New hybrid cycle jump approach for predicting low-cycle fatigue behavior by a micromechanical model with the damage induced anisotropy concept

Abstract

In this work, a new hybrid cycle jump approach was proposed to predict the low-cycle fatigue (LCF) response using a well-established micromechanical model. The damage activation/deactivation phenomenon was formulated at the macroscale. Using the concept of rate independent plasticity with the small strains assumption, the damage deactivation effect on the LCF response and its lifetime was accurately described by the model under simple and complex cyclic loading paths. Three algorithms namely, Taylor 1st and 2nd order and predictor-corrector were utilized for estimating the model response after each cycle jump. An optimum precision factor (ρ) value was defined and then tested via a suitable compromise between computation times and accuracy. In order to numerically determine the cyclic jump length (ΔN), a new hybrid approach was developed based on the overall von Mises stress and the overall damage together with monitoring the accumulated slip. Then, a comparative study revealed that the predictor-corrector algorithm was opted due to its robustness compared to the other algorithms. The numerical study was carried out using two different cyclic loading configurations. One was under simple tension-compression and the other under complex tension-torsion with 90° out-of-phase angle. The model response was described under two distinct states: (i) cycle-by-cycle and (ii) cycle jump concept. For each numerical state, the overall and local responses were recorded. With the new hybrid cycle jump approach, the model response highlights that the greater, the number of cycles, the greater, the gain in computation time but the less, the accuracy. For example, in TC for Nf = 10,000 cycles, a recorded gain in computation time was attained 77% (against 95% by means of cycles number) with a relative error of 4%; whereas in TT90 of Nf = 2124, the computation gain became 43.4% (and 73.4% with respect to number of jumped cycles) with a relative error of 3%.