SE9700039 IMPLEMENTATION OF CONSTITUTIVE EQUATIONS FOR CREEP DAMAGE MECHANICS INTO THE ABAQUS FINITE ELEMENT CODE - SOME PRACTICAL CASES IN HIGH TEMPERATURE COMPONENT DESIGN AND LIFE ASSESSMENT P. Segle, L. A. Samuelson, P. Andersson and F. Moberg SAQ/FoU-Report 96/03 JAQ.KONTROLLAB Box 49306 • S-100 29 Stockholm • Sweden Telefon +46 8 617 40 00 Telefax+46 8 651 70 43 Pre-print of paper to be presented at SYMPOSIUM ON "Inelasticity and Damage in Solids Subject to Microstructural Change" - in honour of the late Professor L.M. Kachanov - September 25-27,1996, St. John's, Newfoundland, Canada Implementation of constitutive equations for creep damage mechanics into the ABAQUS finite element code - Some practical cases in high temperature component design and life assessment Peter Segle1, Lars A Samuelson1, Peder Andersson1 and Fredrik Moberg2 'SAQ Kontroll AB, Box 49306, SE-100 29 STOCKHOLM, Sweden 2Semcon Engineering AB, Box 11T7, SE-131 27 NACKA STRAND, Sweden Abstract. Constitutive equations for creep damage mechanics are implemented into the finite element program ABAQUS using a user supplied subroutine, UMAT. A modified Kachanov-Rabotnov constitutive equation which accounts for inhomogeneity in creep damage is used. With a user defined material a number of bench mark tests are analysed for verification. In the cases where analytical solutions exist, the numerical results agree very well. In other cases, the creep damage evolution response appears to be realistic in comparison with laboratory creep tests. The appropriateness of using the creep damage mechanics concept in design and life assessment of high temperature components is demonstrated. 1. Introduction In the late fifties, Kachanov[l] introduced the continuum damage mechanics concept. Since then the area has been further developed by researchers such as Rabotnov[2], Lemaitre and Chaboche[3] and Murakami[4] and today it is a well known discipline in the field of mechanics of solid materials. Continuum damage mechanics is a pheno- menological way of describing the damage evolution between the virgin state and macroscopic crack initiation. The mechanisms creep, fatigue and ductile fracture of homogeneous materials as well as damage of concrete and damage effects in composite materials can all be described by use of this concept. In the present paper, the appropriateness of using continuum damage mechanics in high temperature design and life assessment of a circumferential weldment is demonstrated. Constitutive equations, describing the deterioration mechanism of creep, are implemented into the ABAQUS finite element code and by studying three typical weldments, an improved understanding of how a weldment behaves when subjected to creep is achieved. With this knowledge, a more accurate prediction of creep damage evolution, advisable positions for non-destructive examination, time to rupture for the weldment, etc., is given. 2. Constitutive equations A modified Kachanov-Rabotnov constitutive equation which accounts for inhomogeneity in creep damage is used[5]. Neglecting plasticity and primary creep the total strain rate is (1) dt dt dt ' where de'f l + v At ~ rr I J^ I -i . ..I J- fti! » \2) (3) and dD A [ao>+(l-g)o-.]" ^T = g77T n n\* ' (4) at <p +1 (1 — 1)) In the equations given above e'", £y, e", <y and s are the total i} tj strain, elastic strain, creep strain, stress and stress deviator tensor, respectively, a, and a are the maximum principal stress and von t Mises stress, E and v the modulus of elasticity and Poisson's ratio, D and D the damage variable and critical damage where the material cHl creep life is assumed to be fully utilised when D/D , reaches the value cri one. a is the material constant relating to the multiaxial rupture criterion which ranges from zero to unity, B, n, A and v are the material constants relating to the minimum creep strain rate and rupture behaviour, g, <f> and p the constants accounting for the inhomogeneity of the damage where p represents the volumetric ratio of the damaged phase. 3. Implementation into ABAQUS In the present work the finite element program ABAQUS/Standard[6] is used to perform the creep damage analyses. A modified Kachanov- Rabotnov equation is added to the program library by use of the user subroutine UMAT[7]. The user subroutine is programmed in FORTRAN 77. The user subroutine is called for at each material integration point at every iteration of each increment. When it is called, it is provided with the material state, i.e. stress, solution dependent state variables, temperature etc., at the start of the increment and with the strain increment and the time increment. The subroutine updates the stresses to their values at the end of the increment and calculates the Jacobian matrix, \.^.^.a jdb.s . Since most constitutive models require the ij ij storage of solution dependent state variables, ABAQUS provides possibilities to allocate storage for any number of such variables for each integration point. The damage parameter, D, is treated in this manner and is updated in the user subroutine during each increment. Since the damage and stress increments cannot be expressed in closed form, the non-linear equations are solved numerically, using additional routines. Fig. 1 shows the communication between ABAQUS and the separate files. Input file 1 User subroutine input file Material state, strain inc., time inc. ABAQUS UMAT Additional subroutines Stress, Jacobian matrix, state variables Utility subroutines - Fig. 1. Communication between ABAQUS and the separate files. In the present work UMAT is formulated strictly for three dimensional continuum elements. The first step is a purely elastic step and the second step is the creep damage response step. An explicit time integration scheme is used with a central difference operator[8] according to C?) /+-A/ where / is an arbitrary function, /, its value at the beginning of the increment, Af the change of the function over the increment and A/ the time increment. Discretisation of the constitutive equations (1) and (4), results in a set of six coupled non-linear equations which are numerically solved by using a globally convergent Newton method[9]. With a user defined material a number of bench mark tests were analysed for verification. In the cases where analytical solutions exist, the numerical results agreed very well. In other cases, the creep damage evolution response seemed to be realistic in comparison with laboratory creep tests. 4. High temperature design and life assessment by use of the continuum damage mechanics concept In components where geometrical and/or material discontinuities are present, continuum damage mechanics simulations are particularly useful in order to understand the creep behaviour of the component. Using this concept, stress redistribution due to the damage evolution can be taken into account and a more profound understanding of how the component behaves when subjected to creep is achieved[10-14]. 4.1 Finite element modelling and simulation In the present investigation, the creep damage evolution in a circumferential V-shaped weldment in a piping system is investigated, see Fig. 2. The outer diameter and the wall thickness of the pipe are 500 and 40 mm, respectively. The welded pipe is subjected to an internal pressure resulting in a nominal hoop stress of 110 MPa and an axial stress of 50.2 MPa. Three different combinations of creep properties of the weldment constituents, i.e. parent metal (PM), weld metal (WM) and heat affected zone (HAZ), are studied by altering the characteristics of the weld metal. These three combinations represent three typical weld systems that can be found in power plants of today. 210 250 3.7 PM 3.7 symmetry line i 2.5 22.8 Fig. 2. Geometry of pipe with weldment. Table 1 shows the material parameters used for the three cases. The constants are based on creep tests of weldments carried out at the Swedish Institute for Metals Research[15]. In [16] it was suggested that a equals 0.43 for the ferritic steels 0.5Cr0.5Mo0.25V and 2.25CrlMo why the same value is used for the parent material in the present paper. For the weld and HAZ metals, no data for a were available and hence the same value as that of the parent material is used. Table 1. Constants in constitutive equations for PM, HAZ, matched WM, creep-soft WM and creep-hard WM, respectively. Constant PM HAZ WM WM WM Matched Creep-soft Creep-hard B 1.940e-15 1.540e-13 1.772e-14 8.860e-14 5.907e-15 n 4.354 3.925 3.870 3.870 3.870 A 8.325e-13 4.365e-12 3.800e-13 4.749e-13 3.304e-13 3.955 3.750 4.110 4.110 4.110 V g 0.961 0.955 0.965 0.965 0.965 1.423 2.017 0.6517 0.6517 0.6517 P 0.393 0.280 0.0985 0.0985 0.0985 a 0.43 0.43 0.43 0.43 0.43 E 160000 160000 160000 160000 160000 0.3 0.3 0.3 0.3 0.3 V The structure is modelled by use of the 20-node solid element C3D20R and axisymmetrical boundary conditions are utilised. 4.2 Results from damage simulation of a matched weldment Fig. 3 shows the hoop stress in the matched weldment at different times characterising the stress redistribution that takes place during the life of the weldment. Due to a higher minimum creep strain rate of the HAZ than that of the parent and weld metal, the HAZ region is off-loaded. This off-loading begins as soon as the creep process starts. In case the tertiary creep would be neglected, a steady state stress field would eventually be established. The first plot in Fig. 3 at 4920 hours, shows the hoop stress just before the tertiary effects start to influence the stresses in the weldment region, i.e. the stress field that comes closest to a steady state. The following stress plots show the hoop stress at times when the tertiary effects have a substantial influence. As seen at the end of the creep life of the weldment, the hoop stress in the HAZ starts to increase again. This is explained by the fact that the creep strain rate of the parent and weld metal has become larger than that of the HAZ due to higher creep damage in the PM and WM. In Fig. 4, the creep damage evolution is shown. As seen from the damage plots, the creep damage is least developed in the HAZ region even though the creep rupture strength of the HAZ is lower than that of the other two weldment constituents. The explanation to this is the off- load of the HAZ as mentioned earlier. The damage evolution in the parent material is somewhat similar to that of the weld metal as expected. Fully damaged material is first found in the weld metal close to the HAZ and final rupture occurs at 11020 hours. 4.3 Results from damage simulation of a creep-soft •weldment As for the matched weldment, redistribution of stresses takes place due to differences in creep strain rates between the weldment constituents. In this case, the material discontinuity between the HAZ and the weld metal is reduced compared to the former case. The first stress plot in Fig. 5 shows the 'steady state' stress field where the HAZ and weld metal are off-loaded due to their higher minimum creep strain rate. As S22 +G.OCE-KXL 11-+8.54E+O1 M +9.CSE-O1 +9.CE+01 +1.C0E+CC +1.O7E+C2 +1.12E+C2 +1.1EE-K)2 +1.23E+C2 +1.2SE+C2 +1.34E+C2 +1.4CE-HD2 Time=4920 (h) Time=9050 (h) m illli Illl 11111 SSSSSSSliii Time=9695 (h) Time= 10170 (h) Time= 10600 (h) Time=10880(h) Time=l 1000 (h) Time=l 1020 (h) Fig. 3. Hoop stress in the matched weldment (MPa). sa/2 -INFMn' +4.8CE-01 •6.2E-01 +6.1EE-01 +6.61E-01 +7.07E-01 +7.52E-O1 +7.9EE-01 +6.45E-O1 -H3.89E-O1 •6.34E-O1 -+9.8C£-O1 -+1.0CG+OD =4920 h) Time=9050 (h) v Time=9695(h) Time=10170 (h) Time=10600 (h) Time=10880(h) Time=l 1000 (h) Time=l 1020 (h) Fig. 4. Creep damage ratio, D/D , in the matched weldment. inl