Sandy soils are characterized by negligible cohesion (compared to clay) and limited drainage capacity (compared to gravel), which makes these soils susceptible to excess pore (water) pressure accumulation induced by short term dynamic loading cycles. Pile driving subjects the soil to such type of dynamic loading. Accumulation of excess pore water pressure continuous until contact between the sand particles is lost, a process known as soil liquefaction. This research investigates two soil configurations which are vulnerable to pore pressure accumulation and subsequent liquefaction induced by pile driving: slopes and confined aquifers. Slopes are vulnerable because of their geometry. Limited accumulation of excess pore pressures might be sufficient to induce sliding failure. Confined aquifers are vulnerable because of the lack of drainage boundaries. This can result in significant excess pore pressure accumulation in the aquifer. Pile driving induced liquefaction instability is modelled by considering the pile-soil-plug interaction, the emission, propagation and attenuation of waves into the soil domain and the resulting generation of excess pore pressures. Two models are developed for this purpose. The first is an cylindrically symmetric damped elastic pile-soil-plug model. This model calculates the vertical and radial displacements in the soil domain as a function of space and time. Shear stresses, which are the driving parameter for the second model, are then calculated as a function of the spatial derivatives of the two displacement components. The second (liquefaction) model is a combination of the governing differential equation for cylindrically symmetric soil consolidation and an empirical model describing generation of excess pore pressures as a function of shear stress amplitudes. The output of this model is the steady state relative overpressure distribution. The relative overpressure distribution can then be used as input into slope stability analysis software in order to investigate whether a slope failure will occur given the reduced soil strength. The biggest uncertainty in the second model is the coupling with the first model. This coupling is achieved by adjusting the elastic shear stresses (which are unrealistic and can therefore not be directly used as input for the liquefaction model) calculated in the first model. Three variables are defined for this purpose. Analysis of the sensitivity of these variables show that the variable beta, which is the ratio between yield shear stress and initial effective stress of a liquefied interface, is the most sensitive and therefore most uncertain variable in the model. Experimental validation of beta is recommended to the increase accuracy of the liquefaction model.

Uncertainties in the soil parameters play a major role in the design of quay walls. In the current design approach, partial factors are prescribed to account for uncertainties in the soil, as well as for other types of uncertainties. This semi-probabilistic (level I) design approach needs to result in a reliable design for a range of quay structures and for multiple soil stratifications. It is therefore expected that, in general, this method results in overdimensioning of the structure. Whether this assumption holds, is investigated in this thesis by carrying out a reliability analysis for two quay walls in the Port of Rotterdam. The first case study covers a simple double-anchored combi-wall, whereas in the second case study a quay wall with relieving platform is considered. Only the most relevant failure mechanisms were considered, which are yielding of the combi-wall, yielding of the anchor bar, shear failure of the grout body and soil mechanical failure. These failure mechanisms are complex soil-structure interaction problems. Therefore, both the soil and the quay structure have been modelled with the finite element program Plaxis 2D, using the Hardening soil model. For performing the probabilistic calculations on this model, the probabilistic module ProbAna has been used. This is a package developed by Plaxis which couples several types of reliability methods to the finite element software of Plaxis 2D. As the computational effort is relatively large when using FEM, the First Order Reliability Method (FORM) was used over sampling methods like Directional Sampling and Crude Monte Carlo simulation. The results for the simple quay wall showed that the reliability level was sufficient for all considered limit states. Hence, almost all partial factors derived on this quay wall were lower than currently prescribed by the Eurocode. In the second case study, a quay wall with relieving platform, monitoring data of multiple years was used for calibration of the Plaxis-model. Thereafter, the reliability for the limit states yielding of the wall and soil mechanical failure was evaluated. It turned out that for both limit states, the reliability index was too low compared to the target reliability. Although each case study concerned a different type of quay wall, the results reveal that choices made in the design, either optimistic or pessimistic, can have large influence on the reliability. Perhaps just as important, are the assumptions made regarding the stochastic description of the soil. It is still under discussion up to what distance soil parameters are correlated in space and how spatial averaging should be applied. Reference calculations showed that choices regarding the amount of independent soil layers and the degree of spatial averaging have a large influence on the reliability. More fundamental research to these topics is therefore recommended.

The objective of this thesis is to enhance understanding of the behaviour of a flexible dolphin and its interaction with the surrounding soil in order to determine the optimal embedded depth. A comprehensive field test is conducted to gain deeper insights into the behaviour of the flexible dolphin and the soil surrounding the pile. The test measurements are then analysed to assess the pile's behaviour. Additionally, calculation models are employed to predict the pile's behaviour, and a comparison is made between the measurements and predictions. Four different calculation models, namely Blum, Brinch Hansen, P-y curves, and Plaxis are utilized to predict the pile's behaviour during the test. The pile behaviour is compared across these models. While most models do not account for repetitive loading, the best available inputs are employed to simulate the test as accurately as possible. The majority of the models exhibit similar top displacements of 0.9 meters, except for the drained Plaxis model, which calculates displacements of 1.05 meter, and Blum's model, which calculates a displacement of only 0.8 meters. During the test, multiple instruments are employed to measure various parameters of the pile, including load, displacement, water pressures, and strains at different depths and time intervals. The total station measurements indicate a maximum displacement of nearly 0.7 meters, which differs by 0.2 meters from the predictions. The deviation between the predictions and measurements can largely be attributed to inaccuracies in the hydraulic jack load measurements. The pile load was back-calculated using the strain data obtained from the pile. The loads estimated based on the strain measurements were found to be hundreds of kilonewtons lower than the hydraulic jack load measurements. As the strain measurements were considered more reliable, the actual load during the test differed from the prescribed load scheme. Consequently, the load inputs in the models need to be adjusted to align with the loads derived from the strain data. The displacement measurements are analysed and compared with each other. The Saaf and optical fibre measurements exhibit similar curvatures, enhancing the reliability of the optical fibre strain measurements during load analysis. However, the total displacement measured by the Saaf and the total station do not match due to the incorrect assumption of zero movement at the pile tip. Determining the exact displacement of the entire pile is impossible due to the inadequate number of boundary conditions available to translate the Saafs curvature measurements into displacement. The calculation models utilize the loads derived from the strain measurements to make predictions, and their results align comparably with the measurements. Plaxis is the only model capable of simulating different load cycles, which improves the comparability of the Plaxis results with the test. Load-displacement graphs obtained from the Plaxis results and the optical fibre measurements display hysteresis in load cycles, consistent with findings in existing literature. The initial load exhibits less stiffness compared to the repetitive loads in both datasets. The amount of energy absorption by the soil is determined from the areas of the loadcycles. When considering the most complete load cycles, Plaxis conservatively estimates the absorbed energy. Reducing the length of a pile results in larger displacements of the pile, while concurrently enhancing its capacity for energy absorption. However, it is crucial to strike a balance between these two parameters, while paying close attention to the permanent soil displacements.