PSI - Issue 45

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

30

3

Classification Societies (International Association of Classification Societies 2010) use ‘damage experience’, or the number and location of cracks related to the fleet, as the main source of information for maintenance planning. Similarly, the presence or absence of a crack at a construction detail can be interpreted as a sample of the corresponding fatigue life (Groden and Collette 2017). However, use of in-service load and response data combined with survey reports to update service life predictions has been limited (Hifi and Barltrop 2015). Fricke et al. (2002) compared fatigue analysis approaches applied to a pad detail on a containership. Though this study highlighted the variability in the results, a weakness was that the design life rather than the actual fatigue life was used as the baseline. This paper proposes a refinement of the nominal stress approach for joints typical of aluminium welded ship details. The refinement process leverages maintenance records and stress spectra derived from long-term strain measurements acquired from a hull monitoring system onboard a 56 m naval ship, of marine-grade aluminium alloy construction. The sensitivity of the predicted fatigue life of welded details to the choice of S-N curve, available from a structural design code, and stress parameter extraction is investigated. 2. Method and Materials Welded details of interest on a naval High Speed Light Craft, shown in Table 1, are studied. The ship was constructed from aluminium alloy 5083-H321/H116 for plating, and 6082-T5/T6 for rolled sections. 2.1. In-Service Information Structural fatigue failures were observed at common locations across the fleet. These observations provide an opportunity to combine different sources of data and analysis approaches (Magoga et al. 2019). The maintenance records of the fleet were reviewed to obtain the time from commissioning to first discovery of a crack (FL m ). Table 1 presents the fleet-wide average and standard deviation of FL m for each detail of interest, normalised by that of ID-1. Cracking was first discovered at ID-2 followed by ID-1, ID-4, and ID-5. At the time of the analysis, no defects had been observed at ID-3. The standard deviation ranges between 9.3% and 31%. This variation is considered reasonable, as maintenance data often suffers from issues with accuracy and completeness (Hodkiewicz and Tien-Wei Ho 2016). In addition, the variation of FL m can be attributed to difficulties with accessing some parts of the structure, inconsistent defect reporting, and operational variability across the class. 2.2. Finite Element Analysis The commercially available FEA package MAESTRO (DRS Defense Solutions 2013) is used to calculate the stresses in the ship structure. The global model of the ship is relatively coarse, though represents the overall stiffness adequately. Fine mesh models of the details of interest are embedded into the global model. The mesh density is approximately one thickness (t) of the material, which provides acceptable stress resolution and is commonly adopted for the evaluation of stresses at welded joints (Hobbacher 2008). Linear elasticity is assumed. 2.3. Fatigue Life Analysis The nominal stress approach to calculate the fatigue life of a welded structural detail is the most commonly practiced in the maritime industry (Collette and Incecik 2006, Horn et al. 2009, Soliman et al. 2015), and is employed in the present study. This method uses the stress acting on the location to be assessed, neglecting the stress concentration arising from both the structural configuration and the weld. These effects are inherently taken into account in the S-N curves. Eurocode 9 (Technical Committee CEN/TC 250 1999) provides S – N curves relating nominal applied cyclic stress ranges S to the corresponding number of cycles to failure N. The curves are based on estimates of the mean and standard deviation, assuming a normal distribution, of observed logarithmic cycles for given logarithmic stress values. These statistics are used to obtain a characteristic regression line for a probability of survival of approximately 97.7% from the mean. This is less than 80% of the corresponding mean strength value, and allows for wider variations in