This study addresses seismic vulnerability assessment of historic masonry building stock, starting from the observation that the same seismic event can produce markedly different damage on apparently similar buildings, depending on their specific construction features. The European Macroseismic Scale EMS-98, the main reference for vulnerability classification, assigns each building typology a range of plausible vulnerability classes, but does not specify which construction factors determine a building's position toward the more or less favorable end of that range. This work investigates this interpretive gap through unsupervised statistical clustering techniques, capable of identifying homogeneous building groups based solely on field-recorded construction features, without predefined assumptions on the relevance of individual factors. The analysis was conducted on 400 historic masonry buildings damaged by the 2016 Central Italy earthquake sequence, surveyed across seven historic centers in the Umbria-Marche Apennines, considering masonry typology, quality of strengthening interventions, structural discontinuities, and morphological irregularities. Clustering, performed without any damage information, identified five internally coherent groups. Checked only afterward, their correspondence with observed damage confirmed that construction features alone effectively anticipate damage propensity. The group with lower-quality masonry and interventions showed the highest damage level, while the group with the best-performing construction technique showed the most favorable behavior. Three intermediate groups, despite differing construction combinations, showed equivalent aggregate vulnerability, indicating that intervention quality can substantially offset the intrinsic vulnerability of the base typology. A synthetic vulnerability index, computable for individual buildings, was also developed. It proved useful as a relative measure but unreliable as an absolute measure, given the non-linear interaction among factors. An individual vulnerability estimate derived directly from observed seismic behavior was therefore introduced, more reliably capturing each building's deviation from its group's typical behavior. The method was validated on an independent sample of 59 buildings: comparison between expected and observed damage yielded positive results, provided typological and contextual compatibility with the training sample. The study thus shows that statistical clustering techniques applied to field-recorded construction features can enrich seismic vulnerability assessment of historic building stock, offering a more articulated reading than traditional typological classification alone, while acknowledging limitations tied to sample specificity, addressable through future dataset expansion.
Il presente lavoro affronta la valutazione della vulnerabilità sismica del patrimonio edilizio storico in muratura, muovendo dalla considerazione che uno stesso evento sismico può produrre danni molto differenti su edifici apparentemente simili, in funzione delle loro caratteristiche costruttive specifiche. La scala macrosismica europea EMS-98, principale riferimento per la classificazione della vulnerabilità, associa a ciascuna tipologia costruttiva un intervallo di classi possibili, senza tuttavia indicare quali fattori determinino la collocazione di un edificio verso l'estremo più o meno favorevole di tale intervallo. Il lavoro propone di indagare questo scarto interpretativo mediante tecniche di segmentazione statistica non supervisionata, capaci di individuare gruppi di edifici omogenei a partire dalle sole caratteristiche costruttive rilevate sul campo, senza ipotesi predefinite sulla rilevanza dei singoli fattori. L'analisi è stata condotta su 400 edifici in muratura storica danneggiati dalla sequenza sismica del Centro Italia 2016, rilevati in sette centri storici dell'Appennino umbro-marchigiano, considerando tipologia muraria, qualità degli interventi di consolidamento, discontinuità strutturali e irregolarità morfologiche. La segmentazione, condotta senza informazioni sul danno subito, ha individuato cinque gruppi internamente coerenti. Verificata solo a posteriori, la corrispondenza con il danno osservato ha confermato che le sole caratteristiche costruttive anticipano efficacemente la propensione al danneggiamento. Il gruppo con muratura e interventi di qualità inferiore presenta il danno più elevato; quello con tecnica costruttiva più performante il comportamento più favorevole. Tre gruppi intermedi, pur con combinazioni costruttive diverse, mostrano vulnerabilità aggregata equivalente, indicando che la qualità degli interventi può compensare apprezzabilmente la vulnerabilità intrinseca della tipologia di base. È stato inoltre sviluppato un indice sintetico di vulnerabilità calcolabile per il singolo edificio, utile come misura relativa ma non affidabile come misura assoluta, data la natura non lineare dell'interazione tra i fattori. È stata quindi introdotta una stima di vulnerabilità individuale derivata dal comportamento sismico osservato, più affidabile nel descrivere lo scostamento dal comportamento tipico del gruppo di appartenenza. Il lavoro dimostra dunque che tecniche di segmentazione statistica applicate a caratteristiche costruttive rilevabili sul campo possono arricchire la valutazione della vulnerabilità sismica del patrimonio storico, offrendo una lettura più articolata rispetto alla classificazione tipologica tradizionale, pur riconoscendo limiti legati alla specificità del campione, superabili con futuri ampliamenti della base dati.
Classificazione e valutazione della vulnerabilità sismica di edifici storici mediante tecniche di clustering su dati empirici di danno: il caso Centro Italia 2016
BUTTO', MARIACHIARA
2025/2026
Abstract
This study addresses seismic vulnerability assessment of historic masonry building stock, starting from the observation that the same seismic event can produce markedly different damage on apparently similar buildings, depending on their specific construction features. The European Macroseismic Scale EMS-98, the main reference for vulnerability classification, assigns each building typology a range of plausible vulnerability classes, but does not specify which construction factors determine a building's position toward the more or less favorable end of that range. This work investigates this interpretive gap through unsupervised statistical clustering techniques, capable of identifying homogeneous building groups based solely on field-recorded construction features, without predefined assumptions on the relevance of individual factors. The analysis was conducted on 400 historic masonry buildings damaged by the 2016 Central Italy earthquake sequence, surveyed across seven historic centers in the Umbria-Marche Apennines, considering masonry typology, quality of strengthening interventions, structural discontinuities, and morphological irregularities. Clustering, performed without any damage information, identified five internally coherent groups. Checked only afterward, their correspondence with observed damage confirmed that construction features alone effectively anticipate damage propensity. The group with lower-quality masonry and interventions showed the highest damage level, while the group with the best-performing construction technique showed the most favorable behavior. Three intermediate groups, despite differing construction combinations, showed equivalent aggregate vulnerability, indicating that intervention quality can substantially offset the intrinsic vulnerability of the base typology. A synthetic vulnerability index, computable for individual buildings, was also developed. It proved useful as a relative measure but unreliable as an absolute measure, given the non-linear interaction among factors. An individual vulnerability estimate derived directly from observed seismic behavior was therefore introduced, more reliably capturing each building's deviation from its group's typical behavior. The method was validated on an independent sample of 59 buildings: comparison between expected and observed damage yielded positive results, provided typological and contextual compatibility with the training sample. The study thus shows that statistical clustering techniques applied to field-recorded construction features can enrich seismic vulnerability assessment of historic building stock, offering a more articulated reading than traditional typological classification alone, while acknowledging limitations tied to sample specificity, addressable through future dataset expansion.| File | Dimensione | Formato | |
|---|---|---|---|
|
Butto_Mariachiara.pdf
embargo fino al 14/01/2028
Dimensione
36.79 MB
Formato
Adobe PDF
|
36.79 MB | Adobe PDF |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/110580