Jackup foundation punch-through in clay with interbedded sand

Jackup installation in multi-layered seabed stratigraphies may pose significant risk of punch-through failure. Punch-through failure typically occurs when a strong layer such as sand or stiff clay overlies a relatively softer layer, such as soft clay. Over the last two decades punch-through investigations were mainly concerned with double layer sand overlying clay stratigraphies. Offshore investigations show that multiple layers of varying strength soil, typically where a sand layer is interbedded in a bed of a single clay layer, occurs frequently. These conditions have previously caused fatalities due to rapid punch-through of a jackup leg. An improved understanding of punch-through failure in a multi-layer stratigraphy and identification of the governing mechanisms of punch-through are therefore important. The fundamental outcome of this research is the development of a universal analytical model for routine punch-through assessment in a three layer clay-sand-clay stratigraphy as well as two layer sand overlying clay stratigraphy, with scope for generalisation to additional layers. To achieve this, extensive half-model - with particle image velocimetry (PIV) based observation of the soil flow mechanisms - and full-model drum centrifuge experiments were performed alongside complimentary large deformation finite element (LDFE) analyses.

Punch through modelling involving multiple soil layers in a centrifuge is challenging as tests are conducted in strongboxes of finite and limited dimensions. As the proximity of the strongbox walls (side and bottom) will affect the penetration resistance of the foundation, a comprehensive set of LDFE analyses were undertaken in clay, sand and sand overlying clay with design charts developed guiding the minimum boundary distance required to avoid any boundary effects. It is shown that the bottom wall boundary effects are enhanced in sand on clay compared to single layer clay due to entrapment of a sand plug under the foundation. This enhancement depends on the size and shape of the entrapped sand plug. The side wall boundary effect is a function of both boundary roughness condition (rough or smooth) and the normalised sand thickness (normalised against the foundation diameter). The greater the normalised sand thickness, the larger is the strongbox width requirement. Overall this parametric study showed that some prior work has overlooked the increased boundary distances required when stiff layers and punch-through are modelled, which has potentially led to misleading results. Improved criteria were then adopted throughout the present study.

Observing the entire punch through event, i.e., the measurement of the peak bearing capacity close to the sand layer and the underlying clay bearing capacity, at which the foundation stabilises into the underlying clay layer, is not straightforward in a centrifuge experiment. It requires detailed planning in terms of soil geometry, soil properties and foundation size. For consistent observance of punch-through, with reliable results that span the relevant range of responses, a methodical approach involving prediction of and reflection on the likely outcomes was adopted in the experimental planning, a process termed as ‘step zero’.

Following findings from the LDFE simulation and the ‘step zero’ approach, extensive centrifuge testing was executed in a clay-sand-clay stratigraphy utilising model foundations of prototype size 6-16 m over a wide range of normalised sand and clay layer thickness, and soil properties. High quality digital images were captured in flight and later analysed utilising the PIV technique. The PIV observations identified key soil flow mechanisms on the basis of which a stress-dependent model is developed furthering previous punch-through models. The centrifuge data bank is utilised to verify the newly developed model.

It is shown that the developed model provides a rational approach for predicting the full bearing capacity response in a clay-sand-clay stratigraphy. It performs significantly better in predicting the risk of punch-through when compared with the current models used in industry practice based on punching-shear and load-spread. These models are rather conservative, in that the load at which punch-through will occur is generally under-predicted. However, the models also sometimes fail to predict punch-through in conditions where it will occur. This makes them both over-conservative but with low reliability. The model developed is universal in the sense that it can be applied to both clay-sand-clay and sand on clay stratigraphies (assuming that sand shearing is always under a drained state and clay shearing is always under an undrained state) by simply setting the top clay height as zero, and it is consistently accurate and reliable when compared with the large database of model tests.