Direct cyclic methods for structures under thermomechanical loading
Résumé
A great number of structures are subjected to thermomechanical cyclic loading. For instance, it is the case of automotive engine: components such as cylinder heads or exhaust manifolds experience cyclic thermal loading. It is also the case of rails: in curve, gauge corners are subjected to high rolling contact stresses inducing cyclic plasticity or even ratchetting. The estimation of the lifetime of such structures under cyclic thermomechanical loading requires, on the one hand, the determination of the response until an asymptotic state is reached, and on the other hand, the use of an appropriate damage criterion. The computation of the whole loading history leads to lenghty and expensive incremental calculations, especially for structures with large number of degrees of freedom. Therefore, it is very useful to develop computational approaches for straightforward calculations of the possible stabilized state under repeated thermomechanical loading. Direct Cyclic Methods offer this alternative. The paper is devoted to Direct Cyclic Methods we propose. These numerical strategies, derived from the Large Time Increment Method are alternative to classical incremental methods or cycle jump techniques. Direct Cyclic Methods permit the direct determination of the asymptotic response of inelastic structures subjected to cyclic thermomechanical loadings.The different versions of the Direct Cyclic Method are presented. They differ from the solution of the global stage which is consuming the most CPU time. The first version of the Direct Cyclic Method uses a classical finite element method to compute the SA and KA response and has been used to studying cyclic response of structures subjected to small oscillatory contact displacements for fretting fatigue applications. The second version uses Fourier expansion in solving the global equations. It is this version which is in Abaqus Sofware. The third version adopts a decomposition in a wavelet basis and the SKA solutions are estimated using only the largest coefficients.