In March 2016, an addendum to the Dutch design guideline for micropiles, CUR 236, was published. This addendum elaborated on the ‘axial stiffness of micropiles’. The softening behaviour was here first mentioned as an aspect relevant for the axial stiffness of the micropiles. Due to the slender geometry of the micropiles, this further translates into axial capacity. Up to this moment very little research was done into this phenomenon and to be conservative a ‘best guess’ of 50% of the peak shear stress was made. The goal of this thesis is to have a more profound understanding about this (strain) softening behaviour, leading to the main research question: “How does strain softening manifest for micropiles under tensile loading?”. This strain softening behaviour can be further subdivided into the mobilisation from peak to residual shear stress and the reduction of the residual shear stress relative to the peak. Furthermore, the installation effects play a big role in the bearing capacity. Also, some thought is given to the large scale measurements of the behaviour. Three different paths are chosen to research the strain softening behaviour. Numerical modelling, small scale tests on sand and large scale pile tests are performed. In case of the numerical modelling, the large scale situation is simulated in the numerical software Plaxis, using an axisymmetric model and the hypoplastic constitutive model. Direct shear tests are assumed to represent a small part of the micropile, simulating the local softening behaviour. Numerous tests are performed varying relative density and top pressure to assess the local strain softening behaviour. Last, a method is developed to quantify the occurring ratio between peak and residual shear stress in-situ, extending on the current prescribed testing procedure (CUR 236, 2011). The only constitutive model available that describes the critical state soil behaviour for sand is the hypoplastic sand model. However, the combination of the tensile loading, softening implementation into the hypoplastic sand model in Plaxis and the deformations, induce considerable inaccuracies and physically impossible behaviour. This rejects further elaboration with this approach. The direct shear tests are performed on only sand, based on the results of Uesugi & Kishida (1987), assuming a rough surface. In summary, a clear dependency on relative density and average confining pressure is observed. Varying from a reduction of approximately 30% for samples with the maximum possible density to 0% for a sand with a medium dense packing. Furthermore, the displacements necessary to mobilise from peak to residual shear stress show a uniform trend, independent of relative density or pressure. Quantitatively, the displacement ranges between 2 and 3 mm. In the large scale tests, due to unexpected, structural failure, the majority of the test results, unfortunately, were unusable. The pile test that did produce workable results, behaved as expected. It could be deduced that the used methodology works. The use of the hypoplastic constitutive model in Plaxis, does not give the expected results. It can thus be concluded that a fully coupled stress – strain analysis of the problem is unfortunately impossible. Concluding on the direct shear tests performed, the implementation in the CUR 236 addendum is too simplistic. In terms of ratio between peak and residual shear stress, a dependency based on relative density and average confining pressure should be included. In terms of displacement, the found displacement is larger. It should however be noted that the direct shear tests are only a simplification, approximating the actual situation. The performed large scale tests do show results in line with expectations and could be used in the future to assess the softening behaviour in-situ.

The construction method of an overturning caisson is reconsidered. This particular self-floating caisson was applied for several caisson quay wall projects over a century ago. After the year 1914, the design was abandoned. After this time, rectangular caissons became more economical due to higher qualities, simplified formwork and easier transport. However, the caisson geometry allows efficient material use and a relatively low draught during transport. When the concept is considered for current quay wall projects, these benefits are still present. However, the magnitude of these benefits reduced significantly. When other aspects, such as construction technologies are included, the remaining economic benefits are estimated to become negligible. The economic feasibility of an overturning caisson could therefore not be demonstrated.

The uncertainties in undrained macro-stability analysis of dikes lead to overdimensioned and inefficient designs. The undrained shear strength (su¬), which describes the strength of clay and peat layers, is important information for the macro-stability analysis of a dike. The cone factor (Nc) gives a linear correlation between undrained shear strength (su¬) of the soil and the cone resistance (qc) from a cone penetration test (CPT)¬. In practice the uncertainty on this correlation, given by the variation coefficient, is high, especially it overestimates the su for high values of qc. As a result, a large reduction factor (RF) is taken into account between the expected value of su and the characteristic value of su used in calculations. In literature the cone factor (Nc) is assumed to depend on the soil stiffness, the cone roughness and the initial soil stress state (∆). Including the initial soil stress state, or more specifically the horizontal soil stresses (σ’h0), in the cone factor determination can lead to a reduction of the variation coefficient. Based on literature and a numerical analysis it is found that including the initial soil stress state in the determination of Nc can lead to a difference of 10% in cone factor. However, the difference is less than 10% for most measures due to the relatively similar stress conditions over a large part of the cross section of a dike. In addition, a theoretical application of the initial soil stress state (Δ) shows that the overestimation of su for high values of qc cannot be solved by including it as found in literature. A case study on the Hollandse IJsseldijk project has been conducted to back up the theoretical approach with experimental measurements. As a conclusion it is found that the variation coefficient is reduced slightly taking into account the initial horizontal soil stresses, resulting in a decrease of the reduction factor (RF) of 2-3%. Taking this small improvement and remaining unknowns in the calculation into account, it is concluded that including a measure of the in-situ soil stress state (Δ) in the determination of the cone factor (Nc) does not result in a significant reduction of the present uncertainties in dike assessment with existing approaches.

A time-dependent capacity increase has often been observed for driven displacement piles in sand. Long-term bearing capacity increase after dissipation of excess pore water pressure is referred to as set-up. Significant increases in axial capacity have been reported where the rate of set-up varied between 15 and 75% per log cycle of time. More economic foundation designs are feasible when set-up is incorporated in pile design methods. The objective of this thesis is to expand knowledge on time-dependent capacity of driven displacement piles and make way for a pile design method that incorporates time dependency of pile bearing capacity. The time dependency of virgin pile capacity has been assessed in a database study and a scaled test program. 4 scaled closed-ended steel piles were driven in a man-made sand deposit. The virgin capacity of a pile was assessed 2, 16, 31 and 70 days after installation. The 3 m long, 0.15 m diameter piles were statically tested in compression. Strain was measured at multiple locations along the shaft during the load tests. The applied load and pile displacement were measured at the pile head. The contribution of the pile shaft and pile tip was determined from the strain distribution over the pile length at failure. A trend for capacity increase with time has been found in the database study and the test program. A capacity increase of 13% per log cycle of time was found for the closed-ended piles tested. Both an increase in shaft and tip resistance was found with respect to the reference capacity determined 2 days after installation. The shaft capacity increased on average 18.5% per log cycle of time and an average increase of 9% per log cycle of time has been found for the pile tip capacity. The measured capacity was compared with the calculated capacity determined by the Dutch design code. Considering the measured capacity normalized by the calculated capacity the influence of time is even stronger on the tip resistance than on the shaft resistance for the piles tested. As discussed in the report, there are some uncertainties with the interpretation of the strain distributions over the pile shafts. The accuracy of the CPT-based design method available in the Dutch design code, method Koppejan, has been assessed from the pile test results in the database. The results were compared with other design methods available. The accuracy of all methods was low. The calculated capacity exceeded the measured capacity up to a factor 3. Time-dependent capacity increase is a potential mechanism to incorporate in designs of axially loaded driven displacement piles. A trend for time-dependent increase of virgin capacity was found for all the test results considered in literature as well as for the test program conducted. However, the rate of capacity increase was highly variable. More research is required on the mechanism(s) for a better understanding of time-dependent capacity increase. Understanding of the mechanisms will also help understand what factors, and how these factors, influence the rate of increase.

In this thesis an evaluation of the limit values which are stated in CUR 236 – Micropiles is presented. This evaluation has been done by means of a dataset. The dataset contained information on failure tests performed on four different types of micropiles, and was initially used for research on relations between logged installation aspects and the final capacity of the piles; such relations were however not found. This was mainly caused by the shape in which the data was logged, and the lack of details in the data. During this research, significant differences between raw data and data adapted according the limit values stated in CUR 236, were found. This led to a change in research direction towards the evaluation of the limit values. Limit values are values that are used to build in additional safety in a design, and thus, to prevent unsafe situations. The values are based on the shape of data in which in general a linear relation between maximum mobilized shear stresses (τ_(mob;max)) and cone resistance (q_c). This relation continues up to a certain cone resistance level, after which τ_(mob;max) does not increase significantly anymore. The point where no significant increase in τ_(mob;max) over q_c was seen, was used as the limit value for the cone resistance. The limit for α_t was chosen to be 2.5% and the limit for τ_(mob;max) 2.5% of the cone resistance limit. Data with values above these limits, has to be reduced to the limit values. (CUR236, 2011) Differences were thus found between raw data, the same data but then adapted according the limit value method stated in CUR 236, and values stated in CUR 236 Table 6.1. For further investigation on how these differences were caused, the available dataset was combined with data from the appendix ‘Bijlage A. Proefbelastingen’ of CUR 236. This showed that the current limit values did not suit the shape of the data very well. Investigation on the usefulness of the limit values showed that they are useful for the design process by preventing under- and overestimations of the capacity. Limit values are thus useful, but not in their current form. New limit values were therefore proposed based on a statistical analysis of the available data points. Next to limit values, also design values (expectancy and lower bound) are added. Because no detailed data for all micropile types was available, it is advised to re-elaborate the limit- and design values for the types that were left out of the proposal given in this thesis, in a similar, statistical way. Furthermore is it advised to investigate a different method to estimate the shear stresses that can be mobilized along the anchor body of a micropile, based on a more fundamental approach. Currently are those stresses often based on full-scale tests from which τ_(mob;max) is derived, but bounded by the limit values. A different approach to this might be more time- and cost-effective.