Performance of suction caisson anchors in normally consolidated clay

As the demand on energy rapidly increases, oil companies extend their search for oil and gas into deeper waters in which floating structures are most economical. These structures are tied at the seafloor with anchors that can sustain loads from waves, storms, and currents. Suction caissons are anchors that utilize the large water pressure in deepwater during anchor installation, making it an efficient and economic alternative to driven piles. In places all over the world, suction caissons are widely used as foundation anchors in normally consolidated and lightly overconsolidated clays for a variety of deepwater structures. Suction caissons in offshore applications are subjected to a wide range of loading conditions. Loads are vertical in tension leg platforms, inclined in taut mooring systems, and mostly horizontal in catenary systems. However, the load capacity of suction caissons is not well defined. Several analytical and numerical models have been published to estimate the capacity of suction caisson, but very little experimental data is available to support such models. Laboratory tests were conducted in an experimental facility specially built to study the behavior of suction caissons under axial, horizontal, and inclined loading conditions. The experiments were performed using two 4-inch diameter prototype caissons inserted to a depth of 32 inches in normally consolidated kaolinite. The tested prototypes are representative of caisson geometries commonly used in mooring systems for deep offshore locations having soft seafloor sediments. The first prototype caisson had a padeye bar along its lower half to allow for horizontal and inclined loading below mudline. The second prototype caisson was built from two thin tubes forming a doublewalled caisson capable of providing separate measurements of the components of axial capacity. Instrumentation was used to measure loads, displacements, tilt, and pore water pressure for loads ranging from horizontal to vertical. In most tests, the caissons were inserted into the test bed soil half way using deadweight followed by suction insertion to full penetration. In some axial loading tests, the caisson was inserted by deadweight to full penetration for comparison. The caisson was loaded rapidly after allowing for sufficient setup time. Tests were also conducted with partial setup times to examine the effect of setup on the axial capacity. The caisson top cap was sealed in all horizontal and inclined tests, while axial loading tests were conducted with sealed and vented top caps. Caisson response during insertion, setup, and loading is presented. Measured capacities are compared with analytical and numerical predictions. The axial capacity of the caisson was the same whether the caisson was installed using deadweight or suction. In case of axial loading with a sealed top cap, the limit equilibrium parameters αexternal and Nc were calculated to be 0.8 and 15, respectively. The external side friction measured from tests with a vented top cap was higher than from tests with a sealed top cap. In case of axial loading with a vented top cap, the limit equilibrium parameters αexternal and αinternal were 0.85 and 0.5, respectively. The external side friction during pullout of the caisson was observed to increase with setup time until the external excess pore pressures were practically dissipated, while the end bearing resistance was not affected by the setup time. Results from horizontal loading tests indicate that maximum capacity is achieved when the caisson is loaded at a depth between two-thirds and threequarter of the embedment depth. The failure mechanism and the generated excess pore pressures depend on the position of the load application. Displacements measured in inclined loading tests were predominantly horizontal for loading angles less than 20° from horizontal and were predominantly vertical for loading angles above 30°. Good comparison was found between measured capacities and predictions from a plasticity model.