This thesis contributes to the current state of knowledge on the mechanical characterization and numerical simulation of intact and fractured laminated glass using methodological concepts of 'artificial intelligence'. The findings obtained within this work are embedded in a more comprehensive numerical-methodical simulation approach which pursues the goal of realistically mapping the time- and temperature-dependent load-bearing behaviour of glass laminates in the intact and fractured state. In order to achieve the aforementioned goal of the realization of a numerical-methodical simulation approach for laminated glass in the intact and fractured state in near future, one part of this thesis deals with theoretical, methodical, numerical and experimental investigations on the influence of the thermomechanics of polymeric interlayers on the load-bearing behaviour of laminated glass in the intact and fractured state. A special focus is on a methodical identification of the parameters of constitutive models for the polymeric interlayers in the context of artificial intelligence and machine learning. The incorporation of these modern methods allows on the one hand a stringent quantification of uncertainties in the treatment of the inverse problem of parameter interference from measurement data with simultaneous knowledge of the model evidence or lower limits thereof. Furthermore, the formal procedure of material parameter identification as presented in this thesis can be transferred to any other polymer without loss of generality. Following the Bayesian paradigm, experimental investigations on the interlayer itself or on laminated glass panes are combined with a priori information on chemical, rheological and other physical / thermomechanical properties of the polymeric interlayers or the glass laminate for posterior information inference. This part of the work focuses on the following aspects: investigations on the experimental determination of the parameters of models of linear viscoelasticity for a standard polyvinyl butyral (PVB) and an ethylene-vinyl-acetate (EVA) interlayer polymer in large- and small-scale tests; investigation of the thermorheology of a standard PVB and an EVA interlayer polymer; investigation of the constitutive modelling of linear viscoelasticity for a standard PVB and an EVA interlayer polymer; investigations on the constitutive modelling of the hyperelasticity for a standard PVB and an EVA interlayer polymer; development of practical and normative guidelines and design aids from the findings of this work. The theoretical deficits of known test-evaluation methods encountered in the assessment of the previously described experiments are investigated and overcome by the approaches and models developed within this thesis. For the modelling of the linear-viscoelastic material behaviour of the interlayer polymers in the intact state of the glass laminate the method 'GUSTL' provides a new and fast method for the creation of so-called 'Master Curves' out of data from Dynamic-Mechanical-Thermal-Analysis (DMTA) and the respective model parameter identification. GUSTL' allows an evaluation of the suitability of the determined Prony series for the given data as well as the uncertainty quantification for predictions with the calibrated model using methods of Bayesian machine learning. Further investigations deal with the modelling of the hyperelastic behaviour of interlayer polymers for the fractured state of the glass laminate and evaluate the suitability of the material models for describing the experimental data using Bayesian machine learning methods. These investigations support the modelling for the case of the interlayer being subject to large strains in the fractured state of the glass laminate. The fast and realistic simulation of the fracture pattern of thermally pre-stressed glass is a further building block for the realization of a numerical-methodical simulation approach for laminated glass in intact and fractured state. In the context of this work, a stochastic model called 'BREAK' for the prediction of the fracture pattern of thermally pre-stressed glass is motivated and calibrated with experimental data as a contribution to the development of this computational model for the assessment of the residual load-bearing capacity of glass constructions with the help of artificial intelligence. The use of Voronoi tessellations induced by stochastic point processes represents a methodologically new approach to the prediction of the fracture pattern of thermally pre-stressed glass. The methodological approach for the simulation and prediction of the fracture structure presented and implemented within this thesis is generally valid and can thus in principle be transferred to other brittle materials. Finally, design supporting materials for commonly used polymeric interlayers for the use in engineering practice are provided in form of tables and graphs. Furthermore, the results of this work are provided in the form of instructions for the performance and evaluation of experiments for the determination of thermorheological material properties as well as the deduction of the parameters of constitutive laws for laminated glass interlayer polymers. A proposal for the standardisation of the experimental and methodological procedure for determining the thermomechanical properties of laminated glass with polymeric interlayers finishes the work. In the outlook, certain points of this work as well as additional aspects of the numerical-methodical simulation approach for future research needs in the field of modelling and analysing glass laminates with polymeric interlayers in the intact and post-fracture state are highlighted. In particular, this thesis contributes to the set-up of a numerical alternative to the time- and cost-intensive component tests for the characterization of laminated glass in the intact and post-fracture state.
«
Login
BibTeX