An accurate numerical simulation of a mechanized tunneling process in an urban area must consider the complex interactions between subsoil, the tunnel boring machine, and ground constructions. Of course, due to the natural origin of the geomaterials, their characteristics deal with randomness. The considerably high level of associated uncertainty inherent in geomaterials may lead to notable deviations in the prognosis of geotechnical structures. To address this issue, a model adaptation framework is presented, which intends to minimize the involved uncertainties in the simulation of mechanized tunneling. In this framework, first parameter identification techniques are developed to reach an adequate soil model for numerical simulations based on measurements. Accordingly, the concept for an optimal measurement campaign is introduced. The optimum observation design is supposed to identify a sensor arrangement, which provides the least uncertainty in the parameter identification process. The optimized measurement concept is developed employing sensitivity indices and other probabilistic tools including Bayesian updating methods. The concept of model adaptation is further developed by involving the field data during an intermediate boring phase in the reliability assessments of the other advancement phases. To do this, a Bayesian updating concept is combined with a Markov chain Monte-Carlo method to evaluate the updated reliability measures, considering the ground settlement as the limit state. Nevertheless, due to the infeasibility of performing an extensive geotechnical site characterization in a spacious project as tunneling, some geological alternation might be overlooked. Here, the adaptation framework proposes a supervised machine learning methodology to predict the geological changes ahead of the TBM. The classification is performed based on different supervised learning algorithms, which assign the obtained characteristics to the predefined geological conditions.
In recent years, stochastic simulation has obtained increasing attention in geomechanics as it allows to consider uncertainties in geomaterials’ properties. Spatial uncertainties may influence simulation results and lead to either unsafe or uneconomic designs. By implementing stochastic random field algorithms in Finite-Element simulations, the impact of such uncertainty can be integrated in the simulation results. However, the application of such algorithms implicates new requirements in understanding the relationships between correlation length, mesh coarseness, and required number of Finite-Element simulations. The present study intends to provide a deeper understanding of such aspects. In the presented approach, stochastic random fields analysis is applied in a Finite-Element code by remote scripting. Considering a slope stability analysis and using Monte-Carlo Simulations, the impact of varying mesh coarseness and the correlation length on the obtained safety factor is investigated. As for large numbers of mesh elements, computational costs increase rapidly, the study expands its investigations by employing two further advanced methods, namely mesh adaptivity and subset simulation. These state-of-the-art methods allow to reduce calculation efforts, once by iterative remeshing, once by selective sampling according to failure probability. The results show promising improvements in the concept of computational efforts.
The control of ground movements is critical in tunneling projects due to potential damage to existing infrastructures. The specification of tunnel control parameters requires a good knowledge of the surrounding soil stratum. As boreholes are discretely distributed, interpolations based on obtained information may conduct to approximate the entire overall stratigraphy, which results in inevitable uncertainty levels in the subsoil identification process. For instance, high stiffness blocks, thin weak layers or faults between two adjacent boreholes may not be properly identified. Due to this reason, a pattern recognition approach is proposed in the present study to predict the soil stratum variations through tunnel advancement. The behavior of soil domain subjected to mechanized tunneling is represented in a multi-dimensional feature space and the unexpected soil layer in front of tunnel face can be classified into one of the three classes: a soil type change, an interlayer, or a block with higher stiffness. The potential condition of the unexpected soil stratum in front of tunnel face can be assessed according to the different features in the monitored data, which are extracted by different machine learning methods like TS-Fresh, Support Vector Machine and K-Nearest Neighbor. The training set of data are derived from the numerical simulation results of a 3D finite element tunneling simulation, which are applied together with noisy data to generate the machine learning set up. Afterwards, different techniques are investigated to achieve an optimal estimation of the soil stratum in front of the tunnel face.
To provide realistic predictions of mechanized tunnel excavation‐induced ground movements, this research develops an innovative simulation technique called hybrid modeling that combines a detailed process‐oriented finite element (FE) simulation (submodel) with the computational efficiency of metamodel (or surrogate model). This hybrid modeling approach has three levels. In Level 1, a small scale submodel is cut out from the global model and the continuous simulations are conducted in this submodel. Level 2 deals with identification of uncertain soil parameters based on the measurements (e.g., surface settlements) during tunnel excavation. In Level 3, the tunneling process parameters (e.g., grouting pressure) can be optimized to control tunneling‐induced ground movements or building deformations according to the design criterion. The proposed hybrid modeling approach is validated via a 3D numerical simulation of mechanized tunnel excavation. The results show the capability of the proposed approach to provide reliable model responses in the near field around the tunnel with reduced computational costs.
Tunnel construction in urban areas neccessitates not only a safe construction of the tunnel itself, but also minimising the impacts on the environing infrastructure. Constructing a tunnel always causes changes in the state of stress and deformations in the surrounding soil area that can induce settlements, cracks, or even collapse in buildings at the ground surface. To plan effective countermeasures, an adequate model is necessary that considers all relevant details of the system. However, the main component of this system is the adjacent soil that cannot be modified and which properties are highly uncertain. To reduce the high level of uncertainty, in-situ measurements are preformed and used in a back-analysis to validate the existing model. Reducing the uncertainty of the soil parameters allows more reliable predictions of the system behaviour. However, initially only assumptions can be made what type of measurement might be most suitable to reduce the parameter uncertainty most efficiently. Using the approach of Bayesian optimal experimental design, the present study enables to find an arrangement of sensors that provides data which is most likely to enable an accurate model validation. The application of this approach is performed using a Finite-Element model of a residence building that is underpassed by twin-tubed tunnel.
This paper presents an approach to optimize the reliability-based optimization design of a rock salt cavity. Liquid or gaseous energy resources can be stored in rock salt caverns which they are basically planned to peak shaving the strong fluctuation in the seasonal demands. The validation of their integrity and stability is a prerequisite in the geotechnical design process of such structures. The present study provides a reliability-based analysis to optimize the design of a typical energy storage cavern in rock salt. An elasto-viscoplastic creep constitutive model is applied to a finite element numerical model of rock salt cavern to assess its behavior. The constitutive parameters are represented as random variables. Failure of the system integrity is evaluated by occurrence of dilation in any spot of the surrounding rock salt. In theory, one can adjust the most important design parameter, namely the minimum internal pressure, to prevent any failure cases. On the other hand, maintaining a high internal pressure for long time in the cavity is not economically favorable. To achieve a robust design, a multi-objective optimization method is employed to set design parameters in a manner that the integrity of the cavity is insensitive to the variations in rock salt properties, while it also minimizes the operational costs. In this regards, the failure probability of the cavern is calculated by subset simulation.
In geotechnical engineering, using monitored data for model validation is common practice. However, a model-based optimal experimental design for parameter identification is unusual when planning a monitoring set-up. As soils are subjected to large parameter uncertainties, model validation is of high interest to enable a precise prediction of the system behaviour. Due to the complexity of considered cases and employed FE-models, time-efficient solutions are of interest. For the case of a dike subjected to a rapid drawdown of the current water level, a Monte-Carlo based approach was previously employed. In the present study, it is intended to improve the efficiency by applying the so-called sigma-Points method that substitutes random sampling by defined characteristics of model response distribution to set-up a monitoring design that allows to identify the relevant system parameters.
Numerical simulations become an increasingly utilised tool for predicting soil behaviour in geotechnical engineering. However, these simulations require reliable soil parameters that have to be determined either by conducting laboratory experiments or by model parameter identification using in-situ measurements. To gain reliable and sufficiently informative measurement results for successful determination of the soil parameters while being still economic, an optimal experimental design is of crucial importance. This work presents a procedure of setting up a monitoring concept for a dike with uncertain material properties exposed to a rapid drawdown of the adjacent water level. The illustrative example is numerically modelled using a coupled hydro-mechanical analysis and the parameters that are most relevant for the dike stability are firstly identified via global sensitivity analysis. Subsequently, the global sensitivity analysis with respect to measurable outputs is employed to identify appropriate monitoring areas. Finally, bootstrapping is conducted to make possible to decide on the optimal observation sensor locations for reliable identification of the most influential soil parameters.
The large number of input factors involved in a sophisticated geotechnical computational model is a challenge in the concept of probabilistic analysis. In the context of model calibration and validation, conducting a sensitivity analysis is substantial as a first step. Sensitivity analysis techniques can determine the key factors which govern the system responses. In this paper, three commonly used sensitivity analysis methods are implemented on a sophisticated geotechnical problem. The computational model of a compressed air energy storage, mined in a rock salt formation, includes many input parameters, each with large amount of uncertainties. Sensitivity measures of different variables involved in the mechanical response of the cavern are computed by different global sensitivity methods, namely, Sobol/Saltelli, Random Balance Design, and Elementary Effect method. Since performing sensitivity analysis requires a large number of model evaluations, the concept of surrogate modelling is utilised to decrease the computational burden. In the following, the accuracy levels of various surrogate techniques are compared. In addition, a comparative study on the applied sensitivity analysis methods shows that the applied sensitivity analysis techniques provide identical parameter importance rankings, although some may also give more information about the system behaviour.
Within the framework of the Embankment and Footing Prediction Symposium 2016, a settlement prediction is performed for a trial embankment located in Ballina, New South Wales, Australia. The present study uses a finite element simulation to model the consolidation and creep process in order to predict the transient settlement and pore water pressure dissipation behaviour below the embankment. As the embankment was constructed on natural, soft clay, particular focus is set on the choice of an adequate constitutive model as well as on the determination of the appropriate constitutive parameters. Different features of existing constitutive models are analysed regarding the sensitivity of the response of the FE model towards the constitutive parameters. As a 3D FE simulation using discrete modelling of the employed drains in general is very time consuming, in the present study, in a first step, a validation technique replacing the discrete vertical drains by a soil body of equivalent permeability is studied. The validated approach is applied to a 2D FE model to achieve a reduced but adequate numerical model and doing so to minimise the computational effort. Finally, a probabilistic analysis is performed using Monte-Carlo simulation to identify the reliability and variability of the predicted model responses.
Performing parameter identification for model calibration prior to numerical simulation is an essential task in geotechnical engineering. However, it has to be kept in mind that the accuracy of the obtained parameter is closely related to the chosen experimental set-up, such as the number of sensors as well as their location. A well considered position of sensors can increase the quality of the measurement and reduce the number of monitoring points. This paper illustrates this concept by means of a loading device that is used to identify the stiffness and permeability factor of soft clays. With an initial set-up of the measurement devices the pore water pressure and the vertical displacements are recorded and used to identify the aforementioned parameters. Starting from these identified parameters, the optimal measurement set-up is investigated with a method based on global sensitivity analysis. This method shows an optimal sensor location assuming three sensors for each measured quantity.
In general, the remarkable amount of uncertainties involved in the material parameters is considered as an explicit feature of any geotechnical problems, due to their geological origin. Therefore, in the context of model calibration and validation, conducting a sensitivity analysis is substantial as the first step. It identifies key input factors that have the most contribution to the uncertainty of the model output. This study employs three commonly used sensitivity analysis methods on a sophisticated geotechnical problem. The simulation of the excavation and operation of a solution-mined cavity in the rock salt, used as compressed air energy deals with highly nonlinear phenomena, and includes many constitutive parameters as input. Sensitivity measures of different variables involved in the mechanical response of the cavern are computed by different global sensitivity methods, namely Sobolˊ/Saltelli, Random Balance Design, and Elementary Effect (Morris) method. An interpretation of the sensitivity indices provided by different methods is presented through a comparative study. The obtained results reveal that the applied methods provide identical parameter importance rankings, although not all of them are able to present the same information about the system behaviour
An almost unavoidable consequence of shallow tunnel constructions is settlements at the ground surface. However, a precise prediction of the settlement can only be achieved with knowledge regarding the constitutive parameters of the surrounding soil. One option is using data of surface settlement and perform an inverse analysis to identify these parameters. One aspect that is mostly neglected in this approach is the question which measurement arrangement provides the most precise results. The present work introduces the concept of the " Design of Experiment " to the field of tunnelling. Global sensitivity analysis is employed to identify which parameters are more relevant and where and when measurements should be performed to minimise the uncertainty in the identified parameter. In this regard, a synthetic example of a tunnel, approaching a shallow surface foundation eccentric from the tunnel's centreline, is investigated.