This thesis work addresses the problem of optimal placement of sensors in thermo-hydraulic systems and real-time estimation of pressure, temperature and velocity fields, through the joint use of model order reduction methods and measurements taken on the system of interest, limiting the effect of random noise. The Empirical Interpolation Method (EIM) and its generalized version (GEIM) are non-intrusive, reduced basis model order reduction methods, that have been adopted and modified to achieve the aforementioned objectives. These techniques have been used to extract the characteristic spatial modes of the system and to select a set of points (or functional) corresponding to the optimal locations for the sensors. The collection of experimental measurements in the selected points allows constructing an empirical interpolation of the fields employed to estimate the variable of interest. In this context, this thesis provides three important contributions: i) the development of a new method for vector field interpolation based on the collection of local flow measurements; ii) an effective solution to the problem of reconstructing the system state in the presence of experimental data affected by random noise thanks to the Tikhonov regularization; iii) the improvement of an inverse reconstruction method for estimating in real-time fields from experimental measurements relying only on related physical quantities. The developed methods are tested on a simple thermo-fluid dynamics problem known as "two-sided lid-driven differentially heated square cavity". Finally, the GEIM and the related techniques developed in this thesis are applied to the Molten Salt Fast Reactor with the aim of defining a sensor placement strategy for the reconstruction of the temperature fields in the primary circuit during an accidental transient. Thanks to the inverse reconstruction method, the evolution of the neutron flux is estimated with great precision using few temperature measurements, despite being affected by random noise. The methodological developments and the numerical results achieved in this thesis pave the way to the real-world applications of GEIM for thermo-hydraulic systems.
Questo lavoro di tesi affronta il problema del posizionamento ottimo dei sensori in sistemi termo-idraulici e della stima in tempo reale dei campi di pressione, temperatura e velocità, attraverso l'uso congiunto di tecniche di riduzione d'ordine e misure effettuate sul sistema d'interesse, limitando l'effetto del loro rumore stocastico. Le tecniche di riduzione d'ordine, a basi ridotte e non intrusive, adottate e modificate per raggiungere l'obiettivo sopra menzionato, sono l'Empirical Interpolation Method (EIM) e la sua versione generalizzata (GEIM). Queste tecniche sono state utilizzate per estrarre i comportamenti spaziali caratteristici del sistema e per selezionare un insieme di punti (o funzionali) di misura corrispondenti al set di posizione ottime per i sensori. La raccolta di misure sperimentali nelle posizioni selezionate rende possibile costruire un'interpolazione empirica dei campi che si vuole stimare. In questo ambito, la tesi fornisce tre contributi importanti: in primo luogo viene sviluppato un nuovo metodo d'interpolazione empirica per la ricostruzione di campi vettoriali, basato sulla raccolta di misure di flusso locale (nel caso di velocità, si traduce in misure di portata); in secondo luogo, si propone una soluzione al problema della ricostruzione dello stato del sistema in presenza di dati affetti da rumore stocastico, attraverso la regolarizzazione di Tikhonov come principale strumento di stabilizzazione; infine, si migliora una metologia precedentemente proposta che permette di stimare in tempo reale campi (scalari o vettoriali) a partire da misure sperimentali effettuate solo su quantità fisiche ad esse correlate. I metodi sviluppati sono testati su un semplice problema di termo-fluidodinamica conosciuto come "Two-sided lid-driven differentially heated square cavity". Per ultimo, il GEIM viene applicato al reattore nucleare a sali fusi MSFR con l'obiettivo di stabilire una corretta selezione di misure per la ricostruzione del campo di temperatura nel circuito primario durante un transitorio incidentale. Grazie al metodo di ricostruzione inversa, si mostra come sia possibile stimare con elevata precisione l'evoluzione del flusso neutronico facendo uso di poche misure di temperatura, nonostante inquinate da rumore stocastico. I risultati di questa tesi avvicinano il GEIM ad applicazioni per sistemi termo-idraulici reali.
Non-intrusive reduced order methods for measurement selection and state estimation in presence of noise
CAVALLERI, SIMONE
2018/2019
Abstract
This thesis work addresses the problem of optimal placement of sensors in thermo-hydraulic systems and real-time estimation of pressure, temperature and velocity fields, through the joint use of model order reduction methods and measurements taken on the system of interest, limiting the effect of random noise. The Empirical Interpolation Method (EIM) and its generalized version (GEIM) are non-intrusive, reduced basis model order reduction methods, that have been adopted and modified to achieve the aforementioned objectives. These techniques have been used to extract the characteristic spatial modes of the system and to select a set of points (or functional) corresponding to the optimal locations for the sensors. The collection of experimental measurements in the selected points allows constructing an empirical interpolation of the fields employed to estimate the variable of interest. In this context, this thesis provides three important contributions: i) the development of a new method for vector field interpolation based on the collection of local flow measurements; ii) an effective solution to the problem of reconstructing the system state in the presence of experimental data affected by random noise thanks to the Tikhonov regularization; iii) the improvement of an inverse reconstruction method for estimating in real-time fields from experimental measurements relying only on related physical quantities. The developed methods are tested on a simple thermo-fluid dynamics problem known as "two-sided lid-driven differentially heated square cavity". Finally, the GEIM and the related techniques developed in this thesis are applied to the Molten Salt Fast Reactor with the aim of defining a sensor placement strategy for the reconstruction of the temperature fields in the primary circuit during an accidental transient. Thanks to the inverse reconstruction method, the evolution of the neutron flux is estimated with great precision using few temperature measurements, despite being affected by random noise. The methodological developments and the numerical results achieved in this thesis pave the way to the real-world applications of GEIM for thermo-hydraulic systems.File | Dimensione | Formato | |
---|---|---|---|
Thesis_SimoneCavalleri_875643.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
27.53 MB
Formato
Adobe PDF
|
27.53 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/10589/150833