FATIGUE LIFE ESTIMATION USING FREQUENCY DOMAIN TECHNIQUE AND PROBABILISTIC LINEAR CUMULATIVE DAMAGE MODEL

Resumo

Engineering structures are designed to withstand a variety of in service loading specific to
their intended application. Random vibration excitation is observed in most of the structural
component applications in the naval, aerospace and automotive industry. Likewise, fatigue life
estimation for such components is fundamental to verify the design robustness assuring structural
integrity throughout service. The linear damage accumulation model (Palmgren-Miner rule) is still

largely used for damage assessment on fatigue estimations, even though, its limitations are well-
known. The fact that fatigue behavior of materials exposed to cyclic loading is a random phenomenon

at any scale of description, at a specimen scale, for example, fatigue initiation sites, inclusions, defects
and trans-granular crack propagation are hardly predicted, indicates that a probabilistic
characterization of the material behavior is needed. In this work, the inherent uncertainties of the
fatigue life and fatigue strength of the material are characterized using the random fatigue limit (RFL)
statistic method, which incorporates the maximum likelihood estimation to produce probabilistic S-N
curves. Furthermore, a frequency domain technique is used to determine the response power spectrum
density (PSD) function of a structural component subjected to a random vibration profile excitation.
The fatigue life of the component is then estimated through a probabilistic linear damage cumulative
model, where not only the time to failure is predicted but also its variability. The methodology was
applied to a Titanium alloy structural component, exposed to a specific random excitation, where the
predicted life using the material percentile curves 5% and 95% has shown a significant variability
when compared to the common used percentile 50%. Thus, this characterization might be relevant for
the definition of material design curves.