Estimation of fatigue damage under uniform-modulated non-stationary random loadings using evolutionary power spectral density decomposition

In the context of increasingly complex non-stationary random loadings, accurately estimating fatigue damage poses significant challenges. This study investigates uniform-modulated nonstationary random processes and introduces a modeling approach for truly non-stationary nonGaussian processes based on the modulation of stationary non-Gaussian processes. By integrating the existing evolutionary power spectral density (EPSD) decomposition method with the typical stationary Gaussian and non-Gaussian spectral methods, this study aims to enhance methodologies for effectively estimating fatigue damage for non-stationary processes, thereby addressing the limitations of traditional spectral methods in characterizing non-stationarity. Time-domain fatigue damage is evaluated by applying the rainflow counting method to the generated stress response time series to validate the accuracy of the EPSD-based spectral methods. Comparative analyses are conducted to evaluate the performance of non-stationary spectral methods under different bandwidth parameters and modulation functions. The findings indicate that the applied modeling approach captures the intricate coexistence of non-stationarity and non-Gaussianity, while the EPSD decomposition method yields reliable fatigue damage estimates. Notably, the skewness and kurtosis modulation coefficients emerge as critical indicators of the impact of nonstationarity on fatigue damage. The Gaussian and non-Gaussian Dirlik and Tovo-Benasciutti methods are identified as particularly suitable for integration with EPSD decomposition, offering robust accuracy across various bandwidth contexts.