Probabilistische Modellierung von Versagensprozessen bei Staudämmen

Dams are complex engineering structures whose failures have caused extensive consequences in the past. The internationally prevalent trend to analyze and assess risk as well as to manage the residual risk is increasingly finding its way into conceptual approaches to dam safety in Germany and now codified in the German technical standard DIN 19700 for dams and reservoirs. A comprehensive approach for assessing risk of hydraulic structures in a profound manner, named RAPID: Risk Assessment: Probability, Inundation, Damage is described in this thesis and now available at the Institute of Hydraulic Engineering and Water Resources Management at RWTH Aachen University. Following the well-accepted definition of risk in a technical context, risk is defined as the product of the probability of occurrence for an undesirable event and the resulting consequences. As consequences are within the field of (socio-)economics extensively, within the context of ecologic and psychosocial effects at least partially amenable to quantification, quantifying probabilities will allow for the mathematical calculation of risk, allowing to assess risk within an objective framework. Derived from this, the objective of the thesis is defined in the context of the quantitative determination of probabilities of failure for embankment dams. The method of the probabilistic modelling of failures is a method for the mathematical determination of probabilities of failure which allows the detailed representation of the physical basics of failure processes. This attribute and the utilization of scientifically profound as well as acknowledged conceptual calculation models make it superior to alternative methods used for quantifying or estimating the probabilities of failure. A literature study shows that probabilistic modelling approaches are available in principle. Nevertheless within these, potentially relevant failure modes are only described up to a limited level of detail. Additionally, no model which shows the capability of covering the wide range of relevant and significant hazards for embankment dams is documented in the literature. Within the thesis a new generic probabilistic model named PrEDaF (Probability of Embankment Dam Failure) is developed and coded. It comprises comprehensive and complex failure modes which are formulated as outcomes of internal and external hazards. These modes are mathematically based on analytical, empirical and numerical approaches. The four external hazards hydrological events, seismology, landslides into the reservoir and the failure of an upstream dam allow the representation of the probabilities of failure based on time as the external initiating events are all time-dependent. In the model these hazards are assigned to four main modules. Internal erosion, geostatic processes and influences from dam monitoring are included into the four main modules but considered equal in detail, process-orientation and scientific accuracy. The transformation of aleatory and epistemic uncertainties in parameters and conceptual models to probabilities of failure is performed by applying Monte-Carlo modelling techniques and related methods for variance reduction. In the course of working within the topic of internal erosion on the identification of filter criteria in order to implement these into the model it is recognized that shortcomings exist in the field of drawing sophisticated conclusions for the interaction of mineral dam cores and filter bodies, within so-called base-filter-systems. Such sophisticated conclusions are of particular interest for the purpose of probabilistic modelling of filter performance. It is found especially important to estimate the expected gradual extent of erosion of base material into adjacent filters instead of simply separating erosion from non-erosion status. As well-established filter criteria are of deterministic nature and separate into dichotomous categories, analyses using the method of logistic regression are undertaken. A data basis formed by the results of a large number of laboratory No-Erosion-Filter(NEF)-Tests is available for this purpose. The investigations result in a probabilistic filter criterion for fine-grained base soils. It subdivides the range of expected erosion of base soil and therewith allows for its inclusion into the probabilistic model PrEDaF in order to extensively investigate possible failure modes. Additionally it allows for improved filter design. The application of the developed probabilistic model PrEDaF is presented for a fictitious dam. In doing so it is also shown that the model is very well suited for evaluating the efficiency of constructive and organizational measures within the framework of reducing probabilities of failure of embankment dams. This allows for purposeful mitigation of risk