Sloshing

Abstract

Offshore structures with partially filled storage tanks may experience sloshing of the cargo when exposed to waves. Inventive use on the topside result in storage tanks which are built as an integrated part of the deck structure. Weight control and available space is often a critical issue for offshore projects and can be improved by this application. CB&I have decided to carry out a research related to the occurrence of sloshing and impact pressures for these, so called, in-deck tanks. The sloshing assessment procedure is an important part of the structural strength checks. Sloshing occurs when the natural period of the fluid coincides to the motions of the storage tank. Four factors mainly contribute to the sloshing phenomenon. Namely, tank dimensions, fill, fluid properties and motion characteristics. However, the complex, chaotic and non-linear behaviour of sloshing makes it hard to predict or estimate impact pressures. In-deck tanks are applied at the topside of the Aasta Hansteen SPAR project, carried out by CB&I. The application of these tanks faced difficulties concerning the sloshing assessment procedure. There is no method applicable related to this situation. Therefore, a conservative method has been defined as a temporary solution. For future implementation of these tanks, better understanding and knowledge of fluid behaviour is essential. In order to tackle this problem, a CFD analysis is carried out in two phases and concludes with a statistical analysis in order to estimate sloshing impact pressures. The first phase relates to a general 2D CFD simulation for various cases. The second phase includes 2D long time simulations of sloshing cases extracted from the first phase. The results of the first phase show that no sloshing occurs for the Aasta Hansteen SPAR related cases. Where the motion period of 60 seconds is too far away from the period of the 1st wave mode, which is around 8 seconds. FPSO related cases contain a period around 10 seconds and show sloshing impact behaviour. The impacts occur specifically for longer tank lengths and higher filling levels as these cases coincide better with the motion behaviour of the tank. Noted that the combination of input parameters for which sloshing occurs is highly dependent on the forced excitation on the tank, where sloshing behaviour is sensitive to changes of these parameters. Furthermore, a motion case analysis is added and different sea states are assessed from mild to harsh tank motion excitations. Resulting in sloshing for harsher sea states and higher accelerations. Overall, sloshing impacts conclude in the order of 100 kPa - 300 kPa. The impact area includes the vertical wall and 2.4 meters on the top of the tank. In the event of non-impulsive oscillating behaviour (no sloshing), one can apply the linear theory for an accurate prediction of the pressures. However, when the fluid motion becomes chaotic and non-linear, there is no method able to accurately predict the impact pressures. The results of the second phase contain the sloshing impact order of magnitude for eight individual sloshing cases. With difference in fluid, tank length, fill and motion type. Six of these cases can be compared to one another and resulted in a fill/length ratio of 0.063 for the highest impact pressures. A lower viscosity of the fluid seems to increase the sensitivity to sloshing behaviour. Filling levels of 50\% - 70\% show high sloshing impacts, where 80\% fill does not result in sloshing anymore. The increase of tank lengths results in higher sloshing impacts. Briefly summed up: 7m no sloshing, 12m semi sloshing, 15m sloshing impact order 150 kPa - 200 kPa and 20m sloshing impact order 300 kPa - 500 kPa. Two fitting curves are used in order to establish the Exceedance Probability Function. Namely, the Generalized Pareto and Kernel Smoothing. Both show a good fitting, but present different behaviour in the so called 'tail' of the Probability Density Function. A good distribution of this 'tail' result in a better Exceedance Probability Function. A decision on the best fitting curve is not made due to the lack of sufficient simulated statistical data. The sensitivity analysis proved the Kernel Smoothing fitting more robust compared to the Generalized Pareto. Also, the reduction of statistical data resulted in the highest sensitivities within the sensitivity analysis. Which underlines the need of more simulation and statistical data for improvement of the results.