Displacement control is of utmost importance in deep excavation design and is usually based on numerical modelling, e.g. Finite Element Method (FEM). Numerical methods tend to be more conservative when analysing soil behaviour during deep excavation, whereas for practical and economic reasons this is not favoured. The inverse analysis allows for the identification of the soil parameter set that can provide the measurements observed in the monitoring when it is applied in the model. When performed in a probabilistic concept, it reduces parameter uncertainty and enables the stochastic prediction of future soil behaviour. In this thesis, capabilities and limitations of difference advanced constitutive models are investigated. The Generalized Hardening Soil Small strain model (GHS) presents a positive aspect in modelling soil behaviour during deep excavation with its various stress/strain dependency settings. Because of the uncertainties originating from the size of the domain and limitations of site investigation, the soil parameters can only be shown as probability distributions. In order to make that distribution more accurate, comparative selection of several inverse analysis optimization algorithms is performed. Thereafter, choice of the relevant parameters is done based on the conducted sensitivity analysis and engineering judgement. Having the most competitive optimization approach selected, remote scripting with Python is used to utilise Finite Element (FE) modelling in the 2D Plaxis software. The input parameters are iteratively updated with response observation (diaphragm wall deflections) using the Ensemble Kalman filter optimisation algorithm based on a chosen excavation stage. The re-calibrated parameters are checked with the data, which was used to create synthetic measurements made using the same FE, to perform reliability assessment of the developed Python-based algorithm and investigate its capabilities and limitations. The further development of the presented optimisation method is expected to increase certainty in setting alarm thresholds in the applications of the Observational Method.

This thesis discusses the characterization of the undrained shear strength, S_u from the net cone resistance, q_net of clay in Dutch sites using Hierarchical Bayesian Modelling (HBM). The performance of the HBM is compared with the current practice methods of the site characterization which propose either the use of only site–specific observations (unpooled models) or the whole data simultaneously (pooled models). HBM can incorporate information from multiple sources such as prior knowledge of the engineers and behaviour met in the examined and the neighbouring sites. The use of different sources of information has been proposed by Eurocode-7 without providing a formal / mathematical procedure. Literature studies have highlighted the potential benefits of incorporating the HBM into the characterization of the geotechnical parameter values. Therefore, this thesis aims to assess whether HBM can enhance the geotechnical decision-making by precisely quantifying the uncertainty in the geotechnical parameter values and making more accurate predictions of them. The impact of using input from the HBM results in a reliability analysis of a dike slope is examined as well. First, a considerable number of paired q_net–S_u measurements is collected, and subsequently is divided into groups. Different statistical models are employed to describe this collected data. Two components characterize a statistical model; the functional form which is the relationship between S_u and q_net and the pooling family (pooled, unpooled and HBM), the method followed to train the statistical model parameters. The statistical models are applied in a comparative study to select the fittest one and to compare the behaviour of the HBM to the other pooling families. The comparative study is performed by applying the Bayesian Data Analysis (BDA) whose applicability is ensured by applying it in an artificial example using artificial data. The first result of the BDA with real data is the comparison of the HBM with the current practice pooled and unpooled models suggesting the ln⁡〖S_u 〗-ln⁡〖q_net 〗 HBM as the fittest model. The HBM estimations for the statistical model parameters fall between the current practice’s methods and they experience lower uncertainty by borrowing information from the neighbouring sites to make site-specific estimations. Between the current practice and the HBM, the latter predicts the S_u with lower uncertainty. The reliability analysis using input from the HBM yields different reliability indices than those proposed by the current practice models. This situation combined with the choice of the HBM after following the BDA workflow propose that the HBM can lead to safer and more economic design. Overall, the use of the HBM for predicting the S_u from q_net with grouped data can be beneficial for the engineering practice. First, the HBM reduces the uncertainty of the statistical model parameters without inheriting extreme values and provides more certain prediction for the S_u accounting for the prior engineering knowledge and the behaviour met in neighbouring sites. Additionally, performing reliability analysis of a dike slope exhibits that the use of HBM derived values can suggest safer and more economic design over the standard approach.