Comparação de métodos espectrais para análise de fadiga

Abstract

Fatigue plays a fundamental role in the design of mechanical components. While fatigue analysis is well-understood for constant amplitude loadings, the same does not hold true for random loadings. The main objective of this work is to analyze² and compare different spectral methods for fatigue simulation under random loading, both among themselves and with the results obtained in the time domain. Additionally, the study seeks to verify the influence of the exponent k in Basquin's law, as well as the bandwidth factor on the performance of spectral methods. To achieve these objectives, stress Power Spectral Densities (PSDs) were numerically generated, serving as inputs for spectral fatigue calculations. Approximately 20 methods were employed in this study, involving unimodal and bimodal PSDs. In summary, the results show that the spectral approach, although complex, offers significant potential in terms of time and resource efficiency when compared to traditional methods. Certain spectral methods, such as the methods of Dirlik, Bands, and Huang-Moan, provide good estimates regardless of the case studied. However, the same is not true for other methods, like the Narrow-Band approximation and Lalanne's method, which yield good results only for narrow-band signals. Furthermore, the findings indicate that the error tends to be more conservative as the bandwidth increases, while the absolute error is smaller for lower values of the exponent k.