This conference paper was submitted for presentation at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

Welding is widely used across manufacturing due to its efficient and cost-effective way of joining metals. However, a major issue with welded joints is their lower fatigue strength compared to the base material. This makes the weld seams the most vulnerable regions in a structure under fatigue loads. Therefore, fatigue life assessment should prioritize weld seams. Various methods, such as nominal stress, effective notch, and hot-spot approaches, have been developed for this purpose. Additionally, real structures often face random loading, like those in offshore platforms and railway equipment, which requires fatigue analysis in the frequency domain using statistical moments. This work presents a numerical methodology based on Finite Element Analysis. It evaluates fatigue life of welded components subjected to stationary random loading characterized by a Gaussian probability density function. The approach employs maximum absolute principal stress as the equivalent stress criterion, utilizes the hotspot method for stress calculation at the weld, and applies the Dirlik method to statistically estimate the number of stress cycles. The Dirlik technique is a tool that allows the analysis of broadband Gaussian random stresses based on their characteristics in the frequency domain. Fatigue life calculations are based on fatigue curves from international standards, such as Eurocode 3 and the International Institute of Welding (IIW), calibrated for the hotspot method. The methodology enables fatigue life assessment when the Power Spectral Density (PSD) of the input excitation is known, as often specified by customers in industrial applications. A comprehensive mathematical procedure for implementing this approach is provided. In addition, an experimental validation of the proposed methodology is included. Eight welded specimens of structural steel S355J2+N were subjected to random loading with a known PSD until failure. Four uniaxial strain gauges were positioned at specific points on each specimen to measure local strains and to compute their cycles distribution, showing consistency between the Rainflow and Dirlik methods estimations. The SN curves of the specimens with survival probabilities of 10%, 50%, and 90% were preliminarily measured as per ASTM E466, to properly calibrate the input during the random test and avoid excessive test duration. Finally, a Finite Element model of the tested specimens has been developed and the methodology described here has been applied. The measured experimental fatigue life of the specimens shows to be consistent with the Finite Element-based fatigue life predictions from the proposed methodology, which has demonstrated a conservative approach. For completeness, an analysis substituting the hotspot method with the effective notch method was conducted, revealing that the latter offers even greater accuracy in fatigue life estimation. In conclusion, the presented Finite Element methodology provides a reliable tool for assessing fatigue life in welded components under random loading, with practical applicability in industrial contexts.