In the last two decades, the demand for safer, lighter, and more efficient engineering assets across civil, aerospace, and marine sectors has accelerated the adoption of Structural Health and Usage Monitoring (SHUM) frameworks. By combining permanently installed sensing systems with advanced inference algorithms, SHUM aims to shift maintenance from scheduled interventions to condition-based and predictive strategies. However, in many critical applications, the most severe performance losses and failures are triggered by rare but high-consequence extreme-load events—such as shock-like actions, impact-type scenarios, or strong dynamic excitations—whose effects are difficult to characterize from operational data alone. As a result, SHUM algorithms may suffer from limited “experience” of the most informative regimes of structural response, ultimately constraining diagnostic and prognostic capability. This PhD thesis addresses this challenge by proposing a unified research vision: integrating high-fidelity simulations of extreme loads with SHUM algorithms to improve state assessment, damage identification, and decision support for engineering structures. The work is organized around three complementary pillars. The first pillar focuses on advanced modelling of extreme loads and their interaction with structures, with particular attention to underwater-explosion-driven fluid–structure interaction and to the propagation of shock-induced vibrations. Beyond primary structural integrity, the thesis highlights the importance of vibro-acoustic response and equipment-level effects, and it investigates mitigation solutions aimed at reducing vibration transmission under extreme excitations. The second pillar advances SHUM algorithmic capabilities through the development and extension of inverse Finite Element Method (iFEM) frameworks for shape sensing, load/usage monitoring, and damage detection. Key contributions address practical obstacles that often limit real deployments, including non-ideal and degrading boundary conditions, sparse sensing and the need for robust strain pre-extrapolation, multiscale reconstruction for stiffened structures, and load-independent damage localization via anomaly-based indicators. These developments are validated across heterogeneous structural typologies—ranging from composite aerostructures and stiffened panels to ship structures and long-span suspension bridges under non-trivial dynamic loading—demonstrating iFEM’s versatility when embedded in realistic monitoring pipelines. The third pillar bridges simulation and monitoring by leveraging high-fidelity numerical models to support, accelerate, and augment SHUM inference, especially when full-fidelity analyses are too expensive for iterative use. Within the UNDEX domain, the thesis explores machine-learning-enabled surrogates and hybrid strategies to enable rapid prediction of key event parameters and to correlate fast, uncoupled simulations with higher-fidelity coupled results. More broadly, the thesis promotes a simulation-informed SHUM paradigm in which physics-based models enrich the monitoring system with scenario coverage, synthetic data, uncertainty awareness, and mitigation-oriented insights that are otherwise inaccessible from sensing alone. Overall, the contributions of this thesis span from high-fidelity extreme-load modelling to SHUM algorithmic enhancements and integrated simulation-informed frameworks. While each advancement is valuable on its own, their full impact is achieved when combined into a cohesive methodology, paving the way for more reliable, explainable, and operationally relevant monitoring and decision-making tools for engineering structures exposed to extreme loading conditions.
Negli ultimi due decenni, la richiesta di sistemi ingegneristici più sicuri, leggeri ed efficienti nei settori civile, aerospaziale e navale ha accelerato l’adozione di framework di Structural Health and Usage Monitoring (SHUM). Combinando sistemi di sensori permanentemente installati con algoritmi avanzati di inferenza, lo SHUM mira a trasformare la manutenzione da interventi programmati a strategie basate sulle condizioni operative e sulla manutenzione predittiva. Tuttavia, in molte applicazioni critiche, le più gravi perdite prestazionali e i guasti sono innescati da eventi di carico estremo rari ma ad alte conseguenze — come azioni impulsive, scenari di impatto o forti eccitazioni dinamiche — i cui effetti risultano difficili da caratterizzare basandosi unicamente sui dati operativi. Di conseguenza, gli algoritmi SHUM possono soffrire di una limitata “esperienza” nei regimi più informativi della risposta strutturale, limitando così le capacità diagnostiche e prognostiche. Questa tesi di dottorato affronta tale sfida proponendo una visione di ricerca unificata: integrare simulazioni ad alta fedeltà di carichi estremi con algoritmi SHUM al fine di migliorare la valutazione dello stato strutturale, l’identificazione del danno e il supporto decisionale per strutture ingegneristiche. Il lavoro è organizzato attorno a tre pilastri complementari. Il primo pilastro si concentra sulla modellazione avanzata dei carichi estremi e della loro interazione con le strutture, con particolare attenzione all’interazione fluido-struttura indotta da esplosioni subacquee e alla propagazione di vibrazioni generate da shock. Oltre all’integrità strutturale primaria, la tesi evidenzia l’importanza della risposta vibro-acustica e degli effetti a livello di equipaggiamento, investigando inoltre soluzioni di mitigazione finalizzate a ridurre la trasmissione delle vibrazioni sotto eccitazioni estreme. Il secondo pilastro sviluppa le capacità algoritmiche dello SHUM attraverso lo sviluppo e l’estensione di framework basati sull’inverse Finite Element Method (iFEM) per shape sensing, monitoraggio dei carichi/utilizzo e rilevamento del danno. I principali contributi affrontano ostacoli pratici che spesso limitano le applicazioni reali, inclusi vincoli non ideali e degradanti, sensoristica sparsa e necessità di robuste tecniche di pre-estrapolazione delle deformazioni, ricostruzione multiscala per strutture irrigidite e localizzazione del danno indipendente dal carico tramite indicatori basati su anomalie. Tali sviluppi vengono validati su tipologie strutturali eterogenee — che spaziano da aerostrutture in materiale composito e pannelli irrigiditi fino a strutture navali e ponti sospesi di grande luce soggetti a carichi dinamici complessi — dimostrando la versatilità dell’iFEM quando integrato in pipeline di monitoraggio realistiche. Il terzo pilastro crea un collegamento tra simulazione e monitoraggio sfruttando modelli numerici ad alta fedeltà per supportare, accelerare e potenziare i processi di inferenza SHUM, specialmente nei casi in cui analisi ad alta fedeltà risultino troppo onerose per un utilizzo iterativo. Nell’ambito delle esplosioni subacquee (UNDEX), la tesi esplora surrogate models basati su machine learning e strategie ibride per consentire la rapida previsione dei principali parametri dell’evento e correlare simulazioni veloci disaccoppiate con risultati accoppiati a più alta fedeltà. Più in generale, la tesi promuove un paradigma SHUM informato dalla simulazione, nel quale modelli physics-based arricchiscono il sistema di monitoraggio con copertura di scenari, dati sintetici, consapevolezza dell’incertezza e indicazioni orientate alla mitigazione, altrimenti non ottenibili tramite la sola sensoristica. Nel complesso, i contributi di questa tesi spaziano dalla modellazione ad alta fedeltà di carichi estremi al miglioramento degli algoritmi SHUM e allo sviluppo di framework integrati simulation-informed. Sebbene ciascun avanzamento abbia valore individualmente, il loro impatto massimo si realizza quando vengono combinati in una metodologia coerente, aprendo la strada a strumenti di monitoraggio e supporto decisionale più affidabili, interpretabili e operativamente rilevanti per strutture ingegneristiche soggette a condizioni di carico estremo.
Integrating high-fidelity extreme load simulations, Structural Health and Usage Monitoring for engineering structures
Bardiani, Jacopo
2026
Abstract
In the last two decades, the demand for safer, lighter, and more efficient engineering assets across civil, aerospace, and marine sectors has accelerated the adoption of Structural Health and Usage Monitoring (SHUM) frameworks. By combining permanently installed sensing systems with advanced inference algorithms, SHUM aims to shift maintenance from scheduled interventions to condition-based and predictive strategies. However, in many critical applications, the most severe performance losses and failures are triggered by rare but high-consequence extreme-load events—such as shock-like actions, impact-type scenarios, or strong dynamic excitations—whose effects are difficult to characterize from operational data alone. As a result, SHUM algorithms may suffer from limited “experience” of the most informative regimes of structural response, ultimately constraining diagnostic and prognostic capability. This PhD thesis addresses this challenge by proposing a unified research vision: integrating high-fidelity simulations of extreme loads with SHUM algorithms to improve state assessment, damage identification, and decision support for engineering structures. The work is organized around three complementary pillars. The first pillar focuses on advanced modelling of extreme loads and their interaction with structures, with particular attention to underwater-explosion-driven fluid–structure interaction and to the propagation of shock-induced vibrations. Beyond primary structural integrity, the thesis highlights the importance of vibro-acoustic response and equipment-level effects, and it investigates mitigation solutions aimed at reducing vibration transmission under extreme excitations. The second pillar advances SHUM algorithmic capabilities through the development and extension of inverse Finite Element Method (iFEM) frameworks for shape sensing, load/usage monitoring, and damage detection. Key contributions address practical obstacles that often limit real deployments, including non-ideal and degrading boundary conditions, sparse sensing and the need for robust strain pre-extrapolation, multiscale reconstruction for stiffened structures, and load-independent damage localization via anomaly-based indicators. These developments are validated across heterogeneous structural typologies—ranging from composite aerostructures and stiffened panels to ship structures and long-span suspension bridges under non-trivial dynamic loading—demonstrating iFEM’s versatility when embedded in realistic monitoring pipelines. The third pillar bridges simulation and monitoring by leveraging high-fidelity numerical models to support, accelerate, and augment SHUM inference, especially when full-fidelity analyses are too expensive for iterative use. Within the UNDEX domain, the thesis explores machine-learning-enabled surrogates and hybrid strategies to enable rapid prediction of key event parameters and to correlate fast, uncoupled simulations with higher-fidelity coupled results. More broadly, the thesis promotes a simulation-informed SHUM paradigm in which physics-based models enrich the monitoring system with scenario coverage, synthetic data, uncertainty awareness, and mitigation-oriented insights that are otherwise inaccessible from sensing alone. Overall, the contributions of this thesis span from high-fidelity extreme-load modelling to SHUM algorithmic enhancements and integrated simulation-informed frameworks. While each advancement is valuable on its own, their full impact is achieved when combined into a cohesive methodology, paving the way for more reliable, explainable, and operationally relevant monitoring and decision-making tools for engineering structures exposed to extreme loading conditions.| File | Dimensione | Formato | |
|---|---|---|---|
|
DEF_Bardiani_PhD_thesis.pdf
non disponibili
Licenza:
Tutti i diritti riservati
Dimensione
163 MB
Formato
Adobe PDF
|
163 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/384580
URN:NBN:IT:POLIMI-384580