Today, water infrastructures must face new challenges that affect both their design and management. Many of these strategic infrastructures have reached or are approaching the lower limit of their lifespan, and this mass aging increasingly requires widespread and efficient monitoring of their functionality. In addition, climate change is leading to new operational scenarios. In this context, it is clear that the traditional monitoring approach, based on exceeding a specific statistically determined threshold on the basis of historical data for each control variable, is increasingly inadequate to meet these challenges. In recent years, there has been an exponential growth in the literature of studies proposing innovative Structural Health Monitoring (SHM) solutions based on Machine Learning analytics and on the power of hyperconnected sensor systems enabled by the Internet of Things. However, the dissemination of these concepts in the world of water infrastructure monitoring has been rather slow, and their relationship with traditional monitoring methods, validated and well established over more than half a century of modern management of these works, has yet to be defined. The research aim of this thesis is therefore to explore these innovative tools to propose a novel approach in detecting and classifying anomalies in the operation of water infrastructures, founded on the fusion of heterogeneous multi-sensor information. To achieve this purpose, this thesis focused on the use of Fiber Optic Sensors, an increasingly emerging solution for strain and temperature monitoring in the field of SHM of extensive infrastructures. Competitive features such as small size, high flexibility, high resistance to harsh environments, and immunity to electromagnetic interference, combined with increasingly low cost, and high measurement resolution and accuracy, have made FOS the "smart sensor" by definition. In order to achieve the objective of this thesis, we have developed complex measurement systems specifically designed to provide the volume of data required for the above-mentioned data-driven analysis. This has required a continuous combination of analytical and methodological studies with considerable technological effort, both in the laboratory and in the field. Some of the well-established techniques for sensor fusion and anomaly detection (clustering, Principal Component Analysis, Gaussian Mixture Model, Mahalanobis Squared Distance) have been positively tested in the case studies here presented, highlighting the value of information offered by FOS to reach the objectives of this thesis. By applying the concept of sensor fusion, we have also demonstrated that it is also possible to classify strain anomalies according to the environmental factors that triggered them. This is a strategic tool as it makes it possible to distinguish in real time which anomalies require attention as they are not strictly linked to the variability of environmental or operational factors. At the same time, continuous monitoring of operational anomalies caused by environmental factors provides a useful knowledge framework for studying adaptation strategies to the effects of climate change on water infrastructures. Finally, the field experience and know-how gained with FOS technology has led to the production of a novel set of case studies in the literature on the use of this technology for the implementation of smart SHM solutions for anomaly detection; among them, the ongoing research on smart revetments for levee monitoring can be considered as a perfect synthesis of the work of this thesis.
Le infrastrutture idrauliche sono un bene strategico per qualsiasi Paese in termini di gestione dei corpi idrici, produzione di energia, approvvigionamento e conservazione della risorsa idrica, l’«Oro blu» del nostro Pianeta. La progettazione di nuove infrastrutture e, soprattutto, la gestione di quelle esistenti, devono oggi far fronte a nuovi scenari operativi, in alcuni casi ben diversi da quelli originali di progetto, dettati dagli effetti dei cambiamenti climatici sui processi della catena idrologica. Inoltre, molte di queste infrastrutture strategiche, realizzate in calcestruzzo armato a metà del secolo scorso, hanno ormai raggiunto o si stanno avvicinando al limite inferiore della loro vita nominale di progetto; questo invecchiamento di massa richiede sempre più un monitoraggio estensivo ed accurato della loro funzionalità. In questo contesto, è chiaro che l'approccio di monitoraggio tradizionale di queste infrastrutture, basato sul superamento da parte dei parametri di controllo di una specifica soglia statisticamente determinata sulla base delle serie storiche, risulta sempre più inadeguato a rispondere a queste sfide. Per rispondere a tale esigenza, negli ultimi anni, si è assistito a una crescita esponenziale in letteratura di studi che propongono soluzioni innovative di Structural Health Monitoring (SHM) basate su tecniche di Machine Learning e su sistemi di monitoraggio complessi composti da sensori iperconnessi basati sul principio dell’«Internet of Things». Tuttavia, la diffusione di questi concetti nel mondo del monitoraggio delle infrastrutture idrauliche è piuttosto lenta, così come è ancora da definire il loro rapporto con le tradizionali procedure di monitoraggio, validate e ben consolidate in più di mezzo secolo di gestione moderna di queste opere. L'obiettivo di ricerca di questa tesi è quindi quello di esplorare questi strumenti innovativi per proporre un approccio innovativo nel rilevamento e nella classificazione delle anomalie nel funzionamento delle infrastrutture idrauliche, fondato sulla fusione di informazioni eterogenee multi-sensore. Per raggiungere questo scopo, la tesi si è concentrata sull'uso dei sensori in fibra ottica (FOS), tecnologia sempre più emergente per il monitoraggio delle deformazioni e della temperatura nel campo dello SHM delle infrastrutture estese. Caratteristiche competitive come le dimensioni ridotte, l'elevata flessibilità, l'alta resistenza a condizioni ambientali estreme e l'immunità alle interferenze elettromagnetiche, unite al costo sempre più competitivo e all'alta risoluzione e precisione delle misure, hanno reso i FOS il “sensore intelligente” per definizione. Per raggiungere l'obiettivo di questa tesi, abbiamo sviluppato sistemi di misura complessi, specificamente progettati per fornire il volume di dati necessario per l'analisi data-driven di cui sopra. Ciò ha richiesto una continua combinazione di studi analitici e metodologici con un notevole sforzo tecnologico, sia in laboratorio che sul campo, per acquisire il background tecnico sui FOS necessario alla corretta gestione e interpretazione della misura. Alcune delle tecniche consolidate per la fusione di sensori e la rilevazione di anomalie (Clustering, Principal Component Analysis, Gaussian Mixture Model, Mahalanobis Squared Distance) sono state testate positivamente nei casi di studio qui presentati, confermando il valore delle informazioni offerte dai FOS per raggiungere gli obiettivi di questa tesi. Applicando il concetto della sensor fusion, abbiamo inoltre dimostrato che è possibile classificare le anomalie di deformazione in base ai fattori ambientali che le hanno provocate. Si tratta di uno strumento strategico in quanto permette di distinguere in tempo reale quali anomalie richiedono attenzione in quanto non strettamente legate alla variabilità dei fattori ambientali o operativi. Allo stesso tempo, il monitoraggio continuo delle anomalie operative causate da fattori ambientali fornisce un quadro conoscitivo utile per studiare le strategie di adattamento agli effetti dei cambiamenti climatici sulle infrastrutture idrauliche. Infine, la sperimentazione dei FOS su casi studio reali ha portato alla produzione di una casistica inedita in letteratura sull'uso di questa tecnologia per l'implementazione di soluzioni di SHM intelligenti volte alla rilevazione di anomalie operative; tra le applicazioni presentate, la ricerca sui rivestimenti intelligenti per il monitoraggio degli argini fluviali può essere considerata una perfetta sintesi del lavoro presentato in questa tesi.
Smart monitoring of water infrastructures: a novel approach based on fiber optic sensors and data fusion
BERTULESSI, MANUEL
2024/2025
Abstract
Today, water infrastructures must face new challenges that affect both their design and management. Many of these strategic infrastructures have reached or are approaching the lower limit of their lifespan, and this mass aging increasingly requires widespread and efficient monitoring of their functionality. In addition, climate change is leading to new operational scenarios. In this context, it is clear that the traditional monitoring approach, based on exceeding a specific statistically determined threshold on the basis of historical data for each control variable, is increasingly inadequate to meet these challenges. In recent years, there has been an exponential growth in the literature of studies proposing innovative Structural Health Monitoring (SHM) solutions based on Machine Learning analytics and on the power of hyperconnected sensor systems enabled by the Internet of Things. However, the dissemination of these concepts in the world of water infrastructure monitoring has been rather slow, and their relationship with traditional monitoring methods, validated and well established over more than half a century of modern management of these works, has yet to be defined. The research aim of this thesis is therefore to explore these innovative tools to propose a novel approach in detecting and classifying anomalies in the operation of water infrastructures, founded on the fusion of heterogeneous multi-sensor information. To achieve this purpose, this thesis focused on the use of Fiber Optic Sensors, an increasingly emerging solution for strain and temperature monitoring in the field of SHM of extensive infrastructures. Competitive features such as small size, high flexibility, high resistance to harsh environments, and immunity to electromagnetic interference, combined with increasingly low cost, and high measurement resolution and accuracy, have made FOS the "smart sensor" by definition. In order to achieve the objective of this thesis, we have developed complex measurement systems specifically designed to provide the volume of data required for the above-mentioned data-driven analysis. This has required a continuous combination of analytical and methodological studies with considerable technological effort, both in the laboratory and in the field. Some of the well-established techniques for sensor fusion and anomaly detection (clustering, Principal Component Analysis, Gaussian Mixture Model, Mahalanobis Squared Distance) have been positively tested in the case studies here presented, highlighting the value of information offered by FOS to reach the objectives of this thesis. By applying the concept of sensor fusion, we have also demonstrated that it is also possible to classify strain anomalies according to the environmental factors that triggered them. This is a strategic tool as it makes it possible to distinguish in real time which anomalies require attention as they are not strictly linked to the variability of environmental or operational factors. At the same time, continuous monitoring of operational anomalies caused by environmental factors provides a useful knowledge framework for studying adaptation strategies to the effects of climate change on water infrastructures. Finally, the field experience and know-how gained with FOS technology has led to the production of a novel set of case studies in the literature on the use of this technology for the implementation of smart SHM solutions for anomaly detection; among them, the ongoing research on smart revetments for levee monitoring can be considered as a perfect synthesis of the work of this thesis.File | Dimensione | Formato | |
---|---|---|---|
PhD_Thesis___Manuel_Bertulessi_final.pdf
accessibile in internet per tutti a partire dal 28/04/2026
Descrizione: PhD thesis Manuel Bertulessi
Dimensione
54.45 MB
Formato
Adobe PDF
|
54.45 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/10589/238057