On the accuracy of creep-damage predictions in thinwalled structures using the finite element method

Abstract

The constitutive model with a single damage parameter describing creep-damage behaviour of metals with respect to the different sensitivity of the damage process due to tension and compression is incorporated into the ANSYS ®nite element code by modifying the user de®ned creep material subroutine. The procedure is veri®ed by comparison with solutions for beams and rectangular plates in bending based on the Ritz method. Various numerical tests show the sensitivity of long-term predictions to the mesh sizes and element types available for the creep analysis of thinwalled structures. IntroductionEngineering structures operating at elevated temperatures such as fossil power plants, chemical plants, reactors, etc. are designed with respect to increased requirements of safety and assurance of long-term reliability. One of the main factors which must be considered in the long-term structural analysis is time-dependent material behaviour coupled with damage evolution (Roche et al., 1992). A powerful tool for the lifetime prediction is the continuum damage mechanics approach which is based on the formulation of constitutive and evolution equations for inelastic strains and material damage. Incorporating these material models into the ®nite element code predictions of time-dependent stress, strain and damage ®elds can be performed by numerical solution of nonlinear initialboundary value problems (e.g. . The ®rst problem arising by creep damage analysis is the formulation of a phenomenological material model that is able to describe the sensitivity of creep strain and damage rate to the stress level, stress state, temperature level, environmental effects, etc. Such a model must be able to extrapolate the experimental creep data usually available from uniaxial short-term creep tests and realised for narrow stress ranges to the in-service loading conditions in the real structure. The second problem can be related to the quality of the ®nite element predictions particularly by analysis of structures with complex shapes.The structural analysis of thinwalled components (pressure vessels, pipes, pipe bends, etc.) can be performed using the mechanical models of plane stress (strain) states or equations of shell theory. In the case of plane stress (strain) problems numerous ®nite element simulations considering creep damage effects has been made because of experimental data available for veri®ca-tion. Examples are discussed by Saanouni et al. Since these examples con®rm the ability of ®nite element simulations to predict stress redistributions and failure times with accuracy enough for engineering applications and can be used as benchmark tests by development of user de®ned material subroutines incorporating damage evolution a little effort has been made for the analysis of transversely loaded thinwalled structures. Numerical results for rupture times of rectangular plates in bending are given by Bodnar and Chrzanowski (1994), Bialkiewicz and Mika (1995) based on in-house ®nite element codes and Altenbach and Naumenko (1997) for re...

On the use of solid- and shell-type finite elements in creep-damage predictions of thinwalled structures

“…The stress-state dependence of the damage evolution is considered by setting a 1 in Eqs. (1). According to the results of simulations, the failure will occur on the outer surface along the outer bend radius, Fig.…”

“…The analysis has been performed using the FE code ANSYS after incorporating the material model (1) with the help of a user-de®ned creep material subroutine. In [1], we discussed various examples for beams and plates in bending which verify the FE calculations with the same code and creep-damage model. For details of the timeintegration and equilibrium-iteration methods used for creep calculations we refer to [18,29].…”

“…The sensitivity of the solution to the mesh size has been studied by comparison of the steady-state creep solutions at different meshes. In [1], we demonstrated on various benchmark tests that the mesh justi®ed with the help of the steadystate creep solution is ®ne enough for the creep-damage solution up to failure initiation (critical damage state). For the quarter of the pipe bend, 30 Â 24 elements were used in the case of the shell model, and in the case of the solid model with four elements in the thickness direction there were 30 Â 24 Â 4 elements.…”

“…First, we exclude the stress-state dependence of the damage evolution by setting a 0 in the material model (1). This simpli®cation leads to the r vM -controlled damage.…”

Analysis of Inelastic Behavior for High Temperature Materials and Structures

“…This result agrees well with the data presented in May et al (1994), where the creep failure has been detected at the same position. The processes of time dependent deformation, stress relaxation and redistribution have been illustrated in the literature based on different examples for beams, plates and shells Altenbach et al 2000bAltenbach et al , 2001Altenbach et al 2004). One feature of the example considered here is that the final creep stage is not only the result of the local material deterioration but is additionally governed by the flattening (ovalisation) of the pipe bend cross section.…”