PSI - Issue 45

Mark Mogeke et al. / Procedia Structural Integrity 45 (2023) 36–43 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

41

6

( ) = 1 5 6 2 4 5 [− 5 4 ( ) 4 ]

(3)

4. Results and Discussion 4.1. Comparison of Wave Data

To determine the amount of uncertainty in the wave environment results from hindcast simulations, a comparison was carried out against the wave buoy data (source of truth for ocean wave environment); the precision of the wave buoy is ± 0.2 m. Table 2 compares the metrics Bias (b), Root-Mean-Square Error (RMSE), and Scatter Index (SI) for the modelled H s and T p encountered during sea trials, relative to the wave buoy measurements. Table 2. Comparison of b, RMSE, and SI for modelled H s and T p encountered during sea trials, relative to wave buoy measurements H s, CAWCR H s, CMEMS H s, NOAA T P, CAWCR T P, CMEMS T P, NOAA b -0.34 -0.26 -0.34 7.2 s 0.84 s 0.18 s RMSE 0.41 m 0.37 m 0.41 m 10.3 s 1.4 s 0.75 s SI 19 % 22 % 19 % 147 % 21 % 15 % In general, the three hindcast sources underestimate the significant wave height and overestimate the peak period. The CMEMS hindcast provides the closest agreement to the wave buoy H S . However, although it has the coarsest spatial-temporal resolution, the NOAA hindcast provides the closest agreement to the wave buoy T p . This variability between the hindcast data emphasizes the importance of checking the validity of the data to ensure its suitability for structural response prediction. 4.2. Comparison of stress response Data from 15 trials runs (N = 15), at 15 kn, 20 kn, and 25 kn, are analysed. The measured and computed σ RMS values were calculated by integration of the stress response for three strain gauge locations, SG-2, SG-1, and SG-3 shown in Fig. 3 with linear curves of best fit and a line of exact agreement. The results indicate that, in general, accounting for wave spreading provides slightly more accurate results. Use of CMEMS (Meteo France Wave Model) and NOAA (WaveWatch III) hindcasts yields similar metrics, which are generally smaller than those based on the wave buoy data. The σ RMS values based on the CAWCR dataset are in poor agreement with the measured σ RMS values, which reflects the large scatter index of the peak period shown in Table 2. Across the wave data sources, the models better predict the measured σ RMS for the heads seas and bow quartering runs than in following seas and stern quartering seas. The structural model generally performed better with inputs from the CMEMS and NOAA hindcasts than with inputs from the wave buoy data despite the latter being the ‘ground truth’. The reason for this is not clear, and the number of observations was limited. However, possible explanations include the effects of measurement uncertainty, interdependencies between the wave parameters (height, period, and spectral shape), and structural model bias that is ‘offset’ by the negative bias in the modelled wave height (Table 2). More data over a larger range of conditions is required to provide more definitive conclusions. 4.3. Sensitivity of stress response to wave parameters A sensitivity analysis was carried out to determine the relationship between stress response and the wave parameters principal wave direction (χ), significant wave height (H s ), wave period (T z ), spreading angle (θ), spreading exponent (n), and σ RMS at SG-2 (centre of flange on keel) using the Kendal correlation coefficient τ. The results indicate that σ RMS is most strongly associated with H s and particularly at strain gauge location SG-2, on the centre keel, the association between T z and σ RMS is stronger than that between χ and σ RMS . Interestingly, the association between σ RMS and the spreading exponent is not statistically significant. This result indicates that it is acceptable to