Within the context of vibration-based Structural Health Monitoring (SHM) for historical structures, the present Doctoral Dissertation addresses two main issues: the effective calibration of FE models – based on architectural research and Operational Modal Analysis – and the detection and localisation of structural damages from continuous dynamic monitoring. The effective calibration of FE models relies on the accurate measurements of modal parameters (i.e., natural frequencies and mode shapes), the selection of proper modelling assumptions, and the quality of the updating technique. Modelling existing structures is often challenging due to the number of uncertainties on material properties and boundary conditions. These uncertainties are even more significant for historical constructions: the distribution of mechanical parameters is generally non-homogeneous due the presence of cracks, past repairing interventions, or different construction stages. Consequently, a procedure based on architectural research and operational modal testing is implemented for selecting the appropriate modelling strategies. The developed procedure involves the following steps: (i) historical and documentary research; (ii) visual inspections and geometric survey; (iii) ambient vibration testing and identification of modal parameters; (iv) iterative FE modelling with selected assumptions and surrogate-based model updating. The second issue tackled in this Doctoral Dissertation is damage identification – with damage being defined as a loss of stiffness – for historical masonry towers and masonry arch bridges, under the assumption that only frequency data are available from the continuous monitoring system. Existing applications have demonstrated the possibility of detecting the onset of damage using only the natural frequencies (data-driven approach); however, to move from detection to localisation with a simplified sensor distribution, numerical models are mandatory (model-based approach). The methodology herein proposed is based on both approaches: the problem of damage detection is tackled from a data-driven perspective, while the localisation employs the spatial information given by a sensitivity matrix built from an updated numerical model. The procedure was firstly developed for masonry towers and then extended for masonry viaducts. Firstly, the damage detection is performed as follow: (1) a regression model is defined to remove the effects of environmental parameters on natural frequencies; (2) the variations in the residuals between the predicted and identified natural frequencies are checked using a control chart; (3) in case the adopted control metric exceeds an appropriate threshold for a certain number of times, most likely a damage occurred on the structure (detection). Subsequently, the position of the detected damage is identified (localisation) using the experimental frequency decay and a numerically computed sensitivity-matrix. A series of damage scenarios are simulated from an updated numerical model, and the consequent variations of natural frequencies are collected in a matrix called Damage Localisation Reference Matrix (DLRM). By measuring the similarity between the frequency variations in the DLRM and the ones detected from the continuous monitoring system using the cleaned observations, it is possible to locate in which area the anomaly appeared. The methodology is then applied to two classes of historical structures: the Zuccaro tower in Mantua and the Olla bridge in Gaiola have been selected as exemplifying cases for the two classes. For both structures, the natural frequencies are identified from Ambient Vibration Tests, and the unknown structural parameters are identified using the developed model updating procedure. It is worth noting that in both cases, it was crucial to include in the FE modelling stage the information coming from the architectural research, namely, visual inspections and decay patterns, as well as historical information regarding the evolution of the structure. Subsequently, pseudo-experimental monitoring data are generated using recorded temperatures. Different damage scenarios are simulated, and their identification is performed with the DLRM method. Promising results demonstrate the applicability of the developed procedure for the SHM of historical masonry towers and multi-span masonry viaducts.
Nell'ambito del monitoraggio della condizione strutturale (SHM) di costruzioni storiche, la presente tesi di dottorato affronta due questioni principali: (1) la calibrazione dei modelli numerici basata sulla ricerca architettonica e l'analisi modale operativa e (2) la rilevazione e la localizzazione dei danni strutturali dal monitoraggio dinamico continuo. L’efficace calibrazione dei modelli numerici di strutture esistenti si basa su misurazioni accurate dei parametri modali (vale a dire frequenze naturali e deformate modali), nonché sulla selezione di ipotesi di modellazione adeguate e sulla robustezza della tecnica di aggiornamento. La modellazione delle strutture esistenti è spesso complessa, a causa del numero di incertezze sulle proprietà dei materiali e sulle condizioni al contorno. Queste incertezze sono ancora più significative per le costruzioni storiche: la distribuzione dei parametri meccanici è generalmente disomogenea a causa di fenomeni fessurativi, interventi di riparazione passati o fasi costruttive diverse. Nella tesi viene sviluppata una procedura basata sulla ricerca architettonica e sulle prove di vibrazione in condizioni operative per selezionare le strategie di modellazione appropriate. La procedura sviluppata prevede le seguenti fasi: (i) ricerca storica e documentale; (ii) ispezioni visive e rilievo geometrico; (iii) prove di vibrazione ambientale e identificazione dei parametri modali; (iv) modellazione FEM ottimizzata mediante la selezione iterativa di ipotesi e la calibrazione dei parametri incerti attraverso l’uso dei modelli surrogati. Il secondo problema affrontato in questa tesi è l'identificazione del danno, definito come una perdita di rigidezza, per torri storiche e ponti ad arco in muratura, nell'ipotesi di utilizzare solo i dati relativi all’evoluzione delle frequenze naturali. Le applicazioni esistenti hanno dimostrato la possibilità di rilevare l'insorgenza del danno utilizzando solo le frequenze naturali (approccio data-driven); tuttavia, per passare dalla rilevazione alla localizzazione con una distribuzione dei sensori semplificata, sono necessari modelli numerici (approccio model-based). La metodologia quivi proposta si basa su entrambi gli approcci: il problema dell’identificazione del danno viene affrontato in una prospettiva data-driven, mentre la localizzazione utilizza l'informazione spaziale data da una matrice di sensibilità costruita a partire da un modello numerico aggiornato. La procedura è stata dapprima sviluppata per le torri in muratura e poi estesa per i viadotti in muratura. In primo luogo, il rilevamento del danno viene eseguito come segue: (1) viene definito un modello di regressione per rimuovere gli effetti dei parametri ambientali dalle frequenze naturali; (2) le variazioni dei residui tra le frequenze naturali predette e individuate sono verificate mediante una carta di controllo; (3) nel caso in cui la metrica di controllo adottata superi una soglia opportuna per un certo numero di volte, molto probabilmente si è verificato un danno sulla struttura (rilevamento). Successivamente, la posizione del danno rilevato viene identificata (localizzazione) utilizzando la diminuzione delle frequenze sperimentali e una matrice di sensibilità calcolata numericamente. In base ad un modello numerico aggiornato (della struttura integra) viene simulata una serie di scenari di danno e le conseguenti variazioni delle frequenze naturali vengono raccolte in una matrice denominata, con locuzione anglosassone, Damage Localization Reference Matrix (DLRM). Misurando la somiglianza tra le variazioni di frequenza nella DLRM e quelle rilevate dal sistema di monitoraggio continuo utilizzando le osservazioni ripulite dalle interferenze ambientali, è possibile individuare in quale area si è manifestata l'anomalia. La metodologia viene poi applicata a due classi di strutture storiche: la torre dello Zuccaro a Mantova e il ponte dell’Olla a Gaiola sono stati scelti come casi esemplificativi per le due classi. Per entrambe le strutture, le frequenze naturali sono identificate da test di vibrazioni ambientali e i parametri strutturali incerti sono identificati utilizzando la procedura di aggiornamento del modello sviluppata. Vale la pena notare che in entrambi i casi è stato fondamentale includere nella fase di modellazione FEM le informazioni provenienti dalla ricerca architettonica, vale a dire ispezioni visive e rilievo del degrado, nonché informazioni storiche sull'evoluzione della struttura. Successivamente, vengono generati dati di monitoraggio pseudo-sperimentali utilizzando le temperature registrate da una stazione meteo vicina. Vengono simulati diversi scenari di danno e la loro identificazione viene eseguita con il metodo DLRM proposto. Risultati promettenti dimostrano l'applicabilità della procedura sviluppata per l'SHM di torri in muratura storiche e viadotti in muratura a più campate.
Vibration-based FE model updating and damage identification for historical structures
Borlenghi, Paolo
2020/2021
Abstract
Within the context of vibration-based Structural Health Monitoring (SHM) for historical structures, the present Doctoral Dissertation addresses two main issues: the effective calibration of FE models – based on architectural research and Operational Modal Analysis – and the detection and localisation of structural damages from continuous dynamic monitoring. The effective calibration of FE models relies on the accurate measurements of modal parameters (i.e., natural frequencies and mode shapes), the selection of proper modelling assumptions, and the quality of the updating technique. Modelling existing structures is often challenging due to the number of uncertainties on material properties and boundary conditions. These uncertainties are even more significant for historical constructions: the distribution of mechanical parameters is generally non-homogeneous due the presence of cracks, past repairing interventions, or different construction stages. Consequently, a procedure based on architectural research and operational modal testing is implemented for selecting the appropriate modelling strategies. The developed procedure involves the following steps: (i) historical and documentary research; (ii) visual inspections and geometric survey; (iii) ambient vibration testing and identification of modal parameters; (iv) iterative FE modelling with selected assumptions and surrogate-based model updating. The second issue tackled in this Doctoral Dissertation is damage identification – with damage being defined as a loss of stiffness – for historical masonry towers and masonry arch bridges, under the assumption that only frequency data are available from the continuous monitoring system. Existing applications have demonstrated the possibility of detecting the onset of damage using only the natural frequencies (data-driven approach); however, to move from detection to localisation with a simplified sensor distribution, numerical models are mandatory (model-based approach). The methodology herein proposed is based on both approaches: the problem of damage detection is tackled from a data-driven perspective, while the localisation employs the spatial information given by a sensitivity matrix built from an updated numerical model. The procedure was firstly developed for masonry towers and then extended for masonry viaducts. Firstly, the damage detection is performed as follow: (1) a regression model is defined to remove the effects of environmental parameters on natural frequencies; (2) the variations in the residuals between the predicted and identified natural frequencies are checked using a control chart; (3) in case the adopted control metric exceeds an appropriate threshold for a certain number of times, most likely a damage occurred on the structure (detection). Subsequently, the position of the detected damage is identified (localisation) using the experimental frequency decay and a numerically computed sensitivity-matrix. A series of damage scenarios are simulated from an updated numerical model, and the consequent variations of natural frequencies are collected in a matrix called Damage Localisation Reference Matrix (DLRM). By measuring the similarity between the frequency variations in the DLRM and the ones detected from the continuous monitoring system using the cleaned observations, it is possible to locate in which area the anomaly appeared. The methodology is then applied to two classes of historical structures: the Zuccaro tower in Mantua and the Olla bridge in Gaiola have been selected as exemplifying cases for the two classes. For both structures, the natural frequencies are identified from Ambient Vibration Tests, and the unknown structural parameters are identified using the developed model updating procedure. It is worth noting that in both cases, it was crucial to include in the FE modelling stage the information coming from the architectural research, namely, visual inspections and decay patterns, as well as historical information regarding the evolution of the structure. Subsequently, pseudo-experimental monitoring data are generated using recorded temperatures. Different damage scenarios are simulated, and their identification is performed with the DLRM method. Promising results demonstrate the applicability of the developed procedure for the SHM of historical masonry towers and multi-span masonry viaducts.File | Dimensione | Formato | |
---|---|---|---|
BORLENGHI_PhD Thesis 2021.pdf
solo utenti autorizzati dal 06/06/2022
Descrizione: PhD Thesis
Dimensione
15.55 MB
Formato
Adobe PDF
|
15.55 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/177088