Comparative Study of Fatigue Assessment Methods for Broadband Random Stress

Under combined wind-wave, wind-ice, and wave-whipping loads, the stress responses of ship and offshore structures often exhibit bimodal characteristics. Frequency-domain fatigue assessment methods, which directly evaluate damage using stress power spectral density functions, offer high computational efficiency. However, because the available methods differ in accuracy, a systematic comparison is needed to select the most suitable method for different operational conditions. In this study, based on time-domain simulation results from 400 bimodal rectangular spectra that span a wide range of Wirsching spectral width parameters from 0.0586 to 0.9922, the existing methods are categorized into four groups: narrow-band correction factor methods, rainflow cycle approximation methods, independent summation of combined fatigue damage methods, and nonlinear combination methods for decomposed broadband spectra. Sixteen typical methods are selected, and their prediction accuracies are quantitatively evaluated under three fatigue strength exponents (m = 3, 4 and 5), using the ratio of frequency-domain damage to time-domain damage as the evaluation metric. The results indicate that the Tovo–Benasciutti, Dirlik and Low methods all possess good engineering applicability. Among them, the Tovo–Benasciutti method provides the highest accuracy, the Dirlik method is more suitable for bilinear S–N curves, and the Low method is more robust in engineering applications.