Structural health monitoring (SHM), whether it relates to mechanical, aerospace or civil structures, is a discipline that, after three decades of continuous evolution, offers ready-to-use solutions, which ensure great added value to the structures on which they are adopted and, concurrently, represents a research topic drawing the attention from a great many different fields including data acquisition and processing, statistical and physical modeling, structural dynamics and control system, smart materials. SHM basically consists in implementing a damage detection strategy for a structure in the attempt of preventing failures potentially resulting in life-safety and economic losses. Among the panorama of SHM applications, great interest is shown in the development of solutions specifically conceived for civil structures. As a matter of fact the development of effective SHM strategies would represent an extremely appealing answer to the pressing problem of assessing the actual health conditions of infrastructures that are approaching, or have already gone beyond, the end of their design service life. Even further, in the event that extreme and unexpected weather, as well as landslides and earthquake strikes cause damage to civil structures, SHM could be used for rapid condition screening. This thesis specifically builds on the field of continuative and long-term SHM on large civil structures. A distinctive trait of structures like high rising buildings, bridges, dams, and stadia, to name a few, is represented, especially in presence of audacious architectural projects, by the coexistence of different substructures each of which exhibits its own features. In order to increase the robustness and reliability of SHM strategies conceived for such complex structures, an increasing attention is paid to adopting an approach, called sensor fusion, which consists in integrating data from sensors of various nature. Noticing that a damage, for its own nature, is a local phenomenon, the implementation of a damage detection strategy can basically follow two different approaches. From the one hand, it is possible to assess the presence of a damage by capturing a macroscopic change in the behavior of the entire structure. On the other hand, the onset and ongoing of the damage can be directly inferred by observing a physical phenomenon that locally arises as a consequence of the presence of damage. Among the different techniques contemplated by the SHM community, this work focuses in particular on a couple of them particularly promising to develop a monitoring strategy operating on a whole structure-level. Namely these techniques are operational modal analysis (OMA) and acoustic emission (AE). The former belongs to the group of the so-called vibrating-based approaches and basically consists in assessing the health condition of the structure by continuously monitoring over the long-term period the evolution of structure dynamic behavior as result of the excitation provided by environmental load (i.e. vehicular traffic and wind). The latter aims at capturing short-time, transient events which arise due to the onset or the ongoing of a damage and propagate through the structure as a wave that can be sensed within a certain distance from the source. The sensorization of the structure and the data acquisition and management that are needed when such approaches become part of a SHM strategy pose a variety of challenging issues. With regard to the vibration-based methods, the current work specifically addresses the problems related to the extraction, from the structure measured response, of those parameters providing a synthetic description of its dynamic behavior. These parameters (namely eigenfrequency, non-dimensional damping ratio and modal constants which are referred to as modal parameters) are estimated by mean of over-refined modal identification techniques. The current work goes into depth in the analysis of an OMA algorithm particularly suited for the extraction of modal parameters from structures exhibiting close-in-frequency and highly damped modes as it is the typical case when large civil structures are investigated. The merit of the first part of this work consists in the development of a general purpose Monte Carlo-based simulation method conceived to assess the result intrinsic dispersion and bias effects of a generic OMA algorithm. Spread assessment is clearly a key information towards any attempt to exploit SHM methods based on modal parameter changes; bias assessment, on the other hand, offers a precious information to critically interpret the modal parameters estimates. With regard to the continuative long-term AE-based SHM strategies, the current work deals with the problems arising from the need to size short-transient events over extended period of time. The challenge associated with capturing these events using classical techniques is that very high sampling rates must be used over extended periods of time. The result is that a very large amount of data is collected to capture a phenomenon that rarely occurs. Furthermore, the high energy consumption associated with the required high sampling rates makes the implementation of high-endurance, low-power, embedded AE sensor nodes difficult to achieve. The merit of the second part of this thesis consists in investigating the suitability of a new sampling technique, called compressed sensing (CS), that, by collecting a small amount of compressed measurements, enables to relax the requirements on the sampling rate and memory demands. The CS technique substantially lends itself as a highly promising approach to promote the transition to SHM strategies relying upon wireless sensor networks featuring minimal installation costs. The research activity carried out within the framework of this thesis is supported by an extensive experimental activity. In particular, the Giuseppe Meazza stadium in Milan is used as test case. Such a structure, due to its specificities, offers the opportunity to investigate the suitability of a monitoring solution relying upon both OMA- and AE-based SHM approaches. The Monte Carlo-based simulation framework developed in this work, thanks to its close adherence to reality, allows to study in depth what are the performances of the OMA-based algorithm of interest in extracting the modal parameters of a given structure where is installed a specific measurement setup. Thanks to the developed methodology, the Monte Carlo method allows to test the OMA algorithm on numerical responses exhibiting the presence of those same modes close in frequency and affected by large modal overlap that are featured by the responses collected on a 3rd ring Meazza stadium grandstand where a permanent monitoring system is operating. As practical outcome, the developed methodology has allowed to deeply test the performance of the OMA algorithm used for the analysis of the data gathered on the real structure and tune it for an increased effectiveness of the analysis over the long-term period. The second part of the work, on the other hand, investigates the suitability of a monitoring strategy for the Meazza stadium roof based on the detection of AE events. On the basis of the data gathered during some preliminary tests conceived to assess the transmissibility of AE events through the stadium roof, CS has demonstrated its suitability to size the AE events of interest by exploiting a number of compressed measurements roughly equal to 25% of samples required by conventional sapling techniques. Ultimately, the general purpose methodology developed to assess bias and dispersion intrinsic to a generic OMA algorithm, as well as, the investigations on the innovative compressed sensing technique make this work a valuable contribution to cope with the challenging issues SHM on large civil structures poses.
Il termine Structural Health Monitoring (SHM), sia che riguardi l’ambito di strutture meccaniche, aerospaziali o civili, si riferisce a una disciplina che, dopo tre decenni di continua evoluzione, offre soluzioni di pratico impiego e di grande valore aggiunto per le strutture sulle quali tali soluzioni sono adottate e, al contempo, rappresenta un argomento di ricerca capace di attrarre l’attenzione da campi i più disparati quali acquisizione dati e processamento dei segnali, modellazione statistica e fisica, dinamica strutturale e controllo dei sistemi, materiali intelligenti. Fondamentalmente lo SHM consiste nell’implementare una strategia di identificazione del danno su una struttura nel tentativo di evitare cedimenti potenzialmente pericolosi e tali da causare perdite economiche. Nel panorama delle applicazioni di SHM, grande interesse è rivolto allo sviluppo di soluzioni specifiche per strutture civili. Si osserva infatti come lo sviluppo di strategie di monitoraggio efficaci possa rappresentare una risposta particolarmente attraente al pressante problema di poter valutare le reali condizioni di salute di infrastrutture che stanno raggiungendo, o sono già andate oltre, la fine della loro vita di progetto. Ancor più, nel caso in cui condizioni atmosferiche eccezionali e impreviste così come frane e terremoti causino danni a strutture civili, lo SHM si costituirebbe come strumento privilegiato per valutare rapidamente le condizioni di salute di tali strutture. Il lavoro di tesi svolto è nello specifico incentrato sul monitoraggio continuativo e di lungo periodo di strutture civili di grandi dimensioni. Un tratto distintivo di strutture come grattacieli, ponti, dighe e stadi, per nominarne alcune, è rappresentato, specialmente in presenza di audaci progetti architettonici, dalla coesistenza di differenti sottostrutture ognuna delle quali presenta caratteri propri distintivi. Al fine di aumentare la robustezza e l’affidabilità di strategie di monitoraggio concepite per tali strutture complesse, un’attenzione crescente è rivolta ad adottare un approccio, chiamato sensor fusion, che consiste in integrare dati da sensori di vario tipo. Notando come un danno, per sua stessa natura, sia un fenomeno locale, l’implementazione di una strategia di monitoraggio può fondamentalmente seguire due differenti approcci. Da un lato è possibile riconoscere la presenza del danno catturando un cambiamento macroscopico nel comportamento dell’intera struttura. Dall’altro, la nascita e la propagazione del danno può essere direttamente dedotta osservando un fenomeno fisico che localmente si manifesta come conseguenza della presenza del danno stesso. Tra le differenti tecniche contemplate dalla comunità scientifica dedita allo SHM, questo lavoro si focalizza in particolare su un paio di esse ritenute particolarmente promettenti per sviluppare una strategia di monitoraggio operante sulla struttura intesa nella sua totalità. Nello specifico queste tecniche sono l’analisi modale operazionale (operational modal analysis, OMA) e l’emissione acustica (acoustic emission, AE). La prima appartiene al gruppo di tecniche vibrazionali e fondamentalmente consiste in valutare le condizioni di salute della struttura monitorandone in modo continuativo e sul lungo periodo l’evoluzione del comportamento dinamico a fronte del forzamento naturale (vento e traffico veicolare sono un esempio). La seconda mira a catturare eventi di brevissima durata e di natura transitoria che si originano come conseguenza della nascita o dell’evoluzione di un danno e propagano attraverso la struttura sotto forma di un’onda che può essere catturata entro una certa distanza dalla sorgente. La sensorizzazione della struttura, l’acquisizione e gestione del dato richiesti quando i due approcci presi in esame sono impiegati in una strategia di monitoraggio pongono problematiche impegnative. In riferimento ai metodi vibrazionali, il presente lavoro di tesi affronta nello specifico i problemi legati all’estrazione dalla risposta della struttura di una serie di parametri che permettono di offrire una descrizione sintetica del suo comportamento dinamico. Tali parametri (frequenze proprie, smorzamenti adimensionali e costanti modali a cui si fa riferimento con il termine di parametri modali) sono stimati tramite sofisticate tecniche di analisi modale. Il presente lavoro approfondisce un algoritmo OMA particolarmente adatto per l’estrazione dei parametri modali da strutture caratterizzate da modi vicini in frequenza e molto smorzati, come è tipicamente il caso in presenza di grandi strutture civili. Il merito della prima parte del lavoro svolto consiste nello sviluppo di una metodologia basata su simulazioni Monte Carlo per valutare la dispersione intrinseca e gli effetti di bias di un qualsiasi algoritmo OMA. La dispersione rappresenta un’informazione chiave per poter sfruttare metodi di monitoraggio basati sullo studio del cambiamento dei parametri modali; la valutazione del bias, d’altro canto, offre una preziosa informazione per interpretare criticamente le stime dei parametri modali. In riferimento alle strategie di monitoraggio continuative e di lungo periodo basate sull’emissione acustica, il presente lavoro affronta i problemi che nascono dal bisogno di catturare eventi di natura transiente e di breve durata su un orizzonte temporale esteso. La sfida legata all’acquisizione di tali segnali usando tecniche tradizionali è dovuta al fatto che devono essere usate frequenze di campionamento elevate per periodi di tempo prolungati. Il risultato è che un’elevata quantità di dati viene raccolta per catturare un evento che raramente si verifica. Inoltre, l’elevato consumo energetico associato alle elevate frequenze di campionamento richieste rende difficilmente realizzabili nodi di misura di tipo embedded, a basso consumo energetico e di lunga durata. Il merito della seconda parte del presente lavoro di tesi consiste nell’investigare l’adeguatezza di una tecnica di campionamento di recente introduzione, chiamata campionamento compresso (CS) che, raccogliendo un numero limitato di campioni compressi, permette di ridurre i requisiti del sistema di monitoraggio in termini di frequenza di campionamento e dimensione in memoria del dato acquisito. La tecnica del CS sostanzialmente si presta come un approccio molto promettente per promuovere la transizione a strategie di monitoraggio basate su reti di sensori wireless contraddistinte da costi di installazione minimi. L’attività di ricerca condotta all’interno di questo lavoro di tesi è supportata da un’estensiva attività sperimentale. In particolare, lo stadio Giuseppe Meazza in Milano è usato come test case. Tale struttura, grazie alle sue specificità, offre l’opportunità di investigare l’adeguatezza di strategie di monitoraggio basate sia su tecniche OMA che di AE. La metodologia basata su simulazioni Monte Carlo sviluppata in questo lavoro, grazie alla sua stretta aderenza alla realtà, permette di studiare in dettaglio le prestazioni di un algoritmo OMA ritenuto di interesse per estrarre i parametri modali da una data struttura dove è installato uno specifico set-up di misura. La metodologia sviluppata permette di testare l’algoritmo OMA di interesse su risposte numeriche caratterizzate dalla presenza dei medesimi modi vicini in frequenza e parzialmente sovrapposti che contraddistinguono le risposte di una delle tribune del 3° anello dello stadio dove è operativo un sistema di monitoraggio permanente. Come risultato pratico, la metodologia sviluppata ha consentito di testare in profondità l’algoritmo OMA usato per l’analisi dei dati raccolti sulla struttura reale e settarlo in modo da accrescere l’efficacia dell’analisi di lungo periodo. La seconda parte del lavoro, d’altra parte, studia la possibilità di realizzare una strategia di monitoraggio per la copertura dello stadio Meazza basata sulla rilevazione di eventi di tipo AE. Sulla base dei dati raccolti durante alcuni test preliminari condotti al fine di valutare la trasmissibilità di eventi di AE attraverso la copertura dello stadio, la tecnica del campionamento compresso ha dimostrato la sua efficacia nel catturare eventi di tipo AE di interesse sfruttando un numero di campioni compressi circa pari al 25% dei campioni richiesti da tecniche di campionamento tradizionali. In ultima analisi, la metodologia sviluppata per quantificare il bias e la dispersioni intrinsechi a un generico algoritmo OMA, così come gli studi condotti sull’innovativa tecnica del campionamento compresso fanno di questo lavoro di tesi un interessante contributo per affrontare le problematiche legate al monitoraggio strutturale di grandi strutture civili.
Sensor fusion and data analysis for structural health monitoring
CATTANEO, ALESSANDRO
Abstract
Structural health monitoring (SHM), whether it relates to mechanical, aerospace or civil structures, is a discipline that, after three decades of continuous evolution, offers ready-to-use solutions, which ensure great added value to the structures on which they are adopted and, concurrently, represents a research topic drawing the attention from a great many different fields including data acquisition and processing, statistical and physical modeling, structural dynamics and control system, smart materials. SHM basically consists in implementing a damage detection strategy for a structure in the attempt of preventing failures potentially resulting in life-safety and economic losses. Among the panorama of SHM applications, great interest is shown in the development of solutions specifically conceived for civil structures. As a matter of fact the development of effective SHM strategies would represent an extremely appealing answer to the pressing problem of assessing the actual health conditions of infrastructures that are approaching, or have already gone beyond, the end of their design service life. Even further, in the event that extreme and unexpected weather, as well as landslides and earthquake strikes cause damage to civil structures, SHM could be used for rapid condition screening. This thesis specifically builds on the field of continuative and long-term SHM on large civil structures. A distinctive trait of structures like high rising buildings, bridges, dams, and stadia, to name a few, is represented, especially in presence of audacious architectural projects, by the coexistence of different substructures each of which exhibits its own features. In order to increase the robustness and reliability of SHM strategies conceived for such complex structures, an increasing attention is paid to adopting an approach, called sensor fusion, which consists in integrating data from sensors of various nature. Noticing that a damage, for its own nature, is a local phenomenon, the implementation of a damage detection strategy can basically follow two different approaches. From the one hand, it is possible to assess the presence of a damage by capturing a macroscopic change in the behavior of the entire structure. On the other hand, the onset and ongoing of the damage can be directly inferred by observing a physical phenomenon that locally arises as a consequence of the presence of damage. Among the different techniques contemplated by the SHM community, this work focuses in particular on a couple of them particularly promising to develop a monitoring strategy operating on a whole structure-level. Namely these techniques are operational modal analysis (OMA) and acoustic emission (AE). The former belongs to the group of the so-called vibrating-based approaches and basically consists in assessing the health condition of the structure by continuously monitoring over the long-term period the evolution of structure dynamic behavior as result of the excitation provided by environmental load (i.e. vehicular traffic and wind). The latter aims at capturing short-time, transient events which arise due to the onset or the ongoing of a damage and propagate through the structure as a wave that can be sensed within a certain distance from the source. The sensorization of the structure and the data acquisition and management that are needed when such approaches become part of a SHM strategy pose a variety of challenging issues. With regard to the vibration-based methods, the current work specifically addresses the problems related to the extraction, from the structure measured response, of those parameters providing a synthetic description of its dynamic behavior. These parameters (namely eigenfrequency, non-dimensional damping ratio and modal constants which are referred to as modal parameters) are estimated by mean of over-refined modal identification techniques. The current work goes into depth in the analysis of an OMA algorithm particularly suited for the extraction of modal parameters from structures exhibiting close-in-frequency and highly damped modes as it is the typical case when large civil structures are investigated. The merit of the first part of this work consists in the development of a general purpose Monte Carlo-based simulation method conceived to assess the result intrinsic dispersion and bias effects of a generic OMA algorithm. Spread assessment is clearly a key information towards any attempt to exploit SHM methods based on modal parameter changes; bias assessment, on the other hand, offers a precious information to critically interpret the modal parameters estimates. With regard to the continuative long-term AE-based SHM strategies, the current work deals with the problems arising from the need to size short-transient events over extended period of time. The challenge associated with capturing these events using classical techniques is that very high sampling rates must be used over extended periods of time. The result is that a very large amount of data is collected to capture a phenomenon that rarely occurs. Furthermore, the high energy consumption associated with the required high sampling rates makes the implementation of high-endurance, low-power, embedded AE sensor nodes difficult to achieve. The merit of the second part of this thesis consists in investigating the suitability of a new sampling technique, called compressed sensing (CS), that, by collecting a small amount of compressed measurements, enables to relax the requirements on the sampling rate and memory demands. The CS technique substantially lends itself as a highly promising approach to promote the transition to SHM strategies relying upon wireless sensor networks featuring minimal installation costs. The research activity carried out within the framework of this thesis is supported by an extensive experimental activity. In particular, the Giuseppe Meazza stadium in Milan is used as test case. Such a structure, due to its specificities, offers the opportunity to investigate the suitability of a monitoring solution relying upon both OMA- and AE-based SHM approaches. The Monte Carlo-based simulation framework developed in this work, thanks to its close adherence to reality, allows to study in depth what are the performances of the OMA-based algorithm of interest in extracting the modal parameters of a given structure where is installed a specific measurement setup. Thanks to the developed methodology, the Monte Carlo method allows to test the OMA algorithm on numerical responses exhibiting the presence of those same modes close in frequency and affected by large modal overlap that are featured by the responses collected on a 3rd ring Meazza stadium grandstand where a permanent monitoring system is operating. As practical outcome, the developed methodology has allowed to deeply test the performance of the OMA algorithm used for the analysis of the data gathered on the real structure and tune it for an increased effectiveness of the analysis over the long-term period. The second part of the work, on the other hand, investigates the suitability of a monitoring strategy for the Meazza stadium roof based on the detection of AE events. On the basis of the data gathered during some preliminary tests conceived to assess the transmissibility of AE events through the stadium roof, CS has demonstrated its suitability to size the AE events of interest by exploiting a number of compressed measurements roughly equal to 25% of samples required by conventional sapling techniques. Ultimately, the general purpose methodology developed to assess bias and dispersion intrinsic to a generic OMA algorithm, as well as, the investigations on the innovative compressed sensing technique make this work a valuable contribution to cope with the challenging issues SHM on large civil structures poses.File | Dimensione | Formato | |
---|---|---|---|
2013_03_PhD_Cattaneo.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
4.01 MB
Formato
Adobe PDF
|
4.01 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/10589/74224