Seismic risk assessment is a major challenge in countries with a significant seismic hazard and a highly vulnerable built heritage, such as Italy. Especially when the risk needs to be analyzed at a large scale, the assessment can entail very time-consuming and costly studies, since it is necessary to define numerous variables that can influence seismic exposure and vulnerability. In this thesis, a mechanically based seismic fragility model has been developed for Italian masonry residential buildings. This model is based on the classification of the building stock into macro-typologies, also considering possible seismic retrofit measures. Also exposure needs to be properly assessed: artificial intelligence techniques can be helpful to evaluate it in a quick and efficient way. Specifically, satellite images are used to automatically collect building data, street view photos are extracted for each building and Convolutional Neural Networks are trained to recognize specific features of interest from pictures, particularly the same ones on which the vulnerability model is based. The following step of this work consists in combining vulnerability and exposure with the seismic hazard within a seismic risk calculation platform that can evaluate seismic damage and risk, expressed as repair or reconstruction costs, number of unusable buildings, casualties, and displaced people. This information can be important for carrying out targeted investigations and establishing priority criteria for seismic retrofit measures. These seismic risk prevention and mitigation tools can be used by emergency authorities to manage resources in the pre- and post-earthquake phases, as well as to select effective emergency response and recovery plans.
La valutazione del rischio sismico rappresenta una sfida in Paesi con una significativa pericolosità sismica e un patrimonio edilizio vulnerabile come l'Italia. Soprattutto quando il rischio deve essere analizzato su larga scala, la sua stima può comportare analisi costose in termini economici e di tempo, a causa della necessità di definire numerose variabili che possono influenzare l'esposizione e la vulnerabilità. In questo lavoro è stato sviluppato un modello di fragilità su base meccanica per gli edifici residenziali italiani in muratura. Il modello si basa sulla classificazione del patrimonio edilizio in macro-tipologie, simulando inoltre la possibile presenza di interventi anti-sismici. Anche l'esposizione deve essere correttamente valutata: tecniche di intelligenza artificiale possono rivelarsi utili per effettuare stime in modo rapido ed efficiente. In particolare, le immagini satellitari possono essere utilizzate per raccogliere automaticamente i dati degli edifici, per poi estrarre foto street view per ogni edificio e allenare reti neurali convoluzionali a riconoscere specifiche caratteristiche di interesse dalle immagini, in particolare le stesse su cui si basa il modello di vulnerabilità. La fase successiva di questo lavoro consiste nel combinare la vulnerabilità e l'esposizione con la pericolosità sismica all'interno di una piattaforma di calcolo in grado di valutare il rischio sismico, espresso come costo di riparazione, numero di edifici inutilizzabili, vittime e sfollati. Queste informazioni sono fondamentali per condurre indagini mirate e stabilire criteri di priorità per le misure di adeguamento sismico. Gli strumenti proposti possono essere utilizzati per gestire risorse nelle fasi pre e post-terremoto e per elaborare piani di recupero efficaci.
SEISMIC RISK ASSESSMENT AT A TERRITORIAL SCALE BASED ON MACHINE LEARNING / Carpanese, Pietro. - (2023 Feb 16).
SEISMIC RISK ASSESSMENT AT A TERRITORIAL SCALE BASED ON MACHINE LEARNING
CARPANESE, PIETRO
2023
Abstract
Seismic risk assessment is a major challenge in countries with a significant seismic hazard and a highly vulnerable built heritage, such as Italy. Especially when the risk needs to be analyzed at a large scale, the assessment can entail very time-consuming and costly studies, since it is necessary to define numerous variables that can influence seismic exposure and vulnerability. In this thesis, a mechanically based seismic fragility model has been developed for Italian masonry residential buildings. This model is based on the classification of the building stock into macro-typologies, also considering possible seismic retrofit measures. Also exposure needs to be properly assessed: artificial intelligence techniques can be helpful to evaluate it in a quick and efficient way. Specifically, satellite images are used to automatically collect building data, street view photos are extracted for each building and Convolutional Neural Networks are trained to recognize specific features of interest from pictures, particularly the same ones on which the vulnerability model is based. The following step of this work consists in combining vulnerability and exposure with the seismic hazard within a seismic risk calculation platform that can evaluate seismic damage and risk, expressed as repair or reconstruction costs, number of unusable buildings, casualties, and displaced people. This information can be important for carrying out targeted investigations and establishing priority criteria for seismic retrofit measures. These seismic risk prevention and mitigation tools can be used by emergency authorities to manage resources in the pre- and post-earthquake phases, as well as to select effective emergency response and recovery plans.File | Dimensione | Formato | |
---|---|---|---|
tesi_definitiva_Pietro_Carpanese.pdf
Open Access dal 17/02/2024
Descrizione: tesi_definitiva_Pietro_Carpanese
Tipologia:
Tesi di dottorato
Licenza:
Altro
Dimensione
17.22 MB
Formato
Adobe PDF
|
17.22 MB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.