic ShearBanding (ASB) is known to occur at high strain rates under quasi adiabatic conditions as ther- mal softening prevails against strain and strain rate hardening. Depending on the application considered, this precursor of structural failure is sought – it is typically the case for metal cutting (see Rhim andOh, 2006), or apprehended – it is the case in penetrator-armour steel interaction (seeOdeshi et al., 2006). 204 P. Longère, A. Dragon The knowledge of favourable conditions for ASB localization onset and fur- ther post-critical behaviour is thus ofmajor interest, particularly in numerical simulations of the processes mentioned above. In parallel, most of dynamic engineering finite element calculation codes take into account thermomechanical coupling only through the inelastic heat fraction which characterizes the fraction of plastic work converted into heat which is usually assumed to be constant. However, the inelastic heat fraction can be strongly dependent on strain, strain rate and temperature as observed experimentally by Mason et al. (1994), Kapoor and Nemat-Nasser (1998) for example, and shown theoretically byAravas et al. (1990), Longère andDragon (2006). For abetter evaluation of plastic deformation-induced temperature rise and related thermal softening, a specific ’user material’ routine implemented in the FE engineering code is needed including other aspects appearing in the formulation of the heat equation (see Longère et al. (2005) for computations involving LS-DYNA R© code). According to ASB-devoted experimental studies, the band orientation is mainly controlled by themaximum shear stress (see Bai andDodd, 1992; Pę- cherski, 1998), and a weak localization of deformation, as a consequence of thermal instability, is primarily observed followed by the genuine strong loca- lization of deformation leading to the ASB process (seeMarchand andDuffy, 1988; Liao and Duffy, 1998). Starting from these results, this paper aims at studying the consequences of simplifications assumed to evaluate dissipative mechanisms-induced heating on the determination of dynamic plastic loca- lization occurrence. The latter is obtained herein from the theory of linear perturbation (see Bai, 1982; Clifton et al., 1984) in the form of a criterion relating the resolved shear stress to strain hardening, thermal softening and viscosity (see Batra and Chen, 2001; Longère et al., 2003) in the context of ASB induced localization. The purpose of the present study is not to discuss or extend the theory but to apply the method in order to obtain a practical criterion reproducing experimental observation – the reader can refer to Fres- sengeas andMolinari (1987) andAnand et al. (1987) for extensive analysis. For this purpose, various levels of simplification of the heat equation formulation are considered, yielding various criteria for the ASB onset. In Section 2, thermodynamic fundamentals are recalled leading to the expression of the heat equation in the case of thermo/elastic-viscoplastic mo- delling. The context considered in this work concerns loading at strain rates high enough to allow for the assumption of adiabatic conditions whilemaking it possible to assume the (visco)plastic material behaviour to be pressure in- sensitive. The strain rates range is thus typically comprised between 102 and 104 s−1, and maximum value of temperature remains lower than the melting point. Three levels of simplification are studied: first, due to its weak con- tribution to temperature changes, the thermo/elastic coupling is neglected; Adiabatic heat evaluation for dynamic plastic localization 205 secondly thermodissipative couplings are neglected as well; finally, the tempe- rature rise is supposed to be directly linked to the plastic work rate via the inelastic heat fraction assumed to have a constant value. In Section 3, the consequences of simplifications mentioned above are exa- mined in the context of thermal instability induced dynamic plastic localiza- tion. With this aim in view, the linear perturbation method is applied to the case of simple shearing of a volume element consisting of a thermo/viscoplastic material. Note that in the case of the ASB phenomenon, the extension of the 1D loading to a complex 3D loading can be performed today (see Longère et al. (2003)). Simplifications are assumed in order to reproduce qualitative- ly different stages of deformation localization (weak and strong) as observed notably by Marchand and Duffy (1988). Further parametric analysis shows how the value of the critical shear strain at localization incipience is strongly dependent on the temperature rise evaluation method. 2. Evaluation of temperature rise under adiabatic conditions Irreversible thermodynamics framework is used here to describe the thermo/elastic-inelastic response of a material (see Perzyna, 1966; Bataille and Kestin, 1975; Raniecki and Nguyen, 1984; Petryk, 2000; Mróz and Olife- ruk, 2002). At first, the internal variable procedure is summarized leading to expressions of dissipated energy and thermal equation (see Perzyna (1998) for computational issues for the evolution problem). Different levels of simplifica- tion of the latter are then introduced in connection with the further study of heat evaluation at the dynamic localization onset. 2.1. Thermodynamic context and constitutive framework The state of thematerial is supposed to be described by a thermodynamic potential, namely Helmholtz free energy per unit mass ψ(T,z), where T re- presents absolute temperature and z a set of normal variables. Gibbs relation takes the following form ρ0ψ̇ =−ρ0sṪ +Zż s =− ∂ψ ∂T ∣ ∣ ∣ z Z = ρ0 ∂ψ ∂z ∣ ∣ ∣ T (2.1) where ρ0 represents the mass density in the initial configuration, s – the entropy and Z – the set of thermodynamic (conjugate) forces associated with the state variables z. According to the second law of thermodynamics, the mechanical part of dissipated energy is written as D = τ : d−Zż ­ 0 (2.2) 206 P. Longère, A. Dragon where d represents the rate of deformation tensor and τ is the Kirchhoff stress tensor defined as theCauchy stress tensor σmultiplied by the Jacobian determinant J of thedeformationgradient F (J =detF= ρ0/ρ), i.e. τ = Jσ. A combination of the first law of thermodynamics and the Gibbs relation gives a local form of the heat equation ρ0czṪ + divq+r = τ :d− ( Z −T ∂Z ∂T ) ż (2.3) where cz represents the specific heat, q – the heat flux vector per unit area and r – the heat supply per unit volume. Equation (2.3) relies on thermal terms on the left hand-side including the temperature rate Ṫ , heat conduction divq, andheat supply r, andmechanical terms on the right hand-side including the mechanical work rate τ : d and other work rates Zż, while the last term Tż∂Z/∂T is referred explicitly to thermomechanical couplings. The study is now reduced to thermo-elastic/viscoplastic behaviour in the case of isotropic strain hardening. The set of state variables is assumed in the form z ≡ (εe,p), where εe represents a measure of finite elastic strain (εe = lnVe; Fe =VeRe) and p is the isotropic strain hardening variable. The setof conjugate forces is thusgiven in the form Z ≡ (τ,R),where R represents the isotropic hardening force (affinity). The rate of deformation (symmetric part of the velocity gradient) d = [∂v/∂x]S is furthermore decomposed into a reversible (’elastic’) part d e = ∇ ε e (where ∇ designates the objective Jaumann derivative of a 2nd order tensor) and an irreversible part dp such that d = de +dp (see Sidoroff and Dogui, 2001). Mechanical dissipation (2.2) becomes D = τ : dp−Rṗ ­ 0 (2.4) The constitutive laws are thus expressed as s =− ∂ψ ∂T ∣ ∣ ∣ εe,p τ = ρ0 ∂ψ ∂εe ∣ ∣ ∣ T,p R = ρ0 ∂ψ ∂p ∣ ∣ ∣ T,εe (2.5) The free energy density ψ(T,εe;p) is expressed in the following form ψ(T,εe;p)= ψe(T,εe)+ψth(T)+ψb(T ;p) (2.6) where ψe(T,εe) represents the recoverable energy which includes isotropic li- near thermo-elasticity, ψth(T) – purely thermal energy, and ψb(T ;p) – stored Adiabatic heat evaluation for dynamic plastic localization 207 energy reflecting competition in the material between isotropic strain harde- ning and thermal softening. These contributions are written as follows ρ0ψ e(T,εe)= λ 2 (Trεe)2+µεe : εe−αKTrεeϑ ρ0ψ th(T)=−ρ0c0 [ T ln ( T T0 ) −ϑ ] −h(0)f(T) (2.7) ρ0ψ b(T,p)= h(p)f(T)−h(0)f(0) where λ and µ represent the Lamé elasticity constants, K = λ +2µ/3 – bulkmodulus, α – thermal dilatation coefficient, c0 – constant heat capacity, ϑ = T −T0 – temperature rise, h(p) – stored energy of cold work and f(T) – thermal softening function.For simplification, the thermo-elasticity coefficients are supposed to be temperature independent. From now on, the elastic deformations will be considered as small (infini- tesimal). After partial derivation of (2.6) with respect to z ≡ (T,εe,p), the thermo- dynamic forces Z ≡ (−s,τ,R) are written as −ρ0s =−αKTrεe−ρ0c0 ln ( T T0 ) − [h(0)−h(p)]f ′(T) τ =(λTrεe−αKϑ)δ+2µεe (2.8) R = h′(p)f(T) By rewriting mechanical dissipation (2.4) in the form D = Zż ­ 0, Z = (τ,R) represents a set of thermodynamic forces associated to the set of fluxes ż = (dp,−ṗ). Considering now the time-dependent plasticity and applying the normality rule with respect to the dual Φ(Z) of the dissipation potential, dependingon Z via the loading (yield) function F(Z) (i.e. standard rule for ż), the internal state variable evolution laws are deduced from ż = Λ ∂F(Z) ∂Z Λ = ∂Φ(F) ∂F ­ 0 (2.9) where Λ represents the viscous multiplier governing the dissipative mechani- sms of plasticity. Applying the normality rule to the present case yields d p = Λ ∂F ∂τ − ṗ = Λ∂F ∂R (2.10) The yield function F(Z) in (2.9) is assumed in the form of Huber-von Mises criterion F(τ,R;T)= J2(τ)−g(R,T) (2.11) 208 P. Longère, A. Dragon where J2(τ)= √ 3 2 s : s and s represents the stress deviator tensor s= τ − Trτ 3 δ The strain hardening function g(R,T) in (2.11) representing the Huber-von Mises surface radius is expressed by g(R,T) = R0f(T)+R(T,p) (2.12) The evolution laws in (2.10) are thus detailed as follows d p = 3 2 Λ s J2 ṗ = Λ (2.13) Themodel is completed by the expression of the force potential in (2.9) Φ(F)= Y m+1 〈F Y 〉m+1 Λ = 〈F Y 〉m (2.14) Furthermore, Eq. (2.3) becomes ρ0c0Ṫ + divq+r = τ : d p+T ∂τ ∂T : de− ( R−T ∂R ∂T ) ṗ (2.15) The context considered here concerns the loading at high strain rate excluding heat supply, and for which conditions can be assumed as adiabatic (dynamic loading). Relation (2.15) is thus reduced to ρ0c0Ṫ = τ :d p+T ∂τ ∂T :de− ( R−T ∂R ∂T ) ṗ (2.16) In (2.16), τ : dp represents theplastic partof themechanicalwork rate (plastic work rate), Rṗ – the stored energy rate, the difference τ : dp−Rṗ represents the unrecoverable energy rate dissipated by heating, see (2.4), (T∂τ/∂T) : d e – thermo-elastic coupling contribution which describes cooling during the tensile loading and heating during compression, and Tṗ∂R/∂T – the thermo- dissipative coupling contribution which expresses the stored energy release rate during the temperature rise (see Clayton (2005) for a similar approach). Adiabatic heat evaluation for dynamic plastic localization 209 2.2. Levels of simplification of heat equation As mentioned above, heat equation (2.16) can be decomposed into various contributions. By denoting Q̇ = ρ0c0Ṫ D = τ :d P −Rṗ ­ 0 Ẇτ = T ∂τ ∂T : de ẆR = T ∂R ∂T ṗ Eq. (2.16) can be rewritten as Q̇ = Ẇ with Ẇ = D+Ẇτ +ẆR (2.17) Three cases are now distinguished depending on the terms retained on the right side of (2.17)1,2. In the first case, called ’quasi complete heat evaluation’, thermo-elastic coupling effects are neglected due to their weak contribution to the temperature change: Ẇτ =0. Relation (2.17)2 is thus reduced to Ẇ = D+ẆR = [ J2− ( R−T ∂R ∂T )] ṗ (2.18) In the second case, heat equation (2.16) is further simplifiedbyneglecting both thermo-elastic and thermo-dissipative couplings contributions (see Voyiadjis and Abed (2006) for similar simplification). The resulting equation is called ’simplified heat evaluation’. This means that Ẇτ = ẆR = 0, and (2.17)2 becomes Ẇ = D =(J2−R)ṗ (2.19) The last case is equivalent to the representation involving the constant inelastic heat fraction β – also known as Taylor-Quinney coefficient (seeMason et al., 1994) – as frequently done in engineering problems. The resulting equation is called ’basic heat evaluation’. The expression for the work rate Ẇ in (2.17)2 is then reduced to Ẇ = βJ2ṗ (2.20) 3. Condition for dynamic localization occurrence The condition for localization (in form of adiabatic shear banding) onset is obtained here using the linear perturbation method. In order to analyse the consequences of simplifications assumed to evaluate the temperature rise, the loading path considered is simple shearing. 210 P. Longère, A. Dragon 3.1. Linear perturbation method The linear perturbation method is, in general, applied to the case of simple shearunder constant velocity boundaryconditions.Assumingnegligible elastic effects, laminar viscoplastic flow and adiabatic conditions, the problem can be reduced to a one-dimensional formulation, seeBai (1982), Clifton et al. (1984), Molinari (1985), Batra and Wei (2006), and also Anand et al. (1987) when three-dimensional generalization is presented. Admitting analytical solutions, the linear perturbation method provides in this case a criterion of instability onset, which is interpreted as the incipience of the adiabatic shear banding process, providing in fact the necessary condition for the onset of formation of bands (possibility of the shear band type instability). Starting from the governing equations (momentum balance, energy balan- ce, kinematics and constitutive law) in the case of simple shear in (1,2) plane under adiabatic conditions and neglecting elasticity, one obtains the following system s12,2 = ρ0v̇1 ρ0c0Ṫ = Ẇ ṗ = Λ(s12,p,T) d12 = 1 2 (v1,2+v2,1)≈ dp12 = 3 2 ṗ s12 J2 (3.1) With J2 = √ 3s12, v1 = Γ̇x2, v2 =0 and Γ = Γ̇t, system (3.1) is reduced to s12,2−ρ0v̇1 =0 ρ0c0Ṫ −Ẇ =0 v1,2− √ 3ṗ =0 ṗ−Λ(s12,p,T)= 0 (3.2) A small perturbation δU =(δv1,δs12,δp,δT) is now superimposed on the set of homogeneous solutions U =(v1,s12,p,T) U ⇒ U + δU with δU ≪ U The perturbed system is linearized from (3.2) as follows δs12,2−ρ0δv̇1 =0 ρ0c0δṪ − δẆ =0 δv1,2− √ 3δṗ =0 δṗ− δΛ(s12,p,T)= 0 (3.3) With the notations inAppendix – see in particular (A.2) and (A.3), the above system can be equivalently written as follows δs12,2−ρ0δv̇1 =0 ρ0c0δṪ − (wsδs12+wpδp+wTδT +wṗδṗ)= 0 (3.4) δv1,2− √ 3δṗ =0 δṗ−2P12δs12−Bδp−EδT =0 Adiabatic heat evaluation for dynamic plastic localization 211 where the expressions for wi are given in Table 1 depending on the level of simplification of the energy balance. Table 1.Expressions for wi Basic Simplified Quasi-complete ws √ 3βṗ √ 3ṗ √ 3ṗ wp 0 −Qṗ −(Q−TY )ṗ wT 0 −Sṗ TZṗ wṗ √ 3βs12 √ 3s12−R √ 3s12− (R−TS) Let the perturbation have a wave-like form δU = U exp(̟t+ikx2)= U exp(̟Rt)exp[ik(ct+x2)] (3.5) where U represents theperturbationmagnitude, ̟ –wavepulsation, k –wave number, x2 – wave plane normal, ̟R and ̟I – real and imaginary parts of the wave pulsation ̟, respectively, and c = ̟I/k – wave velocity. According to the right-hand side of (3.5), the case ̟R = 0 points the transition between the stable and unstable states: • if ̟R > 0, the perturbation may grow with time and the instability mentioned is possible; • if ̟R < 0, the perturbation decreases with time. The objective consists thus in looking for the conditions of transition from the stable state to the possible unstable state by studying the sign of ̟R. Injecting (3.5) into system (3.4), yields iks12−ρ0̟v1 =0 (ρ0c0̟−wT)T −wss12− (wp+wṗ̟)p =0 (3.6) ikv1− √ 3̟p =0 (̟−B)p−2P12s12−ET =0 or otherwise      −ρ0̟ ik 0 0 0 −ws −(wp+wṗ̟) (ρ0c0̟−wT) ik 0 − √ 3̟ 0 0 −2P12 (̟−B) −E               v1 s12 p T          =          0 0 0 0          (3.7) (3.7) can also be put as [A]{U}= {0} (3.8) Thedeterminantof thematrix [A] in (3.8),whose components aredenoted aij, is simply det[A] = a11a22a33a44−a11a42a33a24+a12a31(a23a44−a43a24) (3.9) 212 P. Longère, A. Dragon The spectral equation deduced from (3.8) and (3.9) is a 3rd degree polynomial in ̟ det[A] = a3̟ 3+a2̟ 2+a1̟ 1+a0 (3.10) with a3 = √ 3ρ20c02P12 a2 = ρ0[k 2c0+ √ 3(Ews−2P12wT)] (3.11) a1 =−k2(Ewṗ+wT +ρ0c0B) a0 = k2(BwT −Ewp) 3.2. Localization vs. instability As observed byMarchand andDuffy (1988) anddemonstrated byMolinari (1985), the instability does not imply rigorously localization. Thismeans that the use of the linear perturbation method provides the necessary condition only, which means that it represents a ’lower’ bound for the effective locali- zation incipience. The idea is here to ’delay’ the instability onset, i.e. push it towards the ’upper’ bound in the sense of approaching the strong localiza- tion incipience. In other words, we distinguish indeed the ’instability point’ characterizing locally the equilibrium between strain hardening and thermal softening (maximum of the shear stress-shear strain curve in most cases) and the ’localization point’ beyond which the shear stress drops strongly. The li- near perturbation method provides the ’instability point’, and we look, via a pragmatic engineering evaluation approach, for the ’localization point’ which eventually succeeds the ’instability point’. The adiabatic shear banding occurs as thermal softening overcomes the strain hardening. Before the ’instability point’, strain hardening is predomi- nant and the material is necessarily stable, while past the ’instability point’, the thermal softening becomes predominant and the material may become unstable. In the linear perturbationmethod, attenuating thermal softening al- lows consequently for pushing forward the ’instability point’ and approaching the ’localization point’. Let now examine the sources of thermal softening. In the yield function presented previously, see (2.12), the thermal softening source is double – in the function R0f(T) and in the isotropic hardening force R(T,p), see (2.8)3. Ne- glecting the contribution of R0f(T) – by prescribing that R0f(T)= 0 in the linear perturbationmethod solelywhilemaintaining this thermal contribution in the constitutive law – constitutes amanner to attenuate the thermal softe- ning and pushes forward the instability onset (without altering themaximum of the shear stress-shear strain curve). This simplification is studied here by considering the basic evaluation of temperature rise. In the first case, the problem is solved with both sources of thermal softening (assuming that localization and instability are concomitant), Adiabatic heat evaluation for dynamic plastic localization 213 while in the second case, the first source of thermal softening is neglected (the localization onset is delayed regarding instability onset). Employing Table 1, when considering the basic evaluation, the coefficients a0 to a3 in (3.11) become a3 = √ 3ρ20c02P12 a2 = ρ0(k 2c0+3Eβṗ) a1 =−k2(E √ 3βs12+ρ0c0B) a0 =0 (3.12) Spectral equation is, in this case, indeed reduced to a 2nddegree polynomial in ̟ and ̟ > 0 if and only if a1a3 < 0. The condition for possible perturbation growth (instability occurrence) is accordingly given by √ 3s12 > ρ0c0 β B −E (3.13) The instability incipience in the case of ”two sources of thermal softening” is given by injecting (A.4)2 and (A.4)3 (see Appendix) into (3.13) √ 3s12 > ρ0c0 β Q −(R2f ′+S) = ρcy β h′′(p)f(T) −[R2+h′(p)]f ′(T) (3.14) In order to delay the instability onset (i.e. to approach the localization one), we are now considering a ”single source of thermal softening” (f(T) = 0 in (A.4)3). In this case, the localization is supposed to occur as soon as √ 3s12 > ρ0c0 β Q −S = ρcy β h′′(p)f(T) −h′(p)f ′(T) (3.15) Relations (3.14) and (3.15) are available for anymaterial behaviour. In the fol- lowing application, we are interested in a class ofmetallic materials exhibiting strain hardening with saturation stress and linear thermal softening. Stored energy of cold work h(p) and thermal softening function f(T) in (2.7)3 are given by h(p)= R∞ [ p+ 1 k exp(−kp) ] f(T)= 1−AT (3.16) Model constants have been reported in Table 2 below. Table 2.Material constants for the numerical modelling E [GPa] ν ρ [kg/m3] c [J/(kgK)] α [K−1] 200 0.33 7800 420 10−6 R∞ [MPa] k A [K −1] R0 [MPa] Y [MPas 1/m] m 1000 20 10−3 1000 100 6 214 P. Longère, A. Dragon The response of the above material to shear loading is given for various initial temperatures and various strain rates in Fig.1 and Fig.2, respectively. The critical plastic strain rate for adiabatic conditions is supposed to be equal to about 100s−1. Fig. 1. Shear stress vs. shear strain. Influence of initial temperature. Instability criterion deactivated. Γ̇ =10−1 s−1 Fig. 2. Shear stress vs. shear strain. Influence of strain rate. Instability criterion deactivated. T0 =20 ◦C Detection of theplastic localization onset is essential to activate theprocess of softening behaviour (see Longère et al., 2003, 2005) for three-dimensional modellingof a thermoelastic/viscoplasticmaterial incorporatingadiabatic she- ar banding formation and growth). The description of the post–localization behaviour is not the purpose of the present paper. But in order to visualize the drop in stress induced by adiabatic shear banding, the ASB deterioration model developed by these authors has been used. Details regarding numeri- cal calculations and a corresponding algorithmic procedure can be found in Longère et al. (2005). It is notable that instability criteria (3.14) and (3.15) can be applied to any constitutive equations based on irreversible thermody- namics (see nevertheless the recommendations inLongère andDragon (2006)). InFig.1 andFig.2, the criteria have beendeactivated (nodeterioration), while inFig.3 andFig.4 the corresponding criteria are accounted for and considered Adiabatic heat evaluation for dynamic plastic localization 215 as a necessary and sufficient condition for the localization onset (plastic locali- zation inducedASBdeterioration is thusproducing strong softeningbehaviour according to the model by Longère et al. (2005)). The simplification leading to the delay of the instability point is illustrated in Fig.3 and Fig.4 for given shear strain rates and two values for β. Fig. 3. Shear stress vs. shear strain. Instability criterion activated. T0 =300K; Γ̇ =103 s−1; β =1 Fig. 4. Shear stress vs. shear strain. Instability criterion activated. T0 =300K; Γ̇ =103 s−1; β =0.8 For β = 1 (see Fig.3), the value of localization deformation (drop in stress) is close to 25% for the double source of thermal softening, while it is close to 30% for the single source of thermal softening. For β = 0.8 (see Fig.4), the value of localization deformation is close to 27% for the double source of thermal softening, while it is close to 32% for the single source of thermal softening. With the double source of thermal softening, Figures 3 and 4 clearly show that the instability appears at themaximumof the shear stress-shear strain cu- rve. This is in oppositionwith the experimental results obtained byMarchand and Duffy (1988) who distinguished two stages in the deformation localiza- tion (a weak one preceding the genuine strong one in form of adiabatic shear bands). In this sense, the single source of thermal softening seems to bemore 216 P. Longère, A. Dragon realistic, i.e. reproduces the delay between the thermal instability and genuine localization onset. The reader can also refer to works by Estrin et al. (1997) on full non-linear analysis of adiabatic shear band formation in the case of a shear sample containing a geometric defect. 3.3. Condition for delayed dynamic localization onset As mentioned in the previous section, delaying the strong localization onset with respect to the supposed instability onset is favoured by setting f(T)= 0 in the linear perturbationmethod (concept of ’upper’ bound approximation). This approach is consequently used in the following. According to Table 1, the coefficients a0 to a3 in (3.11) become in the ”simplified evaluation” a2 = ρ0 [ k2c0+3ṗ ( E + 2√ 3 P12S )] a3 = √ 3ρ20c02P12 a1 =−k2[E( √ 3s12−R)−Sṗ+ρ0c0B] a0 = k2ṗ(EQ−BS) (3.17) with E +2P12S/ √ 3=0 and EQ−BS =0. The corresponding constants in (3.17) are thus expressed by a3 = √ 3ρ20c02P12 a2 = ρ0k 2c0 a1 =−k2[E( √ 3s12−R)−Sṗ+ρ0c0B] a0 =0 (3.18) The spectral equation is oncemore reduced to a 2nd degree polynomial in ̟. and the condition for perturbation growth is given by √ 3s12 > R− ṗ α +ρ0c0 ∂R ∂p −∂R ∂T (3.19) Following the previous procedure for quasi-complete evaluation, gives a3 = √ 3ρ20c02P12 a2 = ρ0 [ k2c0+3ṗ ( E − 2√ 3 P12TZ )] a1 =−k2{E[ √ 3s12− (R−TS)]+TZṗ+ρ0c0B} a0 = k 2ṗ[EQ−T(EY −BZ)] (3.20) The spectral equation remains a 3rd degree polynomial in ̟ for which the condition of instability onset is not trivial. In order to obtain a practical cri- terion in the sense of Bai (1982), the condition for instability onset is again deduced from the sign of the product a1a3. Adiabatic heat evaluation for dynamic plastic localization 217 The condition for perturbation growth is, in this case, given by √ 3s12 > R−T ( ∂R ∂T − ∂2R ∂T2 ∂R ∂T ṗ α ) +ρ0c0 ∂R ∂p −∂R ∂T (3.21) Figures 5 and 6 give thematerial response to shearing and temperature evolu- tion for three cases of heat evaluation (twovalues for β are given), respectively. According to Fig.5, the value of shear strain at localization (critical she- ar strain) onset is close to 30% for basic evaluation with β = 1.0, to 32% for basic evaluation with β = 0.8, to 35% for simplified evaluation, and to 38% for complete evaluation. In parallel, the influence of heat evaluation is shown in Fig.6, which gives the evolution of temperature. According to the ”basic evaluation”, temperature increases until about 380K, while according to the ”quasi-complete evaluation”, themaximumof temperature remains un- der 350K. This difference of 30K is enough to provoke a difference of 8% for the estimation of the critical shear strain in the conditions prescribed here. Fig. 5. Shear stress vs. shear strain. Instability criterion activated. T0 =300K; Γ̇ =103 s−1 Fig. 6. Temperature vs. shear strain. Instability criterion activated. T0 =300K; Γ̇ =103 s−1 In this example, the shear stress-shear strain curves diverge from each other from the shear strain close to 10%, see Fig.5. As a consequence, the 218 P. Longère, A. Dragon value of shear stress at themaximum of each curve is different (the highest is obtained for the quasi-complete evaluation,while the lowest is obtained for the basic evaluation for β =1). So,we consider next here shear stress-shear strain curveswith the same value of the shear stressmaximumvalue (see Table 3 for new material constants). Numerical results including the instability criterion are given in Fig.7 and Fig.8. Table 3.Material constants Evaluation type β R0 [MPa] A [K −1] Basic 1.0 1000 1 ·10−3 0.8 1015 1.05 ·10−3 Simplified X 1050 1.1 ·10−3 Quasi-complete X 1080 1.15 ·10−3 The values of critical shear strain are lower than the previous ones (see Fig.5), but the tendency is preserved: the critical shear strain is greater for the quasi complete evaluation. Fig. 7. Shear stress vs. shear strain. Instability criterion activated. T0 =300K; Γ̇ =103 s−1 Fig. 8. Temperature vs. shear strain. Instability criterion activated. T0 =300K; Γ̇ =103 s−1 In Figs. 7 and 8, the maximum shear stress is the same for all simulations. Adiabatic heat evaluation for dynamic plastic localization 219 4. Concluding remarks In many boundary value problems involving high strain rates under the hy- pothesis of adiabaticity and employing numerical simulations via engineering finite element calculation codes, thermomechanical couplings are usually ac- counted for by using the approximation of the linear dependency of the tem- perature growth on the plastic work rate. In practice, a constant inelastic heat fraction coefficient, also known as Taylor-Quinney coefficient, is used. Thismethod consists in neglecting genuine thermomechanical couplings expli- citly occuring in the heat equation. In most cases, this simplification appears to be sufficient and conservative from the engineering viewpoint and regar- ding the calculation time. However, if the thermal instability induced plastic localization is concerned – it is the case of adiabatic shear banding as typi- cally encountered in speed machining and impact ballistics for example – the evaluation of temperature growth and its consequences in terms of the related material behaviour softeningmustbemore accurate.Wehave applied here the linear perturbation method to obtain a criterion for adiabatic shear banding initiation taking into account the strain hardening, thermal softening and vi- scous effects. This criterion is suitable for a wide range of thermo/viscoplastic materials. It has been used to study the influence of terms retained in the plasticity-induced heating and related adiabatic shear banding incipience. The analysis shows clearly that the use of the coarsemethod based on the constant inelastic heat fraction (varying in the usual range between 80% and 100%) le- ads to over-estimation of the temperature rate and produces an early onset of adiabatic shear banding instability as compared to a more rigorous ther- modynamic evaluation. If the use of a constant inelastic heat fraction appears conservative for impactproblems(fromthearmourviewpoint, earlier detection via numerical simulation can be interpreted as a safety-oriented approach), it is not when applied to numerical simulations relevant to metal cutting (in this case, later detection is rather researched for a more realistic definition of cutting conditions guaranteeing serrated chip formation under adiabatic shear banding). A. Appendix The perturbed system is linearized from (3.2) as follows δs12,2−ρδv̇1 =0 ρcyδṪ − δẆ =0 δv1,2− √ 3δṗ =0 δṗ− δΛ(s12,p,T)= 0 (A.1) 220 P. Longère, A. Dragon Perturbed viscous multiplier in (A.1)4 is expressed by δΛ =2 ∂Λ ∂s12 δs12+ ∂Λ ∂p δp+ ∂Λ ∂T δT =2P12δs12+Bδp+EδT (A.2) where Pij = ∂Λ ∂sij = α ∂F ∂sij B = ∂Λ ∂p = α ∂F ∂p E = ∂Λ ∂T = α ∂F ∂T α = ∂Λ ∂F (A.3) Furthermore, the coefficients in (A.3) become Pij = 3 2 α sij J2 B =−αQ E =−α[R0f ′(T)+S] (A.4) where Q = ∂R ∂p = h′′(p)f(T) S = ∂R ∂T = h′(p)f ′(T) (A.5) The expressions of the perturbed work rate depend on the heat evaluation type: — basic evaluation In this simple case, the perturbedwork rate is given by δẆ = β(δJ2ṗ+J2δṗ)= √ 3β(δs12ṗ+s12δṗ) (A.6) — simplified evaluation The perturbed work rate is here given by δẆ =(δJ2− δR)ṗ+(J2−R)δṗ =( √ 3δs12− δR)ṗ+( √ 3s12−R)δṗ (A.7) According to (A.4), one obtains δR = ∂R ∂p δp+ ∂R ∂T δT = Qδp+SδT (A.8) Thus, perturbedwork rate (A.7) becomes δẆ =( √ 3δs12−Qδp−SδT)ṗ+( √ 3s12−R)δṗ (A.9) — quasi complete evaluation The perturbed work rate is here given by the following expression δẆ = [ δJ2− ( δR− δT ∂R ∂T −Tδ (∂R ∂T ))] ṗ+ [ J2− ( R−T ∂R ∂T )] δṗ = (A.10) = [√ 3δs12− ( δR− δT ∂R ∂T −Tδ (∂R ∂T ))] ṗ+ [√ 3s12− ( R−T ∂R ∂T )] δṗ Adiabatic heat evaluation for dynamic plastic localization 221 Differentiating (A.5)2, yields δS = ∂S ∂p δp+ ∂S ∂T δT = Y δp+ZδT (A.11) where Y = ∂S ∂p = ∂2R ∂p∂T = h′′(p)f ′(T) Z = ∂S ∂T = ∂2R ∂T2 = h′(p)f ′′(T) (A.12) Finally, the expression for δẆ in (A.10) becomes δẆ = [ √ 3δs12− (Q−TY )δp+TZδT ]ṗ+[ √ 3s12− (R−TS)]δṗ (A.13) References 1. Anand L., Kim K.H., Shawki T.G., 1987, Onset of shear localization in viscoplastic solids, J. Mech. Phys. Solids, 35, 4, 407-429 2. Aravas N., Kim K.S., Leckie F.A., 1990, On the calculations of the stored energy of cold work, J. Eng. Mater. Technol., 112, 465-470 3. Bai Y.L., 1982, Thermo-plastic instability in simple shear, 1982, J. Mech. Phys. Solids, 30, 4, 195-207 4. Bai Y.L., Dodd B., 1982,Adiabatic Shear Localization, PergamonPress 5. Bataille J., Kestin J., 1975, Linterprétation physique de la thermodynami- que rationnelle, J. de Mécanique, 14, 2, 365-384 6. Batra R.C., Chen L., 2001, Effect of viscoplastic relations on the instability strain, shear band initiation strain, the strain corresponding to the minimum shear band spacing, and the band width in a thermoviscoplastic material, Int. J. Plasticity, 17, 1465-1489 7. Batra R.C., Wei Z.G., 2006, Shear band spacing in thermoviscoplasticma- terials, Int. J. Impact Eng., 32, 947-967 8. Clayton J.D., 2005,Dynamic plasticity and fracture in high density polycry- stals: constitutive modeling and numerical simulation, J. Mech. Phys. Solids, 53, 261-301 9. Clifton R.J., Duffy J., Hartley K.A., Shawki T.G., 1984, On critical conditions for shear band formation at high strain rates, Scripta Met., 18, 443-448 10. Estrin Y., Molinari A., Mercier S., The role of rate effects and of ther- momechanical coupling in shear localization, J. Eng. Mat. Tech., 119, 322-330 222 P. Longère, A. Dragon 11. Fressengeas C., Molinari A., Instability and localization of plastic flow in shear at high strain rates, J. Mech. Phys. Solids, 35, 2, 185-211 12. Kapoor R., Nemat-Nasser S., 1998,Determination of temperature rise du- ring high strain rate deformation,Mech. Mat., 27, 1-12 13. Liao S.-C., Marchand A., 1998, Adiabatic shear bands in a Ti-6Al-4V tita- nium alloy, J. Mech. Phys. Solids, 46, 11, 2201-2231 14. Longère P., Dragon A., Trumel H., De Resseguier T., Deprince X., Petitpas E., 2003,Modelling adiabatic shear banding via damagemechanics approach,Arch. Mech., 55, 3-38 15. Longère P.,DragonA., TrumelH., DeprinceX., 2005,Adiabatic shear banding induced degradation in a thermo-elastic/viscoplastic material under dynamic loading, Int. J. Impact Eng., 32, 285-320 16. Longère P., DragonA., 2006, Plastic work induced heating under dynamic conditions: critical assessment, submitted for publication 17. MarchandA., Duffy J., 1988,An experimental study of the formation pro- cess of adiabatic shear bands in a structural steel, J. Mech. Phys. Solids, 36, 3, 251-283 18. Mason J.J., Rosakis A.J., Ravichandran G., 1994, On the strain and strain rate dependence of the fraction of plastic work converted to heat: an experimental study using high speed infrared detectors and the Kolsky bar, Mech. Mat., 17, 135-145 19. Molinari A., 1985, Instabilité thermoviscoplastique en cisaillement simple, J. Méca. Théorique et appliquée, 4, 659-684 20. Mróz Z., Oliferuk W., Energy balance and identification of hardeningmo- duli in plastic deformation processes, Int. J. Plasticity, 18, 379-397 21. Odeshi A.G., Al-Ammeeri, Mirfakhraei, Yazdani F., Bassim M.N., 2006, Deformation and failure mechanism in AISI 4340 steel under ballistic impact,Th. Appl. Fract. Mech., 45, 18-24 22. PęcherskiR.B., 1998,Macroscopiceffects ofmicro-shearbanding inplasticity of metals,Acta Mech., 131, 203-224 23. PerzynaP., 1966,Fundamental problems in viscoplasticity,Adv.Appl.Mech., 9, 243-377, Academic Press, NewYork 24. Perzyna P., 1998, Dynamic localized fracture in inelastic solids, In: Damage Mechanics in Engineering Materials, 183-202,G.Z. Voyiadjis, J.-W.W. Ju and J.-L. Chaboche (Edit.), Elsevier 25. PetrykH., 2000,General conditions for uniqueness inmaterialswithmultiple mechanisms of inelastic deformation, J. Mech. Phys. Solids, 48, 367-396 26. Raniecki B., Nguyen H.V., 1984, Isotropic elastic-plastic solids at finite strain and arbitrary pressure,Arch. Mech., 36, 5-6, 687-704 Adiabatic heat evaluation for dynamic plastic localization 223 27. Rhim S.H., Oh S.I., 2006, Prediction of serrated chip formation in metal cutting process with new flow stress model for AISI 1045 steel, J. Mat. Proc. Tech., 171, 417-422 28. Sidoroff F., Dogui A., 2001, Some issues about anisotropic elastic-plastic models at finite strain, Int. J. Solids Struct., 38, 9569-9578 29. TaylorG.I.,QuinneyH., 1934,The latent energy remaining in ametal after cold working,Proc. Roy. Soc., A413, 307-326 30. Voyiadjis G.Z., Abed F.H., 2006, A coupled temperature and strain rate dependent yield function for dynamic deformations of bcc metals, Int. J. Pla- sticity, 22, 1398-1431 31. Wei Q., Jiao T., Ramesh K.T., Ma E., Kecskes L.J., Magness L., DowdingR.,KzaykhanovV.U.,ValievR.Z., 2006,Mechanical behaviour and dynamic failure of high-strength ultrafine grained tungsten under uniaxial compression,Acta Mater., 54, 77-87 Szacowanie adiabatycznego ciepła w procesie dynamicznej lokalizacji odkształceń plastycznych Streszczenie Adiabatyczne ogrzewaniew stanie plastycznym jestwwiększości przypadkówopi- sywane za pomocą ułamkowego ciepła zwanegowspółczynnikiemTaylora-Quinney’a, zwykle przyjmowanego jako parametr o stałej wartości. Opis taki, z punktu widzenia termodynamiki, zasadza się na pomijaniu (lub uwzględnianiu jedynie pewnej części) sprzężeń termomechanicznych w równaniu definiującym bilans cieplny. Taki zgrubny opismoże prowadzić do przeszacowaniawzrostu temperaturyw obliczeniachdotyczą- cych plastycznego ogrzewania.Wpracy przedstawiono proces adiabatycznego formo- wania pasm ścinaniawynikający z termicznej niestabilności. Pokazano, że dokładność w przewidywaniu korzystnych czynników dla inicjacji procesu lokalizacji odkształ- ceń plastycznych silnie zależy od metody szacowania ciepła powstającego wskutek uplastycznienia. Celem rozważań jest pokazanie wpływu różnego stopnia uproszczeń przyjętychw równaniu bilansu cieplnego nawarunki odpowiadające za pojawienie się lokalizacji odkształceń plastycznych. Te ostatnie wyznaczono za pomocą kryterium opartego na liniowej metodzie perturbacyjnej. Manuscript received June 1 2006; accepted for print December 1, 2006