PSI - Issue 45

Teresa Magoga et al. / Procedia Structural Integrity 45 (2023) 28–35 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

34

7

The results in Fig. 3 suggest that when d/t approximates one, there is reasonable prediction of the fatigue life relative to the historical data when the Eurocode 9 detail judged most applicable is used. Based on this observation, FL p /FL m for each detail of interest is presented in Table 2. Also presented are the values of FL p normalised by FL m at detail of interest of ID-1.

Table 2: K, and FL p /FL M at d/t=1, and FL m /FL M for each detail of interest ID K Eurocode 9 Detail FL p /FL m at d/t = 1

FL p /FL m at ID-1

1 2 3 4 5

3.3 5.0 1.7 3.5 2.7

7.4.3 7.2.1 11.3

0.9 1.0 0.9 1.1

0.9 0.7 1.1 2.0 13.0

4.0* (*FL p /FL d )

7.6

11.3

4. Discussion It is proposed that the nominal stress parameter can be extracted by averaging the normal stresses recovered at a distance of 1t from the weld/intersection, substantiated by: Good agreement between the predicted fatigue life values and those derived from the maintenance records - ranging between 90% and 110% - in Table 2, and correlation between the order of predicted and in-service fatigue failures (the order of FL p /FL m at ID-1 in Table 2 matches the order of FL m /FL m at ID-1 in Table 1). The results in Fig. 3 and 4 demonstrate the variability in fatigue life estimates with selection of Eurocode 9 S-N curve, which was also shown by Aksu et al. (2015). Whilst the proposed application of the nominal stress approach is reliant on the appropriate selection of the Eurocode 9 detail, the requirement for expert judgment in the design and through-life management of structures is not new. Maljaars and Vrouwenvelder (2014) noted that despite reasonable accuracy of the load measurements, expert judgement is often required to account for uncertainty in the applied loads in the structural reliability assessment of bridge structures . Similarly, sound judgement based on the analyst’s practical experience is needed to define the long-term seaway loads that sufficiently represent the fatigue demand that the ship structure will experience, and to design certain welded details (Horn et al. 2009, Wang 2010). The reliance on the appropriate selection of the S-N curve is need not be a limitation. Over a ship’s life -cycle its structural configuration, and analytical models, can be expected to change. These changes lead to updating of the variables incorporated into the structural predictions (Hifi and Barltrop 2015, Magoga et al. 2019). The fatigue life values of the details of interest inferred from in-service information (presented in Table 1) should be understood to be optimistic as it relies on ‘ average behaviour ’ and a limited number of samples. A crack most likely propagated to a critical length (requiring rectification) some time before being reported. At the same time, use of the Eurocode 9 S-N curves is conservative because they correspond to the mean life curve minus two standard deviation from the experimental data. Therefore, it is recommended that the analyst be aware of the different sources of (non)conservatism in the parameters required as input to a fatigue analysis. To build confidence in the proposed modified stress approach, Monte Carlo concepts for statistical estimations may be useful. In addition, there is merit in investigating Bayesian methods that address the small number of cracking reports (Beer et al. 2013, McNeish 2016). The presented work contributes to the ship structures community as it expands on research to evaluate the fatigue life of one detail on an aluminium vessel (Soliman et al. 2015), in-service data (strain monitoring and defect reports) is utilised, and a modified nominal stress extraction process has been detailed to extend the guidance provided by Eurocode 9. Whilst use of techniques such as fracture mechanics could be pursued, significant investment in obtaining the necessary input information (crack growth rate data, initial flaw size) is needed. 4. Conclusion Both naval ship designers and in-service managers require reliable fatigue evaluation approaches that are sufficiently accurate but also time-efficient. The selection of an applicable detail or S-N curve from a fatigue design code can be challenging, particularly when both the geometry and stress field are complex. Also, the determination of the stress parameter is not always clear. Therefore, this paper proposed a refinement of the nominal stress approach for joints typical of aluminium welded ship details, which extends upon the guidance provided by Eurocode 9. The