A comparison of lifetime prediction methods for a thermal fatigue experiment

This paper is dedicated to the comparison of several numerical models for estimating the lifetime in a fatigue experiment. The models simulate the SPLASH experiment, which produces thermal fatigue by locally quenching stainless steel specimens. All models predict first a stabilized mechanical state (plastic shakedown) and then a lifetime prediction using several fatigue crack initiation criteria. The numerical methods are either completely nonlinear or combine approximate elastic solutions obtained from minimizing a potential energy or closed form solutions with a Neuber or Zarka technique to estimate directly the elastoplastic state. The fatigue criteria used are Manson, dissipated energy and dissipated energy combined with a hydrostatic pressure term. The latter had provided a best prediction over a series of anisothermal and isothermal LCF experiments in a classical fatigue analysis. The analysis shows that for fatigue criteria taking into account the triaxiality of the mechanical response we obtain a systematic and conservative error. As a consequence of this work, we show that simplified models can be used for lifetime prediction. Moreover the paper provides a general technique to asses from the point of view of the design engineer the combination between a numerical method and a fatigue criterion.