Today, the paradigm of complexity permeates all scientific fields. The study of complex systems represents a true scientific revolution, arising from empirical evidence that understanding or "explaining" many real-world systems cannot be reduced to simple, deterministic cause-effect relationships. Large-scale civil infrastructures are complex systems featuring nonlinear, interdependent interactions among numerous components across different domains and scales. Each component, though autonomous, influences the behavior of others, often resulting in dynamic equilibria or instabilities that cannot be deduced by merely combining the rules governing individual components. This is not only due to the mathematical intricacy of the problem but also because of the extreme variability in outcomes that can arise from minimal changes or insufficient definition in input data, affected not only by the current state but also by the entire history of the system. Consequently, for such systems, a classical analytical approach fails to provide significant insights or reliable future predictions without an impractically large and precise data volume, rendering the mathematical model inoperable. Furthermore, in current practice the analysis of large-scale civil infrastructures is often confined within isolated, vertical domains, lacking the flexibility needed to address cross-domain and cross-scale challenges. Data indicates that the absence of a systemic view is a major contributor to inefficiencies in infrastructure management. Presently, such interactions are managed in back-office environments through expertise integration processes that are often too slow for real-time response, failing to meet operational and emergency needs effectively. A representation that captures multi-scale, multi-domain relationships and reveals emergent behaviors of the system and its components is therefore required. The aim of Digital Twins (DTs) is to address these challenges. DTs are meant to analyze systems where heterogeneous phenomena interact, and where even minor inputs from one subsystem can significantly impact others across scales and domains. A DT should be viewed as a system of collective intelligence—a shared interpretation of reality that synchronizes multi-domain and multi-scale subsystems by aligning both virtual and real (also human) agents. A DT cannot solely be thought as a traditional simulation tool, nor simply a virtualization of physical systems, although including them. It holds interpretive and decision-making value, which brings a significant technical challenge: representing in real time system agent’s relationships, where they exist, according to a single criterion which is their relevance and significance to management objectives. DTs can thus be used to represent systemic emergent behaviors and diagnose potential issues. Unlike typical inference methodologies (e.g., top-down causal inferences), which suffice for analyzing simple systems, DTs observe effects and diagnose potential critical phenomena with predictive capabilities, supporting informed decision-making. In this thesis, a digital modeling approach is pursued, enabling the representation of multi-scale and multi-domain phenomena interactions, contributing to the development of a multi-model DT for the management of large-scale civil infrastructure systems. The research involved collaborative efforts to develop a Digital Twin cloud platform architecture and its implementation components (tools and microservices). At the core of the developed framework lies a concertation core that orchestrates multiple specialized microservices. We developed an environment where virtual and real agents are synchronized to realize the large-scale DT for civil infrastructure management, focusing primarily on the representation of interactions among multi-domain, multi-scale subsystems by developing digital twinning methodologies, procedures, and technologies for: 1. structuring heterogeneous data semantically, 2. achieving geometric and semantic alignment of multi-domain and -scale data models (registration), 3. establishing automatic, bidirectional (field-model) registration systems for digital inspections, 4. representing relationships among multi-domain and -scale information and agents (contextualization). Qualitative and quantitative evaluations of the implemented procedures were performed in laboratory with subsequent field deployment. Both the digital twinning approach and vertical analytical solutions proposed in this work represent substantial innovations compared to the current state of the art and ultimately helped define a unique DT cloud platform built on methodological choices and technological solutions that overcome the cited critical limitations found in today’s approaches and systems for civil infrastructure management.
Tutti gli ambiti della scienza sono oggi permeati dal paradigma della complessità. Lo studio dei sistemi complessi si pone come una vera e propria rivoluzione scientifica scaturita dall'evidenza empirica che la comprensione o "spiegazione" di molti sistemi reali non possa essere ridotta a relazioni causa-effetto semplici e deterministiche. Le infrastrutture civili su larga scala sono sistemi complessi caratterizzati dalla reciproca interazione non lineare di molte componenti appartenenti a domini e scale differenti, ognuna delle quali, pur essendo autonoma, influenza il comportamento delle altre. Sono di fatto sistemi che manifestano spesso instabilità o equilibri dinamici che non possono essere desunti dalla semplice sovrapposizione delle leggi che governano le singole componenti. Ciò non solo per la rilevante complessità matematica del problema quanto per l’estrema variabilità dei risultati che si otterrebbero a fronte di piccolissime variazioni o mancata definizione dei dati di input che coinvolgono non solo lo stato corrente ma l’intera storia del sistema. Si può dunque affermare che, per tali sistemi, un approccio analitico classico non riesce a spiegarne il comportamento in modo significativo, né tantomeno a prevederne l'evoluzione futura, a meno di trattare una quantità e precisione di dati tale da rendere il modello matematico non operabile. Inoltre, nella pratica attuale, le analisi su infrastrutture civili su larga scala seguono puntualmente approcci tradizionali confinati in ambiti completamente isolati, privi della flessibilità necessaria per affrontare sfide che coinvolgono più discipline e più livelli di scala. I dati più recenti includono l'assenza di visione sistemica tra i principali fattori di inefficienza nella gestione delle infrastrutture civili. Attualmente infatti, analisi multi-disciplinari e multi-livello sono gestite in ambienti di back-office attraverso processi di integrazione delle competenze che si rivelano spesso troppo lenti per una risposta in tempo reale, non riuscendo così a soddisfare efficacemente le esigenze operative e di emergenza. È pertanto necessaria una modellazione che catturi le relazioni multi-scala e multi-dominio e riveli i comportamenti emergenti del sistema e delle sue parti. Il concetto di Digital Twin (DT) entra in gioco proprio per rispondere a queste sfide. I DT sono concepiti per analizzare sistemi in cui fenomeni estremamente eterogenei interagiscono reciprocamente e in cui anche piccoli stimoli provenienti da un singolo sottosistema possono influenzare significativamente altri sottosistemi, su diverse scale e in diversi domini. Un DT deve essere visto come un sistema di intelligenza collettiva, un'interpretazione condivisa della realtà che sincronizza sottosistemi multi-dominio e multi-scala mediante l'allineamento di agenti virtuali e reali, tra cui l’uomo. Un DT non può perciò essere concepito solo come uno strumento di simulazione, né come una semplice virtualizzazione dei sistemi fisici, pur includendoli. Il DT possiede un valore interpretativo e decisionale, che comporta sfide metodologiche e tecniche significative: rappresentare in tempo reale le relazioni tra gli agenti del sistema, laddove esistono, in funzione della loro rilevanza e pertinenza rispetto agli obiettivi di gestione. I DT modellano comportamenti sistemici emergenti e diagnosticano potenziali criticità, con capacità anche previdenziali. A differenza delle tipiche metodologie di inferenza come quelle causali, sufficienti per analisi di sistemi semplici, i DT osservano gli effetti e diagnosticano le potenziali cause, favorendo un processo decisionale informato e predittivo in tempo reale. In questa chiave, in questo lavoro di tesi si persegue una modellazione della realtà orientata verso un approccio digitale, che consente di inquadrare interazioni multi-scala e multi-dominio, contribuendo alla costruzione di un DT multi-modello per la gestione di infrastrutture civili su larga scala. La ricerca ha comportato un impegno collaborativo nello sviluppo di un'architettura di piattaforma cloud di DT e dei suoi componenti implementativi. Al centro del framework sviluppato si trova un nucleo di concertazione che orchestra molteplici microservizi e strumenti specializzati, tra cui anche algoritmi di intelligenza artificiale e computer vision. Si è definito un ambiente in cui agenti virtuali e reali sono sincronizzati per realizzare un DT su larga scala finalizzato alla gestione di infrastrutture civili, con particolare attenzione alla rappresentazione digitale delle interazioni tra sottosistemi multi-dominio e multi-scala, sviluppando metodologie, procedure e tecnologie di digital twinning per: 1. strutturare semanticamente dati eterogenei, 2. allineare geometricamente e semanticamente modelli di dati multi-dominio e -scala (registrazione), 3. automatizzare le registrazioni, anche bidirezionali (modello-campo) per abilitare ispezioni digitali, 4. rappresentare le relazioni multi-dominio e -scala, pertinenti e rilevanti, tra le informazioni e gli agenti presenti nel DT (contestualizzazione). Valutazioni sia qualitative che quantitative delle procedure implementate sono state eseguite in laboratorio con conseguente applicazione sul campo, prendendo come riferimento scenari complessi definiti da sistemi di infrastrutture civili di grande scala. L’approccio di digital twin e le soluzioni proposte in questo lavoro comportano innovazioni sostanziali rispetto all’attuale stato dell’arte e, in ultima analisi, hanno contribuito a definire una piattaforma cloud di DT per la gestione delle infrastrutture civili su larga scala unica nel suo genere.
Large-scale Digital Twin for Civil Infrastructure Management / Binni, Leonardo. - (2025 Mar).
Large-scale Digital Twin for Civil Infrastructure Management
BINNI, LEONARDO
2025-03-01
Abstract
Today, the paradigm of complexity permeates all scientific fields. The study of complex systems represents a true scientific revolution, arising from empirical evidence that understanding or "explaining" many real-world systems cannot be reduced to simple, deterministic cause-effect relationships. Large-scale civil infrastructures are complex systems featuring nonlinear, interdependent interactions among numerous components across different domains and scales. Each component, though autonomous, influences the behavior of others, often resulting in dynamic equilibria or instabilities that cannot be deduced by merely combining the rules governing individual components. This is not only due to the mathematical intricacy of the problem but also because of the extreme variability in outcomes that can arise from minimal changes or insufficient definition in input data, affected not only by the current state but also by the entire history of the system. Consequently, for such systems, a classical analytical approach fails to provide significant insights or reliable future predictions without an impractically large and precise data volume, rendering the mathematical model inoperable. Furthermore, in current practice the analysis of large-scale civil infrastructures is often confined within isolated, vertical domains, lacking the flexibility needed to address cross-domain and cross-scale challenges. Data indicates that the absence of a systemic view is a major contributor to inefficiencies in infrastructure management. Presently, such interactions are managed in back-office environments through expertise integration processes that are often too slow for real-time response, failing to meet operational and emergency needs effectively. A representation that captures multi-scale, multi-domain relationships and reveals emergent behaviors of the system and its components is therefore required. The aim of Digital Twins (DTs) is to address these challenges. DTs are meant to analyze systems where heterogeneous phenomena interact, and where even minor inputs from one subsystem can significantly impact others across scales and domains. A DT should be viewed as a system of collective intelligence—a shared interpretation of reality that synchronizes multi-domain and multi-scale subsystems by aligning both virtual and real (also human) agents. A DT cannot solely be thought as a traditional simulation tool, nor simply a virtualization of physical systems, although including them. It holds interpretive and decision-making value, which brings a significant technical challenge: representing in real time system agent’s relationships, where they exist, according to a single criterion which is their relevance and significance to management objectives. DTs can thus be used to represent systemic emergent behaviors and diagnose potential issues. Unlike typical inference methodologies (e.g., top-down causal inferences), which suffice for analyzing simple systems, DTs observe effects and diagnose potential critical phenomena with predictive capabilities, supporting informed decision-making. In this thesis, a digital modeling approach is pursued, enabling the representation of multi-scale and multi-domain phenomena interactions, contributing to the development of a multi-model DT for the management of large-scale civil infrastructure systems. The research involved collaborative efforts to develop a Digital Twin cloud platform architecture and its implementation components (tools and microservices). At the core of the developed framework lies a concertation core that orchestrates multiple specialized microservices. We developed an environment where virtual and real agents are synchronized to realize the large-scale DT for civil infrastructure management, focusing primarily on the representation of interactions among multi-domain, multi-scale subsystems by developing digital twinning methodologies, procedures, and technologies for: 1. structuring heterogeneous data semantically, 2. achieving geometric and semantic alignment of multi-domain and -scale data models (registration), 3. establishing automatic, bidirectional (field-model) registration systems for digital inspections, 4. representing relationships among multi-domain and -scale information and agents (contextualization). Qualitative and quantitative evaluations of the implemented procedures were performed in laboratory with subsequent field deployment. Both the digital twinning approach and vertical analytical solutions proposed in this work represent substantial innovations compared to the current state of the art and ultimately helped define a unique DT cloud platform built on methodological choices and technological solutions that overcome the cited critical limitations found in today’s approaches and systems for civil infrastructure management.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.