A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Application of machine learning for the generalization of the response of levees to high-water events ; Primjena strojnog učenja za generalizaciju ponašanja nasipa za obranu od poplava tijekom visokih voda
Levees are earthen structures which, along with other hydraulic and geotechnical structures, are part of larger networks designed primarily for flood protection. Extreme floods are rare events which may pass unnoticed if all goes well, or may cause disastrous consequences if any part of the flood protection system fails. Considering the high variability of geotechnical materials, it is advised that usage of full probabilistic analyses is employed, which imply being able to define a probabilistic distribution of the materials’ parameters of interest and calculating the failure in terms of probabilities. Often, geotechnical engineers have to work with a very limited scope of investigation works, which makes it hard to define the parameters probabilistically. This thesis investigates several methodologies which are applicable for the assessment of the stability of river levees with regards to the various failure mechanisms which may occur during high-water events, with the purpose of creating fragility curves and generalizing the levees behaviour in response to such events. The tools used for the analyses presented here are complex numerical analyses, coupled with statistical, probabilistic and machine learning techniques for creating predictive models and fragility curves, as well as to understand the behaviour of levees exposed to various loading situations. This thesis is presented as a compilation of this thesis’ author’s research papers published in scientific journals. The cumulative results of the thesis aim to improve levee management with regards to predicting failure and early warnings to potential failure. This is achieved through the generalization of the levees composed of highly different cross sections found throughout a whole area of interest, and predicting their behaviour based on the most important input parameters. It is found that out of over a hundred parameters required to uniquely define any complex levee section as realistically as possible with numerical models, only about a third are the ...
Application of machine learning for the generalization of the response of levees to high-water events ; Primjena strojnog učenja za generalizaciju ponašanja nasipa za obranu od poplava tijekom visokih voda
Levees are earthen structures which, along with other hydraulic and geotechnical structures, are part of larger networks designed primarily for flood protection. Extreme floods are rare events which may pass unnoticed if all goes well, or may cause disastrous consequences if any part of the flood protection system fails. Considering the high variability of geotechnical materials, it is advised that usage of full probabilistic analyses is employed, which imply being able to define a probabilistic distribution of the materials’ parameters of interest and calculating the failure in terms of probabilities. Often, geotechnical engineers have to work with a very limited scope of investigation works, which makes it hard to define the parameters probabilistically. This thesis investigates several methodologies which are applicable for the assessment of the stability of river levees with regards to the various failure mechanisms which may occur during high-water events, with the purpose of creating fragility curves and generalizing the levees behaviour in response to such events. The tools used for the analyses presented here are complex numerical analyses, coupled with statistical, probabilistic and machine learning techniques for creating predictive models and fragility curves, as well as to understand the behaviour of levees exposed to various loading situations. This thesis is presented as a compilation of this thesis’ author’s research papers published in scientific journals. The cumulative results of the thesis aim to improve levee management with regards to predicting failure and early warnings to potential failure. This is achieved through the generalization of the levees composed of highly different cross sections found throughout a whole area of interest, and predicting their behaviour based on the most important input parameters. It is found that out of over a hundred parameters required to uniquely define any complex levee section as realistically as possible with numerical models, only about a third are the ...
Application of machine learning for the generalization of the response of levees to high-water events ; Primjena strojnog učenja za generalizaciju ponašanja nasipa za obranu od poplava tijekom visokih voda
Rossi, Nicola (author) / Kovačević, Meho-Saša
2023-05-19
Theses
Electronic Resource
English
machine learning , flood protection , fragility curves , levees , strojno učenje , obrana od poplava , krivulje vjerojatnosti otkazivanja , nasipi , TEHNIČKE ZNANOSTI. Građevinarstvo , TECHNICAL SCIENCES. Civil Engineering , Građevinarstvo. Građevinsko inženjerstvo i tehnika. Građevinska tehnika kopnenog prometa (ceste , željeznice) , Civil and structural engineering. Civil engineering of land transport. Railway engineering. Highway engineering , info:eu-repo/classification/udc/624/625(043.3)