This thesis work focuses on the development of an algorithm for the integration between experimental data and model simulation, through a Kalman filter specifically developed for fluid-dynamics applications. The novelty in this algorithm lies in the fact that mass conservation is guaranteed at every time step of the simulation, and in its expansion to heat transfer problems. The algorithm is implemented in the OpenFOAM software, and testing is performed on the lid-driven cavity test case under the Boussinesq approximation for natural circulation. Following this validation, the algorithm is applied to a real case, namely the coolant channel of the TRIGA Mark II reactor core. Aim of this application is to evaluate the capabilities of such an integrated algorithm to improve the performance of non-optimal turbulence models on low-quality grids. For this reason, real experimental data are used for this analysis, in particular measurements of temperature in six locations of the channel. In order to identify those turbulence models that might benefit from the developed algorithm, a comparison of the RAS models available in OpenFOAM is performed with respect to a LES simulation. Both a sensitivity analysis on the numerical grid and a detailed comparison on the selected best mesh are performed, leading to the identification of the best RAS model i.e. k-Omega SST and of the worst one i.e. the RNG k-Epsilon. Accordingly, the latter turbulence model is the best candidate to be improved by the Kalman data-driven algorithm. For the sake of comparison, the same numerical setup adopted for the model comparison is used and the results. The obtained results show that the developed method leads to performances comparable to those of the best model previously identified, confirming the potential of this new method. The thesis outcomes pave the way to different applications of the developed algorithm as the enhanced 3D data-driven simulation of the TRIGA reactor.
Questa tesi si focalizza sullo sviluppo di un algoritmo per l'integrazione tra dati sperimentali e simulazioni numeriche tramite un filtro di Kalman specificatamente sviluppato per applicazioni di fluido-dinamica. La novità di questo metodo risiede nel fatto che il principio di conservazione della massa è rispettato ad ogni iterazione durante la simulazione, e nella sua espansione a problemi di scambio termico. L’algoritmo è implementato in OpenFOAM, e testato rispetto il caso della cavità sotto l’approssimazione di Boussinesq per la circolazione naturale. A questa validazione segue l’applicazione dell’algoritmo a un caso reale, ovvero lo studio del canale di raffreddamento del nocciolo del reattore TRIGA Mark II. Scopo di questo studio è valutare le capacità dell’algoritmo integrato di migliorare la prestazione di modelli di turbolenza non ottimali applicati a griglie di bassa qualità. In quest’ottica, sono stati usati veri dati sperimentali (misure di temperatura nel canale). L’identificazione di quei modelli di turbolenza che potrebbero beneficiare dell’approccio integrato è stata fatta tramite un confronto dei modelli RAS presenti in OpenFOAM, valutati rispetto una simulazione LES. Tramite un’analisi di sensitività rispetto la griglia numerica e un confronto dettagliato sulla griglia ottimale, si è identificato il modello RAS avente le migliori prestazioni, e cioè il k-Omega SST, e quello avente le prestazioni peggiori, cioè il RNG k-Epsilon. Quest’ultimo è il miglior candidato per l’applicazione dell’algoritmo integrato. Per poter confrontare le prestazioni, è stato usato lo stesso caso numerico adottato per il confronto dei modelli RAS. I risultati ottenuti mostrano che l’algoritmo integrato porta ad avere prestazioni simili a quelle del modello migliore identificato in precedenza, confermando le potenzialità di questo nuovo metodo. I risultati di questa tesi aprono la strada verso l’applicazione dell’algoritmo sviluppato all'interno di una simulazione data-driven 3D del reattore TRIGA.
Data-driven computational fluid dynamics : development of a mass conservative Kalman filter algorithm for integration between experimental data and CFD model simulation, and application to the TRIGA Mark II reactor
INTROINI, CAROLINA
2016/2017
Abstract
This thesis work focuses on the development of an algorithm for the integration between experimental data and model simulation, through a Kalman filter specifically developed for fluid-dynamics applications. The novelty in this algorithm lies in the fact that mass conservation is guaranteed at every time step of the simulation, and in its expansion to heat transfer problems. The algorithm is implemented in the OpenFOAM software, and testing is performed on the lid-driven cavity test case under the Boussinesq approximation for natural circulation. Following this validation, the algorithm is applied to a real case, namely the coolant channel of the TRIGA Mark II reactor core. Aim of this application is to evaluate the capabilities of such an integrated algorithm to improve the performance of non-optimal turbulence models on low-quality grids. For this reason, real experimental data are used for this analysis, in particular measurements of temperature in six locations of the channel. In order to identify those turbulence models that might benefit from the developed algorithm, a comparison of the RAS models available in OpenFOAM is performed with respect to a LES simulation. Both a sensitivity analysis on the numerical grid and a detailed comparison on the selected best mesh are performed, leading to the identification of the best RAS model i.e. k-Omega SST and of the worst one i.e. the RNG k-Epsilon. Accordingly, the latter turbulence model is the best candidate to be improved by the Kalman data-driven algorithm. For the sake of comparison, the same numerical setup adopted for the model comparison is used and the results. The obtained results show that the developed method leads to performances comparable to those of the best model previously identified, confirming the potential of this new method. The thesis outcomes pave the way to different applications of the developed algorithm as the enhanced 3D data-driven simulation of the TRIGA reactor.File | Dimensione | Formato | |
---|---|---|---|
2017_10_Introini.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
13.48 MB
Formato
Adobe PDF
|
13.48 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/10589/135899