Refine
Document Type
- Conference Proceeding (15)
- Doctoral Thesis (3)
- Article (1)
Institute
- Graduiertenkolleg 1462 (19) (remove)
Keywords
- Angewandte Informatik (15)
- Angewandte Mathematik (15)
- Computerunterstütztes Verfahren (15)
- Bayes-Inferenz (1)
- Bewertung (1)
- Bewertungsmethode (1)
- Coupled models (1)
- Damm (1)
- Embankment, sensitivity analysis, parameter identification, Particle Swarm Optimization (1)
- Evaluation methods (1)
- Fehlerabschätzung (1)
- Gekoppelte Modelle (1)
- Ingenieurbau (1)
- Kriechen (1)
- Model quality, Model error estimation, Kinematical model, Geometric non-linearity, Finite Element method (1)
- Modellqualität, Modellfehlerschätzer, Geometrisch nicht-lineare Berechnung, Kinematik Modell, Finite Elemente Methode (1)
- Prognosequalität (1)
- Sensitivitätsanalyse (1)
- Unsicherheitsanalyse (1)
- bridge-vehicle interaction (1)
- random vibrations (1)
- sensitivity and uncertainty analysis (1)
Year of publication
- 2012 (19) (remove)
The process of analysis and design in structural engineering requires the consideration of different partial models, for example loading, structural materials, structural elements, and analysis types. The various partial models are combined by coupling several of their components. Due to the large number of available partial models describing similar phenomena, many different model combinations are possible to simulate the same aspects of a structure. The challenging task of an engineer is to select a model combination that ensures a sufficient, reliable prognosis. In order to achieve this reliable prognosis of the overall structural behavior, a high individual quality of the partial models and an adequate coupling of the partial models is required. Several methodologies have been proposed to evaluate the quality of partial models for their intended application, but a detailed study of the coupling quality is still lacking. This paper proposes a new approach to assess the coupling quality of partial models in a quantitative manner. The approach is based on the consistency of the coupled data and applies for uni- and bidirectional coupled partial models. Furthermore, the influence of the coupling quality on the output quantities of the partial models is considered. The functionality of the algorithm and the effect of the coupling quality are demonstrated using an example of coupled partial models in structural engineering.
Numerical simulations in the general field of civil engineering are common for the design process of structures and/or the assessment of existing buildings. The behaviour of these structures is analytically unknown and is approximated with numerical simulation methods like the Finite Element Method (FEM). Therefore the real structure is transferred into a global model (GM, e.g. concrete bridge) with a wide range of sub models (partial models PM, e.g. material modelling, creep). These partial models are coupled together to predict the behaviour of the observed structure (GM) under different conditions. The engineer needs to decide which models are suitable for computing realistically and efficiently the physical processes determining the structural behaviour. Theoretical knowledge along with the experience from prior design processes will influence this model selection decision. It is thus often a qualitative selection of different models. The goal of this paper is to present a quantitative evaluation of the global model quality according to the simulation of a bridge subject to direct loading (dead load, traffic) and indirect loading (temperature), which induce restraint effects. The model quality can be separately investigated for each partial model and also for the coupled partial models in a global structural model. Probabilistic simulations are necessary for the evaluation of these model qualities by using Uncertainty and Sensitivity Analysis. The method is applied to the simulation of a semi-integral concrete bridge with a monolithic connection between the superstructure and the piers, and elastomeric bearings at the abutments. The results show that the evaluation of global model quality is strongly dependent on the sensitivity of the considered partial models and their related quantitative prediction quality. This method is not only a relative comparison between different models, but also a quantitative representation of model quality using probabilistic simulation methods, which can support the process of model selection for numerical simulations in research and practice.
Safety operation of important civil structures such as bridges can be estimated by using fracture analysis. Since the analytical methods are not capable of solving many complicated engineering problems, numerical methods have been increasingly adopted. In this paper, a part of isotropic material which contains a crack is considered as a partial model and the proposed model quality is evaluated. EXtended IsoGeometric Analysis (XIGA) is a new developed numerical approach [1, 2] which benefits from advantages of its origins: eXtended Finite Element Method (XFEM) and IsoGeometric Analysis (IGA). It is capable of simulating crack propagation problems with no remeshing necessity and capturing singular field at the crack tip by using the crack tip enrichment functions. Also, exact representation of geometry is possible using only few elements. XIGA has also been successfully applied for fracture analysis of cracked orthotropic bodies [3] and for simulation of curved cracks [4]. XIGA applies NURBS functions for both geometry description and solution field approximation. The drawback of NURBS functions is that local refinement cannot be defined regarding that it is based on tensorproduct constructs unless multiple patches are used which has also some limitations. In this contribution, the XIGA is further developed to make the local refinement feasible by using Tspline basis functions. Adopting a recovery based error estimator in the proposed approach for evaluation of the model quality and performing the adaptive processes is in progress. Finally, some numerical examples with available analytical solutions are investigated by the developed scheme.
Non-destructive techniques for damage detection became the focus of engineering interests in the last few years. However, applying these techniques to large complex structures like civil engineering buildings still has some limitations since these types of structures are
unique and the methodologies often need a large number of specimens for reliable results. For this reason, cost and time can greatly influence the final results.
Model Assisted Probability Of Detection (MAPOD) has taken its place among the ranks of damage identification techniques, especially with advances in computer capacity and modeling tools. Nevertheless, the essential condition for a successful MAPOD is having a reliable model in advance. This condition is opening the door for model assessment and model quality problems. In this work, an approach is proposed that uses Partial Models (PM) to compute the Probability Of damage Detection (POD). A simply supported beam, that can be structurally modified and
tested under laboratory conditions, is taken as an example. The study includes both experimental and numerical investigations, the application of vibration-based damage detection approaches and a comparison of the results obtained based on tests and simulations.
Eventually, a proposal for a methodology to assess the reliability and the robustness of the models is given.
The present research analyses the error on prediction obtained under different data availability scenarios to determine which measurements contribute to an improvement of model prognosis and which not. A fully coupled 2D hydromechanical model of a water retaining dam is taken as an example. Here, the mean effective stress in the porous skeleton is reduced due to an increase in pore water pressure under drawdown conditions. Relevant model parameters are ranked by scaled sensitivities, Particle Swarm Optimization is applied to determine the optimal parameter values and model validation is performed to determine the magnitude of error forecast. We compare the predictions of the optimized models with results from a forward run of the reference model to obtain actual prediction errors.
The analyses presented here were performed to 31 data sets of 100 observations of varying data types. Calibrating with multiple information types instead of only one sort, brings better calibration results and improvement in model prognosis. However, when using several types of information the number of observations have to be increased to be able to cover a representative part of the model domain; otherwise a compromise between data availability and domain
coverage prove best. Which type of information for calibration contributes to the best prognoses, could not be determined in advance. For the error in model prognosis does not depends on the error in calibration, but on the parameter error, which unfortunately can not be determined in reality since we do not know its real value. Excellent calibration fits with parameters’ values near the limits of reasonable physical values, provided the highest prognosis errors. While models which included excess pore pressure values for calibration provided the best prognosis, independent of the calibration fit.
Long-span cable supported bridges are prone to aerodynamic instabilities caused by wind and this phenomenon is usually a major design criterion. If the wind speed exceeds the critical flutter speed of the bridge, this constitutes an Ultimate Limit State. The prediction of the flutter boundary therefore requires accurate and robust models. This paper aims at studying various combinations of models to predict the flutter phenomenon.
Since flutter is a coupling of aerodynamic forcing with a structural dynamics problem, different types and classes of models can be combined to study the interaction. Here, both numerical approaches and analytical models are utilised and coupled in different ways to assess the prediction quality of the hybrid model. Models for aerodynamic forces employed are the analytical Theodorsen expressions for the motion-enduced aerodynamic forces of a flat plate and Scanlan derivatives as a Meta model. Further, Computational Fluid Dynamics (CFD) simulations using the Vortex Particle Method (VPM) were used to cover numerical models.
The structural representations were dimensionally reduced to two degree of freedom section models calibrated from global models as well as a fully three-dimensional Finite Element (FE) model. A two degree of freedom system was analysed analytically as well as numerically.
Generally, all models were able to predict the flutter phenomenon and relatively close agreement was found for the particular bridge. In conclusion, the model choice for a given practical analysis scenario will be discussed in the context of the analysis findings.
Methods for model quality assessment are aiming to find the most appropriate model with respect to accuracy and computational effort for a structural system under investigation. Model error estimation techniques can be applied for this purpose when kinematical models are investigated. They are counted among the class of white box models, which means that the model hierarchy and therewith the best model is known. This thesis gives an overview of discretisation error estimators. Deduced from these, methods for model error estimation are presented. Their general goal is to make a prediction of the inaccuracies that are introduced using the simpler model without knowing the solution of a more complex model. This information can be used to steer an adaptive process. Techniques for linear and non-linear problems as well as global and goal-oriented errors are introduced. The estimation of the error in local quantities is realised by solving a dual problem, which serves as a weight for the primal error. So far, such techniques have mainly been applied in
material modelling and for dimensional adaptivity. Within the scope of this thesis, available model error estimators are adapted for an application to kinematical models. Their applicability is tested regarding the question of whether a geometrical non-linear calculation is necessary or not. The analysis is limited to non-linear estimators due to the structure of the underlying differential equations. These methods often involve simplification, e.g linearisations. It is investigated to which extent such assumptions lead to meaningful results, when applied to kinematical models.
Bridge vibration due to traffic loading has been a subject of extensive research in the last decades. The focus of such research has been to develop solution algorithms and investigate responses or behaviors of interest. However, proving the quality and reliability of the model output in structural engineering has become a topic of increasing importance. Therefore, this study is an attempt to extend concepts of uncertainty and sensitivity analyses to assess the dynamic response of a coupled model in bridge engineering considering time-dependent vehicular loading. A setting for the sensitivity analysis is proposed, which enables performing the sensitivity analysis considering random stochastic processes. The classical and proposed sensitivity settings are used to identify the relevant input parameters and models that have the most influence on the variance of the dynamic response. The sensitivity analysis exercises the model itself and extracts results without the need for measurements or reference solutions; however, it does not offer a means of ranking the coupled models studied. Therefore, concepts of total uncertainty are employed to rank the coupled models studied according to their fitness in describing the dynamic problem.
The proposed procedures are applied in two examples to assess the output of coupled subsystems and coupled partial models in bridge engineering considering the passage of a heavy vehicle at various speeds.
Ziel dieser Arbeit ist die Entwicklung von Methoden, mit denen die Prognosequalität von Kriechmodellen des Betons bestimmt werden kann. Die Methoden werden in zwei Ausgangsszenarien unterschieden: die Bewertung ohne und die Bewertung mit Verwendung von spezifischen Versuchsdaten zum Kriechverhalten des Betons. Die Modellqualität wird anhand der Gesamtunsicherheit der prognostizierten Kriechnachgiebigkeit quantifiziert. Die Unsicherheit wird für die Kriechprognose ohne Versuchsdaten über eine Unsicherheitsanalyse unter Berücksichtigung korrelierter Eingangsparameter ermittelt. Bei der Verwendung experimenteller Daten werden die stochastischen Eigenschaften der Modellparameter mittels Bayesian Updating bestimmt. Die Bewertung erfolgt erneut basierend auf einer Unsicherheitsanalyse sowie alternativ mittels Modellselektion nach Bayes.
Weiterhin wird eine auf Graphentheorie und Sensitivitätsanalysen basierende Methode zur Bewertung von gekoppelten Partialmodellen entwickelt. Damit wird der Einfluss eines Partialmodells auf das Verhalten einer globalen Tragstruktur quantifiziert, Interaktionen von Partialmodellen festgestellt und ein Maß für die Qualität eines Gesamtmodells ermittelt.