Experimental investigation of global wave impact loads in steep random seas

This paper presents the comparison of two designs of a buoyancy tank mounted on a semi-submersible platform. The global hydrodynamic hinge forces relative to the platform leg were measured during wave impact based on model test in steep irregular waves corresponding to 10-4 annual probability. The main objective of the experimental investigation was to quantify the difference between the two design concepts regarding the hinge forces and includes verification of an efficient test procedure. The model set up is simplified. Only one leg and the buoyancy tank represent the platform and the structure is retained from any motions. Test runs in random waves with 20 different seed number realizations of the Torsethaugen type sea state with Hs = 20.7 m, Tp = 20.1 s were generated with a full scale duration of three hours. From each 3-hrs realization the sample extreme impact event is identified for both designs. Due to the high sea state in this case, most of them included breaking waves. A technique recently developed for obtaining the extreme value statistics without the need for full-duration basin tests was tested and verified. Thus, based on the "top 20 extreme impact events", additional model tests with spliced random events were performed. This was done through the repetition of the wave maker’s control signal around the events, including some "buffer time zone" before and after. Three such top 20 tests sequences were carried out and compared. Two of them used identical control signals for both designs where the time of the events was defined by the load peaks of the first design. The third sequence was realized with the second design where the times of the events were defined by the load peaks of the second design. The resulting extreme value statistical models from this new kind of data sampling show good similarity to the extreme value statistical models built from the 20 full-duration 3-hours realizations. The new method, presented here, can reduce the cost of an experimental ampaign of different designs comparing extreme value analysis.