Jtam-A4.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 55, 2, pp. 595-606, Warsaw 2017 DOI: 10.15632/jtam-pl.55.2.595 MODELING OF CYCLIC THERMO-ELASTIC-PLASTIC BEHAVIOUR OF P91 STEEL Piotr Sulich, Władysław Egner Cracow University of Technology, Faculty of Mechanical Engineering, Institute of Applied Mechanics, Kraków, Poland e-mail: piotrjansulich@gmail.com; wladyslaw.egner@pk.edu.pl Stanisław Mroziński UTP University of Science and Technology, Faculty of Mechanical Engineering, Bydgoszcz, Poland e-mail: stanislaw.mrozinski@utp.edu.pl Halina Egner Cracow University of Technology, Faculty of Mechanical Engineering, Institute of Applied Mechanics, Kraków, Poland e-mail: halina.egner@pk.edu.pl Thermomechanical low cycle fatigue behaviour of P91 steel used in power industry appli- cations has been extensively investigated. The constitutive model of Armstrong-Frederick, extended with temperature rate effects, has been applied to describe the behaviour of the thermo-elastic-plastic material. The proposed model has been successfully implemented in simulation of low cycle fatigue of the examined steel in two different temperatures. Keywords: constitutive modeling, low cycle fatigue, identification of material parameters 1. Introduction Thermomechanical lowcycle fatigueaccompaniedbyelastic-plastic strains is oneof thedominant failure modes in high temperature structural components such as electric power boilers, boiler pipes, engine elements, etc. Extensive research on the behaviour of various engineeringmaterials under low cycle fatigue conditions has been carried out for the last 50 years in order to develop the adequate constitutive modeling as well as the appropriate predictions of the fatigue lifetime (Taleb et al., 2006; Zhang et al., 2008; Ganczarski and Skrzypek, 2009; Taleb and Cailletaud, 2010; Skrzypek andGanczarski, 2015). The aim of this paper is to work out a procedure of comprehensive analysis of thermome- chanical low-cycle fatigue behaviour on the example of P91 steel, widely used in power industry applications (Duda et al., 2016). Firstly, the material is tested experimentally in two test tem- peratures, then thematerial behaviour is described by a constitutive model implemented into a numerical procedure. Next, the material characteristics are identified in different test tempera- tures. Finally, the numerical simulations of fatigue tests are performed and compared with the experimental results. To reduce the number ofmaterial parameters for identification, theArmstrongandFrederick constitutive model extended with temperature rate effects is here applied. The present work is treated as an initial step in the comprehensive analysis of P91 steel behaviour in thermomechanical fatigue conditions. For this reason, classical Armstrong and Frederick constitutive model and only two testing temperatures for parameter identification are considered. However, the procedure of analysis is general and can be applied to amore complex modeling. 596 P. Sulich et al. 2. Behaviour of P91 steel 2.1. Experimental equipment and material investigated Tested specimens (see Fig. 1) were cut out of a boiler pipe of diameter d = 200mm and wall thickness t = 20mm (see Fig. 2). The chemical composition of steel is shown in Table 1. Low cycle fatigue tests were strain controlled with constant total strain amplitude (εac = const, frequency of loading 0.2Hz) and constant temperature in each test. Fig. 1. Shape and dimensions of the tested specimen Fig. 2. Sampling procedure Table 1.Chemical composition of the steel C Si Mn P S Cr Mo Ni Al Co Cu Nb Ti V W 0.197 0.442 0.489 0.017 0.005 8.82 0.971 0.307 0.012 0.017 0.036 0.074 0.004 0.201 0.02 Table 2. Loading scheme, strain amplitude and temperature Loading scheme εaci [%] Ti [ ◦C] εac1 =0.25 T1 =20 εac2 =0.30 T2 =600 εac3 =0.35 εac4 =0.50 εac5 =0.60 Five levels of the total strain amplitude, and two levels of temperature were applied (see Table 2). Experiments were performed on the testing machine Instron 8502 equipped with a heating chamber. Modeling of cyclic thermo-elastic-plastic behaviour of P91 steel 597 The temperaturewas controlled by the use of a thermocouple attached to the samplemeasu- ring section. Thematerial deformationwas determined bymeans of a strain gauge extensometer (gauge length 12.5mm) (see Fig. 3). Fig. 3. The test stand: (a) ekstensometer mounting, (b) heating chamber 2.2. Material behaviour The tested steel exhibits cyclic softening, regardless of the testing temperature (half-stress amplitudedecreaseswith increasing cumulatedplastic strain, cf.Golański andMroziński (2013)). This softening could be divided into three phases, which are: the rapid softening phase during the initial few hundred cycles followed by a slow quasi-linear softening phase, and finally again fast softening till rupture (see Fig. 4) (cf. Bernhart et al., 1999; Mebarki et al., 2004). The first phase is generally explained by a rapid change in the dislocation density inherited from the quench treatment, the second is related to the formation of dislocation sub-structure and carbide coarsening under the action of time, temperature and cyclic load, while the third phase is a consequence of micro-damage development in the material that ultimately causes failure of the tested sample (cf. Seweryn et al. 2008; Szusta and Seweryn, 2010). 3. Constitutive model of a thermo-elastic-plastic material 3.1. Basic assumptions In constitutive modeling, the well-known formalism of thermodynamics of irreversible pro- cesses with internal state variables and the local statemethod are often adopted (Maugin, 1999; OttosenandRistinmaa, 2005;Chaboche, 1997a,b, 1986; SkrzypekandKuna-Ciskał, 2003;Egner, 2012). In this approach, we consider a material as a specific portion of the physical universe, called a system. The current state of a system is entirely determined by certain values of some independent variables, called variables of state, which can be scalars, vectors, or tensors (matri- ces), such as temperature (scalar) or strain (second order tensor). For a thermo-elastic-plastic material exhibiting mixed hardening the following set of state variables is defined {V stα }= {ε e ij;αij,p;θ} (3.1) where εeij are components of the reversible (elastic) strain tensor, αij corresponds to kinematic plastic hardening, p is the accumulated plastic strain p = t∫ 0 √ 2 3 ε̇ p ijε̇ p ij dt (3.2) and θ is the absolute temperature in Kelvins. 598 P. Sulich et al. Fig. 4. Cyclic softening of the steel at a constant total strain amplitude ε ac =0.6% In the case of infinitesimal deformation, the total strain tensor components εij are expressed as the sum of the elastic (reversible) εeij, plastic (irreversible) ε p ij and thermal ε θ ij strains εij = ε e ij +ε p ij +ε θ ij (3.3) while thermal strain is expressed as εθij =α θ ij(θ)(θ−θ0) (3.4) with αθij(θ) standing for components of the thermal expansion tensor, and θ0 for the reference temperature at which no thermal strains exist. 3.2. State potential and equations of the state The constitutive behaviour is defined by the specification of two potentials: energy potential and dissipation potential. The state potential is a closed, convex, and scalar-valued function of the overall state variables. The Helmholtz free energy ψ is here used, decomposed into thermo- elastic ρψte and thermo-plastic ρψtp terms (ρ is mass density) ρψ(V stα )= ρψ te(εeij;θ)+ρψ tp(αij,p;θ) (3.5) The following classical functions are here adopted (cf. for ex. Ottosen and Ristinmaa, 2005; Egner and Egner, 2015) Modeling of cyclic thermo-elastic-plastic behaviour of P91 steel 599 ρψte = ρh(θ)+ 1 2 (εij −ε p ij)Eijkl(θ)(εkl−ε p kl )−βij(θ)(εij −ε p ij)(θ−θ0) βij = Eijkl(θ)α θ kl(θ) ρψtp = 1 3 C(θ)αijαij +Q(θ) [ p+ 1 b(θ) e−b(θ)p ] (3.6) where h(θ) is a function of temperature, Eijkl(θ) denote components of the elastic stiffness tensor, and C(θ), Q(θ), b(θ) stand for temperature dependent material parameters. The thermodynamic forces conjugated to state variables (3.1) result from the assumed form of state potential (3.5) and are defined by the following state equations σij = Eijkl(θ)(εkl−ε p kl )−βij(θ)(θ−θ0) Xij = 2 3 C(θ)αij R = Q(θ) ( 1− e−b(θ)p ) (3.7) σij is here the stress tensor (thermodynamic force conjugated to elastic strain ε e ij), Xij de- notes the back stress (conjugated to plastic hardening variable αij), and R is the drag stress (conjugated to accumulated plastic strain p). If we now define the components of the thermodynamic conjugate force vector {Jα} and the flux vector components {Ṗα} as {Jα}= { Jmechα ;J θ α } = { σij,Xij,R; θ,i θ } {Ṗα}= { Ṗmechα ; Ṗ θ α } = { ε̇ p ij,−α̇ij,−ṗ;−qi } (3.8) then the dissipation inequality can be expressed as the scalar product of Jα and Ṗα as follows (Krajcinovic, 1996) π = JαṖα ­ 0 (3.9) where π is the dissipation function. 3.3. Dissipation potential and evolution equations A constitutive model that fulfills the Clausius-Duhem inequality fulfills all formal require- ments. However, this does not guarantee that the model provides a good approximation of the real material behaviour. If the internal state variables chosen in the modeling are not identified with underlying physical mechanisms responsible for dissipation, the theory may be physically empty (Maugin, 1999). There are various approaches for the establishment of the rate laws, so that thedissipation inequality is fulfilled.Themost often used is thepotential approachbased on the assumption of the existence of a dissipation potential F, being a closed, convex, and scalar- valued function of thermodynamic forces (3.8)1, and someother possible variables. Thepotential of dissipation F is here assumednot equal to plastic yield surface anddependent on temperature (non-associated thermo-plasticity). This allows obtaining non-linear plastic hardening rules (cf. Ganczarski et al., 2010) F(Jα,θ)= f(Jα,θ)+ 3γ(θ) 4C(θ) Xij(αij,θ)Xij(αij,θ) f(Jα,θ)= √ 3 2 [sij −Xij(αij,θ)][sij −Xij(αij,θ)]− [σy(θ)+R(p,θ)] (3.10) where f(Jα,θ) is the vonMises plastic yield surface, sij is the stress deviator, σy(θ) denotes the yield stress, and γ(θ) is another temperature dependent material parameter. 600 P. Sulich et al. According to the generalized normality rule (cf. Chaboche, 2008), the following classical rate equations are obtained ε̇ p ij = λ̇ p sij −Xij√ 3 2 [sij −Xij(αij,θ)][sij −Xij(αij,θ)] α̇ij = ε̇ p ij − 3γ 2C Xijṗ ṗ = λ̇ p (3.11) The kinetic equations of force-like variables are obtained by taking time derivatives of functions (3.7) σ̇ij = Eijkl(θ)(ε̇kl− ε̇ p kl )−Pijθ̇ (3.12) where Pij =− ∂Eijkl(θ) ∂θ (εkl−ε p kl )+ ∂βij(θ) ∂θ (θ−θ0)+βij(θ) Ẋij = 2 3 C(θ)α̇ij + 2 3 dC(θ) dθ αijθ̇ Ṙ = Q(θ)b(θ)e−b(θ)pṗ+ [dQ(θ) dθ ( 1− e−b(θ)p ) +Q(θ) db(θ) dθ e−b(θ)pp ] θ̇ (3.13) 3.4. Consistency condition Calculation of theLagrangemultiplier λ̇p(θ) for rate-independentmaterial needsmaking use of the consistency condition ḟ(Jα,θ)= ∂f ∂σij σ̇ij + ∂f ∂Xij Ẋij + ∂f ∂R Ṙ+ ∂f ∂θ θ̇ =0 (3.14) The classical Kuhn-Tucker loading/unloading conditions have the form f ¬ 0 and ḟ    < 0 and λ̇p =0 ⇒ passive loading =0 and λ̇p =0 ⇒ neutral loading =0 and λ̇p > 0 ⇒ active loading (3.15) Substituting equations (3.7) and (3.11) into equation (3.14) leads to the following ḟ = ∂f ∂σij σ̇ij − λ̇ p [2 3 C ∂f ∂σij ( ∂f ∂σij −γαij ) − ∂f ∂R Qbe−bp ] − { 2 3 dC dθ ∂f ∂σij αij − ∂f ∂R [dQ dθ ( 1− e−bp ) +Q db dθ e−bpp ] − ∂f ∂θ } θ̇ = ∂f ∂σij σ̇ij − λ̇ pH − θ̇S =0 (3.16) In the above equation, H is the generalized hardeningmodulus H = 2 3 C ∂f ∂σij ( ∂f ∂σij −γαij ) − ∂f ∂R Qbe−bp (3.17) and S reflects the sensitivity of the yield surface on temperature changes (cf. Egner, 2012) S = 2 3 dC dθ ∂f ∂σij αij − ∂f ∂R [dQ dθ ( 1− e−bp ) +Q db dθ e−bpp ] − ∂f ∂θ (3.18) Expression (3.16) determines the consistency multiplier λ̇p(θ)= 1 w [ ∂f ∂σij Eijklε̇kl− ( ∂f ∂σij Pij +S ) θ̇ ] (3.19) where w = ∂f ∂σij Eijkl ∂f ∂σkl +H (3.20) Modeling of cyclic thermo-elastic-plastic behaviour of P91 steel 601 3.5. Heat balance equation To determine the temperature distribution within the body, the heat balance equation is used, derived from the first law of thermodynamics by substituting into it the internal energy density together with Fourier’s law. The heat balance equation takes the form (cθε is the specific heat capacity at a constant strain and r is the distributed heat source per unit volume) ρcθεθ̇ =−qi,i+r−θPij(ε̇− ε̇ I ij)+σijε̇ p ij − ( R−θ ∂R ∂θ ) ṗ− ( Xij −θ ∂Xij ∂θ ) α̇ij (3.21) 4. Numerical implementation and results 4.1. Numerical algorithm In the case of uniaxial loading (tension/compression), the tensorial quantities: stress tensor, strain tensor and back stress tensor may be presented in the following matrix forms [σij] = σ   1 0 0 0 0 0 0 0 0   [sij] = σ   2 3 0 0 0 −1 3 0 0 0 −1 3   [εij] = ε e   1 0 0 0 −ν 0 0 0 −ν  +εp   1 0 0 0 −1 2 0 0 0 −1 2  +αθ(θ−θ0)   1 0 0 0 1 0 0 0 1   [Xij] = X   2 3 0 0 0 −1 3 0 0 0 −1 3   [αij] = α   1 0 0 0 −1 2 0 0 0 −1 2   (4.1) Thenumerical procedure implementingEAFmodel has been built by the use ofMathematica 10 software according to the algorithms based on the classical backward Euler scheme and the Newton-Raphson method (cf. Chaboche and Cailletaud, 1996) (see Fig. 5). The vector ∆S = [∆λ,∆εe,∆θ,∆α]T contains increments of the unknowns, and is iteratively calculated according to ∆S(k+1) = ∆S(k)− [J(k)] −1R(∆S(k)) (4.2) In the above equation [J] = ∂R/∂∆S is the Jacobian matrix and R(∆S) is the residual vector containing the components Ri = ∆Si − ∆̂Si, where ∆Si is a variable while ∆̂Si denotes the function resulting from the evolution rule for the i-th variable Si. It is evident that the condition R(∆S)= 0 defines the solution. If we expand this condition into aTaylor series, we obtain (4.2). The iteration procedure is stopped when the norm of R is sufficiently small. 4.2. Identification of model parameters The identificationofmodelparameters hasbeenperformedwith theapplicationof SIMULIA- -Isight package (cf. SIMULIAAbaqus Extended Products, 2014), which provides a platform for automatic optimal selection of material parameters. For this purpose, two components offered by the program: “DataMatching” and “Optimization” have been used. “Data Matching” com- ponent allows one to calibrate the model by analyzing different error measures between the experimental data and numerical simulation results. The following vector of material parame- ters Pi is searched {Pi}= {σy,E,γ,C,b,Q} Pi ∈ 〈Li,Ui〉 (4.3) 602 P. Sulich et al. Fig. 5. Numerical algorithm The parameters are bounded between their respective lower bounds Li and upper bounds Ui. Before the identification starts, the parameters are subjected to normalization, so that P i ∈ 〈−1,1+〉 Pi = 1 2 [(1−P i)Li+(1+P i)Ui] (4.4) In thepresent analysis, the errormeasureused is the square root of the sumof squareddeviations Fobj(P i)= m∑ k=1 [σk(Pi)−σ e k] 2 (4.5) where σek denote the experimental stress data and σk(Pi) are the stress data calculated numeri- cally by the use of current values of the model parameters Pi. The “optimization” component is applied to find an optimal solution in the user-defined field, with the assumed constraints and the objective function defined by (4.5). The constraints limit the search field to the range of acceptable physical values. To reduce the dimension of the field in which the optimal solution is searched, the following procedure has been applied: the initial yield stress σy and the elastic modulus E are identified manually, considering the Modeling of cyclic thermo-elastic-plastic behaviour of P91 steel 603 monotonic tensile part of the first hysteresis loop. Next, considering the whole first hysteresis loop, the approximate values of kinematic hardening parameters (γ and C) are looked for on the assumption that the isotropic hardening is negligible (b =0 and Q =0) min{Fobj(P i) : E = const ∧ σy = const ∧ b =0 ∧ Q =0} (4.6) Then the parameters related to the isotropic hardening are searched with the use of several chosen hysteresis loops (loop 2 to 20) min{Fobj(P i) : E = const ∧ σy = const ∧ γ = const ∧ C = const} (4.7) As a result, the approximate values of material parameters, constituting the starting po- int for optimization are set, and the identification of all the unknown material parameters (E,σy,γ,C,b,Q) is performed once again, but in a substantially limited range around the initial point. The results of the applied identification procedure are presented in Table 3. Please note that these data ensure the best fit of the results simulated numerically into the experimental results for the chosen objective function Fobj(P i). However, the constitutive model does not account for nonlinear elastic effects, therefore the yield stress is clearly underestimated relative to the offset yield stress Rp02 for the considered material (cf. Mroziński and Piotrowski, 2013). Table 3.Material parameters (εac =0.60%) 20◦C 600◦C σy [MPa] 278 184 E [MPa] 198000 159000 γ 595 752 C [MPa] 130420 89120 b 1.02 1.88 Q [MPa] −39 −69 The experimental observations indicate that steel is not stable during fatigue, and that the microstructure can be modified by the thermal cycle. However, such a case takes place when temperatures reach or exceed the tempering temperatures, even for a short time (cf. Zhang et al., 2008). In other words, there are two ranges of temperature, in which the fatigue behaviour of the steel is different: (1) Above the tempering temperature, a sharp ageing is observed while the fatigue test addi- tionally enhances themicrostructural evolutions related to ageing. In such a case, changes of mechanical properties of the steel are induced by factors related independently to the history of temperature. As a consequence, it is necessary to include additional parameters (state variables) for correct description of such an influence of temperature in a constitu- tivemodel. The identification of model parameters, being temperature history dependent, should be done for all temperatures by identifying with all tests together with the cho- sen material functions of temperature (cf. Cailletaud et al., 2000). For this purpose, an experimental thermomechanical fatigue tests should be performed and used for parameter identification. (2) On the other hand, below the tempering temperature (the case considered in the present analysis) the ageing remains nearly constant, so that themechanical properties depend on the current temperature and not on the history of temperature. In such a case, there is no need to include in the identification procedure all temperatures together, but to use several isothermal tests in different temperatures, and introduce the influence of temperature on the material parameters by interpolation techniques with polynomial or spline functions. 604 P. Sulich et al. 4.3. Simulation of the tests under strain control Kinematic hardening equations (3.11)2 and (3.13)2, proposed by Armstrong and Frederick (1996), introduce a recall term,which is collinearwith the back stressXij. For amonotonic unia- xial loading, the evolution of Xij becomes exponential with a saturation value C/γ. Integration of (3.13)2 with respect to ε p ij (for the isothermal case) yields Xij = ν C γ ( 1− e −νγε p ij ) (4.8) where ν = ±1 indicates the flow direction. Function (4.8) is shown in Fig. 6a for two testing temperatures. It can be seen that the translation of the yield surface for the same strain level is more pronounced in lower temperatures. Fig. 6. (a) Evolution of back stress versus plastic strain. (b) Evolution of R versus accumulated plastic strain The R-p curve obtained on the basis of Eq. (3.7)3 is shown in Fig. 6b. The parameters Q and bdependon thematerial and temperature.For cyclic loading, the value of b is usually placed in the range between 0.5 and 50. In the context of monotonic loading, the value of b should be much higher (cf. Chaboche, 2008). Fig. 7. Comparison between experimental and calculated (a) responses with a strain rate of 10−2 and (b) maximum stress vs. cycles at temperatures of 20◦C and 600◦C Figure 7a presents a comparison between the test results used in the identification process and their simulations by the described model using the material parameters for Table 3. As shown in Fig. 7a, the model simulates the first hysteresis loop with a very good accuracy. Also, stage I of the tests used in the identification process (the rapid softening phase during the initial few hundred cycles) is well reflected, see Fig. 7b. Modeling of cyclic thermo-elastic-plastic behaviour of P91 steel 605 The identification is performedwith the use of chosen hysteresis loops only, and notwith the whole fatigue curve. However, when the optimal values of model parameters are obtained, the (numerically obtained)maximum stress on cycle versus cycle number exhibits a good agreement with the results of all the first 100 cycles of the experiment (see Fig. 7b). 5. Conclusions In the present paper, the algorithm for a comprehensive analysis of thermomechanical fatigue behaviour is presented. Such an analysis consists of fivemain steps: (1) experimental testing in several test temperatures, (2) constitutivemodeling ofmaterial behaviour regarding the efects of temperature change, (3) numerical implementation of themathematicalmodel, (4) identification of model parameters in different test temperatures to obtain temperature-dependent material characteristics, and (5) validation of the analysis by comparison between the experimental and numerical results. The classical Armstrong and Frederick constitutive model used in the presented analysis is not capable of describing different physical mechanisms able to produce material nonlinearities on a macro-scale. For this reason, a more advanced constitutive model should be adopted to properly refelect the material behaviour under different loading paths (cf. Besson et al., 2009; Saanouni, 2012; Egner andEgner, 2014). However, such advancedmodels involvemanymaterial parameters that need to be identified. This is why in this research a simplemodel is considered, but the described procedure is general and can be applied also to amore compehensivemodeling (cf. Velay et al., 2006; Säı, 2011). References 1. ArmstrongP.J., Frederick C.O., 1966,Amathematical representation of themultiaxial Bau- schinger effect, Report RD/B/N731, CEGB, Central Electricity Generating Board, Berkeley, UK 2. Bernhart G., Moulinier G., Brucelle O., Delagnes D., 1999, High temperature low cycle fatigue behaviour of amartensite forging tool steel, International Journal of Fatigue,21, 2, 179-186 3. Besson J., Cailletaud G., Chaboche J.L., Forest S., Blétry M., 2009,Non-Linear Me- chanics of Materials, Springer 4. Chaboche J.L., 1997a, Thermodynamic formulation of constitutive equations and application to the viscoplasticity and viscoelasticity of metals and polymers, International Journal of Solids and Structures, 34, 18, 2239-2254 5. Chaboche J.L., 1997b, Viscoplastic constitutive equations for the description of cyclic and ani- sotropic behaviour of metals, Bulletin de L’Academie Polonaise des Sciences, Série des Sciences Techniques, XXV, 1, 33-39 6. Chaboche J., 1986, Time independent constitutive theories for cyclic plasticity, International Journal of Plasticity, 2, 2, 149-188 7. Chaboche J.L., 2008, A review of some plasticity and viscoplasticity constitutive theories, Inter- national Journal of Plasticity, 24, 1642-1693 8. Chaboche J.L., Cailletaud, G., 1996, Integration methods for complex plastic constitutive equations,Computer Methods in Applied Mechanics and Engineering, 133, 125-155 9. Cailletaud G., Depoid C., Massinon D., Nicouleau-Bourles E., 2000, Elastoviscopla- sticity with aging in aluminium alloys, [In:] Continuum Thermomechanics: The Art and Science of Modelling Material Behaviour (Paul Germain’s Anniversary Volume), Solid Mechanics and Its Applications, Kluwer Academic Publishers, 75-86 10. DudaP., Felkowski Ł., Dobrzański J., PurzyńskaH., 2016,Modelling the strain and stress state under creep conditions in P91 steel,Materials at High Temperatures, 33, 85-93 606 P. Sulich et al. 11. Egner H., 2012, On the full coupling between thermo-plasticity and thermo-damage in thermo- dynamic modeling of dissipative materials, International Journal of Solids and Structures, 49, 279-288 12. Egner H., Egner W., 2015, Classification of constitutive equations for dissipative materials- general review, [In:] Mechanics of Anisotropic Materials, J.J. Skrzypek and A.W. Ganczarski (Edit.), 247-294, Springer 13. Egner H., Egner W., 2014,Modeling of a temperedmartensitic hot work tool steel behavior in the presence of thermo-viscoplastic coupling, International Journal of Plasticity, 57, 77-91 14. Ganczarski A.W., Egner H., Muc A., Skrzypek J.J., 2010, Constitutive models for analy- sis and design of multifunctional technological materials, [In:] Innovative Technological Materials. Structural Properties by Neutrons, Synchrotron Radiation and Modelling, F. Rustichelli and J.J. Skrzypek (Edit.), 179-220, Springer 15. GanczarskiA., SkrzypekJ., 2009,Astudyoncoupled thermo-elasto-plastic-damagedissipative phenomena:models and application to some innovativematerials,Journal of Thermal Stresses,32, 698-751 16. GolańskiG.,Mroziński S., 2013, Low cycle fatigue and cyclic softening behavior ofmartensitic cast steel,Engineering Failure Analysis, 12, 35, 692-702 17. Krajcinovic D., 1996,Damage Mechanics, Elsevier 18. MauginG.A., 1999,TheThermomechanics of Nonlinear Irreversible Behaviors. An Introduction, World Scientific Series on Nonlinear Science, Series A, Vol. 27 19. Mebarki N., Delagnes D., Lamesle P., Delmas F., Levaillant C., 2004, Relationship between microstructure and mechanical properties of a 5%Cr tempered martensitic tool steel, Materials Science and Engineering A, 387-389, 1/2, 171-175 20. Mroziński S., Piotrowski M., 2013, Influence of temperature and loading program on the fatigue life of steel P91,Acta Mechanica et Automatica, 7, 2, 93-98 21. Ottosen N. S., Ristinmaa M., 2005,The Mechanics of Constitutive Modeling, Elsevier 22. Saanouni, K., 2012, Damage Mechanics in Metal Forming: Advanced Modeling and Numerical Simulation, ISTE/Wiley, London. 23. Säı K., 2011,Multi-mechanismmodels: Present state and future trends, International Journal of Plasticity, 27, 250-281 24. SewerynA.,BuczyńskiA., Szusta J., 2008,Damage accumulationmodel for lowcycle fatigue, International Journal of Fatigue, 30, 756-765 25. SIMULIA Abaqus Extended Products, Abaqus 6.14-AP Isight 5.9, Dassault Systemes 1994-2014, http://www.3ds.com/products/simulia 26. Skrzypek J., Ganczarski A., 2015,Mechanics of Anisotropic Materials, Springer 27. Skrzypek J.,Kuna-CiskałH., 2003,Anisotropic elastic-brittle-damageand fracturemodels ba- sed on irreversible thermodynamics, [In:]Anisotropic Behaviour ofDamagedMaterials, J. Skrzypek andA.Ganczarski (Edit.), LectureNotes inApplied andComputationalMechanics, vol. 9, 143-184, Springer 28. Szusta J., Seweryn A., 2010, Low-cycle fatigue model of damage accumulation – The strain approach,Engineering Fracture Mechanics, 77, 1604-1616 29. Taleb L., Cailletaud G., 2010, An updated version of the multimechanism model for cyclic plasticity, International Journal of Plasticity, 26, 859-874 30. TalebL., CailletaudG.,Blaj L., 2006,Numerical simulation of complex ratcheting testswith a multi-mechanismmodel type, International Journal of Plasticity, 22, 724-753 31. VelayV., BernhartG., Penazzi L., 2006,Cyclic behaviormodeling of a temperedmartensitic hot work tool steel, International Journal of Plasticity, 22, 459-496 32. Zhang Z., Bernhart G., Delagnes D., 2008, Cyclic behavior constitutive modeling of a tem- pered martensitic steel including ageing effect, International Journal of Fatigue, 30, 706-716 Manuscript received August 12, 2016; accepted for print December 6, 2016