JOURNAL OF THEORETICAL AND APPLIED MECHANICS 44, 3, pp. 667-689, Warsaw 2006 MODELLING CREEP, RATCHETTING AND FAILURE IN STRUCTURAL COMPONENTS SUBJECTED TO THERMO-MECHANICAL LOADING Fion P.E. Dunne Department of Engineering Science, Oxford University, UK e-mail: fionn.dunne@eng.ox.ac.uk Jianguo Lin Department of Mechanical and Manufacturing Engineering, University of Birmingham, UK e-mail: j.lin@bham.ac.uk David R. Hayhurst John Makin School of Mechanical, Aeronautical and Civil Engineering, University of Manchester, UK e-mail: d.r.hayhurst@manchester.ac.uk Experimentshavebeendescribed inwhich copper components havebeen subjected to combined cyclic thermal and constant mechanical loading. Two thermal cycles were employed leading to predominantly cyclic pla- sticity damage and balanced creep – cyclic plasticity damage loading cycles. The combined loading led to component ratchetting and ultima- tely to failure. Continuum damage-based finite element techniques have been develo- ped for combined cyclic plasticity, creep and ratchetting in components subjected to thermo-mechanical loading. Cycle jumping techniques have been employed within the finite element formulation to minimise com- puter CPU times. The finite elementmethods have been used to predict the behaviour of the copper components tested experimentally and the results compared. Steady-state ratchet rateswere found tobewell predictedby themodels. Modes of failure and component lifetimes were also found to be reasona- blywell predicted.The experimental results demonstrate the importance of isotropic cyclic hardening on the initial component ratchetting rates. Key words: creep damage, cyclic plasticity damage, thermo-mechanical loading, FEmodelling, thermo-mechanical testing 668 F.P.E. Dunne et al. 1. Introduction Over the last two decades, much effort has been devoted to the development of design criteria for structural components operating under extreme mecha- nical, thermal and environmental loading conditions, but it still remains a challenging problem. Components of this type arise, for example, in nuclear and conventional power plant, in pressure vessels and piping, and in aerospa- ce applications (Blackman et al. 1983). Components that operate below the creep regime, may be subjected to regions of high stress which may be indu- ced as a result of severe thermal gradients and mechanical loads, which may lead to the initiation and propagation of damage and micro–cracks due to cyclic plasticity (Remy and Skelton, 1990). Components that operate at high temperatures in the creep regime under cyclic mechanical or thermal loading may, in addition to the above, be subjected to time-dependent deformation and damage processes due to creep, which are dependent on a range of factors such as temperature, rate effects andmean stress (Blackman et al., 1983). At the material level the evolution of time-dependent creep damage and cyclic plasticity damage have been shown to interact (Blackman et al., 1983; Inoue et al., 1985; Lemaitre and Plumtree, 1979), leading to significant re- ductions in component lifetime. Because of this interaction it is not possible to make realistic predictions of material behaviour based on the analysis of the individual processes alone. Hence, material models have been developed over the past two decades based on sets of internal variables (Miller, 1979; Kujowski andMróz, 1980; Bodner andPartoum, 1975; Liu andKrempl, 1979; Chaboche and Rousellier, 1983; Chaboche, 1987; Benallal and Ben Cheikh, 1987; Benallal andMarquis, 1987; Dunne and Hayhurst, 1992a; Dunne et al., 1992), which represent creep and cyclic plasticity damage. At the component level the evolution of creep and cyclic plasticity da- mage leads to material weakening, to stress redistribution, and eventually to material or component failure (Lemaitre and Chaboche, 1990). It is therefore important to be able to model the growth and interaction of creep and cyclic plasticity damage to enable theprediction of component lifetimes.This hasbe- en achieved through the advancement of single damage state variable theories developed at thematerials level by Hayhurst (1972; 1973) and at the structu- ral level, through the development of ContinuumDamage Mechanics (CDM), by Hayhurst et al. (1984a,b) and for creep rupture in 3D welds (Hayhurst et al., 2005a), and by Chaboche (1988) for cyclic plasticity. Suchanalysis techniques fordesignhave beendeveloped for awide range of structural components, includingnotched and crackedmembers inplane stress Modelling creep, ratchetting and failure... 669 (Hayhurst, 1973), plane strain (Hayhurst et al., 1984a), and axisymmetric situations (Hayhurst et al., 1984b);weldments (Hall andHayhurst, 1991;Wang and Hayhurst, 1994) and creep crack growth (Hall et al., 1996; Lesne and Chaboche, 1988) and for components subjected to combined mechanical and thermal loading. More recently, Johanssan et al. (2005) have developed accurate modelling techniques for ratchetting undermultiaxial stress states, which have been ap- plied to cyclic loading of rail steel. Manonukul et al. (2005) have coupled a physically-based creep model with combined isotropic and kinematic harde- ning to examine cyclic plasticity and multiaxial creep in thermo-mechanical fatigue in which failure is dominated by creep damage evolution. Due to high computational cost, these approaches could not be used for components subjected to large numbers of operating cycles over which creep and cyclic plasticity damage interact. To overcome this difficulty, some years ago Dunne and Hayhurst (1992b, 1994) developed cycle jumping techniques which enabled calculations to beperformed in structures and components over many thousands of cycles ofmechanical and thermal loading in the absence of incremental plastic straining or ratchetting. This development enabled large finite element calculations to be carried out using modest computer power (Dunne and Hayhurst, 1992b). More recently, Lin, Dunne and Hayhurst (1997) extended this capability to enable the behaviour of structural components to be calculated when rat- chetting takes place. They demonstrate the effectiveness of the technique by comparing the predictionsmade, using two different methods, of the behavio- ur of a thermally loaded copper slag tap. The first method involved Finite Element analysis of the slag tap component studied by Dunne and Hayhurst (1992b), and the second method involved the modelling of the same slag tap using a multi-bar equivalent of the minimum loading section of the tap. Fol- lowing this approach Lin et al. (1997) showed very close agreement betwe- en the two methods, hence validating the multi-bar modelling method for zero mechanical load. They also predicted, using the multi-bar model, the ratchetting behaviour of a slag tap subjected to a superimposed mechanical load. In this paper, continuum damage – based finite element models are de- scribed for the analysis of components undergoing creep, cyclic plasticity, and ratchetting undercombined thermo-mechanical loading.Experimental tests on copper components are reportedand the results comparedwith thepredictions obtainedusing the finite elementmodels. In the next section, the experimental tests are described. 670 F.P.E. Dunne et al. 2. Slag tap ratchetting test facility and thermal fields for cyclic plasticity dominated and balanced creep-cycle plasticity conditions The experimental setup used to carry out ratchetting behaviour on model copper slag taps is discussed in this section. The model slag tap design is based on a previous design due to Dunne andHayhurst (1992b) and reported in detail byDunne et al. (1993) whichwas used to study thermal loading only in the absence ofmechanical loads. The principal variant in the present design is the addition of a mechanical loading facility. 2.1. Copper slag tap specimen The specimen designmay be seen in Figure 1. To provide an indication of the scale, the diameter of the copper water cooling pipe is 12mm, the width of theminimumcross-section is 70.2mm, thewidth of the larger cross-section, defined by the parallel sides of the specimen, is 74.2mm, and thewidth of the loading shanks which contain loading pins is 74.2mm. The overall height of the specimen is 255mm.The depth of the specimen close to and parallel with the water cooling pipe is 56mm, and the larger cross-section parallel to the loading pins is 74.6mm. Fig. 1. Copper slag tap specimens with copper water cooling pipes thermally expanded into specimen, (a) thermocoupled temperature calibration specimen and (b) virgin specimen showing the locations of point A on the cooling duct and point B on the heating radius Modelling creep, ratchetting and failure... 671 The copper water cooling pipes are thermally expanded into the specimen byheating the specimensuniformly to 150◦Candby cooling thepipes to liquid Nitrogen temperature. In this way an excellent mechanical contact is achieved whichhas goodheat transfer properties. Inpractice, the cooling ducts are long relative to the width of the load bearing cross-sections, and so the conditions of stress approximate closely to those of plane strain. The ratio of specimen width to thewater cooling pipe diameter is 70.2/12.0=5.6 hence justifying the existence of plane strain conditions. 2.2. Design and operation of the thermal loading facility Heat is supplied to the specimen by means of two banks of three 500W quartz line lamps, which are mounted symmetrically on either side of the specimen, and reflectors are used to direct heat onto the specimen. Each bank of three heaters and its reflector is supported by two heater support plates, which in turn are connected to the specimen load train by a parallelogram mechanism of linkages. The purpose of thismechanism is to ensure that as the specimenextendsaxially due to ratchetting, and the specimenwidthcontracts, then the two heater banks move together in sympathy so ensuring that the heat radiation remains focussed on the specimen. Full details of experimental set-up are given byHayhurst et al. (2005b). The steady test load is applied by a 100kN servo-hydraulic testing machine. The power output from the lamps (maximum total power of 2.25kW) and the rate of flow of cooling water, are controlled by a programmable micro- processor, thus enabling the selected thermal loading history to be repeated accurately. 2.3. Establishment of thermal field histories The objective of this research is to verify that it is possible to use vi- scoplastic constitutive equations, which embody the interactive evolution of combined cyclic plasticity and creep damage to predict, using continuumme- chanics theories, the ratchetting behaviour which results from the damage interactions. Two areas are of principal interest. They are ratchetting with dominant cyclic plasticity damage, and ratchetting with balanced creep-cyclic plasticity damage. The types of thermal loading cycles required to achieve these are now discussed. 2.3.1. Cyclic plasticity damage dominated cycle In order to achieve these conditions two requirements have to be met. Firstly, creepdamage formationmustbe avoided andsecondly, cyclic plasticity 672 F.P.E. Dunne et al. damage is to be maximised. The former is achieved by maintaining specimen temperaturebelow the creep regime, and if this cannotbe completely achieved, thenbyreducing the time toaminimumforwhich testpiece temperatures enter the regime. The latter is achieved by maximising the temperature differences between the points A and B given in Figure 1 (and Figure 3, see later) at some point within the cycle. This has been brought about bymaximising the cooling water flow rate whilst achieving optimum heat transfer conditions. The selected thermal cyclic history is shown in Figure 2a. To avoid si- gnificant creep, temperatures are maintained below 300◦C, and to maximise cyclic plasticity damage, a thermal down shock of (θB−θA)= 86 ◦Chas been achieved. The overall cycle time has been kept short so that the increment of cyclic plastic damage per cycle exceeds the contribution from creep damage. The details of the cycle are given as follows. Location B is maintained below 300◦C throughout the cycle. For time t in the range 0s < t ¬ 62.1s heating is provided with no cooling; and, for 62.1s < t ¬ 92.1s both heating and co- oling are provided. The maximum temperature difference between A and B is 86◦C which occurs at 70.1 seconds. The overall cycle time is 92.1 seconds. The effectiveness of this selection will be verified by the results of the analysis presented later in the paper. 2.3.2. Balanced creep damage – cyclic plasticity damage cycle To achieve this cycle the temperatures of the heating radius, point B in Figure 1 is permitted to fluctuate; but, in particular to exceed 300◦C and move well into the creep regime. In addition, the cycle contains a hold period of almost 300 seconds, at the temperature of 340◦C as shown in Figure 2b. It is these conditions which allow the creep damage to be similar in magnitude to the cyclic plasticity damage over the cycle. The details of the cycle are given as follows. For 0s < t < 1308s heating is provided with no cooling; for 1308s < t < 1366.1s, both heating and cooling are provided; and for 1366.1s < t ¬ 1562.5s, cooling only is provided. The maximum temperature difference between A and B is 97.8◦C, which occurs at 1366.1 seconds. The overall cycle time is 1562.5 seconds.Once again, the effectiveness of the chosen thermal conditionwill be verified by the results of the analysis presented later in the paper. 2.4. Application of mechanical load The mechanical load was applied to the testpieces using a 100kN servo- hydraulic test machine operated under load control. The testpieces shown in Modelling creep, ratchetting and failure... 673 Fig. 2. Thermal histories at locations A and B defined in Figure 1 for (a) cyclic plasticity dominated cycle and (b) balanced creep-cyclic plasticity cycle Figure 1 were connected to the testingmachine load train through shear pins, which passed through the larger hole in the thicker section testpiece ends, and pinned the loading ties. 3. Constitutive equations Theelastic-viscoplastic damage constitutive equations employed in thepresent work are those described in detail elsewhere (Dunne and Hayhurst, 1992a, 1994) but for completeness, are summarised in Appendix A. The total strain is given by ε, and the plastic and thermal strains by εp and εθ, respectively. The plastic strain rate term, ε̇p, given in (A.1) follows 674 F.P.E. Dunne et al. from the normality hypothesis of plasticity with an appropriately defined vi- scoplastic potential. The internal variable, x, models the kinematic hardening that occurs in cyclic viscoplasticity, and D is a scalar state variable represen- ting combined irreversible material damage due to creep and cyclic plasticity. Thematerial parameters K, n, Ci and γi are all temperature-dependent. The parameter k, in (A.1) is the cyclically hardened yield stress, and is dependent on both temperature and strain range. The physical basis for the temperature and strain range dependence of the cyclically hardened yield stress and of the temperature dependences in the constitutive equations is described in detail elsewhere (Lin et al., 1996), where in addition, equations for the temperature sensitivities are given. Thescalar state variable, D, describingmaterial damage inequations (A.1) to (A.3) is composed of damage due to creep and cyclic plasticity. Themulti- axial form of the evolution equation for creep damage, ω, is given in the present formulation for creep-cyclic plasticity damage interaction (Dunne and Hayhurst, 1992a) in (A.4), where A, v and φ are temperature-dependentma- terial parameters, σ1 and σe are themaximumprincipal and effective stresses respectively, and α describes the multi-axial creep rupture criterion and the corresponding isochronous locus (Hayhurst, 1972). Here α has been assumed to take thevalue 0.7, reportedbyHayhurst et al. (1984a) for commercially pure copper, and so the creep rupture behaviour is dominated by σ1. The damage Dc contributes to the subsequent evolution of creep damage and ismade up of contributions from the creep damage, and the cyclic plasticity damage.When cyclic plasticity damage is zero, as in the steady load creep, Dc becomes ω and it then represents the combined processes of cavity nucleation and growth as discussed byGreenwood (1973). The form of the interaction between creep and cyclic plasticity damage is described below. The multi-axial equation employed for the evolution of cyclic plasticity damage evolution given in (A.5) is that proposed by Chaboche (1988), where AII is the maximum effective stress range in a cycle, p is defined in terms of AII, M is a function of mean stress, and β is a material parameter. Full details of the equations and of the methods used to determine the material parameters have been given by Dunne and Hayhurst (1992a, 1994). The creep-cyclic plasticity damage interaction law proposedbyDunne and Hayhurst (1992a) is adoptedhere, andgiven in equations (A.6)-(A.9). The ela- stic viscoplastic damage constitutive equations for copper have been validated for conditions of creep, cyclic plasticity, and creep-plasticity interactions un- der both cyclic mechanical and cyclic thermal loading (Dunne and Hayhurst, 1992a, 1994). In addition, the material model has been employed successfully Modelling creep, ratchetting and failure... 675 to simulate the behaviour of a structural component subjected to cyclic ther- mal loading (Dunne and Hayhurst, 1992b) leading to cyclic plasticity. It is implemented into a finite element boundary and the initial value solver de- scribed elsewhere (Dunne and Hayhurst, 1992b). For uniaxial cyclic loading, the effect of damage on stress and viscoplastic strain rate is not considered in compression by setting D = 0, Eq. (A.8). This is used to model microcrack closure under compressive loading. For general multiaxial cases, the damage value is set to zero as the elemental hydrostatic stress becomes negative. The enhancement of the model to include the yield dependence on plastic strain range is described in the next section. 4. Finite element modelling of cyclically hardened yield stress k The constitutive equations discussed above are based on the cyclically harde- ned stabilised state, inwhich the transientmaterial behaviour that occurs over comparatively small numbers of cycles due to cyclic hardening is not model- led in detail. This enables economic computer analysis to be performed. The cyclically hardened yield stress, k, in the constitutive equation (A.1) includes two parts: the first is the initial yield stress, and the second is the isotro- pic hardening stress, which is related to the cyclic strain range as discussed by Lin et al. (1996). At the structural level, materials may be subjected to non-uniform stress and strain conditions. As a result, the cyclic strain range over the structure varies spatially, and the yield stress k varies accordingly. A problem therefore arises initially in that since pointwise strain ranges are not known a priori, it is not possible to determine the yield stress k. It is therefo- re necessary to develop a method to obtain the stabilised cyclically hardened yield stress field efficiently and to avoidmodelling in detail the full transitional cycles necessary to achieve the isotropic hardened stress state. Themethod is described below. Estimation of initial yield stress field Thedetermination of the yield stress field is equivalent to the evaluation of the strain range field for the component under strain controlled cyclic loading, since for a given temperature the yield stress is directly related to the cyclic strain range (Lin et al., 1999). The initial cyclic strain range field is estimated according to the applied strain range and the distribution of ratios of elemen- tal effective stress and volume averaged effective stress for the initial elastic solution. 676 F.P.E. Dunne et al. The volume averaged elastic effective stress, Σev, is defined by Σev = ne ∑ i=1 Σei V ei ne ne ∑ i=1 V ei (4.1) where ne is the number of elements in the finite element mesh, Σ e i is the elemental effective stress for the ith element and V ei is the volume of the ith element. In order to estimate the cyclic strain range field, a factor Fi is introduced Fi = √ Σei Σev i =1,2, . . . ,ne (4.2) The initial cyclic effective strain range field is obtained from ∆εei = Fi∆ε i =1,2, . . . ,ne (4.3) where ∆εei is the estimated elemental cyclic strain range for the ith element, and ∆ε is the applied strain range for a strain-controlled cyclic loading test. Given the relationship between the yield stress and the cyclic strain range, the initial yield stress field can be estimated using the empirical relationship (Lin et al., 1996), as ki =153.3[1.139− exp(−1.319∆ε e i)]exp(−2.0429 ·10 −3T) (4.4) where i =1,2, . . . ,ne and ki are the cyclic hardened elemental yield stresses, and T the temperature [K]. Subsequent yield stress field The modelling of the first loading cycle is based on the yield stress field estimated above. The yield stress field for subsequent cycles is estimated from the effective strain rangefield obtained fromtheprevious loading cycle and the initial estimated strain range field. Therefore, the subsequent effective strain field, ∆εi,j+1e , which is used for obtaining the yield stress field is estimated by ∆εi,j+1e = 1 2 (∆εi,je +∆ε i,j−1 e ) i =1,2, . . . ,ne (4.5) where ∆εi,je and ∆ε i,j−1 e are the elemental effective strain ranges for the jth cycle andthe (j−1)thcycle.Theestimated effective strain rangefield, ∆εi,j+1e , can be used to calculate the yield stress field usingEq. (4.4) for the next cycle. Modelling creep, ratchetting and failure... 677 5. Cycle jumping in combined creep and cyclic plasticity It is well known that the determination of lifetimes of structural components requires extremely long computer run times (Dunne and Hayhurst, 1994). In order to overcome this problem, a cycle jumping techniquewas developed and used formodelling deformation and failure behaviour to avoid the necessity of carrying out detailed calculations around all loading cycles, and yet mainta- ining the required solution accuracy. Cyclic plasticity tests carried out on cast copper (Dunne andHayhurst, 1992b) have shown thatmaterial softening, due to creep and cyclic plasticity damage, does not have a significant effect on the load-carrying capacity or on the stress distribution within the material when the variation of the damage is small. Therefore, a cycle jumping technique has been developed as follows. The cyclic behaviour of the stabilised stress, strain and yield stress is obtained by completing calculations over five full cycles. The resulting stress, strain and yield stress fields are then assumed to be constant, and only the creep damage and the cyclic plasticity damage are allowed to evolve andare integrated over a number of loading cycles according to the damage rates, dψ/dN and dω/dN. The number of cycles for which jumping takes place is limited in such a way that the increment of total damage for the element with the highest damage rate does not exceed 0.05, and individually the cyclic plasticity damage ¬ 0.014. These damage increments have been chosen empirically to ensure that the stress, strain and yield stress fields are not influenced significantly, and so that differences of the calculated lifetimes, for a uniaxial plane bar model, with and without the cycle jumping technique are ¬ 0.5% for a wide range of temperature. The automatic selection of the number of cycles jumped and related numerical integration methods for the creep and plasticity damage were reported byDunne andHayhurst (1994). A cycle jump is allowed to take place, if possible, when every five full loading cycle calculations are completed. This ensures the stabilised stress, strain andyield stress fields canbe obtained. The cycle jumping method has been implemented into the finite element solver, Damage XX, to enable the rupture behaviour and lifetimes of cyclic plasticity testpieces to bemodelled accurately and efficiently. For instance, in studies carried out for a uniaxial plane bar model, only 1/15 of the full cycle computation time is required if the cycle jumping technique is used at 20◦C, with strain range 0.6% and strain rate 0.006% s−1, and the difference in the number of cycles to failure with and without the cycle jumping method is approximately 0.2%. 678 F.P.E. Dunne et al. 6. Finite element analysis of components and comparison with experimental results The copper component discussed above has been analysed using the continu- um damage based finite element formulation presented above. The two cyclic thermal loading histories have been considered, together with the axial load applied.The componenthas beenanalysed in twodimensionsunder conditions of plane strain for reasons that will be discussed later. Because of the planes of symmetry, only a quadrant of the component needs to be considered for modelling purposes, and the finite element mesh employed is shown in Figu- re 3. Mesh sensitivity studies have been carried out for thermal and coupled thermal and mechanical loading analyses. A high density FE mesh was cho- sen for regions where stress and thermal gradients are high. Gradients were assessed by the maximum differences of stress and temperature in adjacent elements; and, the overall number of elements was limited by the CPU time. It was first necessary to establish the thermal field that the component was subjected to experimentally. A transient finite element analysis was carried out using as boundary conditions the experimental temperature variations at points A and B shown inFigure 2. The spatial variations of temperaturewith time were then obtained for both thermal cycles which were used as input to the mechanical analysis of the component. The component was in addition subjected to a constant axial load shown schematically in the finite element mesh in Figure 3. The results for the cyclic plasticity damage dominated cycle shown inFigure 2a are considered first, forwhich a constantmechanical stress of 30MPa was applied. 6.1. Cyclic plasticity damage dominated cycle Effective stresses have been calculated over a stabilised cycle, and the results are shown for these locations on the component in Figure 4. The effective stresses are shown (1) near to the cooling duct, (2) near the centre of the quadrant of the component shown inFigure 3, and (3) near the heating region. The results show that the highest effective stress occurs near the cooling duct. The rapid decrease in the stress that occurs at the beginning of the cycle results from the corresponding decrease in temperature difference shown in Figure 2a. The effective stresses at the cooling duct are smallest when the temperature difference is smallest. The stresses at the quadrant centre point remain fairly constant over the cycle reflecting the uniformity of the stress field in this vicinity generated by the applied constant stress of 30MPa. Modelling creep, ratchetting and failure... 679 Fig. 3. Finite element mesh of the slag tapmodel used for the numerical analysis of the component under combined thermal andmechanical loading Fig. 4. Variation of effective stress over a stabilised cycle for the three locations (1) near the cooling duct, (2) the mid-point of the quadrant, and (3) near the heating region. The results are shown for the cyclic plasticity dominated cycle with an average applied stress of 30MPa at the minimum section Figure 5 shows the predicted and experimentally measured component di- splacement with the number of cycles. It should be noted that the predicted displacements are calculated as those occurring at the top boundary shown in Figure 3. The experimental displacements aremeasured over the full length of the specimen; that is, at the top andbottom component free surfaces just abo- 680 F.P.E. Dunne et al. ve and below the sets of extensometer holes at the four extreme corners of the testpiece, as shown in Figure 1. However, because of the much larger sections of the component in the regions remote from the heating and cooling regions, the majority of the vertical strain occurs locally to these regions. For this re- ason, it is reasonable to assume that the displacementsmeasured over the full length of the specimen are close to those that would be measured over the length corresponding to the finite elementmodel in Figure 3, since the thicker parts of the testpiece are likely to be undergoing rigid body displacement only without significant straining. Fig. 5. Predicted and experimental variation of component loading direction displacement with cycles. The results are shown for the cyclic plasticity damage dominated cycle The results show that very reasonable agreement in life (predicted:∼ 3500 cycles, experimental: ∼ 4300 cycles) is obtained, and in addition, that the steady-state ratchet rate is well predicted. Themajor difference in the results comes from the assumption in themodel of an instantaneously fully-cyclically hardened state of thematerial. The transient process of cyclic hardening that occurs over the first 100 cycles in the material is ignored in the model. The consequence is that initially, the model assumes a much harder material than is actually the case, and therefore fails to predict the primary ratchetting that occurs over the first few cycles which leads to displacements of approximately 0.1mm, shown in Figure 5. The overall quality of the prediction is therefore reasonable and the trend of the secondary and tertiary ratchetting is well predicted. Figure 6 shows the variation of total strain in the loading direction with number of cycles at the heating radius and cooling duct. The strain can be seen to be the largest at the cooling duct with initially very low ratchet rates. The ratchet rates subsequently increase progressively with increasing levels Modelling creep, ratchetting and failure... 681 Fig. 6. Variation of total strainwith number of cycles at locations A and B specified in Figure 3. The results are shown for the cyclic plasticity damage-dominated cycle Fig. 7. (a) Predicted spatial variations of creep (1) and cyclic plasticity (2) damage at component failure and at a life fraction of 0.92, (3) and (4) respectively, shown for the cyclic plasticity damage-dominated cycle, and (b) failed slag tap (plan view) tested under an applied averagemechanical stress of 30MPa for the cyclic plasticity-dominated cycle, showing the fracture surface and chill-cast grain structure and stable cyclic plasticity damage evolution and crack growth of damage. The distribution of damage predicted by the model can be seen in Figure 7a. The figure shows the creep and cyclic plasticity damages at failure ((1) and (2) respectively) and at a life fraction of 0.92 ((3) and (4) respectively). The comparatively low levels of creep damage at failure can be seen to occur removed from both the heating region and cooling duct, and to occur in the mid-section of the quadrant shown. This results from the strong 682 F.P.E. Dunne et al. dependence of the creep damage rate on the maximum principal stress in copper. The maximum principal stresses remain fairly uniform and constant in the quadrant mid-section, leading to the creep damage shown. The cyclic plasticity damage initiates in three separate locations: at the cooling duct, at the heating radius, and at the quadrantmid-section. At failure the three zones of damage link leading to rupture along the component minimum section. This was observed in experiments, and an example is given in Figure 7b, which shows a plane view of the fracture surface, the chill-cast grain structure and stable cyclic plasticity damage evolution and crack growth through the testpiece minimum section. 6.2. Balanced creep and cyclic plasticity damage cycle The thermal cycle corresponding to the generation of balanced creep and cyclic plasticity damage is shown in Figure 2b. The component is subjected to this cycle repeatedly, together with a constant appliedmechanical stress of 25MPa. The results of the analysis for these loading conditions are discussed in this section. Fig. 8. Variation of effective stress over a stabilised cycle for the three locations (1) near the cooling duct, (2) near the mid-point of the quadrant, and (3) near the heating region. The results are shown for the balanced creep-cyclic plasticity damage cycle with an average applied stress of 25MPa at the minimum section Figure 8 shows the variation of effective stress with time for a stabilised cycle at points (1) near to the cooling duct (location A in Fig.1), (2) at the centre of the quadrant and (3) near the heating radius (location B in Fig.1). The highest stress occurs at the point at which the temperature difference Modelling creep, ratchetting and failure... 683 between the cooling duct and heating radius is maximum, and occurs at the cooling duct. The effective stress at the quadrant mid-point remains fairly constant through the cycle at approximately 25MPa. Fig. 9. Predicted and experimental variation of component loading direction displacement with cycles. The results are shown for balanced creep – cyclic plasticity-damage dominated cycle Figure 9 shows the predicted and experimentally determined component displacement with cycles. Failure is predicted to occur at 134 cycles and is found experimentally to occur at 240 cycles. For the reasons given earlier, the model does not predict the primary ratchetting which can be seen to lead to large accumulated strains in the earlier cycles. However, the predicted steady state ratchet rates can be seen to be in reasonable agreement with the experimental data. The variations of predicted values of loading direction strain with cycles at both the cooling duct and the heating radius are shown in Figure 10. Again, the ratchet strains are largest at the cooling duct, and showa progressively increasing ratchet rate for themajority of the component life. The levels of creep and cyclic plasticity damage at failure, (1) and (2), and at a life fraction of 0.92, (3) and (4), are shown in Figure 11a for the balanced creep-cyclic plasticity damage cycle. The levels of creep damage can be seen to be considerably higher than the cyclic plasticity damage and arises predominantly in the quadrantmid-region. Failure is again predicted to occur by the propagation of damage through the componentminimum section. This is also observed to occur experimentally (Hayhurst et al., 2005), as shown in Figure 11b where surface damage in the heating radius can also be seen as predicted by the model. 684 F.P.E. Dunne et al. Fig. 10. Variation of total strain with number of cycle at locations A and B specified in Figure 3. The results are shown for the cyclic plasticity -dominated cycle Fig. 11. (a) Predicted spatial variations of creep (1) and cyclic plasticity (2) damage at component failure and at a life fraction of 0.92, (3) and (4) respectively, shown for the balanced creep – cyclic plasticity damage-dominated cycle, and (b) failed slag tap tested under an applied averagemechanical stress of 25MPa for the balanced cyclic plasticity-dominated damage cycle showing surface damage at the external heating radius, predicted by the model 7. Conclusions A continuum damage-based finite element model has been presented for the analysis of components suffering creep, cyclic plasticity and ratchetting due to combined cyclic thermal andmechanical loading. Themodel employs cycle jumping techniques to reduce computer CPU times. Modelling creep, ratchetting and failure... 685 Experiments have been described in which copper components have been subjected to combined cyclic thermal loading togetherwith constantmechani- cal load. The resulting component ratchetting behaviour was quantified, and the components tested through to failure. The ratchetting, failuremode and lifetimes of the componentswere predic- ted using the finite element model developed and the results compared with the experiments. Although the primary ratchetting is not predicted due to the assumption in the model of an instantaneously fully cyclically hardened state of the material, steady state ratchet rates were found to be well predicted, and lifetimes for both cyclic plasticity and balanced creep and cyclic plasti- city conditions were reasonably well predicted. The mode of failure for both loading conditions was correctly predicted. A. Damage constitutive equations ε̇p = 3 2 { 1 K [J(σ−x) 1−D −k ]}n σ′−x′ J(σ−x) (A.1) ẋi = 2 3 Ciε̇p(1−D)−γixiṗ+ C′i Ci xiθ̇ (A.2) ẋ = n ∑ i=1 ẋi (where n =2) (A.3) ω̇ =A [ασI +(1−α)σe] v (1−Dc)φ (A.4) dψ dN = [1− (1−Dp) β+1]p [ AII M(1−Dp) ]β (A.5) Dc = ω+α1z(ω)ψ (A.6) Dp = ψ+α2z(ω)ω (A.7) D = ω+ψ (A.8) z(ω) = 1 2 + 1 π tan−1µ(ω−ωI) (A.9) σ =(1−D)E(ε−εp−εθ) (A.10) inwhich K, n, Ci, γi, A, α, β, v, φ, p, α1, α2, ω1 and µ have been determined for copper (Dunne and Hayhurst, 1992a; Lin et al., 1996). 686 F.P.E. Dunne et al. References 1. BenallalA.,BenCheikhA., 1987,Constitutive equations for anisothermal elasto-viscoplasticity, In: Constitutive Laws for Engineering Materials; Theory and Applications, C. Desai, E. Krempl (edit.), 607-674 2. BenallalA.,MarquisD., 1987,Constitutive equations fornon-proportional cyclic elasto-viscoplasticity, J. Eng. Mater. Technol., 109 3. BlackmanD.R., SocieD.F.,LeckieF.A., 1983,Applications of continuum damage concepts to creep fatigue interactions, ASME Symp. on Thermal and Environmental Effects on Fatigue, Portland, 45-47 4. Bodner S.R., Partoum Y., 1975, Constitutive equations for elasto- viscoplastic strain hardeningmaterials, J. Appl. Mech., 97, 385-489 5. Chaboche J.L., 1987, Cyclic plasticity and ratchetting effects, In” 2nd Inst. Conf. on Constitutive Laws for Engineering Materials: Theory and Applica- tions, University of Arizona, Tucson, USA 6. ChabocheJ.L., 1988,Continuumdamagemechanics.Part II.Damagegrowth, crack initiation and crack growth, J. Appl. Mech., 55, 65-72 7. Chaboche J.L., Rousellier G., 1983, On the plastic and viscoplastic con- stitutive equations. Part 1. Rules developed with internal variable concept, J. Press. Vessel Technol., 105, 153-158 8. Dunne F.P.E., Hayhurst D.R., 1992a, Continuum damage based constitu- tive equations for copper under high temperatures creep and cyclic plasticity, Proc. R. Soc. Lond.,A437, 454-556 9. Dunne F.P.E., Hayhurst D.R., 1992b, Modelling of combined high tem- perature creep and cyclic plasticity in components using continuum damage mechanics,Proc. R. Soc. Lond.,A437, 567-589 10. Dunne F.P.E., HayhurstD.R., 1994, Efficient cycle jumping techniques for themodelling of materials and structures under cyclic mechanical and thermal loading,Eur. J. Mech., 13, 5, 639-660 11. DunneF.P.E.,Makin J., HayhurstD.R., 1992,Automated procedures for the determination of high temperature viscoplastic damage constitutive equ- ations,Proc. R. Soc. Lond.,A437, 527-544 12. Dunne F.P.E., Puttergill D.B., Hayhurst D.R.,Mabbutt Q.J., 1993, Experimental investigation of cyclic plasticity continuum damage evolution in an engineering components subjected to thermal loading, J. of Strain Anal., 28, 4, 263-272 13. GreenwoodG., 1973,Creep life andductility.Microstructures and the design of alloys, Int. Conf. on Metals, 2, 91-98 Modelling creep, ratchetting and failure... 687 14. Hall F.R., Hayhurst D.R., 1991, Continuum damagemechanics modelling of high temperature deformation and failure in a pipe weldment,Proc. R. Soc. Lond., A433, 383-403 15. Hall F.R., Hayhurst D.R., Brown P.R., 1996, Prediction of plane strain creep-crack growthusing continuumdamagemechanics, Int. J. DamageMech., 5, 353-383 16. Hayhurst D.R., 1972, Creep rupture under multi-axial states of stress, J. Mech. Phys. Solids., 20, 381-390 17. Hayhurst D.R., 1973, Stress redistribution and rupture due to creep in a uniformly stretched thin plate containing a circular hole, J. Appl. Mech., 1, 244-250 18. Hayhurst D.R., Brown P.R.,Morrison C.J., 1984a, The role of continu- um damage in creep crack growth,Phil. Trans. R. Soc. Lond.,A311, 131-158 19. Hayhurst D.R., Dimmer P.R., Morrison C.J., 1984b, Development of continuum damage in the creep rupture of notched bars, Phil. Trans. R. Soc. Lond., A311, 103-129 20. Hayhurst D.R., Hayhurst R.J., Vakili-Tahami F., 2005a, CMD predic- tions of creep damage initiation and growth in ferritic steel weldments in ame- dium bore branched pipe under constant pressure at 590◦C using a 5-material weld model, Proc. R. Soc. Lond., A461, 2303-2326 21. Hayhurst D.R., Makin J., Wong M.T., Xu Q., 2005b,Modelling of com- bined creep and cyclic plasticity in amodel component undergoing ratchetting using continuumdamagemechanics,Philosophical Magazine, 85, 16, 1701-1728 22. Inoue T., Igari T., Yoshida F., Suzuki A., Murkami S., 1985, Inelastic behaviour of 2.25Cr-1Mo steel under plasticity-creep interaction conditions, Nucl. Eng. Des., 90, 287 23. Kujowski P., Mróz Z., 1980, A viscoplastic material model and its applica- tions to cyclic loading,Acta Mechanica, 36, 213-130 24. Johanssan G., Ekh M., Runessan K., 2005, Computational modelling of inelastic large ratchetting strains, Int. J. Plasticity, 21, 5, 955-980 25. Lemaitre P.M., Chaboche J.L., 1990,Mechanics of Solid Materials, Cam- bridge University Press 26. Lemaitre P.M., Plumtree A., 1979, Applications of damage concepts to predict creep-fatigue failure, J. Eng. Mater. Technol., 101, 284-292 27. LesneP.M.,Chaboche J.L., 1984,Predictions of crack initiationunder ther- mal fatigue and creep, In: 2nd Int. Conf. on Fatigue and Creep, Fatigue 84, Birmingham, UK 688 F.P.E. Dunne et al. 28. Lin J., Dunne F.P.E.,HayhurstD.R., 1996,Physically-based temperature dependence of elasto-viscoplastic constitutive equations for copper between 20 and 500◦C, Philos. Mag., 74, 2, 359-382 29. Lin J., Dunne F.P.E., Hayhurst D.R., 1997, Approximate method for the analysisof componentsundergoingratchettingand failure,J. of StrainAnalysis, 33, 1, 55-65 30. Lin J., Dunne F.P.E., Hayhurst D.R., 1999, Aspects of testpiece design responsible for errors in cyclic plasticity experiments, Int. J. of Damage Me- chanics, 8, 2, 109-137 31. Liu M.C.M., Krempl E., 1979,A uni-axial viscoplasticmodel based on total stress and overstress, J. Mech. Phys. Solids, 7, 377-391 32. ManonukulA., Dunne F.P.E., Knowles D.,Williams S., 2005,Multia- xial creep and cyclic plasticity in nickel-base alloy C263, Int. J. of Plasticity, 21, 1-20 33. Miller A., 1979, An inelastic constitutive model for monotonic cyclic and creep deformation. Part 1. Equations development and analytical procedures, J. Eng. Mater. Technol., 98, 97-105 34. Remy L., Skelton P., 1990, Damage assessment of component experiencing thermal transients, In:Proc. Int. Conf. onHigh Temp. StructuralDesign, CEC, Venice. Bury St Edmunds: MEP 35. Wang Z.P., Hayhurst D.R., 1994, The use of super-computer modelling of high temperature failure in pipe weldments to optimise weld and heat affected zonematerial property selection,Proc. R. Soc. Lond., A446, 127-148 Modelowanie pełzania, ratchetingu1 oraz zniszczenia elementów konstrukcyjnych poddawanych obciążeniom termo-mechanicznym Streszczenie W pracy przedstawiono doświadczenia, w którychmiedziane elementy poddawa- no złożonym obciążeniom zawierającym cykle termiczne i obciążenia stałe. Zasto- sowano dwa cykle termiczne prowadzące do uszkodzenia z przewagą cyklicznej pla- styczności i zrównoważonego pełzania oraz cykle obciążeńwywołujących uszkodzenie typowe dla cyklicznej plastyczności. Obciążenia złożone wywoływały ratcheting ba- danych elementów, a w efekcie końcowym zniszczenie. Techniki metody elementów 1W literaturze polskiej stosuje się ten termin bez tłumaczenia, gdyż brak jest jednegookreślenianaprzyrostowenarastanieodkształceńpodwpływemprzyłożonego obciążenia Modelling creep, ratchetting and failure... 689 skończonych kontynualnejmechaniki uszkodzeń zostały rozwinięte dla cyklicznej pla- styczności w złożonym stanie naprężenia, pełzania oraz ratchetingu w elementach poddawanych obciążeniom termo-mechanicznym. W sformułowaniu metody elemen- tów skończonych, aby zminimalizować czasy centralnegoprocesora (CPU)komputera, zastosowano techniki skokówcyklicznych.Metody elementów skończonych zastosowa- no w przewidywaniu zachowania się elementów miedzianych wcześniej badanych do- świadczalnie, a otrzymanewyniki porównano.Zaproponowanemodelewsposób satys- fakcjonujący pozwalają przewidywać prędkości stanu ustalonego ratchetingu. Także przewidywania dotyczące sposobów zniszczenia oraz żywotności badanych elementów są zadowalające.Wyniki doświadczalnepokazują istotnywpływ izotropowegowzmoc- nienia cyklicznego na początkowe składowe prędkości ratchetingu. Manuscript received August 22, 2005; accepted for print March 15, 2006