Advanced Bayesian networks for reliability and risk analysis in geotechnical engineering

Show simple item record

dc.contributor.advisor Beer, Michael DE He, Longxue ger 2020-03-11T08:46:40Z 2020-03-11T08:46:40Z 2020
dc.identifier.citation He, Longxue: Advanced Bayesian networks for reliability and risk analysis in geotechnical engineering. Hannover : Gottfried Wilhelm Leibniz Universität, Diss., 2020, viii, 111 S. DOI: ger
dc.description.abstract The stability and deformation problems of soil have been a research topic of great concern since the past decades. The potential catastrophic events are induced by various complex factors, such as uncertain geotechnical conditions, external environment, and anthropogenic influence, etc. To prevent the occurrence of disasters in geotechnical engineering, the main purpose of this study is to enhance the Bayesian networks (BNs) model for quantifying the uncertainty and predicting the risk level in solving the geotechnical problems. The advanced BNs model is effective for analyzing the geotechnical problems in the poor data environment. The advanced BNs approach proposed in this study is applied to solve the stability of soil slopes problem associated with the specific-site data. When probabilistic models for soil properties are adopted, enhanced BNs approach was adopted to cope with continuous input parameters. On the other hand, Credal networks (CNs), developed on the basis of BNs, are specially used for incomplete input information. In addition, the probabilities of slope failure are also investigated for different evidences. A discretization approach for the enhanced BNs is applied in the case of evidence entering into the continuous nodes. Two examples implemented are to demonstrate the feasibility and predictive effectiveness of the BNs model. The results indicate the enhanced BNs show a precisely low risk for the slope studied. Unlike the BNs, the results of CNs are presented with bounds. The comparison of three different input information reveals the more imprecision in input, the more uncertainty in output. Both of them can provide the useful disaster-induced information for decision-makers. According to the information updating in the models, the position of the water table shows a significant role in the slope failure, which is controlled by the drainage states. Also, it discusses how the different types of BNs contribute to assessing the reliability and risk of real slopes, and how new information could be introduced in the analysis. The proposed models in this study illustrate the advanced BN model is a good diagnosis tool for estimating the risk level of the slope failure. In a follow-up study, the BNs model is developed based on its potential capability for the information updating and importance measure. To reduce the influence of uncertainty, with the proposed BN model, the soil parameters are updated accurately during the excavation process, and besides, the contribution of epistemic uncertainty from geotechnical parameters to the potential disaster can be characterized based on the developed BN model. The results of this study indicate the BNs model is an effective and flexible tool for risk analysis and decision making support in geotechnical engineering. ger
dc.language.iso eng ger
dc.publisher Hannover : Institutionelles Repositorium der Leibniz Universität Hannover
dc.rights CC BY 3.0 DE ger
dc.rights.uri ger
dc.subject Bayesian networks eng
dc.subject imprecise probability eng
dc.subject uncertainty eng
dc.subject stochastic model updating eng
dc.subject sensitivity eng
dc.subject Bayesianische Netzwerke ger
dc.subject ungenaue Wahrscheinlichkeit ger
dc.subject Unsicherheit ger
dc.subject stochastische Modellaktualisierung ger
dc.subject Sensitivitätsanalyse ger
dc.subject.ddc 624 | Ingenieurbau und Umwelttechnik ger
dc.title Advanced Bayesian networks for reliability and risk analysis in geotechnical engineering eng
dc.type doctoralThesis ger
dc.type Text ger
dc.description.version publishedVersion ger
tib.accessRights frei zug�nglich ger

Files in this item

This item appears in the following Collection(s):

Show simple item record


Search the repository


My Account

Usage Statistics